Broadband Access Networks - Telenor

Thomas George | Download | HTML Embed
  • May 24, 2005
  • Views: 10
  • Page(s): 292
  • Size: 10.74 MB
  • Report



1 Contents SECTION I: INTRODUCTION SECTION IV: TOWARDS DEPLOYMENT: OPERATOR STRATEGIES, IMPLEMENTATION Guest editorial, Leif Aarthun Ims ................................. 1 ISSUES AND OPERATIONAL ASPECTS Introduction to broadband access networks, Achieving global consensus on the strategic broad- Leif Aarthun Ims ........................................................... 2 band access network: The Full Service Access Network initiative, Alan Quayle and Jeff Stern ......... 166 SECTION II: MARKET AND REGULATORY ENVIRONMENT Towards broadband access in Europe the view from EURESCOM, Umberto Ferrero ...... 176 What is the willingness to pay for broadband services? Georg Moe and Jan-Petter Sther ........... 23 Towards broadband access in Japan ATM access for Mega-Media Services, Kenji Okada ................... 184 Long term forecasts for broadband demand, Kjell Stordahl and Lars Rand ..................................... 34 Towards broadband access in Norway the view from Telenor, Leif Aarthun Ims, Trygve Jarholt, Broadband demand survey in the residential and SOHO Kjell Stordahl, Frode B. Nilsen, Markku Lhteenoja, market in Norway, Synnve Istad and Kjell Stordahl . 45 Borgar Trre Olsen, Dagfinn Myhre and Stig Lken . 191 Regulation of broadband access networks, The challenge of civil works, Giovanni Ciochetto .... 202 Per Mognes and Terje Nord ...................................... 50 Measurements and maintenance of the future access networks, Angelantonio Gnazzo .................. 208 SECTION III: ACCESS NETWORK ARCHITECTURES AND TECHNOLOGIES Towards a practical implementation of DSL preparing for new technology and regimes, IP or ATM in the access network? Frode B. Nilsen ... 60 Per Klepsland .......................................................... 213 Wireline broadband access networks, Calculation of cable parameters, Per Klepsland ...... 226 Leif Aarthun Ims ......................................................... 73 Fixed broadband wireless access, SECTION V: BROADBAND ACCESS NETWORK Harald Loktu and Erwan Bigan .................................. 88 TECHNO-ECONOMICS Mobile broadband access, Techno-economic guidelines for telecommunication Rune Harald Rkken and Stein Wegard Svaet ......... 96 networks and services, Markku Tahkokorpi Satellite networks and stratospheric platforms, and Markku Lhteenoja ........................................... 236 Agne Nordbotten ...................................................... 107 OPTIMUM a techno-economic tool, The passive optical network (PON), Borgar Trre Olsen .................................................. 239 David Faulkner ......................................................... 113 Design of access network case studies, Flexible wavelength multiplexing techniques for Leif Aarthun Ims ....................................................... 251 broadband fibre access networks, Ton Koonen ...... 119 Broadband Internet Access Inverse multiplexing, Einar Edvardsen .................... 129 a techno-economic study, Ilari Welling ................. 254 Transmission on power line cables, The economics of broadband service introduction, Dermot Collins, Lucien Budry, Gbor Gerdai, Dagfinn Myhre ......................................................... 262 Angelantonio Gnazzo and Ahmed Madani .............. 134 Risk methodology for evaluating broadband Component technologies for wireline broadband access network architectures, Kjell Stordahl, access networks, Daniel Lecrosnier ........................ 138 Leif Aarthun Ims and Borgar Trre Olsen ................ 273 Management of access networks trends and Techno-economic risk assessment of PNO access challenges, Tor Breivik and Hkon Lnsethagen .... 145 network evolutionary paths, Nils Kristian Elnegaard, Leif Aarthun Ims and Kjell Stordahl ......................... 286 Home networks: New challenges for network operators, Markus Wyss and Frdric Pythoud ....... 160 Telektronikk Volume 95 No. 2/3 1999 ISSN 0085-7130 Editor: Ola Espvik Tel: (+ 47) 63 84 88 83 Status section Per Hjalmar Lehne Editorial board: editor: Tel: (+ 47) 63 84 88 26 Ole P Hkonsen, Senior Executive Vice President Oddvar Hesjedal, Vice President, Research & Development Editorial Gunhild Luke Bjrn Lken, Director assistant: Tel: (+ 47) 63 84 86 52 Graphic design: Editorial Telenor AS,Telenor R&D Design Consult AS office: PO Box 83, N-2027 Kjeller, Norway Tel: (+ 47) 63 84 84 00 Layout and illustrations: Fax: (+ 47) 63 81 00 76 Gunhild Luke, Britt Kjus, se Aardal [email protected] Telenor Research & Development

2 Editorial LEIF AARTHUN IMS 1999 is in many respects the market vision with substance. first year of broadband access Secondly, the recent technol- introduction on a reasonably ogy developments, in particu- large scale. The number of lar of high speed copper trans- cable modem installations mission systems and digital world-wide will probably compression technologies exceed one million this year. enable the provision of broad- Deutsche Telekom, Tele band as an overlay on the Denmark and Telenor are existing copper access plant at now offering ADSL (asym- a reasonable cost. Thirdly, metric digital subscriber line) new broadband services are services, as the first operators now gaining momentum as the in Europe. Interactive broad- prerequisite for maintaining band services and broadband margins with an increased portals are now emerging, competition. following the recent dramatic growth in Internet and mobile In this feature section we will services. Simultaneously, the cover a range of broadband rapidly increasing competi- access network related tiveness in the telecommuni- aspects, hopefully providing a cations market forces the valuable insight into the main actors in the field to minimise challenges of broadband costs and to maximise rev- access delivery. We start with enues. Broadband services an introductory chapter for the are widely recognised as reader not too familiar with potentially decisive for the the material. Initially, surveys capability of the actors to of market development and defend and eventually expand the current revenue base as they analyses of future regulatory environments are presented. This position for the future service battle. set of papers addresses the crucial questions of what will be the future demand for broadband services and how much are the The access network is the part of the telecommunications net- customers willing to pay for these services. work most closely related to service demand, but also the most cost sensitive network segment. Providing an access network The future broadband access network will be different from the with interactive broadband service capability requires infra- existing narrowband copper based network in three respects, structure investment levels expected to be comparable to the namely technology variety, open provisioning and service inte- investments associated with the establishment of the current gration. The second section of the feature issue gives a compre- access network for telephony. Moreover, the future interactive hensive coverage of the key issues related to architectures, start- broadband arena, and in particular the residential market, is ing with the impact of the IP development on the access net- characterised by a high uncertainty both with respect to service work evolution. The section includes an overview of wireline, take rates and willingness to pay. Accordingly, high risks are terrestrial wireless, mobile, satellite and stratospheric platform associated with upgrading this cost sensitive network segment. concepts and architecture options, including currently available Hence, developing a positive business case for reaching the cus- technologies, emerging technologies and migration paths tomers with interactive broadband access is probably one of the towards the future broadband access networks. Separate contri- most challenging tasks for the actors in the industry at the butions are devoted to some concepts of particular interest, like threshold of a new millennium. the passive optical network, wavelength division multiplexing, inverse multiplexing and power cable transmission. The crucial The vision of a multiservice broadband access network dates issue of access network management is addressed, as well as back at least to the early eighties, when the first optical fibre concepts for broadband home networks. fever spread. In fact, around 1990 most of the experts expected an evolution toward fibre to the home (FTTH) connections in As we move towards actual deployment of broadband access the access network. A lot of different field trials were launched technologies, implementation issues and operational aspects get and so-called killer-applications have been singled out, but increasingly important. We address these issues through some very little has happened when it comes to commercial broad- of the operators current access network strategies. Finally, in band access products for the mass market. So, what are the main view of the underlying fundamentals of where and when to reasons for this? The main inhibitors have obviously been the invest in order to create a positive business case for broadband high risks associated with placing the significant upgrade services, the last section gives a comprehensive coverage of investments in a market in which future services and revenues techno-economic methodology and application results, with the are highly uncertain. There simply has been no business case focus on work carried out within ACTS and EURESCOM. built so far for broadband access. So why should broadband delivery happen now? First of all, the explosive growth in Internet subscriptions and the likely evolution of the current Internet services capacity requirements into a demand for higher bandwidth has fuelled the broadband Telektronikk 2/3.1999 1

3 Introduction to broadband access networks LEIF AARTHUN IMS Currently the evolution of the access are presented and discussed, primarily during the last two decades. New tech- network towards broadband (capacity in the context of providing broadband nologies like the Internet, the explosion per customer 2 Mbit/s) is one of the service connectivity to the residential in mobile communications, the introduc- most crucial and demanding challenges market and small and medium enter- tion of optical fibres, development of in the battle between the actors in the prises. The intention is first and fore- satellite personal services are revolution- telecommunications industry. The issue most to introduce the major concepts ising the whole field of telecommunica- of establishing access networks for of broadband access networks to tion services. The market has recently broadband services is complex, both readers not too familiar with the topic, seen a great impact from new services in terms of the broadband service mar- through an overview with a modest and technologies coupled with the exten- ket, which is highly uncertain at pre- level of detail. sive market deregulation, for new and sent, and in terms of the future regula- even the traditional voice services. The tion of the customer access to telecom- explosive evolution of the Internet in par- munications infrastructure. Moreover, 1 Introduction ticular during the last years has a great there is a wide range of technologies impact on the telecommunications mar- available for broadband access, which Telecommunication networks and ser- ket in general. Both the traffic and the further complicates the issue of broad- vices constitute the backbone in the number of subscribers are now increasing band access. Starting from sometimes emerging information society. Indeed, exponentially. In addition, the number of very different existing networks the almost all recent technological and applications on the Internet are increas- access network providers have to select organisational innovations are dependent ing, and the functionality and content of the appropriate migration paths and on and sensitised by easy access to ser- the applications are under continuous broadband architectures and ensure vices provided by telecommunication development. New service providers a successful deployment, management networks. The information superhigh- entering into the marketplace are chal- and operation of the new access net- ways, like the motorways in the fifties, lenging the established service providers work infrastructure. And of course, electricity at the beginning of the cen- and threatening even the core business the bottom line is that the shareholders tury, railways in the nineteenth century, of traditional public network operators require return on investments, which are becoming the vital and prerequisite (PNOs) as they emphasise meeting cus- at present probably is the most chal- instruments for economic growth in a tomers needs as the key to success and lenging aspect of upgrading the very modern society. survival. cost sensitive access network to broad- band. In this article these main issues The telecommunications sector has ex- The customers are connected to the related to broadband access networks perienced dramatic technology changes telecommunications network and the information superhighways via the access network, which is the network infrastructure between the customer premises and the nearest local exchange (LEX), point of presence (POP) or cable network hub. The access network is often called the last mile or the local loop, as well. The dominating access networks today are: GEO satellite Twisted pair based networks, initially built for telephony services; Cable networks, intended for broadcast services; Mobile network base station Cellular radio networks for mobile telephony; Satellite networks for broadcasting. CATV Coaxial cable Headend Figure 1 gives a simplistic rendition of these four types of access networks. Interactive broadband services are now emerging, and none of the mentioned Twisted pair cable access network types have the capability PSTN local of providing interactive broadband con- exchange nectivity without a network upgrade. Thus all the access network providers with the intention of offering broadband services in some way or another must prepare their network infrastructures for these new Figure 1 The main access interactive broadband services. The access network types deployed today network is the most cost sensitive part of 2 Telektronikk 2/3.1999

4 the telecommunications network, and in tomer demands are crucial for the opera- such as the twisted pair based telephone addition the one most closely related to tors, service providers and equipment network, the coaxial cable network and service demand. Thus, the access network manufacturers. In this introductory article satellite networks for television distribu- operators of today are faced with the chal- we will address the main areas involved tion and the cellular network for mobile lenge of how to develop the existing net- in the evolution of the access network telephony. In particular the existing work infrastructure into a broadband towards broadband. The article is sec- twisted pair based local loop represents access network. Simultaneously, a rapidly tioned into eight main chapters, of which a significant asset for telecommunication increasing competitiveness in the telecom- the second one briefly sketches the scene operators, and is regarded as the key en- munications market forces the actors in of what is happening right now with abler for provisioning of new advanced the field to minimise costs and to max- respect to broadband access the most services. In chapter 5 the existing access imise revenues, most likely through a aggressive actors, some of the hypes and network infrastructure is described. simultaneous defence and expansion of rumours heard through the grapevine. the current revenue base. This initial status report will hopefully A large variety of access network archi- enable the reader to more easily relate tectures are available for the operators Ten years ago most of the experts ex- the following, more general chapters to and must be rigorously examined in pected an evolution towards fibre to the current situation. order to determine the most appropriate the home (FTTH). However, economic ones for the different area types and ser- analyses indicated that the overall project Chapter 3 addresses the main driver for vice demands. Chapter 6 gives a brief values of network architectures for the broadband network deployment, namely introduction to the access network tech- residential market are reduced signifi- the demand for new services. Provision- nologies and architectures that enable cantly when the fibre is installed close to ing of new, advanced services through the broadband service delivery, including the customer premises. FTTH solutions introduction of modern technology is wireline, wireless and satellite alterna- have proved very expensive, due to the commonly expected to be a crucial pre- tives. high cost of civil works and the low cus- requisite as the operators position for the tomer share of optics and electronics future service battle. However, at present The main inhibitors for the roll-out of equipment (in general, the cost of net- the telecommunications arena, and in par- broadband access networks have obvi- work elements decreases during time, ticular the residential domain, is charac- ously been the high capital investments whilst there is no reduction in civil terised by a high degree of uncertainty required in order to upgrade the existing, works costs). Even Nippon Telegraph with respect to the rapid technology evo- cost-sensitive access infrastructure or and Telephone Corporation (NTT) last lution, market development and regula- build new access networks, and the high year dropped their plan for fibre to every tory environment. Uncertainties in service risks associated with placing these signif- Japanese home by the year 2010, a pro- take rates, willingness to pay, future tech- icant investments in a market in which ject estimated at a total cost of USD nology capabilities, cost levels and tech- future services and revenues are highly 240 billion [1]. So, why then the current nology and market convergence introduce uncertain. No business case has simply increasing interest in broadband access? new and significant risk elements into been built so far for broadband access. Much of this increase in interest is due to telecommunications investment projects. In chapter 7 the economics and corre- the recent development of cable modems The key question is: how will the market sponding financial risks of delivering and digital subscriber line (DSL) modems. for residential broadband services evolve? broadband access are discussed. Access These new technology developments The policy of regulatory bodies will to a network upgrade strategies for emerging enable the utilisation of the existing large extent impact the evolution of the new broadband services have been evalu- twisted pair and coaxial cable base for broadband access delivery to the residen- ated in several studies, with technologi- high capacity transmission to the cus- tial and small and medium enterprise cal options ranging from enhanced cop- tomer as alternatives to the FTTH solu- market. The regulatory environment as per to hybrid fibre coax and broadband tions. The costs of cable modems and such is not addressed in this introductory passive optical networks. The main DSL modems have already decreased chapter. The topic is treated in more issues of broadband access economics significantly compared to a couple of detail in [2]. are summarised in this chapter. years back, and the price reduction is expected to continue over the next years The broadband access network evolution And finally, having taken the major as- due to mass production. This will impact must be considered in view of the current pects of broadband access into account, the broadband service prices, which trends in network evolution. Some of the challenge for the operators is to de- accordingly will fall to a level close to these trends even suggest that the old rive suitable minimum-risk strategies for the current narrowband connection tariffs local exchange network structure is a part either a migration of existing network and give the opportunity to extend the of the past and is fading into oblivion. infrastructures or for deployment of a present narrowband applications on What are the major trend projections, and completely new access network infra- Internet to broadband applications. how might they eventually impact the structure. Chapter 8 introduces the main evolution of the access network? Chapter criteria on which an access network strat- Cost effective, future proof broadband 4 briefly summarises the aspects related egy for broadband migration is devised. access networks accommodating a wide to network evolution which impact the range of demographic diversity for a set evolution towards broadband access net- The intention of this introductory chapter of services with different bandwidth works. is neither to provide a complete overview requirements will be required. Hence, of the complex aspects of broadband strategies for developing the access net- The future broadband access network access, nor to give a thorough description work, along a cost effective path, flexible infrastructure will largely have to be de- of the respective fields of market, regula- enough to serve a complex set of cus- veloped from the existing infrastructure, tion and technology addressed herein. Telektronikk 2/3.1999 3

5 140 US [7, 8]. In May this year AT&T bought MediaOne, almost doubling its XDSL 120 number of cable customers, to reach Cable Modem Millions of subscribers approximately 25 millions. MediaOne 100 Satellite was bought following the acquisition 80 LMDS announced in the summer of 1998 of TCI, then the largest cable operator in the 60 US. This is now known as AT&T Broad- 40 band and Internet Services. The cable networks of TCI and its affiliates pass 33 20 million homes in the US. In February this year AT&T and Time Warner announced 0 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 a joint venture, which will enable AT&T Source: Pioneer consulting to reach directly into another 20 million US homes [9, 10]. This additional joint venture with Time Warner will give Figure 2 Global broadband access subscribers by technology, 1998 2007 AT&T access to cable networks which in total pass over 70 million homes in the US, that is around 60 % of the house- holds in the country. Over the past few The main objective is to provide a brief in relatively large volumes during 1999, years, up until AT&Ts seemingly deter- introduction to broadband access, at the it will of course still be at rather modest mined entrance into the local loop market level of detail considered sufficient for penetration levels compared to traditional by the use of HFC, the interest world- the reader unfamiliar with the subject or services such as POTS (plain old tele- wide in HFC technology has decreased. for the reader in need of a condensed phony services) and CATV (community These moves by AT&T are expected to synopsis on the matter. For a more de- antenna television). At the end of 1998 re-fuel the development of the HFC tech- tailed treatment of the aspects, we refer there were approximately 800,000 sub- nology (hybrid fibre-coaxial) and may to the articles in the remainder of this scribers globally with a broadband access prove decisively important for the cable special issue of Telektronikk on broad- connection [6], with 570,000 cable operator and manufacturer industrys band access networks. modem subscribers, 200,000 DSL sub- capability to enter the residential broad- scribers and approximately 30,000 broad- band markets on a large scale. band customers with satellite access. 2 Broadband access One forecast for broadband access con- Simultaneously, the DSL market in the right now nections for the next ten years is shown US continues to develop with rapid pace. in Figure 2. According to this forecast, Currently US West has by far the largest This year 1999 is pretty much set by the year 2007 the number of broad- base of installed DSL connections in the to be the first year of broadband access band access subscribers world-wide on US, probably more than 80 % of the total introduction on a reasonably large scale XDSL, cable modems, LMDS (Local number of lines [11]. In March this year in some developed markets throughout Multipoint Distribution System) or satel- Covad Communications announced the the world. In the United States in particu- lite will be around 140 million. first US nation-wide DSL network, offer- lar the number of installed cable modems ing services in the range from 144 kbit/s will reach really significant volumes this Of the many forces and incidents impact- to 1.1 Mbit/s symmetric access [12]. year, already an estimated 650,000 to ing the broadband access market, proba- SBC Communications Corp. plans to 750,000 such modems have now been bly four of the most influential ones right offer high-speed Internet access to 9.5 installed in North America, mainly for now are: million potential customers across most Internet Access services [3]. Several of of the operating region by the end of this AT&Ts acquisition of TCI (Telecom- the larger telephone companies in the US year [13]. Microsoft recently began trials munications Inc.), MediaOne and joint have plans for large-scale DSL (digital of their high-speed digital subscriber line venture with Time Warner; subscriber line) roll-outs, and this year in four large US cities, providing con- we will for the first time see broadband ADSL offerings on a commercial basis sumers with speeds of up to 8 Mbit/s connections being offered via ADSL in North America and Europe; [14]. The DSL roll-out in Europe has (Asymmetric Digital Subscriber Line) on been somewhat slower than in the US, Personal computers with factory a commercial basis in Europe, after years but this year ADSL offerings on a com- installed DSL modems; of unfulfilled promises of broadband ser- mercial basis in Europe are expected by vice delivery by the incumbent operators Broadband portals emerging on the some of the major incumbent operators [4, 5]. Internet, offering broadband content. [15]. Telenor, Tele Denmark and Deutsche Telekom are already offering This happens after years of marketing of Also the access arena has lately been commercial ADSL services. BT plans to broadband technologies and services dominated by the general trend in the install ADSL equipment in 400 of their without any significant materialisation in telecommunications industry of rather exchanges within the next year [16]. On terms of service offerings in the residen- frenetic merger and acquisition activities. July 6 this year Oftel, the UK telecom- tial and SME markets. And one should During the last year AT&T through munications regulator, proposed to open bear in mind that even if new broadband mergers and acquisitions has grown to up BTs twisted pair local loop network technologies are expected to be installed become the largest cable operator in the by the year 2001, implying that BT is to 4 Telektronikk 2/3.1999

6 provide other operators with high capac- 3 The market evolution The future consumption patterns are ity access via a point-to-point data ser- likely to have a great impact on the evo- vice between the end user and the service Even if the accelerated development lution of broadband applications and ser- provider [17, 18, 19]. In Australia, the of broadband applications is the most vices and correspondingly the need for operator TransAct in a field trial already important factor for the evolution of a broadband access. But where can we get uses the next-generation DSL, VDSL, to broadband market, and the current status indications on the future behaviour of the offer telephone, data and video services of broadband access seems to indicate consumers? The recent Internet and to some hundred users with connection that the somewhat turbulent market is mobile communication developments capacities of up to 36 Mbit/s [20]. developing rather fast, the future market have shown that the early adopters of evolution is still quite uncertain. In the the new ways of communication and the The integration of broadband capacity mid-nineties there was a strong belief in enabling technologies typically are peo- modems into end-user equipment such as the need for one killer application, and in ple characterised by a high acceptance PCs, is likely to boost the broadband particular video on demand (VOD) was of new solutions combined with a strong access market. The Universal ADSL con- focused as a potential killer application. desire and need to get rid of constraints sortium was formed in January 1998, led However, the market, including the resi- caused by limitations in space and time by Intel, Microsoft and Compaq Com- dential broadband market, has proved to [32]. And in the last years developments puter [21, 22]. The consortium developed be more sophisticated than earlier indicate that the impacts and benefits of the G.Lite specification, which trades assumed. It is obviously difficult to pre- new services for the larger, general pub- transmission capacity for splitterless dict the killer application(s). Moreover, lic market in turn are identified by these installation and thus enables plug-and- there are reasons to believe that in the somewhat entrepreneurial consumers. play solutions in terms of eg. PCs with future the demand for residential broad- The general development, including factory installed DSL modems. PCs with band connections will be created by a enabling technologies, in the information DSL modems are expected to reach the wide range of applications rather than and communications industry is likely to market in significant numbers this year. a single one. In this chapter, we will dis- have a large impact on how, when and Dell Computers now markets a new per- cuss three main issues related to the where broadband services are applied sonal computer with a pre-installed broadband access market evolution, [33]. Very soon the content and distribu- ADSL modem, capable of speeds of up namely the future consumption or usage tion are expected to become all-digital. to 768 kbit/s [23]. The Bell companies patterns, the application demand and The exponential growth in micropro- project the number of installed DSL lines willingness to pay for new broadband cessor power, memory size and storage this year to be 200,000 [24]. But also application and services. capacity will probably continue for the other actors are now making their moves into broadband tailored end-user equip- ment. In May this year Microsoft bought a 2 % stake in AT&T, simultaneously reaching an agreement with AT&T on the use of Windows CE software in up to ten million of the set-top boxes to be in- stalled in connection with the broadband upgrade of AT&Ts cable networks [25]. And finally, the content providers are now seemingly at the entrance into the broadband era. Broadband portals such as and Snap! stream live and on-demand audio and video pro- grams over the Internet [26, 27] amongst delivery of other content (Figure 3). They even offer dedicated services to users with connection speeds of 128 kbit/s or higher [28]. A next likely move is part- nerships between such portals, including the dominating ones like Yahoo! and access network operators with broadband access roll-out plans [29, 30]. It may be worth noting that one of the owners of Snap!, NBC, entered into an agreement with AT&T Broadband and Internet Ser- vices in June, paving the way for distri- bution on NCBs programming proper- ties on AT&Ts cable networks [31]. Figure 3 Broadband portals such as are now emerging on the Internet Telektronikk 2/3.1999 5

7 Communication Present Future demand Change in type demand capacity Real-time demand Human-to-human Speech videoconferencing, interactive video-programming, interactive work, messages with video telemedicine Human-to-machine Download of text Movie downloading, and still pictures distance eductaion, financial transactions, teleshopping, video-on- demand Machine-to-machine File transfer File transfer fig.4 (increased volumes), backup, EDI Figure 4 The development towards a demand for higher access capacity next ten years. This implies that the per- 3. Entertainment services are increasingly Given these indications of future applica- formance of the end-user systems will important. tion and access capacity requirements, continue to increase, possibly to 100 the key question is how large a growth times the performance of the current sys- Figure 4 illustrates a possible develop- in demand for transmission capacity can tems. Resulting new consumption pat- ment from the present towards the future be expected within the next five to ten terns and needs are thus likely to be seen, demand for applications. The types of years? New service forecasting and driving a demand for new services and communication are classified as human- demand projections have recently been applications. to-human, human-to-machine and reported, derived from current spending machine-to-machine. The present appli- patterns of households or from market However, forecasting the new broadband cation demands and usage patterns are surveys. According to several studies, services with respect to both application different for these three types of commu- like the RACE/TITAN and ACTS OPTI- type and demand is very difficult, mainly nication, and so might also the future MUM Delphi surveys [35, 34], the work due to the lack of historical data. In addi- application demands be. However, they by the Bureau of Transport & Communi- tion most of the applications are new and all have that in common that expected cations Economics, Canberra [37] or the it is difficult to predict growth directly. evolution of applications leads to a FSAN initiative [38], and several con- Nevertheless, some surveys have in a demand for higher access capacity, even sulting reports [30, 39] there will be a systematic way aimed at identifying the if the differences in usage patterns are significant future demand for asymmetric new applications [34, 35]. Despite such likely to remain also in the future. and symmetric broadband services in the market surveys, the current market status residential and small business market and recent field trials, the new broadband The different usage patterns will imply segments. These findings are supported customer applications, in particular in the variations in bandwidth requirements of by main findings in several other reports, residential and small business markets, the future services, as illustrated in Fig- which conclude that broadband services remain ill-defined. Nevertheless, some ure 5 for residential services [36]. Some will reach the mass market soon after the trends can be recognised: applications and services are likely to turn of the century with cable modems as be highly asymmetric in nature, whilst the dominant access method (at least in 1. The integration of telecommunications others will require a more symmetric set- North America) at the same time as con- and information processing is escalat- up of capacity between consumer and net- tent and service providers will create new ing. work provider, and vice versa. Further- services for the broadband customers 2. The use of interactive video services, more, this might to some extent govern [40]. Forrester Research in a recent such as videoconferences, is growing. which infrastructure platform the ser- report estimates that almost one third vices are provided on. of all on-line subscribers in the US will 6 Telektronikk 2/3.1999

8 have cable modem or ADSL access in 2003 [30]. High (1.5 Mbit/s or Symmetric greater) broadband networks The third main market issue is the will- ingness to pay for new broadband appli- Video Interactive Games cations and services, and the crucial conferencing video programming question of is there any incremental will- Two-or-more simultaneous ingness to pay for new broadband, multi- interactive sessions Bandwidth required from media services? In this context multi- consumer to provider media services comprise voice, video Asymmetric and data services. The demand level will broadband networks to a large extent depend on the price and Financial Video on the pricing schemes, in particular in the transactions demand very important mass market, which Home shopping mainly is discussed here. Some studies indicate that the incremental willingness to pay compared to current services may Telephone lines Satellite and cable be very limited, particularly in the con- Basic Information Pay-TV Pay-per-view Public TV sumer market [34, 41, 42]. This is proba- telephony services bly the main challenge in the access Commercial TV operators strife for achieving returns Low on the huge investments required in net- Low Bandwidth required from provider to consumer High work upgrades for broadband access delivery. In addition, not only the price Figure 5 Bandwidth requirements of current and future residential services [36] level, but also the pricing structures are likely to be very important. Based on a reasonably large scale experiment mea- suring the demand for Internet as a func- tion of bandwidth, applications and pric- ing structure, the Internet Demand Experiment (INDEX) confirms that applications; work components and the expected Internet users are sensitive towards dif- future cost evolution. The price paid by technology; ferent pricing structures. The INDEX the users for the given applications and project proposes usage-based pricing as a network platform; services in turn depends on investment fair way to charge customers [43]. costs, operation and maintenance costs service quality; and the revenue considerations of the The future Internet market, including cost evolution; network operator. The application de- an eventual emergence of wide-scale mand is determined by factors such as demand; offered broadband Internet connections, the expected competition, the market is also expected to be segmented into price; potential for the applications, expected quality-differentiated service portfolios, market shares, substitution effects be- environment; implying a mixed set of price structures tween applications, penetration as a func- for the segments. Customer group seg- strategy/policy. tion of time and the price the service mentation and segmentation into geo- quality. In addition there are interactions graphic areas are the two other important The strategy of the network operator is between the main factors, as shown in segmentation aspects related to broad- governed by revenue estimates and the the figure. band access delivery. targeted and expected return on his investments. In general, the introduction The current network trends towards mul- of new applications, new technology, timedia broadband are influenced by the 4 Network evolution new network platforms, new architec- general technology evolution, which typ- tures, etc. are likely to depend on the ically characterised by digitalisation, This telecommunications network devel- long term revenue prospects. However, miniaturisation, high capacity and mobil- opment in general obviously impacts the with increasing competition strategic ity, the latter both in terms of terminal, evolution towards broadband access net- decisions play an important role also in user and service mobility. And indeed, works. Thus, some of the major trends in the near term positioning. Figure 6 illus- we are currently witnessing a change in network evolution will need to be briefly trates the network platform options for network technology with respect to price introduced here. Many complex and introduction of new applications, namely and functionality, for instance through interacting factors have an impact on the further utilisation of the existing network the introduction of WDM (wavelength demand for new services and accordingly platform, expansion of the network plat- division multiplexing) ATM (asyn- the network evolution. Figure 6 gives a form or by introduction of new network chronous transfer mode) and IP technol- brief overview of the complexity [44]. technology. ogy, often considered as key enabling The main factors, as illustrated in the technologies for the future full service figure, may be considered to be: The preferred alternative will among integrated network, which is a vision that other factors depend on the cost of net- dates several years back. Furthermore, Telektronikk 2/3.1999 7

9 APPLICATIONS hierarchy (PDH) or synchronous digital hierarchy (SDH) transmission systems in Video applications addition to radio links in rural areas. But Data transfer still, network operators commonly intro- Enterprise network duce new services by means of a new Information services Transactions service network with dedicated equip- Combined applications ment. This may very well be the case for several years to come, resulting in an incoherent mixture of technologies remaining also in the future [45, 46]. TECHNOLOGY NETWORK PLATFORM SERVICE QUALITY Digitalisation PSTN / ISDN MTTR However, in this paper the evolution Optical Network PSDN / CSDN Performance towards broadband access will be dis- Technology Frame Relay Reliability cussed in view of a development towards DSL-technology ATM Operation and the future telecommunications network SDH-technology Internet maintenance which may be considered as representing ATM-technology Leased line a paradigm shift from dedicated service Radio-technology CATV networks towards a common service pro- Satellite-technology Satellite duction platform utilising a set of parallel access network technologies, such as twisted pair, coaxial cable, cellular, COST EVOLUTION DEMAND PRICE radio and satellite. This is illustrated in New technology Market potential Market share Figure 7. This evolution represents a Standardisation Penetration Cost based shift towards network independent ser- Competition Demand curves Market based vice production, mainly based on soft- Production Substitution effects ware. The result will most likely be a improvements Competition fundamental change of roles within telecommunication, in which product development and customer relations will be taken over by other than todays suppliers and operators. ENVIRONMENT STRATEGY / POLICY Regulation Strategic positioning Thus, given this shift towards a common Standards Introduction of applications service production platform utilising a set USO Investment strategies of parallel access network technologies, ONP Return on investments LIberalisation New network platforms the future broadband access network will Competition Revenue be different from the existing narrow- Interconnect band copper based network in three respects, namely technology variety, openness and service integration [47]. Figure 6 Key factors for demand and network evolution [44] Hence, the architectural requirements of the access network are likely to change along these three dimensions. This will most likely lead to the end of a single access network architecture in terms of the network evolution towards multi- port network at long distance and junc- underlying technologies. Future broad- media broadband is influenced by the on- tion level and also in parts of the access band multiservice access networks will going convergence of the traditional mar- network. The technology used today is probably be built on combinations of sys- kets for broadband services, namely the mainly fibre and plesiochronous digital tems and technologies, including fibre data communications market, the broad- cast/video market and the telecommuni- cations market. The rigid boundaries between these traditionally separated service markets are already blurred, and Service production/IT look set to disappear over the next few PSTN/ISDN Cable-TV years. Thus, both the technology evolu- Service network Satellite Mobile WWW Data tion and the service market evolution may be taken as indicators of networks Transport evolving into integrated broadband ser- vices networks. Access The telecommunications network opera- tors have already integrated the traffic Figure 7 The shift towards a common service production platform utilising from the different platforms in the trans- a set of parallel access network technologies 8 Telektronikk 2/3.1999

10 solutions, satellite systems, cable-TV 5 The existing access The above network infrastructure types networks, broadband radio systems and are either broadband access networks DSL systems on the existing copper net- network infrastructure without interactive capabilities at present work. However, in a competitive envi- (eg. the coaxial cable networks and satel- The future broadband access network ronment with a variety of network opera- lite networks), or narrowband networks infrastructure will to a large extent have tors the access operators will be required with interactive capabilities (eg. twisted to be developed from the existing infra- to establish open network architectures. pair networks or cellular radio networks). structure such as the twisted pair based And lastly, the vision of the integrated That is, the cable operators usually have telephone network, the coaxial cable net- multiservice access network requires ser- a starting situation for a migration work and satellite networks for television vice integration in terms of introduction towards interactive broadband access distribution, and the cellular network for of statistical multiplexing in the access which is significantly different from the mobile telephony. The dominant tele- part of the network. situation of the telephone operator. The communication access network infra- twisted pair network of the telephone structure is still based on twisted pair Before we briefly introduce the different operator has a point-to-point topology copper cables for telephony services and broadband access technology options, we and the coaxial cable network has a dis- Internet access. Coaxial cable networks will in the next chapter address one of tributive topology. Nevertheless, inter- and direct to the home (DTH) satellites the most influential aspects with respect active broadband services may not be support the residential market with dis- to selection of the future broadband implemented in any of the current access tributive broadband applications. access architecture, namely the existing networks without an upgrade of the infra- Recently some cable operators have infrastructure. structure. However, the differences in upgraded their coaxial cable networks existing access networks may call for with return capabilities in order to pro- In summary, the access network operator quite different upgrade strategies. In this vide Internet access. Cellular networks will be faced with a bewildering choice chapter the main features and differences for mobile services have been deployed of alternatives. The preferred architecture in existing access networks will be out- in large numbers over the past years, and will depend on considerations of several lined. constitutes a powerful starting point for factors, of which the key ones are: the development towards broadband The regulatory regime; access. In summary, the main access 5.1 Twisted pair access The competitive environment (com- network platforms applied today are: networks moditisation, price reduction; Twisted pair networks for PSTN, The telecommunication access network ISDN and leased lines services; connects the telephone set at the cus- The services to be provided; tomer premises to the local exchange Coaxial cable networks for digital Market segmentation; (LEX) physically through a pair of cop- and analogue broadcasting; per cables, referred to as a twisted pair. The existing plant; Geostationary satellite networks for Approximately 600 million twisted pair Replacement strategy; digital and analogue broadcasting and access lines have been installed world- business communications; wide [48]. A typical structure of the System costs; existing twisted pair access network is Cellular radio networks for mobile Financial strength in terms of cash illustrated in Figure 8. services; flow and capital funding. Wireless networks for PSTN. MDF NT1 ISDN LEX POTS Primary Secondary Customer Local exchange flexibility point (PF) flexibility point (SF) premises

11 The twisted pair based access network nection between the cable side and the 2 Mbit/s provided on optical fibres. This consists of a main distribution frame exchange side. Thus, from the customer is mainly due to the digitalisation and the (MDF), a primary distribution cable, sec- premises to the exchange, each customer high demand for primary rate integrated ondary distribution cable and drop cable. has his own pair of copper cables dedi- services digital network (ISDN), leased Flexibility points (connection points), cated for his usage. lines and data communication services. often located in street cabinets, are intro- duced in the network in order to provide At present several access network opera- flexibility in network evolution. The drop tors have completed their digitalisation of 5.2 Coaxial cable networks cable (usually two or more pairs) is in- the exchanges. However, the access net- The cable operators typically have a co- stalled between the customer premises work infrastructure which connects the axial cable distribution network which and is terminated in a flexibility point, or customers to the local exchanges has not distributes television and radio pro- distribution point. The flexibility point is changed significantly during the digitali- grammes from a central receiver loca- located in a small cabinet installed either sation process. In the original roll-out of tion, often referred to as CATV (commu- at street corners, on poles, or in the base- the network the reach of the network was nity antenna television) networks. The ment of building blocks. Usually be- typically set to between four and six kilo- signals are fed from a receiver station, tween ten and 30 customers are con- metres, with allowances for longer dis- called headend, which passes some tens nected to this flexibility point. Also ter- tances in more sparsely populated areas, of thousands of homes, to the distribution minated in this flexibility point is a larger using thicker cables. In urban areas today network. The headend may serve several single cable with a cable pair size depen- fibres are deployed not only in the junc- hubs which typically pass some thou- dent on the number of customers con- tion networks but also in parts of the sands of homes. The distribution network nected to the flexibility point. This larger access network. Fibre loops with service is usually a coaxial cable network, but cable, secondary distribution cable, con- access points (SAPs), transmission more and more often fibreoptic cables nects the secondary flexibility point to equipment and remote subscriber units and combinations of coaxial and fibre- a primary flexibility point closer to the are established in order to increase relia- optic cables are used. Optical fibres are local exchange. Several secondary distri- bility and flexibility. The SAP refers to usually used for the transmission of sig- bution points (in the order of tens) are the localisation of the concentrator and/or nals from the headend to the hubs, whilst terminated in this primary flexibility add and drop multiplexers in the net- the network between the hub and the sub- point. The primary flexibility points are work, as commonly used in European scriber is a coaxial cable distribution net- terminated in the local exchange via pri- countries with advanced access network work, with bandwidths of 300 MHz, 450 mary distribution cables, with cable pair infrastructures. The development of the MHz, 606 MHz, 750 MHz or 860 MHz. sizes dependent on the number of flexi- new architecture is advancing by the Figure 10 shows a coaxial cable network bility points connected to each primary introduction of SDH technology and add for distribution of television services. In flexibility point. In the local exchange, and drop multiplexers in the local loop. the coaxial cable part of the network the the primary cable pairs are terminated in Shown in Figure 9 is a typical public signals are amplified, selected and fre- the main distribution frame, which is a switched telephone network (PSTN) quency converted by the use of coaxial rack connecting the primary cables to the access network infrastructure with ser- amplifiers. All of the customers on the local exchange equipment. The customer vice access points. In addition, many coaxial cable branch receive the same is connected to subscriber equipment business customers are now connected signal. such as a line card through a cross con- to the access network with multiples of SDH-ring MDF LEX NT1 ADM ISDN ADM RSS/ RSU POTS 155-622 Mb/s Primary Secondary Customer Local Service access point (SAP) flexibility point (PF) flexibility point (SF) premises exchange

12 D1- amplifiers D2- ODF amplifiers 300-860 MHz CATV EO OE Router D0 D1 D2 D3 network network network network Customer Headend Hub D2 D3 premises Figure 10 Existing coaxial cable network for distributive services The distribution network depicted in Fig- lites are located at the geostationary earth good starting point for the development ure 10 consists of four segments, the D0- orbit (GEO), approximately 42,000 km towards broadband access service deliv- D3. The D0 network distributes the opti- from the earths centre, and normally ery. Figure 11 shows a common access cal signals from the headend, typically function as analogue transponders [51]. network infrastructure established for over long distances, to the network hubs. GSM and NMT networks. The base The D1 network consists of coaxial The main application so far has been station controller (BSC) depicted on the cables and amplifiers and may be con- broadcasting. The satellite systems are left-hand side may be co-located with sidered as corresponding to the primary very different from cable based access the PSTN local exchange. The traffic be- network in the twisted pair based access networks, since they offer coverage over tween the base station controller and the network infrastructure. The number of very large geographical areas and may base station (BS) may be transmitted by amplifiers in this network segment is operate without any terrestrial infrastruc- the use of fixed lines (twisted pairs or determined from the distance in the D1 ture [52]. Each satellite typically has fibreoptic cables) of capacities in the segment. The D2 network, which similar some hundred MHz of available spec- range of up to 2 Mbit/s, or alternatively to the D1 segment contains coaxial cable trum, with cell sizes of commonly some by the use of radio links. Several base sta- amplifiers, may be compared to the sec- hundreds to thousands of kilometres in tions may be connected to the same mul- ondary network in the twisted pair based diameter. Thus, the average individual tiplexer (DXX), which is typically located access network. The D3 segment often channel user density supported is low at a network level comparable to the pri- consists of individual coaxial cables, compared to terrestrial access networks. mary or secondary flexibility point of the connecting the customer premises to the Furthermore, the two-way delay in satel- twisted pair access network. Present cell distribution point. The D3 distribution lite systems is about 0.25 seconds, sizes in GSM 900 networks in densely point serves in the order of ten to 30 cus- caused by the long transmission distance. populated areas are typically some few tomers. The D3 segment may be looked In addition there is a significant propaga- hundred metres. The development of the upon as corresponding to the drop cable tion path loss. By the use of very small current mobile networks towards broad- in the twisted pair network. Today an aperture terminals (VSAT) geostationary band is described in more detail in [53]. increasing number of cable operators are satellite networks may be established for offering Internet access over their coaxial wideband access, point-to-point commu- cable networks. This requires an upgrade nications or for star or mesh type of net- 5.5 Wireless local loops of the described coaxial cable network in works. However, the user terminals are Today wireless local loop (WLL) and order to enable return channel signalling, expensive and high bitrates require very radio in the local loop (RLL) systems are as described in [49]. large antennas. Satellites can be launched deployed for the provision of narrowband into orbits that are closer to the earth in or broadband services intended as a re- order to overcome these problems. These placement for the conventional copper 5.3 Satellite networks systems are described in more detail in loop. In countries with a developed Geostationary communication satellite [52]. twisted pair infrastructure point-to-multi- networks at present include fixed service point radio access systems are mainly satellites (eg. for television distribution, used where network rehabilitation or distance learning and data communica- 5.4 Mobile access networks extension by the use of twisted pair cable tions), DTH broadcast service satellites The cellular networks deployed over the based solutions require high investments. and mobile service satellites (voice ser- past years for mobile services, such as the However, in terms of installed systems in vices and digital communications) [50]. GSM networks, comprise access network developed countries, wireless local loop The geostationary communication satel- infrastructures which may constitute a alternatives are very few compared to Telektronikk 2/3.1999 11

13 GSM LEX DXX MDF Modem Modem GSM BSC Modem NMT HDSL HDSL Modem NMT Local Primary flexibility point (PF) Base station Customer exchange Secondary flexibility points (SF)

14 commercially available for some time ity per simultaneous user, rather than the The DSL system technologies include already. maximum available capacity. IDSL (ISDN digital subscriber line), ADSL, VDSL (very high-speed digital The technology variety is illustrated in Some of the most challenging issues of subscriber line), HDSL (high bit-rate dig- Figure 13 in which some of the relevant broadband access networks are related to ital subscriber line), SDSL (symmetric technologies are sorted by transmission management systems and deployment of digital subscriber line) and CDSL (con- medium. The five main wireline upgrade new technology. These two aspects are sumer digital subscriber line, also known alternatives are shown in the figure, not treated here, but addressed in more as G.Lite or ADSL Lite) [63, 64]. At pre- namely power line modems to the far detail in [55] and [56, 57, 58] respec- sent only ADSL and HDSL of the broad- left, coaxial cable modems and HFC tively. band alternatives are commercially avail- technology, digital subscriber line (DSL) able, and have already been installed by modems and fibreoptic systems. As we several operators. VDSL and SDSL are move further up into the air there are sev- 6.1 Wireline broadband access expected to become available quite soon. eral terrestrial wireless alternatives, such There are basically four types of wireline ADSL Lite offers lower capacity than as local multipoint distribution system transmission media and associated sys- ADSL with a potential cost reduction (LMDS) and mobile systems such as tems available for broadband access net- obtained through simpler and more UMTS. Several satellite systems and works today: robust customer premises equipment concepts have been presented, including which the customer himself may install. Twisted pair cable systems; geostationary satellites (GEO), mid earth The concept of parallel DSL systems, Coaxial cable systems; orbit (MEO) systems and low earth orbit called inverse multiplexing, has been Fibreoptic cable systems; (LEO) systems. The so-called stratos- proposed as a further development of Powerline systems. pheric platforms such as HALO and DSL technology, enabling an aggregate SkyStation are based on high altitude capacity of up to 155 Mbit/s or even 622 Wireline broadband access systems are balloons or aeroplanes. Mbit/s and can be transmitted between treated in more detail in [49, 59, 60, 61 two network locations by combining a and 62]. In this chapter only a brief intro- The assessment of the different access set of twisted copper pairs [62]. duction will be given. Until recently cop- network architecture alternatives must be per pair has been considered to be a sig- based on a consistent evaluation of a set The existing coaxial cable network may nificant bottleneck with respect to capac- of criteria, among other things including be upgraded to interactive broadband ity. Recent developments of complex performance, cost effectiveness (installed capability by the use of cable modems. modulation schemes have enabled the first costs and running costs), technologi- A cable modem is installed at the cus- extension of the line capacity by order of cal maturity and flexibility. The trans- tomer premises, and the coaxial network magnitudes. In the short term the main mission capacity is probably the most is upgraded with return amplifiers in advantage of copper is a variable cost important aspect regarding the perfor- order to provide two-way transmission. option, alleviating the need for high and mance evaluation, ie. which capacity in This is commonly combined with a seg- risky upfront investments. DSL (Digital Mbit/s may or will be provided to the mentation of the coaxial cable network Subscriber Line) deployment consists of customer on the chosen architecture? into smaller segments by introduction of fitting DSL modems at the customer Some of the architecture alternatives, optical fibre cables. This is called HFC premises and at the local exchange side, such as the satellite systems, coaxial technology (HFC: hybrid fibre and co- and utilising the installed twisted pair cable modem systems and some broad- axial cable network). The coaxial cable base. The enhanced copper or DSL tech- band radio systems, are based on a shar- network segment is shared among the nologies differ with respect to transmis- ing of the transmission capacity between sion capacity, transmission distance and the customers. For other alternatives, the number of copper pairs used. In gen- however, such as ADSL (asymmetric eral for the DSL options, there is a trade- LEO digital subscriber line) and VDSL (very off between distance and capacity [48]. high-speed digital subscriber line), the Globalstar Teledesic specified transmission capacity is exclu- Skybridge MEO sively available to each simultaneous user. The maximum available capacity Skystation GEO Iridium per user under ideal circumstances may HALO Odyssey be very high for eg. satellite systems and Astrolink coaxial cable modem systems, with up GSM DECT to several tens of Mbit/s. The costs asso- APON ciated with a network dimensioning FTTC UMTS/UTRA which guarantees this maximum avail- FTTN BPON MMDS MBS DCS1800 able capacity per simultaneous user are MVDS FTTB LMDS likely to be prohibitively high. Thus, the PON network is likely to be dimensioned for Powerline FTTH available capacities per simultaneous modems FSAN ADSL user which are significantly lower than HDSL HFC IDSL G.lite the maximum capacity. Hence, a consis- SDSL tent evaluation of the different access Cable modem IMUX network architecture alternatives must be ISDN VDSL CDSL based on the guaranteed available capac- Figure 13 Access network alternatives Telektronikk 2/3.1999 13

15 connected users [49]. Several hundred 6.2 Wireless broadband radio service), which depending on the thousand cable modems have already access number of time slots and the coding been installed in North America for scheme used, may enable maximum data The remarkable growth of wireless com- Internet access. AT&Ts acquisition of rates of 170 kbit/s. GPRS may be intro- munications over the past years, includ- TCI and MediaOne, and the joint venture duced without significant changes of the ing terrestrial radio systems and satellite with Time Warner may have a significant radio interface in the currently deployed systems, has led to an increased interest impact on the development of broadband GSM networks. The GPRS service may in wireless technologies for broadband access technology for coaxial cable net- be further enhanced by EDGE (enhanced access [66]. The interest in wireless works. data rate for GSM evolution), with the broadband solutions is very high at pre- potential of 384 kbit/s packet services sent, and a variety of concepts and sys- Introduction of fibre in the loop (FITL) using GSM. However, EDGE will re- tem alternatives have been proposed and requires the deployment of fibreoptic quire an upgrade of the radio interface, are currently under research and develop- cable. Both the associated civil works and may not become a reality before ment. In the short to medium term LMDS costs and equipment costs are inhibi- third generation mobile systems are (local multipoint distribution system) is tively high at present, and only to a very introduced. At present the next genera- expected to be the most relevant technol- limited degree have fibreoptic systems tion mobile system is in the making, ogy for wireless asymmetric broadband been introduced in the access network called UMTS (universal mobile telecom- services [67]. LMDS requires line of over the past years. There is a variety munication system). UMTS may enable sight. Analogue LMDS systems are amongst FITL architectures regarding several hundred kbit/s of transmission available today, which utilise the 27.5 node configurations, ie. the location of broadband capacity to outdoor mobile 29.5 GHz frequency band yielding a ser- the optical transmission terminal equip- terminals, and up to 2 Mbit/s symmetric vice area radius of between 5 and 10 ment. The most common configurations transmission capacity indoors. The intro- kilometres. Future digital LMDS systems are fibre to the cabinet (FTTCab), fibre duction of UMTS will require the estab- will have higher capacity than the ana- to the node (FTTN), fibre to the curb lishment of new base stations in the net- logue LMDS systems. At present only (FTTC), fibre to the building (FTTB) and work, most likely with a higher density pilot systems are in operation in the 42 fibre to the home (FTTH). Fibreoptic than the currently deployed GSM base GHz spectrum. Commercial systems are transmission systems are utilised in con- stations. Mobile broadband access sys- expected to be available during 1999. nection with ATM and SDH and may be tems are treated in more detail in [53]. Currently the maturity of the broadband configured as point-to-point connections, radio access systems lags behind their ring structures or as point-to-multipoint connections (SDH PON or BPON). The wireline counterparts, at least if signifi- 6.4 Satellite systems for FSAN initiative, in which all the major cant broadband take rate capabilities are broadband services considered. Nevertheless, the broadband telecommunications operators in the Satellite systems and concepts for inter- radio access solutions may in due time world are participating together with the active broadband services include geo- prove to be a key technology in the largest equipment manufacturers, has stationary satellite systems (GEO), mid emerging broadband market, mainly worked out specifications for an access earth orbit (MEO) systems and low earth because they enable potentially very network based on a fibreoptic transmis- orbit (LEO) systems. The satellite sys- rapid network roll-outs, low capital costs sion in combination with DSL technol- tems for broadband services are de- compared to wireline alternatives in ogy [65]. The most aggressive vendors scribed in more detail in [52]. The major- sparsely developed areas, and flexibility plan to deliver FSAN compliant equip- ity of these satellite systems for wide- in planning and deployment. Cell size, ment this year. band and broadband access are probably capacity and return channel capability are better suited for asymmetric, distributive the main differentiating attributes of the Currently there is a significant interest in and downloading services than continu- wireless access solutions. However, the systems for transmission of data over the ous wideband or broadband services with interactivity is one challenge of these low voltage electricity distribution net- significant requirements for upstream broadcast tailored systems. The return work, known as power line communica- capacity. The lack of symmetric capabil- channel may either be provided by the tions [59]. A power line modem is in- ity and the limited total system capacities use of the existing twisted pair cable net- stalled at the low voltage transformer, compared to the wireline alternatives work, by the use of an overlaid cellular and an additional power line modem is makes satellite more of a complementary radio technology or through the use of installed at the customer premises. The system than full-scale, alternative infras- a return channel in the high frequency customers connected to the same low tructures to the wireline systems within band. Wireless broadband access systems voltage transformer share a data trans- the next five to ten years. This is reflect- are treated in more detail in [67]. mission capacity of about 1 Mbit/s. The ed in the fact that only a few of the LEO technology may enable the power utility systems now under development are tar- companies to enter the Internet access 6.3 Mobile broadband access geted at offering interactive broadband market utilising the existing low voltage services in the range of Mbit/s. Currently the mobile access networks electricity distribution network. Field have very limited capability of transmit- trials have been running since 1992/93; The DirecPC system has been in opera- ting high speed data traffic. GSM net- however, power line communication will tion from its geostationary position for works may transmit data rates of 9.6 probably not be implemented on a wide some time already. Two-way Internet kbit/s or 14.4 kbit/s. The data transmis- scale in the short term, but may turn out access is offered with a 400 kbit/s down- sion capacity of mobile networks will to be an alternative within three to five stream capacity combined with a stan- be increased over the next years by the years at access speeds of up to some few dard dial-up modem and ISP connection. introduction of GPRS (general packet hundred metres. The Motorola led international consor- 14 Telektronikk 2/3.1999

16 tium Iridium put its service into opera- Investment in broadband access delivery The studies include extensive techno- tion last year. Iridium is a global LEO must generate positive returns, consistent economic evaluations of upgrade tech- satellite voice and data communication with typically short term expectations of nology options, ranging from the wire- system with 66 satellites. Each of the investors. The decision to invest in new line options based on DSL, ATM PONs satellites is equipped with a switching technology is subject to the risks and and HFC systems to wireless alternatives system. The intention is to offer services uncertainties inherent in the competitive like LMDS and point-to-multipoint radio like data, fax, paging, real-time voice, marketplace. Thus, identifying the key access. In the studies access network messaging and position location. The economic issues related to broadband upgrades are addressed both from the available data rate is 2.4 kbit/s. Global- access investment are of utmost impor- point of view of the incumbent public star, another LEO system, is expected to tance. In this chapter, we will outline network operator (PNO), the community be up and running soon, with a total of some of these key issues which all may antenna television (CATV) operators as 48 satellites in operation. The intention have a significant impact on the overall well as from the point of view of new is to offer the same service portfolio as economics, namely: entrant access network operators. Iridium, however with a slightly higher The investment cost level. What kind available data rate of 9.4 kbit/s. The The methodology and tool developed of investment levels might be expected switching is performed at earth stations. within TITAN and OPTIMUM, and now for various technologies deployed for Skybridge and Teledesic are two of the under further development in the ACTS broadband access in different markets LEO satellite systems intended for broad- TERA project have been applied in the and areas? band data services and voice. The satel- techno-economic analyses in the studies lite system Teledesic with 256 satellites The service take rate. How will the [75, 76, 77, 78]. Typically a study period in LEO has been designed for data rates service take rate level affect invest- of ten years is considered, including the between 16 kbit/s and 2 Mbit/s, with a ment costs? expected price evolution of network design objective of 20,000 world-wide components and development of tariff Ducts and civil works costs. What is simultaneous 1.5 Mbit/s Internet links. levels throughout that particular period the impact of ducting and civil work Teledesic is planned to go into operation of time. The network element costs are costs as a consequence of the broad- in 2002. The features of Skybridge are extracted from a database developed band access upgrade? comparable to Teledesic. within the various projects. The database Fibre penetration and capacity. Given includes costs at a given reference year a migration towards fibre in the access for components, installation and civil 6.5 Stratospheric platforms network, how deep into the access net- works costs. The demand forecasts for The so-called stratospheric platforms aim work should the fibre be deployed? the selected bearer services for the resi- at providing high capacity broadband ser- dential and small business market as used The timing of the upgrade. How will vices over a limited geographical area of in the described studies, are extracted the required investment level change approximately 3,000 km2. The HALO from the RACE 2087/TITAN and the over time? (high altitude long operation) system is ACTS OPTIMUM Delphi surveys [35, based on aeroplanes circulating approxi- Revenue, payback and cashflow. What 34]. The business market services and mately 16 kilometres above earth. Sky- project values in terms of cash flows penetrations are predicted based on avail- Station utilises high altitude balloons and payback periods can the access able statistical material. The tariffs used located around 22 kilometres above the network operator expect as the revenue are European averages from the Delphi surface of the earth. The stratospheric streams from the new broadband ser- survey and other sources. The relation platforms are something in between high vice delivery arrive? between penetration and tariffs tariff radio towers and satellite systems, offer- elasticity has been incorporated in the ing potential coverage advantages com- The costs and economic viability of economic analysis. The effect of compe- pared to the former. The first SkyStation broadband upgrades have been studied tition is modelled through appropriate platforms are planned for operation over in several international projects. The pre- adjustments of market shares. Rome, Lisbon and Singapore in year sentation in this chapter is based on key 2000. The aim is to provide broadband findings from different studies carried services with 2 10 Mbit/s access capac- out over the past years within the projects 7.1 The investment cost level ity. The stratospheric platform concepts RACE (research and development in Initially, the range of expected invest- for broadband services are described in advanced communications technologies ment levels for broadband access deliv- more detail in [52]. in Europe) 2087/TITAN (tool for intro- ery should be set in absolute terms. The duction scenario and techno-economic required investments will of course vary evaluation of access network), ACTS from technology to technology, and 7 Economics and risk 226 OPTIMUM (Optimised architectures between markets and geographic areas. for multimedia networks and services), However, the studies performed all indi- The main inhibitors for the roll-out of EURESCOM (European institute for cate a level of installed first costs (IFC) broadband access networks have obvi- research and strategic studies in telecom- per new broadband connection which ously been the high capital investments munications) P306 (Access network evo- vary from just below 400 euro to be- required in order to upgrade the existing lution and preparation for implementa- tween 2,000 and 3,000 euro per new con- access infrastructure or build new access tion), P413 (Optical networking) and nection, depending on factors like tech- networks, and the high risks associated P614 (Implementation Strategies for Ad- nology choice, dwelling distribution and with placing these significant invest- vanced Access Networks) [68, 69, 70, civil work costs. In urban areas the in- ments in a market in which future ser- 71, 72, 73, 74]. stalled first costs per new switched ser- vices and revenues are highly uncertain. vice connection for plain old telephone Telektronikk 2/3.1999 15

17 service (POTS) and narrowband inte- investments are needed. The drivers for included for two degrees of duct avail- grated services digital network (N-ISDN) the fibre introduction are slightly differ- ability, 0 % and 100 %. The FTTB costs in the existing access network is typically ent, and hence the different cost struc- are plotted for each ONU serving eight in the range of 400 500 euro. This tures indicate that the effect on costs (BPON-8), 32 (BPON-32), 64 (BPON- underlines the fact that a further upgrade of the service take rate need to be taken 64) or 128 (BPON-128) potential cus- of the access network probably will into account in cost comparisons of the tomers. Please note that the potential require huge investments for any technol- different technology alternatives. capacity offerings to the customers are ogy selected, and that the operators will much higher for the FTTB solutions than be faced with investment projects of sim- Figure 14 shows the IFC per 2 Mbit/s for the ADSL and cable modem alterna- ilar or even higher financial burdens than access line for a ten year linear upgrade tives. the ones of establishing the narrowband project, as a function of broadband pene- access networks of today. tration in 2005. Line costs are analysed The alternative upgrades based on ADSL for the main system alternatives for the modems or cable modems both have three relevant wireline transmission prospects of line costs close to 500 USD 7.2 The service take rate media, namely fibre cable, coaxial cable for moderate take rates. The costs per The total upgrade investments may be and twisted pair copper cable. Results are line of upgrading existing twisted pair split into fixed and variable costs. The shown for an urban, apartment block area networks and coaxial cable networks fixed costs must usually be placed ini- with an average copper loop length of with DSL and cable modem technologies tially before the first customer sub- 400 metres between the optical node respectively are comparable for take rates scribes, and are to a large extent service (service access point) and the buildings. of up to 30 %. This illustrates the differ- penetration independent. The variable The IFC for BPON, or alternatively ence in the cost structure of the two alter- costs are typically service specific invest- ATM PON, in the FTTB configuration is natives. The upfront costs of the cable ments such as DSL modem pairs, and depends on the take rate. The different broadband upgrade alternatives have a different cost structure with respect to required upfront investment levels and 3000 service penetration dependent or variable costs: ADSL upgrades in many cases only involve service penetration dependent investments, which the network opera- 2500 tor need not take before the customer BPON-8 is connected. This alleviates the need for risky initial investments, although at the expense of a limited possibility 0% d.a. Installed First Costs per line (USD) for capacity offerings beyond 2 Mbit/s. 2000 BPON-32 Typically, architectures based on fibre systems such as broadband passive optical networks in combination with DSL technology incur significant upfront costs in terms of an initial 1500 HFC,ASB deployment of fibre infrastructure, with associated risks of loss due to BPON-64 100% d.a. failing future service take up and cor- responding revenues. The gain is the DS A capability to offer capacities higher L, 0% d.a. 1000 fee than 2 Mbit/s to the customers. der in HFC, no ret path c. 100% d.a. Similarly, HFC upgrades imply BPON-128 ADSL 0% d.a. upfront costs both in terms of return 100% d.a. amplifiers and deployment of some 0% d.a. fibre infrastructure to feed the different 100% d.a. 500 HFC, cable modem coaxial cable branches as the network is segmented. Also, HFC networks may in the long term provide capaci- ties beyond 2 Mbit/s. For both the two latter types of architec- 0 tures the level of upfront costs depends 0 20 40 60 80 100 on the capacity to be provided to the cus- Broadband connection demand tomers. The higher the capacity, the more Figure 14 The cost per line in an urban area for the main wireline upgrade fibre, and correspondingly more upfront alternatives as a function of broadband connection demand (d.a.: duct availability) 16 Telektronikk 2/3.1999

18 modem upgrades are higher than the ators are preparing the business cases for copper loop distance and capacity avail- almost negligible upfront costs of ADSL, broadband service introduction [80, 62]. able. The cable infrastructure costs vary whereas the variable costs of the cable significantly between network areas as modem upgrades are lower than the vari- previously discussed, and hence quite able costs of ADSL. Thus, for high ser- 7.4 Fibre penetration and different fibre penetration levels are vice take rates, cable modem upgrades capacity likely to be seen. have a good potential for reduction in Given the intention of a fibre-in-the-loop line cost due to sharing of infrastructure upgrade, the establishment of the fibre This is reflected in particular in the and low variable costs. Thus, for higher infrastructure is in itself probably a results from a study on cable infrastruc- penetrations the cable modem technology strategically more important decision ture costs, carried out in EURESCOM seems to have a cost advantage, but a than the choice of system technology, project P614 (Implementation strategies limitation in traffic capacity compared provided a system independent infra- for advanced access networks). The anal- to the ADSL technology. structure is rolled out. The reason behind ysis covers four network area types, a this is two-fold: The fact that the fibre downtown area, an urban area, a sub- A higher degree of coaxial network seg- infrastructure costs are likely to be sig- urban area and a rural area. The areas mentation or use of dedicated channel nificant and quite similar for all kinds have been segmented and characterised HFC upgrades, illustrated in the figure by of fibre systems, and that the difference according to average copper loop length HFC ASB (asymmetric switched broad- in system costs can be rather marginal in the existing access network (also band) or HFC symmetric switched broad- between different technological options. reflects the density of living units in the band (SSB), may increase the traffic Secondly, the expected technical lifetime area), availability of existing ducts and capacity in HFC networks. In the same of the fibre infrastructure is long com- surface conditions with corresponding way the twisted pair networks may be pared to equipment. Thus, rolling out cable deployment type and civil works upgraded to higher capacities by the use fibre remains a key decision of great costs. Representative ranges of the char- of fibre systems in combination with strategic importance in access network acteristic parameters have been assigned VDSL, shown as BPON in the figure. upgrading. But given a migration to each network area segment. Figure 15 However, the upfront costs increases towards fibre in the access network, how shows the asymmetric capacity provided significantly in this case, and the results deep into the access network should the after the upgrade as a function of cable indicate that the operator will have to fibre be deployed? This is one major infrastructure investment per homes rely on take rates of 50 % or higher in question in the long term perspective for passed for the four network areas. Cable order to reach line cost levels below access network operators, and PNOs in infrastructure investments encompass 1,000 USD. particular. The recent developments in civil work costs (digging, ducting and digital subscriber line (DSL) technology surface reinstatement), costs of cable enable the PNOs to provide broadband installation and cable costs. The upper 7.3 Ducts and civil work costs capacities on the existing copper net- and lower curves in each area represent As mentioned in the previous section, the work. However, in general for the digital minimum and maximum levels of civil level of civil work costs will in many sit- DSL options, there is a trade off between work costs. uations have a significant impact on the upgrade costs. In fact, the costs of civil works remain as one major obstacle for extensive introduction of fibre in areas with an established access network of 60 good quality. Fibring the upper part of Downtown the access network is already likely to be 52 Mbit/s cost effective in some cases, as illustrated 50 in Figure 14 by the lower costs of the fibre alternatives with a high number of Capacity provided (Mbit/s) Urban potential broadband customers per ONU. 40 Suburban However, the lower part of the network is very sensitive to civil works costs. Fibre deployment beyond the main flexi- 30 bility point in the network increases the 26 Mbit/s overall costs significantly if available ducts are scarce. Some of the studies 20 Rural indicate that civil works costs may con- stitute more than 30 % of the total invest- 13 Mbit/s ment costs for high capacity broadband 10 8 Mbit/s access network upgrades [79]. In conclu- sion, the costs of broadband access up- 2 Mbit/s 0 grading and correspondingly extensive fibre deployment is strongly related to 0 100 200 300 400 500 civil works costs. This makes new meth- Investment per homes passed (euro) ods for cable deployment as well as inno- vative concepts for utilisation of the Figure 15 Asymmetric capacity provided after upgrade existing cables crucial issues as the oper- as a function of investment per homes passed Telektronikk 2/3.1999 17

19 In general, for densely populated areas upgrade of the network to the next higher cost evolution embedded in the study (downtown, urban and suburban areas) asymmetric capacity level are between results. The expected decrease in cost the investment levels required for up- twice and three times the total initial during the next years is clearly seen. grades to higher capacities have an investments. increasingly strong dependency on the The equipment cost of the three upgrade loop lengths for increasing asymmetric alternatives will most likely have a quite capacities. 7.5 The timing of the upgrade similar cost reduction potential, as they The appropriate timing of broadband are all emerging technologies in this mar- The cable infrastructure cost levels pre- upgrading is crucial, not least in order ket. The total ADSL line costs are ex- sented here indicate a need for differenti- to ensure that the operator maintains his pected to be reduced by two-thirds of ating between network areas with respect customer relationship, but also in order to the 1996 cost level during the ten-year to service (capacity) offerings at asym- possibly reduce the investments. In some period, whereas the resulting fibre and metric capacities beyond 2 Mbit/s. of the studies this latter aspect has been coaxial cable upgrades are expected to analysed in more detail. The costs of up- experience a 50 % cost reduction. The Another aspect of this is the potential grade strategies involving mass deploy- difference in cost reduction is attributed capacity increase gained with additional ment of residential access fibre are anti- to the fact that the latter two upgrades investments on top of the initial upgrade cipated to become significantly lower if involve civil works costs in addition to investments, ie. a further upgrade of the the system introduction is delayed the equipment costs. network to the next higher asymmetric enough to benefit from component cost capacity level, indicated in the right side reductions. Figure 16 shows the broad- In conclusion, postponing the fibre roll- of the figure, eg. from 2 Mbit/s to band line cost as a function of roll-out out may result in a cost advantage com- 8 Mbit/s, or from 13 Mbit/s to 26 Mbit/s. year and penetration for selected broad- pared to the HFC and ADSL upgrades. The gain in available capacity per in- band PON (BPON), HFC and ADSL In addition, prospects of future operation vested euro decreases as a function of the alternatives. and maintenance savings might motivate population density, illustrated in Figure for an extensive fibre deployment. So far 15 by the decreasing slope of the invest- It is assumed that the indicated penetra- however, there is no clear evidence of ment level areas as we move from dense- tion is obtained at time of roll-out. The decreased operation and maintenance ly populated areas to more scarcely pop- BPON alternatives are calculated with costs to offset the huge investments ulated areas. Nevertheless, for all areas 50 % duct availability. The figure illus- required. the additional investments for a further trates the assumptions with respect to 7.6 Revenue, payback and cash flow It is obvious that the time frame and cor- 1600 responding payback periods of the up- grades to a large extent relies on the rev- 1400 enue levels. Figure 17 shows the payback period as a function of average annual access network related revenue per line 1200 for a ten year linear upgrade project. The payback period is defined as the period 1000 from the start of the project to the time Line cost (euro) when the cash balance (cumulative sum 800 of the retained cash flows) turns positive. The indicated penetration is the satura- 600 tion level of 2 Mbit/s asymmetric switched broadband service penetration in 2005. The BPON alternatives are cal- 400 culated with 50 % duct availability. The results indicate that the cost level of 200 broadband access upgrades is likely to be within the investment range in which 0 payback periods in the order of five years may be expected with annual revenues 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 per access line of 400 euro. However, it must be emphasised that these calcula- Roll-out year tions are based on an early (1996) and Cable modem, 10 % pen. BPON-64, 50 % pen. quite extensive roll-out. A more gradual deployment which in addition is post- BPON-64, 30 % pen. Cable modem, 30 % pen. poned in time, will require a lower turn- ADSL BPON-64, 80 % pen. over per access line in order to achieve Cable modem, 20 % pen. Cable modem 100% pen. acceptable payback periods. Hence, the payback periods of broadband access is Figure 16 Line cost as a function of roll-out year and penetration likely to be in the order of five years with 18 Telektronikk 2/3.1999

20 annual revenues per access line of 400 Development of the existing infra- access projects, and thus setting targets euro. This illustrates that access network structure and platforms; and accordingly strategies and decisions, broadband upgrades are likely to turn out are complex in the broadband access arena. Investment profile and economic risk; to be long term projects with payback periods in the range of five to ten years. Technology risk; The competitive situation in the short and long term may impact when, how and for Technology alternative performance The upgrade effects on the prospects which market segments the access net- and merits. looking beyond the upgrade period are work operator rolls out his broadband not evident from the NPV and the pay- access network. The impact of the com- The definition of the broadband access back period. The cash flows in the final petition on the operators actual selection network strategy depends on the overall year is useful as a predictor for the long of technology to be used for broadband strategic targets and imperatives, which term effect of expanded revenue base. access is somewhat more unclear and for the telecommunication operators and With the given tariffs and penetrations may in the end turn out to be rather lim- thus broadband access operators under all of the upgrade projects have gained ited. competition will be to optimise the net strength through the broadband upgrade present value of the broadband access compared to the initial annual cash flow, Market segments and geographic areas investment project(s). This very often ie. the final year cash flow is higher than are crucial aspects to consider in working implies minimising the risk of losses, the pre-upgrade cash flow. The results out strategies for broadband access. This exploiting revenue opportunities, limiting illustrate that access network broadband is mainly due to expected capacity de- the time for return on investments, mini- upgrades represent long term projects mand variations in the market, the local mising life-cycle costs and minimising which are assumed to create future cash nature of the access network segment and installed first costs. Moreover, the opera- flows. As such, the establishment of a the associated high investment levels. tor will often aim at achieving economies broadband access network platform rep- The capacity demands and willingness to of scale and economies of scope [81], resents a challenge comparable to the pay of large business users are very dif- which are closely related to the market build-up of the present narrowband ferent from the needs of residential cus- situation and degree of competition in the infrastructure: heavy investment projects tomers and small and medium enter- area being considered. These targets are with associated investment levels which prises. Thus, a variation in service offer- not always easily combined in broadband in the short and medium term result in a weakened net present value. However, at the same time these investments are a requirement in order to maintain the revenue level in the long term; in other words, a necessity in order to establish 10 an effective, broadband platform the basis for the future money machine. 9 8 Broadband access mi- 8 Payback period (years) gration and strategies 7 The challenge for the access network 6 operators is to derive strategies for mi- gration towards broadband access, in- 5 cluding evolutionary paths for either a migration of existing network infrastruc- 4 tures or for deployment of a completely new access network infrastructure. A dis- 3 cussion on the different migration paths for wireline access networks is found in 2 [49]. The main aspects (most of which have been addressed in the previous 1 chapters) related to devising broadband access migration strategies may be con- 0 sidered to be: 121 182 242 303 364 424 485 525 583 The overall strategic targets and im- Average annual access network related revenue per subscriber (euro) peratives; The competitive situation; Cable modem, 10 % pen. BPON-64, 50 % pen. ADSL Cable modem, 30 % pen. Market segments and geographic areas; Cable modem, 20 % pen. BPON-64, 80 % pen. Existing and future product portfolio; Cable modem 100 % pen. Figure 17 Payback period as a function of average annual revenue per line Telektronikk 2/3.1999 19

21 transmission capacity, quality of service, Investment profile flexibility and operational stability, in and economic risk addition to the economics. Existing and future service portfolio Acknowledgements Market segments Several of my colleagues at Telenor have & geographic areas inspired and through advice and discus- The overall sions contributed to the work on which Broadband strategic this article is based. In particular contri- Technology Access alternatives butions including graphics and edited Network targets and text from Markku Lhteenoja, Dagfinn Strategy imperatives Myhre, Frode B. Nilsen, Borgar Trre Existing infrastructure and Olsen and Kjell Stordahl are acknowl- platforms edged. Parts of this document are based on results achieved in the project 2087 Technology TITAN of the European RACE program, risk and the project 226 OPTIMUM of the Figure 18 Devising The competitive ACTS program. The author gratefully situation the broadband access acknowledges the support of the Euro- network strategy pean Commission and the project mem- bers in carrying out this work. This paper is also based on results achieved in EURESCOM projects P306, P413 and P614; this does not imply that it reflects a common technical position of all the ings between the market segments are As discussed in the previous chapter the EURESCOM shareholders/parties. The needed. Furthermore, most access opera- investment profile and associated eco- author gratefully acknowledges the sup- tors upgrading to broadband delivery will nomic risk of broadband access upgrades port of EURESCOM, EU and the mem- have to do a geographic segmentation of have so far possibly been the major in- bers of the various projects in carrying their network, mainly driven by the cost hibitor of a large-scale roll-out of such out this work. structure of the access network, with lim- networks. In general, the operator will try ited possibilities of cost sharing between to reduce the economic risk both through customers and network areas. The chal- a gradual deployment and by increasing References lenge is to combine the market segmenta- the share of variable costs relative to tion with the geographical segmentation. fixed installed first costs. However, the 1 Communications Week international most influential economic risk will any- online. (1999, 04,05) [online]. URL: Appropriate timing and design of the way be associated with the market uncer- secure/. existing and future product portfolio is tainty. critical for broadband access network 2 Mognes, P, Nord, T. Regulation of strategies, since the operators cash flow The technology risk is generally first and broadband access networks. Telek- and hence investments for the future ser- foremost associated with the maturity of tronikk, 95 (2/3), 5059, 1999. (This vices are so dependent on proper accom- the technology, and the timing of com- issue.) modation of the cash cow services during mercial available equipment in relation to the migration period. On the other hand, planned service launches. Other general 3 TELE.COM (1999, 04,05) [online]. the access network operator must avoid issues related to technology risks are the URL: that competitors through their broadband evolution of new coding techniques and directlink.cgi?TLC19990405S0039. access product offerings both gain cus- the impact of the ongoing convergence tomers for basic services and simultane- between the broadcast, data communica- 4 Clark, E. Pricing the last mile. Net- ously win the majority shares in the new tions and telecommunications industry. work magazine, Feb, 3641, 1999. broadband access market. In addition there are several technology specific risks involved, such as the actual 5 Gotta wear shades., 12 (4), Typically, the access network operator reach in operational conditions of DSL 5265, 1999. as a part of a larger telecommunications systems, the frequency licences of broad- company has considerable capital in- band radio access systems and the physi- 6 Mason, C. Can you find the wireless vested in an existing network infrastruc- cal condition of existing plant and future runners? Americas Network, 1 Feb, ture, and will seek to exploit as much increases in capacity demand connected 2438, 1999. as possible of this infrastructure in the with coaxial cable modem upgrade tech- migration towards broadband access. nologies. 7 Sloan, A, Sandberg, J. Bell of the Thus the development of the existing Brawl. Newsweek, 17 May, 5254, infrastructure and platforms is a very And finally, the performance and merits 1999. important criterion to take into account of the technology alternatives are crucial. in broadband access strategies. The evaluation of the technology alterna- 8 (1999, 06, 13) [online]. tives must at least address the aspects of URL: q6302.htm. 20 Telektronikk 2/3.1999

22 9 Waters, R. AT&T links up with Time 24 Biagi, S. Helping Copper Grow. 38 FSAN Requirements Specification, Warner. Financial Times, 02.02.99. Telephony, 9 Nov, 2637, 1998. Draft G. 10 AT&T (1999, 04, 05) [online]. 25 NEW YORK TIMES (1999, 05, 07) 39 OVUM. Interactive Multimedia Ser- URL: [online]. URL: http://www. vices to the Home, 1997. 0,1193,330,00.html. biztech/articles/07cable.html. 40 Forrester Research, Inc. Broadband 11 Flanagan, P. DSL and the access hits home, August 1998. race. Telecommunications Americas 26 BROADCAST.COM. (1999, 07, 05) edition, May, 3539, 1999. [online]. URL: http://www.broad- 41 Barrecca, S L. Internet Access Mar- kets: Marketing and strategic impli- 12 TELECHOICE (1999, 04, 05) cations for the next millennium. [online]. URL: http://www.tele- 27 SNAP! (1999, 07, 05) [online]. Unpublished. choice. com//inthenews/tele- URL: choice/89.asp. 42 Moe, G, Sther, J P. What is the 28 SNAP! (1999, 07, 05) [online]. willingness to pay for broadband ser- 13 TELECHOICE (1999, 04, 05) URL: vices? Telektronikk, 95 (2/3), 2333, [online]. URL: http://www.tele- 1999. (This issue.) choice. com//inthenews/tele- 29 Owen, T. choice/26.asp. deal sets broadband stage. RED 43 Rupp, B et al. Internet user reactions HERRING (1999, 04, 02) [online]. to usage-based pricing. In: Proc. Sec- 14 Telephony (1999, 04, 05) [online]. URL: ond Berlin workshop on Internet URL: http://www.internettelephony. insider/1999/0402/news-yhoobcst.html Economics, May 1999. com/archive/BetweenEditions/ 3.31.99/betweentop.htm. 30 Forrester Research, Inc. From Dial- 44 Stordahl, K, Olsen B T, Ims, L A. Do Up To Broadband, April 1999. we need a Pan-European network 15 POINT TOPIC (1999, 07, 01) and what is the need for new applica- [online]. URL: http://www.point- 31 AST&T (1999, 06, 11) [online]. tions? In: Proc. 22nd European Con- topic. com/. URL: ference on Optical Communication 0,1193,510,00.html. (ECOC 96), Oslo, 1519 Sep 1996, 16 BT. (1999, 07, 10) [online]. URL: 2, 2-3 2-10. 32 Palsson, H, Fortea-Laguna, S. The ment/nr9949.htm. human wants for freedom and its 45 Kazovsky, L G, Khoe, G D, van influence on telecommunications. In: Deventer, O. Future telecommunica- 17 (1999, 07, 10) [online]. Proc. 11th International Symposium tions Networks. IEEE Communica- URL: on Subscriber Loops and Services tions Magazine, 36 (11), 122127, q10de5e.htm. (ISSLS 96), Melbourne, 1996, 1998. 1014. 18 OFTEL (1999, 04, 05) [online]. 46 Evagora, A. The end of architecture. URL: 33 Vittore, V. Networking comes home., 1 (9), 5360, 1996. petition/llu1298.htm. Telephony, 22 Feb, 2634, 1999. 47 Nilsen, F B. IP or ATM in the access 19 BT (1999, 04, 05) [online]. URL: 34 Stordahl, K, Rand, L. Long term network? Telektronikk, 95 (2/3), forecasts for broadband demand. 6072, 1999. (This issue.) regulatory/response/bandwidth/ Telektronikk, 95 (2/3), 3444, 1999. index.htm (This issue.) 48 Kelly, D. Digital Subscriber Lines (DSLs): expanding the use of the 20 Wired News. (1999, 07, 09) [online]. 35 Stordahl, K, Murphy, E. Methods for copper loop plant. In: Proc. EURES- URL: forecasting long term demand for COM Workshop on Optical Multiser- news/email/explode-infobeat/ wide and broadband services in the vice Access Networks (WOMAN), rtechnology/story/20448.html residential market. IEEE Communi- Turin, Italy, 2324 Sep 1996. cations magazine, 13 (2), 4449, 21 Kwok, T C. Residential broadband 1995. 49 Ims, L A. Wireline broadband access architecture over ADSL and G.Lite networks. Telektronikk, 95 (2/3), (G.992.2) : PPP over ATM. IEEE 36 OVUM. Interactive Multimedia Ser- 7387, 1999. (This issue.) Communications magazine, 17 (5), vices to the Home, 1997. 8489, 1999. 50 Seidman, L P. Satellites for wide- 37 Luck, D P. Broadband To The Home: band access. IEEE Communications 22 Steinke, S. A lineup of local-loop Evolution Scenarios for Australia. In: Magazine, 34 (10), 108111, 1996. contenders. Network magazine, Feb, Proc. 13th Annual Conference Euro- 3035, 1999. pean Fibre Optic Communications 51 EURESCOM. Opportunities for and Networks, EFOC&N 95, 2730 broadband radio technologies in the 23 The Telechoice report on XDSL, June 1995, Brighton, UK. access network. Heidelberg, Ger- March 1999. Telektronikk 2/3.1999 21

23 many, 1998. (EURESCOM project 62 Edvardsen, E. Inverse multiplexing. 71 Ims, L A et al. Multiservice access P614, Deliverable 4.) Telektronikk, 95 (2/3), 129133, network upgrading in Europe : a 1999. (This issue.) techno-economic analysis. IEEE 52 Nordbotten, A. Satellite networks Communications Magazine, 34 (12), and stratospheric platforms. Telek- 63 Maxwell, K. Asymmetric Digital 1996, 124134. tronikk, 95 (2/3), 107112, 1999. Subscriber Line : Interim technology (This issue.) for the next forty years. IEEE Com- 72 Ims, L A et al. Upgrading the PNO munications Magazine, 34 (10), and cable operator access networks 53 Rkken, R, Svaet, S W. Mobile 100106, 1996. to advanced broadband services : broadband access. Telektronikk, 95 technology alternatives and their (2/3), 96106, 1999. (This issue.) 64 ADSL Forum (1999, 02, 18) [online]. economic implications in a competi- URL: http://www.adsl. com/gen- tive environment, In: Proc. Telecom 54 Clarke, D E A et al. Emerging broad- eral_tutorial.html. America 96, Technology Summit, band access technologies. BT Tech- Rio de Janeiro, Brazil, 1015 June, nology Journal, 16 (4), 187194, 65 Quayle, A, Stern, J A. Achieving 1996, 243249. 1998. global consensus on the strategic broadband access network : the full 73 Olsen, B T et al. Technoeconomic 55 Breivik, T, Lnsethagen, H. Manage- services access network initiative. evaluation of optical broadband ment of access networks trends and Telektronikk, 95 (2/3), 166175, access network scenarios for the resi- challenges. Telektronikk, 95 (2/3), 1999. (This issue.) dential and business market. In: 145159, 1999. (This issue.) Proc. 21st European Conference on 66 Krapf, E. Are changes in the air for Optical Communications (ECOC 56 Klepsland, P. Towards a practical local access? Business Communica- 95), Brussels, Belgium, 1417 Sep implementation of DSL preparing tions Review, Feb, 2730, 1999. 1995. for new technology and regimes. Telektronikk, 95 (2/3), 213225, 67 Loktu, H, Bigan, E. Fixed broadband 74 Myhre, D. The economics of broad- 1999. (This issue.) wireless access. Telektronikk, 95 band service introduction. Telektro- (2/3), 8895, 1999. (This issue.) nikk, 95 (2/3), 262-272, 1999. (This 57 Gnazzo, A. Measurements and main- issue.) tenance of future access networks. 68 Ims, L A et al. Economics of broad- Telektronikk, 95 (2/3), 208212, band access network upgrade strate- 75 OPTIMUM (1999, 07, 04) [online]. 1999. (This issue.) gies : the European perspective. In: URL: Proc. Access Network Mini Confer- 58 Ciochetto, C, Esposto, F. The chal- ence, Globecom 98, Sydney, Aus- 76 TERA (1999, 07, 04) [online]. lenge of civil works. Telektronikk, 95 tralia, 812 Nov 1998. URL: (2/3), 202207, 1999. (This issue.) 69 Olsen, B T, Ims, L A, Myhre, D. 77 Ims, L A (ed.). Broadband Access 59 Collins, D et al.Transmission on Techno-economic analysis of broad- Networks introduction strategies power line cables. Telektronikk, 95 band access network upgrading. In: and techno-economic analysis. Chap- (2/3), 134137, 1999. (This issue.) Proc. Networks 96, Sydney, Aus- man-Hall, January 1998. tralia, 2529 Nov 1996, 1, 373378. 60 Koonen, T. Flexible wavelength mul- 78 Olsen, B T. OPTIMUM a techno- tiplexing techniques for broadband 70 Ims, L A, Olsen, B T, Myhre, D. econoic tool. Telektronikk, 95 (2/3), fibre access networks. Telektronikk, Economics of residential broadband 239250, 1999. (This issue.) 95 (2/3), 119128, 1999. (This issue.) access network technologies and strategies. IEEE Network, 11 (1), 79 Ims, L A, Myhre, D, Olsen, B T. 61 Faulkner, D W. The passive optical 5864, 1997. Costs of upgrading the residential network (PON). Telektronikk, 95 telecommunications infrastructure to (2/3), 113118, 1999. (This issue.) broadband. In: Proc. Globecom 98, Sydney, Australia, 812 Nov 1998. 80 Ciochetto, C. The challenge of civil works. Telektronikk, 95 (2/3), 202207, 1999. (This issue.) Leif Aarthun Ims (33) is Senior Research Scientist in the Strategic Network Development group of Telenor R&D. Since 1994 he has been working on access network 81 Reed, D P. Residential fiber optic development in several RACE, ACTS and EURESCOM networks an engineering and eco- projects, and recently as project manager for long term nomic analysis. Artech House, Nor- access network development in Telenor. He is now project wood, MA, USA, 1992. leader of the EURESCOM project P901. He is editor of the book Broadband Access Networks published last year and has published more than 50 papers in international journals and conferences. email: [email protected] 22 Telektronikk 2/3.1999

24 What is the willingness to pay for broadband services? GEORG MOE AND JAN-PETTER STHER 1 Introduction researchers the possibility to evaluate the In order to get the general understanding price/bandwidth ratio to some extent. of the future we had a look at what con- The next major task for telcos is the sultants at Ovum, Forrester, Analysys deployment of broadband access, the In the study we distinguish between two and Telenor Research & Development magnitude of investments implies that main market segments, small and had written about the future in addition one has to be very careful about when medium sized enterprises (SME) with to other relevant literature [1, 2, 3, 4, 5, and where to start. For the incumbents it 1100 employees and household/SOHO 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, is a change of regime at the same time. (Small Office, Home Office), in addition 18, 19, 20, 21, 22, 23, 24]. From that a Previously, when there was no competi- to population and business density. The couple of scenarios were designed. tion, one might be tempted to argue that main reason for this simplified segmenta- new technology could be deployed acc- tion is that we are looking at the roll-out For the quantitative part we made two ording to a 5 year plan without having to of new infrastructure, we assumed that sets of questionnaires, one for each main take care of customer needs. The old net- population and business density were the segment. For the SME market we had a work planning offices made their fore- two most important variables for this traditional approach, about 1000 compa- casts based on history, in a stable market purpose. The reason why we excluded nies were contacted. The person respon- environment. Pricing was seldom based larger companies, is that they usually sible for IT in the company was asked. on the customers willingness to pay; have a more differentiated demand for Because these people are mainly oriented rather it was based on some costing mod- telecomms services, in addition they are towards the near term challenges we lim- els, or set to control the demand with few in numbers in Norway relatively ited the future time horizon to three years. respect to the speed of deployment. In speaking. They were asked about current and future the present competitive environment all telecomms and IT preferences and spend- has become different, new operators do ing patterns, in addition to their willing- things differently, and pricing has be- 3 Approach ness to pay. Regarding their willingness come an important issue. What expecta- to pay we tried to distinguish between An unlimited number of factors deter- tions do we have for willingness to pay fixed and mobile broadband access (main mine the success of goods and services for new services based on broadband focus on the applications level). We anal- both in the professional market as well as access? This is the central question we ysed the responses, and made our own in the household market, however some are going to address here. To illustrate forecasts based on the expressed willing- factors are crucial. There has to be a need different aspects we will refer to a survey ness to pay, and our perception of how for the product, and there must be a conducted in the Norwegian market the market would develop. trade-off between the utility and the price (summer/autumn 1998). and ability to pay for it. In the profes- For the residential market we performed sional market this utility is characterised We are not going into details regarding a Delphi survey, where several experts by the ability to perform necessary activ- the results, because actual prices, elastic- answered a questionnaire. Due to their ities better. In the household market the ity values and so on are less relevant for expert nature, the time horizon was utility is described by the direct welfare the discussion. So we will focus on the defined to 15 years. The SME and Delphi to the individual as final (or in connec- challenges, and methodologies in our survey are not completely compatible, tion with) consumption but also the value approach. But first of all, we would like because the experts were asked to make as signal in a social context. to present the case; just to show why we forecasts on their own. These forecasts have been doing it this way rather than may be inconsistent with their perception Price and income are the more solid vari- another. of what people in general are willing to pay. ables in terms of a quantitative approach. Even only with these two variables and 2 Background a set of defined products we will see a 4 Broadband applications complicated system for determining the The purpose of the study was to get a potential demand for a new product. Sub- It makes little sense to talk about willing- picture of the broadband market, before stitution and complementarity, dynamic ness to pay unless one says something any investment in infrastructure was effects, time budgets and product hier- about the applications involved. The made. That means that the services on archies are in short terms examples of applications do in fact play a major role top of the broadband infrastructure were the complexity in this field. regarding the deployment strategy. If the only described in general terms, just to applications require some sort of critical give the respondent some hints of what Because we were investigating the roll- mass of subscribers, then you would was about to come. A major weakness of out of new technology in the local loop, expect some heavy initial investments to the study is the fact that the respondents in a new competitive environment and cover a broad area. Otherwise one would hardly have any references for an evalua- looking at products not very well inter- expect a more selective approach, where tion of broadband services, like quality of preted among most of the customers we the expected utilisation of the network is service, distribution etc., apart from the needed a methodology to analyse the higher. fact that broadband Internet will be market. We needed both a general faster. Price, or price level, is the only understanding of the future and a way In our study of the SME segment we element that says something definite to quantify this future. Needless to say, made a ranking of applications, based on about the product. However, the applica- the future situation has many possible a short-term interest (one year from now) tions described to the respondent im- outcomes, so we had to choose a set of and a longer-term interest (three years plicitly require a given bandwidth and scenarios. from now). The following applications a level of symmetry, which gives us as were involved: Telektronikk 2/3.1999 23

25 E-commerce (providing this for cus- tomers, and as a tool for ordering from the suppliers); Video on Demand; - E-commerce - Information gathering - Advanced home office Advanced home office; - Videoconferencing - Video on demand - Major file transfers Major file transfers, related to out- sourcing (symmetrical bandwidth); Information gathering, related to Inter- net and multimedia (asymmetric band- width); Randomised call pattern Dedicated call pattern Video conferencing. (depends on critical mass) (does not depend on critical mass) Within all segments the most wanted Hybrid solution applications were: Advanced home office: semi ran- Figure 1 Categories of broadband applications domised demand, both in terms of bandwidth symmetry and time of day; Major file transfers: needs for sym- metric transmission, mainly at given points in time (outsourcing); Information gathering: randomised demand with respect to time of day, asymmetrical bandwidth. The respondent interprets s the question Interestingly enough, these applications ate Th do not require a critical mass of sub- ul e re scribers from a network deployment rm s r fo th po point of view. Initially this is important, e on e nd r ch esti qu e because the equipment needed will be a u es nt expensive. As the equipment prices fall, se q tio an re a n sw one has the possibility to provide the e es Th randomised customer with cost effec- tive services. Figure 1 indicates how we categorised various applications. For the SME segment this categorisation The researcher needs The researcher interprets helps us somewhat to understand the responses in our survey, at the same time to know the answer it creates some sort of confusion. Let us Does the researcher and start by looking at the confusing part. respondent understand each other? The problem did arise during the inter- pretation of the results of the question- naire. We may ask a series of questions, but it is not necessarily so that the respondent has the same understanding of the questions as we have. One might say we have a chain of misunderstanding and misinterpretation as indicated in Fig- ure 2. The importance here is that we were too optimistic about the willingness to pay, initially. The respondent may have inter- preted our questionnaire so as to say that The researcher draws all the described applications of interest conclusions were available from day one, and the respondent did not make any distinctions Figure 2 Communication between respondent and researcher as context-sensitive whether the applications are suitable for in the interpretation process randomised call patterns or not. Finally, 24 Telektronikk 2/3.1999

26 the respondent may express a willingness Price to pay, based on the expectation that everybody else has access, and use appli- cations that require broadband access. So, we ended up saying that we do not immediately believe in the respondents willingness to pay, therefore we had to modify results that indicate willingness to pay in order to make forecasts. Simplified demand curve Having estimated the demand curves, we made a major correction downwards in our forecasts the first couple of years, because we do not expect to be able to cover the whole market from day one anyway. From the researchers point of view there is another element complicating the inter- pretation of the expressed willingness to pay. The problem is of a fundamental nature when estimating demand func- tions/curves, and is related to the prob- Quantity lems mentioned above. Depending on Figure 3 Simplified demand curve in the price-quantity dimension how one approaches the estimation of demand curves, one may find a model that describes the data well, but the re- sulting function does not necessarily de- scribe the demand in the market. Suppose we have an estimated (and generalised) Price demand function like the one in Figure 3. The respondents perception of the supply side This aggregate curve is based on the answers of all the respondents. But, the respondents have given their answers based on what they perceive the supplied product/service is. Instead of being just Actual demand curves another respondent among other respon- dents, they may just as well be a repre- sentative of another subset of potential customers with a completely different demand behaviour (perhaps because they perceive the supply side differently). The demand curve shown is just an average based on all respondents within a seg- ment as defined by the researcher. The Estimated demand actual set of demand curves may as well curve look like shown in Figure 4 [29]. One could argue that we have not defined our segments properly before estimating the demand curve. However, the cost of Quantity the study would increase to prohibitive Figure 4 Possible underlying perception of supply curves, individual levels if all kinds of segments were de- demand curves and a misinterpreted estimate of a demand curve fined and taken care of. In addition, one has the problem regarding what the respondent thinks about the supply of products and services. Failing to realise this can have serious consequences with equal. The distinction between price and IT industry have an understanding of respect to the estimated price elasticities. elastic and inelastic demand is crucial broadband services. They do not exist The estimated demand curve may indi- in a competitive environment (Figure 14). today, and it is difficult to relate to some- cate a price inelastic demand, whereas thing you do not know about, and proba- the actual demand curves represent a A lot of these problems have to do with bly even more difficult to express a will- price elastic demand, all other things the fact that few people outside the telcos ingness to pay. Telektronikk 2/3.1999 25

27 5 The complexity of in question increases, and at the same would then be average annual price lev- time developing costs are being re- els within each of the mentioned demand covered. In turn new suppliers are enter- segments. ing the market and the product tend to be To indicate the complexity of demand, a so-called commodity, a standard prod- let us consider an individuals consump- uct with many competing suppliers. The 6 Results from the SME tion. Using one of the simplest models possible including prices, income, a set price will decrease continuously during market survey this process, so also the value of the price of n commodities and a utility function and income elasticities for that particular Different models were tested with the with substitution gives us a set of equa- product. For a price close to zero, the retrieved data, indicating that a log-linear tions with ((n(n + 1) / 2) 1) degrees of price elasticity will also be close to zero, one was matchless in producing good test freedom, which is the minimum number ie. a further decrease in price will not values. The model specification is as of elasticities that have to be estimated have any remarkable effect in the follows: empirically, ie. a considerable amount. demand. Typically the traffic per tele- The price elasticity of for instance tele- x = Apb which can be derived to phone line per day has been at a level of phone usage is defined as the relation ln x = ln A + b ln p, 25 minutes per subscriber line in USA. between the relative change in demand The level in Europe has been between a a format suitable for using linear re- for an infinitesimal relative change in half and a third of that level, to a high gression in order to estimate parameter price ((d log(x) / (d log(p))), or as a prac- extent reflecting that zero-priced local values. The number of accesses is repre- tical approximation the change in percent traffic (ie. flat-rated) has been more sented by x, p is the price per access per of the quantity (minutes) for a one per- prevalent in the US than in Europe, year, A and b are parameters estimated in cent price change. The cross elasticities where the main principle has been to the model. b will implicitly represent the are in the same manner defined as the use traffic-based tariff paying per minute value of the price elasticity, which is change in percent for a commodity when or per pulse in addition to subscription ranging from 0.3 to 0.6 for the differ- the price of another commodity increases tariffs. If the price is zero, the equilib- ent company size segments; ie. from a by one percent. The income elasticities rium level of minutes per telephone line certain level a ten percent price reduction are defined as (approximately) the per- per day will be finite. In practice we must would increase the demand for accesses centage change in demand for a com- accept to use rather rough estimates for by three to six percent. In other words modity, as a result of an income change elasticities. broadband access is inelastic with respect of one percent. All relations indicated to price, which seems to be in conflict could be estimated by varying the prices Let us start with some general results, with the assumption that unique new and incomes in a controlled manner, based on the literature studies. The products should be very elastic with however a market economy would not results have been categorised according respect to price. allow such an experiment. to the segments we use and a time scale in Figure 5. In the SME study we dis- We have already mentioned in this article The smallest change in the model indi- covered that applications related to home that the life-cycle of a unique product cated above to simulate a market oriented offices were perceived as very interest- that gradually becomes a commodity as economy should be the introduction of ing, which at the same time raises the time elapses is characterised by a high new products in the market and the de- question as to how to describe the resi- absolute value of the price elasticity in letion of obsolete products, ie. a more dential market. As long as the residential the initial phase. The price elasticity will dynamic approach of the indicated market only focuses on residential appli- decrease as the price decreases and out- framework. Here a crucial question is cations, like games, music on demand, put increases. This is obviously in con- how new products comply with the cus- or video on demand, then the picture is trast to the results from the customer sur- tomers taste, their utility and their ability clear. But, the concept of the home vey, which indicates a constant and low to afford them. How do new products office is based on professional applica- absolute value of the price elasticity. On interfere with the existing line of con- tions (and the company is paying). How- the other hand we have indicated that sumer products and the pricing of these ever, given that the broadband access is the assumption of an aggregate demand products? In a short description we in place at home (for professional use), curve that represents the whole market would suggest the following: A new then there is no reason why it cannot be might be incorrect. Furthermore the product tends to be priced according to used for entertainment. This raises some answers could be tactical in the sense its uniqueness (however in the telecom- questions regarding price, distribution that customers might want to move the munications world so far recovery of and quality of service. prices to a convenient level through their sunk cost is an aspect that has affected response. Last but not least, the cus- the pricing scheme). A unique product One of the conclusions of our study tomers might not have the realistic idea with the potential of conveying a high was that there is a big difference in the of the utility of broadband access due to degree of utility to the customer is ordi- willingness to pay for broadband access the fact that computers/terminals today narily highly priced in the introductory between the SME (between the smallest seem to run sufficiently well on narrow- phase. As competitors offering the same SMEs) and household segment, all things band services in combination with com- or a very similar product approach there equal. So, the non-price variables need a pression. However, increasingly power- will be a competition in which price lot of attention. As mentioned earlier, the ful end-user equipment tends to fill in- tends to be focused. Price and income main purpose of our analysis is to focus creased capacity as it approaches. Figure elasticities tend to be high, which in on how we were going to deploy new 6 indicates demand curves for 2 Mbit/s effect boosts demand when the prices are infrastructure; the willingness to pay access for companies of different sizes decreasing. Total revenue for the product and price levels. 26 Telektronikk 2/3.1999

28 General Applications for - ISDN is still increasing, ISDN has reached comments wireless access exist. but revenues per subscriber maturity. Only a price matter. is falling. Internet is the primary - Huge investments regarding reason why customers digital entertainment and switch technology and shopping. operator. Choice of distribution channels will be essential - Wireless access could be - VDSL in early stages. the technology of choice for new entrants. - Wireless access is also con- sidered as complementary to fixed access, not a substitute. SME - Media, finance, travel - Cable modems are not regarded agencies in addition to as interesting for SMESs. business related to con- - ADSL is considered as a better sumer electronics are choice. probably early adopters. - Wireless is not considered as a real choice. SOHO It is critical for the network - Primary segment for ADSL. operator to gain access to - Plug & play is essential for the SOHO and residential this segment. market. - Wireless only for certain applications. Residential Cable modems It is critical for the network - Cable modems become ADSL has a huge have limitations. operator to gain access to mature. potential in this seg- the SOHO and - ADSL only for densely ment, but faces fierce residential market. populated areas. competition based on - Plug & play is essential other technologies. for this segment - Wireless only for certain applications. Present Near future Future Figure 5 Market segments and distribution channels at present and in the future We have made a presupposition regard- In addition we have made a judgement estimate of the degree of bandwidth ing the market adoption rate. If the of historical values for the prices of data migration was completed by the use of results presented in the curve in Figure 6 communications services network com- what we could denominate a calibration should be converted directly into fore- ponents and Customer Premises Equip- model [25]. The calibration model was casts, initial demand in the year of intro- ment (CPE), which indicates that data applied for estimating access network duction would constitute about 50 per- communications will tend to have an capacity, and it indicates a much higher cent of the demand in the final year of annual decrease of about 15 to 20 percent absolute value of the price elasticity than the study, which surely is quite unrealis- per year. This seems to be a realistic what was derived from the user survey tic. Market adoption rate is not very judgement according to historical figures. and the model applied to it. The elastici- predictable. Therefore we have made a With the proposed assumptions, a de- ties from the calibration model should, judgement of such a rate since it will take layed adoption rate and a considerable however, be interpreted as quasi-elastic- time for the companies involved to adopt decrease in the price of broadband ities which are composites of elasticities broadband access. There are different access, the estimate of the total market and technical trends. Both values exceed reasons for this, eg. that company penetration is presented in Figure 8. three in absolute numbers. That model planning, technology platform restric- was based on the simple assumption that tions etc. will be obstacles for the imme- The results from the survey may be trans- the demand for data communications diate demand. The adoption curve is formed to total market revenue for the capacity in the access network depended presented in Figure 7. broadband service in question. An earlier on the price of data communication Telektronikk 2/3.1999 27

29 100 90 80 70 access (weighted index) and gross 1-6 employees domestic product (GDP) in real terms as Percent (%) 60 7-25 employees in the following relation: 26-100 employees 50 dp dr dqt+k = qt e t + E t , pt rt 40 in which q, p and r represent total data 30 communications access capacity (q), price for data communications services 20 (p) and GDP (r). e and E represent quasi- 10 elasticities for price and income includ- ing technical change. t represents time 0 and t+k represent the time lag from a change in either price or GDP before the 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000 6500 7000 7500 8000 8500 market responded, typically 1.5 years. Historical annual figures from 1980 to Figure 6 Percentage of companies indicating adoption of 2 Mbit/s broadband 1993 were used in the calibration pro- access for different price alternatives cess, and to a very large extent the cali- bration model was able to reproduce historical figures. This is no proof for the validity of the model, however there is 100 reason to believe that price and income to a large extent explains the demand for 80 non-voice communications capacity for the time period from 1980 to 1993. This Percent (%) Degree of adoption is illustrated in Figure 9. 60 Revenue estimates based on 1) the 40 survey among small and medium enter- prises, and 2) the calibration model, give results as presented in Figure 10. 20 Despite the quite different approaches chosen, we have constructed two time 0 series of revenue figures for the Nor- wegian market in which the calibration 2000 2001 2002 2003 2004 2005 2006 2007 2008 model should represent the total non- voice communications market, while the Figure 7 A tentative adoption curve as a correction for demand curves user survey only includes the market for derived from user survey 2 Mbit/s access, which is part of the total non-voice communications market. The results from the two methods seem to support each other. In the final year the 70 results from the calibration model indi- cate an approximately four billion NOK 60 revenue which is one third of the total 1-6 employees fixed telephone revenue today, not neces- 50 7-25 employees sarily exaggerated. The results indicate 26-100 employees that a possible value of the 2 Mbit/s mar- Percent (%) ket will be 700 million NOK in 2007. 40 If we assume more than one access per company among the larger companies the 30 2 Mbit/s market might be considerably larger, up to 1.6 billion NOK. 20 The user survey included a question 10 about the potential demand for applica- tions that need support from 25 Mbit/s 0 access, however the response to that question was low and did not logically 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 make any sense. In Figure 10 an indica- tion of higher bandwidth revenue is Figure 8 Penetration of broadband access (2 Mbit/s) in companies plotted in order to parallel the trend of of different size (introduction in year 2000) the total revenue in non-voice communi- 28 Telektronikk 2/3.1999

30 cations. There is no reason to believe that 2100 the total market for broadband access will slow down after 2007. 1900 1700 Observations 7 Results from the resi- 1500 Results of model dential market studies 1300 calibration In our study for the residential market we 1100 used two different approaches, one based 900 on a Delphi survey and another based on how much a consumer spends on tele- 700 com, newspapers, entertainment and 500 related applications. These two app- roaches were used in order to examine 300 the presence and level of consistency. 100 A very timely question, if not always 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 given too much attention, is how much time we will spend on different new goods and services. An increasing diver- Figure 9 Demand for access capacity in packet and line switched data networks sity of products is introduced in the mar- and leased lines 19801993. Observed values and calibrated values according ket. Consumption takes time, therefore to the model described above the use of time for every activity involv- ing consumption of goods and services should be analysed in order to see whether there is a trade-off between the different factors. Attempts have been 4000 made at introducing a general theory of the economics of time, for instance as 3500 >2 Mbit/s (proposed) presented in [26]. Time for the consump- User survey, 2 Mbit/s (estimate) tion of goods and services is in [26] 3000 "Calibration model" results explicitly introduced in a model includ- ing prices and income as well. As we 2500 have already indicated, the simplest pos- sible models including price, income and 2000 a utility function are rather complex. A model including time will be even more 1500 complex. As far as we have experienced, nobody has made estimates of price and 1000 time elasticities using such an approach. However the concept is theoretically very 500 valuable in understanding that to focus on price alone is an oversimplified app- 0 roach. An inquiry into this field was 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 made by [27]. The principle of relative constancy tells Figure 10 Total revenue from demand for access capacity 19942010 according to us that the spending on media compared results from the calibration model. Estimated revenue from demand for 2 Mbit/s to the total of consumer spending is more access according to results from user survey. Proposed revenue from demand for or less constant [28]. In the USA this bandwidths beyond 2 Mbit/s. (The input in the calibration model for annual price relationship has been more stable than in reductions is 15 %, similar to survey based forecast. The GDP in the calibration Norway. The spending on telephony has model is assumed to have an annual growth rate of 2 %.) increased somewhat in Norway in the last few years, even though the prices have been reduced quite a lot. The in- crease can be attributed to the tremend- ous increase in cellular telephony and the growth of the Internet. The Internet has certain shops, why not save time by The fact that the level of spending on probably another consequence: migration accessing similar shops on the net. Some media related applications is fairly con- from other media towards the Internet. people might even accept a higher price, stant is important, because it tells us that In addition the Internet may change the given that they save time. Figures 11 and there are some important limits as to how valuation of time. Instead of spending 12 indicate spending patterns in Norway much we are willing to spend on new a lot of time searching for something in and USA, respectively. services. The new applications have to be Telektronikk 2/3.1999 29

31 10 Media in percent of consumption 8 perceived as so valuable to the customer, 6 TV, video, photo, that the customer is willing to shift some sound of his/her expenditure from old and Reading 4 proven applications to new ones. Entertainment PC In the home office case this might be 2 Post viewed from a perspective where the Telephone customer (company and/or employee) 0 is willing to substitute the demand for 1968 1971 1074 1977 1980 1983 1986 1989 1992 1995 transportation to and from the office, to telecomms related demand. This is a pos- Figure 11 Media expenditures in percent of total expenditures, Norway 19681995. sible migration where media spending Source: Statistics Norway (SSB) is increased relative to other costs. The important question here is whether the employer and/or the employee will bene- fit from a positive payoff in terms of time and money when telecommuting substi- tutes physical transport. 12 Media in percent of consumption From an isolated point of view a new 10 product might seem very profitable or Television, radios, attractive because it allows high prices. 8 sound equipment But it should be viewed in the light of 6 Reading possible substitution effects, and how Entertainment much the public is willing to spend re- 4 lative to their total spending. Postage and 2 stationery 0 Telephone 8 Price elasticity 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 The classic approach to decide upon short-run profits is to find where margi- nal revenue equals marginal costs, which Figure 12 Media expenditures in percent of total expenditures, USA 19841996 is not quite valid in the telecomms indu- Source: Bureau of Labor Statistics (BLS, USA) stry. Such a solution may not necessarily maximize the long-run profits, or share- holder wealth. Assuming that the market can be described as monopolistic, one would expect prices to be relatively high (when compared to a competitive mar- ket). By keeping prices high and earning Profit $ monopoly profits, the monopolist en- courages competitors to enter the market. Instead of charging prices that maximize short run profits, one may decide to engage in limit pricing; where the new price is lower than the monopolistic Profit stream when behaving price. It is difficult to determine the like a monopolist correct price, whatever that is. Without delving into the subject, it does focus on competitive intelligence; how do you assess the threats and opportunities in the market? [29] Profit stream when applying limit pricing The following elements have to be con- sidered when looking at limit pricing (Figure 13): High initial price, followed by a rapid decline (assuming initial monopolistic behaviour); time Lower initial price, followed by a slower decline (assuming limit Figure 13 Pricing schemes depending on strategy and opportunity pricing). 30 Telektronikk 2/3.1999

32 In order to compare the approaches one 9 Setting the price level Price/Marginal revenue has to calculate the net present value of |E|>1 the potential profit flows generated by Introducing broadband access services each strategy. In turn, this implies that (based on ADSL technology) to the one has to decide upon the discount rate. P1 |E|=1 existing line of telecommunications Choosing a high discount rate will place |E|

33 vice they have no experience in using Semi-randomised call structure, Information retrieval/huge file transfer might be a hazardous basis for making a medium/high bandwidth: xDSL. may require high bandwidth; business and roll-out plan. On the other Work preparation and completion may hand there are few, if any, feasible alter- Our basic assumption is that the price require high bandwidth. natives except for a general and qualified level for ADSL has to be somewhere judgement. A recent service introduction between ISDN and Leased lines, Within this context it would be reason- in the UK might serve as an illustration: depending on usage patterns measured in able to focus on bandwidth on demand Telepoint in the UK was based on CT2 hours per day or month. In addition one rather than the use of a specific access cordless telephony standard. In the late has to differentiate between the demand technology. The access technology does eighties four operators were licensed to for switched and non-switched traffic. It only provide the user with the possibility introduce a public cordless telephony ser- is difficult to make such comparisons, to have a better peak performance, most vice Telepoint or Phonepoint. The ser- but the points made are illustrated in a of the time a lower bandwidth will suf- vice handled outgoing, though not in- fairly simple way in Figure 15. fice. At the same time this will probably coming calls from a telepoint base reduce the network load and thereby station within a range of 100 metres. At The low end user is assumed to spend reducing network costs. However, the least one of the operators made a com- few hours and have limited needs for local loop will be more expensive. prehensive market survey for the service, speed, and therefore ISDN will suffice. Instead of having a relatively high price including demonstrations of a dummy Further up the road, be it either more level for new broadband services, and a and thorough interviewing of potential hours or higher speed requirements, relatively slow penetration speed in the customers. The demonstrations and inter- ADSL or similar technologies will satisfy market, bandwidth on demand may give viewing took place in subway terminals, the user. Larger requirements are solved the opportunity to add a price premium if railway stations, etc. A pricing scheme by leased lines. In order to achieve this it is charged according to usage, and still was also developed according to what the structure it seems natural to have fairly be attractive to more customers. Stretch- respondents indicated about willingness high subscription prices, and a variable ing it a bit further, it makes sense to say to pay. The prices indicated by the price component based on time and/or that you are paying a price premium for respondents were fairly high. It turned data volumes. the bandwidth flexibility and still have a out to be more base stations than cus- cheap connection for everyday purposes. tomers when the expectations to the ser- From the customers point of view it is Figure 16 illustrates the bandwidth re- vice began to erode. Thoroughly per- more or less irrelevant whether the name quirements for the typical home office formed market surveys might very well of the access technology is ISDN, xDSL, user described above. fail. Success histories are often referred cable modem or whatever. The band- to, the unsuccessful ones are harder to width requirements will probably vary Most of the day 12 ISDN B-channels trace. However, the telepoint type of ser- according to the tasks that are being will do the job, whereas higher require- vices has been successful in Hong Kong performed. Let us consider the home ments are limited to a short time of day. and Singapore, among other things due to office for a moment. Most of the tasks In Figure 16 we have assumed a two- dense population and relatively few pub- performed at the home office do prob- price structure. However, it is probably lic phones. ably not require a high bandwidth. But a natural to introduce more flexible prices, few tasks do require a lot of it. The fol- ie. both bandwidth and time-of-day based One way to establish an initial price level lowing generalisations might be useful: prices. is to perform a substitution analysis. For Ordinary tasks, like writing, making the SME and SOHO segments it is natu- phone calls and sending faxes are not ral to look at the price/performance lev- critical in terms of bandwidth; els of ISDN and leased lines relative to usage per unit of time, or bandwidth- Work group document processing and hours. In Figure 15 it is assumed that video conferencing in particular do ISDN has a fairly simple cost structure require high bandwidth; for the customer. However, one has to take into account that the call duration per call may differ, and hence the slope of the line. Our simplification is based upon an assumption that the underlying traffic consists of large amounts of data, Bandwidth requirement and therefore we have assumed a long duration per call (on average). When looking at alternative local loop infrastructure, the customer has the Professional Non-professional Price premium following choices: Normal price use use Randomised call structure, low band- width: ISDN; Fixed call structure, high bandwidth: 08:00 16:00 23:00 time Leased lines; Figure 16 Bandwidth requirement during the day with a dual-price structure 32 Telektronikk 2/3.1999

34 References Communications, Oslo, 1996, vol. 2, residential market. IEEE Communi- 2.3 2.10. cations Magazine, 33, (2), 4449, 1 Riezenman, M J. Communications. 1995. IEEE Spectrum, 35 (1), 29-36, 1998. 16 Sykes, P, Sewell, R. Broadband Internet for the Mass market. 23 van Landegem, T, Prycker, M, van 2 Wright, D. Analysis of the Market for Telecommunication Journal of Aus- den Brande, F. 2005: A Vision of the Access to Broadband Telecommuni- tralia, 46 (1), 4752, 1996. Future Network. Electrical communi- cations in the Year 2000. Computer cation, (3), 231, 1994. Operations Research, 25 (2) 1998. 17 Burton, F. Residential Broadband Services: An European Study. British 24 Wasem O, J, Gross, A M, Tlapa, G 3 Analysys. ADSL Megabit Internet Telecommunications Engineering, 14 A. Forecasting Broadband Demand Access Arrives. Cambridge, 1997. (3), 250257, 1995. between Geographic Areas. IEEE Communications Magazine, 33 (2), 4 Bernstein, L. Managing the last mile. 18 Heer, D N, Maher, D P. The Hearth 5057, 1995. IEEE Communications Magazine, 35 of the New Information Appliance, (10), 7276, 1997. IEEE Transactions on Consumer 25 Svendsen, B, Sther, J P. Datakom Electronics, 41 (3), 869, 1995. 2000. Kjeller, Telenor Research 5 Forrester. DSLs Field of Dreams. Department, 1994. (TF report R Cambridge, Mass., 1997. 19 Furht, B. Where are Information 43/94.) Superhighways Headed? ACM Com- 6 Hamilton, S. E-Commerce for the puting Surveys, 27 (4), 554556, 26 DeSerpa, A C. A theory of the eco- 21st Century. IEEE Computer, 30 1995. nomics of time. Economic Journal, (5), 4447, 1997. 81, 828846, 1971. 20 Hopkins, M et al. A Multi-Faceted 7 Honcharenko, W. Broadband wire- Approach to Forecasting Broadband 27 Batt, C E et al. Consumer spending less Access. IEEE Communications Demand and Traffic. IEEE Commu- behavior and telecommunications Magazine, 35 (1), 2026, 1997. nications Magazine, 33 (2), 36, 1995. services. A multi-method inquiry. Elsevier Science Ltd. Telecommuni- 8 Khasnabish, B. Broadband to the 21 Newstead, A. The Australian cations policy, 22, (1), 2346, 1998. Home (BTTH): Architectures, Access Telecommunications Network Past, Methods, and the appetite for it. IEEE Present and Future. Telecommunica- 28 McCombs, M, Eyal, C. Spending on Network, 11 (1), 5869, 1997. tion Journal of Australia, 45, (2), 5, Mass Media. Journal of Communica- 1995. tion, 30 (1), 153158, 1980. 9 Kohn, D M. Providing Global Broad- band Internet Access using Low- 22 Stordahl, K, Murphy, E. Forecasting 29 McGuigan, Moyer. Managerial eco- Earth-Orbit Satellites. Computer Net- long-term demand for services in the nomics. 1989. works and ISDN Systems, 29 (15), 17631768, 1997. 10 Kwok, T C. Residential broadband Internet Services and Applications Georg Moe (38) has been financial advisor and market ana- Requirements. IEEE Communica- lyst in Telenor since 1993. He received a Bachelor of Busi- tions Magazine, 35 (6), 7683, 1997. ness Administration from the Norwegian School of Manage- ment in 1986, with main focus on managerial economics, 11 Mathias, C J. Satellite Services: financial analysis and marketing management/ strategy. He Broadband-Ready - Eventually. was engaged in developing decision support systems for Business Communication Review, 27 major transportation companies and the tax reform program (11), 5660, 1997. during the late 1980s and early 1990s and was part time teacher in managerial economics and econometrics at the Norwegian School of Management between 1990 and 1993. 12 OVUM. Broadband Wireless Market email: [email protected] Strategies. London, 1997. 13 OVUM. Interactive Multimedia Ser- vices to the Home. London, 1997. Jan-Petter Sther (52) has his background from the Uni- versity of Oslo as political economist with a specialisation 14 OVUM. Digital Subscriber Line in macroeconomic planning. He has worked at the Institute The Route to Broadband. London, of Transport Research, Norwegian Railways, Telenor Re- 1997. search and Development, Norwegian Post and Telecom- munications Authority and Telenor Nett. Demand for 15 Stordahl, K, Olsen, B, T, Ims, L A. established and new telecommunication services and Do we need a Pan-European Net- growth in bandwith demand have been the major fields work and What is the Demand for of interest in recent times. New Applications? In: Proc 22nd email: [email protected] European Conference on Optical Telektronikk 2/3.1999 33

35 Long term forecasts for broadband demand KJELL STORDAHL AND LARS RAND 1 Introduction Fibre to the building (FTTB); the speed in million instructions per sec- ond (MIPS) has increased proportionally Fibre to the home (FTTH). What are the market drivers for future [6]. In 1983 the cost per Mbyte was USD broadband demand? This paper shows 300, while in 1995 the cost was reduced In addition new multiplexing techniques, that the long term demand for broadband to 15 cents. Future exponential develop- access protocols for point-to-multipoint services depends on a set of different ment of the storage capacity will enable configurations and modulation tech- market drivers. Some of the market software decompression of MPEG-2 video niques are developed. Also the digital drivers like application evolution, devel- streams and direct computer storage. subscriber line (DSL) technologies, like opment of new technology and network ADSL (asymmetric digital subscriber architectures, terminal equipment tech- nology, mass production of network line), HDSL (high bit rate subscriber 1.3 Market drivers line), VDSL (very high speed digital components and tariff evolution are The new technologies, the mass produc- subscriber line) and SDSL (symmetric presented. tion of network components and low digital subscriber line) are of great transmission costs are continuously cre- importance for utilising twisted pairs [1, The long term demand for broadband ating new applications. At the same time 2]. The technologies may substitute each services is estimated based on an interna- an extraordinary expansion of the Inter- other or may be deployed as supplements tional Delphi survey. The experts partici- net has occurred. It seems that it is not a in different parts of the network. pating in the Delphi survey used infor- killer application for the broadband mar- mation about market drivers as a basis ket, but that Internet is a killer network. Introduction of passive optical network for their evaluation of the evolution of a From 1998 wideband services were components as TPON and ATM PON future broadband market. A comparison offered on the Internet, and broadband and the use of ATM- and SDH technol- is done between the expectations the ex- services are also expected to emerge ogy will increase the transmission capac- perts had about the market developments soon. At the same time some CATV ity and reduce the costs. Wireless broad- in 1994 and 1997. The results from the companies are installing cable modems band access is a technology currently Delphi survey are used to model analyti- and are offering broadband services on under development. The access radio cal forecasting functions for broadband their networks. Some of the main drivers technology is expected to evolve from demand. The aggregated forecasts for for the development of the broadband carrying narrowband services to transport specific broadband capacities are split market are: new technology, new appli- of services up to 2 Mbit/s capacity into asymmetric and symmetric broad- cations, increased computing power and through local multipoint distribution band forecasts. storage, mass production, price reduc- service (LMDS) and multipoint multi- tions, the Internet revolution and the channel distribution (MMDS) [35]. competition [7]. 1.1 Technology development Another relevant architecture in the and new network architec- future is the universal mobile telephone tures system, UMTS. 2 Prediction of network In the transport network deployment Other alternatives are satellite communi- component cost trends strategies for substitution of PDH trans- cation combined with a wireline return mission equipment with SDH transmis- Within the European programs RACE channel. The cable operators will up- sion equipment are now being carried and ACTS the projects RACE 2087/ grade their networks with return channels out. In parallel the fibre capacity is ex- TITAN, AC 226/OPTIMUM and AC offering both POTS/ISDN, Internet and panded by the introduction of wavelength 364/TERA have developed a methodol- broadband services together with CATV. division multiplexing (WDM). Over the ogy and tool for calculation of the overall The most relevant architecture is a com- last years the development of new tech- financial budget of any access architec- bination of passive optical network and a nology has dramatically reduced costs ture. The tool handles the discount cost coax droop called hybrid fibre coax sys- by significant expansion of the system system, operations, maintenance, life tem, HFC. The access technologies men- capacity. During a 20 year period the cycle costs and the cash balance. This tioned may substitute each other or may transmission cost per capacity unit has enables a comparison of various optical be deployed as supplements in different been reduced from 10,000 to 1. However, or hybrid architectures through a global parts of the network. the technical problem of high capacity system assessment. The tool has the abil- switching is not yet solved. One possibil- ity to combine low level, detailed net- ity is to use ATM, another possibility is 1.2 Terminal equipment work parameters of significant strategic to use IP, and a third one is to implement technology relevance with high level, overall strate- ATM over the IP platform. gic parameters for performing evaluation The terminal equipment is evolving of various network architectures [1, 811]. rapidly into several future options includ- In the access network a wide range of ing specific electronic interfaces/termi- fibre architectures are relevant, of which The TITAN project developed a method- nals which may be used together with a deployment depends on factors such as ology based on an expansion of the TV, like a network computer. Another the subscriber area: Wright and Crawfords learning curve possibility is the use of a PC. There are models to predict future cost of the net- Hybrid fibre coax (HFC); several drivers connected to the termi- work components [1214]. In the OPTI- nals. During the last 20 years, from the Fibre to the cabinet (FTTCab); MUM project, Wright and Crawfords 8080 to the Pentium processor, the num- learning curve models for cost predic- Fibre to the node (FTTN); ber of transistors per chip has doubled tions were examined. The models for every 18 months (Moores law), while Fibre to the curb (FTTC); cost predictions were extended not only 34 Telektronikk 2/3.1999

36 to estimate the costs as a function of changed following a consideration of 4 Private communications services number of produced units, but also as a results from the previous round. The (Symmetric) function of time. The cost prediction of procedure can be repeated a number of Videophone each network component is described by times and usually leads to a reduction in expansion of the learning curve given as the variance of the answers received. Teleconferencing a function of the parameters: Medians are used as a measure in the 5 Teleworking Delphi survey because they are more f(0) the predicted costs at time 0; (Symmetric and asymmetric) robust estimators than the mean value n(0) the relative proportion of pro- and standard deviations, and also less Videophone duced components at time 0; affected by extreme answers. Joint editing/publishing t the time interval between 10 % In 1994 the TITAN project carried out Teleconferencing and 90 % penetration; an international postal Delphi survey Teleparticipation K the learning curve coefficient on broadband service demands among (relative decrease in the cost by experts in ten European countries Information retrieval the double production). [1415]. An additional comprehensive Multimedia applications; two-round, on-site Delphi survey was The extended learning curve function is: carried out during the OPTIMUM work- 6 Telelearning shop Techno-economics of Multimedia (Symmetric and asymmetric) f(t) = f(f(0), n(0), t, K, t) Networks in Aveiro, Portugal in Octo- = f(0)[n(0)1 (1 + exp[ln(1/n(0) Video on demand ber 1997. The following countries were 1) 2t ln9/t])1]log2K represented in the survey: Belgium, The Videophone Czech Republic, Denmark, Finland, Virtual reality; France, Switzerland, Germany, Greece, The parameters in the learning curve: Holland, Hungary, Ireland, Italy, Nor- f(0), n(0), t and K are given in the way, Portugal, Spain and Sweden. The OPTIMUM cost database, which con- number of participants were 36 in the tains more than 200 different network first round and 32 in the second round. components. The components are grouped in volume classes. The values used for the various volume classes are 3.1 Broadband applications shown in Table 1. In the same way the K Table 1 Variation in n(0) and t for each volume class The Internet development and the new parameter is estimated based on type of technology continuously create new component. The K value indicates how Volume class n(0) t applications. To be able to evaluate the much the component price is reduced different broadband applications, they are by a doubling of the production. 1 0.5 5 divided into groups. The questions in the survey do not address single applications, 2 0.1 5 In the cost database all components are but the main group of applications. The listed with a given n(0), t and K value 3 0.01 5 main groups of applications in the study in addition to the estimated cost f(0) at are: 4 0.5 10 time 0. Then the extended learning curve is uniquely defined and the prediction of 1 Tele-entertainment 5 0.1 10 the costs is determined. (Symmetric and asymmetric) 6 0.01 10 Multimedia telegame Table 2 shows that new components 7 0.001 50 based on electronics or advanced optics Virtual reality experience a significant price reduction. Video on demand When the production is doubled, the price is reduced by 20 % and 30 % Audio/music on demand; respectively. An additional doubling of 2 Information services (Asymmetric) Table 2 The K values for component groups the production will reduce the cost by 36 % and 51 % respectively. Information retrieval Component group K value Electronic magazines 3 The Delphi survey Information retrieval by intelligent Civil work 1 agents Copper 1 A Delphi survey is a method by which the opinions of experts are canvassed, in Electronic newspaper; Installation 1 order to achieve consensus on a particu- 3 Teleshopping (Asymmetric) Sites and enterprises 0.95 lar issue. The methodology involves asking a set of questions, analysing the Teleshopping Fibre 0.9 results and resubmitting the questions to Advertising; Electronics 0.8 the experts, together with a summary of the first round results. The experts then Advanced optical components 0.7 resubmit their opinions, which may have Telektronikk 2/3.1999 35

37 Example of application: Video on demand and Audio/Music on demand General description: This is an application where a video library is accessed, and programs may be ordered and transmitted to the home. This appli- cation could substitute some part of the time spent on ordinary TV and part of the money spent on hiring videos in video shops. Technical assumptions: The user may use either an advanced telephone or a PC to communicate with the video library. The transmission of the video may be done either via a Cable-TV network or a telecommunication network. The access capacity will be in the range of 2 4 Mbit/s. Television Video library Cable-TV Telephone PC outlet Tele- communication Service- network access Video Decoder Home The video-transmission can either be done over the cable-TV network, or the telecommunication network. Given the following alternative prices per hour (1997 ECU), what do you believe will be the expected use of this group of applications (Tele-entertainment)? Note: We assume that the tele-entertainment applications are supplementary to the traditional TV channels, but there may be some substitution effects. Round 1 Prices per hour: 0.5 ECU 2 ECU 5 ECU 10 ECU 20 ECU Minutes per day: 70 40 12 5 1 Having seen the above results, what would your answers be to the corresponding question today? Round 2 Prices per hour: 0.5 ECU 2 ECU 5 ECU 10 ECU 20 ECU Minutes per day: 4 Comments (if any): Figure 1 An example from the Delphi survey questionnaire 7 Telecommunity 3.2 Access capacity 2 4 Mbit/s asymmetric access includ- (Symmetric and asymmetric) ing a 384 kbit/s symmetric upstream The technology and network components capacity; Telesurveillance are rather expensive today, but mass pro- duction may exponentially reduce pro- 25 Mbit/s asymmetric access including Videophone duction costs and consequently the a 384 kbit/s symmetric upstream Telediagnostics. prices. The following access capacities capacity; were examined in the Delphi survey: 36 Telektronikk 2/3.1999

38 25 Mbit/s asymmetric access including 3.3 The Delphi questionnaire expensive, depending on factors like new a 6 Mbit/s symmetric upstream capa- technology, sales volume, competition, The questionnaire starts with a short city. etc. Broadband communications costs description of the application, followed can be divided into four elements: by some questions relating to it. The Several factors contribute to an applica- main questions in the survey are: Costs for necessary terminal equip- tions requirements in terms of band- ment; width over the network. In most cases, Usage as a function of charge; high capacity is needed for large volumes Subscription charges; Penetration as a function of charge; of information. Traffic charges; Penetration as a function of time First of all, the type of medium (speech, (forecast); Charges for delivered information (eg. text, graphics, video, or several media charge for hiring/ordering a video). Demand as a function of disposable simultaneously multimedia), may indi- household income. cate the possibility for large volumes of In the Delphi survey we were interested information. The types of applications in how much the customers are willing An illustration of the design of the ques- demanding high bandwidth transmission to pay for the traffic and the subscription tions in the questionnaire which includes are fast transfer of video, high quality (connection) charges. It was assumed a description and an illustrative figure is images/graphics, large data files, or a that the customers already possess the given in Figure 1. simultaneous combination of these in necessary terminal equipment like TV, multimedia applications. PC, etc. Supplementary expenditure for As shown in Figure 1, in the second specific adapters and interface hard- round of the survey the participants were High quality videophone, telecommunity ware for the applications which have to presented the medians from the same and telemedicine are applications bene- be installed, was assumed to be covered questions in the first round of the survey. fiting from a high symmetric upstream by a subscription (connection) charge. The participants took this into considera- capacity. For example, interactive video The costs of the delivered information tion when answering the second round applications transferring moving pictures were not taken into account. questions. like videotelephony or videoconferences, require a minimum image frequency Some information may be financed by In order to use the presented applications, depending on the speed of change in the advertisements. The teleshopping appli- an access in the range of 128 kbit/s pictures transferred. This increases the cation may be financed by the sellers and 25 Mbit/s is needed. The users will have bit rate requirement, and so does the not by the customers. The costs of other to pay more for enhanced performance image resolution and colour richness. information like electronic newspapers and quality generated by higher band- may be substituted by a reduction of width. All the equipment and network costs for buying hard copies (traditional components will gradually become less Table 3 Leading Group of Applications, percentage score Choices 1994 survey *) Answers 1994 *) Choices 1997 survey Answers 1997 Video on demand 28 % Teleworking 28 % Home office 27 % Information services 25 % Videotelephony 18 % Tele-entertainment 24 % Remote education 8% Teleshopping 7% Multimedia telegames 7% Private communications services 6% Home ordering system 4% Telecommunity 4% Interactive TV/specialized channels 4% Telelearning 3% Electronic newspapers 3% Others (Telebanking) 1% Advertising and marketing 1% Telecommunity 0% *) Source [15] Telektronikk 2/3.1999 37

39 newspapers). Nevertheless, it is reason- budget for newspapers, magazines, 4.2 Ranking groups able to believe that the customers have dictionaries, specific books and videos, of applications to pay for some type of information. etc. The respondents were asked to point out the three most important services for the The household has both a time budget 4.1 Household budget and a financial budget, dependent on the future. This makes up 33 % the highest and usage number of persons in the household. The possible score for an application group. Table 3 shows the ranking of the group A household has an annual disposable budgets limit the use of applications. It is of applications in 1994 and in 1997. income, which is the income after tax or reasonable to expect some substitution Teleworking, information services and the part of the income which is available effects between the households use of tele-entertainment stand out as the antici- for purchasing goods and services, for time today and possible use of broadband pated most popular services for broad- savings etc. Part of the service budget is accesses. In the questionnaire we have band. A comparison with a similar Del- related to asked how much additional time a house- phi survey in 1994 [15] shows that the hold would spend on the new broadband budget for entertainment; three most promising broadband applica- applications as a function of additional tions were video on demand, home office budget for telecommunications; payment. and videophony. It looks like information services have become more popular from the 1994 survey to the 1997 survey. In addition, from the first to the second sur- vey other application groups have 250 become more interesting than private communications services (videophony). Teleworking Business 200 A telecommunication access line may Teleworking Private support the use of many of these groups Minutes per day Telelearning of applications, so for each group of 150 applications questions were asked on the demand at different prices in order to quantify the demand. 100 4.3 Potential usage 50 of applications For every service the respondents were asked to indicate the demand in minutes 0 per day for a given set of prices per hour. 0,0 5,0 10,0 15,0 20,0 Hence a demand curve can be con- Price per hour (euro) structed for each application group based on medians from the survey. In Figure 2 Figure 2 Demand curves for broadband access for teleworking and telelearning a distinction is made between company paid teleworking and teleworking paid by the households themselves, since compa- nies are expected to have a higher will- 70,0 ingness to pay than private households. 60,0 The demand for telelearning is trickier Tele-entertainment because it concerns a small share of the Information services households and for a limited time of the 50,0 year. The household usage will be high Telecommunity Minutes per day some days and zero at other times 40,0 Private communications services depending on the type of courses and Teleshopping education frequency. Demand curves for 30,0 other private broadband application groups are shown in Figure 3. The results 20,0 show that tele-entertainment follows the same demand curve as telecommuting 10,0 from a price of five euro per hour, but has a higher saturation level so the 0,0 expected demand is much higher at a 0,0 5,0 10,0 15,0 20,0 low price. Tele-entertainment services, Price per hour (euro) which are defined as video on demand, audio/music on demand, multimedia tel- egame and virtual reality, are very attrac- Figure 3 Demand curves for other private broadband applications groups tive services, but are quite price elastic. 38 Telektronikk 2/3.1999

40 Figure 4 shows medians for round 1 and round 2 and identifies the range from the 20 25 quartile to the 75 quartile of the answers on demand for broadband con- nections for different hourly prices. The Prices per hour (euro) Round 2 figure indicates significant reductions in 10 Round 1 the variance of the answers received in the first round compared to the second round. 5 Teleworking, information services and tele-entertainment stand out as the most 2 promising broadband applications in the future. The interest for information ser- vices can be explained by the rapid 0,5 development of the Internet and the related narrowband applications. The 0 20 40 60 80 100 120 140 160 180 interest for tele-entertainment is caused Minutes per day by a high degree of usage of existing applications. Teleworking is of special Figure 4 Demand curves for broadband access all applications interest and may be one important driver for the broadband market. Teleworking is used by self employed persons with their office at home (SOHO), by one person in the family financed by the company, or by some in the family, but financed inter- nally. Today there is a positive trend 60,0 towards supporting teleworking at home. 2-4 Mbit/s (1997) Percentage of the residential market For employees with qualified and inde- 25 Mbit/s downstream and 384 kbit/s upstream (1997) pendent work there are reasons to predict 50,0 25 Mbit/s downstream and 6 Mbit/s upstream (1997) 2 Mbit/s wideband (1994) that society and the companies are will- 8 Mbit/s broadband (1994) ing to support and finance extensive use 40,0 of teleworking with a broadband connec- tion. In that way the employees can work more effectively and in a more flexible 30,0 way. Society also supports teleworking because of reduced pollution and reduced 20,0 traffic at rush hour times, etc. Some large companies now offer a home office solu- tion combined with a company paid nar- 10,0 rowband access (N-ISDN) for some of their employees. 0,0 2000 2005 2010 2015 Saturation 4.4 Demand forecasts Year The respondents were asked to indicate Figure 5 Forecast for broadband access in the residential market the expected penetration in the residential market for broadband access for the years 2000, 2005, 2010, 2015 in addition to the saturation level. Figure 5 indicates quite a high demand for broadband con- nections in the residential market. The Table 4 Revised broadband forecasts as percentage of the residential market penetration forecasts for 2 Mbit/s have not changed much from the 1994 Delphi survey. In 1994 the experts predicted, for Access capacity 2000 2005 2010 2015 Saturation 2 Mbit access, a penetration rate of 5 %, 2 Mbit/s 2 12 23 40 50 10 % and 15 % respectively for the years 2000, 2005 and 2010. In the 1997 survey 8 Mbit/s 0.5 5.5 14 22 40 the forecasts are 4 %, 12 % and 23 % respectively for the years 2000, 2005 and 26 Mbit/s 0.1 3 9 15 25 2010. The difference in the predictions is Sum 2.6 20 45 75 *) somewhat larger for faster connections, but the results show the same pattern. *) The saturation for the various accesses will occur at different points in time While the experts in 1994 expected a Telektronikk 2/3.1999 39

41 Table 5 Parameter estimates and multiple correlation coefficient for broadband pene- for the first years. Therefore, the para- tration forecast functions meters in the model are estimated by ordinary least squares regression (OLS) for different values of . The OLS esti- Parameter estimates a b g M R2 mation is based on the following trans- formation: 2 Mbit/s - 0.07496 - 0.19266 5 50 98.56 ln((M/Yt)1/ 1) = + t 8 Mbit/s - 4.79468 -0.13249 500 40 99.38 26 Mbit/s - 4.57674 - 0.15775 500 25 99.37 The saturation level M and the parameter are fixed values in the estimation pro- cess. M is found from the Delphi data, while is estimated by systematic calcu- lations of RMSE (root mean square error) for a set of different values. The penetration rate for 8 Mbit/s of 1 %, 2 % model with three parameters give a rather multiple correlation coefficient, R2, for and 5 %, the respective penetration fore- good fitting. The model is defined by the the models is rather high. The estimated casts for 25 Mbit/s in the 1997 survey are following expression: values are given in Table 5. 2 %, 5% and 13 %. Yt = M / (1 + exp( + t)) The broadband penetration forecasts are where the variables are defined as fol- shown in Figure 6. 4.5 Analytical forecasting lows: functions Yt Demand forecast at time t 4.6 Symmetric and asymmetric The development of analytical forecast models for broadband access was a part demand modelling M Saturation level of the OPTIMUM project. The results The forecasts have to be divided into t Time from the Delphi survey contain only asymmetric and symmetric demand. 2 Mbit/s and 25 Mbit/s accesses. There , , Parameters. Introduction of analytical functions are are reasons to believe that also 8 Mbit/s convenient for describing the share of will be a conventional offered access. The parameters , , cannot be esti- asymmetric and symmetric accesses. The Evaluation of the results shows that the mated simultaneously by ordinary least question is how the symmetric demand sum of the two 25 Mbit/s gives about the squares regression since the model is will develop compared to the asymmetric same demand as the 2 Mbit/s. During the non-linear in the parameters. The main demand. The symmetric demand will first ten years the demand for 2 Mbit/s objective in the fitting is not to get the probably be low for the first years com- will probably be significantly higher than best overall fit, but a reasonably good fit pared to the asymmetric demand. After 25 Mbit/s. Since the total demand for 25 Mbit/s seems optimistic, it has been suggested to transfer 25 Mbit/s with 384 kbit/s return demand to a 8 Mbit/s demand. In addition it has been sug- gested to split the given demand in a symmetric demand and an asymmetric demand. Since 8 Mbit/s is a lower cap- 50,00 acity than 25 Mbit/s it has been sug- gested to increase demand by 10 %. 45,00 In addition we will use 26 Mbit/s, which 2 Mbit/s is closer to the new standard than 40,00 8 Mbit/s 25 Mbit/s. The forecasts for 8 Mbit/s Penetration (%) 35,00 26 Mbit/s and 26 Mbit/s for the year 2000 is also reduced to 0.5 % and 0.1 % respectively. 30,00 The revised forecasts are found in Table 4. 25,00 20,00 The demand forecasts in the table in- clude both symmetric and asymmetric 15,00 accesses. The fraction between asymmet- ric and symmetric will change over time, 10,00 but during the first years, there will 5,00 mainly be asymmetric accesses. The models developed in the OPTIMUM 0,00 project are based on the results from the 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 2020 1997 Delphi survey. Different analytical forecasting models for fitting the Delphi data are tested. The extended Logistic Figure 6 2 Mbit/s, 8 Mbit/s and 26 Mbit/s broadband forecasts 40 Telektronikk 2/3.1999

42 some years the symmetric demand 30 % will probably have a relatively higher increase. In the end we assume that the proportion of symmetric subscriptions 25 % will converge to a given level. One 2 Mbit/s important element is how the PCs are 8 Mbit/s 20 % used as broadband terminals, either for communication with specific information 26 Mbit/s sources, or for communication between 15 % users. The behaviour may be modelled by constructing analytical functions 10 % defining market shares as a function of time between the asymmetric demand and the symmetric demand. The analyti- 5% cal functions should be simple. 0% It is suggested to use the Logistic model 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 to describe the evolution of the distribu- tion of asymmetric and symmetric Figure 7 Proportion of the symmetric communication penetration of demand. The following parameters the total broadband penetration demand. The given assumptions lead to are defined: forecasts for asymmetric and symmetric demand as shown in Figure 8 S Saturation level St Share of symmetric demand t Time 20 T Time to 50 % saturation 18 2 Mbit/s asym a Growth per year 16 2 Mbit/s sym , Parameters in the Logistic model 8 Mbit/s asym 14 (Model 2). 12 8 Mbit/s sym The model is given by: 10 26 Mbit/s asym St = S / (1 + exp( + t)) 8 26 Mbit/s sym The Logistic model is symmetric on 6 both sides of S/2. The model is uniquely 4 defined if S, and are defined. Instead 2 of defining the parameters, we have decided to determine the function by 0 the following assumptions: 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 1) Define the saturation level S; Figure 8 Forecasts for asymmetric and symmetric connections (%) 2) Define the time (number of years) T, until half saturation is reached; 3) Define the market share S0 at time 0, which is the year of introduction. 26 Mbit/s is estimated to be eight years, demand curves for 2 Mbit/s, 8 Mbit/s and The parameters in the model are found by: and the starting proportion of symmetric 26 Mbit/s broadband connections. The broadband communication demand is difference between the demand curves is =*T estimated to be 2 %. The distributions very small and indicates that residential = (1/T) * ln(S/S0 1) are shown in Figure 7. users are not willing to pay much more for a high capacity connection despite The degree of symmetric demand de- the better quality. pends on the offered broadband capacity. 4.7 Demand for access The analytical specification differs, de- capacities pending on connection capacity. The sat- 4.8 Analytical demand models The access lines with different capacities uration for 2 Mbit/s symmetric demand may support the use of many of the ear- As a part of the OPTIMUM project ana- is suggested to be 40 % in the long run, lier mentioned applications. Thus, ques- lytical demand models dependent on while the 8 Mbit/s and 26 Mbit/s sym- tions were asked on the demand for price were developed. Based on the same metric demand is suggested to be 30 % broadband access for three different arguments as for analytical forecasting and 25 % respectively. The time to reach access types as a function of annual functions, 2 4 Mbit/s, 25 Mbit/s with a half saturation for 2 Mbit/s, 8 Mbit/s and costs. Figure 9 shows the estimated narrowband return channel and a broad- Telektronikk 2/3.1999 41

43 30 around 10 gives a rather good fit. The framework for the demand curves is Precentage of the residential market % described hereafter. It is important to 25 underline that the tariff in this context 2 Mbit/s consists of both a one year subscription 20 tariff and also a usage tariff based on the 8 Mbit/s expected traffic during one year. The 26 Mbit/s methodology described in the previous 15 sections is used to predict the tariff evo- lution for broadband connections. The 10 predictions are calculated in the follow- ing steps: 5 The tariff p is found by transforming the demand model to the formula: 0 0 1000 2000 3000 4000 5000 6000 pt = [(ln yt)1/ ]/ Annual cost (euro) The parameters , and are found by the above equations. Then the tariff pre- Figure 9 Demand curves for broadband access as a function of annual cost dictions for the years 2000, 2001, ..., 2010 are determined by inserting the demand forecasts {yt} in the same years. The tariffs are found in Table 6. The tar- iff evolution for broadband services in band return channel are transferred to determined by minimising the squared the mass market is shown in Figure 10. 2 Mbit/s, 8 Mbit/s and 26 Mbit/s. The distance between the demand curve and suggested demand model based on three the results from the Delphi survey. parameters is: 4.9 Willingness to pay Now, let the initial values be: for access capacity y = e(+p) Willingness to pay as a function of dis- (yI, pI) and (yL, pL) y Demand; posable household income is estimated for broadband access based on answers p Price; Hence: from the respondents. Disposable income , , Parameters in the model. = [(ln yL)1/ (ln yI)1/] / (pI pL) is the household income after tax, ie. the part of the income that is available for = (ln yI)1/ pI The parameter estimates are found by saving and purchasing goods and ser- OLS regression for a given set of val- vices. Figure 11 shows that households The parameter is found by minimising ues. A variant of this model is based on with an annual disposable income in the the following expression: the assumption that the demand is 100 % 10,000 to 15,000 euro range cannot 2 when the price is 0. Evaluation of the Q( ) = (yi e[( )+( )pi ] ) afford to pay more for a high capacity results show that the fitting is not satis- connection. Incremental willingness to factory. To improve the fit, the para- In the non-linear estimation procedure, pay for broadband access is very small, meters and are determined such that not only the last equation is minimised even for wealthy households. the demand curve passes through the two but also the first years achieve a reason- initial points, while the parameters are ably good fit. For all models equal to Table 6 Assumption tariff evolution for broadband services (mass market) Parameters 2 Mbit/s 2 Mbit/s 8 Mbit/s 8 Mbit/s 26 Mbit/s 26 Mbit/s asym sym asym sym asym sym Demand, year 2000 1.900 % 0.037 % 0.637 % 0.013 % 0.150 % 0.003 % Demand, long run 40 % 40 % 30 % 30 % 25 % 20 % Tariff, year 2000 1800 euro 2700 euro 3240 euro 4860 euro 5192 euro 7788 euro Tariff, long run 500 euro 750 euro 900 euro 1350 euro 1442 euro 2163 euro 42 Telektronikk 2/3.1999

44 For households with an annual dispos- 9000 able income of between 25,000 and 2 Mbit/s asym 60,000 euro the willingness to pay for 8000 2 Mbit/s sym subscription and traffic charges for the 7000 8 Mbit/s asym highest capacity access is only 2 % of the households disposable income. 6000 8 Mbit/s sym 26 Mbit/s asym 5000 26 Mbit/s sym 4.10 Price and capacity 4000 The previous sections have shown a low 3000 willingness to pay for higher capacity and better quality. The questionnaire 2000 also included direct questions on the 1000 households willingness to pay for increased capacity relative to a 128 kbit/s 0 access, ie. an ISDN basic access. Figure 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 12 confirms a low willingness to pay for incremental increased connection capac- Figure 10 Tariff evolution for broadband services (mass market) euro ity. The difference between the 75 % quartile (25 % answered higher) and the 25 % quartile (25 % answered lower) are shown in the figure as a measure of the variation in the answers. The uncertainty increases with increased capacity. 1 400 2 Mbit/s 2% 1 200 5 Conclusions Annual charges (euro) 8 Mbit/s 1 000 The results from the 1997 Delphi survey 26 Mbit/s 2% show that there will be a substantial 800 demand for broadband services in the residential and SOHO markets during the 600 2% next ten years. However, the households are not willing to pay too much more for 400 additional broadband applications and 1% additional capacity. Households with low 200 disposable income will not afford to have 0.3 % a subscription, while households with a 0 reasonable disposable income are willing 10,000 20,000 30,000 40,000 50,000 60,000 to pay up to 2 % of their disposable Disposable income (euro) income. The possibilities for substitution effects between new and old media Figure 11 Income spent on broadband accesses as a function (newspapers, magazines, video rental, of disposable income video games, etc.) are taken into account. The Delphi survey indicates that cus- tomers will be unwilling to pay much 350 more for increased capacity. It is interest- ing to see that the demand curves for 300 300 75 quartil 2 Mbit/s, 8 Mbit/s and 26 Mbit/s are quite 50 quartil 250 Index Annual Price similar. The results are supported by the 25 quartil 240 price/quality question where the experts 220 200 200 200 indicate that the households are willing to 190 176 pay twice as much for a 50 Mbit/s access 150 150 compared to how much they are willing 120 145 to pay for an ISDN BA access, and only 100 113 2.2 times as much for a 500 Mbit/s 100 access. This is a quite important finding, 50 because a 50 Mbit/s access is possible using VDSL modem for customers with 0 short subscriber lines, while 500 Mbit/s is 128 kbit/s 384 kbit/s 2 Mbit/s 50 Mbit/s 500 Mbit/s impossible because of the physical limita- tion on the twisted pair. To offer 500 Capacity Mbit/s access an FTTH solution will Figure 12 Willingness to pay for increased capacity relative to 128 kbit/s Telektronikk 2/3.1999 43

45 probably be needed, but the customers are In addition analytical broadband fore- 5 Welling, I. Internet access and Video not willing to pay more than 2.2 times the casting functions and demand functions on demand. In: Proc. IBC Broadband price of an ordinary ISDN access! are developed together with forecasts strategies the battle for customer for asymmetric and symmetric demand. access. London, UK, 45 Dec 1997. Comparison of the results from the 1997 The analytical forecasting functions and Delphi survey and the one carried out in demand functions are modelled based 6 Moore, G. Electronic Materials Sym- 1994 shows that the results are rather on the results from the Delphi survey. posium, Santa Clara, USA, Mar 1998. similar. It is interesting to note that the penetration forecasts for a 2 Mbit/s 7 Stordahl, K, Olsen, B T, Ims, L. Do access for the years 2000, 2005 and 2010 References we need a pan-European network are quite close, with the 1997 forecasts and what are the demand for appli- 1 Ims, L A et al. Evolution of technolo- being a little bit higher. For higher access cations? Invited paper 22nd Euro- gies and architectures to a full service capacities we see the same pattern. Usu- pean Conference on Optical Commu- network. In: Proc. IBC Broadband ally, the experience when comparing old nications. Oslo, Norway, Sep 1996, strategies the battle for customer forecasts of new telecommunication ser- 2, 2.32.10. access. London, UK, 45 December vices to new forecasts is that the old fore- 1997. casts have been too optimistic. The fore- 8 Olsen, B T et al. Techno-economic casts of the Internet evolution is of evaluation of narrowband and broad- 2 Ims, L A (ed.). Broadband access course an exception. band access network alternatives and networks : Introduction strategies evolution scenario assessment. IEEE and techno-economic evaluations. Like in the 1994 Delphi survey, the vari- Journal of Selected Areas in Commu- London, Chapman-Hall, 1998. ation in the answers among the experts in nications, 14 (6), 11841203, 1996. the 1997 Delphi survey was significantly 3 de Passoz, G, Clausse, J L, Karam, reduced from round 1 to round 2. The 9 Ims, L A et al. Multiservice access G. Can digital techniques give a new variation was measured by 25 % and network upgrading in Europe : a boost to MMDS? In: Proc. 11th 75 % quartiles. The results indicated that techno-economic analysis. IEEE International Symposium on Sub- it was unnecessary to carry out an addi- Communications Magazine, 34 (12), scriber Loops and Services (ISSLS). tional round. 124134, 1996. Melbourne, Australia, 49 Feb 1996, 134139. To realise the potential broadband de- 10 Olsen, B T et al. PNO and Cable mand, a key option is the development Operator broadband upgrade technol- 4 Nowak, R J. Wireless : part of the of the broadband drivers mentioned in ogy alternatives : a techno-economic broadband future. In: Proc. 11th the introduction. Models to predict cost analysis. In: Proc. Optical Fiber International Symposium on Sub- trends for network components show that Conference 1996 (OFC 96). San scriber Loops and Services (ISSLS). increased production gives significant Jose, USA, 25 Feb 1 Mar 1996. Melbourne, Australia, 49 Feb 1996, reduced cost. 122127. 11 Ims, L A, Stordahl, K, Olsen, B T. Risk analysis of residential broad- band upgrade in a competitive envi- ronment. IEEE Communications Kjell Stordahl (54) received his M.Sc. degree in statistics at Magazine, 35 (6), 96103, 1997. the University of Oslo in 1972. He worked with Telenor Re- search Dept. for 15 years, seven of which as manager for 12 Wright, T P. Factors affecting the the teletraffic field. He joined Telenor Nett in 1989 and was cost of airplanes. Journal of Aero- manager in the Planning Department until 1996. From 1997 nautic Science, 3 (4), 122128, 1936. to 1999 he was manager for Market analysis in the Market Division, Telenor Nett. He has participated in various Euro- pean projects and authored or co-authored more than 90 13 Crawford, J R. Learning curve, ship papers in international journals and conferences. curve, ratios, related data. Lockheed Aircraft Corporation, 1944. email: [email protected] 14 Stordahl, K et al. Delphi survey : forecasting the demand for wide- and broadband services. From RACE 2087/TITAN. In: Proc. 6th Interna- Lars Rand (34) is a graduate engineer from the Norwegian University of Science and Technology from 1990. He tional Workshop on Optical Access worked for seven years as researcher at the Institute of Network. Kyoto, Japan, Oct 1994, Transport Economics with modelling and analysis of private transport. Since 1997 he has been working as forecast analyst in Telenor Nett. 15 Stordahl, K, Murphy, E. Methods for email: [email protected] forecasting long term demand for wide and broadband services in the residential market. IEEE Communi- cations Magazine, 13 (2), 449, 1995. 44 Telektronikk 2/3.1999

46 Broadband demand survey in the residential and SOHO market in Norway SYNNVE ISTAD AND KJELL STORDAHL 1 Introduction 3 Technical prerequisites 4 Broadband applications A Delphi survey has been carried out to and facts The different broadband applications make demand forecasts of broadband ser- under consideration are in the question- The applications in the questions concern vices in the Norwegian residential and naire divided into groups. The questions the use of different media such as voice, SOHO (Small Office Home Office) mar- in the survey do not address single appli- data/text, graphics, live images, or a ket for the next five to ten years. A Del- cations, but the main groups of applica- combination of these, so-called multime- phi survey is a method where the opin- tions. The main groups of applications in dia. To be able to offer transmission of ions of experts are canvassed, in order to the study are listed in the following (the all of these media on the same subscriber achieve consensus on a particular issue. same list as used in [1]): line we need a rather high capacity. The The methodology involves asking a set capacity on the copper subscriber line 1 Tele-entertainment of questions, analysing the results and can be upgraded by introduction of (Symmetric and asymmetric) resubmitting the questions to the experts, ADSL modem and VDSL modems. together with a summary of the first Multimedia telegame The capacity can be expanded to 2 round results. The experts then resubmit 50 Mbit/s depending on the length of Virtual reality their opinions, which may have changed the subscriber line. The introductory following consideration of results from Video-on-demand information given on the different tech- the previous round. The procedure can be nologies was similar to the one used in Audio/music on demand repeated a number of times, and usually the 1997 Delphi survey [1]. it leads to a reduction in the variance of 2 Information services (Asymmetric) the answers received. The Delphi The access technology opens for high Information retrieval methodology involving experts was used capacity. In the questionnaire the follow- instead of performing a market survey Electronic magazines ing access capacities are introduced: including family members. The reason is Information retrieval by intelligent that we expected it would be difficult for 2 Mbit/s asymmetric access including agents residentials to give accurate answers to a 384 kbit/s symmetric upstream rather complicated questions about future capacity; Electronic newspaper broadband demand. 8 Mbit/s asymmetric access including 3 Teleshopping (Asymmetric) a 384 kbit/s symmetric upstream Teleshopping 2 The Delphi survey capacity; Advertising 25 Mbit/s asymmetric access including The Delphi survey was carried out in a 384 kbit/s symmetric upstream 4 Private communications services September/October 1998. Questionnaires capacity. (Symmetric) for the first round were sent by post to experts, intending to participate at a con- Videophone In addition to the technical access solu- ference on Broadband communication tions different mechanisms influence the Teleconferencing and multimedia organised by the Nor- demand for broadband communication in wegian Research Board (NFR). The par- 5 Teleworking the residential and SOHO market. Estab- ticipants were experts from the sectors of (Symmetric and asymmetric) lishment of broadband connections for telecommunications, research and uni- teleworking will mainly be paid by com- Videophone versities. Approximately 90 people par- panies. In Norway a lot of companies are ticipated at the conference. In the first Joint editing/publishing now offering a PC to their employees. round of the survey 28 answers were The PC can be used as a broadband ter- Teleconferencing received. During the conference the minal in the future. Telelearning is an experts were asked to fill in the Delphi Teleparticipation application which could be partly questionnaire for the second round. Only financed by public means. Traffic costs Information retrieval 17 questionnaires were received. Because for teleshopping are supposed to be in- of the high non-response and also the Multimedia applications cluded in the product costs and not as limited number of returned question- communication costs. The traffic cost 6 Telelearning naires, there are significant uncertainties can be paid by the companies through an (Symmetric and asymmetric) related to the results. The questionnaire 800 number. consisted of 17 questions, very similar to Video-on-demand the questions included in the Delphi sur- A households willingness to pay is Videophone vey carried out in the OPTIMUM work- related to communication costs, ie. sub- shop on techno-economics at the Univer- Virtual reality scription costs and traffic costs. Costs sity of Aveiro in October 1997 [1]. The related to the terminal equipment or the 7 Telecommunity list of relevant application groups is iden- information content are not included in (Symmetric and asymmetric) tical in the two surveys. In that way, it is the questionnaire. possible to compare the results from the Telesurveillance surveys. Videophone Telediagnostics. Telektronikk 2/3.1999 45

47 Table 1 Leading groups of applications The survey results on expected broad- band communication per day for a house- Sum Score hold are shown in Figure 1. The broad- band connection has a downstream Tele-entertainment 15 29 % capacity of 2 Mbit/s and an upstream capacity of 364 kbit/s. Necessary termi- Information services 13 25 % nal equipment for multimedia applica- Teleworking 9 18 % tions is supposed to be available. Telelearning 6 12 % The figure illustrates how the usage per Teleshopping 5 10 % day is expected to be two hours if the price per hour is five NOK and one hour Private communication services 2 4% if the price is 15 NOK. The usage time is Others 1 2% reduced significantly, to 0.1 hour, if the price per hour is 40 NOK. We can see an equal trend from the results achieved in the 1997 Delphi survey. Assuming that the number of active usage days per year is 300, we get usage costs or traffic costs 5 Demand driving 6 Expected use of broad- per year for the two surveys as shown in Table 2. applications band applications We can see from the table that expected The experts were asked to point out the All households have a time budget and traffic costs for an average Norwegian three leading groups of broadband appli- an economic budget which are dependent household will not exceed 4,500 NOK. cations in the year 2010. Table 1 shows on the number of persons in the house- This means that a household is not will- that tele-entertainment, information ser- hold and the households disposable ing to take on higher annual costs for vices and teleworking got the highest income. Today, when the broadband broadband services than they pay today scores and are supposed to be the highest applications are not available, the time for narrowband services. Based on the broadband demand driving application budget and also the economic budget are European survey we can estimate annual group. spent on other services. When the broad- traffic costs for a household up to 4,000 band applications are introduced, substi- NOK. The subscription costs are not In the European survey [1] the same tution effects between the traditional ser- included in the above estimates. three demand driving groups of applica- vices and the new broadband services are tions were indicated. expected. In the near future the tradi- tional services are expected to dominate. 7 The broadband The objective with the Delphi question- naire is to estimate to what extent a forecasts household will use the new broadband For each of the broadband connections applications as a function of additional 2 Mbit/s, 8 Mbit/s and 25 Mbit/s the payment, taking into account the poten- experts were asked to estimate the broad- tial substitution effects. band demand forecasts for the years 2000, 2005, 2010, as well as the satura- tion level. The results from the survey show that the experts are much more 2.5 optimistic than the experts in the 1997 Delphi survey [1]. There may be differ- ent causes for the deviations. One reason 2 may be high non-response and few received questionnaires; another reason 1.5 may be a too homogeneous group of experts in the Norwegian survey, mainly Hours consisting of people related to research 1 on multimedia and broadband evolution. On the other hand, it seems reasonable that the results from the Norwegian sur- 0.5 vey should give more optimistic results compared to the results on a general European level, since Norway has a 0 rather advanced infrastructure and high 5 15 40 80 150 demand for new services like ISDN and Price per hour (NOK) GSM. Figure 1 Expected broadband communication per day 46 Telektronikk 2/3.1999

48 8 Demand curves for Table 2 Estimated traffic costs per year in the Norwegian Delphi survey compared to the European survey access capacity Different access capacities may be used Price/hour Price/hour Hours Active user Traffic costs for offering the described applications. ECU (NOK) days per year (NOK) For a set of given annual costs for 2 Mbit/s, 8 Mbit/s and 25 Mbit/s services 5 2 300 3000 the experts were asked to estimate the 15 1 300 4500 fraction of the residential market ex- pected to ask for the service. The ques- 40 0.1 300 1200 tion of willingness to pay was related to the year 2010. The results indicate that within a period of ten to 15 years it is 0.5 4 2 300 2400 expected that a substantial part of the subscribers will have a broadband con- 2 16 0.75 300 3600 nection which in turn means that the sub- 5 40 0.33 300 4000 scribers have many alternatives for com- munication. This is called a high exter- nality effect. The figure shows that the expected demand is quite similar for the three access capacities. The experts believe Demand as a function of annual charges that the households are not willing to pay 45 much more for higher capacity and better 40 quality. These results support the opinion 2 Mbit/s, 384 kbit/s of the experts in the international Delphi 35 8 Mbit/s, 384 kbit/s Percentage survey. An interpretation of the results is 30 25 Mbit/s, 384 kbit/s that the households de facto are unwill- 25 ing to pay more for a higher capacity. 20 Another interpretation is that the experts 15 today have difficulties understanding 10 what an increased bandwidth really means for broadband communications. 5 Today, a 2 Mbit/s connection is a rather 0 high capacity. The demand curve for 2,000 7,000 12,000 17,000 22,000 access capacity in the European survey is Annual charges (NOK) parallel with the expected Norwegian demand curve, but at a lower level. At an Figure 2 The broadband demand in percentage of the residential market annual cost of 4,000 NOK the Norwegian as a function of annual charges experts expect a demand of 25 %, while the European experts estimate 15 %. With annual costs of 8,000 NOK the numbers are 10 % and 5 %. 10,000 9 Disposable income and willingness to pay 8,000 Annual charges Figure 3 shows the expected willingness 2 Mbit/s, 384 kbit/s to pay for broadband connections as a 6,000 8 Mbit/s, 384 kbit/s function of a households disposable 25 Mbit/s, 384 kbit/s income. The results indicate that the 4,000 household only to a limited degree is willing to pay more for higher capacity. 2,000 In the disposable household income bracket between 200,000 NOK and 600,000 NOK the households are willing 0 to use about 1 % of annual expenses on 0 200,000 400,000 600,000 800,000 1,000,000 broadband communication. With dispos- Disposable income (NOK) able income above 600,000 NOK the willingness to pay as percentage of dis- Figure 3 Willingness to pay for broadband communication posable income decreases slightly. In the as a function of disposable household income Telektronikk 2/3.1999 47

49 Table 3 Willingness to pay in percentage related to ISDN costs as a function of increased bandwidth to download and store large volumes of data like videos, a new demand for ex- Capacity 128 kbit/s 2 Mbit/s 8 Mbit/s 25 Mbit/s 50 Mbit/s 500 Mbit/s change of large volumes of data is likely to be created. The evolution of broad- Index annual charges 100 135 150 160 167 175 band terminals ie. PCs will support this development. Figure 4 shows the expected growth of symmetric 2 Mbit/s accesses and 8 Mbit/s accesses, according to the sur- European Delphi survey the experts esti- household in Norway has about 5,000 vey results. In 2005 the share of the sym- mated that the households were willing NOK in annual telephone expenditure. metric access capacity is estimated to be to pay between 1 % and 2 % of their The table indicates that the households 5 % for 2 Mbit/s and 2 % for 8 Mbit/s. In yearly disposable income. are unwilling to pay more than 9,000 the long run it is assumed that the satura- NOK for an extremely high bandwidth tion level of symmetric broadband access (ie. 500 Mbit/s). These results are quite 10 Price versus comparable to the results achieved in the will be 30 % for 2 Mbit/s and slightly lower for 8 Mbit/s. The uncertainty in the broadband capacity European survey [1]. answers increases significantly with the prediction time. The difference between The experts were asked to estimate a price for different access capacities rela- 11 Symmetric the 25 % quartile and the 75 % quartile increases from 3 % in 2005 to 26 % in tively to ISDN as a reference access. The broadband access 2015. The experts assume that from 2005 yearly costs for the reference access were to 2015 between 25 % and 35 % of those set to 100. The experts should add an In the coming years the broadband com- who demand 2 Mbit/s access will additional percentage in price for in- munication is assumed to be highly influ- demand symmetric capacity. creased capacity. enced by the Internet. The communica- tion generated is rather asymmetric since The table confirms that the households the customers mainly communicate with 12 Mobile are unwilling to pay much more for very some Internet data sources. When the high access capacities. Today an average subscribers get terminals which are able broadband access The respondents were asked to estimate in percentage how much more the sub- scribers are willing to pay for a mobile 35 broadband access rather than a fixed con- nection. The question was repeated for 30 the years 2000, 2005, 2010, 2015, and for the saturation year. Table 4 shows that 25 2 Mbit/s the subscribers are not willing to pay 8 Mbit/s more than 30 % in addition to the fixed Percentage 20 broadband access price to get a mobile broadband access. 15 According to the survey results the expected demand for a 2 Mbit/s mobile 10 broadband access increases from 8 % of the households in 2005 to 20 % in 2010. 5 At the saturation point it is expected that 30 % of the residents will demand a 0 2000 2005 2010 2015 Saturation mobile access. Year 13 Access technology Figure 4 Expected demand of 2 Mbit/s symmetric access and 8 Mbit/s symmetric access The experts were asked to give a percent- age distribution of the different access technologies envisaged to be the most relevant ones in the future. The relevant access technologies listed in the ques- Table 4 Increased willingness to pay for mobile broadband access relative to fixed tionnaire were: broadband access (in percentage) Coaxial cable modems; Year 2000 2005 2010 2015 Saturation Copper line modems (digital sub- scriber line, DSL); Percentage 30 30 30 20 30 Fibre (fibre-to-the-home, FTTH); 48 Telektronikk 2/3.1999

50 Broadband radio access systems; 35 Satellite systems; 33 75 quartile 30 30 Power line modem. 50 quartile 25 quartile 25 25 Percentage The results are presented in Table 5. The experts expect that the two dominating technologies will be coaxial cable 20 20 20 modems and copper line modems (DSL), assuming that each technology will cover 15 16 15 35 % of the households by the year 2010. 13 The respondents also estimate that the 10 10 10 power line modem technology will cover 8 about 10 % of the broadband subscribers. 5 5 3 14 Conclusions 0 1 The Norwegian Delphi survey on broad- 2000 2005 2010 2015 Saturation band demand shows that there will be a Year significant demand for broadband ser- Figure 5 Expected demand for a 2 Mbit/s mobile access vices in the residential market during the next ten years. Tele-entertainment, infor- mation services and teleworking are the applications expected to be driving the demand. cable modems will be the preferred Reference The households are unwilling to pay access technologies. much more for higher connection capac- 1 Stordahl, K, Rand, L. Long term ity. They expect the technical evolution In comparing the results from the Norwe- forecast for broadband demand. to reduce prices significantly and enable gian Delphi survey with the results from Telektronikk, 95 (2/3), 3444, 1999. the operators to offer better quality and the European survey [1], we find very (This issue.) higher capacity at reasonable prices. Fur- similar results, except for the estimated thermore, the households are only willing broadband forecasts, in which the Nor- to pay marginally more for a mobile wegian experts expect a much higher broadband access connection than for a demand than the European respondents. fixed broadband access connection with similar capacity. It is expected that copper line modems (DSL) and, more surprisingly, coaxial Synnve Istad received her B.A. with Honours from the University of Strathclyde in Glasgow in 1991, specialising in economics. She worked as advisor in The Norwegian Table 5 Expected coverage (in percent- Competition Authority from 1991 to 1997, and has since age) of access technology to Norwegian then been working in Telenor Nett as market analyst. households in year 2010 email: [email protected] Technology Percentage Coaxial cable 35 % modems Copper line modems 35 % (digital subscriber Kjell Stordahl (54) received his M.Sc. degree in statistics at line, DSL) the University of Oslo in 1972. He worked with Telenor Re- Fibre (fibre to the 5% search Dept. for 15 years, seven of which as manager for the teletraffic field. He joined Telenor Nett in 1989 and was home, FTTH) manager in the Planning Department until 1996. From 1997 Broadband radio 5% to 1999 he was manager for market analysis in the Market access systems Division, Telenor Nett. He has participated in various Euro- pean projects and authored or co-authored more than 90 Satellite systems 10 % papers in international journals and conferences. Power line modem 10 % email: [email protected] Telektronikk 2/3.1999 49

51 Regulation of broadband access networks PER MOGNES AND TERJE NORD Introduction context of dimensions we believe are the tering competition and compatibility in a fundamental ones. For instance, will pub- complex network industry like telecom- This article deals with future regulation lic policy decisions regarding (the weight munications is of paramount importance of broadband access networks in general. given to) the generic issues described in in future regulation, and we also devote Even though this is an issue of immense points 13 in the panel below be influen- some space to this issue (point 2). The importance to the strategic planning of tial to the further direction of the regula- last dimension mentioned (point 3) is not new and old industry players, it is safe to tory regime, and also have impact on the directly discussed in the article, but is say that it is not an issue in the present actual market development? inherent in present regulation of the regulatory framework applying to the industry, and the outcome regarding telecommunications sector. Bluntly Before getting into the specifics we points 1 and 2 will decide the focus of stated, the only regulatory discourse should make it clear that we have de- this issue. regarding broadband access networks has limited the subject of regulation of been within the context of whether broadband access networks to imply reg- The article is organised in four sections. mandatory access to the dominant play- ulation of public networks and public First we give the historical background ers copper lines is within the scope of services only.1) Furthermore, we keep for regulating the telecommunications the present interconnection regime. A possible political decisions and public industry and a short description of the natural point of reference for an article subsidisation of broadband access net- regulatory development towards its such as this one is thus the present status work investments outside the scope of present state. In section two we give an of policies regarding interconnection and the article. We have kept the focus to the overview of anticipated regulatory pro- access and the most likely developments European setting, and thus given the reg- cesses relevant to the formation of a in that regime. We will, however, also ulatory processes within the European future regulatory regime applying to discuss other regulatory developments Union much weight. In this setting a broadband access networks. The last two relevant for possible future regulation of more precise definition of broadband sections represent the discussion part of broadband access networks. access networks is not necessary (eg. the article. Section three deals with the transmission rates) since the EU legisla- issue of securing access to bottleneck It is not possible to accurately predict the tion does not define broadband networks. network facilities, the question of when outcome of the regulatory process in the and why something becomes or ceases to years to come neither by extrapolation of We devote considerable effort to the be a bottleneck facility, and discusses the the existing regulatory regime by some understanding of the balance between most likely regulatory solutions regard- form of scenario building, nor by analy- sector specific regulation and general ing broadband access networks as we see sis of the balance of power between dif- competition policy (point 1), because it it. In the final section, we take a closer ferent institutions and ideologies. Fur- will denote the regulatory toolbox avail- look at a few other critical issues regard- thermore, it is not possible to isolate a able to regulators as broadband access ing future regulation of broadband access regulatory discussion to the issue of networks become a reality. We also networks. broadband access networks alone. We believe that finding optimal access rules thus readily accept that a description of (who should be given access at what a future regulatory regime that may seem price) that combine the objectives of fos- Regulation of tele- plausible at present may be next to communications in a worthless tomorrow as market and tech- nology may fundamentally alter the retrospective glance underlying assumptions. We have there- 1) We also refer to broadband access net- For decades telecommunications (net- fore chosen to take a broad approach to works possible of transferring two-way works and services) have been provided the matter, and will discuss it within the simultaneous services only. by a secure monopolist. Until the late 1980s public enterprises holding a legal monopoly was the common rule in most countries around the world.2) The ab- sence of competition was motivated by the existence of large fixed costs in sev- 1. The scope and extent of sector specific regulation eral parts of the network, whose duplica- Market Regulation tion was neither privately profitable nor socially desirable. The telecommunica- Competition policy Sector specific regulation tions industry was deemed to be a natural monopoly and the services public utilities 2. Fostering competition or compatibility a period often referred to as the good old days by incumbent telecommunica- Competition Compatibility tion operators around the world. Access to bottleneck facilities Standardisation, services regulation, seamless networks 3. Technology dependent or technology neutral regulation Vertical focus Horizontal focus 2) In the USA AT&T was a private, regu- Traditional approach Convergence approach lated corporation. 50 Telektronikk 2/3.1999

52 Over the ten year period 198897, the Table 1 Overview of key developments in EC Telecoms Policy European Community enacted an exten- sive package of telecommunications leg- islation designed to enable Europe Period Key measures Relevant milestones to respond to the challenges of rapidly evolving and converging technologies First phase: ( 1987) Standardisation BT case and the globalisation of the information First Community measures Public procurement economy. A few factors contributed to Competition rules do apply the reform movement.3) The growing awareness of the inefficiencies of the Second phase: (1987 1992) Liberalisation of value Green Paper on incumbent monopolists (poor incentives Initial market opening added services and telecommunications to reduce costs and a severely distorted terminals markets price levels and structures), and the tech- Creation of an Open Terminals directive / nological change made it evident that Network Provision Services directive some segments could be served equally framework well, and sometimes better by new players than by the incumbent telephone ONP Framework / operator. Furthermore, the impact of the Leased lines developments in the United States, in directive particular the AT&T divestiture consent decree and the resulting transformation Third phase: (1992 1998) Co-ordinated Review (1992/93) / of the US market began to be felt in Full liberalisation liberalisation Infrastructure Green Europe. At the same time the progressive Paper (1994/1995) deregulation of the telecommunications sector and the privatisation of British Interconnection regime Extension of the Telecom in the UK since 1982 made Services directive Europe more receptive to the concept (satellite, cable, of market deregulation.4) In addition, the mobile, full comp.) European Court of Justice confirmed in Competition rules Reform of ONP the British Telecommunications case that (interconnection, EU competition rules applied to the licensing, USO) telecommunications sector [3]. Fourth phase: (1998 ) Process of defining new 1999-Review Table 1 gives an overview of the main Beyond full liberalisation regulatory principles and developments in EC Telecoms policy options and its provisions. Convergence Green Paper on As the informed reader will know the convergence telecommunications industry is still a heavily regulated industry. The focus European Case-law of regulatory attention has however changed over the period. While in the first phase (1987) one focused on issues like standardisation activities, public procurement procedures and the implica- tions of above referred to decision in the framework), the second phase (1987 from (92/93) where member states BT case (often called the legal corner- 1992) was about progressive market agreed upon the decision to liberalise the stone of the EU telecommunications opening and issues like market by 1 January 1998 (including the remaining public voice telephony and Full liberalisation of markets for value telecommunications network infrastruc- 3) The so-called reform movement has added services in order to introduce ture monopolies). This time lag was competition [4]; implemented at least two separate meant to give the incumbent companies processes all around the world. First, The separation of regulation and oper- time to prepare for the forthcoming com- incumbent operators are being privat- ations, a pre-requisite for the develop- petition. Furthermore, there was agree- ised and are provided with better in- ment of an open market, and open, ment to adjust the ONP framework to centives to minimise cost, as well as transparent and non-discriminatory fully liberalised market conditions and more flexibility to rebalance rates in regulatory decision making; to establish a regulatory framework for conformance with business and eco- interconnection and access to services The definition of a harmonised set of nomic principles. Second, markets and networks. This implied an asymmet- access conditions, best known as Open have been deregulated or at least ric regulatory regime, where former Network Provision (ONP) [5]. legally liberalised. monopolists are subject to more stringent 4) For good surveys of the UK reforms, regulatory requirements (mandatory pro- Phase 3 (1992 98) full liberalisation visions, non-discrimination, cost orienta- see reference [1] and [2]. was initiated by the EU Telecom Review Telektronikk 2/3.1999 51

53 tion, price regulations, etc.) than do other bottleneck holders are needed, the Networks are multiplied and are also operators. The scope and extent of the stronger the argument for an approach multipurpose information infrastruc- requirements do however differ substan- based on general competition law prin- tures, and due to network externali- tially across nations. ciples will be. From this line of argu- ties6), must be interconnected. Gener- ment, the development path should in ally speaking, firms with large existing With full liberalisation, and the emerging the long run be towards a competitive networks tend to be against compati- sector specific EU framework, the defini- regime, which is based on competitive bility, even when welfare increases tion of access and interconnection within access markets, and the application of with compatibility. Securing of inter- the ONP framework acquired more and general competition rules. connection then becomes both a tech- more importance. This was refined par- nological compatibility and a regula- ticularly with the adoption of the ONP In the meantime striking the right bal- tory harmonisation issue. Interconnection directive in 1997 [6]. ance between sector specific regulation, Large fixed cost is an inherent charac- At the same time, recent developments competition rules and structural solutions teristic of the industry, and some seg- in EU Competition Law made it easy to will be the regulatory challenge. Indus- ments may even be natural monopo- relate access to bottleneck facilities in try, on the other hand may find this regu- lies. These segments become bottle- telecommunications more explicitly to latory framework too complex and uncer- necks to which other operators must the essential facilities concept [7]. tain. have access in order to compete. Inter- connection policies must be designed The European Unions experience of reg- ulation for securing access to so-called Regulation in the making so as to allow efficient entrants to come in and keep out inefficient ones. network bottleneck facilities in the The former and the present regulatory The price signals must be the right telecommunications sector is however framework in telecommunications still ones (make or buy decision to the still in its early stage. It is at the moment very much have the properties of entrants) and give the bottleneck shaped by a three pillar approach, based monopoly regulation as introducing com- owner a reasonable compensation so on the interplay of hands-on sector spe- petition has been and still is the regula- that they have incentives to build and cific ex ante regulation of access, an ex tors major objective. That is, primary maintain the bottleneck and not to post use of the competition rules, and, focus on regulation of the bottleneck exclude their rivals from access to the to some extent, the search for structural facilities of the dominant operator and bottleneck. Lastly, interconnection solutions aimed at the development of the securing of one-way access of input prices must induce an efficient use of competitive access markets, ie. cable-TV for entrants. Furthermore, incentive regu- the network. network investments in merger notifica- lation of the incumbents is sustained. By tion processes. The location of the bottlenecks incentive regulation we understand the changes with the evolution of technol- regulatory schemes offered to the incum- In a stable environment (as the traditional ogy, and furthermore (as competition bent operators securing performance- telephone market is often assumed to be) in the sector increases and substitute based returns and more freedom to set the cost associated with direct regulatory facilities evolve) bottlenecks are rarely rates in accordance with standard busi- intervention may be minor compared to pure bottlenecks, but most ness practices. The economic discussion the benefits of assuring efficient and often incomplete bottlenecks. is then how incentive regulation must open access and interconnection to the trade off cost efficiency and the limita- incumbent bottleneck provider. However, We will return to these features in the tion of operators rents, and how service the more the situation is one of rapidly discussion part of the paper. The rest of pricing should be structured in order to changing markets, the limitations and this section is devoted to giving a brief attain economic efficiency, etc. The costs of sector specific approach becomes overview of the regulatory processes advent of competition should reduce the more apparent.5) As markets are con- anticipated by the EU Commission that attention paid to the incentives and raises verging, and the more rapidly innovation may have relevance for the future regula- a set of new issues, such as efficient proceeds and new investments by the tion of broadband access networks. The competitors, the co-ordination of invest- following tables give a summary descrip- ments in facilities and new technology tion of the key provisions in the EU 1999 between operators, the duplication of Regulatory Review, the EU regulatory networks, and so on. approach to convergence, electronic 5) Sector specific regulation, particularly commerce and next generation mobile From an economic point of view some with regard to price regulation, is a systems (UMTS). important general features of the deep intervention in market mecha- telecommunications industry must be nisms, with a high risk and responsi- taken into account: bility for the regulator. It becomes highly dependent on definitions, which implies a high degree of technicality, 7) Separating the concepts of unbundling and therefore has a great potential for and interconnection is important for legal conflict. And it inevitably leads to 6) Externalities imply that one firm or designing regulation, because, for the substantial intervention in the day-to- customer affects others without com- purpose of fostering an efficient com- day business practices and strategies pensation being paid. Other things petitive environment, the process of of the bottleneck holder, with the dan- being equal it is better for customers setting guidelines for unbundling and ger of heavy handed regulatory to be connected to a large network interconnection is driven by different approach. than a smaller one. economic considerations. 52 Telektronikk 2/3.1999

54 Table 2 EU 1999 Review issues and schedule Focus and time schedule Issues relevant for broadband access regulation 19992000: Generic issues: Analysis, investigation, consultation and policy formation Sector specific regulation vs. competition rules Market convergence horizontal, technology neutral regulation of access European harmonisation vs. national subsidiarity Specific issues FMC End-user access (unbundling requirements) Facilities vs. services competition Regulation of scarce resources (frequencies, numbers, domain names) USO Licensing Internet (telephony, Internet access conditions) 20002001: EU Commission proposals for new legislation submitted to the European Parliament Formation of a revised EU legislation and the Council. Consolidation and simplification of EC rules Extension of infrastructure regulation Introduction of Sunset Provisions (efficient competition test for removal of some provisions of sector specific regulation) Institutional set-up at EU level, relationship between different regulatory levels 20012006: A political process that may be very time consuming, but much effort will be put into Adoption of new legislation by EP the harmonisation of European regulation, and the idea of one-stop shopping and Council, and implementation into Member States national laws Access to bottleneck both in the context of network externali- stream telephony market, may effectively ties (compatibility) and the development squeeze an operator with a small market facilities of competition. The present interconnec- share. Thus, interconnection services are tion services are related to the public mandatory to provide at a reasonable As mentioned in the introduction a switched telephony service (call origina- price and quality (soon including carrier thorough analysis of the scope and extent tion, call termination and transit as speci- pre-selection and number portability), in of the present interconnection regime and fied services). From an economic point order to avoid practices of foreclosure. its likely development is probably the of view the physical access to the cus- best way to say something meaningful tomer is always a source of market power Regulatory requirements regarding inter- about regulation of broadband access net- (mobile operators for instance control the connection services other than those works. Regulators can greatly affect the only access to their mobile customers). enabling a seamless service in a competi- nature and development of competition Call termination is thus an input that tive telephony market (fixed and mobile), through their choices of scope and extent could effectively be used for restriction will depend upon market development. of the ex ante requirements of intercon- of competition in the retail market. Addi- For instance, what kind of services the nection and unbundling in the sector spe- tionally, the operator has incentives for incumbent operator offers his own cus- cific regulation.7) price discrimination between on-net and tomers and what services competitors off-net calls. If one operator dominates would like to realise for their customer We define interconnection as an agree- the market entirely by its size (market base at that point in time. In a competi- ment that gives access to competing share), refusal to supply the input or tive market termination of circuit operators customer base and vice versa. unreasonable pricing of it combined switched broadband services in each Interconnection is an important issue with price discrimination in the down- others networks could become manda- Telektronikk 2/3.1999 53

55 Table 3 Convergence, electronic commerce and UMTS Regulatory process Issues relevant for regulation of broadband access networks Convergence Regulation of access to networks and digital gateways Policy formation regarding a regulatory framework Technology neutral regulation (horizontal approach) for convergence Adaption of the existing framework Electronic commerce Internet access conditions Setting the principles and removal of regulatory uncertainty Removal of regulatory risk may help boost supply of services requiring broadband capabilities UMTS Part of a global standard capable of providing broadband services over a wireless platform Standardisation activities, securing sufficient allocation of frequencies, co-ordinated introduction, criteria/principles Licensing and regulatory framework in place ahead of market. May for regulation [8] set a standard for regulation of broadband access networks tory if commercial negotiations fail and The option of using the ONP framework The appropriate level of unbundling is a such failure would seriously impede the to secure broadband service offerings has function of the network segments degree value of the service to end-users. If the been discussed by the EU Commission in of being an essential facility. Unbundling market situation on the other hand still is the past [9]. Recent revisions for adapta- of essential facilities is a means to pre- one of total dominance for the incumbent tion of the directives to a competitive vent an incumbent from foreclosing effi- operator, regulatory requirements regard- environment [10] however show that reg- cient entry. Forced unbundling of non- ing the whole range of interconnection ulatory obligations will be restricted to essential facilities, however, overrides services, and restrictions regarding the the provisioning of additional types the role of competition in allocating net- pricing of such services, could probably of leased lines, taking market demand work investments and can needlessly be a plausible requirement. We do not and progress with standardisation into stifle innovation. What essential facilities however find this a very likely scenario. account. A thorough assessment of the are, though, depend on the characteristics need for continuation of the directives of the network, the availability of com- The issue of bringing the logic behind or the need for further measures will be petitive services and what is perceived as regulation of interconnection into the done within the 1999 Review process. the evolution of technology. Indeed what Internet domain is still premature. Only today are believed quite generally to be market development and the future Requiring the unbundling of network essential network facilities may not be industry structure can give any answers elements so that the competitor can pur- regarded essential once other network regarding the need for such public inter- chase piece parts as needed to supple- facilities represent viable alternatives. vention. However, the links between ment or replace its own network, would different Internet Service Providers net- in addition allow for competitors to lease works (ISPs) bring about concerns simi- access lines, thus avoiding costly dupli- Access to the local loop lar to the ones in traditional telecommu- cation of facilities in that network seg- The case of local loop unbundling nication. In a new and dynamic market ment. In this way, competitors can build (LLUB) is thus based on the benefit general competition rules would nor- their own networks much more quickly of stimulating local competition (in the mally be regarded a better tool than than they could by providing their own telephony service and in high speed data direct regulation. loops. The regulatory risk is as Pablo access),8) making the assumption that the Spiller [11] puts it, that regulators local loop is at the moment an essential Resale of an incumbents service is a underestimate the costs associated with facility. In a newly released study, means for new entrants to enter a market, extensive unbundling. Indeed, extensive Ovum, defines LLUB as an ... intercon- build a customer base, and compete with unbundling and the resale requirement nect service provided at a point between the incumbent at the retail level. It pro- may provide a disincentive for entrants vides entrants with the quickest form of to build competing facilities, thereby im- entry, since it requires no up-front capital peding the facilities based competition investment. An obligation to provide regulators hoped to achieve in the long such access to all services the incumbent run. Finally, coercing the incumbent to 8) Other less robust arguments that are itself provides at retail level is not likely, share its technological innovations can used in favour of LLUB is 1) avoiding unless infrastructure competition is destroy or unfairly award to competitors access network duplication, 2) full totally absent. Setting the right price for the returns from research and develop- competition rather than oligopoly, such wholesale services is a complex ment through competitors use of the 3) strengthening of existing competi- issue. incumbents network. tors to the incumbent. 54 Telektronikk 2/3.1999

56 the customer premises and the line-side 1999 Review process (see table above). have the capabilities to develop multi- of the access network operators local The Norwegian regulatory authority has, functional broadband access. The public switch. LLUB thus gives an entrant dedi- for example, reached the decision [13] telephone networks of the incumbents, cated access to the customer via the local that now is not the right time for such an and the many cable-TV networks (with a loop. [12]. obligation.11) Telenor, on the other hand, total penetration of 30 % in the European has during the process decided to Union). Two variations of LLUB are being dis- develop direct access products with cer- cussed: tain usage restrictions.12) If a regulator is Cross-ownership of the cable-TV net- requiring bitstream access, the regulator works by the incumbent telephone opera- 1 Direct access to the transmission will need to determine the particular ser- tor may constitute a barrier to the invest- medium in the local loop. This would vices which should be provided. Agree- ment incentives to upgrade these net- imply copper loop9) rental with the ing upon the set of designated bitstream works with full multi-functional access point of interconnect being at the dis- services and the conditions attached (the capabilities. In the Cable-TV Review tribution frame which marks the end issues of (i) who may require access, and [14], the EU Commission discuss the of the copper loop, either at the local (ii) at what price), is not trivial. The Nor- investment incentives given to the local switch or in a remote concentrator unit. wegian regulatory authority says it will bottleneck provider under different regu- This allows the entrant to operate its continuously evaluate the level of com- latory conditions like: 1) extension of the own transmission system to provide petition in the Norwegian market, and the ONP regime to cover cable-TV net- the customer access service. regulatory and market experiences with works, 2) legal separation, to establish a 2 Bitstream access, which compromises LLUB from other countries. Regulatory minimum separate development base for both the transmission medium and the intervention in the access network is both networks, and 3) full scale divesti- transmission system. When requiring therefore not excluded as an alternative ture of the cable network by the incum- bitstream access, regulators will need neither in the European framework nor in bent telephone operator, to establish a to determine the particular intercon- the Norwegian regulatory regime. business case for both networks to de- nect service that should be provided. velop full future capabilities. In fact, the We do however assume that the probabil- EU Commission chose option 2 as a min- Ovum recommends the regulators to ity of such an ex ante regulation of the imal solution for the European Union as require some form of local loop un- bottleneck holder decreases over time. a whole. They went on to state that in bundling to be offered by the incumbent Regulators may find that regulation and certain circumstances it might be that the operators.10) The EU Commission has the efficiency of access is the right only means which would allow the cre- made it clear that LLUB falls outside approach in a stable environment (as ation of a competitive environment con- the requirements of the interconnection when the voice telephony market is the sists in the divestment of the cable tele- directive in its present form, but leaves focus). However, the more convergence vision network .... How far a divestiture it open to national authorities to incorpo- of markets becomes the dominant fea- of bottleneck facilities could be enforced rate such requirements in their national ture, and the more rapidly innovation under EU competition rules will be an legislation. The issue of access to the proceeds and innovative investments by issue for future case law. However, the access network (in effect LLUB as a use- the bottleneck holders are needed, the EU Commission has made it crystal clear ful tool to jump-start competition) will more an approach based on general com- that: 1) sector specific regulation is not also probably be further discussed in the petition law principles should become the right means to create competitive preferred. access markets, 2) companies that enjoy a dominant position in two markets must take particular care not to allow their Ownership of cable-TV conduct to impair genuine undistorted 9) It is not at the moment considered networks competition (Art 86), and 3) an extension technically feasible to offer direct As mentioned earlier the development of of an operator dominant in both telecom- access to the transmission medium in viable competitors in the access market munications and cable-TV networks into the case of fibre loops, as capacity is building their own access networks either related fields could raise serious competi- then shared between many customers based on fixed line technology or alterna- tive concerns (merger-regulation). which makes separating a physical tive access technologies (substitutes), path for unbundling impossible. may fundamentally alter the essentiality 10) Ovums stand in this matter is actu- of the incumbents copper lines. Such Regulating mobile ally considerably revised from earlier developments must carefully be taken communication services reports. Confront Ovums report for into consideration when making the regu- In the mobile communications field, next the Norwegian regulatory authority latory decision concerning the need for generation technology does also promise (Pt) from 1997 where they unam- public intervention. On the other hand, next to broadband capabilities. Such biguously recommend copper rental. the regulatory framework may also great- broadband capabilities mixed with 11) This conclusion is confirmed by a ly influence the development of new mobility could over time make the markets (for instance the incumbents mobile operators a serious competitor to Parliamentary decision in spring 1999. incentives to develop new ways of access the existing potential broadband access 12) Usage restrictions on the transmis- or the incumbents and the cable-TV networks. In the EU Councils common sion medium, that is restrictions on the operators incentives to make the neces- position paper [8], UMTS is described to rental of dedicated subscriber lines sary infrastructure investments). At pre- be a system for mobile multimedia: A related to standard termination of sent there are two mass distribution sys- third generation of mobile communica- either PSTN, ISDN or ADSL. tems available in the local loop, which tions system capable of providing, in par- Telektronikk 2/3.1999 55

57 ticular, innovative wireless multimedia The process of selecting eligible candi- Other critical issues services, beyond the capability of current dates must be open and non-discrimi- second generation systems such as GSM, natory. The mechanism for picking Negotiation versus direct regu- and capable of combining the use of ter- winners (auction, beauty contest or lation of access agreements restrial and satellite components. If the other procedures) and the likely crite- Negotiating interconnection is a complex UMTS vision is to come through, the ria and conditions attached to a licence undertaking, and as such it is extremely regulatory framework must secure the must be decided and communicated to difficult to attempt to set all economic availability of frequency spectrum, a stakeholders prior to the licensing pro- and technical conditions via direct regu- scarce resource, as well as providing for cess. lation. On the other hand, commercial an enabling regulatory environment. It Efficient utilisation of spectrum negotiations have shown to produce sub- seems to us that much of the same regu- resources implies finding the right bal- stantial delays, particularly when there latory logic as the one described above ance between giving each operator is no clear framework for resolving dis- for fostering investments from cable-TV enough spectrum to enable sufficient putes. Most countries have therefore operators will be used in this context. service capabilities and the need to chosen to let charges be a result of com- Thus, general competition law should be have enough operators within the mercial negotiations between the parties, used as the regulatory framework for available spectrum to maintain and with the regulator looking over their UMTS. Mobile communications has, promote a fully competitive market. shoulder accepting the terms in the however, some characteristics that will According to the Licensing directive, agreement. A key ingredient of a clear complicate part of the picture. We will any limitation to the number of opera- framework for resolving interconnection return to the specifics of mobile regula- tors should only be made on the basis disputes is thus a cost methodology to tion, but first also mention the impor- of the scarcity of the spectrum. A har- apply should negotiations fail. So far, tance of the standardisation process, monised European approach regarding most countries have chosen to apply a which will remain a key factor in pro- finding available frequencies is already cost based methodology. The rationale viding quality services at an affordable proposed. for such an approach is straightforward: cost and enable roaming and interwork- cost oriented rates attempt to replicate ing between systems. The flexibility of The requirements in the license or competitive conditions, and as such pro- interfaces and the capacity to evolve in other regulation concerning 1) network vide the right signals to entrants and parallel with technology is as crucial for quality and service capabilities, 2) roll- incumbents in terms of investment and UMTS as it has been for GSM. However, out and coverage, and 3) provisioning network expansion. There is less agree- such standardisation has an inherent ten- of roaming, facility sharing, etc. ment, though, on how to measure costs dency to become industrial policy, and Private networks and other use of and which cost should be measured. what we have observed is a tug of war uncoordinated spectrum. between different geographical regions. With true competition market forces will drive prices to their efficient levels. Policy is an integrated part of regulation, Alternative broadband access Thus, in the case of true competition the and especially the relationship between networks regulator should allow the parties to frequency administration, the choice of In addition to the traditional access net- negotiate prices freely. In the transition licensing techniques and the attached work systems (copper-, coax-, radio- process toward this competitive market licensing conditions may be problematic based systems), adoption of novel access regulators want to prevent anti-competi- as different public policy aspects have technologies like broadband radio (for tive behaviour of the incumbent (price conflicting interests in due matter. Har- instance broadband DECT and LMDS) well above cost to discourage competi- monisation is not compatible with com- or powerline modems running on the tion or collect the majority of rents) by petition and the European solution tradi- electric transmission grid may turn local establishing a framework for pricing. The tionally differs somewhat from the infrastructure into a truly competitive process of agreeing upon the right princi- American. Especially the issue of tech- segment. When such solutions represent ples regarding charging methodology is nology competition is problematic as viable alternatives (if technology and at the moment a hot issue in the regula- Europe in order to promote pan-Euro- demand develop), the local infrastructure tory discourse (sector specific regula- pean markets and services is supporting segment will become less of a bottleneck tion). Although fully allocated historical one standard, and spectrum resources are segment. Thus, the rationale for asym- cost has been mostly used, consensus is being dedicated to that particular stan- metric regulation of the incumbent being built concerning the advantages of dard. The outcome of the different initia- provider of copper line access in order to using a forward-looking incremental tives in ETSI, 3GPP, ITU and work jump start competition, will vanish. Fur- cost. Thus, a form of Long Run Incre- within autonomous industrial co-opera- ther development of infrastructure regu- mental Cost (LRIC) methodology, tives will be decisive for strengthening lation will be a less relevant issue. On the already implemented in some countries the further development of the mobile other hand, it could make regulation of and strongly advocated by the EU Com- multimedia market. services even more relevant. In order to mission, will probably be the chosen cost secure for instance the provisioning of standard in most countries regarding The following regulatory aspects for a quality broadband service to the end- input facilities pricing. licensing UMTS13) are important: users, transmission and quality standards might be necessary. Future regulation The extent to which we must expect sec- should therefore imply more emphasis tor specific regulation to require this kind on technology neutrality to avoid giving of charging policies to apply also in the 13) See for instance [15]. biased investment incentives to the case of broadband access networks, industry players. 56 Telektronikk 2/3.1999

58 depends on the actual state of the market The need for structural separation of cer- developments, and that this is not at that time. That is, how competitive the tain network segments that occasionally counter-balanced by additional radio market is, and how far one has come in is brought into the regulatory debate, like spectrum becoming available through the the direction of applying general compe- splitting the incumbents infrastructure introduction of new and more efficient tition rules to the sector. It will also be activities (the bottleneck segments) from technologies, the consequence is that the influenced by the actual experiences of (potentially) competitive segments like scarcity of radio spectrum is increasing. different cost based pricing schemes. The service provisioning activities, should, When in addition to that, the existing cost of regulatory oversight will increase following the same line of logic, be an mechanisms may be insufficient to phase as the market develops. We would expect even less relevant regulatory instrument. out or relocate existing systems to other that the more blurred the relationship We will, however, remind the reader that parts of the radio spectrum in time (as in becomes between services and the under- this is a popular approach in other net- the case of developing countries), the lying delivery system the stronger the work industries like electricity and rail- conflicting interests become apparent at regulatory cost becomes associated with ways. Thus, one cannot totally ignore the WRCs in terms of differing frequency requiring the necessary oversight. Addi- possibility of regulatory decisions like requirements. tionally, the flaws of asymmetric regu- (although very unlikely) the Internet lation14), that is regulation that either backbone network becoming vertically The spectrum administration, consisting exclusively or differentially applies to disintegrated regulated utilities, or regu- of the process for assignment of radio one or a group of providers, will be more lators trying to encourage the develop- spectrum to individual users and the apparent once multimedia markets de- ment of an ADSL local loop that would licensing of radio communications opera- velop. Regulation designed in a world be providing services to all operators on tors, is done at the national level, subject with a one-to-one correspondence be- complementary segments. Under struc- to certain conditions agreed in the World tween delivery system and service will tural separation, the utility in general Trade Organisation (WTO) and for the no longer be appropriate as the number sells wholesale services to other firms European Economic Area, the EU Com- of services offered over different delivery who then market final services to the petition Law. The aim is to establish a systems is expanding rapidly. The need consumers. Price regulations would still regulatory level playing field within the to focus more on dynamic efficiency be necessary. WTO member countries for all users of (that is giving the low cost producers the radio spectrum which is based on open, right entry or investment incentives for objective, non-discriminatory and trans- innovation with a focus on long term Radio spectrum allocation parent ground and which supports tech- efficiency) will be given more weight in A vast array of radio communications nological innovation and competition15). the future regulatory discourse. We sup- techniques and services has become vital port the hypothesis proposing that the to the industrialised worlds economy With regard to internal market considera- greater the differences between the sunk and safety of people. Given the increas- tions in the EU, radio spectrum availabil- cost of the potential competitors, the ing dependence of society on the provi- ity affects the scope for the pan-Euro- weaker the argument for asymmetric reg- sion of information and communication pean provision and free movement of ulation [16]. If our expectations regard- by wireless means (in satellite-, broad- services and equipment. The harmonisa- ing technology- and market development casting-, mobile- and other terrestrial tion of the use of radio spectrum is there- are correct, regulation of access to broad- based radio communications systems), fore considered particularly important in band network facilities will be an issue spectrum matters are becoming critical this context to facilitate the introduction for the competition rules only. We will from an economic, political, consumer, of pan-European and global systems and then have commercial negotiations and and public welfare point of view. to realise the economies of scale neces- price setting. Prices for such facilities sary to make European industry compe- must however still be reasonable, and The planning of the usage of radio spec- titive in world markets. With regard to the only practically available benchmark trum by services at a national, regional this, standardisation and type approval will probably become other operators and global level depends on the decisions policies are central means in the overall charges for the same input product or taken at World Radio-communications EU policies. any relevant retail tariffs. Conferences (WRCs). At WRCs the 186 member countries of the International The issue that dominated the negotiations Telecommunication Union (ITU), decide at the WRCs in 1995 and 1997 concerned whether, how and under what conditions frequency availability for the provision frequency requirements for existing and planned radio communications systems 14) Two general forms of asymmetric reg- can be accommodated. WRCs therefore ulation exist. Line of business restric- result in legally binding international tions prevent a delivery system from commitments. The development and 15) This includes ensuring that choices providing a particular service (eg. negotiation of European positions for with regard to the attribution of radio fixed network operators to enter the WRCs and the voluntary adoption of har- spectrum do not privilege technical cellular market). A second, more indi- monised measures, are done by the Euro- solutions at the expense of competing rect form of asymmetric rules are pean countries in the framework of technologies, and that no blocking or those that impose on one group of CEPT (European Conference of Postal unreasonable limitation of market providers (deemed dominant) the and Telecommunications administra- access for operators from other WTO requirements to post tariffs, to provide tions). Considering that the demand for member countries are taking place evidence for cost orientation, to supply radio spectrum is increasing steadily due (ref. UMTS licensing controversy certain services, etc. to technological, market and regulatory between US and EU). Telektronikk 2/3.1999 57

59 of satellite-based broadband services band access networks alone. Predicting References which will allow for high-speed Internet the overall future regulation, we would access and video conferencing to take say that: 1 Armstrong, M. Competition in place anywhere in the world. At the Telecommunications. Oxford Review For the foreseeable future, sector spe- WRC in 1997 spectrum was opened up of Economic Policy, 13, 6482, 1997. cific rules will be required in addition for the provision of such services. The to normal competition law to ensure forthcoming WRC in 1999 will take fur- 2 Armstrong, Cowan, Vickers. Regula- universal service, interconnection and ther decisions on spectrum requirements tory reform : economic analysis and consumer protection. Concerning uni- for the mass market broadband commu- British experience. Cambridge, MIT- versal service the question needs to be nications systems, eg. satellite broadband Press, 1994. asked whether the existing universal systems (as proposed by for example service model fits the goal of deliver- Teledesic and SkyBridge) and terrestrial 3 Commission Decision No ing the benefits of the information mobile broadband systems (eg. UMTS). 82/861/EEC, Official Journal, society to all. The limited amount of frequencies avail- L41/83 (1985), (1983) 1 CMLR 487 able for the provision of satellite broad- In the medium term, sector specific (British Telecommunications). The band services and the huge capital invest- regulation will be needed in the tele- issue was presented on appeal in Italy ments required to develop such wireless communications sector to promote v. Commission, Case C-41/83 Internet infrastructures will have conse- competition in the following circum- (1985). quences for the level of global competi- stances: tion in this important area of the emerg- 4 Commission Directive of 28 June a) when dominant operators have an ing information society. 1990 on competition in the markets interest in refusing interconnection/ for telecommunications services access; (90/388/EEC; Official Journal, User access through private b) when there are network externali- L192/10, 24.07.90). networks ties; The EU regulation framework provides 5 Council Directive of 28 June 1990 on c) when a supplier controls a bottle- conditions for non-restrictions in the use the establishment of the internal mar- neck. of or access to public telecommunica- ket for telecommunications services tions networks. It may include harmon- through the implementation of open The following set of general principles ised conditions like technical interfaces, network provision (90/387/EEC; will hopefully underpin the formation of usage conditions, tariff principles and Official Journal, L192/1, 24.07.90). any future regulatory framework apply- access to numbers. The actual possible ing to broadband access networks. network segments between the public 6 Directive 97/33/EC of the European networks termination points and the user 1 Regulation shall only cover areas that Parliament and of the Council of 30 terminal systems, which we may call pri- cannot be left to competition (the prin- June 1997 on Interconnection in vate networks, are however not included. ciple of lightness). Obeying the basic Telecommunications with regard to These network segments are historically economic principle that the existence ensuring universal service and inter- provided and owned by most incumbent of large fixed costs and large returns to operability through application of the public operators. According to the ONP scale contradicts marginal cost pricing principles of Open Network Provi- principles, the location of network termi- regulation will be kept in mind. This sion (ONP) (Official Journal, nation points shall, however, be defined recognising the fact that broadband L199/32, 26.07.97). by the national regulatory authority and access networks will never be built if should represent a boundary, for regula- their owners are allowed to charge 7 Commission Notice of 31 March tory purposes, of the public only marginal costs (LRIC). The com- 1998 on the application of the com- telecommunications network. This putation of marginal cost also leaves petition rules to Access agreements boundary is in many cases drawn at the the regulators in charge of setting in the telecommunications sector foundation wall or in connection with prices and is discretionary. It will (Official Journal, C 265, 22.8.1998, other private property boundary lines. lead to heavy-handed regulation. p.2). Thus to ensure user terminal access to 2 Policy objectives must be clearly iden- public broadband networks, it may be 8 Draft Decision of the European Par- tified, by which regulation must be necessary to put regulatory requirements liament and of the Council on the related. The objective of creating a on private networks. coordinated introduction of a third- level playing field is thus only an generation mobile and wireless com- intermediate objective. New objectives munications system (UMTS) in the Concluding remarks for instance related to the building of Community (98/0051 (COD). Brus- a very costly information superhigh- sels, 20 Nov 1998. Current regulation does not, as described way should require totally different in this article, address broadband access regulatory means, if any. 9 Report on the findings of the EU- networks. Apart from discussing broader 3 Regulation must be technologically financed project: The application of access to local loop elements of the neutral. The current differences be- ONP to MAN, Frame Relay and incumbent operator, the issue has not tween the regimes for fixed and mobile Advanced Transmission Networks been discussed in any detail at all. Fur- should be abandoned. and their services. OVUM & Fischer thermore, it is not possible to isolate the & Lorenz, 1993. future regulatory development to broad- 58 Telektronikk 2/3.1999

60 10 Directive 97/51/EC of the European Parliament and of the Council of 6 October 1997 amending Council Directives 90/387/EEC and 92/44/EEC for the purpose of adapta- tion to a competitive environment in telecommunications (Official Jour- nal, L295/23, 29.10.97) 11 Spiller, P. Value-creating intercon- nect : survey of interconnection poli- cies. 1997. (SNF-report 50/97.) 12 OVUM. 1998. Unbundling the local loop : a regulatory and market assessment. Core study report. 13 St.meld. nr 24 (199899). Om enkelte regulatoriske sprsml i telesektoren. 14 European Commission Communica- tion (SEC (97)2390) of December 16, 1997 concerning the review under competition rules of the joint provision of telecommunications and cable TV networks by a single opera- tor and the abolition of restrictions on the provision of cable TV capacity over telecommunications networks. 15 Considerations of licensing condi- tions for UMTS network operations. London, OVUM, 1998. (UMTS Forum Report No. 4.) 16 Waverman, Shankerman, Pupillo. Asymmetric regulation of converging markets : problems, pitfalls and potential. Paper prepared for the TPRC Meetings, Washington, Octo- ber 1998. Per Mognes (48) was, at the time of writing this article, Senior Advisor at Telenor R&D. He has been employed by Telenor since 1977 doing research in data communication services, value added services and future telecom regula- tory regimes. Since March 1999 he has been employed as Senior Advisor for the Corporate Management of Telenor. email: [email protected] Terje Nord (32) was, at the time of writing this article, Research Scientist at Telenor R&D. He has been employed by Telenor since 1995, most of the time working with regulatory affairs and new competitive challenges in the telecommunications industry. Since March 1999 he has been employed as Advisor for the Corporate Manage- ment of Telenor. email: terje[email protected] Telektronikk 2/3.1999 59

61 IP or ATM in the access network? FRODE B. NILSEN This paper discusses the applicability This paper focuses on service integration 1.1 Organization of the paper of IP and ATM in the future broad- and the crucial point is how to perform The rest of the paper is organized as fol- band access network. These are the statistical multiplexing with QoS gua- lows. Section 2 presents a generic archi- two possible statistical multiplexing rantees. The key question is whether IP tectural model of the future broadband technologies that can be used to or ATM should be used. We argue that access network. It includes a discussion accomplish service integration. The ATM is the most appropriate multiplex- of fundamental issues concerning hous- development of a service integrated ing technique for the access network. ing facilities, hierarchical network levels network is required in order to pro- This is based on an assumption that and basic transmission technology. The vide efficient and cost-effective broad- guaranteed QoS is indeed a critical factor model is in turn used as a reference for band access. The main conclusion of and that ATM currently is a more mature the discussion of IP and ATM in the the paper is that ATM has beneficial technology in this respect. access network. However, the need for properties in the access domain and service integration is first discussed in it involves less risk when it comes to Otherwise our view rests on a general general terms in section 3. The focus is deployment. However, it is likely that argument that deploying an ATM based the rationale for service integration and the access network will operate in an access network involves less risk than how this is related to both the core net- IP oriented context. Then the combi- using IP. The point is that an ATM work and access network domains. Then nation of IP and ATM becomes an assumption is compatible with both in section 4 the general battle between IP important issue. This combination cre- views traditionally carried forward by and ATM as possible technologies to ates new opportunities that affect the the telco and datacom industries, respec- realize service integration is discussed. role of the access network. The paper tively. The telco view considers ATM as The two different views being represen- asks if the traditional provision ori- the universal protocol supporting end-to- tative for the telco industry and the data- ented role carried forward by ITU is end user communication. In the datacom com industry, respectively, are central to obsolete. view the role of ATM is reduced to a the discussion. Equipped with the general transmission technology with local sig- knowledge from section 3 and section 4 nificance. An overlaid IP layer is con- we are ready to discuss specifically how 1 Introduction sidered to be the proper level for uni- IP and ATM can be applied in the access versal end-to-end user communication. network. This is the subject of section 5 The future broadband access network Note, however, that the application of which provides an answer to the question will be different from the existing nar- ATM is still beneficial since the QoS posed in the title of the paper. The same rowband copper based network in three capabilities can be exploited by the IP applies for section 6 discussing the role respects: level. of access network. The key issue is if the Technological variety; role should be redefined as a result of the We argue that the datacom view is most new opportunities that arise when IP and Open provisioning; likely to prevail in the future and discuss ATM are introduced. The paper is con- how an ATM based access network fits Service integration. cluded in section 7. into an IP oriented context. This is in par- ticular related to the interface between Hence, we are facing paradigm shifts the access network and core network 1.2 Acronyms along three dimensions that change the operators. Since both IP and ATM are architectural requirements of the access A number of acronyms are used through- equipped with rich functionality con- network. Technological variety means out the paper. They are summarized in cerning control and switching new that both fibre solutions, satellite sys- the following table for the convenience opportunities arise. Hence, we ask if the tems, cable-TV networks, radio systems of the reader. provision oriented role of the access net- and xDSL solutions over the existing work as traditionally carried forward by copper network can be part of a unified ADSL Asynchronous DSL ITU is obsolete. Different industrial architecture. An open architecture is groups like ATM Forum and ADSL AR Access Router required for operating in a competitive Forum propose novel solutions implying environment with different core network AS Access Switch a new understanding of the access net- operators. Service integration refers to work. The first evolutionary step is ses- ATM Asynchronous Transfer Mode an access network in which all kinds of sion oriented free selection of core net- traffic are mixed by means of statistical BAP Broadband Access Point work provider. In the most radical case multiplexing1). This is in contrast to the access network can be equipped with CATV Community Antenna TV static multiplexing and dedicated connec- switching capabilities, thus working tions which are currently being used. The CLIP Classical IP over ATM closely together with the core network. objective is to provide a common access In an IP oriented context the ATM based CoS Class of Service solution replacing the set of specific access network will comprise an under- solutions being used today for platforms CR Core Router lying subnet in a larger IP network. Then like ISDN, FR, IP and ATM. The ratio- an assessment of techniques like CLIP, CS Core Switch nale for service integration is efficient NHRP, MPOA and MLPS for providing and cost-effective provisioning of broad- DSLAM Digital Subscriber Line Multi- IP over ATM is required. band access. The challenge is to maintain plexer QoS in the mixed traffic stream. ENET Ethernet 1) Also called dynamic multiplexing. FR Frame Relay 60 Telektronikk 2/3.1999

62 HDSL High Speed DSL 2 Broadband access representative for a CATV solution [10]. The local feeding point for the cable net- IDSL ISDN DSL requires a new work will then constitute a LAP. An IETF Internet Engineering Task architecture alternative interpretation of the same case Force is a radio solution where the base station Figure 1 is a generic illustration of how constitutes a LAP. In any case access IP Internet Protocol the future broadband access network takes place over a shared medium so that ISDN Integrated Services Digital Net- architecture is likely to be. The specific traffic is aggregated all the way to the work model and terminology used here is a customer premises. The case in the mid- result of a study performed by Telenor dle is representative for a hybrid solution ISP Internet Service Provider [1]. However, the model is perfectly in with a tree-structured PON in combina- ITU International Telecommunica- line with similar work carried out by tion with VDSL over the last drop [11]. tion Union other operators and equipment vendors If the fibre network covers a moderate [28]. number of customers it is natural to let L2TP Level 2 Tunneling Protocol the root of the tree, denoted an OLT, LAN Local Area Network The trend towards open provisioning define a LAP. If the fibre network has means that different administrative extended coverage it might be more LANE Local Area Network Emulation domains will be involved as marked with appropriate to let the optical termination LAP Local Access Point vertical dashed lines. In addition to the point in each leaf, denoted ONUs, consti- access operator marked with a yellow tute LAPs. LIS Logical IP subnet color, it is natural to distinguish between LMDS Local Multipoint Distribution transport providers and service providers. The point denoted broadband access Service In the case of a switched network service point (BAP) is central in the architectural we use the term platform operator in- model. A BAP links a number of local MPLS Multi-Protocol Label Switching stead. As suggested by the figure cross- areas on one side to one or more service MPOA Multi-Protocol over ATM domain management is an important nodes on the other side. A LAP is linked issue for the future broadband access net- to only one BAP whereas an SN can be NHRP Next Hop Resolution Protocol work. Management is beyond the scope connected to several BAPs. Hence, a NT Network Termination of this paper, though. BAP defines the termination of a local access network with the corresponding OLT Optical Line Termination The trend towards technological variety interface to the service nodes denoted ONU Optical Network Unit is illustrated by three different cases to SNI. The model facilitates use of inter- the right in the figure. The topmost case mediate transport providers at both sides OSPF Open-Shortest Path First corresponds to the existing narrowband of the BAP. This may happen either in PDH Plesiosynchronous Digital solution with copper lines terminating in terms of leased lines or over switched Hierarchy an RSU. The vertical lines branching off networks. In the latter case it is assumed at this point indicate how access to vari- that permanent virtual connections are PMD Physical Medium Dependant ous platforms is currently realized. The used as indicated by the dashed lines. PON Passive Optical Network existing solution is characterized by The point is that arbitrary switching is having dedicated connections and no not allowed, thus prohibiting direct con- POP Point Of Presence concentration of traffic in the access net- nection between two LAPs (or BAPs). PPP Point to Point Protocol work. In the future broadband architec- ture access will take place in terms of The physical location of a BAP depends PPTP Point to Point Tunneling service nodes (SN) as shown to the left on the customer base in the connected Protocol in the figure. The significant difference LAP areas, and also the location of the PSTN Public Switched Telephone from the existing solution is that the SNs. One extreme case is to co-locate the Network depth of the access network increases. BAP with one of the connected LAPs. Further, traffic will be concentrated in The other extreme is to co-locate a BAP QoS Quality of Service several steps between the end user and with an SN. In practice, the optimal loca- RSU Remote Subscriber Unit the service node. tion will be determined by a trade-off of the transport costs carried by the access RSVP Resource Reservation Protocol The term local access point (LAP) is operator and the service providers, RTP Real-time Transport Protocol used for the natural aggregation point respectively. The same trade-off will also covering the customers within a geogra- determine the number of LAPs and BAPs SDH Synchronous Digital Hierarchy phical area of moderate size. Compared in a nation wide access network. SN Service Node to the existing access network a LAP may correspond to a point where an RSU SNI Service Node Interface is located. Assuming that deployment 3 The Internet drives TC Transmission Convergence of ADSL [9] is the initial step towards service integration broadband access, a DSLAM will be TE Terminal Equipment located at the same point. The next step The vision of a world-wide service inte- VDSL Very High Speed DSL is to concentrate traffic even before the grated network was originally associated LAP. This is illustrated by the triangles with the work on B-ISDN and ATM as XDSL (Any) Digital Subscriber Line in the figure. The bottommost case is the enabling technology [10]. However, Telektronikk 2/3.1999 61

63 SNI UNI Service Transport Access Transport/Access Access Private domain domain domain domain domain domain (null) (null) Management Existing PSTN, ATM, IP SN NT TE RSU BAP NT SN LAP NT SN NT BAP LAP SN NT NT LAP ? ATM ? Net element Housing Aggregation/concentration ATM assumption No protocol assumption Figure 1 An architectural model of the future broadband access network the mass market has failed to request any It is natural to consider the development 3.1 The short-term solution large-scale service except ordinary tele- towards service integration in two does not affect the access phony up till now. For this reason there phases. In the short term the objective is network has been no driving force towards service to provide an efficient solution for Inter- Using the existing line switched architec- integration. The recent growth of the net access that exploits the existing line ture, in terms of dial-up connections or Internet as the killer application in addi- switched architecture. This does not leased lines, is the normal way to connect tion to telephony has led to a new situa- imply true service integration since the to the Internet today. In either case the tion. The challenge today is to develop existing telephony service is operating in resulting point-to-point link runs IP be- an overall network architecture which parallel. In the long run the objective is tween the customer premises and an can efficiently provide both a real-time to provide the telephony service over a access router as illustrated in Figure 2. service (telephony) and a best-effort data new integrated platform which can also These routers are located at selected service (Internet) all the way to the cus- be used for any other service, including places in the core network and admini- tomer premises. Internet access. Such a platform is pre- strated by various ISPs. It is customary sumably based on either IP or ATM. to denote such a connection point a POP. The session protocol normally being used 62 Telektronikk 2/3.1999

64 between the customer and the POP is PPP [11]. This protocol has mechanisms Access Net for accounting, authentification and authorization. Core Net The problem with the solution outlined CR in Figure 2 is that the bursty nature of IP traffic gives a low utilization of the dedi- cated link between the customer and the POP POP. For dial-up connections the holding AR time is also much longer than is the case for an ordinary telephony session. The effect is that the existing access and transport networks, which to some extent are tailored to telephony, are increasingly being used to carry IP traffic with widely different characteristics. The problem can be solved as illustrated in Figure 3 by moving the access routers CR Core Router (IP) IP/PSTN closer to the customers. Concentration of IP traffic close to the customers results in a multiplexing gain and gives improved AR Access Router (IP) IP/Leased Line overall link utilization. As the customer base grows the number of access routers will increase and the routers being Figure 2 The prevailing solution for Internet access today moved closer to the customers. An alternative to the IP solution is to use ATM or any other packet switched plat- form like FR or X.25 to achieve concen- tration of traffic before the POP. This is illustrated in figure 4 for the case of ATM. As suggested it requires that access switches are located close to the customers. Compared to the IP solution multiplexing takes place at a lower level but the gain in terms of improved link utilization is otherwise the same. The dis- advantage of using ATM for transport of IP is increased overhead and packet loss. Access Net This results in a throughput reduction of about 10 % compared to the pure IP solu- Core Net tion. One advantage of the ATM solution CR is that it is possible to change the ISP connection in a flexible way. Another point is that ATM can be used to carry other data services than IP. One impor- tant example is transport of Ethernet frames between remotely connected LANs. Note however, that it is possible to implement dynamic ISP selection and POP POP generic data transport also on an IP plat- AR AR form. The protocol suite PPP, PPTP [12] and L2TP [13] can be used to establish transparent tunnels in an arbitrary IP net- work. In general, the IP solution is preferable CR Core Router (IP) IP/PSTN when the ISP itself is taking responsibil- ity for connecting the customers to the AR Access Router (IP) IP/Leased Line POP. The ATM solution is a more natu- ral choice when the ISP buys a generic access service from a secondary pro- Figure 3 Pure IP solution for concentration of Internet traffic Telektronikk 2/3.1999 63

65 nating, with IP and ATM as parallel net- Access Net works. The platforms are realized on top of a synchronous transmission infrastruc- ture. Figure 6 is a net projection of this Core Net representation showing a stylistic view of POP CR CR POP how the transport network is being used for different purposes. The dashed lines indicate that access to the various plat- forms happens in terms of dedicated links as discussed in section 3.1. CS An expected evolution is that either IP or ATM will gain popularity and eventually constitute a unified platform for all ser- AS AS vice production. This will initially happen in the transport network domain as illustrated in Figure 7. Eventually the same evolution will take place in the access domain so that concentration of traffic can take place even closer to the CR Core Router (IP) CS Access Switch (ATM) customers. The last step is indicated by the arrows in the figure. CS Core Switch (ATM) IP/ATM It is important to note that the develop- ment of a universal service integrated Figure 4 ATM solution for concentration of Internet traffic platform represents a paradigm shift that affects the current understanding of the terms access network and transport net- work. When IP or ATM constitutes a layer on top of the existing infrastructure, a new type of transport and access net- work is formed. A side effect is that the existing clear distinction between the two domains gets blurred. We return to this Services Phone WEB, E-mail, ... LAN-LAN issue in section 6. Note also that the development of an integrated platform moves the focus from network aspects Platform PSTN IP ATM to service aspects. It should be added, however, that a fully paradigm shift is probably 10 years into the future. Synch. transmission Transmission ptp (PDH, SDH, IDSL, HDSL, ...) 4 There is a general battle between IP Figure 5 The existing situation with IP and ATM operating as parallel platforms along with PSTN and ATM In order to discuss the role of IP and ATM in the future broadband access net- work we need to have a general opinion on whether IP or ATM will be the pre- ferred choice to accomplish end-to-end service integration. This is the most con- vider. In any case it is important to note 3.2 The long-term solution troversial issue in the combined telco and that the access network itself is not implies a service inte- datacom industry today. It is important to affected. The connection to the closest grated access network keep in mind though, that the two indu- switch or router is dedicated for each stries approach this issue in different The transport infrastructure in terms of customer and operates in parallel with ways. Historically, the datacom industry dark fibre and PDH/SDH links is in- the telephony service. There is no con- has been data oriented without any creasingly being used to build switched centration of IP or ATM in the access regard to real-time requirement. It is only platforms like ATM or IP in addition to network domain. recently that this industry has gained telephony. Figure 5 is a pictorial repre- interest in supporting real-time applica- sentation of the situation today in terms tions. The telco industry has traditionally of a three-layered model. The figure been oriented towards telephony and suggests that the PSTN platform is domi- 64 Telektronikk 2/3.1999

66 Access Net real-time applications. Even if they have also developed efficient solutions for Core Net Core Net data communications it is only recently POP POP that this has become equally important in terms of traffic volume and revenue basis. The growth of the Internet is a IP main contribution in this respect. PSTN ATM The next two sections describe the views on IP and ATM that these two industries represent. The telco industry is in general IP/ATM ATM oriented, whereas the datacom industry is IP oriented. In section 4.3 we IP/PSTN argue that the IP oriented datacom view is most likely to win the ongoing battle. 4.1 The telco view is ATM oriented Figure 6 A network projection of the platform level in Figure 5 The key issue regarding the question about IP or ATM is how the Internet and the associated IP protocol is considered. The traditional view of the telco industry is shown in Figure 8 and considers Inter- net as an important service, but still as a service along with telephony, VoD, etc. Access Net The Internet is regarded as the killer application that has been missing in order Core Net to stimulate the development of a service integrated platform based on ATM. Note also that ATM defines basic transmission as indicated by the lower level in the fig- ure. A number of standards exist that IP/ATM define cell transmission for point-to- point and broadcast media. The main point of the telco view is that ATM is designed for QoS, thus enabling service integration. In contrast, the exist- ing IP technology does not have any sup- port for real-time traffic. Nevertheless, it is still possible to handle real-time re- quirements at a higher level by end-to- end control [14]. Some examples of this approach are the RTP protocol and also Figure 7 The long term development towards a service integrate platform the H.323 protocol family used for IP based on IP or ATM telephony. The overall idea is to monitor the end-to-end quality and exploit the available resources in the best possible way without actually giving any guaran- tees about the service quality. Experience has shown that this works satisfactorily as long as the network is not heavily WEB, E-mail, ... Services IP loaded. Phone, LAN-LAN, ... IP 4.2 The datacom view is IP Platform ATM oriented ATM The prevailing view of the datacom ATM transmission Transmission industry is that the IP technology will be ptp (PDH, SDH, HDSL, ADSL, VDSL, ...), bcast (PON, CATV, ...) further developed to support QoS. This means that resources can be pre-allocated Service integration in the network rather than relying on the end-systems to utilize the available Figure 8 The ATM oriented telco view on service integration Telektronikk 2/3.1999 65

67 switched connection at the lower level. This creates a short-cut bypassing any Services WEB, E-mail, IP-phone, PVN, ... intermediate routers. Any QoS require- ments can in turn be associated with the switched short-cut rather than being Platform IP IP handled at the IP level. The most promi- nent example of an MPLS implementa- Any transmission tion is IP over ATM as discussed in Transmission ATM ptp (PDH, SDH, ...), bcast (ENET, ...),swnet (ATM, ...) section 4.4.2. The disadvantage of the MPLS approach compared to imple- Service integration menting QoS at the IP level is that all involved subnetworks must be QoS en- Figure 9 The IP oriented datacom view on service integration abled. This is not the case for the large installed based of legacy systems like Ethernet. It should be noted that quality can be provided in a best-effort IP network by always providing sufficient resources. resources in the best possible way. The DiffServ and IntServ, respectively [16]. This approach is generally called capa- situation with a service integrated IP The third item is needed when true QoS city over-provisioning. It works within platform is shown in Figure 9. As is to be supported. User-driven resource a single administrative domain where all opposed to ATM it is important to note reservation requires that some sort of the resources are controlled by a single that IP is not a transmission technology. signaling mechanism is introduced. One operator taking responsibility for infra- It is rather assumed that IP packets can ongoing effort in this area is the RSVP structure investments. The problem arises be encapsulated by a transmission unit protocol [17, 18]. at the border to neighboring domains offered by an underlying subnetwork. which are outside the control of the Hence, IP can coexist with other network Implementing QoS and resource reserva- actual operator. layer protocols. There is a number of tion is in general a complex issue for IP standards for encapsulation of IP packets networks since there is no notion of a To summarize, there is today a heated including both point-to-point links, connection. The challenge is to take the debate on what is the proper way to broadcast networks and also switched connectionless nature into account so that provide IP QoS. Different solutions are networks. ATM is an example of the any reserved resources are moved acc- being implemented by operators but lack latter case and is increasingly being used ording to the dynamics of the routing of consensus and standardization compli- for transport of IP traffic. protocols. Two key concepts in this cate interoperability across administra- respect are soft-state and transient flow tive domains. detection. The technical challenge is to 4.2.1 IP QoS is a combine the connectionless nature with controversial issue resource reservation. The aim of RSVP 4.3 The datacom view is The evolution towards IP QoS is a hot is to handle the trade-off between these likely to win the game topic including at least three components: conflicting goals. The basic idea is to As illustrated to the right in Figures 8 associate resource reservation with tran- QoS routing; and 9 IP over ATM is a key element sient traffic flows. In order to co-ordinate regardless of the view taken. The differ- Traffic prioritization; this with the dynamic behavior of the ence is that the telco industry tradition- routing protocols the reservations are Resource reservation. ally has considered ATM as the level of volatile (soft-state). The difference com- service integration. In contrast, the data- pared to making reservations along estab- The first item implies that new routing com industry claims that this should take lished connections in an ATM network protocols supporting QoS metrics must place at the IP level. We argue that the (hard-state) is that the reservation be developed. One example of such an most likely outcome in the long term is requests must be repeated regularly in effort is the extension of the OSPF proto- that IP will be used for building a world- order to be maintained. A reservation is col [15]. The standard IP routing proto- wide service integrated platform with being maintained or moved depending cols being used today are rudimentary as end-to-end guarantees. The are four argu- on what is the current optimal route. they base their optimal route calculation ments in favor of this view [15]: only on minimizing the number of hops. An alternative approach is to handle QoS It is easier to build a world-wide plat- The second item means that packets are at a level subordinate to the IP level. The form based on IP since a large set of scheduled in the routers according to requirement is that the underlying subnet underlying transmission technologies some priority scheme different from first- technology is equipped with QoS fea- can be used. An obvious disadvantage come-first-served. Tagging of packets is tures. Further, there must be a tight inter- of the telco view is that a homogenous either based on statically assigned traffic action between the two levels. The initia- infrastructure based on ATM is re- classes or dynamic assignments set up by tive called MPLS is a standardization quired. the users. The former is generally de- effort in this direction [1922]. It relies noted CoS, whereas the latter facilitates There are implementations of IP for all on tagging traffic flows at the IP level true QoS. The two ongoing initiatives kinds of end-systems and operating and then make an assignment to a corresponding to these cases are called systems, and this SW is normally free. 66 Telektronikk 2/3.1999

68 Further, TCP/IP is the dominant appli- 4.4.1 The classical view is layered lifted up to the IP level. Figure 10 is an cation development environment. illustration of the situation where the log- The classical view is to consider an ical subnets are termed LIS. The third IP is designed for distributed routing underlying ATM network as an opaque solution is based on a distributed proto- protocols so that every packet is for- cloud offering interconnection between col called NHRP. It performs address warded on an individual basis. This arbitrary end-points. The functionality at resolution beyond subnet boundaries. gives fault-tolerance and simplified the ATM level is otherwise isolated from The rationale is that intermediate routers administration. In particular, it is easy the functionality at the IP level. Con- can be bypassed by providing a direct to build networks crossing administra- sequently, there are duplicated functions ATM connection. This gives improved tive domains. at the two levels, the primary example performance. The initiative called MPOA being the relationship between ATM sig- IP is a simple protocol assuming a [27] is an embracing standard including naling and IP routing. The standardized minimum of functionality from the all the three listed solutions as special solution which are classified as layered network elements. The main responsi- cases. A common feature of the classical are as follows [23]: bility for implementing reliable com- layered solutions is that IP is transported munication is left to the end-systems. LANE [24] over ordinary ATM connections between end systems as illustrated in Figure 10. CLIP [25] In sum these features make IP a flexible The figure indicates how NHRP supports networking technology that scales well. NHRP [26]. connections crossing subnet boundaries. A key property compared to ATM is the connectionless nature. Since every packet The first two are simple technologies 4.4.2 An integrated view is forwarded on an individual basis there with bad scaling properties. This is due improves efficiency is no state information associated with to centralized servers being responsible the network elements; ie. the network for address resolution. The result is that As already mentioned the disadvantage does not have any notion of which pair a large ATM network need to be parti- of a layered view is duplicated function- of stations are communicating at any spe- tioned into a number of smaller logical ality. In particular, IP routing and ATM cific time. The stateless property ensures subnets so that the communication at the signaling cover the same basic require- scalability, flexibility and fault-tolerance. ATM level does not cross the logical ment. An interesting development trend At the same time it represents a challenge subnet boundaries. To pass subnet today is integration of IP and ATM in with respect to QoS. The most important boundaries intermediate IP routers are terms of systems that use IP routing on argument in favor of ATM is that introduced and the communication is top of a raw ATM switching fabric resource reservation is simplified when it can be associated with established connections between the communicating parties. 4.4 IP over ATM is an interesting case The argument that IP in the long-run will be best suited to build a world-wide service integrated network rests on the scalability feature in addition to the ability to work over heterogenous under- lying networks. However, it is reasonable to anticipate that ATM will play an important role as a carrier of IP traffic within restricted areas and administrative domains. The access network can be con- sidered as an example of this as we will discuss in section 6. Hence, the interac- tion between IP and ATM becomes an important issue. We distinguish between a layered view and an integrated view as discussed in the next two subsections. It should be emphasized, however, that the interaction of IP and ATM is an issue only when the IP level sees the under- lying ATM infrastructure as a true net- work. Transporting IP over an estab- IP router LANE/Classical IP lished ATM connection between two end-points is straightforward. ATM switch NHRP Figure 10 Classical transport of IP traffic over established ATM connections Telektronikk 2/3.1999 67

69 like PON, xDSL or LMDS, we distin- guish between systems using either ATM or IP as the protocol layer above basic transmission: ATM/X; IP/frame/X. In the first case ATM cell transmission is defined for the actual medium by a TC layer above the PMD layer. For IP based systems a variable length frame format needs to be defined first. The frames are in turn used to carry encapsulated IP packets. The two parallel development tracks reflect the general battle between IP and ATM that is characteristic for the com- bined telco and datacom industry. The IP switch interest in IP based systems is mainly IP switch driven be the need for broadband access to the existing best-effort Internet. The advantages of ATM based systems are Figure 11 Piecewise short-cuts along a route in an IP switched network QoS support and increased flexibility. The latter refers to the fact that it is straightforward to use ATM for access to overlaid platforms other than IP. Despite our general argument in favor of IP from section 4.3 we recommend that ATM should be used for statistical multi- [2831, 23]. In contrast to the classical on short-cuts. If it is not desirable or pos- plexing and service integration in the view this can be called ATM under IP. sible to establish a short-cut, traffic is access network. There are two reasons The basic idea is to combine the scalabil- routed at the IP level in the ordinary way. for this. ity features of IP with the high switching Another point is that a short-cut can be ATM has better multiplexing proper- performance offered by ATM. It is inter- restricted to cover only a portion of the ties, in terms of resolution and real- esting to note that ATM switches are route between two nodes. This is illu- time behavior, over low-speed links. more cost effective that IP routers mea- strated in Figure 11. Likewise, there can This is due to the short fixed-length sured in terms of dollar per bytes per be several piecewise short-cut portions cells used by ATM. The generally second [28, 32]. along the route. Note also how this figure longer and variable-length packets is different from Figure 10 that applies to used by IP contribute to both latency Today a number of specific technical the classical case. and jitter resulting in reduced real-time solutions all based on the same integra- behavior. This is in particular true for tion idea are being studied. An effort to co-ordinate the various initiatives is 5 An ATM based low-speed links as is the case in the access network. called MPLS [1922]. It is also custom- access network ary to use the terms IP-switching or tag It is still uncertain whether the IP ori- switching as generic designations for the is recommended ented datacom view will eventually same idea. The essential difference from prevail. A service integrated access Section 2 outlined the structure of the the classical layered view is that the stan- network should also be able to support future broadband access network but did dard signaling protocols in ATM are not other platforms than IP in the short- not discuss how the anticipated service used to set up end-to-end connections. term and mid-term migration period. integration being discussed in section 3 Instead specialized light-weight signaling Hence, we argue that it is a significant can be realized. Equipped with the gene- protocols are developed that facilitate risk to base the access network deploy- ral discussion from section 4 we are establishment of short-cut connections. ment too strongly on the assumption ready to discuss how IP and ATM could These protocols are tailored to be used that a universal end-to-end IP platform be applied in the access network to together with IP. will be developed and standardized. accomplish this task. It is important to emphasize that the Note that the significance of the first Vendors of broadband access network rationale for creating a short-cut is to argument is weakened as the line speed equipment normally have two types of gain performance by avoiding processing increases. Hence, in high-speed back- systems in their product portfolio today. at the IP layer. However, the operation of bone networks it is likely that IP will be Letting X denote an arbitrary technology the network is not critically dependent an appropriate statistical multiplexing 68 Telektronikk 2/3.1999

70 SNI UNI Service Transport Access Transport/Access Access Private domain domain domain domain domain domain (null) (null) SN NT TE BAP LAP (a) ATM ATM ATM ATM IP IP IP (b) ATM ATM IP IP IP IP (c) ATM (d) ATM ATM Figure 12 Elaborated view on how an ATM based access network may fit into a wider context technology which can also support traffic dance with the layered view from Figure is that the BAP is transparent for IP in with real-time requirements. The key 9 we distinguish between a basic trans- the former case. Consequently, ATM point about the second argument is that mission level and an overlaid platform transmission must be used also between the ATM oriented telco view is not level. the BAP and the SN. Letting the BAP excluded. At the same time the protocol interpret IP as in the latter case gives assumption is compatible with the IP Figure 12 gives an elaborated view on improved flexibility with regard to the oriented datacom view. how an ATM based access network can transmission technology being used be- fit into a wider context. We distinguish tween the BAP and the SN. It is also pos- The view that ATM is the preferred pro- between four cases denoted (a)(d), each sible to let the BAP be equipped with a tocol for statistical multiplexing in the being characterized by the operating richer set of functionalities as discussed access network is illustrated by the environment at the SNI and UNI inter- in the next section. Case (d) corresponds bottommost double arrow in Figure 1. faces. to a situation in which no specific proto- The small vertical crosslines signify col view prevails. Rather arbitrary points at which ATM cells are being pro- Case (a) corresponds to the pure ATM parallel platforms are being developed as cessed; ie. the figure tells that both BAP oriented telco view with cell transmission illustrated by the pair of dashed lines. It and NT will participate in interpretation at every stage. No platform level is indi- is reasonable to anticipate that ATM is of the ATM cell stream. Depending on cated since ATM is also a networking being used between the BAP and the SN the actual transmission technology it technology supporting end-to-end user in this case so that the BAP becomes might be additional equipment between communication. The cases denoted (b) platform independent. An obvious dis- these points that also processes ATM and (c) correspond to the datacom view advantage of coexisting platforms is that cells. The typical example is equipment in which IP is used at every stage to it is necessary to support a larger set of located in a LAP. To simplify the figure accomplish internetworking across protocols at the UNI and SNI interfaces. any such intermediate points are sup- underlying transmission technologies. pressed. The dashed horizontal arrows Hence, IP provides an overlaid platform In accordance with the discussion in sec- and cross lines at the bottom of Figure 1 supporting end-to-end user communica- tion 4.3 we anticipate that the operating indicate open protocol issues. In accor- tion. The difference between (b) and (c) environment will be dominated by IP in Telektronikk 2/3.1999 69

71 the long run. This applies to both the ser- the need for manual intervention. Pro- dynamic scenario. By letting the access vice domain and the private customer vided there exists a universal overlaid network be equipped with routing capa- domain. Hence, case (c) is considered to platform it is also possible to let the bilities it can interact closely with core be the most likely outcome. The ATM access network interact closely with con- networks. based access network will then act as a nected core networks. As already stated carrier of IP traffic. This can be accom- we expect the latter to be based on IP. A common feature of all cases in Figure plished in a number of ways depending Together these two features open for a 13 is the existence of an association be- on what is considered to be the role of different understanding of the role of the tween the customer and the BAP. It is the access network as discussed next. access network. By putting extended realized in terms of an ATM connection functionality like routing in the network which is semi-permanent in the first two and enable user controlled management cases and switched in the latter case. Any 6 The role of the access it is natural to consider more complex QoS guarantees are linked to the under- network is affected scenarios than provisioning which is lying ATM connection. Otherwise there dominating today. is a significant difference between cases The traditional role of the access network (A) and (B) on one side and case (C) on has been to connect end customers to Figure 13 shows three interpretations of the other side. In the latter case there is service providers on a semi-permanent the role of the future access network. In no association between the BAP and the basis. We call this provision oriented in accordance with alternative (c) in Figure SN. The key issue is what kind of func- the sense that the customer relationship 12 we assume that IP will constitute an tionality the BAP is equipped with at the is associated with a subscription that lasts overlaid layer and that ATM is used for IP level. for a period of time. Physically the underlying transmission in the access access is realized either in terms of a domain. The cases denoted (A)(C) in In case (A) end customers are associated leased line or a dial-up connection to the Figure 13 are distinct in the way cus- with service providers on a semi-perma- closest switching point for the actual tomers are associated with SNs. A solid nent basis. This is the conservative inter- platform. Changing the customer rela- line style is used to indicate fixed asso- pretation carried forward by ITU and is tionship normally involves manual inter- ciations whereas a dashed line style is called provision oriented. It implies that vention like updating registry systems, used to signify an association that is the access operator configures the net- physical reconnection or installation of established on-demand. Case (A) corre- work elements so that all traffic from a new transmission equipment. sponds to the traditional provision ori- given customer follows a specified path ented view with a fixed association. Case all the way to the service node; ie. any An access network based on ATM is (B) represents a session oriented view ATM cross-connect equipment between more flexible and facilitates automized with a dynamic selection of service pro- the customer and the BAP and also the IP management and control, thus reducing vider. Case (C) is the most radical and forwarding table in the BAP are statically SN UNI Service Transport Access Transport/Access Access Private domain domain domain domain domain domain (null) (null) SN (A) NT TE (B) (C) SN BAP LAP IP IP IP IP ATM Figure 13 Different interpretations of the role of the access network 70 Telektronikk 2/3.1999

72 configured by management. There is no (B) is comparable to a dial-up connec- next step will be to implement free selec- dynamics in terms of user control or rout- tion. The difference is that the BAP takes tion of service provider per session. The ing. A customer can have provisioned the role of a generic access server and ultimate step is to let the access network connections to several SNs at the same that appropriate signaling mechanisms become a true IP subnet interworking time, of course, and he is free to pick are being developed. The PPP protocol is closely with arbitrary subnets in the core which one to use at any time. Even if central in this respect. domain. case (A) is called provision oriented it is important to note that each customer To summarize, the introduction of IP and does not have a dedicated channel. Traf- ATM in the future broadband access net- 8 References fic from different customers is statisti- work opens for increased complexity and 1 Ims, L et al. Aksess 2005 aksessnet- cally multiplexed on the way towards the added functionality which affect the role tarkitekturer. Kjeller, Telenor R&D, SN. This happens at the ATM level with- of the access network. A likely evolution 1998. (R&D report R 44/98.) in the access network and at the IP level scenario is to start from the provision ori- in the BAP. ented case (A), then turn to case (B) and 2 Donnelly, A, Smythe, C. A tutorial free selection of service provider. The on the Digital Audio-Visual Council Case (C) illustrates the most radical ultimate step is (C) which considers the (DAVIC) standardization activity. interpretation of the role of the access access network as a sub network in a Electronics & Communications En- network and assumes that a full fledged larger IP network. Note that the tradition- gineering Journal, 4656, 1997. router is located in the BAP. This gives ally distinct border between core and full dynamics in the sense that IP packets access gets blurred in cases (B) and (C). are routed individually without any asso- The point is that the access operator takes 3 ADSL Forum. ADSL Forum System ciation to an SN. Neither are there any common responsibility of tasks which are Reference Model. Fremont, 1996. pre-established connections at the cus- otherwise left to the various core network (ADSL Forum TR-001.) tomer side. Rather, connections are providers. Hence, the access network will established as needed depending on the expand towards the service providers forwarding decision at the IP level. rather than the other way round, which is 4 ADSL Forum. Requirements & Ref- Among the different solutions for run- the case today. erence Models for ADSL Access Net- ning IP over ATM, both CLIP and works: The SNAG Document. Fre- LANE can be used. Hence, the access network can be considered as an IP sub- 7 Conclusion mont, 1998. (ADSL Forum TR-010.) net using ATM as underlying transport. The main conclusion of the paper is that Note that if ATM is used also between ATM is considered the best technology 5 ATM Forum. Residential Broadband the BAP and the SN, or at the customer to obtain statistical multiplexing and ser- Architectural Framework. 1998. side, it is possible to use either NHRP, vice integration in the access network. (ATM Forum af-rbb-0099.000.) MPOA or MPLS. The gain is that the This is partly due to an argument that routers at the two ends of the access net- ATM gives better QoS performance over work can be bypassed in terms of a short- low-speed links. But it is also due to the 6 FSAN. Full Services Access Network cut at the ATM level. fact that an ATM based access network Requirements Specification, draft G. is applicable regardless of whether IP or Ipswich, 1997. An important point about case (C) is that ATM will constitute the universal plat- com/profsoc/access/ the access network operator takes respon- form for end-to-end user communication. sibility for administrating a router in the It involves less risk to start deploying an 7 DVB. A guideline for the use of DVB BAP which provides an interconnection ATM based access network since it can specifications and standards Digital point for other network operators. One also be used in an IP oriented context. Video Broadcasting (DVB). (DVB immediate consequence is that traffic can Report tr-101-200.) http://www. be routed also within the access network Despite the uncertainty about what will domain. This breaks fundamentally with be the preferred platform for end-to-end the classical understanding of the access user communication, we argue that the 8 ITU. Framework recommendation on network. Nevertheless, such solutions are IP oriented view carried forward by the functional access networks (AN). being studied by industrial groups like datacom industry is the most likely out- Geneve, 1995. (ITU-T G.902.) ATM Forum [5] and ADSL Forum [3]. come. The connectionless nature of IP The obvious advantage is increased flexi- makes it a scalable, flexible and fault-tol- 9 Ims, L. Wireline broadband access bility. The challenge is handling of secu- erant networking technology. A pertinent networks. Telektronikk, 95 (2/3), rity, accounting and charging. question is how an ATM based access 7387, 1999. (This issue.) network can fit in a wider IP oriented Case (B) is an intermediate solution context. The point is that the combination 10 de Prycker, M. Asynchronous Trans- facilitating free selection of service of IP and ATM opens for new opportuni- fer Mode. Ellis Horwood, 1993. provider. This happens per session and ties affecting the role of the access net- the customer must go through a proce- work. We do not make any conclusions 11 IETF. Point-to-Point Protocol dure every time to handle authorization, at this point. A likely evolution is that the (PPP). 1994. (IETF RFC 1661/STD authentification and configuration. Com- broadband access network will initially 51.) pared to case (C) this solution gives be provision oriented in the same way as reduced dynamics. In some sense case the existing narrowband network. The Telektronikk 2/3.1999 71

73 12 IETF. Point-to-Point Tunneling Pro- 23 Klovning, E, Ingvaldsen, T, Clemet- 32 Keshav, S, Sharma, R. Issues and tocol PPTP. 1998. (IETF Internet sen, M. Internett nettarkitektur og trends in router design. IEEE Com- Draft, draft-ietf-pppext-pptp-04.txt.) funksjonalitet relatert til ulike bred- munications Magazine, 36 (5), bndsteknologier. Kjeller, Telenor 144151, 1998. 13 IETF. Layer Two Tunneling Protocol R&D, 1998. (R&D report R 10/98.) L2TP. 1998. (IETF Internet Draft, draft-ietf-pppext-l2tp-11.txt.) 24 ATM Forum. LAN Emulation over ATM. 1995. (v.1.0. ATM Forum af- 14 Kostas, T. Real-time voice over lane-0021.000.) http://www.atmfo- packet-switched networks. IEEE Net- work, Jan/Feb, 1998, 1827. 25 IETF. Classical IP and ARP over 15 Huitema, C. Routing in the Internet. ATM. 1998. (IETF RFC 2225.) Prentice Hall, 1995. 26 IETF. NBMA Next Hop Resolution 16 Crowcroft, J et al. A rough compari- Protocol (NHRP). 1998. (IETF RFC son of the IETF and ATM service 2332.) models. IEEE Network, 1216, Nov/Dec 1995. 27 ATM Forum. Multi-Protocol Over ATM. 1997. (v. 1.0. ATM Forum af- 17 Zhang, L et al. RSVP : a new mpoa-0087.000.) http://www.atmfo- resource ReSerVation protocol. IEEE Network, 818, Sep 1993. 28 White, P P. ATM switching and IP 18 RSVP and Integrated Services in the routing integration : the next stage in Internet : a tutorial. IEEE Communi- Internet evolution. IEEE Communi- cations Magazine, 35 (5), 100106, cations Magazine, 36 (4), 7983, 1997. 1998. 19 IETF. A Framework for Multiproto- 29 Newman, P, Lyon, T, Minshall, G. col Label Switching. 1997. (IETF Flow Labelled IP : a connectionless Internet Draft, draft-ietf-mpls-frame- approach to ATM. In: Proc. IEEE work-02.txt.) Infocom, San Francisco. 1996, 1251 1260. 20 IETF. Multiprotocol Label Switching Architecture. 1998. (IETF Internet 30 Guarene, E, Fasano, P, Vercellone, Draft, draft-ietf-mpls-arch-02.txt.) V. IP and ATM Integration Perspec- tives. IEEE Communications Maga- 21 IETF. Use of Label Switching With zine, 36 (1), 7480, 1998. ATM. 1998. (IETF Internet Draft, draft-ietf-mpls-atm-00.txt.) 31 Dumortier, P. Toward a new IP over ATM routing paradigm. IEEE Com- 22 Davie, B et al. Switching in IP net- munications Magazine, 36 (1), works. Morgan Kaufman, 1998. 8286, 1998. Frode B. Nilsen (32) holds a Masters degree (1993) and a Doctoral degree (1998) in computer engineering and data communication from the University of Oslo. The academic research focused on methods for performance evaluation of high-speed networks. He is currently employed at Telenor R&D as research scientist working on network architectures. email: [email protected] 72 Telektronikk 2/3.1999

74 Wireline broadband access networks LEIF AARTHUN IMS 1 Introduction 2 Service characteristics architectures like high speed digital sub- scriber line (HDSL) and optical fibre The future broadband access network and network architec- point-to-point systems. architecture needs to be flexible enough tures to efficiently support the provision of a The point-to-multipoint, or multicast and full set of broadband and narrowband The selection of the network architecture broadcast communication configuration services with a wide range of capacity is strongly related to the services and are most readily implemented in shared demands in a highly diverse demographic applications to be provided over the net- medium architectures as passive optical environment. work [8, 9, 10]. Hence, the service and networks (PONs), tree-and-branch co- application attributes are of great impor- axial cables and point-to-multipoint radio The selection of transmission medium tance in the selection of architectures. systems. In other types of architectures, or sets of transmission media and corre- The relationship between the applications including synchronous digital hierarchy sponding network topologies, node con- and the proposed bearer services is (SDH) rings, the point-to-multipoint con- figurations and system technologies is addressed in more detail in [11]. The figurations require some additional func- among the key strategic issues in the terms application and service are not tionality in the active equipment. How- development of the access network. unambiguously defined in the telecom- ever, point-to-point, or narrowcast con- Capacity demand, service types, flexibil- munication environment. In the context figurations require additional capabilities ity and costs are the main differentiators of this paper the application refers to the like filtering, blocking, and encryption in in the evaluation of the various access use of the telecommunication service shared medium architectures. network architectures. In the following made by the customer and the fact that the term architecture denotes (a) specific the service is the product sold to the cus- The communication establishment im- combination(s) of transmission medium, tomer by the service provider. Addition- pacts the mix of capacity permanently network topology, node configuration ally, the telecommunications services can allocated in the access network and and system technology. be divided into information services and capacity allocated on demand. Video communications services. Information distribution services such as television A variety of broadband access technolo- services permit the access to a service broadcasting, pay per channel (PPC), gies are available. The five main wireline centre for access to data bases, movies pay per view (PPV) and near video on upgrade alternatives are shown in Figure on demand, etc. Communication services demand (NVOD) require a fixed amount 1, namely digital subscriber line (DSL) allow the interconnection of remote users of bandwidth per customer. Additionally, modems, coaxial cable modems, fibre and the dialogue between them for the establishment of connections on systems, power line modems and the videotelephony, file transfer, etc. The demand and the control of information hybrid fibre-copper systems. The latter impact of the service selection on the flow and format require the presence of comprise all the architectures based on a choice of the access network architecture a signalling channel up to the nearest net- combination of fibreoptic transmission in may be based on the values of the fol- work element capable of realising such the upper part of the access network and lowing attributes: functions. different copper drop technologies in the Modulation technique; lower parts of the network. These hybrid Mobility implies that some category of combinations are probably the most Bit rate for digital services; services can be implemented either on a interesting ones, in particular in the long fixed architecture or on a wireless archi- Bandwidth required; term. This paper describes the main tech- tecture only, at least in the distribution nologies and architectures commercially Symmetry; part. Services which allow mobility are available today, some of the emerging necessarily implemented on wireless Communication configuration; technologies and some alternatives which architectures, but other features for con- are expected to reach the market within Communication establishment; trol of the position of the subscribers and the next years. The paper discusses the call handling are required. Mobility. key parameters that differentiate the vari- ous access network solutions and some The bandwidth and bit rate required to of the key issues faced by the access net- support the service are a fundamental work infrastructure providers in the attribute of services with respect to the migration of the present network to selection of the architecture, since some Coaxial Digital broadband. The likely migration paths types of physical media and transport cable subscriber from the existing wireline access net- systems may not provide the required modems line works towards broadband are outlined. capacity. The bandwidth required is not modems The intention is neither to provide a com- Hybrid an independent parameter as it is deter- plete overview of access network tech- fibre-copper mined by the type of modulation and nologies and architectures, nor to give a systems coding adopted for the information thorough description of the respective transfer. technology and architecture options. The main objective is to provide a brief intro- Power line Fiberoptic The symmetry of the service differenti- modems systems duction to this aspect only. Several of the ates between architectures which have technologies are treated in more detail in been conceived for asymmetric services, the following contributions in this issue like asymmetric digital subscriber line of Telektronikk [1, 2, 3, 4, 5, 6, 7]. Figure 1 The main groups of broadband (ADSL), traditional coaxial cable sys- wireline access network alternatives tems, some radio systems, and symmetric Telektronikk 2/3.1999 73

75 3 Topology, node 4 Digital subscriber standard modem has a 1.5 km trans- mission distance [17]. For both down- configuration and line architectures stream capacity options the upstream system technology Until recently twisted pair has been con- capacity is at least 2 Mbit/s. The sym- metric VDSL versions will have sidered to be a significant bottleneck Today there is a wide range of alterna- 26 Mbit/s (0.3 km) and 13 Mbit/s with respect to capacity. Recent develop- tives for upgrading the physical layer of (1 km) transmission capacities. Ana- ments of complex modulation schemes the access network to broadband [12, logue telephony is to be transmitted have enabled the extension of the line 13]. The different available wireline on the same twisted pair. However, capacity by order of magnitudes. In the transmission media are twisted pair cop- there is an uncertainty with respect short term the main advantage of copper per cable, coaxial cable and fibre cable. to the coexistence of symmetric and is a variable cost option, alleviating the For each of the media there are several asymmetric VDSL systems in the same need for high and risky up-front invest- different options with respect to network cable sheath. ments. In addition the installed base of topology, node configuration and system 600 million lines world-wide constitutes technology. In general for the digital subscriber line an immediate advantage. The enhanced (DSL) options, there is a trade off copper or digital subscriber line (DSL) The physical network topology refers to between distance and capacity available. technologies differ with respect to trans- the physical network links. The network Table 1 shows the transmission distance mission capacity, transmission distance topology can be rather complex, depend- for the various symmetric or asymmetric and the number of twisted pairs used ing on the number of levels of the net- capacities obtainable with the transmis- [14]. The DSL system technologies work itself. However, the main topology sion equipment available today. Trans- include: IDSL (ISDN digital subscriber options are point-to-point networks and mission distances are shown for 0.4 mm line), ADSL, VDSL (very high-speed point- or multipoint-to-multipoint net- twisted pair cable and for asymmetric digital subscriber line), HDSL (high bit- works. In the former each physical link switched broadband (ASB) and symmet- rate digital subscriber line) and SDSL is dedicated to connect two (active star, ric switched broadband (SSB) service (symmetric digital subscriber line) [15, double active star) network terminations. transmission. 16]. At present only ADSL and HDSL The latter connects several network ter- are commercially available, VDSL and minations (single/double passive star, Figure 2 depicts the variations in trans- SDSL are expected to become available bus, tree). mission reach of the different DSL sys- quite soon. tems with reference to a commonly The node configuration refers to the con- IDSL (ISDN digital subscriber line) is found access network, including the local figuration of the network terminations of the transmission system used for ISDN exchange, distribution cabinets, tele- one particular transmission medium, ie. Basic Access. phone cables and local distribution the network nodes at which the terminal cables. ADSL uses one twisted copper pair for equipment is located. For fibre architec- transmission of between 2 Mbit/s tures, several node configurations have Figure 3 shows an ADSL architecture (4 km) and 8 Mbit/s (2 km) down- been proposed: fibre to the cabinet with a multiplexer located in the local stream and up to 640 kbit/s upstream. (FTTCab), fibre to the node (FTTN), exchange. In Figure 3, at a later stage in The transmission distance varies from fibre to the curb (FTTC), fibre to the the network evolution, the access multi- 2 km to 4 km, depending on the building (FTTB) and fibre to the home plexer is co-located with the remote sub- selected transmission capacity. POTS (FTTH) are commonly used, depending scriber unit (RSU) in the service access or ISDN is transmitted on the same on the local network area conditions. point (SAP). An SDH ring between the twisted pair. local exchange (LEX) and the SAP is The system technology attributes are the HDSL usually requires two pairs for depicted. The infrastructures for plain old multiplexing technique, the access proto- symmetric 2 Mbit/s transmission, but telephone system (POTS) and N-ISDN col for point-to-multipoint configurations systems for one or three pair transmis- are included for clarity. and the modulation techniques (basically sion are available. The typical trans- classified in analogue and digital tech- mission distance for HDSL is in the Fibreoptic transmission systems are com- niques). Typical multiplexing techniques range of 3 km. Analogue telephony monly considered to be the only wireline are frequency division multiplexing may not be transmitted simultaneously alternative for providing capacities (FDM), time division multiplexing on the same twisted pair. higher than the ones offered on individ- (TDM), subcarrier multiplexing (SCM) ual DSL systems on the existing twisted SDSL (symmetric digital subscriber wavelength division multiplexing pair network, such as 155 Mbit/s. How- line) is a broader class of systems, (WDM), high density wavelength divi- ever, by combining a set of twisted cop- which still are symmetric, but not nec- sion multiplexing (HDWDM) and optical per pairs, each with DSL systems in- essarily transmit 2048 kbit/s capacity. frequency division multiplexing (OFDM). stalled (ADSL or VDSL), an aggregate SDSL typically uses one twisted pair, Commonly used access protocols are fre- capacity of say 155 Mbit/s or even 622 and may include the option of trans- quency division multiple access (FDMA) Mbit/s can be transmitted between two mission of analogue telephony on the time division multiple access (TDMA), network locations on the existing twisted same twisted pair. code division multiple access (CDMA), copper pair network. Thus, the possibly subcarrier division multiple access VDSL provides both asymmetric and prohibitive high civil works costs associ- (SCMA) and wavelength division mul- symmetric transmission, with asym- ated with installation of fibreoptic cable tiple access (WDMA). metric capacities as high as 26 Mbit/s may be avoided. This concept of parallel (1 km). The 13 Mbit/s capacity VDSL DSL systems is called inverse multiplex- 74 Telektronikk 2/3.1999

76 Table 1 Transmission distance (in km) for the various symmetric or asymmetric 5 Coaxial cable modem capacities on 0.4 mm cable architectures Bitrate 2 Mb 4 Mb 8 Mb 10 Mb 13 Mb 26 Mb 52 Mb The coaxial cable networks, or Commu- nity Antenna Television (CATV) net- ASB 3.5 3.3 2.5 2 1.5 1.0 0.6 works, were originally designed to dis- SSB 2.5 1.6 1.1 0.8 0.6 0.3 0.2 tribute analogue television signals. The coaxial cable network consists of cas- caded unidirectional coaxial amplifiers, usually offering an overall bandwidth of 300 MHz or 450 MHz. These conven- tional CATV networks now constitute an increasingly valuable asset for the cable ing, and is described in more detail in Lite. Additionally, the ADSL-forum and operators as the momentum of the con- [3]. The main motivation behind inverse the Universal ADSL Working Group vergence of services and technology in multiplexing lies in the fact that it may (UAWG) are influential promoters of the broadcasting, data communications be a rather inexpensive way to avoid the ADSL technology and ADSL Lite, and telecommunications sectors in- high civil works costs of fibreoptic cable respectively. installation in the primary access network in particular. In the lower part of the high capacity domain we find the transmission solu- tions IDSL, HDSL and SDSL. IDSL and HDSL are being widely deployed today, and SDSL is also technically mature and manufactured by several suppliers. How- ever, products for the upper part of the high capacity domain, typically based on VDSL technology, still have not reached 5 the market at a significant scale. Today, it is products within the middle part of 4 the high capacity domain which has the 2 Mb highest momentum within the industry 3 with an increasing number of vendors of 6 Mb ADSL technology. The products are now 2 7 Mb further developed towards higher capac- ity limits, flexible configuration of trans- 10 Mb 1 mission parameters, efficient manage- 25 Mb ment of different user groups and Inter- net service providers and improved oper- ator interfaces. Currently, the concept of inverse multiplexing is being developed towards prototyping. Recently significant attention has been devoted to the concept of ADSL Lite, which is also known as UADSL (univer- sal asymmetric digital subscriber line), CDSL (consumer digital subscriber line) or G.Lite. The ADSL Lite is very similar to ADSL from a technological point of view. The primary motivation behind ADSL Lite is the potential cost reduction obtained through simpler and more robust customer premises equipment which the customer himself may install. However, this potential cost advantage Distribution cabinet over ADSL is achieved at the expense of relaxed transmission capacity. Initially Local exchange building the ADSL standards were established in Primary distribution cable ANSI and are now further developed to Secondary distribution cable ITU recommendations, possibly with additional recommendations for ADSL Local distribution cables Figure 2 The reach of the different DSL systems Telektronikk 2/3.1999 75

77 MDF NT1 LEX ISDN DSLAM POTS ADSL ADSL Primary Secondary Customer Local exchange flexibility point (PF) flexibility point (SF) premises

78 most likely to be used in the future. In mission capacity. Today, implementing FITL upgrades have only to a very the remainder of this section, the descrip- cable modem technology in existing limited extent been implemented over the tion of the coaxial cable modem architec- coaxial cable networks with several thou- past years, and it is expected that it will ture is based on systems similar to the sands of customers connected per coaxial still be years before the fibre deployment IEEE 802.14 standard systems [19]. cable segment is first and foremost to be in the access network for residential Typically in such networks the frequency considered as an alternative to narrow- broadband services reaches economically spectrum from 5 to 45 MHz is dedicated band ISDN, if a cost-effective network justifiable levels in non-greenfield areas. to upstream digital transmission. This dimensioning is to be used. Providing 40 MHz band is sectioned into multiple higher capacities, comparable to the Depending on the local network area 1 MHz 6 MHz wide radio frequency capacities enabled by ADSL technology, conditions, there is a variety amongst (RF) channels, each with a capacity will in most cases require a splitting of FITL architectures regarding node con- between 1.6 Mbit/s and 10 Mbit/s. The the coaxial network into smaller coaxial figurations, ie. the location of the optical frequency spectrum between 50 and segments in order to achieve the required transmission terminal equipment. The 450 MHz is used for downstream ana- return path capacity. Typically, the net- most common configurations are: logue broadcast transmission, and the work is split into smaller segments by the Fibre to the cabinet (FTTCab); 450 MHz to 750 MHz frequency spec- use of fibreoptic feeder cables and trum for downstream digital transmis- (hybrid fibre coaxial) HFC technology. Fibre to the node (FTTN); sion. Figure 5 illustrates an architecture HFC is treated in a separate section of Fibre to the curb (FTTC); designed for coaxial cable modems. The this paper. In this section the focus is on figure shows that coaxial cable return the coaxial cable modem architecture. Fibre to the building (FTTB); amplifiers are installed in addition to Fibre to the home (FTTH). cable modems and a cable router termi- nates the coaxial cable network segment 6 Optical fibre system The architectures based on FTTCab, at the hub side. In the case of each sub- architectures FTTN, FTTC and very often architec- scriber being fitted with a 10 Mbit/s tures based on FTTB are hybrid solu- cable modem typically 500 600 sub- Optical fibre is the transmission medium tions, ie. such networks are implemented scribers share 50 60 Mbit/s upstream which offers the highest bandwidth, with by the use of fibre systems in combina- capacity on one coaxial network seg- transmission capacities potentially as tion with twisted pair, coaxial cable or ment, with 10 Base-T Ethernet interfaces high as Tbit/s [21]. The concept of intro- wireless drop technologies. These hybrid at the customer and hub side. ATM-25 ducing fibre-optic cable and transmission architectures are described in more detail interfaces are expected to be available systems in the access network dates back later in the next section of this paper. quite soon from several vendors, and to the early eighties, and several field This section is devoted to the pure broad- there is work going on within IETF on trials have been carried out. However, band fibreoptic access network architec- the issues related to IP transport in such introduction of fibre in the loop (FITL) tures. networks [20]. requires the deployment of fibre cable, which is usually associated with high The majority of the fibre based broad- In coaxial cable modem networks the civil works costs. In addition, the present band access systems installed today are simultaneous users on one coaxial net- fibre optics terminal equipment is rather either rolled-out in an FTTB or an FTTH work segment share 50 60 Mbit/s trans- costly, due to low production volumes. D1- amplifiers ODF D2- 300-860 MHz amplifiers CATV EO OE Cable Router modem D0 D1 D2 D3 network network network network Customer Hub D2 D3 premises Figure 5 Architecture designed for coaxial cable modems Telektronikk 2/3.1999 77

79 configuration. Fibre-optic transmission opment and growth areas, areas where cally 90 % of the copper loop lengths are systems for capacities from 34 368 kbit/s the existing twisted pair network is shorter than approximately twice the and higher are currently being installed replaced, or in greenfield areas. average copper loop length for the whole to customers with the highest capacity exchange area. Thus, a 90 % coverage demand, ie. exclusively business cus- The second generation FITL systems will with 8 Mbit/s DSL requires average cop- tomers. Also available today are fibre- provide broadband capacity, for instance per loop lengths of 1,300 metres. optic transmission systems with four with a twisted pair digital subscriber line channels of 2 048 kbit/s capacity each. (DSL) modem pair between the optical Figure 6 summarises the theoretical Usually point-to-point connections are node and the customer premises. The cumulative copper loop length distribu- used, alternatively ring topologies. The recent developments in DSL technology tions for areas with different average systems with at least 155 520 kbit/s enables the PNOs to provide broadband loop lengths. The coverage of customers capacity are normally SDH systems, capacities on the existing copper network in the area (in per cent) is shown as a which can easily be adopted to carry at a reasonable cost. However, in general function of loop length. The bitrates on ATM traffic. Alternatively, the fibre- for the DSL options, there is a trade off the horizontal loop length axis corre- optic transmission systems can be based between copper loop distance and capac- sponds to the asymmetric capacities that directly on ATM transport. ity available as described earlier in this may be provided on 0.4 mm twisted pair paper. For instance, the transmission dis- copper cables (Table 1). As can be seen More advanced optical technologies like tance on 0.4 mm twisted pair cable for from the figure, for an exchange area wavelength division multiplexing (WDM) 2 Mbit/s asymmetric capacity is 3.5 km, with a 800 m average loop length, may enable future access networks with whereas 25 Mbit/s asymmetric capacity 13 Mbit/s asymmetric capacity may be significantly increased flexibility and has a distance limitation of 1.0 km on offered to 90 % of the customers, since capacity, as presented in more detail in this cable. Thus, in order to offer higher 90 % of the copper lines in the area are [2]. However, the near-term and medium capacities the copper loop length of the below 1500 m. However, in an area with term challenges in the access network existing infrastructure must be shortened, 2000 m average loop length, 13 Mbit/s will most likely be solved by further typically by replacing parts of the current may be provided to only 40 % of the cus- utilisation of the above mentioned set twisted pair cable between the local tomers. of technologies. In the long term, the exchange and the customer with fibre deployment of WDM may become a cost optic cable. Hence, hybrid fibre and Several architectures for a combined effective solution also in the access net- twisted pair copper architectures utilising transmission twisted pair cable and fibre- work for very high capacity demands. DSL systems are likely to constitute the optic cable are relevant, with SDH based future broadband access network for and ATM based solutions as the most operators with an existing twisted pair interesting ones. In the longer term a cost 7 Hybrid fibre and network. One major question in the long effective and flexible roll-out of a fibre twisted pair copper term perspective for access network based access network will require traffic operators, and PNOs in particular, then concentration deeper into the network architectures becomes: How deep should the fibre be than todays local exchange locations, ie. deployed? The cable infrastructure costs at the position of the fibre node. Concen- The potentially very high capacity vary significantly between network areas, tration capability is a very important dis- offered by optical fibre transmission sys- and hence quite different fibre penetra- tinction criterion between the different tems combined with the simultaneous tion levels are likely to be seen. FTTCab, fibre access architectures. In addition, the and prohibitive high costs of civil works FTTN, FTTC and FTTB are probably the solutions may vary with respect to fibre for deploying new fibre cable has spurred most relevant node configurations for topology. Topologies like ring, point-to- a significant interest in particular in resi- these hybrid solutions, and they reflect point and passive optical networks dential architectures based on fibre sys- different degrees of fibre penetration in (PONs) are all relevant, as illustrated in tems in combination with less costly drop the access network. the figures. technologies. In addition to the maximum transmission Today SDH systems are installed in large Telecommunication over a passive opti- distances of the DSL systems, the actual numbers in the transport network, often cal network (TPON) was the first genera- distribution of copper loop lengths within implemented in ring structures. SDH is tion fibre in the loop (FITL) systems in the exchange area is of importance and probably the technology with the highest terms of a hybrid fibre and twisted pair may influence the upgrade strategy. At a maturity and availability also when it copper architecture. These systems utilise strategic level, the average copper loop comes to fibreoptic systems for the a shared fibre infrastructure to deliver a length within exchange areas is likely to access network. An SDH architecture range of services to a set of customers. be a crucial parameter. One key question for the access network may be based on TPON is used to support telephony, is: What average copper loop lengths a ring topology and add and drop multi- ISDN basic access (BA) and ISDN pri- within exchange areas in European coun- plexers (ADMs) located in the optical mary rate access (PRA) and leased lines, tries will be required in order to offer node. Alternatively, introduction of SDH either in FTTCab, FTTN, FTTC, FTTB certain capacities? For typical statistical technology in the access network may be or FTTH configurations. But the concept copper loop length distributions within based on point-to-point transmission and of TPON was conceived prior to DSL exchange areas in some European coun- line terminals located in the optical node. technology becoming widely available, tries, the cumulative distribution typi- However, in these network architectures and as such the TPON is mainly consid- cally fits a linear function up to the 90 the traffic is not concentrated in the opti- ered to be tailored for existing services. percentile, with the average as 50 per- cal node. This is one of the major reasons Usually the TPON is deployed in devel- centile. This is shown in Figure 6. Typi- why conventional SDH technology is not 78 Telektronikk 2/3.1999

80 considered as an obvious access network One key question related to the introduc- tation of the coaxial cable network into alternative in the medium to long term. tion of hybrid fibre and twisted pair cop- smaller segments (eg. with 500 homes per architectures is the degree of network passed per segment), and most often a Several concepts for traffic concentration integration in the primary network. In simultaneous upgrade of the D1 network in the optical node are relevant. An ATM Figure 9 there is no network integration with optical fibre cables. This is called concentrator may be installed in the in the primary access network, with nar- HFC technology (HFC: hybrid fibre and node, including SDH line terminals for rowband services and broadband services coaxial cable network) [25, 26]. The size point-to-point transmission between the being transmitted in parallel on twisted of the coaxial cable segments in terms of node and the exchange. In a typical SDH pair cables and fibre optic cables respec- homes passed, and accordingly the fibre PON architecture the optical network tively. Figure 10 shows a BPON archi- penetration is a crucial question related unit (ONU) contains an ATM concentra- tecture in which the twisted pair cables in to HFC upgrades. tor, whilst a passive optical network is the primary access network are replaced used for the fibre optic transmission. by fibreoptic cables, and the narrowband The cable television industry has been SDH PON may be regarded as a further services and broadband services are trans- deploying HFC technology for a decade development of current local loop or mitted on separate fibre optical cables in now, mainly to support the broadcast access multiplexers, for instance in order the same primary network ducts. video business. HFC networks support to extend the capacity in SDH based a mix of analogue and digital channels DLC (digital loop carrier) systems to Figure 11 depicts a fully integrated by the use of frequency division multi- include DSL modems. Figure 7 illus- BPON architecture for broadband ser- plexing (FDM) techniques. The HFC net- trates an SDH PON architecture for vices and narrowband services, in which works currently being installed often has broadband services and narrowband the twisted pair cables in the primary active components with an available services, with a 155 Mbit/s transmission access network are replaced by fibreoptic bandwidth of 750 MHz. For non-active capacity between optical line multiplexer cables, and all services are integrated on components the available bandwidth is (OLM) and ONU. the same fibre cable system in the pri- in the range of 1 GHz [25]. mary network. A third alternative is to establish a ring Figure 12 shows an HFC architecture structure with concentration capabilities Architectures with a combination of opti- designed for the combined use of coaxial in the optical node, ie. the ONU. Figure 8 cal fibre and twisted pair cables are based cable modems and fibreoptic transmis- shows an ATM architecture for the on technology which currently is rela- sion systems. The fibre nodes are con- access network based on a ring topology, tively mature, as for instance SDH. Ring nected to the CATV head-end via fibre with some optional interfaces indicated. topology and point-to-point SDH sys- trunks equipped with linear lasers. Often Today such systems are often equipped tems are already well-proven technolo- the head-ends are interconnected by the with cards supporting interfaces such as gies and have been available for several use of SDH or ATM based core net- synchronous 2 Mbit/s PDH, n * 64 kbit/s years. The BPON technology combined works. HFC solutions are currently being and ATM, from 2 Mbit/s and up to 155 with VDSL modems is less mature, but installed in significant numbers, and the Mbit/s. the FSAN initiative is a driver for the technology is thus considered to be rela- standardisation of this alternative in the tively mature. In HFC systems the D2- Broadband transport of ATM on a pas- ITU [22]. Several field trials have been and D3 network technology is the same sive optical network (BPON) is currently carried out, some systems are already as the one described in the chapter on the architecture being standardised [22], operational in Japan, and several vendors coaxial cable modems. mainly based on the work carried out by plan to have FSAN and ITU compliant the FSAN initiative [23, 24]. The concept systems commercially available in 1999. of BPON (ATM based broadband passive 9 Transmission on optical network) is already well-known as 8 Hybrid fibre power line cables systems that typically offer 622 Mbit/s or 155 Mbit/s downstream capacity and 155 and coaxial cable Power line communications refers to the Mbit/s upstream capacity [23]. The access transmission of data over the low voltage is shared between 16 or 32 optical net- modem architectures electricity distribution network. The work units (ONUs) with an inherent sta- installed electricity distribution network Similar to upgrading the twisted pair tistical multiplexing capability. The opti- constitutes a significant asset with its network with DSL modems, upgrading cal line terminal (OLT) terminates the near-ubiquitous customer coverage, and the coaxial cable network with cable network at the exchange side. A typical offering telecommunications services modems for high capacity demands, will BPON architecture is shown in Figure 9. over this network may make the very require the access network architectures BPON is combined with for instance strong access network competitors to the to be based on fibre systems in the upper VDSL modem for the transmission of incumbent operators. Power line commu- part of the network in combination with higher capacities on the existing twisted nications are described in more detail in coaxial cable modems in the distribution pair the last few metres towards the cus- [4]. part of the network. As penetration of tomer. Figures 9 11 show typical archi- broadband services increases and the tectures considered possible in the long There are in principle two main classes capacity demand increases, the shared term. Figure 9 illustrates a BPON archi- of systems for transmission on power resources of the coaxial cable network tecture for broadband services in parallel line cables: power line telecommunica- segment must be shared by fewer cus- with a twisted pair network for narrow- tions (PLT) and power line communica- tomers if a comparable level of service band services between the local exchange tions (PLC). The former includes the shall be retained. This implies a segmen- and the customer. most recently developed systems. Two- Telektronikk 2/3.1999 79

81 100 line telecommunication technologies use frequencies in the range from 1 MHz to 30 MHz, which is a challenge in terms of Coverage of customers in area (%) 80 Average loop electromagnetic compatibility [4]. length in area: 2000 m Field trials have been running since 60 1300 m 1992/93, when Norweb connected 25 800 m households to their first trial network 500 m [27]. However, the technology is still 40 rather immature and there are no pub- 300 m lished standards for PLT. Only a limited set of vendors now have commercial sys- 20 tems available, or are targeting a release of commercial systems by 1999, includ- 0 ing Ascom AG, West End and Norweb Communications. Power line modems as 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 an access network solution will probably 52 26 13 10 8 6 2 not be implemented on a large scale in Copper loop length (m) and corresponding available bitrate (Mbit/s) the short term, but within a period of between three and five years the technol- ogy is likely to become an interesting Figure 6 Cumulative copper loop length distribution alternative for customers requesting capacities up to say 300 kbit/s [28]. 10 Migration and evolutionary paths way PLC systems were developed in the In a power line telecommunications sys- Given the previously described large late eighties with transmission capacities tem a power line modem is installed at number of available access network in the range of kbit/s, and has since then the low voltage transformer, which can architecture alternatives, the challenge been applied for purposes such as remote serve from some few homes in rural for the operator is two-fold: the target power meter reading and network load areas to some hundred homes in urban architecture for future broadband access management [27]. Currently there is a areas. A data transmission capacity of delivery must be determined, and significant interest in the PLT systems in about 1 Mbit/s is shared between the cus- equally important the migration or evo- particular, since these may be the means tomers connected to the low voltage lutionary path(s) toward this target archi- by which the power utility companies transformer. An additional modem is tecture must be selected. Thus, access may enter the Internet Access market. installed at the customer premises. Power network migration towards broadband NT1 LEX ODF LT SDH-PON ISDN POTS OLM 1:16 ISDN 155 Mb/s POTS VDSL ATM VDSL switch 155 Mb/s Local Primary Secondary Customer exchange flexibility point (PF) flexibility point (SF) premises

82 155 Mb/s NT1 LEX 155 Mb/s Ringnode ISDN POTS Masternode ISDN POTS VDSL ATM VDSL switch Local Primary Secondary Customer exchange flexibility point (PF) flexibility point (SF) premises

83 MDF RSS/ NT1 LEX RSU ODF ISDN 2-8 Mb/s OLT OLT BOLT 1:16 ONU POTS 622 Mb/s VDSL ATM switch VDSL VDSL 155 Mb/s VDSL Local Primary Secondary Customer exchange flexibility point (PF) flexibility point (SF) premises

84 ODF ODF D2- amplifiers CATV EO OE OE 300-860 MHz Cable Router modem D0 D1 D2 D3 network network network network Customer Headed D1 Hub D3 premises Figure 12 HFC architecture designed for the combined use of coaxial cable modems and fibreoptic transmission systems the fibre node (ADSL + fibre node Fibre node + LMDS with twisted pair RSX is installed in the fibre node for RSX); return: The network architecture Fibre POTS and ISDN in parallel with optical node + LMDS with twisted pair return network units for transmission of broad- Two parallel fibreoptic transmission is a first generation broadband wireless band services with VDSL between the systems in the primary access network, upgrade based on fibreoptic nodes and node and the customer premises. one for feeding an RSX and one for LMDS base stations. The traffic from the the broadband services with VDSL fibre node to the customer premises is Fibre node VDSL: The network archi- drop (Fibre node RSX + VDSL); transmitted via radio, whereas the twisted tecture Fibre node VDSL consists of a Fibreoptic systems in combination pair network is used for the return path. similar establishment of a fibre node with VDSL drop (Fibre node + infrastructure, but with a fully integrated VDSL). Fibre node + LMDS: In the network fibre optic transmission system in the pri- architecture Fibre node + LMDS fibre- mary access network for POTS, ISDN ADSL: In the network architecture ADSL optic nodes and LMDS base stations are and broadband services. In the secondary the broadband customers are served with established. Both the downstream and the network VDSL modems, POTS modems ADSL modems, and POTS and ISDN are upstream traffic from the fibre node to and ISDN modems are used in addition provided on the existing twisted pair the customer premises is transmitted via to combined modems for narrowband infrastructure. An ADSL connection is radio. and broadband services. assumed to be able to transmit capacities in the 2 Mbit/s to 8 Mbit/s range down- ADSL + fibre node RSX: This network In addition to the decision on the target stream, and up to 640 kbit/s upstream, architecture is based on the establishment network itself, the operator must decide with a capacity dependent reach of be- of a fibre node infrastructure, in which an when and how the architecture is to be tween 2 km and 4 km. access network multiplexer and ADSL deployed. Time combinations of the modems are installed in the fibre node in mentioned architectures are likely to be IMUX-node + VDSL: IMUX-node + addition to an RSX (remote subscriber used. This is illustrated in Figure 13, VDSL implies establishment of transmis- system or remote subscriber unit). ADSL where the initial introduction of the sion nodes for transmission of aggregate modems are installed at the customer IMUX node + VDSL architecture allevi- capacity between these local access premises and POTS and ISDN are pro- ates the need for an immediate introduc- nodes and the local exchange. Inverse vided on the existing twisted pair infra- tion of fibreoptic transmission systems in multiplexing (IMUX) is used to establish structure between the customer premises the primary access network. Some years the high capacity connection on the exist- and the RSX in the fibre node. later fibreoptic cables are deployed to the ing twisted pair infrastructure between node where the inverse multiplexer is the local access node and the local Fibre node RSX + VDSL: In the net- located, and the twisted pair cables in the exchange. By using VDSL modems work architecture Fibre node RSX + primary access network may be discon- between the node and the customer VDSL a fibre node infrastructure is estab- nected. In the secondary network the premises capacities in the range of lished. Two parallel fibreoptic transmis- same set of VDSL modems may be used 26 Mbit/s to 52 Mbit/s may be provided. sion systems are used in the primary throughout the period under considera- The architecture is primarily an alterna- access network, one for feeding the RSX tion. tive to architectures with fibreoptic trans- and one for the broadband services. The mission in the primary access network. latter may for instance be a broadband Starting from the large variety of migra- passive optical network system. The tion paths in Figure 13, we have selected Telektronikk 2/3.1999 83

85 1999 2008 term a fibre node infrastructure is estab- Short term Medium/long term lished, with an access multiplexer and ADSL modems installed in the local exchange in addition to an RSX. In the long term the ADSL equipment is gradu- Twisted pair TADSL TADSL ally removed and replaced by VDSL network modems. During the period there is no integration in the primary access network of narrowband and broadband services IMUX node+ IMUX node+ on one fibreoptic transmission system. VDSL VDSL Migration path 4, Twisted pair ISDN Fibre node RSX + VDSL. This migration path results in the same target network as Fibre node + Fibre node + Migration path 3, but does not in oppo- LMDS tw. pair LMDS tw. pair sition to Migration path 3 have ADSL return return + Fibre node RSX as an interim solution. Migration path 5, Twisted pair ISDN Fibre node Fibre node ADSL + Fibre node RSX Fibre node + LMDS + LMDS VDSL. In the short and medium term the same migration path as outlined in Migration path 3 is followed. However, ADSL + Fibre ADSL + Fibre in the long term a fully integrated net- node RSX node RSX work for narrowband and broadband services is implemented. Migration path 6, Twisted pair ISDN Fibre node Fibre node Fibre node RSX + VDSL Fibre node RSX+VDSL RSX+VDSL VDSL. In the short and medium term the same migration path as outlined in Mi- gration path 4 is followed. In the long term though, a fully integrated network Fibre node Fibre node Figure 13 Twisted pair for narrowband and broadband services VDSL VDSL network migration paths is implemented. Migration path 7, Twisted pair ISDN Fibre node VDSL. This migration path in the long term results in a target architec- ture as Migration path 5 and Migration path 6, without the intermediate architec- ture solution. A fully integrated network for narrowband and broadband services some of the ones considered to be most and one for the broadband services. The is implemented initially. relevant, and numbered them, as shown latter may for instance be a broadband in Figure 14. passive optical network system. In the Migration path 8, Twisted pair ISDN fibre node an RSX is installed for POTS IMUX-node + VDSL Fibre node Migration path 1, Twisted pair ISDN and ISDN together with optical network VDSL. In the short and medium term ADSL Fibre node RSX + VDSL. In the units for transmission of broadband ser- the network architecture IMUX-node + short and medium term the broadband vices with VDSL between the node and VDSL is implemented, omitting installa- customers are served by ADSL modems, the customer premises. tion of fibreoptic cable in primary net- and POTS and ISDN are provided on the work in the initial phase. In the next existing twisted pair infrastructure be- Migration path 2, Twisted pair ISDN phase the twisted pair cables in the pri- tween the local exchange and the cus- ADSL Fibre node VDSL. This alterna- mary access network are disconnected, tomer. An access multiplexer and ADSL tive has the similar evolution as Migra- and fibreoptic cables are deployed to the modems are installed in the local tion path 1 initially. However, in the long location of the inverse multiplexer. The exchange. In the long term the ADSL term the narrowband and broadband ser- same set of VDSL modems may be used equipment is gradually removed from the vices are fully integrated on the same pri- in the secondary network throughout the network, and a fibre node infrastructure mary access network transmission sys- period. is established. However, narrowband and tem. broadband services are not fully inte- Migration path 9, Twisted pair ISDN grated. In the primary access network Migration path 3, Twisted pair ISDN Fibre node + LMDS with twisted pair two parallel fibreoptic transmission sys- ADSL + Fibre node RSX Fibre node return Fibre node + LMDS. In the tems are used, one for feeding the RSX RSX + VDSL. In the short and medium short and medium term LMDS base sta- 84 Telektronikk 2/3.1999

86 tions are established, fed with fibreoptic 1999 2008 cables. The traffic from the fibre node to Short term Medium/long term the customer premises is transmitted via radio, whereas the twisted pair network initially is used for the return path. In the long term all traffic is integrated in the Twisted pair TADSL LMDS network. network Migration path 10, Twisted pair ISDN Fibre node + LMDS. This migration IMUX node+ path in the long term results in a target VDSL architecture as Migration path 9 without the intermediate solution Fibre node + LMDS with twisted pair return. Initially all traffic is integrated in the LMDS net- Fibre node + work. LMDS tw. pair 9 return Figure 15 illustrates some migration paths for the coaxial cable network. Three different solutions are indicated Fibre node Fibre node + LMDS 10 + LMDS for the end of the period under study: Cable modem, twisted pair return; 1 Cable modem, return amplifiers; 8 ADSL + Fibre Fibre node cable modem (HFC). node RSX 2 3 Cable modem, twisted pair return: In the network architecture Cable modem, 5 twisted pair return coaxial cable modems Fibre node Fibre node RSX+VDSL 4 RSX+VDSL are installed at the customer premises. The existing twisted pair network is used 6 for the return channel. A transmission capacity of between 10 30 Mbit/s is Figure 14 Selected migration Fibre node Fibre node shared among some thousand cable sub- paths for the twisted pair net- VDSL 7 VDSL scribers. There is no need for installation work of return amplifiers in the coaxial cable network. At the hub the return channel is terminated in a router port for further transport of the broadband services. Cable modem, return amplifiers: The network architecture Cable modem, return amplifiers is similarly based on installation of cable modems at the cus- tomer premises. In addition return ampli- fiers are installed in the coaxial cable dis- tribution network. 1999 2008 Fibre node cable modem (HFC): The Short term Medium/long term network architecture Fibre node cable modem (HFC) is similar to the Cable modem, return amplifiers architecture Cable modem, Cable modem, except that the existing coaxial cable net- tw. pair return tw. pair return work is segmented into smaller coaxial cable branches. Parts of the D1-network and eventually the D2-network is up- Coaxial cable Cable Cable graded with fibreoptic cable and trans- network modem return modem return mission systems. This is known as HFC amplifiers amplifiers technology (HFC: hybrid fibre and coax- ial cable network). In such a network a Figure 15 Coaxial transmission capacity of between 10 Fibre node, Fibre node, cable network evolu- 30 Mbit/s is typically shared among 500 cable modem cable modem tionary paths cable subscribers, as opposed to the two (HFC) (HFC) Telektronikk 2/3.1999 85

87 point for limiting the choices and select- 1999 2008 Short term Medium/long term ing the appropriate architecture for enter- ing the broadband service delivery mar- ket. Coaxial cable Cable Cable modem return 1 modem return network amplifiers amplifiers 12 References 2 1 Faulkner, D W. The passive optical network (PON). Telektronikk, 95 Figure 16 Selected evo- Fibre node, Fibre node, (2/3), 113118, 1999. (This issue.) lutionary paths for the cable modem 3 cable modem coaxial cable network (HFC) (HFC) 2 Koonen, T. Flexible wavelength mul- tiplexing techniques for broadband fibre access networks. Telektronikk, 95 (2/3), 119128, 1999. (This issue.) 3 Edvardsen, E. Inverse multiplexing. Telektronikk, 95 (2/3), 129133, 1999. (This issue.) 4 Collins, D et al. A. Power line com- former solutions, in which some thousand ing coaxial cable network is segmented munications. Telektronikk, 95 (2/3), cable subscribers shared this capacity. into smaller coaxial cable branches and 134137, 1999. (This issue.) an HFC network is established. Starting from the large variety of migra- 5 Quayle, A, Stern, J. A. Achieving tion paths in Figure 15 once again we The above presented migration paths are global consensus on the strategic have selected some of the ones consid- considered to be among the most inter- broadband access network : the full ered to be most relevant, and numbered esting ones for the evolution of the services access network initiative. them, as depicted in Figure 16. Note that twisted pair networks and coaxial cable Telektronikk, 95 (2/3), 166175, the architecture Cable modem, twisted networks, and they illustrate that in the 1999. (This issue.) pair return neither is considered as a evaluation of access network migration viable long term solution nor as an inter- towards broadband both the existing net- 6 Okada, K. Towards broadband access mediate alternative. work, the target architecture, and the cor- in Japan : ATM access for Mega responding intermediate infrastructure Media services. Telektronikk, 95 Migration path 1, Coaxial cable net- changes have to be carefully examined. (2/3), 184190, 1999. (This issue.) work Cable modem, return amplifiers. In the short and medium term coaxial cable modems are installed at the cus- 11 Summary 7 Ferrero, U. Towards broadband Access in Europe : the view from tomer premises, and return amplifiers are This paper has briefly described the main EURESCOM. Telektronikk, 95 (2/3), introduced in the coaxial cable network. broadband access network architectures 176183, 1999. (This issue.) currently being considered. The key Migration path 2, Coaxial cable net- parameters that differentiate the various 8 Orth, B. Broadband Access Net- work Cable modem, return amplifiers access network solutions have been dis- works: Network Architectures and Fibre node cable modem (HFC). In cussed, as well as some of the key issues Service Aspects. In: Proc. Access the short and medium term the same faced by the access network infrastruc- Networks Mini-Conference, Globe- migration path as outlined in Migration ture providers in the migration of the com 1998. Sydney, Australia, 812 path 1 is followed. However, in the long present network to broadband. Finally, November, 1998, 5762. term the existing coaxial cable network some of the likely migration paths from is segmented into smaller coaxial cable the existing wireline access networks 9 Kwok, T C. Residential Broadband branches and parts of the D1-network towards broadband have been outlined. Internet Services and Applications (and eventually the D2-network) is Requirements. IEEE Communica- upgraded with fibreoptic cable and trans- In conclusion, no single network archi- tions Magazine, 35 (6), 7683, 1997. mission systems. In the long term coaxial tecture seems to be the obvious choice cable segments with 500 cable subscribers at present. Instead, boundary conditions 10 LeBel, P, Oliver, R. Residential are targeted. given by the existing infrastructure and broadband architectures strategies customer segment characteristics may and economics. In: Proc. The 7th Migration path 3, Coaxial cable net- possibly require simultaneous roll-out International IEEE workshop on work Fibre node cable modem (HFC). of several technology alternatives. How- Optical Access Networks (OAN 95). This migration path in the long term ever, the operator in terms of his knowl- Nuremberg, Germany, 2427 Sep, results in a target architecture as Migra- edge of existing network, his network 1995, 4.4-1 4.4-10. tion path 2 without the intermediate solu- area characteristics, his customer base tion Cable modem, return amplifiers. In- and business target, has a good starting stead, already in the short term the exist- 86 Telektronikk 2/3.1999

88 11 Stordahl, K, Rand, L. Long term 23 van de Voorde, I, van der Plas, G. forecasts for broadband demand. Full service optical access networks : Telektronikk, 95 (2/3), 3444, 1999. ATM transport on Passive Optical (This issue.) Networks. IEEE Communications Magazine, 35 (4), 7075, 1997. 12 Pugh, W, Boyer, G. Broadband access : comparing alternatives. 24 Faulkner, D et al. The Full Services IEEE Communications Magazine, 33 Access Networks Initiative. IEEE (8), 3445, 1995. Communications Magazine, 35 (4), 5868, 1997. 13 Kopf, D. Internet Race xDSL vs. Cable Modems. Americas Network, 25 Paff, A. Hybrid fiber/coax in the 1 August 1996. public telecommunications infra- structure. IEEE Communications 14 Kelly, D. Digital Subscriber Lines magazine, 13 (4), 1995. (DSLs) : expanding the use of the copper loop plant. In: Proc. 26 van Driel, C L et al. The (R)evolu- Eurescom Workshop on Optical tion of access networks for the Infor- Multiservice Access Networks mation Superhighway. IEEE Com- (WOMAN). Turin, Italy, 2324 munications Magazine, 35 (6), September 1996. 104112, 1997. 15 Maxwell, K. Asymmetric Digital 27 Konstali, T, Sinkerud, K. Accessing Subscriber Line : interim technology the utilisation of the power network for the next forty years. IEEE Com- for telecommunication and solutions munications Magazine, 34 (10), for home communication and auto- 100106, 1996. mation. Kjeller, Telenor Research and Development, 1998. (R&D note 16 ADSL Forum (1999, February 18) N 44/98.) [online]. URL: http://www.adsl. com/general_tutorial.html 28 Nortel (1999, February 18) [online]. URL: 17 Chow, J S, Jacobsen, K. Very-high- broadband/powerline speed Digital Subscriber Line (VDSL) workshop. Globecom 1998. 29 Ims, L A et al. Access 2005 access Sydney, Australia, 812 Nov 1998. network architectures. Kjeller, Telenor Research and Development, 18 EURESCOM. Techno-economic 1998. (R&D report R 44/98.) evaluation of multiservice access net- work architectures, scenarios and business cases. Heidelberg, Ger- many, 1998. (EURESCOM project P614, Deliverable 10.) 19 IEEE. Cable-TV access method and physical layer specification. 1998. (IEEE Project 802.14/a Draft 3 Revi- sion 1. http://www.walkingdog. com/catv/.) 20 IETF. Logical IP Subnetworks over IEEE 802.14 Services. 1998. (IETF Internet Draft draft-ietf-ipcdn-ipover- 802d14-01.txt.) Leif Aarthun Ims (33) is Senior Research Scientist in the Strategic Network Development group of Telenor R&D. 21 Cook, A, Stern, J. Optical Fiber Since 1994 he has been working on access network Access Perspectives Towards the development in several RACE, ACTS and EURESCOM 21st Century. IEEE Communications projects, and recently as project manager for long term Magazine, 32 (2), 7886, 1994. access network development in Telenor. He is now project leader of the EURESCOM project P901. He is editor of the 22 ITU. Name is missing here. Geneva, book Broadband Access Networks published last year 1997. (ITU-T, Draft recommendation and has published more than 50 papers in international G.983.1.) journals and conferences. email: [email protected] Telektronikk 2/3.1999 87

89 Fixed broadband wireless access HARALD LOKTU AND ERWAN BIGAN 1 Introduction Broadband wireless local loop is also (POTS, ISDN, fractional E1 leased lines) becoming a reality through another which can be delivered through multi- Wireless local loop (WLL) emerged approach. Broadband symmetrical point- plexers. When connecting businesses or some years ago as an attractive alterna- to-multipoint (PMP) systems are opti- mobile base stations, only a fraction of tive to copper for POTS services. The mised for n * 64 kbit/s services (circuit- the E1 capacity may be required. Due to related advantages were fast deployment, switched POTS and ISDN, fractional E1 limited availability of spectrum, these because civil work is minimised and pro- leased lines) to small- and medium-size radio systems should eventually offer gressive investments, as radio equipment businesses (SME). A number of fre- bandwidth-on-demand, although some is only installed for those prospects quency bands may be used for broadband initial versions only support fixed band- which actually do subscribe to the ser- symmetrical PMP depending on national width allocation. These systems are vice. Narrowband WLL is now a mature regulations and industrial development typically available in the frequency bands technology with several million POTS being 3.5 GHz, 10.5 GHz, 26 GHz, or 3.5 GHz, 10.5 GHz, 26 GHz, or 28 GHz. connections world-wide. Different fre- 28 GHz. The distinction between LMDS The lowest frequency bands yield the quency bands are used in different coun- and broadband PMP is not always clear- longest range (eg. 1020 km at 3.5 GHz) tries, ranging from VHF/UHF up to 3.5 cut. at the expense of limited capacity (total GHz. WLL has mainly been used in aggregated capacity in the order of 100 developing countries and eastern Europe Bi-directional LMDS or broadband PMP Mbit/s). On the other hand, the highest to improve teledensity. It has also been systems are being deployed by several frequency bands offer the highest capac- used by a few operators in the UK, vendors for the purpose of conducting ity (total aggregated capacity in the order Atlantic being one of them. field trials. The number of connected of 1 Gbit/s), at the expense of reduced customers world-wide probably lies in range (eg. a few km at 26 or 28 GHz). Broadband wireless local loop offers the the range of hundreds. According to the same advantages, but still lacks the matu- consulting firm Ovum, the broadband 2.1.3 Local Multipoint rity of its narrowband counterpart. Most wireless market is expected to reach Distribution System broadband wireless local loop systems todays narrowband wireless market deployed to date are one-way. They are size of a few millions by 2002. The acronym LMDS (Local Multipoint Multiservice Multichannel Distribution Distribution System) designates bi-direc- Systems (MMDS), which offer a wireless tional asymmetrical point-to-multipoint means to carry CATV services. MMDS 2 State-of-the-art of fixed systems able to deliver multi-services to systems are available off-the-shelf, and broadband wireless either residential (TV, Internet access, several million homes are connected POTS) or business customers (leased through MMDS world-wide. Most of access lines, ISDN, Internet access etc.). The these systems operate in the 2.5 GHz fre- digital telephony, data and video signals quency band. New systems operating in 2.1 Overview of broadband wire- are initiated at central offices, head-ends higher frequency bands (28 GHz in the less access technologies or satellites. The signals are combined USA, 40 GHz in Europe) have been and transmitted, using optical fibre or 2.1.1 Point-to-point introduced to increase the number of TV microwave links, to multiple LMDS channels. The acronym MVDS (Multi- Point-to-point radio relay systems have nodes placed on towers or on top of high channel Video Distribution System) is been readily available off-the-shelf for a buildings. The radio signals are distri- used to designate TV distribution in the number of years from a large number of buted to households and businesses 40 GHz band. vendors. Initially deployed in core net- where they are received using a rooftop works (medium to high capacity), they antenna. An up/down-converter is The use of higher frequency bands also are now mainly used in access networks mounted close to the antenna. A set-top permits the introduction of a radio return (small capacity, ie. n * 2 Mbit/s), either box or multiplexer is installed indoors channel. The typical range is reduced to connect business customers or for to connect the various customer terminal from a few tens of kilometres for MMDS mobile base station backhaul. These equipment. A coaxial cable connects the at 2.5 GHz, down to a few kilometres for compact systems operate in the fixed outdoor and indoor units. The down- LMDS (Local Multipoint Distribution service frequency bands between 7 and stream bit rate capability is up to 25 50 System) at 28 GHz. LMDS systems were 38 GHz. Their range typically lies be- Mbit/s per customer and the upstream bit first designed to offer packet-switched tween 2 and 20 km depending on the rate is up to several Mbit/s per customer. multimedia services to residential cus- frequency band, the required availability, Operation at high frequency bands (eg. tomers (fast Internet access). However, and the rain zone. There is no doubt that 28 GHz) is necessary to handle such they have the potential for becoming point-to-point radio relay is the most large bit rates. The typical range is a truly multiservice (POTS, ISDN, leased mature broadband wireless access tech- few kilometres. lines, TV, fast Internet access, ATM) nology. because bandwidths in excess of 1 GHz The traffic asymmetry depends on the are available in the 28 or 40 GHz band. service mix to be delivered This is the 2.1.2 Point-to-multipoint In particular, the support of leased lines reason why LMDS systems have been has become a priority in order to satisfy Symmetrical broadband point-to-multi- designed to operate in frequency bands the needs of alternative operators in point (PMP) systems have been designed where great flexibility is left to the opera- developed countries. to connect either small- and medium-size tor regarding which spectrum capacity business customers or mobile base sta- should be allocated to the down- and tions, at E1 rates or N * 64 kbit/s services uplink. This is the case of the 28 GHz 88 Telektronikk 2/3.1999

90 frequency band, which was auctioned areas. Because of the limited available limited to 1 2 Mbit/s gross bit rate to be by the FCC in the USA early 1998. In bandwidth, UMTS is not expected to shared. Furthermore, these systems oper- Europe, LMDS systems may be deployed handle large numbers of broadband cus- ate in unlicensed frequency bands (eg. in the 26, 28 or 40 GHz band. The 26 tomers. On the other hand, broadband 2.4 GHz band) for which no quality of GHz frequency band is well suited for PMP or LMDS systems will be increas- service (QoS) can be guaranteed because symmetrical applications, whereas the 40 ingly used in conjunction with point-to- the overall bit rate or transmission delay GHz band is dedicated to broadcast point radio relay systems to connect is affected by interference. applications.1) The European status of mobile base stations (second generation the 28 GHz band is still unclear because mobile systems, UMTS) to core net- the old CEPT frequency band plan for works. 2.2 Radio specific issues symmetrical applications will need to be affecting the choice of reworked to take into account sharing technology 2.1.6 Terrestrial broadcasting of the band with satellite systems. 2.2.1 Propagation Terrestrial broadcasting being MMDS, digital terrestrial TV in the VHF/UHF Most broadband wireless access systems 2.1.4 Broadband satellite systems band may also play a role in broadband operate at frequencies above 10 GHz for Geo-stationary Earth Orbit (GEO) satel- access networks. Some MMDS networks which clear line-of-sight (LOS) is re- lite systems are widely used for broad- are already used to offer fast Internet quired between the transmitter and the cast as well as transmission applications. access using POTS return channel. receiver. This reduces the coverage pro- They can also be used to offer fast Inter- Future digital terrestrial TV networks vided by broadband systems for terres- net access, using POTS return channel. will offer the same possibility, but to a trial wireless access. The actual coverage However, they have only found limited smaller fraction of customers because depends on the base station antenna use in access networks. Low Earth Orbit of the larger cell size. There is on-going height, the cell size, and the environment. (LEO) satellite systems have been pro- research towards low bit-rate radio return As a crude indication, the typical cover- posed in order to reduce the large propa- channels for these systems, which could age LOS probability is between 30 and gation path loss and delay. These propos- alleviate the need for POTS return in 70 %. Accurate planning tools are needed als (Teledesic, Skybridge) target asym- case such systems are used to offer fast to identify those prospective customers metrical broadband wireless access at bit Internet access to residential customers that can be connected through radio. If rates up to several Mbit/s per customer. in the future. 100 % coverage is required, alternative access techniques must be used in con- Compared to broadband terrestrial sys- junction with radio. 2.1.7 Radio Local Area Networks tems (broadband PMP, LMDS), satellite (RLANs) systems are best suited to handle very The second propagation issue is rain low customer densities because of the RLANs are used to avoid cabling prob- attenuation, which becomes significant much larger spot size. These systems lems and/or to provide limited mobility. above 10 GHz and reduces the range. could thus be used to offer broadband Off-the-shelf equipment allows bit rates The range then depends on the required access to the upper tier customer cate- up to 1 2 Mbit/s and future generations availability and the climatic zone. Range gory as for instance international busi- target 10 20 Mbit/s. Although RLANs reduction usually increases deployment ness travellers, international news corre- are used mainly indoors within business costs, except when base stations are uti- spondents or to selected business cus- premises, they can also be used lised at full capacity. tomers. These systems are still at the marginally in access networks. A few design stage, with service opening being vendors propose license-free point-to- 2.2.2 Spectrum availability planned in a few years. point radio relay equipment based on RLAN technology and they also envisage Spectrum is a prerequisite for successful point-to-multipoint systems packaged for operation of any radio based telecommu- 2.1.5 Universal Mobile outdoor use. The bit rate capability is still nication service. The large bandwidths Telephone System (UMTS) Third generation mobile systems (UMTS) will handle voice as well as data traffic. Micro-cells will provide bit rates up to 2 Mbit/s in hot spots. Up to a few hundred bit/s will be offered in most Table 1 Spectrum in Europe Frequency band Total available bandwidth 3.40 3.60 GHz 2 * 90 MHz 1) The UK has proposed a frequency band plan permitting highly asym- 10.15 10.65 GHz 2 * 150 MHz metrical applications in the 40.5 24.50 26.50 GHz 2 * 1 GHz 42.5 GHz, and CEPT is currently working towards extending this band 27.50 29.50 GHz 2 * 1 GHz up to 43.5 GHz which would give this 40.5 42.5 GHz (+42.5 43.5) 2 3 GHz band a truly multiservice capability. Telektronikk 2/3.1999 89

91 Table 2 Wireless deployment scenarios highly competitive environment from a technological point of view. Further- Residential SME (2 Mbit/s) more, with such inherent service flexi- bility, both the residential and business Alternative wired A/VDSL, HFC, HDSL, FTTB FTTB customers may be served with the same techniques FTTC/H access technology platform. This de- velopment is for the moment also Urban/suburban LMDS PMP, LMDS PP strongly encouraged by the ongoing deregulation of the previous monopolist Rural MMDS, DVB-T PP, PMP PP markets for services. The highly dynamic and competitive future market for delivery of such ser- vices will require short time to market and ability to reconfigure the access net- work according to the service configura- required for broadband applications can relatively low density areas, but deploy- tion requested by the customers. Within only be found at high frequencies, most ment in lower frequency bands should this context, broadband wireless access often above 20 GHz. be preferred in that case (eg. 3.5 GHz). (BWA) systems will be one of the most Isolated customers may be best con- interesting candidates for delivery of true There is a number of candidate frequency nected using PP systems. MMDS or multi-media services combining fixed bands for broadband radio access. Table DVB-T may be used to offer fast Internet and broadcast services in a cost-efficient 1 lists those frequency bands that are access to residential customers, with way. most likely to be used for this application POTS return channel. The longer range in Europe, along with the total available of MMDS or DVB-T compared to bandwidth. LMDS makes them most favourable for 3.2 Wireless access networks rural areas. 3.2.1 Access network architectures Not all these frequency bands are avail- able for broadband access in every coun- Wireless access networks for fixed try. However, summing up the total 3 Evolution of fixed broad- broadband services are currently evolv- available bandwidth in any country usu- band wireless access ing along three major routes towards the ally yields several gigahertz of spectrum. future network. The far most developed Therefore, lack of spectrum for broad- 3.1 Evolution of broadband one so far seems to be the broadcasting band radio access applications is not a service requirements oriented approach, which in a European key issue. The most serious issue is the perspective is dominated by the DVB The future evolution of broadband access lack of harmonisation: although most group within EBU. Originally intended networks is characterised by a broad of these bands have been harmonised by to provide specifications for digital video range of very dynamic and rather com- CEPT, different bands are available in broadcasting, it has expanded the scope plex scenarios. Within the long term different countries. This may to some of its work to include bi-directional virtually all electronically based informa- extent inhibit cost reduction through vol- asymmetric access as well. Furthermore, tion requested by people and computers ume production. there is also a telecom-oriented route in at least developed countries, is pre- towards the future based on bi-directional dicted to be both accessed and processed symmetric access. P-P and P-MP systems 2.2.3 Deployment scenarios in a digital format. Together with the low are at the moment the most evident cost related to the storage and processing The following analysis is purely qualita- examples of this trend. In Europe the of digitally based information, this trend tive. A quantitative analysis should rely progress along this path is led by the provides the necessary framework for upon techno-economic evaluation. Such Broadband Radio Access Network cost-effective development of an increas- techno-economic analysis can be found (BRAN) project within ETSI. Finally, a ingly large variety of new services and in a different article of this journal issue. datacom oriented route is also present of service combinations. Consequently, the Table 2 outlines possible deployment which Wireless LAN (WLAN) systems traditional categorisation of services into scenarios for point-to-point (PP) radio based on IP presently is the most suc- telecom, datacom and broadcast is no relay, broadband symmetrical point-to- cessful example. longer appropriate and is gradually being multipoint (PMP), and LMDS. converged into a combined service sce- The reference model for interactive ser- nario often referred to as multi-media LMDS is the only radio technology that vices adopted by DVB is shown in Fig- services. can handle both business and residential ure 1. It proposes a broadband broadcast- customers in principle. However, ing medium in the direction towards the An immediate consequence of this de- answering the needs of alternative opera- user and in addition a bi-directional sym- velopment is that broadband access net- tors in developing countries, the current metric medium for the interactive part works have to be able to support a large generation of LMDS systems primarily with considerable lower capacity. The variety of service configurations. Hence, targets business customers. Besides, the reference model itself includes in prin- the different broadband access technolo- short range favours LMDS for dense ciple both wireless and wireline tech- gies must to a large extent be capable of urban or suburban areas. PMP may be nologies but clearly assumes that the offering the same services generating a used to connect business customers in access mode is asymmetric. An impor- 90 Telektronikk 2/3.1999

92 Broadcast Channel Broadcast Set top box (STB) (DVB Transmissions Systems) channel Broadcast Broadcasting Broadcast Delivery Network Broadcast Service Media Adapter Interface End Provider Module User Set Top Return Unit Interaction path user Interactive (STU) Interactive Interactive Interface Service Interaction Network Module Provider Network Adapter (can be Network external Interface to STB) UNIT (NIU) Interaction Channel Forward Interaction Interaction channels path network independent network dependent network independent Figure 1 DVB interactive reference model tant inherent feature of the reference The interactive network is according to network is interfaced to the TEs and the model is that it promotes implementation the DVB reference model provided as an local networks through adequate inter- of hybrid access networks using a com- overlaid network. This is appropriate if a working functions (IWF). bination of different media for service reduced service access network is as- delivery. Both wireless-wireless and sumed with for instance PSTN or ISDN Unlike the DVB reference model, the wireless-wireline configurations are en- as interaction networks. However, to be described reference model defines a visaged and the hybridisation could be able to establish a full service access flexible multipoint-to-multipoint (MP- both between the forward and return network in an efficient way, the forward MP) or mesh architecture combining media as well within each of the direc- broadcast and interaction network should point-to-point and point-to-multipoint tions themselves. in the future be merged into a single for- connections. This allows for dynamic ward network. Furthermore, dynamic routing functionality being implemented Due to the focus on broadcast services, switching or routing functionality must both at the RRs and APC as well as at the typical access architecture is a point- be provided. At present this is being done the radio terminations (RT) providing to-multipoint (p-mp) or star configura- by encapsulating IP or ATM traffic into support for both distributed and central- tion. The DVB standard prescribes a the MPEG-2 transport stream in the for- ised routing of traffic. As such it defines multiplexing scheme in the forward ward direction while pure IP or ATM is an architectural framework for imple- direction based on 188 byte time slots or employed in return direction. mentation of a re-configurable service cells being referred to as an MPEG-2 delivery mechanism. This framework is transport stream (TS). However, no A general reference model adopted by also believed to provide the necessary addressing or routing/switching scheme the ETSI BRAN project is shown in support for implementation of close to is devised and hence, such functionality Figure 2. Besides defining a set of inter- 100 % wireless coverage of the customer is often implemented in a centralised faces, it describes a general access net- locations in a given area. manner. The major networking operation work consisting of access points (AP) is performed by the MPEG-2 TS multi- and access terminations (AT). The ATs The BRAN model is basically estab- plexer and demultiplexer mainly adding are connected directly to one or more lished to specify delivery of telecom and and dropping time slots from the trans- access point transceivers (APT) or via datacom type services within the scope port stream in a static way. There is no one or more radio relays (RR), which of a switched full service access network multiplexing hierarchy defined for the together with the access point controller for broadband services. As such it will MPEG-2 TS which prevents a seamless (APC) constitutes an AP. Optionally, ter- support switching and multiplexing interconnection with standard core net- minal equipment (TE) may communicate based on asynchronous (ATM) and most works. directly via radio relays without passing likely a pure IP based scheme as well. through an external switch. The access For the time being the working assump- Telektronikk 2/3.1999 91

93 Terminals Local networks W.3 RT TE IWF W.1 W.2 B.1 IWF ISDN AP W.1 W.2 APT IWF ATM W.3 W.1 TE IWF APC RT/ RR W.2 W.3 B.2 IWF PSTN TE IWF B.1 W.1 W.1 APT W.2 IWF IP W.1 W.3 RT/ TE IWF RR W.2 B.1 IWF etc. AT W.1 W.1 W.3 TE IWF RT B.3 B.1 EMS W.1 RR TE W.3 IWF RT Hiperaccess network Element management B.1 system Figure 2 ETSI BRAN general reference model for fixed service broadband access networks Table 3 ACTS projects with broadband access trials or demonstrators tion in the specification work is ATM Max Channel Frequency Range which employs a 53 byte cell structure. PROJECT ACCESS ENVIRONM. The access mode is symmetric but does Bit Rate Band [km] [Mbit/s] [GHz] not preclude delivery of asymmetric ser- vices. It provides a switching hierarchy Up Down Up Down Up Down allowing an efficient interconnection between the core and access network. As FRANS-2 40 622 29 Fixed 0.5 Outdoor opposed to the DVB approach it offers inherent dynamic routing functionality. ATMmobil 155 38 Fixed 0.3 Outdoor The BRAN model only considers wire- FRANS-1 N*2 155 2.2 42 Fixed

94 The study revealed a trend towards There is a close relationship between or alternatively in a lower part of the higher capacity and higher frequency different features given in Table 3. Maxi- spectrum. Conversely, time division bands. Channel or information bit rates mum range of radio path length is traded duplex (TDD) will be employed for in the range 2 to 622 Mbit/s were tar- against range as well as increasing radio short-range indoor access. geted and spectrum in the range 2.5 to frequency both for indoor and outdoor 60 GHz were considered. Two major cat- environments. Hence, if the combination These emerging technologies may be egories were found: portable or nomadic of large capacity and range is envisaged, mapped onto both the BRAN and DVB access for indoor environments and fixed asymmetric network access must be reference model. The evolution towards a outdoor access. The vast majority of expected if cost-effective solutions shall full service network most likely implies projects had chosen ATM as preferred be provided. Consequently, to separate an ATM multiplexing scheme. To pro- multiplexing scheme and both solutions the forward and return connection a vide close interconnection to the ATM based on asymmetric and symmetric frequency division duplex (FDD) will core networks, the access maximum bit access are proposed. be used to provide long-range outdoor rates should comply with standard rates access, either in the same frequency band of 25.6, 51.8 or 155.5 Mbit/s. Due to the expected merger of telecom and broad- PAST fig.2 FUTURE PREVIOUS NOW SHORT-TERM MEDIUM-TERM LONG-TERM MOBILE UMTS UMTS + Enhanced Wireless WLAN IP NOMADIC IEEE 802.11 Enhanced WLAN WLAN WLAN IEEE 802.11 HIPERLAN FWA Wireless ATM WIRELESS HIPERACCESS Full Service P-P P-MP Access Network MVDS MWS LMDS FIXED MVDS Hybrid MVDS LMDS PSTN/ISDN xMDS Bi-directional MMDS MMDS Bi-directional MMDS Hybrid MMDS MMDS PSTN/ISDN WIRELINE PSTN/ ISDN Figure 3 Migration paths towards a wireless FSAN Telektronikk 2/3.1999 93

95 cast services, asymmetric access may be effort being conducted to standardise 3 Loktu, H et al. Status and future evo- the preferred solution in the short term. IP on the UMTS platform and in the lution of broadband radio access. In: In the long term symmetric access will medium term there will most likely be Proceeding of NOC98. Manchester, be preferred due to the larger inherent a family of wireless IP systems for com- England, May 1998. service flexibility. bined nomadic and fixed access avail- able. 4 ETSI BRAN. Requirements and architectures for broadband fixed 3.3 Migration towards future Moving towards a future full service radio networks. Sofia Antipolis, systems access network (FSAN) a the conver- France, May 1998. (ETSI Technical A possible migration towards future sys- gence of cell (ATM) and packet (IP) Report TR-11177.) tems for fixed wireless access is shown oriented access into a generic delivery in Figure 3. At the moment, no true bi- mechanism for fixed wireless access directional broadband system for wire- (FWA) is bound to happen in the end. Abbreviations less access has reached a large scale The broadcast oriented route towards a ACTS Advanced Communications, deployment in the market. There is a generic approach is in the European per- Technologies and Services number of technologies available in the spective being established in CEPT under AP Access Point market however, mostly offering near- the heading of multimedia wireless sys- broadband capability in terms of avail- tems (MWS). It is unlikely however that APC Access Point Controller able information bit rate. Hybrid MVDS these two routes will arrive at single sys- APT Access Point Transceiver and MMDS systems offer broadband tem definitions in a long-term perspec- capacity in the forward direction but tive of ten years. An achievable objective AT Access Termination have so far only offered narrowband could be to establish a generic frame- ATM Asynchronous Transport Multi- wireline capacity in the return path. P- work for standardisation of broadband plex MP and WLAN systems offer symmetric wireless access allowing for a range of capacity above 2 Mbit/s capacity but are different system implementations BRAN Broadband Radio Access Net- still not widely deployed. adapted to the particular service con- work figurations required. BWA Broadband Wireless Access Within the short term perspective of 23 years true asymmetric broadband access CATV CAble TeleVision systems will be launched in the market 4 Conclusion CEPT European Post and Telecommu- of which MMDS or LMDS like systems In this paper we have addressed wireless nications Consultative Commit- seem to have the highest potential for access to broadband services. An tee large scale deployment. Within the overview covering the present state-of- medium term of up to five years, these DVB Digital Video Broadcast the-art of wireless systems for fixed broadcast oriented systems will probably access is presented, and the evolution ETSI European Telecommunication merge into a more generic system de- towards a future system has been dis- Standards Institute noted xMDS where the major difference cussed. At the turn of the millennium, would be the radio aspects. FCC Federal Communication Com- broadband wireless systems show indica- mission tions of reaching the maturity necessary The P-MP technologies will mostly for volume production and large-scale FDD Frequency Division Duplex evolve into ATM based solutions in the deployment. A part or full merger of dif- short to medium term. From a European FTTB Fibre To The Building ferent technologies will happen on the perspective this is promoted by the ETSI way towards a competitive future im- FTTC Fibre To The Curb BRAN project in close collaboration with poses increasingly more complex the ATM Forum. Their fixed service FWA Fixed Wireless Access requirements on system design and en- P-MP like solution will offer at least gineering. FSAN Full Service Access Network 25.6 Mbit/s capacity and is denoted HIPERACCESS (HIgh PErformance GEO Geostationary Earth Orbit Radio ACCESS system). In parallel, the References HDSL High speed Digital Subscriber BRAN project is establishing a standard Line for nomadic access name HIPERLAN 1 EURESCOM P614. Opportunities (HIgh PERformance LAN) based on for Broadband Radio Technologies HFC Hybrid Fibre Coax ATM. The two BRAN systems will have in the Access Network, EURESCOM HIPERACCESS a common core partly defining a generic P614 Deliverable 4, vol 1 : Main HIgh PErformance Radio wireless ATM system. Report. Heidelberg, March 1998. ACCESS network In the short term a new generation of 2 EURESCOM P614. Opportunities HIPERLAN WLAN technology enhanced to at least for Broadband Radio Technologies HIgh PErformance Radio Local 10 Mbit/s capability will emerge. Also in the Access Network, EURESCOM Area Network UMTS will be launched mostly offering P614 Deliverable 4, vol 2: Annex. ISDN Integrated Services Digital Net- in-door broadband access limited to fixed Heidelberg, March 1998. work or nomadic services. There is strong 94 Telektronikk 2/3.1999

96 IP Internet Protocol IWF InterWorking Function LEO Low Earth Orbit LOS Line-Of-Sight MMDS Multi-channel Multi-point Dis- tribution System MVDS Multi-channel Video Distribution System MPEG Motion Pictures Expert Group LMDS Local Multi-point Distribution System MWS Multi-media Wireless System POTS Plain Old Telephone Service PSTN Public Switched Telephone Network SME Small and Medium Enterprise RLAN Radio Local Area Network (see WLAN) RR Radio Repeater RT Radio Termination TE Terminal Equipment TDD Time Division Duplex TV TeleVison UHF Ultra High Frequency UMTS Universal Mobile Tele- communication System VHF Very High Frequency WLAN Wireless Local Area Network WLL Wireless Local Loop Harald Loktu (36) is Research Scientist at Telenor R&D, Kjeller. He is working in the field of broadband wireless communication for fixed services, with a special interest for overall system design and analysis. email: [email protected] Erwan Bigan graduated from Ecole Polytechnique in 1986, Ecole Nationale Suprieure des Tlcommunications in 1988, and received a Doctorate in Physics from Universit dOrsay in 1991. From 1988 to 1992 he was research en- gineer with France Telecom CNET, developing high- speed modulators for long haul optical fibre transmission systems. From 1992 to 1995 he was Assist. Professor at Northwestern University, Evanston, Illinois. Since 1996 he has been in charge of broadband wireless local loop re- search and development in France Telecom - CNET. email: Telektronikk 2/3.1999 95

97 Mobile broadband access RUNE HARALD RKKEN AND STEIN WEGARD SVAET This article focuses on mobile broad- calls to them regardless of position mobile communications for use through- band access networks. Our viewpoint within the operation area. This is known out Western Europe. The standardisation is to look at systems emerging from the as roaming in the mobile societys lan- work was later on taken over by ETSI, mobile communications society, where guage. The system should also be able to and the resulting system was the well- the vision of wireless communications automatically switch the users access known GSM. as the customers first choice is highly point from one base station to another alive. We foresee the transition to per- during a conversation. This is denoted a As for the first generation cellular sys- sonal communications based on tailor- handover. Such characteristics of a tems, it was demanded that the system ing of services, service mobility, user mobile communications system seem should route calls automatically to a mobility and terminal mobility. The inevitable today, but especially the two mobile in any position inside the cover- work on third generation mobile com- last paragraphs were quite revolutionary age area. The idea this time was that ser- munications systems is now in an at the end of the seventies [2]. vices should be based on services for the intense stage, aiming at finalising the digital integrated services network ISDN first sets of specifications by the end and allow for encryption of user data of the century. Our intention is to look 1.2 First generation cellular over the air interface. beyond the first phase of third genera- In 1981 the worlds first automatic cellu- tion mobile communications systems, lar international mobile telephone system It was also demanded that it should be making visible the concept of mobility was commercially put into operation in possible to have several GSM systems in into multimedia networks. Norway, Sweden, Denmark and Finland. one country, making mobile communica- The system was named NMT Nordic tions the first area of telecommunications Mobile Telephone. The system was ana- exposed to full competition. 1 Introduction logue and was mainly designed for speech communications. The launch of Characteristics of second generation 1.1 From Morse telegraphy to the NMT system was a great enhance- systems are digital transmission, basic cellular services ment in making public telephony avail- speech service, supplementary services, able outside the wireline network. limited data capabilities and continental Morse telegraphy between land and ship roaming. GSM has later on spread to dif- relying on short wave radio was probably The first versions of NMT terminals ferent parts of the world, hence living up the first useful application of communi- weighed more than 10 kilos and were to its name Global System for Mobile cation to mobile users. Land mobile mainly manufactured for installation in Communications. There are by February applications have been present since the cars. At the end of the eighties hand 1999 some 140 mill. GSM users world- 1930s, in the form of closed user groups portable mobile phones started appear- wide. GSM and its sister systems D- connected to a base station. Police forces ing. Miniaturisation has continued and AMPS in USA and PDC (Personal Digi- were among the pilot users of such ser- today true pocket phones are available tal Cellular) in Japan are digital by nature vices. Later on, several similar services for the analogue mobile systems. and hence more suited for data communi- were launched to the public, like the cations than the first generation cellular Norwegian service named OLT (offentlig Characteristics of first generation cellular systems. However, for the digital cellular landmobil tjeneste = public land mobile systems are: analogue transmission, basic systems operating today the vast majority service) which was made available in the speech service, data capabilities not built of the connections made is for speech late 1960s and was up and running until into systems, limited roaming capabili- telephony. Maybe not a big surprise, as the beginning of the 1990s. ties. only very low data rates are available, eg. 9.6 kbit/s as the highest available data During the operation of OLT it soon In addition to NMT there are several rate for GSM. Besides, the available ter- became evident that this was not well other systems belonging to the so-called minals are not well suited to data com- suited for offering mobile communica- first generation cellular mobile phones, munications, and the users are not yet tions to the mass market. Capacity prob- like the AMPS (Advanced Mobile Phone quite familiar with messaging services. lems were envisaged due to limited avail- Services), TACS (Total Access Commu- able radio spectrum. Hence it was nication Systems), NETZ-C, MATS-E regarded advantageous to use a cellular and others, totally holding some 93 mil- 1.4 Improvements of second layout of the base stations to allow reuse lion subscribers world-wide [3]. generation systems of the radio frequency spectrum to raise 9.6 kbit/s cannot be considered as wide- the total system capacity. In the same band wireless access. Hence, within the period, during the 1970s, switchboard 1.3 Next step digital cellular standardisation bodies work is going on operators were removed and calls in the During the 1980s standardisation work to improve the data capabilities within wireline networks were set up automati- was going on to design a harmonised the second generation systems. Figure 1 cally. Hence the idea of forming an auto- mobile communications system for use shows GSM terminals optimised for matic international mobile network took throughout Western Europe. Reservation speech and data services. The first step place in the Nordic Telecommunications of a common frequency band throughout is the possible introduction of high speed Administrations. The basic idea was to Western Europe, together with the Euro- circuit switched data (HSCSD) into make as many as possible of the teleser- pean Post and Telecommunications GSM. Using four time slots per user and vices offered via the wireline network Union (CEPT) as a co-ordinator of this with the new rate of 14.4 kbit/s per available to mobile users, and the mobile work made it possible to set time pres- timeslot, this may give a user rate of system should be capable of automati- sure and national industry policies aside 57.6 kbit/s. Technologically the service cally tracking the users and forwarding and agree upon a unified standard for is expected to be available in 1999. If and 96 Telektronikk 2/3.1999

98 1.5.1 Increased mobility One trend that has been around since mobile telephony was introduced is head- ing for increased mobility and ubiquitous access to the wanted services. With the introduction of the S-PCN (Satellite Per- sonal Communications Networks) based on Low Earth Orbit satellites the cover- age will be ubiquitous. It is however expected that in the future there will be different coverage areas for the different services, due to the fact that increased data rates give decreased coverage area due to radio propagation mechanisms. Four types of mobility are being referred to: User mobility: One access number is connected to a user, regardless of ter- minal and access network used. One sort of user mobility is SIM card roam- ing (plastic roaming) within GSM, which can also be used to extend GSM PHOTO: NOKIA WEBSITE coverage by use of LEO satellite sys- tems; Figure 1 GSM terminals optimised for both speech and data services Service mobility: Uniform access to the same set of services across termi- nal and access network used; Terminal mobility: Continuous mobil- ity across locations, relying on radio based mobile terminals that also give the possibility of changing access point during conversation; Session mobility: During a communi- cation session the user can move be- when the operators will introduce (EDGE) is also being considered for tween terminals and access networks. HSCSD for GSM is of course dependent enhancing the data rates for GSM ser- One example is a person being alerted on how they consider the market for such vices. The operation of 384 kbit/s packet on his radio pager that someone wants a service. services using GSM is being considered a video conference with her. Then the within the standardisation bodies. user has the possibility of transferring The next step in tailoring GSM to data the session to a terminal and access communications is to introduce packet Since none of those services are imple- network offering the requested ser- data services into GSM. The standard mented yet, it is difficult to forecast what vices. called General Packet Radio Service the market for GSM-based data services (GPRS) is being specified at the moment. will be. What we know is that there is an Depending on the number of time slots emerging market for wireless access to 1.5.2 Demand for higher and the coding scheme used the maxi- broadband services. Improving terminals bandwidth mum data rate will be approximately 170 and man-machine interfaces will proba- kbit/s, but it is believed that the rate of bly be the most important factor to make Telecommunications is no longer re- 115 kbit/s will be more widespread. The mobile data take off. We do however see stricted to person to person voice calls. introduction of packet data services in that the terminal manufacturers are quite Users also require access to information GSM will make it necessary to introduce innovative at the moment, giving a cer- services, messaging, teleconferencing, new network nodes into the GSM net- tain pressure on the data capabilities of collaborative working, databases, work. The technology for GPRS is the mobile networks. Intranet access, WWW, etc. This has led expected to be commercially available in to an ever-increasing demand for band- the year 2000. Whether the different width, and the trend will continue. Data operators will choose to introduce GPRS 1.5 Trends applications tend to become more and is then dependent on how they consider Several trends are appearing, which in a more bandwidth consuming, and multi- the market for a cellular packet switched serious manner will affect the develop- media information and communications service. It is even possible to enhance the ment of wireless broadband access over technologies will be more and more inte- GPRS service. A service named the next decades. We will pay attention grated into business processes rather than Enhanced Data Rate for GSM Evolution to some of those trends. just used as a support tool. Telektronikk 2/3.1999 97

99 Traditionally, the available bandwidths video telephony; been designed for powerful PCs. To fulfil have been higher in the fixed networks the demand for mobility without limiting video conferencing (128 768 kbit/s); than in the mobile ones, due to the fact the communication capabilities, future that radio spectrum is a very limited messaging services; application platforms must handle com- resource. However, when people get used munication platforms with lower trans- telefax group 4; to access services from the fixed network mission capabilities, using less powerful they also want access to the same set of data base access; terminals like smartphones, micro- services using mobile terminals. This browsers, PDAs (Personal Digital Assis- broadcasting services. leads to the demand for higher band- tants), palm top computers, etc. Within widths also within mobile communica- year 2000 the prognosis is 22 million These services become available to tions. users world-wide using other types of mobile users, either by designing systems terminals than PCs to access the Web [4]. for higher bit rates or by accommodating WAP Forums aim is to standardise a 1.5.3 Fixed and Mobile for protocols to tailor the services to the protocol for wireless Web access. WAP Convergence bit rates being offered by the different Forum has through the published Wire- radio systems. One consequence of mul- Fixed and Mobile Convergence (FMC) less Application Protocol (WAP) speci- timedia services is also that the has been a buzzword for some time. fication identified a set of protocols and demanded bandwidth might be different FMC can be the bundling of mobile and programming languages which will allow in the different directions, giving asym- fixed services done by the service pro- further development of mobile phones metric links. This is typically the case vider, or even in the simplest case issuing into microbrowsers. Figure 2 shows a when there is a need for a wideband one bill for mobile and fixed subscrip- WAP terminal. The idea is to offer wire- information transfer channel in one di- tions. For the user, this will lead to ob- less applications using Web browsers rection and a narrow channel for control vious benefits such as single point of utilising protocols filtering and adapting information in the opposite direction. contact and access to the wanted set of the information from the Web server to a services independent of terminal and format suited for the thin client (mobile access network, offering communications 1.5.5 Internet based services terminal). based on personal addressing. The ser- to thin clients vice providers possibilities of bundling The introduction of the WAP is based on The design of advanced data applications services is, however, dependent on the the assumption that there is a demand for based on Internet Web technology in regulatory situation. Internet access from portable terminals mobile environments has recently started. and has paved the way for making a vari- Traditionally, Internet technology has FMC at a service level, as described ety of terminals, including the mass-mar- above, can be implemented in several ket handsets then becoming true informa- ways, ranging from use of interworking tion appliances. functions within the network to full inte- gration based on one common service 1.5.6 The telecommunication platform, transport network and switch- terminal as a module in ing/routing platform. FMC at the infra- a multifunctional device structure level implies the latter option, ie. integration of platforms and networks. Bringing the cellular phone and the elec- Obviously, the integration will ease the tronic personal organiser more or less provision of FMC services. However, wherever one goes is quite usual among the optimum degree of integration must people today. Products containing both always be subject to attention based on an organiser and a mobile phone in one benefit for the customer and cost and package are already on the market. As revenue for the network operators. Also electronic commerce is becoming more Mobile IPv6 provides for enhanced and more widespread the next step of mobility suitable for mobility handling integration could be that also the wallet within or across access networks. Inter- turns electronic and becomes a part of a net technology is a very promising candi- communicating enhanced PDA. Keys date for core and access network technol- and physical access control could also be ogy in future mobility systems, possibly made electronic and integrated in a mul- paving the way for true implementation tifunctional device. Figure 3 shows two- of Fixed and Mobile Convergence. slot terminals which give space for both the SIM card and a credit card. These are already on the market. There is a trend 1.5.4 Multimedia access for telecommunication companies enter- Users are more commonly requesting a ing banking and banks entering telecom- variety of services requesting different munications, hence diminishing the bor- service qualities. Examples of such ser- ders between the different players. vices are: A development in the direction of inte- 7 kHz audio (AM quality); PHOTO: NOKIA WEBSITE grating what people usually carry in their 20 kHz audio (CD quality); Figure 2 WAP terminal pockets into one electronic device will 98 Telektronikk 2/3.1999

100 2 Overview over system concepts for wireless broadband access 2.1 Emerging concepts 2.1.1 UMTS and IMT-2000 As early as 1985, the CCIR Interim Working Party 8/13 (IWP 8/13) was established to study a third generation system concept known as FPLMTS (Future Public Land Mobile Telecommu- nications System). This working group is now known as ITU-R task Group 8/1, and the system concept has changed its name to IMT-2000. The concept is the near term realisation of the ITU vision of global wireless access in the 21st cen- tury, including mobile and fixed access. PHOTO: MOTOROLA At the end of the 1980s the European Commission boosted the research acti- vities in the area of mobile broadband Figure 3 Two-slot terminals giving space for both the SIM card and a credit card systems through the RACE programme (Research into Advanced Communica- tions for Europe). The term UMTS was in fact introduced by the RACE Mobile project [1]. Later on, these research activities have been continued within the ACTS programme (Advanced Communi- depend on the level of security that can directing the antenna beam towards the cations Technologies and Services). be obtained. Satisfactory security solu- communication partner only, more users tions to prevent fraud, eavesdropping or can be assigned to the same base station. ETSI followed the ITU in 1991 with the unauthorised use are essential. Also, An additional capacity gain is achieved formation of a Sub Technical Committee acceptable ways of handling lost or by reduced interference levels, making called Special Mobile Group 5 (SMG 5) stolen devices have to be found. smaller frequency reuse distances pos- responsible for the standardisation of a sible. Hence, smart antennas can increase third generation mobile system called both capacity and QoS in wireless broad- 1.5.7 Smart antennas Universal Mobile Telecommunications band access networks. System (UMTS). Although being de- Personal and mobile communications veloped by ETSI up to now, the UMTS systems have experienced an overwhelm- 1.5.8 Things that think and link development has enjoyed the support not ing increase in the number of users in only from Europe. The development of recent years. With an increasing number There is a tendency to include intelli- UMTS has therefore now become more of users and also additional services like gence in all kinds of electric and elec- global, under the responsibility of 3 GPP enhanced data services being introduced tronic devices. For instance there could (Third Generation Partnership Project), there is a growing need for capacity. be a communication device in the coffee which is a global organisation. Increased spectrum efficiency is one way machine, so that it could be remotely of meeting the growing need for capacity controlled via the Internet. Other applica- A general feature that is common for in cellular systems. Traditional base sta- tions are toys that have communication IMT-2000 and UMTS is modularity, in tion antennas are omnidirectional or sec- capabilities enabling them to educate the sense that access networks are clearly tored, resulting in a waste of power each other, like the Furby dolls that can separated from core networks at specified because most of it will be transmitted in be trained as well as share their knowl- interfaces. The standards will open for a other directions than towards the desired edge with other Furbys [6]. Hence, there many-to-many relation between access user. In addition, the power radiated in will be communications needs not only and core networks. This concept is moti- other directions will be experienced as connection people, but increasingly there vated by the need for different access interference by the rest of the users. One will also be communications needs networks in different environments (eg. very promising technique for increasing between electronic equipment, maybe not satellite, wide-area cellular, cordless or spectrum efficiency is the use of the even including the user in the communi- fixed) and different core networks. The smart or adaptive antennas [5]. This cations. There are several candidates for concept also invites to competition be- technique adds a new way of separating such kinds of communications. Two of tween different standards and imple- users on one base station, namely by the best known are the initiatives Blue- mentations. Within ITU, this concept is space, introducing the concept of SDMA Tooth [7] and HomeRF. referred to as the IMT-2000 Family of Space Division Multiple Access. By Telektronikk 2/3.1999 99

101 Systems. The current understanding is that network segments of UMTS (eg. access networks) may become members of the IMT-2000 family. The modularity Access Core Fixed principle is illustrated in Figure 4. network network Network The access network functionality is mainly concerned with radio trans- mission and radio resource management Figure 4 Modularity principle of third generation mobile communications systems including functions for local mobility. The core network functionality is con- cerned with the functions for control of the calls or transactions, the subscriber data, global mobility and provision of mobile specific services. The fixed net- tight link to one second generation stan- HIPERACCESS (HIgh PErformance work may be of any relevant type, eg. dard is neither technically nor politically Radio ACCESS network), which pro- (B)ISDN or a Public Data Network as for feasible since there are several competing vides remote radio access to broadband instance Internet. It should be noted that second generation standards world-wide. applications, supporting a range of the separation of the mobile communica- data communications services. The tions system into an access network and radio spectrum for this application a core network may lead to terminology 2.1.2 Broadband Radio comes from almost anywhere in the confusion, since in a general context the Access Networks 2 GHz to 60 GHz region. sum of the two may be referred to as an Broadband systems for local radio based access component to the fixed network. HIPERLINK (HIgh PErformance access may be classified into two cate- Radio network LINK), which is a net- gories, namely The first implementations of third gener- work-to-network radio interconnect ation systems concentrate first of all on Systems providing LAN functionality; which will support ATM and possibly the cellular application, where a cellular other protocols. Systems providing fixed radio connec- access network component in the 2 GHz tions to customer premises. band is developed. The allocation of 230 Each of the above BRAN Family mem- GHz in the 2 GHz band was endorsed by bers will support ATM transport and In the R-LAN category, the main func- WARC 92, and is one of the important signalling protocols. Support for other tionality is to provide communication early results of third generation standard- protocols, eg. Internet Protocol, is not between portable computing devices isation. There is, however, a notable dif- precluded. and broadband core networks, giving ference in emphasis on the applications telecommunications access and being for IMT-2000 and UMTS. Whereas UMTS in the first phase focuses solely capable of supporting multimedia appli- 2.2 Characteristics of on the cellular application, possibly cations in the future. Local user mobility emerging concepts within the service area is supported. The enhanced with indoor or campus applica- RLL category focuses on replacing the 2.2.1 UMTS tions, the IMT-2000 also includes wire- wireline to the customer premises, and less local loop application and the satel- Easy to use and customisable services has its strengths compared to the wired lite access component. This is probably together with prices competitive with access in that it allows rapid and flexible due to the stronger influence within the fixed access are the key success factors deployment. The systems are intended to ITU from countries with less developed to UMTS. Similarly, there will be a need be able to compete with and complement telecommunications infrastructure. The for a wide range of terminals which are other broadband wired access systems wireless local loop application is a viable affordable to the mass market and which including xDSL and cable modems. solution for rapid deployment of fixed still support the advanced capabilities of telecommunications, particularly appli- UMTS. Figure 5 depicts a possible ETSI BRAN (Broadband Radio Access cable in developing countries, where the UMTS terminal. The majority of the Network) is developing specifications market is believed to be at least as large surface of the terminal is allocated to for a family of broadband wireless access as the market for mobile communica- the display rather than the keyboard. systems that support various applications. tions. This reveals the emerging idea that in the The BRAN Family members are: future data services will be of increased There is also a difference between HIPERLAN/2 (HIgh PErformance importance to the user. UMTS and IMT-2000 with regard to Radio Local Access Network), which the relation to GSM. UMTS is tightly provides local access with controlled Market studies show that speech will linked to GSM. The first phase of stan- QoS to broadband applications and remain the dominant service up to year dards will be based on the same type of services as well as to telecommunica- 2005 for existing fixed and mobile tele- fixed infrastructure, so UMTS may be tions services, eg. Internet and video phone networks, including GSM [8]. seen as an evolution of GSM. This tight conferencing. The services are Hence, the main motivation for users to linkage is of course due to the fact that accessed through wireless terminals, move from second generation cellular to UMTS has come about from a European including portable computers, using UMTS will at the first stage be demand- initiative supported by GSM operators unlicensed radio spectrum in the ing advanced data and information ser- and manufacturers. For IMT-2000, a 5 GHz band. vices. Long term forecasts for UMTS 100 Telektronikk 2/3.1999

102 shows a strongly growing multimedia subscriber base. The potential to support 2 Mbit/s data sets UMTS clearly above second generation mobile systems. In addition, the inherent Internet Protocol (IP) support of UMTS is a powerful com- bination to deliver interactive multimedia services as well as other new wideband applications such as video telephony and video conferencing. It is also considered if UMTS should be improved by allow- ing BRAN type of access networks to increase the available data rates in certain environments. UMTS is also being de- signed to provide both connection-less PHOTO: NOKIA BROCHURE and connection-oriented data services Figure 5 Possible UMTS terminal with data rate on demand, depending on the users needs and the current status of the network. Hence it is possible to put up a table on what the introduction of UMTS means to the user (Table 1). work standard. At the time of implemen- be added. These access networks are tation, the GSM type core network will intended for high speed Radio LAN or UMTS offers the user a consistent set of consist of a connection-oriented part as RLL applications, and are discussed sep- services even when she roams from her known today and a connectionless GPRS arately. Also shown in Figure 7 is the home network to other UMTS operators part under development. Figure 7 depicts USRAN (UMTS satellite radio access a Virtual Home Environment (VHE). a UMTS architecture. It has also been network). After phase 1 there may be VHE will ensure the delivery of the same discussed to include other new access developments of the core networks that service profile independent of the users networks in phase 1, and it is likely that are alternatives to the GSM/GPRS infra- location or mode of access (satellite or access networks of the BRAN type may structure. This is illustrated in Figure 7. terrestrial). The ultimate goal is full mobility into multimedia networks, and that the different underlying technologies should be invisible to the user. In the Table 1 What UMTS means to the user [8] future there will be a mixture of access networks giving the user access to her basic and advanced services information. Some of the access net- ever-increasing range of services built around virtual works will give access to services via a Far more than home environments public network operator; other networks second generation attractive multi-mode terminals for access to second will be made available eg. within campus generation services areas. Figure 6 shows a reference config- uration for UMTS phases 1 and 2. In future proof for the 21st century later stages of UMTS there will be a pos- sibility of accessing the UMTS services a full third generation global mobile and wireless system via a multitude of access networks. 2 Mbit/s capability in diverse radio environments UMTS 2.3 General architecture highly personalised mass market and functionality new and innovative interactive and multimedia services 2.3.1 Architectures UMTS architecture a full member of the IMT2000 family ITU identified spectrum for both terrestrial and satellite As previously stated, the UMTS architec- UMTS access via ... radio ture is in line with the modular concept, UTRA, a revolutionary air interface optimised for both where access networks and the core net- FDD and TDD spectrum work are clearly separated and may evolve rather independently. For the first phase of UMTS, the development of a new access network called UTRAN build on the footprint of the evolving GSM core network (UMTS terrestrial radio access network) UMTS networks are compatible with Internet Protocols is the main focus. UTRAN will be con- and services support convergence of fixed and mobile services nected to a core network of the GSM type, and will probably also be compar- access via mobile or fixed, public or private networks able with the American IS-46 core net- Telektronikk 2/3.1999 101

103 Service IP, VHE, Mobile IP platforms IN, IP, VHE Service Mobile IP MSC, SGSN, HLR/VLR networks Transport IP, ATM networks ISDN, IP, X25 GSM/ USRAN UTRAN BRAN * GPRS ATM-LAN DAB Access ATM/ networks ATM-LAN DVB WATM Blue- LMDS SDH FDDI tooth Home CATV PLC WDM RF xDSL rural (sub)urban indoors, campus local wide area *) HIPERLAN, HIPERLINK Will be part of UMTS phase 1 Will probably be part of HIPERACCESS and enhanced phases of UMTS enhanced phases of UMTS Figure 6 Reference configuration for UMTS phases 1 and 2 Future core network technologies are Broadband Radio ATM [9]. Its purpose is to obtain a com- currently under debate in standardisation, Access Networks mon understanding of reference models, and several options exist. These are services, features and interface specifica- ETSI and ATM Forum are working on a based on two main technologies, namely tions. The current common reference Common Reference Model for broad- IP technology and ATM technology. model is illustrated in Figure 8. band radio access networks supporting As seen, the WACS (Wireless ATM Access Systems) Node consists of the WACS Terminal and the WACS Termi- GSM Access GSM based nal Adapter. The WACS Access Point Network Core Network is decomposed into the WACS Access IWF Point Transceiver and Controller. The UMTS Terrestrial WACS Access Point is connected to an ISDN Radio Access Network End-user Mobility Supporting ATM switch, which is the gateway to the ex- UMTS Satellite ternal network. The reference point W.1 Radio Access Network IWF IP-networks is the radio interface and contains the functions for transparent ATM transport, Broadband Radio Other Core support of mobility and security func- Access Networks Networks tions. It also contains the UNI protocol IP based X.25 with mobility enhancements. The refer- ATM based ence point W.2 between the WACS Other Access Networks Hybrid IP/ATM Access Points specifies the signalling Phase 1 related to establishment and release of connections and handover between Phase 2 WACS Access Points. R.1 is a standard Figure 7 Illustration of the UMTS architecture interface for connection to external net- 102 Telektronikk 2/3.1999

104 W.1 WACS - WACS - AP Mobility WACS - WACS - AP WACS - AP Enhanced External Terminal Controller TA Tranceiver Controller -Switch Network W1.1 W1.2 W.2 R.1 WACS Node WACS - Access Point Wireless ATM Subsystem Figure 8 WACS reference model works, and enhancements to support Call-, connection- and bearer control, Call-, connection- and bearer control: terminal mobility may be included. required to perform the set-up and Connection set-up and release in Examples of interfaces are (M-)UNI or release of calls. For connection ori- accordance with ATM signalling (M-)NNI. The above reference model ented services, the functionality is well specifications. Traffic management is general for all BRAN networks. In case developed and described. For connec- is performed to control QoS. Device of HIPERACCESS, the interface between tionless services, necessary function- addressing shall be consistent with the WACS Terminal and the WACS Ter- ality for end-to-end control must be world-wide roaming. minal Adapter may be specified, since it specified; Radio resource management: Monitor- is unlikely that these are integrated in the Interworking with other networks. ing of the radio conditions, and same equipment. The interface is Interworking with ISDN, B-ISDN, dynamic allocation of radio link capac- assumed to be a standard UNI. X.25 PDN and IP data traffic will be ity to fulfil the traffic contract. specified; Mobility management: Local mobility 2.3.2 Functionality Access, service and security control management functionality in case of regulating the access to services and HIPERLAN. The mobility function of UMTS networks; handover is also supported by HIPER- The required network functionality for ACCESS, but is more intended for the Network management, necessary to UMTS is under continuous development. purpose of providing better quality of ease and support tasks such as plan- The functionality falls into the following the radio link than providing mobility ning, installation, provisioning, opera- categories: for the users. tion, maintenance, administration and Radio resource management, neces- customer service. sary for allocation and control of radio communication resources; 2.4 Frequency spectrum BRAN allocated to UMTS Mobility management, including func- The broadband radio access networks The frequency allocation for UMTS was tions for location registration, paging, provide the following functionality: given by WARC 92, resulting in a functions for supplying routing infor- 220 MHz allocation in the 2 GHz fre- mation, and functions for handover; quency band. The frequency allocation for UMTS is shown in Figure 9. 1900 1920 1980 2010 2025 2110 2170 2200 Terrestrial applications Satellite applications Figure 9 Frequency allocation for UMTS Telektronikk 2/3.1999 103

105 2.5 Radio interfaces for UMTS services might only support lower data that satellite and terrestrial systems are rates in remote or heavily congested harmonised, ensuring that roaming and 2.5.1 UTRAN areas. In the early stages of UMTS handover between satellite and terrestrial The UMTS radio access system UTRAN deployment, UMTS services will pro- networks will be possible. No unique (UMTS Terrestrial Radio Access Net- bably predominantly be offered in high USRAN (UMTS Satellite Radio Access work) is aimed at supporting operation traffic locations. To allow users to gain Network) specification exists, and it with high spectral efficiency and service access to their services everywhere probably never will. Instead it is reason- quality in the different environments in operators may wish to arrange roaming able to believe that systems like Iridium, which wireless and mobile communica- agreement with second generation opera- Globalstar and ICO which are already tions take place. UTRAN will offer both tors (for example GSM operators) or satellite components to GSM through connection oriented and connectionless UMTS satellite operators to increase the interworking and roaming agreements, services to the user. The maximum bit coverage area. Seamless operation and also might become satellite components rate offered via UTRAN is 2 Mbit/s, the roaming between private and public of UMTS. In addition, there are other available bit rate at a given location is UMTS networks as well as access to ser- system concepts that might fit into the however dependent on a number of fac- vices via fixed and mobile will be possi- UMTS/USRAN concept, like the tors, like, ble to support these users. The access Inmarsat Horizons concept. Those sys- scheme chosen for UTRAN is W-CDMA tems will probably provide speech ser- The mode of operation, as the core net- (Wideband Code Division Multiple vices and 144 kbit/s to laptop PC type work rate for circuit switched services Access). Hence, the operation of terminals. will be limited to 64 kbit/s, whereas UTRAN will be a bit different from 2 Mbit/s will be available in packet GSM, for example. Characteristic for mode; 2.5.3 BRAN CDMA systems are that all other users The distance between the base station appear as noise to the desired user. This It is assumed that HIPERACCESS and the user. The cell size will de- means for instance that the coverage area systems are primarily deployed in a crease as the data rate increases. This from a base station will decrease as the licensed spectrum. However, this does implies that the maximum bit rate of number of active users within the cell not specifically preclude their use in a 2 Mbit/s in packet mode will be avail- increases. licence-exempt spectrum where there able only in certain environments. may be little or no co-ordination of fre- quency use. 2.5.2 USRAN In practical implementations of UMTS some users may be unable to access the Satellite technology can relatively easily highest data rates at all locations. For provide global coverage and service. 3 Mobile broadband example, the physical constraints of radio Hence it is expected to play an important system (MBS) propagation and the economics of operat- role in the extension of UMTS coverage. ing a network will mean that the system UMTS is being standardized to ensure The MBS (mobile broadband system) concept may be viewed as a cellular sys- tem providing very high user bit rates, typically tens of Mbit/s. The fact that the system is cellular is an important distinc- tion from the ETSI BRAN concept, and implies that the radio link and the mobi- UNI lity management functions are designed for high speed terminal mobility. The Mobile MBS concept has been studied within Terminal 1 Base Station research for several years, and pilot sys- Transceiver 1 tems have been implemented to demon- strate the concept. Due to the high bitrate, the envisaged applications of Base Station MBS are many, including for instance Controller video telephony, teleworking, city guid- ance and TV broadcasts. Of course, more conventional applications as found in Mobile Base Station R-LAN and cellular systems with lower Terminal 2 Transceiver 1 bit rate are also possible for MBS. The ATM biggest challenge for MBS is to provide interconnection the high bit rate radio link with sufficient quality. The system operates at NNI microwave frequencies, where there are NNI: Network-to-Network Interface available frequency bands around 40 and UNI: User-to-Network Interface Video Server 60 GHz. In many environments there are ATM Network unavoidable distortion effects of multi- path propagation, due to delay spread on the required broadband signals, and Figure 10 SAMBA trial system strong Doppler effects and path losses. 104 Telektronikk 2/3.1999

106 As mentioned earlier, the MBS concept 4 The phased approach necessary for UMTS, where a completely has been studied within research bodies. new radio infrastructure and terminals Within the EU research programme The introduction of UMTS will proceed will be needed. Customers can be RACE, the project R2067 MBS studied through a number of pre-operational and attracted onto these intermediate net- the concept and realised a demonstrator operational phases taking place over the works by the provision of attractive ser- at 60 GHz to verify reliable transmission years 2002 2005. The aim of this vices from new content and service including handover. In the EU research phased approach is to reduce risk and providers. These customers will then be programme ACTS, a follow on project cost for operators and ensure early adop- willing to invest in new UMTS terminals called AC204 SAMBA (System for tion of services by end users. A UMTS on the anticipation of better and more Advanced Mobile Broadband Applica- Phase 1 development schedule as en- efficient delivery of enhanced services. tions) has developed the concept further visaged in [8]. It represents the consensus In turn, this provides the incentive for and implemented a trial system operating on the timetable at that time. Backlogs in network operators to invest in UMTS at 40 GHz. Both the MBS and SAMBA some tasks are already visible today and infrastructure in order to satisfy the need projects have based the system concept may require revision of the schedule. for capacity demanded by a successful on ATM, with a target user bit rate of mobile multimedia mass market. 155 Mbit/s. For the implemented demon- Many people within the industry believe strators, the user bit rate has been limited that an important step along the way will The current situation is less than optimal, to around 30 Mbit/s, however, still high be the widespread deployment of packet particularly with regard to licensing cer- enough to demonstrate HDTV on a cellu- radio services being developed for sec- tainty. The EU Proposal envisages publi- lar system. The trial system implemented ond generation systems, such as GPRS cation of procedures for UMTS licencing in the SAMBA project is shown in Fig- (general packet radio service) for GSM. in all member countries by 1 January ure 10. These systems will give valuable experi- 2000. This represents a significant delay ence for the operators with connection- with respect to the schedule originally Both the MBS and SAMBA projects less systems and could provide a plat- proposed by the UMTS Forum in [8], have made significant contributions to form for the development of service while maintaining the commercial key issues such as radio link characteri- interworking functions and service pro- deployment phase beginning in 2002. sation, antenna design, mobility manage- vider interfaces as well as a core of The primary phases of the development ment functions and development of spe- mobile multimedia services. This can be of UMTS are: cialised hardware. done with less initial investment than is Table 2 UMTS timeline [8] Task name 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 UMTS revised vision Co-operative research: ACTS Regulation: UMTS Forum report Regulation: EC, ECTRA measures Regulation: National licence conditions Regulation: Operators identified Operators commitment: Drafting Operators commitment: Signature ETSI: Basic standards studies ETSI: Freezing basic UMTS parameters ETSI: UMTS Phase 1 standards UMTS Phase 1: System development Pre-operational trials UMTS Phase 1: Planning, deployment UMTS Phase 1: Commercial operation Telektronikk 2/3.1999 105

107 Extension of GSMs capability to 5 Conclusions UMTS is aimed at giving access to multi- include packet and high speed data media applications. It predominantly operation, as described above; Until now the main cellular service has evolves from GSM, and will interwork been speech, even though data services with GSM. In the longer term, integra- Pre-UMTS Trial Phase, during which have been present in the operational net- tion between UMTS and IP networks prototype UMTS base stations will be works for several years. However, in offering differentiated quality of service tried out either in subsets of real GSM later years the focus of the mobile society will be the next evolutionary step. UMTS networks, or in isolated trial packet- has changed from speech to data services may then give global access to multime- based networks; due to the tremendous growth in the dia services across platforms like mobile, Basic deployment phase beginning in demand for information services and fixed and satellite-based networks. 2002, which includes the first incor- Internet access. poration of UTRA base stations into Mobility across terminals, locations and live networks and the launch of satel- The next step will be integration of the infrastructure together with tailoring of lite-based UMTS services; new ser- different fixed, cellular and Internet ser- high quality services will make the vision vices based exclusively on UTRAs vices. Mobility functions in both wireless of communicating anytime, with anyone, capabilities; and support of both and fixed networks, and flexible service anywhere come true. The availability of narrowband and broadband services creation and management make the dif- broadband wireless access will substan- over the same UTRA interfaces; ferences between fixed services, cordless tially affect the lives of individual telephony and cellular diminish. Intro- citizens as well as the functioning Full commercial phase, beginning duction of packet switched data into the of society. shortly after 2002 and approaching GSM network, like the General Packet fruition in 2005 incorporating Radio Service (GPRS) is turning the enhancements to its performance and original circuit switched GSM network References capability, and involving the introduc- into a hybrid network. tion of new, sophisticated UMTS 1 Berg, O et al. RACE Mobil Prosjekt. based services. Felleseuropeisk satsing p neste The standardisation of a third generation generasjon mobilkommunikasjons- cellular system Universal Mobile system. Telektronikk, 86 (1), 4855, Telecommunications System (UMTS) 1990. has been going on for several years. In addition, real broadband mobile systems 2 Lvsletten J, Rkken, R H. From offering bit rates of up to 155 Mbit/s mobile telephony to personal com- are being developed. munications. Telektronikk, 94 (2), 28, 1998. 3 Helme, S. Theres still life in the old dog yet. Mobile Communications International, 50, 1998. 4 ComputerWorld Norge, 37, 1997. Rune Harald Rkken (37) is Senior Research Scientist at Telenor R&D, where he has been employed since 1987. 5 Pettersen, M, Lehne, P H. Smart He has been working with standardisation of the GSM sys- antennas the answer to the demand tem, with radio propagation aspects, FMC and mobile for higher spectrum efficiency in per- communications in general. During the last years he has sonal communications systems. been in charge of the mobile and personal communica- Telektronikk, 94 (2), 5464, 1998. tions group at Telenor R&D. He is currently heading the technological aspects of the Telenor UMTS project. 6 Furby (1999,05,18). [online]. URL: email: [email protected] 7 BlueTooth (1999,05,18). [online]. URL: 8 The Path towards UMTS. Technolo- Stein Wegard Svaet (40) is Senior Research Scientist at gies for the Information Society. Lon- Telenor R&D. He has been with Telenor since 1986, and don, 1998. (UMTS Forum report No. with Telenor R&D since 1988. He has been working with 2.) radio planning for NMT, system simulation of GSM and network design aspects of future mobile communications 9 Common ETSI ATM Forum refer- systems. The work has included participation in inter- national activities within RACE, ACTS, EURESCOM and ence model for Wireless ATM ETSI. His current involvement is within the standardisation Access Systems (WACS), TR040001 of UMTS in 3GPP. v 0.3.1 (1998-02). email: [email protected] 106 Telektronikk 2/3.1999

108 Satellite networks and stratospheric platforms AGNE NORDBOTTEN This paper presents the network con- The capacity of these systems is however the satellite broadcast systems in combi- cepts and operational principles of limited which means that they are not nation with the digitalisation of moving new broadband satellite networks suited for servicing of a high number of pictures and digital satellite broadcasting based on GEO and LEO platforms individual users per km2. Within the made possible through the work of and the stabilised stratospheric plat- satellite domain there are now three MPEG (Moving Pictures Expert Group) forms proposed for operation from important areas for broadband and multi- and DVB (Digital Video Broadcasting). 20 km altitude. media access under rapid development: The new broadband digital broadcast services now introduced form the back- Broadband multimedia satellites ope- ground for development of digital inter- rating from a GEO position. The first 1 Introduction ongoing phase of this development is active services. Digital TV transmission normally operates at bitrates of 4 8 a consequence of the development of Satellite systems operated from the geo- Mbit/s. The programs are time division more general interactive digital plat- stationary orbit are today used for com- multiplexed into the MPEG-2 transport forms from the digital broadcasting munication to remote areas (represented stream with a capacity of 34 Mbit/s for systems already in operation. The sec- by the Arctic island Spitzbergen, ferries, the bandwidth used for a PAL or ond phase includes new systems with and oil installations in the North Sea), D2MAC transmission. The satellite on-board processing operating also at communication to mobile vehicles (land transmits in the frequency band 10.7 higher frequencies. and sea), business type communication 12.75 GHz. The RF unit at the receiver (VSAT) and first of all broadcasting, High capacity LEO systems for multi- antenna converts the input to the IF fre- which up to now has been the killer media services under development and quency range 950 2050 MHz which is application. Entirely new and improved production and ready for operation in the input frequency range of the inte- satellite networks are now under de- 2002 2003. Teledesic and Skyway grated receiver detector (IRD) commonly velopment and introduction. The ongoing typically represent them. called a set top box, which interfaces to development involves a multiplicity of the normal TV receiver. The set-up is Satellite systems for UMTS services systems, new services, new operator illustrated in Figure 1. In this way 4 8 operated from LEO, MEO and GEO groups, new technological solutions and digital TV channels are transmitted per orbits are now being specified and the use of higher frequency bands to transponder with a capacity of one ana- developed in close co-ordination with increase available capacity. logue TV channel. the ongoing development of terrestrial UMTS with a focus on obtaining A new system group based on the use of Transmission of TV channels is the dom- world-wide coverage for UMTS. stratospheric platforms has many simi- inant satellite application today, and it larities with the satellite systems, but illustrates the strength of satellite trans- The present status is that some tests are operate in more local areas with a higher missions in broadcast and point-to-multi- running on the GEO systems, and ser- capacity per km2. point applications. The total transmission vices will be gradually offered. capacity at Ku-band from a satellite posi- The new systems which are expected in tion in the geostationary orbit using operation during the period 20002005, 2.1 Broadband multimedia QPSK modulation is approximately represent a total renewal of satellite com- GEO satellites 5 Gbit/s corresponding to 1,000 digital munication and broadcast and take the TV programs. If this capacity is divided The next generation GEO systems are full step into the world of multimedia among 1 million households in Norway follow-up systems of the VSAT and offering possibilities for broadband com- on an individual basis, they would have broadcast systems of today. The basis for munication all over the world regardless 5 kbit/s per family on average. With the rapid development is the success of of previous infrastructure. In this article the focus will be on the potential, possi- bilities and development trends of the new generation satellite and stratospheric networks. 2 Satellite networks Figure 1 Digital TV broadcasting by satellite Satellite systems have some important advantages: They offer total coverage and may operate without any terrestrial infras- tructure. However, for most new sys- Uplink tems interoperability with other net- User location works is a strong requirement; They are very cost effective on point- Set to-multipoint connections and broad- top cast. box Outdoor Indoor Telektronikk 2/3.1999 107

109 Broadcast Channel Broadcast Set top box (STB) (DVB Transmissions Systems) channel Broadcast Broadcasting Broadcast Delivery Network Broadcast Service Media Adapter Interface End Provider Module User Set Top Return Unit Interaction path user Interactive (STU) Interactive Interactive Interface Service Interaction Network Module Provider Network Adapter (can be Network external Interface to STB) UNIT (NIU) Interaction Channel Forward Interaction Interaction channels path network independent network dependent network independent Figure 2 Generic model for a broadcast based interactive system used by DVB and ETSI 100,000 households as users on a 10 % Digital broadcast systems are converted ure 2, which shows the generic interac- basis the available capacity per house- into interactive systems by functionally tive model adopted by DVB and ETSI. hold increases to 500 kbit/s on average. adding an interactive point-to-point com- This is still low for broadband communi- munication channel as illustrated in Fig- This model is based on the use of the cation. This example illustrates that the DVB/MPEG-2 based digital broadcast satellite technology has its strength in channel with a capacity of approximately broadcasting and point-to-multipoint 34 Mbit/s for the down link. The MPEG- applications. The capacity from a satel- 2 transport multiplex may contain digital lite position in the geostationary orbit is TV or data including ATM or IP type of increased through the use of higher fre- data transmissions. The satellite delay of quency bands and more efficient modula- 250 ms does represent complications for tion methods. protocols like TCP, reducing the trans- k Lin d ar ks rw turn Lin Fo DV B B DV Fo rw SIT Re ar d Lin SIT k Re tur nL ink Broadcast s Service Broadcast Feeder Provider Network D Station D Adapter Interactive Service Interactive SIT Network Hub Provider Station SIT Adapter SIT SIT Figure 3 Interactive satellite system 108 Telektronikk 2/3.1999

110 mission capacity to some hundreds of operation late 1999 / early 2000. The Table 1 Tentative requirements for standardisation of kbit/s if proper precautions are not taken. next generation system with on-board interactive terminal processing will have a mesh capability The choice of return channel is more allowing for direct communication be- Parameter Value flexible, different technologies and tween users. capacities may be used. The interactive Antenna diameter Less than 1.2 m channel consists of an up link part and a The down link capacity of the broadcast down link part. It is used for signalling type GEO platforms focused on in Output power Less than 2 W and ID purposes, service request and up Europe is limited to 34 Mbit/s per trans- Modulation QPSK link information transfer. Today ISDN is mission beam. Downloading of videos commonly used for the return channel. and movies will require much higher Max data capacity 2 Mbit/s capacity, at least 155 Mbit/s. This may Access system MF-TDMA The preferred solution is return by satel- lead to development of systems where lite, and this is now being tested and the TV programs are included in high implemented. A proposal for standardisa- capacity ATM or IP based transmissions. tion has been worked out by a group of The new GEO systems proposed by dif- European satellite operators co-ordinated ferent US companies may move in such through ESA [1]. In the proposed system, directions in particular since they are not which is referred to as a Satellite Inter- TV based. Broadcast or point-to-multi- 2.2 Broadband LEO active Terminal (SIT), the up link is point operation will however be the effi- satellite systems proposed for operation in the frequency cient mode of operation for satellite net- Broadband LEO satellite systems are band 29.5 30.0 GHz. The total inter- works. The increased storage capacity of under development and are expected into active system is referred to as a Ka/Ku- PCs used for home servers will favour operation in 2002 2005. The high band solution. An illustration of the sys- this development. capacity LEO systems are typical multi- tem is shown in Figure 3. This proposal purpose satellites not dedicated for TV has been taken up by DVB and the finali- Market penetration depends on the cost broadcasting. Transmissions may be IP- sation of the standard will hopefully take of user equipment, the number of users or ATM based. Their interoperability place in 1999. sharing the transmission costs and a with terrestrial networks is less compli- higher transmission capacity per satellite. cated than for GEO systems since the The main reason for the combined solu- It has been estimated that a geostationary time delay over the satellite hop is negli- tion is the more favourable link budget gigabit satellite with on-board switching gible. While the emphasis for GEO satel- obtainable at Ka-band allowing for can be realised with a capacity of about lites is on modest interactivity, LEO 2 Mbit/s with reasonable antenna size. 45 Gbit/s corresponding to 80,000 satellite constellations represent more The tentative requirements set for the 64 kbit/s circuits [2]. The transmission efficient solutions to highly interactive standardisation work were as listed in cost is highly dependent on the capacity services with a short round trip time over Table 1. of the satellite. For the high capacity the satellite. The low propagation delay system with a space segment cost of 400 has a strong impact on response times The proposed solution is based on ATM million euro the cost per minute per using protocols such as TCP/IP and for the up link. The access system of the 64 kbit/s can be less than 0.01 euro simplifies interoperability with terrestrial interactive channel is based on MF- which would then also be the cost for networks. For more local traffic the LEO TDMA with a high degree of flexibility a 2 Mbit/s circuit shared by 30 users system can be considered as a RLAN or with regard to up link bitrate. The SIT located in the same down link beam spot. an integrated radio based part of a terres- will operate at bitrates up to 2 Mbit/s trial network. with an output power not exceeding 2 W and an antenna diameter less than 1.2 m in diameter. The MF-TDMA multiplex, however, should have a capacity of at least ATM-25, preferably ATM-50. Table 2 Some characteristic parameters for Skybridge and Teledesic The first generation of interactive sys- tems will be based on the use of trans- parent satellite transponders. For the next Parameter Skybridge Teledesic generation it is expected that on-board processing will be used allowing for Satellite constellation 2 x 32 LEO 12 x 24 LEO direct communication between users in Total coverage area 68 90 different antenna beam areas. An option for Ku-band return preferred by some Altitude 1,457 km 1,400 km operators is included in the specification proposal. This option is preferred for Down link frequency (DL) 10.70 12.75 GHz 18.9 19.3 GHz installations with a large number of users Up link frequency (UL) 12 18 GHz (part of) 28.6 29.1 GHz (like SMATV). The first generation interactive satellite systems based on Max DL capacity per user 60 Mbit/s 64 Mbit/s transparent transponders will operate in a Max UP capacity per user 2 Mbit/s 2 Mbit/s star network. The first systems will be in Telektronikk 2/3.1999 109

111 The main types of terminals discussed 3 Stratospheric platforms for S-UMTS are hand-held, vehicular, transportable, fixed and paging receivers. The philosophy of the stratospheric plat- The different terminal types will nor- forms is to find a solution for providing mally provide a bit rate which is lower high capacity broadband services, which than the maximum S-UMTS bit rate typi- can be preferable to satellite and radio cally obtainable using fixed transportable solutions. Platforms kept in stable posi- uplink/ control terminals. Standardisation must specify tions 1624 km above the surface of the 691 cells the bit range and services offered by each earth are claimed to represent very attrac- type of terminal. The portable terminal tive possibilities. They will cover an area must be cheap, compact and lightweight of approximately 3000 km2 without line- with low power consumption. The first of-sight problems and thus represent a generation of this group of terminals is solution with coverage advantages rela- the Iridium terminal. They will exist both tive to terrestrial radio based solutions as dual mode (multi-mode?) and for like LMDS. The up- and down link satellite use only. power requirements will be much lower 60 - 80 km than for a satellite system even when it is For all terminal types handover terres- operated in a LEO orbit. Table 3 shows trial/satellite and between satellite beams a comparison between different radio Figure 4 Basic principle for balloon borne are a requirement. These problems are systems based on spot beam numbers and multimedia platform now being studied in several ACTS pro- sizes. jects (SINUS, SUMO, THOMAS) and the main conclusions from these projects In addition the available frequency range will be made available during 1999. Pre- and its reuse potential are important liminary tests, which are now performed parameters when discussing capacity. In At present satellite Internet access seems by operators, indicate problems with con- general a system based on small cells to be the market focused on by these sys- nections over several minutes using the with a high reuse potential will have the tems. Problems associated with handover IRIDIUM, which is in a pre-operational highest capacity which means LMDS, and line of sight blockage may compli- phase of testing, and system evaluation. stratospheric platforms, LEO satellites cate for the user. Electronically steerable and GEO satellites as shown in Table 3. antennas are required and they are not The mobile satellite systems under de- This is however a simplification not tak- available at an acceptable cost yet. The velopment are mainly of the personal ing into account the traffic between cells Teledesic system operated over a 500 communication type. A system like and the amount of broadcast/multicast MHz band in Ka-band has been con- Inmarsat Horizon will have higher capac- traffic. sidered the most ambitious of these ity, but is not so far meant to become systems. The number of satellites has compliant with the UMTS standard. For A stratospheric platform may be estab- recently been reduced to 256. The spot services requiring higher capacity than lished in different ways; the proposals, beam area of 3,000 km2 will have a voice / low rate data using hand held ter- which today seem reasonably close to communication capacity of approxi- minals, the terminals fall into two cate- realisation, are based on the use of sta- mately 300 Mbit/s. Frequency resources gories; portable and mobile. The mobile bilised balloons at an altitude of 23 km or for operation of new satellite systems is a terminals will be mounted in vehicles aircraft circling at an altitude of 16 18 km. problem. A European LEO system, Sky- and trucks, on-board ships and aero- bridge, planned for operation at the Ku- planes, while the portable terminal band frequencies of the GEO satellites, typically represented by a laptop will will have to turn off transmissions in the be personal. direction of a GEO system to avoid inter- ference. Table 2 gives a comparison be- tween the two systems Skybridge and Teledesic. The parameters are very simi- lar, but Teledesic has a better coverage Table 3 Capacity of different radio based systems for broadband distribution towards the polar regions. Technology Spot beam capacity Spot beam area Capacity per km2 2.3 Multimedia mobile satellite systems Terrestrial 4 6 beams 3 15 km2 500 Mbit/s The strong focus on UMTS also includes LMDS per tower mobile satellite systems; S-UMTS. In UMTS, the satellite part of the network GEO satellites Regional spot Up to semi-global Low has been foreseen to provide mobile mul- LEO satellites 1 spot beam 3000 km2 100 kbit/s timedia services up to 144 kbit/s. It may per town be discussed however whether this could be increased to at least 384 kbit/s, possi- Stratospheric 700 spot beams 5 10 km2 1 Mbit/s bly 2 Mbit/s. Some standardisation work platforms per platform still remains. 110 Telektronikk 2/3.1999

112 3.1 Balloon borne platforms Table 4 Some characteristic platform parameters A Washington DC based company, Sky Station International, promotes the sys- Parameter Value tem consisting of balloon borne plat- forms [3]. Several European companies Height of operation 21 23 km are contributing to the system by making both balloons and electronic equipment. Platform length 150 160 m The concept is illustrated in Figure 4. Diameter 50 m The main data for the balloon borne plat- Volume 170,000 m3 form are listed in Table 4. Weight 11,000 kg The platforms will be stabilised and operated using power from solar cells Payload 1,000 kg and storage in fuel cells. This requires the platforms to be operated in areas Coverage area 5,000 km2 max somewhat to the south of the polar circle to produce energy enough from the solar cells. The first platforms are planned for operation over Rome, Lisbon and Singa- pore in year 2000. The critical issues, which are not too well documented, are the problems of stabilisation and the Table 5 Electrical parameters for proposed platform payload amount of energy required for keeping the platforms in a stable position. The total communication capacity may be Parameter Value quite large; approximately the same as for a Ku-band satellite position while the Allocated frequencies 47.2 47.5 and cost of the platform is approximately 80 47.9 48.2 GHz million euro. It may be taken down for repair and there is a possibility for reuse Frequency reuse factor 9 of equipment. Thus the system may rep- Total capacity 7.68 Gbit/s resent a success if stable operation is established. This type of system has Number of beams 691 obtained its own frequency allocation at the WRC 97 conference. The main elec- Covered area per beam 7.3 km2 trical and operational parameters for the platforms proposed are listed in Table 5. Capacity per cell 11.114 Mbit/s The data in Table 5 indicate that this is a Capacity per household* 3 kbit/s system which can be used for telephony, Internet access, conferencing and more Number of channels 100,000 general data exchange. It seems to be a very flexible system. Since only local Available user rates 64 2048 kbit/s areas are covered however, it is strongly Up link access MF-TDMA dependent on other networks for connec- tions outside the coverage area. The sys- Down link TDM tem as presented is not well suited for broadcast or point-to-multipoint delivery. * With 500 households per km2 That would involve a large number of antenna beams. Introducing an overlay antenna beam covering the whole area would solve this problem. 3.2 Aircraft based platforms but cost effectiveness may become a crit- The system has not been allocated any ical issue. The High Altitude Long Oper- frequency bands for operation, but it is Another system is based on the use of ation (HALO) Aircraft will operate for 8 assumed that it may operate anywhere manned aircraft circulating in orbit above hours before it is replaced by another air- from 2 GHz up to at least 50 GHz and in the coverage area at a radius of 3 8 km craft and another crew consisting of two particular in the LMDS bands at 30 (US) at an altitude of 16 km which is well persons. Figure 5 illustrates the principle and 40 GHz (Europe). The name of the above the corridors for commercial air of operation. It is obvious that the mov- company, Angel Technologies, is some- traffic [4]. The concept is based on ing platform automatically leads to han- what ambitious. known and proven aircraft technology, dover or antenna steering problems. Telektronikk 2/3.1999 111

113 Several types of interesting and promis- ing types of platforms are now being developed and introduced. Some of them will be winners, other losers. Good busi- ness concepts and cost effective user equipment will be important factors. Standardisation of equipment and basic operational principles are definitely re- quired. References 1 Definition of a Satellite Interactive HUB Terminal for use in a DVB system. 1999, 05, 11. [online]. URL: http:// Illuminated areas 2 Witting, M. Large-Capacity Multi- media Satellite Systems. IEEE Com- Figure 5 Principle of operation for HALO Network munications Magazine, 35, (7), 4449, 1997. 3 Rothblatt, M, Frohbier, J, Ye, H. Stratospheric altitude : the key to delivering broadband wireless local loop service to consumers world- wide. 1999, 05, 11. [online]. URL: 4 Concluding remarks Total coverage globally, as required in UMTS, favours the use of satellite net- press/WRC97/Skystation.html. The increasing need for efficient broad- works for most regions of the world from band access has led to the proposal and an area point of view. 4 Djuknic, G, Freidenfelds, J. Estab- development of different types of new lishing wireless communication ser- access systems. The capacity of the dif- The stratospheric platforms are focusing vices via high-altitude aeronautical ferent systems referred to area or popula- on densely populated areas with insuffi- platforms : a concept whose time has tion density decreases with increasing cient infrastructure for their first genera- come? IEEE Communications Maga- cell size, resulting in low individual tion. With operational success during this zine, 35 (9), 128135, 1997. capacity for satellite systems with semi- introductory phase, they may represent global coverage. However, the larger the strong competition for terrestrial net- cell, the better the broadcast ability of a works in southern parts of the world. system. The different systems are com- They have also been discussed for use as plementing each other more than com- base stations for high capacity mobile peting. networks. Agne Nordbotten (62) is Research Manager for Satellite and Radio Systems at Telenor R&D, Kjeller. He is working on the development of interactive broadband services by both satellite and cellular radio systems. At present he is project co-ordinator for the ACTS project CRABS Cellu- lar Radio Access for Broadband Services. email: [email protected] 112 Telektronikk 2/3.1999

114 The passive optical network (PON) DAVID FAULKNER This paper presents passive optical Broadband passive Telecommunications over network (PON) concepts, architecture options and key components for optical networks a passive optical network narrowband PONs and broadband Cable TV networks on the other hand This and other experiments led to more PONs. It traces the history of PONs, have traditionally been broadcast in interest in the technology and it was relates experiences from field imple- nature and exploited the greater band- Keith Oakley of BT Network Strategy mentations, and examines the future width of coaxial cable and line fed am- who set the challenge of building an role of PONs. plifiers to enable a very high degree of experimental telephony system. None of resource sharing. If the same architec- us thought that we would achieve a cost- tural approach could be applied to fibre effective solution for single lines but Introduction then the potential for the loop would be maybe for multiple line services, ISDN enormous, at least for broadcast services. and cable TV the technology could prove In the early 1980s, the technology for itself. long-haul fibre networks made the tran- Work was focused on the design of sin- sition from multi-mode to single mode. gle-mode star couplers as a step forward Our first experimental telephony system This opened up the possibility of almost from those developed for multi-mode over a passive optical network was unlimited bandwidth. Not only was fibre LANs, which tended to suffer loss vari- carried out with existing primary multi- dispersion low enough now to allow ability due to mode selectivity. Much of plexers (PMUXs), which were intended gigabit transmission on one wavelength the credit for this pioneering work on sin- for point-to-point transmission as shown but also seemingly unlimited bandwidth gle mode couplers goes to my colleague in Figure 2 [5]. Whilst the downstream could be made available by wavelength Dave Payne and his team [1, 2]. His idea multiplex of 30 telephony channels could division multiplexing. At that time, costs was to use WDM as a means of gaining be broadcast easily over a PON even were too high to justify extensive use of independence of transmission format with the standard ternary (HDB3) line fibre in the local loop, nevertheless the from one channel to another and to use code, a certain amount of reverse engi- potential of single mode fibre as a ubiq- WDM to enable other logical topologies neering was needed to enable point-to- uitous high capacity transmission to be set up over the network by wave- multi-point transmission in the upstream medium was appreciated. What was length routing. It was therefore necessary direction. The problems were to needed was a way of reducing the costs to produce some experimental HDWDM so that all users of the local loop could prevent signals from separate upstream components to maximise the number of benefit. The approach taken at BT Labs channels from overlapping with each channels available. was to maximise resource sharing so that other; the most significant cost per customer My approach was to investigate a some- ensure channels arriving at the head- was the customer drop and terminal. what less ambitious solution by maximis- end demultiplexer were in separate Until then the local loop had relied on ing the use of digital transmission and byte-wide time slots. point-to-point metallic pairs with origins TDM, which was already successful for over a century ago. A typical topology core transmission but had yet to gain HDB3 encoded signals from separate for the existing access network is shown acceptance for TV or telephony applica- PMUXs would interfere. It was therefore in Figure 1 with flexibility points, lengths tions. This called for less innovative opti- necessary first to revert to un-encoded up to 90 % of the cumulative distribution cal component technology except for the binary transmission on the passive opti- function of cable lengths, and cable sizes, passive power divider. Our first experi- cal network by picking off individual into which fibre access networks would ment in 1986 was the demonstration of NRZ telephony channels from the back need to fit. cable TV with 8 channels of 140 Mbit/s plane to drive the laser at the customer- TV, which led to a multiplex line rate of end. These channels were found con- Loss and bandwidth constraints offered 1100 Mbit/s [3, 4]. This was fed through veniently in separate time slots of one little scope for multiplexing and shared a 64-way power divider to the receivers. byte duration with 2.048 Mbit/s line rate. lines had not been popular because of the lack of privacy. Even though the distribu- tion network already exhibits sharing of the cable duct in a physical tree topology, the wiring inside the cable and duct routes is point-to-point. Introducing both a shared access architecture and fibre transmission was going to be a major culture shock for traditional telcos. Distribution Cabinet Cabinet However, the term passive could be Point, 8-18 lines 250-800 lines 250-800 lines seen as one way of softening the blow since the local loop has always been Switch /CO passive to ease whole-life maintenance 2-4 pair 40 pair 1000 pair 1000 pair costs. Fibre could then be offered as a cable cable cable cable 20000 lines multi-service delivery system, which avoids problems of electrical surges, 100m 900m 3500m EMC and water ingress. Figure 1 Layout of a typical service area Telektronikk 2/3.1999 113

115 6km Head End TDM Customers Terminal Coupler Coupler PMUX PMUX POTs Laser Laser channel Optical Power POTs channels Divider NRZ/HDB3 Receiver Receiver Delay OTDMA Figure 2 Experimental system using PMUXs Secondly, since the round trip delay was over a passive optical network. The new specification a range of 0.1 10 km was not compensated for at this experimental system would have an automatic ranging specified. In the diagram the difference stage, a variable delay line was needed system, 256 bi-directional telephony in round trip delay between farthest and to enable bytes to be delayed at the cus- channels, sufficient power budget for 128 nearest customers is 2(Tmax Tmin). If tomer ends. This is explained below. way splitting and would be implemented the same time slots are to be used for A simple bit-stepping circuit was used to in CMOS technology to save power [6]. both upstream and downstream channels, move the bytes into a valid time slot in This system was designed and built this delay needs to be added to the vari- the multiplex as seen at the output of the under the supervision of John Balance able delay line in each customer terminal head-end receiver. A further expediency who was responsible later for the first to build out artificially the round trip was the use of a NRZ to HDB3 encoder ATM over a passive optical network delay so that the upstream data arrives after the head-end receiver to ensure that (APON) system running at 155 Mbit/s [7]. in its correct time slot. A description of the input signal appeared normal to the the automatic ranging system used to head-end primary multiplexer. Using this measure and compensate for the round arrangement a number of two-way calls Outline of TPON ranging and trip delay is given in [6]. over the passive optical network were pulse amplitude control demonstrated. The diagram illustrates the optical time The need for an automatic ranging sys- division multiplex in the upstream direc- Following this demonstration, ambitious tem is illustrated in Figure 3, which tion as having channel number 3 inactive. plans were put into place to purpose- shows a typical TPON system with byte Hence in NRZ form, the head-end re- build a telephony system and it was at interleaving. Nearer customers will ceiver detects zeros in this time slot. The this time that the acronym TPON was receive the downstream multiplex earlier absence of an upstream line-code and the coined to describe telecommunications than more distant ones. In the TPON variability of upstream pulse amplitudes Propagation delay T min Propagation delay T max Broadcast downstream ax channels Customer ONU-1 0 1 2 3 4 .. .. n Power Divider Reference 1 1 Channel Exchange Power OLT Divider Upstream channels Customer Power ONU-n Divider 1 2 4 n n Figure 3 Diagram showing the need for ranging 114 Telektronikk 2/3.1999

116 according to distance put special con- transmission systems. The device is fab- 1 3 straints on the design of the head-end ricated from two or more fibres [1]. receiver. The receiver design chosen used a zero voltage restoration circuit During construction the aim is to bring and referenced the binary decision the cores into proximity so that the fields threshold from this. In the system design, interact. A furnace with a pulling and 4 2 provision was also made to control the twisting rig is used in the fabrication. amplitude of the upstream pulses by Upon stretching, the fibre cores come Figure 4 Fused bi-conical taper showing remote control from the head-end re- close together and the evanescent fields ports and core interaction region ceiver, which also included too high begin to couple. At a certain distance, and too low thresholds. power transmitted into port 1 or 2 appears at ports 3 and 4, split by 50 %. Further pulling causes total cross cou- Technology for FITL pling at certain wavelengths. The compo- nent then forms the basis of a wavelength Single mode fibre is preferred over division multiplexer. Advantages of this multi-mode because it has lower disper- technology are ease of jointing to trans- A further benefit has been the ability of sion and is therefore capable of higher mission fibre and low excess loss (less the PON to survive adverse weather con- bit rates for a given reach. The lowest than 1/2 dB). A disadvantage is the ten- ditions, such as water ingress and light- dispersion on G652 fibre is in the 1300 dency for the power ratio to be wave- ning damage, at times when the metallic nm band. Although the need for low dis- length dependent. network is returning a higher than normal persion in the access network is not fault rate. paramount, the need for minimum cost Monolithic Power Dividers have waveg- is. By choosing an initial operating wave- uides formed on the surface of a glass Although PONs have proved themselves length of 1300 nm costs of fibre and substrate by ion deposition using fabrica- technically in the field and user reports lasers were minimised. Products were tion methods similar to silicon integrated on TPON have been favourable, some easier to make and more plentiful at this circuits. Using this technology 1 by n practical issues remain which have wavelength. A step index profile with 8 devices can be fabricated (depending limited their widespread use: m core and 125 m cladding charac- upon substrate length). These devices terises single mode fibre. A broad emis- The cost of civil works and technology exhibit good wavelength flatness and sion bandwidth was chosen to suit sim- restricts their use except in new-build triple window operation. Reliable fibre pler, and hence cheaper, buried het- situations where a number of narrow jointing, polarization sensitivity and high erostructure lasers. An emission band band channels are required such as excess loss have been the chief draw- between 1285 1330 nm, 1 mW power might be found in new business parks. backs. output is specified in ITU-T Recommen- Services requiring 2 Mbit/s can often dation G981. Transmitters such as light The power budget available in a typical be met with HDSL, which is a more emitting diodes as used for point-to-point fibre system is 40 dB. Access networks recent technology not available at the access networks could be even cheaper are typically up to 5 km long and fibre time TPON was developed. TPON is but do not launch sufficient power into exhibits an installed loss of around 5 dB more suitable if the distance to be cov- single mode fibre to allow power over this distance. Allowing for 5 dB ered exceeds the range of HDSL. dividers to be used with adequate margin transmission margin, there is around and dynamic range. Demand for broadband services often 30 dB power budget available in the comes from isolated businesses on a access network for passive power divi- Receivers for fibre in the loop have been piecemeal basis. This demand is more sion. specified with less than state-of-the-art obviously met with point-to-point sys- sensitivity to allow low cost technologies tems. The PON architecture is point- such as PIN-bipolar to be used although Experiences with TPON to-multipoint. most suppliers have opted for PIN-FET Although TPON can carry 8 primary which can achieve higher sensitivity. The The first experimental TPON systems multiplexes if used as a point-to-point large power budget available (40 dB or showed that it was relatively straight- system, PDH or SDH technology is more) on optical systems operating at forward to design and build narrowband usually chosen, probably because the bit rates of 155 Mbit/s or less can easily PONs and produce the necessary optics management systems match that being make the short reach from the central and electro-optics for the local loop. The used with existing core and private office with sufficient budget to allow field trial and subsequent use of TPON networks. This may change with power distribution to a large number in quantities of approaching 40,000 lines FSAN APONs [8], which will enable of customers. has also shown no technical problems. greater capacity to be delivered more The key benefit found in practice has flexibly than TPON or SDH. Single mode optical power dividers are been the ability of the PON to act as a available from two family types: fused single multi-service access network Some customers need alternative bi-conical taper and monolithic. A simple offering flexible service delivery of access paths. SDH has already solved two-by-two port fused bi-conical taper POTS, as well as basic and primary rate this problem using dual self-healing coupler has the same functionality as ISDN. Both public and private circuits rings and two alternative paths. Al- an electrical reflectometer, hybrid trans- can be delivered. Prior to this, separate though expensive, it works. The former or return loss bridge in electrical solutions existed for different services. parenting of customer units on two Telektronikk 2/3.1999 115

117 head-ends is possible with PONs via Service was provided with a mixture of amplifier would only offer the benefit of alternative routes. This adds to the PON and ADSL technology for the final extending the upstream range of the PON overall complexity of planning a PON drop. In the case of APON the system to enable it to transmit to a distant head- system, which then becomes multi- was connected directly to the ATM end receiver. point-to-multi-point. switch, whereas ADSL required an addi- tional SDH demultiplexer to spit off the The search for more range, split and Systems now require more capacity signals to the ADSL cards. upstream capacity has now moved a than TPON offers. stage further with work on the ACTS NTT in conjunction with several supp- PLANET project at Alcatel Central Re- APON trials liers are developing APON systems [9]. search Labs. The aim has been to make These systems can provide bi-directional a 2.5 Gbit/s downstream and 310 Mbit/s and deployment switched broadband services and with upstream amplified-PON [12]. An optical additional fibres or WDM can provide split of 2048 and range of 100 km is The APON system is similar to a TPON telephony services (TPON), interactive achievable. Such networks are often system from an optical transmission broadband, and broadcast services referred to as superPONs or transparent viewpoint. It has the same power budget (BPON). As systems evolve, the APON optical networks. To allow amplifiers to but runs at 155 Mbit/s. The system is system should be able to offer all ser- be cascaded in the upstream direction designed to transmit ATM packets which vices in the same multiplex. NTT has semiconductor optical amplifiers (SOAs) are 53 bytes long bi-directionally to carried out multimedia trials of FTTH are used which can be turned on only allow broadcast interactive broadband using PON systems in three areas where when traffic is flowing. SOAs can switch services. These may be variable bit-rate 900 customers are connected. TPON on and off rapidly but fibre amplifiers do or constant bit-rate within the constraints with BPON is installed in Tachikawa not. The upstream media access protocol of the system capacity. and APON is installed in Yokosuka and ensures that only one upstream amplifier Urayasu. Leased service of ATM was attached to a passive splitter is operating BT ran a successful trial of interactive introduced commercially using APON at a time. TV to 2000 homes in and around Colch- in June 1997 for business users. ester using either ADSL links or 155 Mbit/s APON technology. Video and The future role of PONs audio signals were digitally encoded and Optical amplifiers compressed prior to storage on 6 very The future of PONs may be in the use of large multiple access disk drives. Cus- in combination with the FSAN/APON in situations where the tomers used a remote control hand set to passive power dividers capacity of ADSL and HDSL is insuffi- navigate around the menus appearing on cient and access SDH is uneconomic. their TV set. The trial offered a range of The emergence of fibre amplifiers Narrowband PONs would then be super- education, communications, information, opened up new possibilities for both seded by broadband PONs which can entertainment, home shopping and bank- access and core networks. Broadcast deliver both cell based (ATM) and circuit ing services and has enabled BT to learn networks with fibre amplifiers, power switched capacity at primary rate or about both the potential market for inter- dividers, WDM and TDM led to some higher. For this technology to become active TV services and the ability of the very impressive experiments demonstrat- widespread it will be necessary for plan- network to support these services. ing the possibility of national coverage ners to move actively from expedient from a single cable TV head-end with point-to-point solutions to target clusters The trial enabled marketing information huge numbers of channels [10]. The of customers. Both incumbent and sec- to be obtained through the billing of a experimental system offered 40 Gbit/s ond operator networks could be served variety of services. Video on demand did (using 16 wavelengths) over 44 million using this technology. One of the key not provide sufficient revenue alone to way split and 527 km range. Using to- determining factors will be the expected justify the costs but other services in- days MPEG2 video codecs the number penetration and density of customers cluded in the trial such as: tele-shopping, of 2 Mbit/s video channels would be within a given geographical area and the education and other on-line services cou- 22,000. ability of alternative technologies such as pled with fast Internet access may be cost broadband radio to offer similar service effective in the future as the cost of Whilst cable TV systems of this sort are at lower cost. ADSL technology falls. possible, they are not yet competitive with satellite or conventional cable TV. Apart from direct fibre entry systems, The feeder system included SDH (syn- What might tip the balance in favour of PONs may also find application in fibre chronous digital hierarchy) equipment fibre would be the presence of a corre- feeder systems for other forms of cus- to bring the broadband services within sponding upstream path. tomer drop such as hybrid fibre twisted reach of the central office. This was pair (HFTP), coax (HFC) or radio (HFR). configured as a bi-directional ring to Experiments and analysis revealed that Of these HFTP is of interest to incum- give resilience. An ATM (asynchronous noise-funnelling [11] led to a signal to bent operators with a large twisted pair transfer mode) cross-connect allows noise ratio degradation, which was legacy network. HFC appeals to cable customers to be connected to broadband dependent on the number of amplifiers operators and HFR could find a range of services. Upstream signalling was routed feeding the upstream power divider. This applications for new entrant operators if via the same cross connects. limited the upstream split ratio to that both narrowband and broadband access achievable on a single PON with no are required. By choosing a suitable fibre amplification. The presence of a fibre feeder system the option may be avail- 116 Telektronikk 2/3.1999

118 able for future direct-entry fibre systems future. The ACTS FRANS project [13] PONs, perhaps using plastic optical using PONs. has explored the use of hybrid fibre-radio fibre (POF), may find a place in a systems with PONs feeding a number of future generation of LAN using the Achieving end-to-end transparency with base stations. This brings together two power of PCs rather than centralised fibre systems has remained an ultimate flexible technologies for broadband routers to achieve fast file transfer. goal for many operators who recognise point-to-multi-point access. With further However POF needs further develop- WDM as a very powerful service en- development, a choice of fibre or radio ment to achieve adequate range, power abler. Whilst this has been achieved for drop could be offered according to the budget and improved practicality. undersea or core systems, access remains market demand. Radio could fulfil the PONs with high-power upstream lasers a problem because of cost. What is requirements for a profitable entry-sys- As the split ratio of a PON increases, needed is either an evolutionary or a rev- tem, which could later be overbuilt with the duty-cycle of the upstream trans- olutionary approach. At present, evolu- fibre to reduce the demand for radio mitters decreases. This could allow tion seems to favour point-to-point and spectrum and offer an economical dual transmitters to operate at higher peak hybrid-fibre solutions, whilst revolution access system for those customers re- power whilst maintaining the same can be associated with PON. Although quiring additional reliability [14]. average power and so compensate for TPON, APON and BPON have great the loss of the power divider. This ben- appeal, no one entry-system or combina- Whether or not PONs will find a place in efit could allow more power budget tion has yet succeeded in driving invest- the forthcoming data-wave still remains and more ambitious split ratios in the ment hard enough to offer ubiquitous uncertain. The FSAN APON promises future without the need for upstream solution in the UK although APON the next step. If successful, amplified amplifiers. The same principle is used is going ahead for business customers in successors could follow to give greater in TV remote control handsets. This Japan. economy and bandwidth flexibility. arrangement could also allow larger Versions with WDM could also be ex- switches of the type described above. Within the UK, there has been massive pected at that time. investment in broadband access by cable PONs with concentration and/or by operators with HFC passing almost 50 % statistical multiplexing of homes. At the outset of the construc- Stones left unturned When large split-ratios PONs are tion of cable networks, in the mid 1980s, sought, the power budget and capacity In the hunt for a cost-effective fibre PON technology was immature and the requirements also increase. These access network, a number of technical anticipated new market could be more effects conflict. To make better use of features of PONs have been passed-by to economically served by twisted pair and the available bandwidth, concentration enable focused solutions such as TPON coax in the final drops in Siamese and/or by statistical multiplexing may and APON to prove themselves. In the cables. Cable companies have struggled be used. future ideas, which have been shelved, to see return on this huge investment could later prove advantageous if signifi- because their broadband services are In a circuit switched network such as that cant investment in fibre takes place. As competing with satellite and their nar- used with TPON concentration requires already mentioned, there are many possi- rowband services are competing with call by call capacity assignment and bilities for new services via WDM, espe- fixed and mobile access. possibly more user ports than the pre- cially cable TV. Other technical possibil- provisioned approach would allow. ities are outlined below. Incumbent operators would like to enter the mass broadband market but are PONs with active loop-back The issues with a cell based ATM system unwilling to invest heavily in another Experiments were performed at BT are even more complex because of the access infrastructure against an uncertain Labs with TDM PONs, which allowed statistical nature of the cell-based system, market. DSL offers a much less risky loop-back at the head-end so that the which could be studied further. Early approach. upstream channels could be reflected experiments with APONs at BT Labs back into the downstream direction. made use of the Orwell protocol for Second operators may find a use for This topology can turn the PON into a this purpose [7]. PONs in the future. Strategically, owning distributive switch, which is controlled an access network is important and fibre, by time slot selection at the customer- rather than metallic transmission, offers end [15, 16]. Although these experi- Conclusion better future proofing. However, radio ments were successful, the drive for Point-to-point fibre access technology is systems represent an attractive alterna- ATM centralized switching has been already well established for businesses tive to PON for narrow-band services very strong and the use of this tech- and buildings requiring in excess of because no civil works are required and nique probably depends upon centrally 2 Mbit/s capacity. For users requiring bi-directional broadband may soon be switched PONs appearing first. An iso- 2 Mbit/s or less, technologies such as possible for mass deployment via satel- lated LAN has been demonstrated but HDSL can offer a lower cost alternative lite, high altitude aircraft and broadband further work at BT Labs was shelved if the twisted pairs are available. For cellular systems. in favour of industry solutions. new-build situations, point-to-point fibre PONs in LANs and PON offer capacity for future-proof- Whilst second operators favour fibre and Distributed switching has been the ter- ing but more rapid deployment and lower radio systems as separate technologies ritory of LANs with technology such costs may be achieved with radio if suit- now, fibre and radio technology together as Ethernet being capable of address- able radio spectrum is available and ade- in a single system could make a powerful ing different nodes on the network. quate coverage can be achieved. claim for second operator traffic in the Telektronikk 2/3.1999 117

119 The FSAN APON in the future should 6 Faulkner, D W et al. Optical net- Globecom, Sydney, Australia, 812 support both existing narrowband and works for subscriber loop applica- Nov, 1998. Mini conf. on access net- broadband services efficiently at a low tions. IEEE JWLT, 7 (11), 1741 works, 5156. cost relative to point-to-point alterna- 1751, 1989. tives. The future of PONs now depends 15 Smith, P J. A high speed optically upon the successful deployment of this 7 Ballance, J W et al. ATM Access amplified TDMA distributed switch technology in a market segment demand- through a passive optical network. network. In: Proc. ECOC91, 1991. ing capacity which is unsuitable for DSL. Electronics Letters, 26 (9), 1990. 16 Culverhouse, S et al. Optically For new-entrant operators, fibre and 8 Stern, J R et al. Access network evo- amplified TDMA distributed switch radio systems offer a good technical lution to fibre to the cabinet. In: network with 2.488 Gbit/s capacity starting point now, although HFC has Proc. Globecom, Sydney, Australia, offering interconnection to over 1000 been used in the UK in recent years. 812 Nov 1998. Mini conf. Sessions customers at 2 Mbit/s. Electronics The evolution of these networks towards 5 and 6 FSAN, 112. Letters, 28 (17), 16721673, 1992. second operator status with full service capability may be served with HFC or a 9 Faulkner, D W et al. The Full Ser- combination of PON and radio systems, vices Access Networks Initiative. perhaps in a new generation of HFR. IEEE Communications Magazine, 35 (4), 5868, 1997. References 10 Forrester, D S et al. 39.81 Gbit/s, 43.8 million-way WDM broadcast 1 Mortimore, D B. Wavelength flat- network with 527 km range. Elec- tened fused couplers. Electronics tronics Letters, 27 (22), 20512052, Letters, 21 (17), 742743, 1985. 1991. 2 Payne, D B, Stern. Technical options 11 Mortimore, D B. Modelling of opti- for single mode local loops-TDM or cally amplified splitter based net- WDM? In: Proc ECOC, Barcelona, works. In: Proc. European Fibre 1986, 465468. Optic Communications and Networks Conf., Brighton, UK, 1995, 158161. 3 Faulkner, D W. Broadcast Television via Passive Optical Networks. In: 12 Phillips, A J. Redundancy Strategies Proc. ECOC, 1987, Helsinki, for a SuperPON. In: Proc. NOC, 283286. 1998. IOS Press, 1, 7380, 1998. 4 Faulkner, D W et al. Novel sampling 13 Schmuck, H, Heidemann, R. Fibre technique for gigabit cable TV re- Optic mm-wave source configuration ceivers. Electronics Letters, 24, for hybrid fibre radio systems. Proc. 654656, 1988. NOC, 1998. IOS Press, 1, 238244, 1998. 5 Stern, J R, et al. Passive optical local networks for telephony applications 14 Faulkner, D W et al. Opportunities and beyond. Electronics Letters, 23 for broadband access in green and (24), 12551257, 1987. brown field situations. In: Proc. David Faulkner (47) is manager of BTs R&D projects on Broadband Transmission Technology and Broadband Access Standards. He received his first degree in Elec- trical Engineering at Bristol University and his M.Sc. in Telecommunications Systems and Ph.D. in Electrical Engi- neering at the University of Essex. He is involved in running international conferences and is a visiting lecturer at the University of Essex. email: [email protected] 118 Telektronikk 2/3.1999

120 Flexible wavelength multiplexing techniques for broadband fibre access networks TON KOONEN After having reached an established ous increase in the capacity of networks. vide alternative paths to route the data position as the transmission medium In core transport networks, optical fibre streams, thus increasing the networks of choice in core transport networks, provides ample bandwidth for digital sig- availability. The most pronounced advan- fibre is penetrating ever deeper into nals. A single wavelength may carry a tages, however, emerge when the wave- the subscriber access networks. data stream of 10 Gbit/s or more, and length routing can be done dynamically Todays access networks are mostly by deploying multiple wavelengths this in response to varying communication of a hybrid nature. Fibre is used in the capacity can even be stretched beyond needs. Feeder capacity can be directed feeder part of the network, but in the 1 Tbit/s. Ultra-wideband optical fibre to those locations in the network where last drop to the residential user there amplifiers are able to handle more than there is a temporally high traffic demand, is a variety of media, such as twisted 100 wavelength channels in an 80 nm ie. to the so-called hot spots. Thus the pair, coaxial cable, and wireless drops. band in the 1.5 m wavelength window network operator can most efficiently As the capacity demand by the sub- [1]. Commercially available systems exploit the network resources, and hence scriber grows, the capacity in the fibre offer 80 wavelength channels with an maximise his revenues. Moreover, the feeder part has to keep up with it. aggregate capacity of 400 Gbit/s. The wide variety of services and affiliated Wavelength multiplexing is a power- data streams are commonly built accord- transport requirements can be adequately ful technique to upgrade the fibre net- ing to the digital Synchronous Digital met by carrying those services via differ- work capacity without having to resort Hierarchy (SDH), by multiplexing into ent wavelength channels. Separation at to (expensive) new fibre plant installa- transport modules as high as STM-64 the wavelength level by wavelength tion. Wavelength routing techniques (9.95 Gbit/s). To increase the network leasing allows independent service are receiving much interest for path reliability and availability, wavelength providers and different types of services restoration in core transport networks. routing techniques by means of optical to be transported in the same fibre net- Similarly, in access networks wave- crossconnects and add-drop multiplexers work, thus maximising the exploitation length routing provides higher net- are being explored for fast path restora- benefits of this infrastructure. work availability. But even more pow- tion; thus failing or congested links can erful, by dynamic wavelength routing be circumvented. In this paper, two novel hybrid access fibre feeder capacity can be directed to network architectures deploying wave- those spots in the access network The growing need for capacity is simi- length multiplexing will be described. In where there is a temporally high traf- larly felt in access networks. As com- the European R&D programme ACTS fic demand. Thus network resources pared to the core transport networks, [3a], project AC028 TOBASCO has can be optimally deployed in response there is a huge variety in information sig- studied the application of WDM in to traffic conditions, and the operator nal formats and transport media: copper fibre-coax networks; the project AC349 can maximise the efficiency of these twisted pair networks for telephony, co- PRISMA [3b] is addressing WDM in resources. Moreover, different inde- axial cable networks for CATV distribu- fibre-wireless networks. The basic con- pendent wavelength channels offer tion, wireless GSM and DECT networks cepts of these projects will be outlined, independent transmission paths for for mobile telephony, etc. In each of as well as some experimental results and hosting a multitude of service pro- these networks, both the number of users future prospects. viders and/or different types of ser- connected and the range of services vices in one single infrastructure. The offered are growing steadily. The concept of wavelength leasing thus demand for more capacity pushes fibre 2 WDM in fibre-coax provides another way of getting maxi- ever deeper into access networks, closer networks: the mum benefits out of the infrastruc- to the customers. Justification of the tural investment. installation of optical fibre up to the TOBASCO project home still requires considerable cost In this paper, two hybrid subscriber The project TOBASCO (Towards Broad- reductions in optical components and access network architectures applying band Access Systems on CATV Optical fibre installation practices to be made flexible wavelength routing will be networks) aims at upgrading fibre-coax [2]. However, in hybrid access network introduced. Firstly, the application in CATV networks in order to provide architectures fibre has established its fibre-coaxial CATV networks will be broadband interactive services with position in the feeder part of the network; described; this has been developed and speeds in the order of 2 Mbit/s bidirec- from there the signals are handed over to put into a field trial in the ACTS pro- tionally at the customer. The user appli- the variety of media in the last drop to ject AC028 TOBASCO. Secondly, the cations foreseen are tele-working and the customer. In this way Hybrid Fibre- application in fibre-wireless networks tele-learning, by LAN emulation, fast Coax (HFC) networks are delivering will be discussed, which offers even Internet access, video conference, etc. CATV broadcast services, augmented more fertile opportunities for network The project started in September 1995 with narrowband interactive services reconfiguration. The concepts being and finished in October 1998. (voice telephony, ISDN). In wireless net- worked on in the ACTS project AC349 works, fibre is making its way towards PRISMA will be introduced. the base antenna stations. Without affect- 2.1 TOBASCOs system ing the installed fibre plant (in which architecture huge investments have been made), the 1 Introduction increasing demand for capacity in the The CATV fibre-coax infrastructure taken as the starting point is shown in feeder network can be met by introduc- The introduction of new telecommunica- Figure 1. The CATV distributive services ing multiple wavelength channels. Like tion services as well as the increasing (DS) are put on a wavelength 0 in the in core transport networks, wavelength usage of existing ones requires a continu- 1.55 m wavelength window. The ana- multiplexing techniques may also pro- Telektronikk 2/3.1999 119

121 and the huge number of users, the upstream capacity is limited to narrow- TV band services like Plain Old Telephone Service (POTS) and ISDN. coax To enable broadband interactive services 1 1 (BB-IS) delivery to the customer, the 0 OA 0 TV TOBASCO project has proposed the CATV CATV analog analog system concept shown in Figure 2 as an DS HE Tx Rx coax upgrade of the CATV distribution net- work in Figure 1. Clearly, the fibre plant N P installed in the field remains the same, thus protecting the infrastructure invest- ments. Also the analogue CATV distri- Headend Local Splitting Centre Optical Coax user bution system on wavelength 0 is still in Network access netw. place. But on top of that, a number of multi-wavelength optical transceivers Unit Figure 1 Fibre-coax CATV distribution network have been added at the headend in the Optical Line Terminations (OLTs). Each OLT supports a bidirectional 622 Mbit/s ATM-based BB-IS communication chan- nel, for which two wavelengths carry the data in up- and downstream direction across the tree-and-branch split optical logue optical signal is split at several amplifiers to the customer residences. fibre network. ATM-PON technology stages, using unidirectional optical The modern mini-fibre node networks from the ACTS BAF project and its suc- amplifiers (OAs; preferably erbium- may have a number of users as low as 40 cessor BONAPARTE is used [3a]. Eight doped fibre amplifiers, EDFAs) to over- per ONU; together with N = 4 and P = 16 wavelengths are deployed, with four in come the splitting losses. Typically, the this yields at least 2560 users to be fed by downstream and four in upstream direc- split factors N and P are 4 to 16. At each a single fibre from the headend. In tion. The wavelengths are chosen in the Optical Network Unit (ONU), the ana- Hybrid Fibre Coax systems, the unused lower part of the EDFA operation win- logue optical signal is converted to an spectral bands are deployed for convey- dow (1.53 1.54 m), in order to allow electrical one. This RF signal usually lies ing data upstream from the customer easy multiplexing with the CATV distri- in the range of 42 to 900 MHz, and is fed homes to the headend. Because of the bution wavelength in a coarse wave- via a coaxial cable plant with electrical limited amount of spectrum available, length demultiplexer (CWDM). The lat- Network Mgt. 1.. 8 & Control @ 622 Mbit /s ATM Network Mgt. & Control (odd down, even up) control HDWDM 1.. 8 BB APON OLT -switched cable 1 1 TRX contrl. CM TV IS 1 , 2 IS CWDM CWDM coax exchange CATV analog bidir . analog N P Rx HE Tx OA 1:40 LUP CATV 0 0 5 to 31 MHz up DS 42 to > 450 MHz down Headend Local Splitting Centre Optical Network Unit Coax user access network Figure 2 TOBASCOs multi-wavelength upgrade scenario 120 Telektronikk 2/3.1999

122 upstream downstream ONUs ONUs ONUs ONUs 1 1 1 1 1 2 Headend Headend Headend NM&C NM&C NM&C 3 3 3 4 6 OLT OLT 5 5 5 OLT 8 7 7 7 N N a) b) Figure 3 Network reconfiguration a) rearranging the wavelength allocation for symmetric services b) different virtual topologies for the up- and downstream directions of asymmetric services ter wavelength is preferably put in the modems (CMs) enable broadband inter- ate asymmetry in capacity between up- higher part of the EDFA window, in the active data streams to be transported via and downstream direction may be intro- range 1.55 1.56 m. The BB-IS wave- the coaxial cable plant. Up- and down- duced, as shown in Figure 3.b. This may lengths travelling in the same direction stream data traffic streams are positioned be useful for asymmetric services; eg. are spaced at 200 GHz. The upstream in different spectral bands. The upstream for static downstream video broadcast wavelengths are interleaved with the traffic is usually put below the lowest in combination with irregular upstream downstream ones; wavelengths travelling CATV channel (so mostly below 40 customer-generated traffic. in opposite directions are spaced at 100 MHz), and the downstream traffic in the GHz. In the field, the original unidirec- empty frequency bands between the The flexible allocation of wavelengths to tional optical amplifiers in the CATV CATV broadcast channels. the ONUs allows to provide capacity on distribution-only network obviously have demand. Each wavelength channel can to be replaced by bi-directional multi- handle 622 Mbit/s on an ATM packet wavelength ones. At the ONU, the ana- 2.2 Dynamic wavelength basis. As illustrated in Figure 4, when a logue CATV distribution services are assignment certain ONU requires more capacity than firstly separated from the interactive ser- By flexibly assigning the wavelength actually available within its currently vices by a coarse wavelength demulti- channels to the ONUs, the operator is in assigned wavelength channel, it may plexer (WDM). The analogue signal is control of the virtual topology of the be moved, by changing its wavelength treated in the same way as in the original fibre network, without intruding the fibre setting, to another channel where enough CATV network. The BB-IS wavelength plant. He can thus optimise the wave- spare capacity is still available. Several channels are fed by the CWDM device to length grouping of the ONUs such as to strategies may be followed for the wave- a wavelength-switchable transceiver. optimise the network operation effi- length reallocation [4]: a semi-static one Which of the channels is to be processed ciency. As shown in Figure 3.a, each (for failure circumvention or mainte- by this transceiver can be chosen via the wavelength channel may feed the same nance), a long-term one (for determinis- network management and control number of ONUs in case the traffic load tic user behaviour patterns), and a short- (NM&C) system. By providing the is equally spread among them. If, how- term one (for randomly varying user wavelength selection signal from the ever, the load is heavier at particular behaviour). The latter strategy needs to NM&C system at the headend, the net- ONUs, the wavelength allocation can assign bandwidth on demand; it con- work operator can remotely control the be rearranged in such a way that these siderably reduces the blocking proba- wavelength selection at each ONU. ONUs have to share the wavelength bility of the overall network when com- channel capacity with less other ONUs. pared with a fixed wavelength allocation. In the coaxial cable user access network, For instance, in Figure 3.a wavelength The results from a traffic analysis are the cable infrastructure also remains 5 is dedicated to a single ONU, thus shown in Figure 5, where the system largely the same when upgrading the providing it with the full 622 Mbit/s blocking probability is plotted versus the CATV distribution system. The electrical ATM capacity. As the wavelength setting average traffic intensity (ie. the total amplifiers obviously need to be replaced for the transmitter at the ONU may differ traffic load normalised on the total avail- by bi-directional ones. Advanced cable from that for the receiver, even a deliber- able network capacity of 4 x 622 Mbit/s Telektronikk 2/3.1999 121

123 B may be concluded that the dynamic max wavelength assignment strategy extends the capacity available to the subscriber by more than 50 %; or in other words, for ONU x a given subscriber capacity demand fewer wavelength channels will suffice for the operator. Bandwith etc. 2.3 Novel system modules A number of optical modules has been developed within the TOBASCO project. ONU 2 In Figure 6, the circuit layout of the bi- directional optical amplifier is shown. It is composed of two branches. In one ONU 1 branch, the CATV signal (positioned in 0 the 1.55 1.56 m range) is amplified in 1 3 7 downstream direction only in a relatively 5 long erbium-doped fibre, yielding a high wavelength channel output power and a low noise figure. In the other branch, the interactive signals Figure 4 Bandwidth according to need (in the 1.535 1.542 m range) are amplified in a short erbium-doped fibre; together with a gain-equalising filter, a relatively flat gain over the wavelength range is obtained. Two pump laser diodes emitting 120 mW at 980 nm are de- ployed. In this so-called Two-Window 1E+0 Optical Branching Amplifier (TWOBA), coarse WDM devices (CWDM) separate 1E-1 the CATV signal from the interactive signals, and a 1 : 4 splitter distributes the signals to four outlets. The net gain of the 1E-2 TWOBA is 11.5 to 12 dB for the CATV signal (at 5 dBm input level), and P_blocking 1E-3 17 1 dB for the interactive signals. Figure 7 shows a multi-wavelength re- 1E-4 Pb_stat_2M ceiver, which uses a phased-array wave- Pb_dyn_2M guide grating for demultiplexing up to Pb_stat_3M 1E-5 eight wavelength channels; the photo- Pb_dyn_3M diodes have been integrated on the same indium-phosphide (InP) chip. The wave- 1E-6 length-switchable ONU transceiver mod- 0.7 0.8 0.9 1 1.1 1.2 ule pictured in Figure 8 contains 4 wave- Average Traffic Intensity length-specific DFB laser diodes and the control electronics. The use of multiple Figure 5 System blocking probability fixed-wavelength lasers instead of a tune- able laser allows a make-before-break procedure when switching to another wavelength; this eases the data buffering problems during switch-over. A phased- array demultiplexer integrated in planar = 2.5 Gbit/s). It has been assumed that the ONUs, with each wavelength feeding glass and four individual photodiodes are each subscriber generates calls requiring 16 ONUs, the blocking probability for used for the receiver functions; the per- either 2 Mbit/s or 3 Mbit/s. The call 3 Mbit/s calls is obviously worse than the formance of this configuration is still arrival process follows a Poisson distri- one for 2 Mbit/s calls. With the dynamic somewhat better than that of the inte- bution, where the call duration is ex- wavelength assignment strategy, how- grated InP multi-wavelength receiver. ponentially distributed. Each ONU ever, the blocking probabilities are sig- connects 40 subscribers; in total there nificantly lowered. As shown from the Advanced cable modems have been are 64 ONUs, fed by 4 wavelengths in graphs, by using this strategy the block- developed deploying synchronous either direction which each can handle ing probability for 3 Mbit/s calls is even CDMA techniques, for optimum robust- 622 Mbit/s on an ATM basis. When the lower than the one for static wavelength ness against ingress noise, in combina- wavelengths are statically assigned to assignment with 2 Mbit/s calls. Thus it tion with QAM-16 (or -64) techniques 122 Telektronikk 2/3.1999

124 Bi -directionality for IS-signals pump 1 CWDM Er3+ Er3+ CWDM -filter pump 2 1.5 m 50/50 1:4 Er3+ coupler a) Uni -directionality for CATV b) Figure 6 Two-window optical branching amplifier a) circuit layout, b) module with control electronics and optical parts for better spectral efficiency. Per 6 MHz frequency slot a data signal of net 8.2 Mbit/s can be accommodated. 2.4 Cost analysis In co-operation with the ACTS OPTI- MUM project, a first cost comparison of the TOBASCO system with alternative system architectures has been made. These alternatives were a Space Division Multiplexed (SDM) architecture, and a static wavelength multiplexed (static WDM) architecture. In the SDM one, multiple fibres are used instead of mul- tiple wavelengths. In the static WDM one, the ONUs were not wavelength- flexible, and hence were only equipped with a single laser diode. The installed first costs have been assessed, and their Figure 7 Integrated evolution in time, under the assumption multi-wavelength that the fibre network is already in place receiver and that the prices of the various modules decrease due to growing market volumes. The study results are shown in Figure 9. Although at the present price levels the TOBASCO solution is clearly the most expensive one, the cost difference with the other solutions will tend to diminish within the next ten years, mainly due to the growing market volumes and the price erosion of optical components expected. Note, however, that the TOBASCO solution is the most versatile one. 2.5 Field trial The TOBASCO system was put into a field trial in the city of Ghent, Belgium, for six months (from 15 April to 15 October 1998). The system was installed in a part of an operational fibre-coax Figure 8 Wavelength- CATV network run by TeveOosT, the switchable transceiver regional CATV operator. The network at ONU Telektronikk 2/3.1999 123

125 1.4 upstream per subscriber. Measurements of the round-trip times for 64-bytes Cable Modem system packets showed that the TOBASCO 1.2 Network system performance is close to that of a 10 Mbit/s Ethernet LAN. The user ONU Installed First Costs (a.u.) 1.0 experiences with the system were HE/LSC evaluated by means of questionnaires. 0.8 The results showed that about 80 % of the users were very satisfied with the speed of the system, and would favour it 0.6 above their own PC connections at home or in the office. 0.4 3 WDM in fibre-wireless 0.2 networks: the PRISMA 0 project dyn. stat. dyn. stat. dyn. stat. In wireless access networks, the users WDM WDM SDM WDM WDM SDM WDM WDM SDM with their mobile terminals are connected 1998 1998 1998 2003 2003 2003 2008 2008 2008 to antennas at the base stations via microwaves. The information transport to Figure 9 Evolution of Installed First Costs per user, estimated at five-year intervals and from these base stations is carried via a fixed wired network, in which fibre is steadily penetrating ever deeper. The traffic load at the base stations may fluctuate even stronger than at the ONUs in the fibre-coax network considered before, as users not only vary their communication needs in time but also in Internet space when moving around. Having the ATM Cable Cable User advantages of flexible wavelength alloca- Switch Controller Modem PC tion in fibre-coax networks in mind, the ACTS project AC349 PRISMA (Pho- Router WDM- APON tonic Routing of Interactive Services for Mobile Applications) studies the aspects of flexible multi-wavelength techniques Local Servers in fibre-wireless networks. The project TOBASCO started in April 1998, and will finish end of January 2000. IP packets In general, two types of mobility can be Format discerned: IP SAR ATM Cells Converter Cells SAR Ethernet Frames Nomadic services, where connections are set up from a certain location, run AAL5 Frames/RFC1483 Encapsulation for some time and are subsequently torn down; the user moves to another location and sets up a connection Figure 10 IP bridging over ATM (according to IETF RFC 1483) again, etc. These services usually have a broadband nature, such as wireless LAN applications for nomadic com- puting, and do not employ in-travel communication with its associated call hand-over issues. management and control information was ATM optical network (WDM APON), IP Mobile services, where users roam transported via some spare fibres in the packets could be transferred as shown in throughout the network while staying cables. The interactive services (fast Figure 10. The system was used by stu- in connection; this entails call hand- Internet, fast file transfer, video confer- dents, university staff members, sales over processing. encing, etc.) were offered by the Uni- persons in a travel agency, and visitors at versity of Ghent, where an ATM switch a local Internet caf. The operation of the Roughly speaking, mobility is traded connected to file servers and to the Inter- system was quite successful. Monitoring against capacity. For nomadic services, net was located. Via the coaxial cable of the performance showed IP peak rates high information transfer rates (several modem system and the multi-wavelength of 6.2 Mbit/s downstream and 3.6 Mbit/s 124 Telektronikk 2/3.1999

126 1 2 ONU BSC 3 3 OLT 4 4 hot 5 5 fibre spot 6 Base * = Base 7 Station Transceiver Controller Station a) b) Figure 11 Wiring a microwave cellular network with optical fibre, using multiple wavelengths a) feeding the Base Transceiver Stations (BTSs) b) wavelength allocation among the ONUs of the BTSs Mbit/s) are needed, which implies a The PRISMA project is mainly targeting 3.1 PRISMAs system smaller cell size. Also hand-over pro- nomadic applications, run from eg. port- architecture cessing gets more difficult, and hence able PCs. Three user scenarios are being Basically, a similar fibre feeder network mobility is limited to low speeds. In con- explored: university campus (eg. group as in TOBASCO is considered to feed trast, mobile services requiring lower bit- exercises), hospital (eg. bedside diagnos- the Base Transceiver Stations (BTSs) [5]. rates (10 kbit/s for GSM, up to 2 Mbit/s tics, remote X-ray pictures retrieval), The mapping on a microwave cellular for UMTS) allow a much larger cell size. and industrial park (eg. flexible office network is illustrated in Figure 11. To Hand-over processing is easier, and as rooms). avoid interference problems, each cell the hand-over frequency is reduced by needs to use a microwave frequency the larger cell size at a given user move- which differs from the one used at the ment speed, the mobility is higher. Network Mgt. Network Mgt. & Control & Control 1.. M IS 1.. M 1 1 control exchange portable HDWDM -switched base BB APON OLT TRX station PC IS 1 , 2 bidir N P OA Local Exchange Local Splitting Centre Fibre-Wireless Wireless mobile and Base BTS terminals Station Controller Figure 12 PRISMAs system architecture Telektronikk 2/3.1999 125

127 neighbouring cells, which implies that at least 7 different frequencies are needed (indicated by different colour shadings in Figure 11.a). The frequency pattern of a fibre 7-cells cluster can be repeated to cover a Network larger geographical area. The fibre net- optical BSC work feeding the BTSs can be laid out as -layer a split fibre network, where the BTSs are accessed via their Optical Network Units. When the traffic load in a cell becomes very high (ie. if a cell becomes a hot spot), an extra microwave carrier may be switched on to handle it. At the ONU extra feeder capacity needs to be made Network electrica available, and for this a flexible wave- -wave layer length allocation strategy like in TO- BASCO can be deployed. In contrast, a larger region with less intense traffic may be covered by a macro-cell, served by a single microwave carrier and BTS. As Figure 13 Mapping microwaves to wavelengths shown in Figure 11.b, the wavelength channels can be allocated among the ONUs in such a way that the varying demands are adequately met. The architecture of the PRISMA system Depending on the traffic load per cell, the system. In principle, the same alloca- is depicted in Figure 12. In a similar way the mapping can be adjusted. The inter- tion strategy as depicted in Figure 4 can as in the TOBASCO system, the wave- position of application filters, mobile- be followed. The results of a first analy- length allocation at the BTSs can be aware agents and proxies at the wireless- sis are shown in Figure 14, where the remotely controlled from the central local fixed network interface may even lead to system blocking probability has been exchange/base station controller site. higher traffic loads at the ONUs, affect- plotted versus the normalised traffic load, This allows the operator to optimally ing the mapping choices. assuming 7 wavelength channels in the direct the networks capacity resources to fibre network in either direction (ie. the the BTSs in such a way that the locally total traffic load offered at the BTSs, nor- offered traffic load is adequately met. 3.2 System performance malised on the total available network The total splitting factor in the optical analysis capacity of 7 x 622 Mbit/s = 4.35 Gbit/s). network, enabled by the bidirectional Traffic studies have been made to assess The call arrival process is assumed to fol- amplifiers, may be as high as 73 = 343, the impact of the wavelength reallocation low a Poisson distribution, and the call corresponding to three stages of 1 : 7 strategy on the blocking performance of duration a uniform distribution and band- split (the factor 7 allows easy mapping on the cellular network topology). M = 8 wavelength channels are to be used, of which 4 are for the upstream and 4 for the downstream direction. Per wave- length, 622 Mbit/s ATM-based data can 1.000 be transported. At the wireless end, the HS = 49 system from the ACTS project AC085 WAND [3] is used. This system provides static WDM, Blocking Probability 20 Mbit/s capacity per microwave carrier 0.100 all HS on 1 on an ATM basis (total of up- and down- stream traffic), to be shared by the dynamic WDM mobile terminals connected. It can support as much as 5 carriers in the 5 GHz region, and deploys Orthogonal 0.010 Frequency Division Multiplexing static WDM, (OFDM) technology. 7 HS per The mapping of the microwaves in the 0.001 individual cells to the wavelength chan- 0.4 0.5 0.6 0.7 0.8 0.9 1 nels in the fibre network is flexible, and can be changed by reallocating the wave- Normalized Load lengths at the BTSs. For example, a par- ticular wavelength allocation may yield Figure 14 Impact of wavelength reallocation on the system blocking probability the mapping as shown in Figure 13. in the presence of hot spots (HS) 126 Telektronikk 2/3.1999

128 width. Hot spots (HS) are defined as cells among the wavelength channels. The the end-to-end IP-in-ATM connectivity where the call arrival rate is twice that of latter strategy (load balancing) may using the flexible wavelength allocation a normal cell; it is assumed that there are imply a less complicated wavelength strategy. This set-up will integrate a 49 of these hot spots in the network. allocation process, and thus may be eas- number of novel and adapted optical When the allocation of the wavelength ier to implement. modules, including the wavelength- channels to the BTSs is fixed, the block- switching functionality, and will be ing probability is obviously the highest the basis for the field trial. if all hot spots are located such that they 3.3 System trials are served by the same wavelength. The The PRISMA project builds on the flex- The project is planning to have a small- blocking probability is the lowest if the ible multi-wavelength ATM PON system scale field trial during the last three hot spots are evenly spread among the realised in the TOBASCO project, and months of 1999 in Ghent, Belgium. wavelength channels, ie. 7 hot spots per the wireless ATM system for nomadic Facilitated by the IMEC group at the wavelength. In reality, it is unknown computing from the WAND project. In a University of Ghent, the potential of the where the hot spots will occur, so for a first laboratory set-up, a successful inte- system will be assessed in a realistic user fixed wavelength allocation the blocking gration of both systems has been accom- environment. The layout of the trial net- probability may vary anywhere between plished. In a LAN emulation (Ethernet work is shown in Figure 15. Within the these two extremes. Following the over ATM) set-up, error-free transport of university premises in the new Urbis dynamic wavelength allocation strategy, data packets has been achieved between building complex, network interconnec- however, the capacity is directed to the a server PC at the headend and a user PC tions will be made with students, locations where traffic loads demand it. at the wireless mobile terminal, at data researchers and other university staff This results in the blocking probability rates of 4 Mbit/s downstream and members via an IP-in-ATM LAN curve labelled dynamic WDM. Clearly 2 Mbit/s upstream. Later on in the pro- testbed, which comprises an ATM switch this strategy yields the lowest blocking ject, a more comprehensive laboratory and a router providing fast Ethernet probability, independent of the actual system is planned, with a network man- access to the backbone network of the location of the hot spots, and thus allows agement and control system overlooking university. The university network (RUG the network operator to employ its capac- ity resources with optimum efficiency even for unknown hot spot distributions. The graphs also suggest that a second best strategy is to equalise the traffic load over the various wavelengths, in such a way that the hot spots are equally spread Technicum 1 out of 6 fibres ONU/BS RUG net 1000 Mbit/s Academic Computer "Urbis" complex Centre 3 fibres IBM 8265 OLT ONU/BS server IP router 100 Mbit/s ATM 155 Mbit/s IP router switch interface portable PC BELNET 34 Mbit/s Figure 15 PRISMAs field trial Telektronikk 2/3.1999 127

129 Net) has a high-speed connection to the ing). Both in the laboratory and in the References Internet via the BELNET national field, the TOBASCO project has proven research network. In the Urbis complex, the technical feasibility of the flexible 1 Sun, Y et al. A gain-flattened ultra a number of lecture halls and laboratory multi-wavelength concept for upgrading wide band EDFA for high capacity rooms will be connected, providing a hybrid fibre-coax CATV network WDM optical communications sys- broadband wireless access to the univer- towards the provisioning of broadband tems. In: Proc. of ECOC98, Madrid, sitys network. By means of a fibre run- interactive services. A number of novel Sepember 1998, vol. 1, 5354. ning via the Academic Computing Cen- system modules have been realised and tre, the same access capabilities are successfully tested. Traffic studies have 2 Ferrero, U. EURESCOM Project offered to students and university staff at pointed out that a significant increase in P614 : Implementation strategies for the Technicum building. A number of network efficiency can be obtained by advanced access networks. Telek- businesses are also located in the Urbis deploying flexible wavelength assign- tronikk, 94 (2), 8083, 1998. complex, such as a computer ment to the optical network units, where- hardware/software supplies shop, music/ as a first analysis of the economics shows 3a European Commission DGXIII-B, video shops, a supermarket, and some that the extra costs incurred in compari- ACTS 97 Project Summaries, Ref. other traditional mall shops. In addition, son to alternative (less flexible) network No. AC971392-PS. ISBN 92-827- a consultancy company has moved in and architectures are only marginal. In fibre- 0997-3. more companies are expected to follow. wireless networks, as studied in the The services planned to be offered in the PRISMA project, even larger increases in 3b European Commission DGXIII-B, trial network are fast Internet access, network utilisation can be gained with ACTS 98 Project Summaries Sup- video conferencing, remote access to the flexible wavelength allocation strat- plement covering 3rd Call, Ref. No. the universitys Academic Computing egy. First traffic studies show that local AC980872-PS. ISBN 92-828-4212-6. Centre, etc. temporal concentrations of traffic load Website (hot spots) can be handled very effec- tively with a limited set of resources. The 4 Koonen, A M J et al. TOBASCO : 4 Conclusions projects focus is on nomadic broadband an innovative approach for upgrading applications (such as nomadic computing CATV fibre-coax networks for After having proven the benefits for on portable PCs); the scope of the studies broadband interactive services. IEEE capacity increase and traffic routing in will be extended to mobile UMTS appli- Communications Magazine, 35 (4), core transport networks, wavelength mul- cations shortly. 7681, 1997. tiplexing techniques have a large poten- tial to extend the capabilities in optical 5 Koonen, A M J et al. Reconfigurable access networks also. Flexible wave- Acknowledgements broadband fibre-wireless network length multiplexing allows the network employing dynamic wavelength allo- operator to deploy his resources in the The author would like to thank the part- cation. In: Proc. of ECOC98, most efficient way, by adapting his net- ners in the TOBASCO project and the Madrid, September 1998, vol. 1, works virtual topology in order to pro- PRISMA project for their contributions, 577578. vide capacity on demand; he may also and the European Commission for partly provide wavelength channels as indepen- funding both projects in the framework dent transport pipes in order to offer a of the ACTS programme. wide variety of services and to host other independent operators (wavelength leas- Ton Koonen (44) is Technical Manager in the department For- ward Looking Work, part of Bell Laboratories, Lucent Tech- nologies, located in Huizen, The Netherlands. He has been working in optical fibre communication since 1979. His cur- rent field of interest is broadband access networks. He was project manager of the ACTS TOBASCO project, and cur- rently manages the ACTS PRISMA project. He has also been part-time professor in Photonic Networking at the Uni- versity of Twente, Enschede, The Netherlands since 1991. email: [email protected] 128 Telektronikk 2/3.1999

130 Inverse multiplexing EINAR EDVARDSEN This paper describes a new alternative buried in the ground. Even in cases were with one hundred twisted pairs is in this for transforming the existing telephone such ducts do exist, the cost of fibre context a small cable. The graphs indi- access network into a real broadband cables and related installation work may cate that within reasonable cable lengths, network. The objective is to introduce be considerable and it will take time to the capacity of one hundred twisted pairs a low-cost network evolution scenario, accomplish time that telecom operator is in the range of several gigabits, ie. a which makes maximal use of the huge companies possibly might not have. capacity that today is far beyond what investments connected to this kind of New competitors entering the market are is needed in residential access networks. networks. The existing network is sub- offering broadband access over power- The lower curve indicates the capacity of divided into small cell networks, each lines, radio, satellites and CATV net- 100 copper pairs using the performance covering an area with an approximate works. To maintain their market shares, of standard modems. The upper one indi- radius of 1 kilometre. Inverse multi- the owners of the telephone cable net- cates the theoretical upper value for the plexing of xDSL systems is used to work will have to meet this challenge aggregated capacity. aggregate bandwidth in the 100 to by upgrading their networks. 1000 Mbit/s range between nodes in These brief calculations are of course not the network. VDSL modems able to exact. Noise and cross talk influence on carry about 25 50 Mbit/s, are used 2 The capacity of tele- the obtainable bandwidth, resulting in on the last section from the node to the phone cables less performance. Nevertheless do the customer premises. These modems results indicate that the telephone cables provide all households within each Though each single twisted pair in a tele- are much more powerful than we are geographical cell with real broadband phone cable has a limited bandwidth, the used to believe. The question is whether access to a large common bandwidth. total capacity of a telephone cable can the bandwidth of this network can be ex- nevertheless be very large. This is due ploited in an efficient way, thus opening to the fact that telephone cables contain a new possibility of how to provide real 1 Background large numbers of copper pairs, each of broadband access to the general public. which by utilising xDSL technology may At the present time there is a common provide a relatively large bandwidth over understanding that the vision of Fibre-To- a certain distance. If the capacity of each 3 Inverse multiplexing The-Home (FTTH) still belongs to an single pair could be added up, the aggre- Inverse multiplexing is used to aggregate undefined future scenario. This is both gated capacity would be huge, as seen bandwidth from a number of lines with due to the fact that the demand for broad- from Figure 1. Inverse multiplexing is smaller bandwidths. The principle of in- band services has not yet matured, that the technology that enables such aggre- verse multiplexing is shown in Figure 2. the related investments are supposed to be gation of capacity from a number of very high, and that such huge civil works lower bandwidth digital channels. Figure 2 shows the basic principle behind will take a long time to accomplish. Other inverse multiplexing. The incoming traf- intermediate solutions have therefore The figure shows the bandwidth of a tele- fic stream from the left is sequentially been looked upon as more realistic. phone cable with 100 twisted pairs as a distributed over a number of lower bit function of line length. A telephone cable Various adaptions to the Fibre-To-The- Xx (FTTX) architectures have therefore been given more attention. FTTC (Fibre- To-The-Cabinet) is one of the approaches adopted by the FSAN (Full Service Access Network) group. The idea behind FTTC is to utilise the existing telephone access network from the cabinet to the customer, but use optical fibre cables on sections between the cabinets and the 100 broadband switch. By doing it this way, optical fibre cables will only be installed Shannon on sections were the costs can be shared Standard xDSL by large numbers of customers, while the Gbit/s existing infrastructure will be used in the other parts. In practice this means that 10 ADSL or VDSL technology is supposed to play a major role in the access network in order to make available the necessary bandwidth on the sections between the cabinets and the customers. Even if an FTTC architecture is flexible 1 and less expensive than the FTTH archi- 0 0.5 1 1.5 2 tecture, it still involves significant invest- km ments. In areas were cable ducts do not exist, optical cables will have to be Figure 1 Capacity of a telephone cable as a function of its length Telektronikk 2/3.1999 129

131 xDSL lines rate lines and reassembled at the other plexing of xDSL modems is therefore so end. Inverse multiplexing is standardised far not standardised. There are several 28.4 Mbit/s by a number of standardisation organisa- methods for performing inverse multi- 34.2 Mbit/s tions, such as ITU, ETSI and ATM plexing of packet based traffic (ATM), 148 148 Forum. The available standards cover for instance by the use of sequence num- Mbit 27.8 Mbit/s Mbit inverse multiplexing of channels with bering, as illustrated in Figure 3. 29.5 Mbit/s equal bandwidth, for instance inverse multiplexing of four times 2 Mbit/s to Each cell arriving from the left is given a 28.1 Mbit/s form a channel of 8 Mbit/s. Although sequence number before it is transmitted from a technical point of view it is also over one of the available lines. The band- Figure 2 The principle of inverse multiplexing possible to multiplex lines/channels run- width of each line may differ from line ning various bit rates, it is not covered by to line. One line can perform 13 Mbit/s, any official standards yet. Inverse multi- while another can perform 19.3 Mbit/s. The cells will therefore arrive out of order at the receiver end, and have to be intermediately stored in the buffer on the right side in order to re-establish the sequential order. Since it is mandatory for inverse multiplexers to be transparent for traffic streams, and the ATM cell itself does not contain any field useful for carrying the sequence number, it must be 3 transferred as a tag to the cell. The cell length will therefore deviate from the 5 ATM standard (53 bytes). However, the 4 interface between the two terminals of the inverse multiplexer can be looked SEQ. 7 upon as an internal interface, and a devi- ation from the standards is therefore 1 acceptable. Mixing of lines with different 6 bandwidths may create delay and cell delay variation (CDV), which are unac- 2 ceptable for certain types of traffic. To meet requirements from the various traf- request fic types, each of them may have to use dedicated line groups with properties that match their demand. As mentioned above, the method of using sequence Figure 3 Inverse multiplexing by use of sequence numbering over xDSL lines numbering to perform inverse multiplex- ing of lines with different bit rates is only one way to do it. Other methods also exist. A method called Deterministic Inverse Multiplexing is illustrated in Fig- ure 4. The traffic coming from the left is distri- buted over the four included lines in the same relation as their bit rates. In the 9 above example, two units of data are sent 9 Mbit/s over the 2 Mbit/s line, six units over the 6 Mbit/s line, 18 units over the 18 Mbit/s 18 line, and nine units over the 9 Mbit/s line 18 Mbit/s during the same period T. The method requires that the lines are phase-locked to 6 Mbit/s 6 each other. Compared with the first described method, the latter one is in- dependent of data format. It can transfer 2 Mbit/s 2 ATM as well as bit synchronous traffic. T Figure 4 The principle of Deterministic Inverse Multiplexing 130 Telektronikk 2/3.1999

132 4 The network concept they will complement each other in a IDMU way that both the operators and the pub- The network concept is based on using lic may profit from. IMA-DSL to lower level node the existing telephone access network infrastructure. In the general structure of the new network, new nodes (Figure 5) 5 The ACTS project IMA-DSL to higher level node V D IM IMA-DSL to lower level node will have to be installed in connection AC309 ITUNET A- D L L- M U with the street cabinets. These network X nodes must perform both inverse multi- To promote the network concept and to VDSL to end users plexing and statistical multiplexing. Fig- pave the way for commercial products ure 5 contains a functional description of needed to implement this kind of net- such a network node. work, the ACTS project AC309 ITUNET was initiated. The project started in Control & Management Figure 6 shows the principal structure of March 1998 and will continue for two the new network. The nodes in the net- years. One of the main objectives of the Figure 5 A network Node work are placed geographically close to project is to study how the existing the cabinets in the old telephone network access network infrastructures can be in order to have easy access to the neces- upgraded using xDSL technologies to sary copper pairs. The distance between form a cost-efficient integrated service the nodes should not be more than about network providing the necessary capacity two kilometres. Each node covers an area and functionality for broadband services. with an approximate radius of one kilo- By conducting field trials the project metre, which makes it possible to provide seeks to prove the viability of the 25 50 Mbit/s to each of the customers. adopted network topology for the realisa- multiplexing and statistical multiplexing. Inverse multiplexing over a number of tion of an end-to-end ATM network pro- The equipment will be used in various copper pairs running VDSL modems, is viding a selection of broadband applica- field trials aiming to visualise the poten- used to aggregate the requested band- tions. Through the trials, ITUNET will tial usage areas of the technology. Three width between the main ATM switch gain valuable experience, both with areas are identified: located at the telephone exchange build- regard to installation and operation of the In a Full Service Access Network ing, and nodes in the network. With proposed platforms, and to the end-users based upon the existing telephone access lines of 25 50 Mbit/s the needed perception of the supplied broadband ser- access network; bandwidth between the nodes will be in vices. the 100 1000 Mbit/s range. 100 As above, but in combination with 1000 Mbit/s can be obtained by inverse One key component in the proposal is the other fibre technology; multiplexing up to a few tens for lines. development and evaluation of a network In corporate networks. Each network node performs both in- node being able to perform both inverse verse multiplexing and statistical multi- plexing of traffic flows from the individ- ual users. The users are connected to the nearest node by VDSL modems, thus giving them access to a large aggregated bandwidth, which has to be shared among them. The nodes will be equipped with signalling according to standards, thus enabling SVCs (Switched Virtual Connections) with QoS (Quality of Ser- vice) as defined by relevant standards. Distributor cabinets The structure of this network is more or less identical with the recommendation from the FSAN (Full Service Access Network) consortium. The difference is that FSAN recommends optical fibres Telephone between the nodes, while this approach is Exchange based upon copper. However, due to the similarities between the two concepts, it is easy to adopt the technology that best fits the need in each case. On sections were optical cables can be installed at low cost, fibre cables are the natural choice. But on sections where optical Areas covered cables cannot be installed at a reasonable by each distr. cabinet cost, inverse multiplexing is the choice. The two approaches go hand in hand Figure 6 The architecture of the new network Telektronikk 2/3.1999 131

133 In order to demonstrate services in the consequence of using bit rate adaptive VDSL modems, which are a basic compo- network, ITUNET has chosen two ser- modems. The capacity of the physical nent both in this concept and in the vices that will be partly developed in the lines may vary over time due to varying FSAN recommendation, are under stan- project: line conditions. The project also aims to dardisation and partly available today. perform in-depth calculations and evalu- Switched broadcast TV (SBC-TV) ation of the cost related to establishing an Inverse multiplexers adapted to the stan- based upon use of the UNI 4.0 sig- access network based on DSL. An evalu- dardised transmission systems (T1, T3, nalling protocol (multi-cast); ation with regard to service provision in ...) are available. With regard to inverse Video Surveillance enabling a number the proposed access network will also be multiplexing over DSL technology, this of video channels to be multiplexed performed. kind of equipment is not available today. and transferred over the network. However, work has begun to find indus- trial partners being interested in making The new network concept will be a rather 6 Evaluation commercial products. complex network. Signalling and re- The general complexity of a network source management may have to be A case study is being performed in order node is comparable to what is imple- implemented further out in the network to have a rough evaluation of the imple- mented in existing ATM switches and than todays recommendations prescribe. mentability of such a network and the multiplexers. The concept does not The technology also makes it possible to cost to establish it. A local network in a involve unknown technology. All the utilise Adaptive Asymmetrical Systems residential area (NorVillageA) has been basic technology is more or less avail- (AAS) a method of adapting upstream chosen for this purpose. Figure 7 shows able. VDSL modems and inverse multi- and downstream capacity to the actual an overview drawing of the converted plexing technology are key elements in traffic profile. The management of Float- network. NorVillageA has 728 telephone the network concept. The present status ing Transmission Capacity (FTC) is also subscribers. 384 of them live closer to for these elements is: an important issue to manage. FTC is a the telephone exchange building than A3-5 A3-25 3200m H1-13 H1-1B nnHb/55p 41 user nnHb/49p nnHb/29p nnHb/17p 1000m 1230m 1350m 17 user H1_A H1 5 user 16 user nnHb/93p nnHb/98p 93m 800m H1-1 nnHb/18p 0 user 17 user 1550m nnHb/26p 11 user 1080m A3 10 user A3-33 nnHb/123p 975m GI-15 28 user D1-2 D1-20 D1-20-1 nnHb/39p nnHb/52p nnHb/43p nnHb/50p D1-7 910m 1000m 400m 1390m GI nnHb/84p 17 user 21 user 25 user 26 user nnHb/97p 400m GI-4 31 user D1-7 D1-35 600m 42 user nnHb/45p nnHb/10p nnHb/26p 1400m 1040m 2570m 24 user 38 user 9 user Exchang_bulding 384 user < 1000 m ITUNET node Figure 7 An overview drawing of the converted network 132 Telektronikk 2/3.1999

134 1,000 metres, thus one can provide 25 50 Mbit/s to each of them with single user VDSL modems. The remaining subscribers, 244, must be connected via inverse multiplexed equipment. The case study revealed that: 1 It is not possible to place the new nodes at optimal locations. The nodes have to be installed at central points in the cable network, ie. at cabinets which form distribution points in the cable network structure. As a result of this, the number of network nodes will Figure 8 The FSAN (left) and ITUNET (right) network structures be higher than expected. 2 Relatively few users will be connected to each network node, ie. less statisti- cal gain. 3 Due to non-optimal location of the net- work nodes, the case study shows that The ITUNET concept uses the existing An optical cable has more capacity one needs up to 4 cascaded nodes in local telephone cables between the net- than a telephone cable, but the band- order to reach the most distant users. work nodes. Investments as mentioned width of the latter will in most cases be above are not relevant, but inverse sufficient. Bandwidths in the range of multiplexers (an integral part of the Gbit/s can relatively easily be 7 Comparison of the remote node) are used to establish the achieved. FSAN and the ITUNET needed bandwidth between the nodes. None of the concepts need to influence concepts The number of VDSL modems an on the telephony service. Both ana- often heard objection against the logue and digital telephony can be The technology and the network concept ITUNET approach. The modems used transmitted in a separate frequency of ITUNET are introduced as an inde- for inverse multiplexers are a common band on the twisted pairs. pendent strategy for building a broad- resource for a number of users, and in The complexity of the remote network band access network. The concept is well total it is expected that the ITUNET nodes in the two cases will differ adapted to other approaches, such as the concept only will use approximately slightly since inverse multiplexing will FSAN approach. These two strategies are 10 15 % more modems than the have to be integrated in the ones used complementary in the sense that one eas- FSAN strategy. for the ITUNET concept. ily can combine sub-nets based on the two technologies without large influence on the planned network structure. Both of the two approaches rely on utilisation of the existing telephone network. Optical fibres can supplement a network based upon the ITUNET ideas, as well as in- verse multiplexers can supplement a net- work based on the FSAN concept. Since there are a number of similarities be- tween the FSAN and the ITUNET con- cept, it is of special interest to compare them. The principal structure of the two networks are shown in Figure 8. In summary, here is a very brief over- view of the two concepts: The FSAN concept is based upon fibre Einar Edvardsen (54) is research scientist at Telenor R&D, optic communication between network Kjeller, Broadband Technology Group. His current focus nodes. Installation of optical cables is on broadband provision over existing telephone cable often involves large investments both infrastructure. He is the project manager of the EU project in optical cables and civil work. The AC309 ITUNET. His research interests are technologies connected to network architectures and evolution. operation, administration and mainte- nance costs however, are assumed email: [email protected] to be very low for fibre compared to copper. Telektronikk 2/3.1999 133

135 Transmission on power line cables DERMOT COLLINS, LUCIEN BUDRY, GBOR GERDAI, ANGELANTONIO GNAZZO AND AHMED MADANI 1 Introduction 2 Power network structure 3 Power cable charac- Transmission on power line cables is The architecture of the electrical power teristics and trans- currently the subject of much research network is one of the most important fac- mission techniques and development activity, constituting tors in determining whether it is suitable a potentially very attractive alternative for data transmission. Figure 1 shows the The suitability of a power distribution access technology as the power utility generic European power network struc- network for PLT use is affected by the companies are considering to enter the ture. physical characteristics of the cable. A residential telecommunications market. number of different cable types may be Power Line Communications refers to Power is transmitted to customers via the found. Typical cable construction for two the transmission of data over the low high voltage transport network at 220 kV types of modern distribution cable is voltage electricity distribution network. 800 kV and the low voltage distribution shown in Figure 2. Figure 2a shows a This is achieved by injecting the data sig- network at 220 V 240 V. The low cable with sector-shaped conductors and nal at the low voltage transformer via a voltage distribution network is used for Figure 2b shows a cable with circular power line modem. Another modem in PLT. In urban areas in European coun- conductors. the customer premises extracts the signal. tries approximately 200 homes are served The transformer serves a number of by the final transformer and the distance Practical testing on low voltage distribu- houses and consequently, the data signal from each house to the transformer is tion networks in the UK has shown that is shared between that number of cus- usually from 50 m to 400 m. The distri- for the frequency range 1 to 10 MHz the tomers. Two types of transmission capa- bution network is a mixture of overhead attenuation can vary anywhere between bilities are distinguished, namely, Power and underground cabling with a majority 30 dB and 90 dB for a network of length Line Communication (PLC), which of underground cable in urban areas. 250 m. The variation in attenuation is operates in the range of kbit/s and Power caused by reflections from impedance Line Telecommunication (PLT), which mismatches at the end of each spur and is able to transmit data in the range of from any point on the cable where its Mbit/s [1]. electrical parameters change. The distri- bution cable size reduces as distance from the transformer station increases typically being 185 mm2 on leaving the station and reducing to 95 mm2 after some distance. There are several obstacles to be over- come in attempting to transmit high bit rate data signals over the electrical power network. Its bus architecture is hetero- HV/MV HV/HV geneous (in electromagnetic terms) and Power MV/LV the impedance is not well defined. In Station addition, the noisy nature of the network between the transformer and the cus- tomer premises, and inside the home, requires sophisticated but feasible tech- Figure 1 Generic European power network structure Users niques. Among the most promising are Coded Orthogonal Frequency Division Multiplexing (COFDM) and Spread Spectrum techniques such as Direct Sequence Spread Spectrum (DSSS), and Frequency Hopping Spread Spectrum (FHSS). 4 Electromagnetic Com- patibility standards Power line telecommunication technolo- gies use frequencies above 1 MHz (up to 30 MHz) to deliver broadband services over not dedicated cables. So it is impor- tant to consider EMC (Electromagnetic Compatibility) issues. At the moment no a) b) standards exist regarding this specific subject over power lines. The only exist- Figure 2 Typical modern electricity distribution cable structure; ing standard concerning the power lines (a) sector-shaped conductors, (b) circular conductors is the CENELEC 50065 that takes into 134 Telektronikk 2/3.1999

136 -20 account Signalling on low-voltage elec- -40 trical installations in the frequency range 3 kHz to 148.5 kHz. -60 VDSL PSD [dBm/Hz] Electromagnetic Compatibility concerns the ability of electronic equipment to function correctly in a disturbed environ- -80 ment (immunity), while limiting the dis- ADSL turbing effect (emission), of the equip- -100 ment on its environment. The evaluation of EMC performance should include equipment, cabling, outlets, connecting -120 ISDN hardware, electrical interfaces and cable HDSL termination at the equipment. -140 1 10 100 1000 10000 100000 5 Digital power Frequency [kHz] spectral density Figure 3 PSD for digital technologies In order to make a comparison with the Digital Power Line technologies, it is useful to collect the values of the differ- ent digital technologies that have been already standardised. Here, the Power Spectral Density (PSD) for copper digital technologies such as ISDN, HDSL, indoor and outdoor applications where switching and remote meter reading [2, ADSL and VDSL are presented. Figure 3 the power meter is the boundary since it 3, 4, 5, 6, 7]. Telecommunications ser- shows the PSD for copper digital tech- blocks the PLT signal. These two net- vices which could be provided are voice nologies. works are illustrated in Figure 4. telephony and data services such as Inter- net provision. The latter, in particular, is Indoor applications refer to the use of the seen as the killer application for PLT. 6 Services and indoor power network for communica- Table 1 compares access technologies in Applications tion with household appliances. Outdoor terms of limitations and functionalities. applications are those which use the low Services and applications fall into two voltage distribution network between the Current field trials suggest that a capacity general classes. Firstly, low speed customer premises and the electricity of about 1 Mbit/sec per final transformer (kbit/sec) telemetry applications and sec- company transformer station. This is cur- may soon be available. It is indicated that ondly, new high speed (Mbit/sec) appli- rently considered a low bit rate field of a penetration rate of about 10 % will be cations for telecommunications. Another application and includes such tasks as necessary to make the service commer- distinction to be drawn is that between power supply load management, tariff cially viable. Then, with 200 customers Base station Electric Toward power lines core Network ADSL Coax or fibre Meter ADSL box POTS Residential gateway etc or PC Coax Adaptors Transformer HV/LV Power line cables Figure 4 Indoor and outdoor powerline networks Telektronikk 2/3.1999 135

137 Table 1 Comparison of technology features Media Twisted pair Coaxial cable Power line Method POTS ISDN ADSL ADSLlite Cable Modem PLT Media type point-to-point shared Switching circuit packet packet to be type determined User interface 33 144 ~ 640 512 2000 to be up (kbit/s) determined User interface 56 144 8000 1500 25000 1000? down (kbit/s) (shared) (shared) Point phone S interface specific phone cable power meter, of access plug cabling plug plug specific cabling Required -- NT, digital modems modems, modems, modems upgrade from line card distributed access to POTS other filters cable than POP network access per transformer and an activity rate of 9, 10, 11]. The Internet trial connected There are currently no published stan- 0.1, a customer could have up to 500 12 computers using a shared 1 Mbit/sec dards for PLT. The CENELEC 50065 kbit/sec available. This would, clearly, power line link. A larger Internet trial is standard will not cover PLT since it represent a substantial improvement planned. Additional field trials are in specifies a maximum transmission fre- over current PSTN-based access bit progress, planned or under consideration quency of 140 kHz. High bit rate PLT rates. However, at the moment dedicated in some European countries, such as applications will require a spectrum of cabling from the electricity meter to the ENEL Spa in Italy (planned Internet 1 to 20 MHz or more. Electromagnetic modem is required. access trial) and Tesion Communications Compatibility at these high frequencies in Germany/Switzerland (meter reading, may be problematic. The Radio Society Internet access). Meter reading field of Great Britain believes EMC targets for 7 Power line products trials are in place in Spain and Italy. PLT will not be met and so it opposes PLT completely. A number of products are available which use the power line for communica- 9 Conclusions and The role of the domestic power meter tion. Most are intended for low bit rate telemetry applications such as automatic open questions in the power line communications circuit is unclear. Currently, it acts as a barrier meter reading and power supply manage- The message for the Network Operator to PLT signals and consequently there ment. Some products are intended for is one of wait and see. The number of exists separate products for outdoor and voice communication but also at low power line products available is now indoor use. The need to access the PLT bitrates. For telecommunication applica- quite limited and in the case of outdoor signal from the meter is inconvenient for tions, products are available for indoor Power Line Telecommunication (PLT) the consumer and the PLT service pro- use which are designed to communicate services there is only a single product. vider. However, there is no obvious between household appliances and be- Power line technology is not now a com- reason why this situation should continue tween computer equipment in the home. petitive threat to the incumbent operator. and the development of a power meter The only known outdoor high speed which does not block PLT signals can product is the Nortel/Norweb power line The low voltage electricity distribution be foreseen. modem. network is not now an alternative to the Network Operators access network for There is no inherent physical barrier to 8 Field trials telecommunications services. This is due the further development, improvement to the immature nature of PLT technology. and commercialisation of PLT technol- Independent verification of field trial ogy. It can be expected that PLT will be information is very difficult to obtain and The PLT infrastructure is owned by an driven by business needs and that if it is the information presented here is taken actual or potential competitor in most seen as viable, then higher performance largely from the press and press releases European countries. Consequently, it is PLT systems will surely appear. For this of the companies involved. Nortel/Nor- difficult for a traditional Network Opera- reason and due to its near-ubiquitous cus- web have conducted field trials in the UK tor to trial or properly measure the per- tomer coverage, PLT technology has the using their Digital PowerLine technology formance of PLT products. potential to being a serious competitor to to provide voice and Internet services [8, traditional Network Operators. 136 Telektronikk 2/3.1999

138 References Dermot Collins (32) obtained his Electronic Engineering degree in 1993 from Dublin Institute of Technology and 1 Propp, M. The use of existing electri- Trinity College Dublin and his Masters degree from Dublin cal powerlines for high speed com- City University in 1997. He has worked in the Network munications to the home. In: Hurley, Planning and Switching departments of Telecom Eireann D, Keller, J H (eds.). The first 100 and is participant in the Eurescom project Implementation feet, options for Internet and Broad- Strategies for Advanced Access Networks. band Access. Cambridge, Mass., MIT email: [email protected] Press, 1999. (Conference at the Free- dom Forum Arlington, Virginia, 2930 October 1996.) 2 Hooijen, O G, Han Vinck, A J. On the channel capacity of a European- style residential power circuit. In: Lucien Budry works for the Swisscom Research and Proceedings of 1998 international Development department as a Senior Engineer with Net- symposium on power line communi- work Evolution Strategy has a main activity. He is currently cations and its applications, Tokyo, Project Leader for Long Term Access Network Scenarios. Since 1994 he has been working on core and then access Japan, Soka University, 1998, 3144. network development, and has been involved in several ACTS and EURESCOM projects. 3 Brown, P A. Some key factors influ- encing data transmission rates in the email: [email protected] power line environment when utilis- ing carrier frequencies above 1 MHz. In: Proc. 1998 International Sympo- sium on Power-line Communications and its Applications, 1998. Gbor Gerdai (51) received his degree from the Technical 4 Brown, P A. Telecommunications University of Budapest in 1972. His first job was at the services and access provision. IEE Hungarian Optical Works as development engineer. Colloquium, Jan 1996. Between 1975 and 1977 he did postgraduate studies in digital electronics at the Technical University of Budapest. 5 Power deal puts Britain ahead. The Since 1975 he has been working at the Telecommunica- tions Development Institute of the Hungarian Telecom Times, 01.04.1998. Company (MATV), where he now works on powering issues in telecommunications networks. 6 Allbritton, C. Electric utility announces Internet access through email: [email protected] power lines. Philadelphia Newspa- pers Inc., 09.10.1997. 7 Automated meter reading article for Angelantonio Gnazzo (36) received his degree in physics World Water. Schlumberger Mea- from the Universit degli studi di Torino. In 1988 he joined surement & Systems, April 1997. CSELT and became involved in the design and realisation by MCVD of special optical fibres, such as dispersion shift- 8 Gautam, N. English inventor puts a ed, dispersion flattened, polarisation maintained and rare spark into the net. News EDGE/Web, earth doped fibres. From 1994 to 1996 he moved his activ- 30.06.1998. [online]. URL: ity into the fields of integrated optics and Bragg grating devices. Since 1996 he has been working in the plant topology and maintenance department of CSELT. 980630english_inventor.htm email: [email protected] 9 Norweb shows utilities whos got the power. Network News, 01.04.1998. 10 Ayres, C. NOR.WEB aims to plug Ahmed Madani (59) joined France Telecom CNET Lannion communication gap Nortel Labs. in 1975 after 10 years as researcher at Rennes Uni- United Utilities. The Times, versity. He has been involved in ionospheric propagation 26.03.1998. studies, expertise and specification for the optical French programme, as well as the CATV optical trunk network. He 11 Scales, I. Volts and amps joined by has been involved in the introduction of optical amplification IP packets. CommunicationWeek for video transmission and has been CNET project leader International, (193), 31, 1997. in two RACE programs. Became head of technical DVS dept. in CNET Lannion; since 1998 in core transmission and access networks. email: [email protected] Telektronikk 2/3.1999 137

139 Component technologies for wireline broadband access networks DANIEL LECROSNIER Infrastructure costs represent one of 1 Introduction if a high volume (above 1 million pieces) the most important factors in the glo- market is reached. However, this app- bal cost of an access network construc- With the explosion of Internet-based roach of upgrading copper networks tion. Advances in this field can strong- services, there is an increasing demand presents limitations with respect to band- ly influence the practical feasibility of for access networks supporting high width, distance and quality performances. new network deployment. The purpose bandwidth. This demand is now clearly of this paper is to address the basic identified for business customers and will On the other hand, the laying of fibre elements including key components, appear in the near future for profes- rich infrastructure appears as the most installation techniques and powering sionals, and later for residentials. For promising solution to overcome the needed for the construction of wireline network operators a broad diversity of above mentioned copper limitations and FTTx infrastructures supporting implementation is becoming available for serving new-built areas. Optical fibre broadband services. Up to now, most to meet the great variety of customer is now widely used for long distance optical technologies used in access net- demands. trunks, and in the access network it is works have been developed for the assumed that fibre could also be the basic long haul application for which cost On the one hand, the full exploitation of medium to construct a core infrastructure is not a prominent factor. However, in existing infrastructures, ie. re-use of supporting full wideband services. At the last few years, the specific access copper pairs, coaxial cables and, under present, the real implementation of op- network requirements have been certain circumstances, even powerlines tical infrastructures require heavy invest- stressed and made clearer both in the only need new electronic-based equip- ments, not only due to the use of opto- framework of the FSAN initiative and ment to be installed at the local exchange electronic equipment but mainly due in EURESCOM P614 project Imple- and at the customer premises. Consider- to fibre deployment. As a consequence, mentation Strategies for advanced ing ADSL as an example, a technology cost evaluation of fibre-based infra- access networks. As a result, the analysis of currently available modems structures is a very relevant issue for net- maturing of optical technologies is in reveals that they rely mostly on conven- work planners. rapid progress to meet access network tional CMOS circuits, therefore no major requirements. A few examples will be technical barrier has to be overcome for In the following sections the various ele- presented indicating how technical their fabrication. Modem cost evaluation ments needed to construct an FTTx and economical barriers should be can be easily predicted using the well infrastructure are reviewed and discussed overcome in the near future. known silicon industry economic trends, with the main emphasis on recent ad- and the low cost target will be achieved vances and cost reductions in optical technologies. 2 FTTx infrastructure for access networks Figure 1 shows a variety of ways to build broadband access networks based on fibre deployment. Today, the basic archi- tectures under study in most countries FT TB (HFR) are: LE X ONU FTTF/O: Fibre to the floor/office; O NU O NU FTTB: Fibre to the building; FTTC/Cab OM MDF FTTH FTTCab: Fibre to the cabinet; O (HFC) O NU FTTH: Fibre to the home; ONU L HFC: Hybrid fibre coax; FTTF /O ONU T HFR: Hybrid fibre radio. O NU The construction of a fibre rich access network relies on the implementation of Duct several major building blocks. In the Duct / direct buried local exchange, the OLT (Optical Line Manhole Termination) ensures the interface be- tween the switching equipment and the Cables : Connectors : ODN (Optical Distribution Network). The OM (Optical Monitoring) module Splices in enclosure : Splitter : of which the functionality is to survey the ODN quality and an MDF (Main Dis- MDF : main distribution frame OM : optical monitoring tribution Frame) which provides a con- nection point between equipment and Figure 1 FTTx infrastructure outside cables. For outside plant con- 138 Telektronikk 2/3.1999

140 Table 1 Optical access network requirements struction, it is necessary to consider the Optical transmission 2 fibres 1 fibre 1 fibre hardware parts (cables, splices, splitters, WDM WDM connectors and enclosures) together with 1.3 m 1.3 1.5 1.3+ 1.3 civil work and installation techniques. The optical network terminates at the Bit rate (Mbit/s) between Downstream Upstream ONU (Optical Network Unit) whose OLT and ONU location depends on the chosen archi- Symmetric 155 155.52 155.52 tecture. Finally, powering equipment is needed to supply all active equipment. Asymmetric 622 622.08 155.52 Optical path loss 10 15 dB (G.982 Class B) 3 Optoelectronic 15 30 dB (G.982 Class C) conversion modules Differential path loss 15 dB For both OLT and ONU equipment, the Overall reflectance at S/R points 32 dB optoelectronic conversion is a critical functionality with regard to cost. As large Temperature FTTH ONU 0 to 60C (case temperature) volume production is a prominent factor 40 to +85C (storage) for cost reduction, common physical media layer requirements have recently been agreed on within the FSAN initia- tive [1] and ITU-T [2]. Table 1 indicates basic parameters for ATM/PON systems. called micro-optics. It relies on assem- a consequence, automation of the assem- From the FSAN study it has been demon- bling several individual parts as sketched bling process is limited, which severely strated that bi-directional transmission in Figure 2: a laser diode, a photodiode, a impacts fabrication costs. over one fibre using 1.5 m (down- wavelength dichroic filter, etc., which all stream) and 1.3 m (upstream) windows need to be mechanically assembled with To reduce the cost, a new generation of is more cost-effective than two fibre lenses and fibre with less than a micro- modules is under development in many transmission [3]. meter tolerance with respect to optical companies. It is based on hybrid inte- axes. This requires the use of high pre- gration of optoelectronic chips with pas- The presently available technology for cision materials and components and sive optical components (splitter, filter) production of O/E modules is generically time consuming operator assembling. As implemented with the Planar Lightwave c) Integrated optical module /InP 1.3 m 1.5 m fibre Si platform WDM mux/dmux b)Hybrid optical module Laser with spot size converter 1 , 2 1 fibre 2 Waveguide Photodiode PLC Si platform (Planar Lightwave Circuit) a) Micro-optics selfoc WDM LD lens filter fibre Figure 2 Technologies for opto- electronic bi-directional trans- mission module lensed PD 1990 2000 Years Telektronikk 2/3.1999 139

141 Finally, monolithic integration of the links. In comparison to the long distance optical and optoelectronic functions in a network, the access network has several single semiconductor chip is at a research specific requirements: stage and could lead to further cost sav- In urban areas, the topology presents ings. Yields and material compatibility many bends and the ducts are usually for different functions are the main issues occupied by other cables, therefore at present. A production status is unlikely installation of a new infrastructure to be reached before year 2000. is often contracted in small volumes leading to a small bending radius for Prices of optical modules produced with fibre and cable; different technologies for different pro- duction volumes are presented in Figure Cables are installed in various condi- 4, taking as a reference a bi-directional tions: outside, underground (in duct transmission at 155 Mbit/s for G.982 or buried), aerial, indoor (in operator class B standard. Cost evolution over the building and in customer premises). last two years clearly shows a dramatic Consequently many thermal, mechani- decrease for the so-called micro-optic cal and other stresses have to be taken technology while PLC hybrid integration into account such as humidity, tempe- is still at prototype level. For mass pro- rature variations, UV exposure, wind, duction level (one million pieces a year) frost, snow and shots; a target value as low as USD 40 seems Figure 3 Detachable pigtail packaging The links usually comprise many con- realistic both with micro-optic and hybrid for automated board mounting nections, so it is desirable that installa- technologies. tion does not require specially trained technicians; easy preparation of fibre 4 Optical cabling for splicing and for connector mount- ing is recognised as a critical issue; Circuit (PLC) technology. PLC technol- ogy uses dielectric waveguides (Silica on 4.1 Access network The cost of the global link has to be low. Silicon) deposited on the same silicon requirements motherboard that performs chip and fibre During the last ten years, optical fibre self alignment. In addition, the planarity has demonstrated great qualities for the 4.2 Low diameter, low weight, of this assembly technique permits a long-haul applications; among them high density optical cables packaging suited for automated surface capability to transmit very high bit rates, Figure 5 presents geometrical and weight mounting on printed boards. Examples insensitivity to electromagnetic pertur- characteristics for typical copper and of a detachable pigtail device are shown bations leading to very low bit error optical cables of a quite similar capacity. in Figure 3. rates, and a reliability better than coaxial It clearly shows that optical cable tech- nology leads to a tremendous reduction in weight as well as in diameter para- meters when compared to the copper cables and therefore offers large poten- 500 tials for lower installation costs. Small diameter and high density cables really Data rate Micro-optics trend open the door to a re-use of existing civil 155 Mbit/s PLC Hybrid Integration work in case of overlay fibre deploy- 400 ment. Moreover, ultra high density cables can be realised by using the newly devel- oped multicore fibre [4]. 300 From recent cost analysis it appears that 1996 todays price of bare G.652 standard FSAN data fibre is quite uniform world-wide: it 200 ranges from USD 40 to USD 60 per kilo- metre. It is about twice the copper pair price. Following the fibre prices, which have been continuously dropping during 100 the last years, quite similar values for 1998 micro-optic fibre and copper pair prices are expected in the short term. 10 2 10 3 10 4 10 5 10 6 Production volume/year Figure 4 Optical module cost dependency on volume 140 Telektronikk 2/3.1999

142 4.3 New installation techniques 62 mm In addition to the usual pulling and push- 5500 kg/km ing methods, the blowing technique is now recognised as a technology break- 21mm through for low cost installation of opti- cal cables. Taking advantage of the tiny 363 kg/km nature of fibre, British Telecom re- searchers first proposed this method. The fibre unit is placed in a machine 448 copper pairs 648 optical fibres where it is pushed forward by a little caterpillar or by a series of wheels. On the fibre unit two kinds of pushing forces Figure 5 Comparison of currently available copper and optical cables are therefore present: the concentrated force due to the caterpillar and the distri- buted force along the duct due to the compressed air. The distributed force limits the effects of the curves on the asphalt of road or pavement (see Figure fore, in addition to the general require- maximum installation length, so that air 7). As a result the cable is required to ments of optical links, some specific blowing techniques could be particularly meet exact demands as to crush resis- features become critical. For example, useful if the installation route is quite tance and, in particular, temperature the connectors set up inside the customer tortuous. resistance which is needed when sealing home must be secure, robust and friendly the cable in the groove with hot bitumen. to be handled by unskilled people. Significant improvements have been This new technique is significantly more added to the initial blowing concept cost effective and more rapidly deploy- As many optical devices which have ini- allowing the installation of a great vari- able than a standard solution. tially been designed and produced for the ety of cables over long distances. To the long haul transmissions, connectors have head of the cable a small piston is con- to be hardened and adapted to fulfil nected (see Figure 6) in order to produce 4.4 Connecting technologies access requirements. The main specific an additional pulling force. If the piston requirements are: is an airtight one, the compressed air 4.4.1 Connectors pushes it and consequently the cable Reliable coupling mechanism both Connectors used in access networks is pulled by the piston inside the duct. between the plug and the pigtail and could be installed in the outside plant as In this case, the flow of compressed air well as at the customer premises, there- does not produce a distributed pushing force along the cable length because it is stopped by the piston, but it contributes to reduce the friction between cable and duct. If the piston is not an airtight one both contributions are present: a small pulling force on the head of the cable and a force due to the pressure drop along the con- duit, distributed along the cable. Recent Air tight piston improvements in the piston design called Joint box sonic head [5] allows the extension of the trajectory of the cable. With both techniques, according to cable weight and installation route, it is pos- Compressed air sible to install cable lengths with a maxi- mum of 15002000 m in one step. This Cable technique is applicable for both outdoor and indoor installations. Pushing and blowing head For buried cables, a Micro Cabling Sys- tem concept has recently been proposed by Siemens [6] in order to reduce install- ation costs and installation time. This Microduct is achieved by means of a simple but Cable rugged design cable which is laid in a very low depth groove dug into the Energy Figure 6 Fibre blowing principle scheme Telektronikk 2/3.1999 141

143 4.4.3 Enclosures The purpose of an enclosure in commu- nications systems, whether employing copper or fibre-optic cables, is to protect the splices and all other components of a cable at a jointing or branching point. General requirements for optical cable splices have been proposed for en- closures: PE sheat copper tube optical fibres Mechanical strength: The enclosure must have sufficient mechanical strength to withstand external forces acting on the laid cables, such as vibra- tion, tensile and compressive forces. Its lifetime should be at least as long Figure 7 Low depth buried cable as that of the cable. Pneumatic seal: Generally the en- closures are installed by placing them in manholes. So the device must be tightly sealed to prevent the effect of water or moisture which would dam- between the plug and the adapter in basic performances (insertion loss and age fibres. order to face accidental pulls; reflection). A cost target in the range of Enclosures function: In order to easily USD 4 to USD 10 seems within reach at Reliable optical performances; maintain and upgrade the network the high volume production. enclosures should be readily accessible High mechanical strength against and simple to use (no special training, accidental crashes or drops; 4.4.2 Splices short intervention time, no special Good endurance to the temperature tools). The installation of a future proof and changes (above all for connectors set stable infrastructure could represent a Chemical resistance: The enclosure up in outside plants); key factor in a competitive market. The material should have a strong capacity Easy and quick field installation; splices and the enclosures as joint ele- to resist the attack from chemicals, ments have a strong impact in this en- such as acid, base, solvent, etc. Protection against eye-damaging laser vironment as they must guarantee high radiation; quality, low maintenance and low instal- With todays technologies, the above re- Low cost. lation cost. quirements are usually fulfilled by manu- facturers. However, large cost variations One major critical point has been stressed Concerning splices, two technologies are found mainly due to small volume by EURESCOM P614 members: con- are still competing at present: fusion and production. Fully standardised products nector field assembly capabilities appear mechanical. At present, fusion splice will be beneficial for the operators. not yet ready for true mass application. using ribbon fibre arrangement appears Fibre preparation, fibre positioning into as the preferred jointing technique. This the plug, fibre ending still require well- technique provides very low losses 5 The splitter trained technicians and expensive tools. (0.1 dB) and low reflected light (50 dB) The splitter is a basic device for PON In fact, at present, most operators buy in the whole temperature range for out- systems, since it permits not only a pigtailed connectors and splice the tails side operation. Cost investigations of sharing of the fibre in the network but in. Significant improvements in con- fibre splicing have revealed a high sen- also a sharing of optoelectronic con- nector and fibre technology are expected sitivity to specific factors such as field verters in the central office. Three basic to solve this problem in the very near labour cost and splicer amortisation technologies exist today: future. which are liable to strong variations from country to country. Fused fibres (a): This technology In addition, to reduce cost some innova- appears well suited for low splitting tions are also desirable: Mechanical splices are usually related to ratios (up to four) but suffers from a the temporary situation, as in the case of rather large package when splitting Ceramic ferrule replaced by glass or emergency restoration. From a technical ratios increase. This is mainly due to plastic; point of view the mechanical splices the splicing technique used for con- Ferrule-less connector; have some weaknesses as compared to catenation of 1:2 couplers. Example: the fusion splicing: lower tensile 1:16 splitter, dimensions 150 x 100 x Multi-fibre connectors allowing minia- strength, higher return loss and the use 20 mm. turisation. of index matching gels which are tempe- Planar glass (b): Waveguides are rature sensitive and less suitable for mul- With present technologies, prices range obtained by diffusion techniques tiple fibre arrays. from USD 8 to USD 25 depending on into glass substrate. 142 Telektronikk 2/3.1999

144 Planar Lightwave Circuits (PLC) (c): cable break). To implement the monitor- allowing a high quality of service offer. Waveguides are made of silicon oxide ing of an optical infrastructure three However, the implementation of on-line layers deposited either by flame hydro- methods can be envisaged: use of a dedi- monitoring based on OTDR (Optical lysis (which was initially developed cated branched fibre monitoring, dark Time Domain Reflectometry) as for fibre fabrication) or by plasma fibre and on-line monitoring [9]. On-line sketched in Figure 9 needs a heavy techniques. Substrates commonly techniques refer to the monitoring of investment which should be justified used are silicon or glass. every transmission fibre while the ser- when high quality transmission gua- vices are running and consequently rantees are required. Both b) and c) technologies apply some of the well matured silicon microelec- tronic processes (eg. lithography, etch- ing) to delineate the various branches of a splitter; devices with ratios up to 16 (and even 32) are now fabricated in a very compact package. Example: 1:16 Point to Point splitter, dimensions 90 x 10 x 6 mm. The cost is about USD 50 per port for 5,000 ONU pieces, which is quite a high figure. By OLT contrast to connector or splicing devices ONU which serve only to joint fibres, splitters have a higher functionality. They have a strong impact on network implementa- tion; their installation requires a careful Point to Multipoint analysis taking into account many para- meters such as installation planning, ONU impact on maintenance, network up- OLT gradability. One basic point is its loca- ONU tion: in the field or in the central office. These two options have been studied in Splitter in the exchange EURESCOM P614 together with point- to-point links as sketched in Figure 8. Point to Multipoint This comparison highlights the advan- tage of point-to-point solution when con- ONU OLT sidering network functionality (O&M, service provision, quality and upgradabil- ONU ity) as a strategic parameter [7]. Splitter in the outside plant Figure 8 PON and point-to-point network topology 6 Hardware maintenance optical monitoring ITU-T gives the following definition of maintenance: Maintenance involves the whole of operations required for setting up and maintaining, within prescribed Local exchange Outside plant limits, any element entering into the s F setting up of a connection [8]. Starting OLT from this generic definition, EURES- COM P614 proposes the concept of hard- ware maintenance: Hardware mainte- Rm F nance covers all the means to guarantee OS ONU the performances of the physical carrier m s (fibres, twisted copper pairs, coaxial, OTDR splices, connectors, passive components, m enclosures), and specifically to detect To other PONs and locate any fault in the access net- Control work. Some guidelines have been estab- lished and a procedure is given based on three parts: acceptance of components, OS : Optical selector Rm Reference reflection network qualification and network moni- WDM toring. As regards the optical infrastruc- ture, monitoring should identify prob- Filter to stop m m ( 1600-1650 nm window ) lems such as slow and uniform degrada- tion as well as abrupt fault (for example Figure 9 Optical maintenance scheme based on reference reflection and optical selector Telektronikk 2/3.1999 143

145 7 Powering mean that power concentration is im- References possible unless Controlled Environ- Contrary to the copper pair, the fibre mental Vaults (CEVs) are developed, 1 Okada, K et al. Overview of full ser- does not transport energy, therefore the but these are difficult to install in vices optical access networks. Full powering of ONUs is a hot topic in the heavily populated urban areas. Services Access Network Conference, network planning due to the use of active London, 20 June 1996. (http://www. equipment in the field. To solve the prob- A cost analysis of powering reveals ex- lem some solutions exist based on two tremely large variations according to the index.htm) major architectures, namely local or chosen architecture and to the customer remote powering. The decision to opt for requirement with regard to service quali- 2 ITU. High speed optical access sys- a particular powering architecture relates ty. In case of an FTTH architecture with tems based on Passive Optical Net- not only to its cost. Field constraints of a local powering, a value of about USD 60 work techniques. Geneva, 1998. non-technical nature mean that very diffi- for ONU powering equipment with three (ITU-T Rec. G982.) cult decisions have to be taken regarding hours battery back-up has been estimated. the place where the powering equipment 3 ITU. ATM Passive Optical Network is installed. Experience with installing Specification. Geneva, 1998. (ITU-T classical CATV networks has shown all 8 Conclusions Rec. G. 983.1.) the difficulties of performing mainte- Major building blocks needed for the nance on equipment in the access net- 4 Zylbersztejn, A et al. Low cost com- construction of fibre rich infrastructure work, especially for installations in pri- ponents for full service access net- have been addressed in the scope of this vate buildings. Although no specific works. In: Proc. Eighth International paper, and the main conclusions drawn solution can be rejected a priori, each Workshop on Optical/Hybrid Access are: option has its own particular difficulties, Networks, Georgia, Atlanta, 25 Mar mainly in heavily populated urban areas: Recent advances in optoelectronic 1997. module technology show great poten- Installing equipment in a private build- tials for dramatic cost reductions in 5 Le Noane, G et al. Bunched multi- ing: This is the preferred solution for a short term time scale. core fibres : a key for the future equipment environmental conditions. FTTH networks. In: Proc. Twelfth However, many old buildings do not For outside plant construction, basic Annual Conference on European have plant rooms; moreover, installing hardware parts (cables, splices, con- Fibre Optic Communications and equipment in one building to supply nectors and enclosures) are in progress Networks, Heidelberg, 2124 Jun other buildings raises problems of in order to meet the specific access 1994, 1, 3135. accessing the building playing host network requirements. to the equipment if maintenance needs Very significant innovations (blowing 6 Griffioen, W, Plumettaz, G. Current to be carried out. technique, micro-civil work concept) development in cable-in-duct blow- Installing equipment in a street box: are now providing cost effective instal- ing technique. ANCIT workshop, The equipment environment conditions lation techniques, which in turn makes Torino, 3031 Mar 1998, 1. are much less favourable (reduced the construction of fibre rich access capacity for the batteries at low tem- networks more and more attractive as 7 Finzel, L. Micro cabling system : a peratures and reduction of battery life a means of providing full wideband new trench-less technique. ANCIT at high temperatures). Furthermore, services to a large spectrum of cus- workshop, Torino, 3031 March operators of various networks find it tomers. 1998, 1. difficult to obtain permission to locate more street equipment in major con- 8 Pousa, M et al. Why not use of point urbations for aesthetic reasons. Acknowledgement to point solutions in FITL? In: Proc. NOC, Manchester, 2225 Jun 1998, Installing equipment in a telecom man- This paper is based on results achieved 95101. hole: Active equipment must be placed in task 4 of the EURESCOM P614 pro- in costly leakproof cases requiring ject. The author wishes to express a 9 ITU. Maintenance philosophy for civil works to widen existing man- special thanks to all task 4 participants. telecommunication networks. holes. Problems of heat dissipation Geneva, 1997. (ITU-T Rec. M.20.) 10 Gnazzo, A et al. Access network maintenance : an overview from Daniel Lecrosnier (59) is a staff member of strategic plan- ning studies in the Transmission and Distribution Division EURESCOM project P614. In: Proc. of CNET in Lannion. He has been involved in opto-elec- NOC 98, Manchester, 2225 Jun tronic component research and more recently in optical 1998, 3541. outside plant installation. He participates in EURESCOM P614 project as task leader on Enabling technologies for broadband access networks. email: [email protected] 144 Telektronikk 2/3.1999

146 Management of access networks trends and challenges TOR BREIVIK AND HKON LNSETHAGEN This article provides an introduction design and choice of management inter- which represent major shifts upon man- to telecommunications management faces to network elements. Furthermore, agement architectures. While section 4 technologies, also identifying emerging the management frameworks of IETF2) provides an overview of a few important technologies and solutions challenging and OSI respectively both have strengths generic standards, section 5 identifies the traditional ones. From a broad that must be considered carefully with management information and function- management perspective, important respect to the network technology and ality related to the ADSL and PON net- properties and challenges related to architecture they are best suited to. work technologies. Software technology the introduction of modern access net- trends, opportunities and issues are work technologies are identified. Operations, administrations and mainte- further discussed in section 6, whereas nance of network resources is only one section 7 presents issues related to inte- area of telecommunications management. gration of management systems and 1 Introduction The scope of telecommunications man- interworking among management tech- agement (or TMN) also covers service nologies. Access network operators are facing management. That is, in our case, man- many questions, challenges and opportu- agement of access network transport nities in their effort to remove the band- (bearer) services. Efficient and reliable 2 Telecommunications width bottleneck of the access network. service provisioning is a key goal in this management The future success of an access network area. Clever management systems must operator will largely depend on how suc- make their service provisioning activities The concepts and ideas behind telecom- cessful his telecommunications manage- based on adequate network level resource munications management have been ment solutions become. Whereas man- models. These information models must around for quite some years, first being agement1) of network resources largely be able to support multi-service and introduced in the mid-1980s. Standards is a non-issue in relation to old access multi-technology networks. Moreover, providing the setting for basic manage- network facilities, the forthcoming access user-friendly customer access enabling ment technologies have been stable for network resources fully depend on man- on-line management of services may several years. Still, however, the commu- agement interfaces to accomplish the very well become the facility distin- nity of telecommunications management management functions integral to the guishing the successful operator from does have several interesting issues to corresponding network technologies. the others. address, as will be discussed below. This section gives a brief introduction to the Several network technologies will enter Along with the emerging management basics of telecommunications manage- the access network in the future. A systems for new access network tech- ment technologies, and provides an over- paramount goal will be to enable man- nologies, integration and architectural view of OSI4) Systems Management agement of the network resources in an challenges also appear regarding integra- (OSIsm), TMN from ITU-T and SNMP5) integrated fashion. If operations systems tion with and migration of legacy opera- from IETF. Important differences be- operators must learn and handle several tions and administrative systems. Strate- tween the IETF approach and the OSI different management systems and user gies to handle these challenges must also approach will be addressed. interfaces, the added costs induced by cope with evolving and changing busi- inefficient operational procedures will ness processes and the uncertainties significantly reduce the competitiveness related to the choices, timing, and the 2.1 The manager-agent of the network operator. Likewise, evolution of new network technologies. paradigm maintenance of numerous but similar Some way or another, the various solu- management applications and systems Emerging and maturing software tech- tions available for telecommunications must be avoided. nologies for distributed computing, like management are based on the manager- object request broker platforms (eg. agent paradigm. A system or an applica- The management infrastructure should be CORBA)3) and component based tech- tion in the manager role communicates suited to the needs and characteristics of nology and corresponding middleware, with an application in the agent role the access network architecture. Know- provide a means of handling the chal- being managed by the manager applica- ledge and insight related to the access lenges identified above. tion. Thus, the system containing the network technologies as perceived from a agent (the application in the agent role) management perspective is important to This article will address the issues iden- is taking the managed role. The agent make the right judgements related to the tified above and provide insight to the provides a representation of the system underlying challenges. The following or resource being managed, and performs section will introduce TMN and related the management tasks as instructed by management infrastructure technologies, the manager. These ideas are illustrated 1) In this article, the term management is also covering some emerging technolo- in Figure 1. A fundamental goal is to used analogous to telecommunications gies. Section 3 will identify some of the management, and with the same scope main characteristics related to the future as adopted by TMN (see below). Un- role and architecture of access networks, 4) Open Systems Interconnection (jointly less otherwise noted the term manage- from ISO/IEC and ITU-T). In this ment of network resources assumes article, only the ITU-T recommenda- interworking between managers and 2) Internet Engineering Task Force. tions numbering will be used. network elements (their agents) across management interfaces involving a 3) Common Object Request Broker 5) Simple Network Management Proto- data communication facility. (ORB) Architecture. col. Telektronikk 2/3.1999 145

147 Managed system/resource which may contain a set of the just men- Manager system tioned properties. The aim is to increase Application the potential reuse of specification Application Management operations: Agent role Manager role get, set, etc. pieces. The properties associated with a class are specified via the identified set of packages constituting the class defini- Operation responses, tion. The GDMO language consists of a Notifications set of templates such as the Managed Object Class template, the Package Management communication template, and the Attribute template. The Managed resource semantics associated with the specifica- (representation of) tion templates is explained in a separate Figure 1 The Manager-Agent Paradigm recommendation describing the Manage- ment Information Model [4] associated with GDMO. The syntax associated with GDMO tem- plates has a mapping to ASN.16) syntax [5]. ASN.1 enables abstract specification provide a means for specification of 2.2 OSI systems management (independent of a specific transfer syn- interfaces between manager systems tax) of the structure and type of data ele- The OSIsm solution was developed for and managed systems in a complete and ments to be communicated across a net- the management of complex OSI data well-defined way to enable vendor in- work. GDMO may utilise the full power networks (OSI systems) and telecommu- dependent interface specifications. The of the ASN.1 language allowing for nications facilities. Reliable management specification should only constrain what instance an attribute to be of a complex communications has been emphasised, is necessary to ensure interoperability, structure. Furthermore, each specification thus the choice of a connection-oriented leaving a considerable amount of free- element (of a standard) is registered with solution. To handle the complexities of dom regarding the implementation. a unique value whose type is the Object the resources to be managed as well as Furthermore, hierarchical structures of Identifier (OID) ASN.1 type. This is used to keep the management traffic low a management systems may also be pos- by the management protocol ensuring powerful management protocol has been sible. This implies that a management that the management information is developed as well as an expressive speci- system that takes the agent role with properly typed. fication language. However, this results respect to one interface may take the in complex functionality in the agent manager role with respect to another Managed object instances controlled by increasing the cost and the processing interface. and clustered under one agent constitute load of the managed system. the Management Information Base A few essential aspects can be associated (MIB). Each instance has a name and to The specification language of OSIsm with this paradigm. First, the capabilities ensure an efficient and unique naming of is GDMO (Guidelines for the Definition of the protocol enabling management managed object instances, they are of Managed Objects [3]). GDMO is a communication between the manager organised in a tree structure referenced powerful object-oriented specification (the application in the manager role) and as the Management Information Tree language supporting classes, inheritance, the agent will to a great extent settle the (MIT). Naming is based on the contain- and allomorphism (see below). A class basic properties of a management solu- ment relation specified relative to Man- specification identifies properties such as tion. Secondly, the capabilities of the aged Object Classes using the name- the attributes (data elements), actions management information specification binding template. In this fashion, every (operations relating to the object as a language further determines additional (subordinate) instance is named relative whole) and notifications associated with important properties on which the man- to and contained within a superior objects (instances) of that class. Notifi- agement solution will be based. A third instance, where the root of a MIT is cations are an effective capability of aspect relates to architectural issues and associated with a directory object7) OSIsm-based solutions allowing the choices, both with respect to structures of representing the agent. agent to spontaneously report events re- system internal functional entities as well lated to the managed object. Operations as possibly hierarchical structures of re- The Management Information Model on the attributes (get, set) as well as ac- lated management systems. In this sec- identifies rules for compatibility, that is, tions may be associated with result or tion, the first two aspects will get more how one class can be a compatible exten- error information as a response to the attention. sion of another. A subclass derived from operation. Inheritance allows a class to a superclass is compatible with the super- inherit the properties of possible multiple Telecommunications Management Net- class. However, a class may not be a sub- parent classes, provided by the derived work (TMN) from ITU-T, have adopted class to be a compatible class. The notion from construct. Unlike object-oriented the core standards from OSIsm [1, 2]. of allomorphism is related to and based programming languages, GDMO allows Thus, before giving an overview of attributes, actions and notifications to be TMN, we will first take a closer look at identified as self-contained specification OSIsm, addressing the capabilities of the 6) Abstract Syntax Notation One. pieces outside the scope of the class de- management protocol and the manage- finition. The same is true for packages, 7) OSI Directory Service X.500. ment information specification language. 146 Telektronikk 2/3.1999

148 on that of compatibility, and is provided for by OSIsm. If an instance of a class Operations Systems Function (OSF) block: can be managed as an instance of another related to the telecommunications management manager role, for the purpose of class, such other classes are termed its monitoring, co-ordinating, configuring and controlling telecommunications resources allomorphic classes. By using allomor- as well as the TMN itself. phism, an agent may support both an old and a new version of a class, thus allo- Workstation Function (WSF) block: morphism may be considered as a means provides means to interpret TMN information for the human user (operator), and vice to support systems evolution. versa. A more in-depth tutorial of GDMO and Network Element Function (NEF) block: information modelling related to OSIsm a function block representing telecommunications equipment (or parts of equipment), may be found in [6]. communicating with the TMN for the purpose of being monitored and controlled. Mediation Function (MF) block: 2.2.1 CMIS/CMIP mediates and/or processes information passed between an OSF and NEF (or QAF). The application layer protocol supporting An MF block may store, adapt, filter, threshold and condense information. communication between manager and agent, and closely related to GDMO, is Q Adapter Function (QAF) block: specified in a pair of standards. That is, translates between a TMN and a non-TMN reference point. For example between the Common Management Information non-TMN network elements and TMN (MF or OSF). Service Definition (CMIS) [7] defining the service elements and the capabilities Box 1 TMN function blocks of the protocol; and the protocol itself, the Common Management Information Protocol (CMIP) [8]. This protocol allows reliable asynchronous message passing and an elaborate set of capabili- ties. It provides means for basic opera- tions associated with single objects as specified by GDMO, such as Get and 2.2.2 Generic management 2.3 TMN Telecommunications Set attribute, Action, Create and Delete information and functions Management Network object instance, and Event-report, the OSIsm defines a set of generic managed The principles for a TMN [9] are related later supporting notifications. In addition, object classes that can be used and re- to four main topics or classes of architec- these protocol services (except Event- fined by other standards. Of particular tures, that is functional, information, report and Create) can be used with an importance is the top managed object physical, and logical layered architecture. expansion or query mechanism known as class. Every other class must inherit this The TMN functional architecture identi- scoping and filtering. These mechanisms class, which identifies properties that fies and defines function blocks and their allow an operation to be performed on a every managed object must support, for content in terms of functional compo- selected set of objects. Thus, the func- example an attribute telling which class nents. Several solutions exist regarding tionality of the agent is expanded. it belongs to. Furthermore, classes such which functional components the various Scoping provides various ways of identi- as Log, Log record, Alarm record, Event functional blocks contain. Reference fying a subset of the MIT below a record and Object creation record are points between relevant pairs of function given base object. The selected set of defined, representing generic capabilities blocks are also identified. The TMN objects may further be narrowed or fil- useful in many contexts. function blocks are shown in Box 1. tered using logical constraints. When using this facility of performing an oper- Systems Management Functions8) repre- The q class of reference points identifies ation to a set of objects, the manager sents another group of standards. These the association between NEF and OSF or must choose either the best effort or the define generic functionality and corre- MF, and between OSF and MF or QAF, atomic synchronisation property. The sponding managed object classes useful and between MF and QAF. The f refer- protocol provides means to handle mul- in many settings. Examples of this kind ence point identifies the association be- tiple replies resulting from this kind of of general functionality are object man- tween WSF and OSF or MF. The x refer- operations. agement function, state management ence point identifies the association be- function, event forwarding discrimina- tween OSFs of TMNs belonging to dif- This set of protocol services results in tion function, alarm reporting function, ferent administrative domains. powerful capabilities, which keeps the log control function, and security audit network traffic at a minimum, however, trail function, to name a few. The information architecture describes at the price of increasing the complexity an object-oriented approach for transac- and cost of the agent. SNMP, on the tion-oriented information exchanges. In other hand, has chosen the other end essence, this is covered by the adoption of the spectrum. This will be discussed of the OSIsm means of information below. exchange (ie. GDMO, CMIS/P). How- 8) The X.73x to X.75x series of ITU-T ever, it is noted that other concepts recommendations. supportive of location transparent fully Telektronikk 2/3.1999 147

149 distributed management applications are appropriate view of the network level works and related equipment such as being considered. This is further dis- resources (services) to the service routers, servers, workstations and other cussed in [10]. In addition, the informa- management layer. network resources [11]. tion architecture also considers the notion Service Management Layer (SML) of shared management knowledge and The following provides a brief overview This layer is responsible for the con- context negotiation to establish this. of SNMP, identifies enhancements in tractual aspects of services provided Naming and addressing of TMN SNMPv2 and SNMPv3, and comments to customers, such as service order resources is also considered. upon major differences compared to handling, co-ordination of services, OSIsm. A more in-depth coverage of complaint handling, QoS data, and in- The TMN physical architecture identifies SNMP can be found in [12], and [13] voicing. various options for configurations of provides a more in-depth comparison of physical systems or equipment (TMN Business Management Layer (BML) SNMP and OSIsm (as well as CORBA). building blocks), in terms of Operations This layer has responsibility for the System (OS), Mediation Device (MD), entire enterprise. It may relate to all The SNMP MIB specification language Q Adapter (QA), Data Communication the other layers. It should support the [14, 15] is significantly simpler than Network (DCN), Network Element (NE), decision-making process for optimal GDMO. It does not provide the notion of and Workstation (WS). These physical investments and use of new resources. class nor inheritance. Thus, it is not pos- entities contain the corresponding func- Management information in this layer sible to specialise one class from a tional block. Optionally however, they is not subject to standardisation. generic class. An SNMP object is a sim- may also contain some of the other func- ple data value or row entry whose num- tional blocks as well. As an example, the The transition from one logical layer to ber of elements is fixed and the elements OS may contain MF, QAF and/or WSF. another will often occur inside an OS. are of simple types. ASN.1 is also used Correspondingly, an interface between by SNMP; however, only a subset of Interfaces between physical entities are OSs will often address one logical layer, ASN.1 is used. Ultimately, object values identified by capital letters corresponding although one still has to consider the or row elements are of the ASN.1 basic to the letter identifying the reference manager (client) vs. the agent (server) types INTEGER, OCTET STRING and point between the functional blocks con- side of the interface.10) OBJECT IDENTIFIER [13]. Although tained in the respective physical entities. tables can be specified as part of the For example, the interface between NE MIB, whose lengths can dynamically and OS is Q (or more specifically Q3). 2.4 IETF management and be varied, SNMP does not provide any Further requirements regarding commu- SNMP vs. OSI systems notion of composite object of several nication protocols are also identified. management attributes manageable as a whole. As The management solution from IETF, the such, a table is not a named and manage- The TMN logical layered architecture set of SNMP standards, is also based on able object as such. SNMP objects are (LLA) was developed as an attempt to the concept of manager-agent (manage- singly instantiated. Multiple instantiated deal with complexity. LLA implies ment station and management agent objects can only be achieved through grouping specialised OSFs including respectively), as well as management tables and multiple table entries. clustering of related management infor- information base (MIB) and management mation into layers. TMN has identified protocol. It was first issued in 1988 and SNMP registers MIB object types using the following layers (in addition to the designed to provide a low complexity ASN.1 OIDs in a similar fashion as network element layer: cost efficient solution for vendor-inde- OSIsm. However, naming of objects pendent management of IP-based net- (instances) is significantly different from Element Management Layer (EML) OSIsm, as the OIDs are also used for Manages individual network elements naming. Thus, naming of a singly instan- possibly on a group basis. As an objec- tiated object is sufficiently distinguish- tive, this layer will provide a vendor- 10) It is stated in M.3010 [9] that a man- able by the OID itself. Table entries are independent view of network elements agement information model is asso- named by the OID of the table suffixed to the layers above.9) ciated with a management layer, and typically by the value of one or more ele- Network Management Layer (NML) may furthermore be used for the ex- ments whose fields can be considered as Manages and controls (abstract) change of information at the inter- the table key. This registration and resources associated with the network layer interface. We note that this last naming strategy prohibits the opportunity view related to the network elements part is a matter of definition. It is to define generic objects that can be within a domain. Must provide an clear, however, that the translation of imported and used in different contexts. information corresponding to the transition from one management layer The SNMP protocol [16, 17] does not to another does not occur as part of provide create, delete or action opera- 9) It is interesting to note (considering the management protocol. Manage- tions. This limits the possibility to spec- the q3 reference point between the ment information exchange is only ify complex operations with precise NEF and the Element-OSF of figure concerned with establishment of semantics. In addition, SNMPv1 was 18/M.3010 [9]) that the previous goal shared state knowledge between the limited by the maximum packet size of of providing a vendor-independent manager and agent (ie. synchronising the protocol data unit (PDU). This can view related to the management inter- the knowledge of the state of the same be a problem with respect to atomic table face to network elements, has been information entities at both sides of entry creation. SNMPv2 has a more relaxed. the interface). elaborate interaction scheme to remedy 148 Telektronikk 2/3.1999

150 this limitation. In addition to get and set is increased due to the lack of support Open Distributed Management Archi- operations, a get-next operation is pro- of multiple replies, in combination with tecture (ODMA) [20] will exploit these vided to ease the access to greater limited packet size. The increased net- opportunities. ODMA is further based on amounts of information. The order infor- work traffic can be a significant draw- the concepts from RM-ODP [21]. mation related to the naming scheme and back when the communication resource the column-by-column ordering of table is shared and the distances between man- A primary goal of the WBEM initiative entries, is used by the get-next operation. agers and agents are long. is to facilitate integrated management of The lack of support of the scoping and an enterprise wide range of resources, filtering facility also limits the power of The proposed SNMPv2 standards of from user applications, desktop systems the protocol. SNMPv2 does however 1993 did not get the anticipated accep- and servers, to systems, network devices support a get-bulk operation. It uses tance [11]. The functional enhancements and networks, as well as corresponding ordering knowledge to read more than have been welcomed, however, the secu- services. Solutions are based on cost effi- one object. This get-operation allows rity mechanisms were found too complex. cient general purpose computing plat- multiple variables or objects to be The SNMPv2 workgroup was not able to forms and engines (eg. web-browser, requested. In SNMPv1, this must be an resolve this problem, and as SNMPv2 web-server). Last summer, the ownership atomic operation. In SNMPv2 however, was progressed to draft Internet standard to WBEM was transferred to the Desktop non-atomic responses are allowed [11]. as of 1996, no security solution was part Management Task Force (DMTF),14) of it. The result has been delayed accep- who will be in charge of issuing WBEM The trap facility provides a means for the tance of SNMPv2. An SNMPv3 working standards. Currently, the only WBEM agent to notify the manager. In SNMPv1, group was charted in 1997, and a less standards are the Common Information this facility is associated with the agent complex and a more accepted solution to Model (CIM) standards, partly originat- itself rather than with objects. This limits the security challenge has been provided, ing from the Microsoft Hyper Media the use of this facility. SNMPv2 allows a as well as a few minor fixes related to Management Schema (HMMS). The notation to associate traps with objects. general SNMPv2 functionality. CIM standards cover a notation for spe- In addition, SNMPv2 extends the limita- cifying management information in a tion of having just one manager-agent UML-oriented way, as well as a specifi- level. A middle level of hybrid man- 2.5 Trends cation of a core set of management infor- ager/agents has been appointed. In addi- The primary focus thus far has been vari- mation and furthermore, some general tion, multiple top-level management sta- ous aspects relevant to specification and domain specific models. Currently, tions (management servers) may exist in communication related to the manager- WBEM is focused on developing an one management domain. This hierarchi- agent interface and interworking. Solu- XML15) vocabulary for CIM, enabling cal structure distributes and makes the tions in this area contribute to the basic XML-based encoding and communica- processing burden on each manager node management infrastructure. However, the tion. lower, as well as reduces the total man- Web-Based Enterprise Management agement network traffic. Corresponding (WBEM)12) initiative also provides an When considering relevancy for access to this scheme, an element manager tak- interesting set of solutions, with their network management, WBEM technol- ing the manager as well as the agent role, web-based management infrastructure. ogy can be used for several reasons and can send inform messages to the manage- in various architectural contexts. It can ment server. The management server Before taking a brief look at WBEM, provide a low cost solution for user inter- must provide a response to the inform we will note that advances in distributed faces and uniform access to heteroge- message, making this a reliable notifica- object technology also play an important neous management applications and tion. A manager-to-manager MIB has role. While TMN and SNMP are focused information. Thus, it is particularly rele- been developed for the support of man- on the communication interface between vant between a client and the middle tier ager-to-manager communication.11) the manager and agent, the ORB-based of a multi-tier architecture. Furthermore, distributed object technology is focused it can provide means for applications and Due to the goal of achieving low cost on APIs between applications. The systems management related to manage- simple agents, the choice of the connec- CORBA architecture and related specifi- ment of the access network management tionless unreliable user datagram proto- cations from OMG13) provides the basis systems and applications. A more open col (UDP) as required protocol has been for a set of important technologies, tech- question however, is whether WBEM or made. The unreliable nature of the trans- nologies considered important for XML technology are also suitable for port protocol increases the burden on the telecommunications management [18, system-system interworking and manage- applications as well as the total network 19]. Characteristic aspects and interesting ment communication with NEs. traffic. Considering that traps are unreli- topics related to these technologies will able, the manager must still resort to be identified and discussed further be- The idea of a platform supporting the polling to ensure that problems have not low. The work by ITU-T related to the facility of deploying and delegating man- occurred. Furthermore, network traffic agement functionality to agents in a dynamic fashion, have been considered 12) 11) It is not clear form [11, 13] if these 13) Object Management Group. An indu- 14) facilities are general in the sense that stry consortium of more than 800 they also support communication hori- companies, spanning a variety of 15) eXtensible Markup Language, zontally between mid-level man- businesses from IT vendors to telecom developed and specified by W3C. agers, or between top-level managers. operators. Telektronikk 2/3.1999 149

151 by the research community for several Above, we have pointed out that efficient 3 Evolving access years. The Java solution and the recently use of network resources for the purpose announced JMX16) may provide a viable of management is an important goal, in network architectures solution in this direction and deserve particular as communication distance The adoption and introduction of modern close consideration. increases. This is however somewhat access network technologies represents debatable as network bandwidth con- a giant change for the (access) network tinues to increase. The related topic of 16) Java Management Extensions. operator. This section will identify a few DCN design and technologies is anyhow characteristic factors related to adminis- important, although not getting much tration and management of access net- JavaManagement/ attention here. works, considering the transition from yesterdays technology to the future access network technologies and ser- vices. Although access to leased line ser- vices is an important area that also in- volves management, here the focus is the provisioning of access to end-user tele- AN N:1 N:M com services. However, it should be con- 1:N sidered whether provisioning and man- agement of leased line services should be more integrated with management of access network resources supporting on- N:M N:1 demand telecom services. UNI SNI SN 1 1 The access network of today or yesterday consists of passive twisted pair copper N N wires, their cables and distribution AN - Access Network frames at various sizes and levels 1:1 SN - Service Node throughout the local loop. Provisioning LUP LSP UNI - User Network Interface of a telephone service can be associated SNI - Service Node Interface with an instance of a switch port and its LUP - Logical User Port associated pair of wires manually cross- N:1 LSP - Logical Service Port connected in the distribution frames to reach the customer premises equipment Figure 2 Access network related entities (CPE). The administration of this physi- and relationships cal infrastructure is a challenging task in itself, but this network architecture does not involve active network management via management interfaces to network elements (NEs). The introduction of extended concentration nodes does not change this picture as this kind of NE is managed as part of the switch (Service Node) itself. Access service Tele service transport This situation is not significantly product type stream# type changed as self-contained multiplex systems are introduced to free copper pairs part of the way between the cus- tomer and the extended concentration node. The multiplex system will require SN remote management; however, its basic CPE type Access path* type type/role configuration is static and requires man- ual intervention to change. Likewise, the one-to-one physical association be- tween the switch port, the telephone ser- Access server vice and the copper pair at the customer layer type *, # premises is still valid. Path and stream, to be interpreted in a wide sense The introduction of new technologies client/server All relations are N : M like PON and VDSL, and for instance related ATM technology, implies intro- duction of new types of entities and rela- Figure 3 Relationships among access network resource tionships, and the vanishing of the one- types and service types 150 Telektronikk 2/3.1999

152 to-one relationship between the service supporting systems evolution and flexi- transmission media layer. A network port and the physical infrastructure. The bility will be important in order to handle layer is defined independently of the introduction of several technologies and the uncertainties associated with which other layers and each is decomposed into network layers and corresponding logical network architecture to adopt. three basic functions: Adaptation, Termi- user ports and service ports makes a nation and Matrix Connection, and may more complex picture. This is illustrated have its own operations and maintenance in Figure 2. Figure 3 illustrates the multi- 4 Generic architectures capabilities. The model is useful in de- tude of possible technologies and service and models fining the managed objects related to the combinations. While it is a great chal- management interface of the access net- lenge to decide the network architecture, Although different technologies are work, and the goal is that different tech- it is an equally great challenge to come available for offering broadband solu- nologies from different vendors can be up with the best systems architectures to tions to end users, these have a lot in managed in a uniform way. Although support administration and management common concerning management. In the being focused toward OSIsm by identi- of this complex setting. An essential following, central standards are described fying a Q3-interface, G.902 can still be objective should be to provide for each that are suitable for handling the considered as general with respect to service node type (role), a consistent and resources of different technologies in a management if one allows other manage- integrated high level view of the access uniform way. G.902 [22] provides a gen- ment paradigms as well. network, hiding differences in technolo- eral architecture of access networks gies and technology specific details. Per- while G.805 [23] provides a generic G.902 establishes a functional architec- formance issues and load balancing will functional architecture of transport net- ture and how each of the functional become important issues as dynamic con- works. In addition to these general net- groups are interconnected (see Figure 4). figuration changes of the access services work architectural models, common The AN is broken down into functional and network will be possible. On-line models within specific management groups (some examples of functions are user access to such configuration man- areas are also available and will be given). They are: User Port Function agement services will result in a high briefly described. Management solutions (UNI termination, signalling conver- volume of configuration transactions. based on these standards will help in sions), Service Port Function (protocol achieving a common view of the network mapping, testing of SNI), Core Function These changes result in a need for the resources and parameters to be managed, (bearer channel concentration), Transport introduction of a new network level man- and thus, providing a basis for integrated Function (physical media functions, mul- agement system keeping track of the management solutions. tiplexing, cross connect), and AN-Sys- associations and connectivity (eg. VPCs) tem Management Function (FCAPS17)). between the LUPs and LSPs of the Q3 Agent and MIB are located in the access network. This system must inter- 4.1 Generic architectures face with subordinate management sys- G.902 provides an architecture of an tems for the particular technology spe- Access Network from a high level per- 17) Fault, Configuration, Accounting, cific subnetworks and corresponding spective by identifying and defining the Performance, and Security manage- NEs, including management of SNIs, the functions and requirements above the ment. Defined in X.700 [1]. later co-ordinated with the management systems of the SNs. This new network level management sys- tem(s) will be the system to inquire for access network resources and status, and furthermore, to request for configuration Q3 operations, in the case of service provi- sioning. These systems must also provide Access Network essential information to the access net- work planners. The system administrat- AN System Management Function ing the physical resources will most likely be separate from the former, and must be augmented to cover administra- tion of the new physical resources in- cluding PON and xDSL physical Core Core resources, to provide a user friendly view Funct. Funct. User User of the physical network infrastructure. Port Transport Port Function Function Function UNI SNI The strategic introduction of new man- agement systems as well as the transition of the role of the system administrating the physical infrastructure, represent a T1304240-95/d03 significant challenge. The access network operator must develop a strategy to en- User Bearer and User Signalling information able a step-by-step introduction of new Figure 4 Example of functional architec- technologies. Mechanisms and strategies Control and Management ture of an Access Network (Fig. 3/G.902) Telektronikk 2/3.1999 151

153 AN-SMF. The AN-SMF acts as an Agent G.805 describes how to decompose a (eg. included alarm objects, alarm status, to the TMN and as a Manager to the AN network into specific path layer networks severity, or probable cause), and man- functions such as UPF, CF, etc. How- which are likely to be independently agementOperationsSchedule to allow ever, G.902 does not advice a manage- managed and where paths across the the reporting of alarm summaries to be ment architecture within the AN domain. specific path layer network will be set up performed periodically. independently from the set-up of paths in A user port function has a fixed associa- other specific path layer networks. A net- Alarm report notifications as defined in tion with one and only one SNI through work can be decomposed into a number X.733 [26] are used to indicate that an provisioning. This concerns all the bearer of independent layer networks with a exceptional event or condition has been capabilities at the user. In case of shared client/server association between ad- detected in the NE. These notifications UNI, which may support more than one jacent layer networks. Each layer net- are emitted from objects that represent logical user port function at the same work can be separately partitioned in a the affected resources or model the time, eg. using ATM, more than one SN way that reflects the internal structure of detecting resource. The alarm reports may be accessed through a single UNI that layer network or the way that it will provide a very detailed alarm event de- (eg. one VP for each logical access). be managed. The actual decomposition scription (eg. event type, perceived used to generate the specific path layer severity, probable cause, physical loca- The objective of this Recommendation networks is dependent on the technology. tion of the affected resource, and addi- is to describe an access network concept While G.902 asserts adoption of the tional specific information on the event) that provides flexibility towards future G.805 principles, the latter does give according to the information specified in access types, eg. for interactive video valuable additional input to the mod- the object classes representing the services. It is VB5 focused, deals with elling of access networks. G.805 presents resources and their associated attributes. semi-permanent connections, and con- diagrammatic conventions for producing nects logical ports at Service node to graphical models. Examples of events are detection of their peers at the customer. transmission data errors, the crossing of a performance threshold, and the detection Access networks with many technologies 4.2 Common models of faulty equipment. Alarm reports may from a mix of vendors may constitute a The Systems Management Functions be stored in alarm logs represented by rather complex network. G.805 describes from OSIsm provide useful functionality specific X.735 [27] Alarm Log objects. networks from the perspective of the generally applicable to many areas. The The default severity, which is assigned to information transfer capability. For trans- standards identified in the following also each alarm by the NE, may also be modi- port networks, this recommendation de- provide common facilities, although each fied from the OS. scribes the functional architecture in a dedicated to a specific management func- technology independent way. It provides tional area. These are alarm surveillance, The purpose of Q.822 [28] on Perfor- means to describe network functionality performance management, and physical mance Management is to provide a set in an abstract way in terms of a small resource management. These standards of application messages and associated number of architectural components. all define Q3 interface information mod- support objects for parameter collection These are defined by the function they els and thus are based on OSIsm. While and thresholding aspects of Performance perform in information processing terms it is not always feasible to require the use Management (PM). Q.822 provides defi- or by the relationships they describe be- of the CMIP protocol, the adoption of nitions of managed objects and attributes tween architectural components. these standards provides a basis for eas- and associated functionality, and further- ier integration of vendor-specific element more, specifications of the services, func- The client/server relationship is funda- managers. tional units, and protocols related to PM. mental and described as the association The PM functions supported are data between layer networks that is performed The ITU-T Recommendation Q.821 [24] storage, thresholding and data report- by an adaptation function to allow the on Alarm Surveillance specifies the Q3 ing. The supporting object classes are link connection in the client layer net- interface requirements for communica- currentData, historyData and thresh- work to be supported by a trail in the tion between an OS and an NE and is a oldData. Examples of monitored param- server layer network. The transport generic framework for the management eters described in Q.822 are: processing functions, adaptation and ter- information and functions related Slip second; mination, give relationships between to alarm handling. It supports the TMN topological components that give an management service component de- Code violation; abstract description of the network. scribed in M.3200, Management Services Errored second (of various types); [25], and defines a generic information A port is given as the representation of model for alarm surveillance. It provides Loss of signal second; the output of a trail termination source a set of application messages and asso- Protection switching count; or unidirectional link connection, or the ciated support objects for the support of input to a trail termination sink or uni- management communication. It covers Protection switching duration. directional link connection. This could fit alarm reporting and logging, manage- into the G.902 port concept, but it seems ment of alarm severity profiles, alarm The duration of the performance interval like G.902 defines functional compo- summary reporting and management can be set, and Scanner objects may be nents for AN covering all layers at the operations scheduling. It defines classes used to aggregate measurements from same time. such as currentAlarmSummaryControl a number of Current Data objects into which is used to specify the criteria for summary reports or to perform statistics the generation of the summary reports on the measurements. 152 Telektronikk 2/3.1999

154 The described generic information model important access network technologies, the most promising approach to achieve for performance management will be ADSL and PON, will be the subject of large-scale full service access network useful for manager/agent communication this section. ADSL is the most standard- deployment that could meet the evolving between OSs and NEs in TMN Manage- ised xDSL technology also regarding service needs of network users. It has ment solutions, and also for SNMP man- management. Even if parameters will be been shown that this APON approach aged NEs using QAs/MDs. different for other xDSL technologies, could support a wide range of FTTx the management regime for these tech- access network architectures Fibre-to- ES 201 097-1 [48] on the other hand, is nologies will be similar to the one estab- the-Building, Cabinet, Curb, or Home. part 1 of a series of ETSI Resource Man- lished for ADSL. PON was combined with copper technol- agement standards to come. It provides a ogy meaning that the ONUs are equipped generic model for the management of NE However, management issues related to with Very-high speed Digital Subscriber resources (hardware and software). It several other network technologies and Line modems. defines the managed objects to be appli- equipment are also important for man- cable at both the OS-NE and OS-OS agement of access networks. While not The FSAN initiative has produced a set interfaces. The functional descriptions of covered in this article, we will just men- of requirement specifications for Opera- physical resource management (func- tion that the following issues are impor- tion, Administration and Maintenance tional requirements and management tant; operator management of CPEs, where it has attempted to incorporate functions) are based on M.3400 [29]. All ATM cross-connect and multiplexer applicable standards where they exist. applicable OSI Systems Management management, and VB5 management. The Optical Access Network workgroup Functions (X.730-X.750 series) are Management related to IP is also relevant in FSAN has created an APON specifica- reused. The information model is fully in the case of IP technology being used tion. It has been presented to several aligned with M.3100 [30]. as an access network technology. standardisation bodies, eg. ITU, ATM Accounting management is not relevant Forum and ETSI. The first draft of the This standard covers the configuration with respect to xDSL and PON technol- ITU G.983 [12] specification was based management and the fault management ogy as such. However, it may be an issue on this specification. areas (based on Q.821), including testing, related to the usage of client layer timing, protection and inventory. Within resources, such as IP or ATM resources. each management view, the standard 5.2 ADSL technology addresses the management of physical An ADSL system consists of modem resources (hardware) comprising a net- 5.1 The Full Service Access pairs and some associated equipment work element. This includes the periph- Network (FSAN) initiative necessary to offer ADSL as a customer eral/accessories parts (eg. disk, power A group of telecommunication network access. All development of ADSL supply, fan equipment, etc.) subject to operators and equipment manufacturers modems is based on the standard T1.413 management actions. have undertaken an international initia- [31], that describes transmission related tive to create requirement specifications details like line code, receive signals, for access systems. Although FSAN is organisation of transmitted and received 5 Management of not a standardisation activity, it has pro- signals into frames, and electrical and duced valuable results concerning AN mechanical specifications of the network specific access through development of a set of require- interface. Further, the interface between network technologies ments specifications. the telecommunications network and the customer installation in terms of their The identification of management An ATM Passive Optical Network interaction and electrical characteristics. aspects and needs associated with two (APON) was identified by the project as ITU-T is preparing a set of standards for Q NT AN TE Management Management Entity Entity Home P P Broadband Network ATU-R ATU-C Network H Loop H Y Y TE T/S T-R U-R U-C V-C V Management Interface Figure 5 G.997.1 System Reference Model Telektronikk 2/3.1999 153

155 ADSL (eg. G.992.1 [32]) for discrete als state that the ADSL NE shall collect ones listed here cover medium specific multitone encoding for transmission usage data at the SNI and UNI for billing layers only: rates up to 8.192 Mbit/s downstream and purposes. 640 kbit/s upstream. ADSL Forum WT-025: CMIP Specification for ADSL Network 5.2.3 Example ADSL Element Management (9/98) [34]. This 5.2.1 Embedded management parameters document specifies a CMIP based man- Operation Channel The ADSL technology is depending on agement framework. The modem at each end of the copper copper lines with varying quality, and line, ATU-C and ATU-R (see Figure 5), therefore threshold settings, threshold The TR-005 [35] and TR-006 [36] MIB communicate with each other over crossing alarms for bit errors, s/n ratio definitions suggest an SNMP agent local- embedded operations channel (clear eoc), and attenuation are all important para- ised in ATU-C and acting as proxy for well documented in T1.413 and G.992.1 meters. Noise Margin parameters are ATU-R. ADSL Forum WT-022 [37] and [32]. The latter describes eoc organisa- defined to control the noise margin to WT-023 [38] support and complement tion and protocol, eoc message structure, assure acceptable BER (

156 TR-002 also addresses the control and VPI = n VPI = 0 CPE management planes related to supporting VCI = y VCI = y ATM in the user plane. It includes the ATM 1 ATM PVC support, signalling for SVC access switch support and operations and maintenance DSLAM 2 functionality to support ATM over ADSL. Concerning ATM Layer Manage- ment, ADSL Forum suggests that ATM management of the CPE shall be based on ILMI 4.0 [42]. ATM A management system should be able One VP to manage the entire ADSL network. DSLAM management interface must per subscriber n therefore handle; VP crossconnect with VPI remapped and Subscriber line (VPI/VCI/port no/cus- tomer id); Figure 7 DSLAM with VCI tunneled NE IP address; ATM Cross-Connect through DSLAM WAN port (no/type); Master profiles and service profile. transferring the required information 5.2.6 DSLAM and DSLAM across the U interface for presentation to management functions 5.3 Passive Optical Network the Q or S/T interface. PON To interconnect multiple ADSL users to A likely configuration of a PON is shown a high-speed backbone network, the AN IETF is also making an effort at ADSL in Figure 8, with a single OLT at the provider uses a Digital Subscriber Line management developing an ADSL MIB exchange side and a number of remote Access Multiplexer (DSLAM). DSLAM in line with the SNMP v2 standard. ONUs. OAM functions that are relevant might connect to an ATM network that for PONs are: Configuration, Perfor- An ADSL system and subnetwork can be can aggregate data transmission at trans- mance, Fault and Security Management. modelled according to the G.805 archi- port network data rates. At the downward tecture. So far, no model is presented in end of each transmission, a DSLAM The Optical Distribution Network standards. This is perhaps due to the fact demultiplexes the signals and forwards (ODN) provides the optical transmission that both DSLAM and ATU-R realise them to appropriate individual ADSL medium for the physical connections and rather advanced network functions. connections. ADSL can be combined consists of passive optical elements; opti- ADSL Forum WT-025 [34] contains a with ATM, SDH and Ethernet. G.992.1 cal fibres and cables, optical connectors, CMIP Specification for ADSL Network describes how ATM and SDH are used passive branching components, passive Element Management. This document as clients for ADSL. optical attenuators and splices. Individual suggests an adslLine (trail) containing ODNs may be combined and extended adslChannels (connection). AdslLineTTP The typical DSLAM might perform a VP using optical amplifiers. and adslChannelTTP objects are defined. cross-connect function and will be trans- The adslChannel is serving the client parent to VCI values (Figure 7). PVCs When modelling a simple PON without layers (eg. ATM). from each customer are effectively tun- amplifiers and multiplexers, the transmis- nelled through the DSLAM to the serv- sion can be split into network layers Other ADSL Forum contributions have ing ATM switch. A unique VPI is according to G.805. Using one Optical slightly different suggestions for an assigned to each customer for the link Medium Layer and one Optical Section architecture. Common for the contribu- between the DSLAM and the ATM Layer, the latter to support the Client tions are that adslTTP shall cover physi- switch. The DSLAM is effectively an Layer, this can be sufficient for mod- cal matters such as lineCoding, support- ATM cell multiplex in the upstream elling purposes. The association of an edChannelTypes, attenuation, rate, s/n direction and a VP based cell router OLT with its ONU will then be modelled margin, output Power, etc. AdslChannel in the downstream direction. through the association with the fibres. shall contain information about how the In such a simple model a PON object line is used concerning type, id, rate, etc. ADSL Forum TR-002 [41] identifies and represents the one or two fibres of the Five types of ADSL Line Type are defined: defines the functional blocks of an ATM- bi-directional point-to-multipoint link based ADSL access network, which are No channels exist; between the OLT and its ONUs, each formally referred to as B-ISDN Network located at a specific fibre, including the Fast channel exists only; Termination (B-NT) for the ADSL associated splitters and other passive modem and Access Node for the access Interleaved channel exists only; components. multiplexer system. The report addresses Either fast or interleaved channels can the layer 2 protocols and specifically Furthermore, one single object may be exist, but only one at any time; describes the implementation of ATM used for the OLT (oltTtp) and one for transport over ADSL links. Both fast and interleaved channels exist. each of the ONUs (onuTtp) to represent Telektronikk 2/3.1999 155

157 Regarding the equipment fragment, ONUs and OLTs are represented by racks, shelves, slots with field-replace- Q3 able units, etc., just like equipment of ODN other technologies. Parameters especially Passive relevant for ONUs and OLTs are: SNI Optical Splitter UNI Temperature/temperature threshold; OLT ONU Bias current/Bias current threshold; IFPON Optical power/optical power threshold; Output frequency/threshold (outside OLT - Optical LineTermination SNI - Service Node Interface pre-set operating frequency window); ONU - Optical Network Unit UNI - User Network Interface ODN - Optical Distribution Network IFPON - PON specific Interface Line width broadening; Tx BER/threshold; Figure 8 PON configuration Optical signal/noise ratio. A main role of the alarm management functionality is the long term proactive prevention of failures, and monitoring is all the functionality of these locations model suggests the PON to be managed performed to control transmitter stability avoiding the need for vendor specific as one element, through a Q3 manage- and component degradation. Frequency detailed modelling of the sublayers. Each ment interface. The information defined stability depends on the temperature, and onuTtp object is in turn associated with for the PLOAM provides the basis for the statistics made on monitored data can a number of userPortTtp objects, which management functions available through identify failure trends. Attenuation must represent user ports. Each ONU may the Q-interface. be monitored also to avoid introduction have a number of user ports. The user of additional splitters. ports interface over UNIs to a terminal There are several ongoing efforts on (CPE). modelling optical networks, but the mod- els are not yet stable and consistent. Even 6 Management platforms Future ODN may have NEs that are all if the models are describing slightly dif- optical and have signals that are pro- ferent functionality in their layers, the and component based cessed in the optical domain. Use of managed objects realised for these layers SW technologies various forms of wave division and sub- must cover functions for termination, carrier multiplexing may be assumed, adaptation and supervision. Management Although OSIsm and SNMP are focused allowing several subgroups of ONUs to functionality that must be covered for the on the interface between the manager and share the same fibre. Then a model using PON is, eg.: agent, software platforms for develop- a wider division of the optical layers ment as well as for run-time use, has Frequency and bandwidth allocations; should be applied. ITU-T has several rec- always been an important topic related to ommendations relevant for optical net- Supervision of optical signal power management systems. In general, plat- works management. G.982 [43] deals level; forms provide ways of making the job with the characteristics of an Optical for the application developer easier. For Monitoring of optical signal/noise Access Network (OAN) based on instance, software services (eg. middle- ratio; 64 kbit/s bearer capabilities up to and ware19) services) provided by a platform including ISDN primary rate services. Alarm by loss of optical signal; can be utilised by an application de- G.983 [44] focuses on a network to veloper, and so he does not have to de- Signal detection and frame alignment. support services with bandwidth require- velop that piece of software. Another ments greater than ISDN basic rate and is facility of a platform can be to provide An ODN may be made more reliable concentrating on ATM over PON. G.872 half-made skeletons or solutions that can using redundant fibre up to first splitter, [45] deals with optical transport network be further specialised and completed by thus making protection switching pos- functionality described from a network the developer. sible. A dedicated wavelength may be level viewpoint using the modelling used for test/management purposes. To mechanisms from G.805. support management, sometimes an extra fibre is deployed, running continuously G.983 describes capabilities of an ODN along with all the fibres that constitute the and related OAM functions that must ODN, thus enabling testing and supervi- 19) Middleware a loose term for soft- exist, and parameters that should be sion. Faults that may occur in an ODN detected in OLT and ONU. It also ware and corresponding services, are fibre breakage and loss/reduction of defines the messages that must be avail- layered between communications- optical signal. Fibre breaks can be local- able in the PLOAM18) channel. The and-operating systems facilities and ised by supervision of reflected signals. the applications. Middleware consti- Loss/attenuation in fibres is supervised by tutes much of the application develop- 18) Physical-layer OAM. monitoring received optical signal. ment environment. 156 Telektronikk 2/3.1999

158 X/Open and TMN (previously NMF) While Java and Java RMI21) and ment technologies will be addressed have developed API20) standards for Microsoft DCOM22) represent technolo- briefly in this section. access to OSIsm or SNMP middleware. gies challenging CORBA (ver.2), the These standards are programming lan- appearance of transaction based server- In general, the mapping from one tech- guage dependent and for OSIsm they are side component models like Enterprise nology to another can be done at the for instance categorised into ASN.1/C++, Java Beans (EJB), and the CORBA Com- manager-side or the agent-side by CMIS/C++, and GDMO/C++ API stan- ponents Model23) go one step further. By embedding a gateway in either the man- dards. formalising the notion of components ager or the agent respectively. The gate- and hiding more software details from way can also be located in a standalone The CORBA initiative by OMG (see the application developer, software de- (mediation) device. On the other hand, above) also provides a basis for plat- velopment is made more efficient and the one could translate an interface model forms, but is even more general as its support for assuring component (applica- from one language to the other, and like- focus is specification of programming tion) interoperability is improved. wise use the corresponding technology at language independent APIs for middle- both the manager and the agent side. ware and ORB services. The ORB-based Based on these emerging trends in dis- However, the more likely situation solution enables full distribution of appli- tributed software technologies, one can involves a gateway, to let investments in cations where access and location trans- hope that a better basis is provided for to existing solutions carry through. Yet parencies are supported. The CORBA enable integrated management, than what another solution is to let the application Interface Definition Language (IDL) and we have experienced so far in the realm developer take care of the mapping, by the CORBA services, specified in IDL, of TMN. Several issues are unsettled. For having him program against several man- enable portability of applications and instance, are NE providers willing to agement technologies. This does make provide a basis for interoperability deliver element managers with well spec- the burden on the application developer between applications. Based on these ified component interfaces, and thus let considerably greater. While this is not properties, the CORBA technology is their element manager be integrated in effective for programming against APIs promoted as an excellent distributed sys- an operators integrated element manage- representing similar interfaces, it may be tems technology and the future technol- ment system? Are operators willing to a reasonable strategy if the technology ogy for heterogeneous systems and appli- take greater responsibility themselves in mapping also involves a conversion from cation integration. Thus, many of the developing integrated solutions based on a model and interface at one management challenges identified above concerning integration of subsystems and software layer to another. For instance, if the integration of systems in the network and components from NE vendors and man- application realises a mapping between service management layer, can possibly agement platforms providers? Will a an NE-level to a network level model, best be handled using CORBA technol- combined buy and build approach like one can envisage that the NE-level inter- ogy. this give greater freedom of choice and face is based on CMIP, while the net- better control of the operators manage- work level interface is based on CORBA Since the CORBA technology addresses ment systems portfolio? In any case, it IDL. many of the challenges experienced with is likely that an operator will acquire an the development of large complex and increasingly complex management sys- Due to the simplicity of the SNMP man- integrated TMN solutions, two issues tems portfolio to manage the access net- agement information model, automatic appear: How to integrate CORBA tech- work, and the management of systems specification translation from OSIsm- nology with OSIsm or SNMP technol- and applications as well as providing or CORBA-based models to SNMP is ogy, and, can OSIsm or SNMP technol- support for systems evolution will not possible without losing important ogy be replaced by CORBA technology become more important. Application semantic information. Mapping from in an advantageous fashion? The first server technology as well as systems SNMP to OSIsm or to CORBA IDL on issues will be addressed in the next sec- and application management solutions the other hand, is achievable. By locating tion. It should be noted that TMN does will become correspondingly important. for instance a CMIP/SNMP gateway near not in itself exclude CORBA technology, the SNMP agent, it is possible to extend and ITU-T SG4 currently focuses on how the functionality of the SNMP agent and to harmonise CORBA and similar tech- 7 Integrated management introduce event reporting, logging and nologies within TMN to take advantage and interworking summarisation functionality closer to the of what these technologies have to offer. NEs. However, the mapping of SNMP While SNMP and OSIsm can be con- The development of systems plans and table entries to separate CORBA IDL sidered as being at opposite ends of systems architectures to enable integrated interfaces may result in a significant a complexity scale, the CORBA tech- telecommunications management is a overhead if the number of interfaces nology has the potential to be profiled great challenge to network operators. becomes large. and appropriately suited for a range of Issues relevant to this challenge have settings. been identified above. Although just Due to the popularity of CORBA, en- being one other part of the challenge abling interworking between CORBA of integrated management, the issue of based environments and OSIsm based interworking between different manage- environments is of particular relevance. Both modelling notations are based on an 20) Application Programming Interface. object-oriented modelling approach. This 21) Remote Method Invocation. provides a good starting point for trans- 23) Supported by CORBA 3.0, becoming lating models. However, while transla- 22) Distributed Component Object Model. available in 1999. tion from CORBA to OSIsm is rather Telektronikk 2/3.1999 157

159 straightforward, the translation in the 9 References Technologies and implementations. opposite direction is more difficult. The IEEE Press, 1998. latter case is also more relevant to access 1 ITU. Management framework for network management scenarios. Most Open Systems Interconnection (OSI) 14 Rose, M T, McCloghrie, K. Structure of the difficulties due to the differences for CCITT applications. Geneva, and identification of management have been resolved and a standardised 1992. (ITU-T X.700.) information for TCP/IP-based inter- way of achieving specification transla- nets. Internet Engineering Task Force tion has been suggested by X/Open [46]. 2 ITU. Information technology : Open (IETF), May 1990. (RFC 1155.) Still, there is semantic information asso- Systems Interconnection : Systems ciated with GDMO conditional packages management overview. Geneva, 15 Case, J et al. Structure of Manage- that may have to be manually resolved. 1992. (ITU-T X.701.) ment Information for Version 2 of the While based on the specification transla- Simple Network Management Proto- tion from X/Open, the final submission 3 ITU. Information technology : Open col (SNMPv2). Internet Engineering to the OMGs CORBA/TMN interwork- Systems Interconnection : Structure Task Force (IETF), Jan. 1996. (RFC ing RFP [47] provides a gateway solution of Management Information: Guide- 1902.) for dynamic conversion between the pro- lines for the definition of managed tocols and associated behaviour of the objects. Geneva, 1992. (ITU.T 16 Case, J D et al. Simple Network Man- two technologies. To support all interac- X.722.) agement Protocol (SNMP). Internet tions possible in CMIP, this solution Engineering Task Force (IETF), May relies on previously specified CORBA 4 ITU. Information technology : Open 1990. (RFC 1157.) services like naming, life cycle, event, Systems Interconnection : Structure property, and collection services. of management information : Man- 17 Case, J et al. Protocol Operations for agement information model. Geneva, Version 2 of the Simple Network Although the just mentioned specifica- 1992. (ITU-T X.720.) Management Protocol (SNMPv2). tion efforts represent a significant step Internet Engineering Task Force in the integration of these two technology 5 ITU. Information technology : (IETF), Jan. 1996. (RFC 1905.) domains, there is much experience to Abstract Syntax Notation One be acquired before one can be sure what (ASN.1) : Specification of basic nota- 18 Object Management Group. Telecom architectures and technologies to go for, tion. Geneva, 1994. (ITU-T X.680.) Task Force. CORBA-based telecom- considering large scale management munications network management solutions. 6 Krsen, A-G. The structure of OSI system. White Paper, Framingham, management information. Telektro- MA, USA, 1996. nikk, 89 (2/3), 9096. 1993. 8 Conclusion 19 SMART TMN : Technology Integra- 7 ITU. Common management informa- tion Map. GB 909. Issue 1.1. TeleM- Several technologies are challenging the tion service definition for CCITT anagement Forum, Oct. 1998. Mor- GDMO/CMIP-based solutions from OSI applications. Geneva, 1991. (ITU-T ristown, NJ, USA. Systems Management. SNMP-based X.710.) solutions will play an important role 20 ITU. Information technology : Open regarding management of network ele- 8 ITU. Common management informa- Distributed Management Architec- ments. However, trade-offs must be tion protocol specification for CCITT ture. Geneva, 1997. (ITU-T X.703.) assessed to choose between the two. applications. Geneva, 1991. (ITU-T CORBA-based solutions will become X.711.) 21 Lnsethagen, H. Introduction to RM- more relevant at the network and service ODP : A reference model of open management layers, and technology 9 ITU. Principles for a Telecommuni- distributed processing. Telektronikk, interworking solutions will be required. cations management network. 94 (1), 5566, 1998. The WBEM initiative may also provide Geneva, 1996. (ITU-T M.3010.) interesting solutions within specific 22 ITU. Framework Recommendation areas. 10 Sidor, D J. TMN Standards: Satisfy- on functional access networks (AN) : ing todays needs while preparing for Architecture and functions, access The development of systems plans and tomorrow. IEEE Communications types, management and service node systems architectures to enable integrated Magazine, 36 (3), 5464, 1998. aspects. Geneva, 1995. (ITU-T telecommunications management is a G.902.) great challenge. Emerging component 11 Stalling, W. SNMP and SNMP v2 : based software technologies provide a The infrastructure for network man- 23 ITU. Generic functional architecture basis for application integration. How- agement. IEEE Communications of transport networks. Geneva, 1995. ever, great challenges prevail in the area Magazine, 36 (3), 3743, 1998. (ITU-T G.805.) of the network and service management layers where new kinds of network and 12 Rose, M T. The Simple book : An 24 ITU. Stage 2 and stage 3 description service resources must be managed in an introduction to Internet management. for the Q3 interface : Alarm surveil- integrated and efficient fashion. 2nd. ed. Prentice Hall, 1994. lance. Geneva, 1993. (ITU-T Q.821.) 13 Aidarous, S, Plevyak, T. Telecommu- 25 ITU. TMN management services and nications network management : telecommunications managed areas : 158 Telektronikk 2/3.1999

160 overview. Geneva, 1997. (ITU-T 38 ADSL Forum. CAP Line Code Spe- 43 ITU. Optical access networks to sup- M.3200.) cific MIB. Feb 1999. (Technical port services up to the ISDN primary Report TR-015/WT-023 revision 5.) rate or equivalent bit rates. Geneva, 26 ITU. Information technology : Open 1996. (ITU-T G.982.) Systems Interconnection : Systems Management : Alarm reporting func- 39 ITU. Splitterless Asymmetric Digital 44 ITU. High speed optical access sys- tion. Geneva, 1992. (ITU-T X.733.) Subscriber Line (ADSL) Transceivers. tems based on passive optical net- Melbourne, 1999. (ITU-T temporary work. Geneva, 1998. (Com 15-R 11- 27 ITU. Information technology : Open document MA-008/draft rec. E/ITU-T determined draft rec. Systems Interconnection : Systems G.992.2.) G.983.) Management : Log control function. Geneva, 1992. (ITU-T X.735.) 40 ADSL Forum. Element Management 45 ITU. Architecture of Optical Trans- Implementation Guidelines for DMT- port Networks. Turin, 1998. (ITU-T 28 ITU. Stage 1, stage 2 and stage 3 de- based ADSL systems. Feb 1999. draft rec. G.872.) scription for the Q3 interface : Per- (Contribution 99-048.) http://www. formance management. Geneva, 46 Open Group Preliminary Specifica- 1994. (ITU-T Q.822.) tion. P509. Inter-domain Manage- 41 ADSL Forum. Interfaces and System ment : Specification Translation. 29 ITU. TMN management functions. Configurations for ADSL : ATM Mar. 1997. ISBN 1-85912-150-0. Geneva, ITU, 1997. (ITU-T Mode. Mar 1997. (Technical Report. M.3400.) TR-002.) 47 Object Management Group. JIDM members.html Interaction Translation. Final sub- 30 ITU. Generic network information mission to OMGs CORBA/TMN model. Geneva, 1995. (ITU-T 42 ATM Forum. Integrated Local Man- interworking RFP. Framingham, M.3100.) agement Interface (ILMI) Specifica- MA, USA, 1998. tion. Ver. 4.0. Mountain View, CA, 31 ANSI. Asymmetric Digital Sub- USA, 1996. (af-ilmi-0065.000.) 48 ETSI. Telecommunications Manage- scriber Line (ADSL) Metallic Inter- ment Network (TMN); Resource face. 1995. (ANSI T1.413). atmforum/specs/aproved.html management. Valbonne, 1998. (ES 201 097-1.) 32 ITU. Asymmetrical Digital Sub- scriber Line (ADSL) Transceivers. Melbourne, 1999. (ITU-T temporary document MA-007/draft rec. G.992.1.) 33 ITU. Physical Layer Management for Digital Subscriber Line (DSL) Trans- ceivers. Melbourne, 1999. (ITU-T temporary document MA-009/draft rec. G.997.1.) Tor Breivik (52) graduated from the Norwegian Univ. of Science and Technology, Electro-technical line in 1972 specialising in Radio. He has been with Telenor R&D from 34 ADSL Forum. CMIP Specification 1973 and has experience from work with testing, reliability, for ADSL NE Management. Sep quality control, QoS and O&M systems. Since 1991 he has 1998. (Working Text WT-025.) been working with TMN and management of Access Net- works in particular. email: [email protected] 35 ADSL Forum. ADSL Network Ele- ment Management. Mar 1998. (Tech- nical Report TR-005.) http://www. 36 ADSL Forum. SNMP-based ADSL LINE MIB. Mar 1998. (Technical Hkon Lnsethagen (36) is Research Scientist at Telenor Report TR-006.) R&D, Kjeller, Network and Service Management Platform members.html Unit. Currently, his focus is access network management and ATM transport network management. His research 37 ADSL Forum. DMT Line Code Spe- interests are systems architecture, systems evolution and cific MIB. Mar 1999. (Technical specification techniques. Report TR-014/Working Text WT- email: [email protected] 022v4.) members.html Telektronikk 2/3.1999 159

161 Home networks: New challenges for network operators MARKUS WYSS AND FRDRIC PYTHOUD 1 Introduction HN is the termination of the broadband medium is electronically contaminated network infrastructure. Many different and has a prohibited frequency band You are at home, sitting in front of your areas are touched by the Home Network (different in Europe than in the US). It desktop, quickly reading the news on the as shown in Figure 1 and discussed in the is still for the time being a very good web. Suddenly, the phone rings. As you following sections. candidate for typical home automation are used to, you click your mouse and get applications. the conversation with your friend. He is very keen on the football match that is 3 Media types Wireless infrared and radio offer the very nice advantage that they do not being played now. In order to share his Twisted pair, plastic optical fibre (POF), require extra cabling. However, the enthusiasm, you display the correspond- power lines, air (for wireless infrared or reach of these systems is most of the ing TV channel on your screen. At this radio) are the most known media types time limited to the room except for moment, you notice that you have for- that may be used for home network some radio systems operating at gotten to switch off the light in the living applications. Each of these media types adapted frequencies. For radio wireless room and a warning on your screen has some advantages and disadvantages communication in the customer pre- informs you that the electric cooker is that are briefly mentioned hereafter: mises currently unlicensed spectrum is ready to explode. By another two mouse used. For available wireless LAN sys- clicks you perform with simplicity ... all Twisted pair is a medium that is very tems in the ISM-Band (2.4 GHz) and necessary actions. popular for business LAN (Ethernet, future HIPERLAN (High Performance ATM 25). The standards available Radio LAN) systems in the 5 GHz This imaginary story is certainly not for specify twisted pair to be able to carry band house coverage can be achieved. today. But it may typically represent up to 100 Mbit/s over several tens of tomorrows home network features. The metres (cat. 5). More often than not, home network represents the last drop of the actual cabling used for POTS 4 Low bitrate trans- communication networks by which the distribution inside buildings is not information gets to or from the user. It is adapted for such bitrates. The cabling mission systems thus a domain with a high degree of com- thus has to be newly installed. home automation plexity that is in touch with different POF (Polymer Optical Fibres) also areas of telecommunications. The imaginary scenario mentioned in offers the possibility of carrying the introduction requires access to many 100 Mbit/s over several tens of metres. appliances and consumer electronic 2 Home Network POF is younger than twisted pair, but equipment in the house, like the oven, may typically compete with twisted scenarios pair in the future. POF offers the lights, heating, air conditioning, TV, the video, security systems, which typically advantages of high flexibility, easy The Home Network (HN) does not require low bitrate. One could think, for handling and processing, EM immu- simply deal with a transmission system, this so-called smart house, of connecting nity, low cost components, high band- or a management system, as the last drop all these systems to a broadband trans- width, and low attenuation. of the telecommunication network covers mission network for all applications all OSI layers. This makes the subject Power lines is a very interesting inside the home like a twisted pair of quite complex. Moreover, it is located at medium for communication inside the optical fibre network. However, a set of the customer premises so that the owner- home since it is already deployed in existing technologies is already available ship issue, responsibility issue, and inter- every room, and sometimes in every to connect the different house equipment face definition between operator and corner of the room. However, this together: customer are problematic. X10 is currently the most widely deployed technology for home automation. It transmits over power- lines using the zero crossing of the AC sinusoid to transmit information bi- directionally. The bitrate is typically a few bits per second. CEBus (Consumer Electronic Bus) is Services - a communication standard for home Applications networks developed by the Electronic Industry Association (EIA) and the Terminal 3 IP/ATM Consumer Electronics Manufacturer Terminal 1 Association. The standard covers Terminal 2 communication via many media like twisted pair (TP) cable, coax cable, RF, Infrared, and AC powerline car- Fiber, Copper, Physical rier. The bitrate is typically a few Coax, Radio Layer kbit/s. LonWorks is a proprietary local operating network communication Figure 1 Representation of the different domains and problems touched by HN 160 Telektronikk 2/3.1999

162 technology introduced in 1990 by Ech- cial cabling where 2 pairs are used for 6.1 GX-FSAN elon Corp. LonWorks can typically transmission and 1 pair for powering. The Network Termination Home Net- work at 45 kbit/s over powerlines, The bandwidth available is 100, 200, work Group of GX-FSAN has produced twisted pair, coax cable, fibre optics, or 400 Mbit/s. Its flexibility makes it a few different reference configurations and wireless RF and infrared. a good candidate for computer and based on the general architecture. Figure telecommunication applications. EIB (European Installation Bus) is 2 represents the general architecture a transmission system optimised for IEEE802.11 is an Ethernet interface according to GX-FSAN for xDSL based each medium: twisted pair, powerline, specially developed for wireless appli- access. It is assumed that the cabling and radio frequency (RF). cations (both infrared and radio). The carrying the broadband signals and the available bandwidth is now 12 Mbit/s cabling carrying the narrowband signals and should be increased in the future to inside the home are separate, although 5 High bitrate trans- 10 Mbit/s. they can run in parallel. One option of mission systems IrDA is a standard developed by the the market deployment reference model is shown in Figure 3. Japanese manufacturer of computers, The candidates generally accepted for communication equipment, and semi- use in home environments are briefly listed below: conductors for short range half duplex 6.2 DAVIC bitrates up to 4 Mbit/s using infrared DAVIC has provided Home Networking ATM25 is a transmission system able wireless transmitters. specifications in its last 1.4 specification. to transmit 25 Mbit/s on twisted pair category 5 cables. ATM25 is a PTP Other wireless systems are being con- DAVICs Home Networking tools have system that is normally used in star sidered to increase the bandwidth to sev- been developed to support the following architectures. However, the use of eral tens of Mbit/s like UMTS, MBS, functional requirements: small switches allows tree-like archi- HIPERLAN, HRFWG, Bluetooth, COM- tectures. MEND projects. These systems are not 1 Secure tunnels (ie. authentication available at the moment but are planned and encryption), secure home owner ATM 50 is a transmission system able for 2000 and later. resources (eg. select video content, to transmit 50 Mbit/s over twisted pair financial data); (category UTP5) and POF. Ethernet 10 Mbit/s is for twisted pair 6 Architectures for the 2 Multiple consumer appliances per home (eg. VCR, STU, PC); cat. 3 or 5 (10 BaseT) as well as for home network fibre optic (10BaseF). 3 Multiple service providers (eg. HFC, FSAN (Full Service Access Network), FTTC); Ethernet 100 Mbit/s can also be used DAVIC (Digital AudioVisual Council), on twisted pair cat. 3 (100Base T4), 4 Inter-room and intra-room communi- VESA (Video Electronics Standards cat. 5 (100BaseTx), or optical fibre cations; Association), HAVi (Home Audio/Video (100Base Fx). Interoperability Architecture) are differ- 5 Capabilities for uncompressed or IEEE1394 (or Firewire) is an interface ent international groups that have their lightly compressed video transmis- originally made for consumer elec- view on the home architecture. The con- sions between consumer appliances; tronics (TV, VCR) and requires a spe- cept of a residential gateway is able to 6 Target at least 100 metre runs; cope with many different home networks and access network interfaces. 7 Target at least an aggregate band- width of 200 Mbit/s; 8 Target at least eight simultaneous isochronous streams (eg. MPEG2 TS, Broadband services lightly compressed video stream); STB Broadband Broadband NT Home Network STB Access network (ADSL, VDSL, FTTx) Splitter NB/BB Narrowband Narrowband NT Home Network Figure 2 General architecture according to Narrowband services GX-FSAN for xDSL based access Telektronikk 2/3.1999 161

163 POTS (FTTH) POTS (ADSL, VDSL, FTTx) B-NT B-UNI STB User Port ATMF25 STB ATMF25 Intra residence Mux T PC communication ATMF50 PC 10BaseT Splitter (only for ADSL, VDSL, FTTx) Coax Broadcast STB Combiner NT1 + NT2 Figure 3 One of the FSAN models: Modular B-NT, with integrated NT2 for Market Deployment 9 Target ease of use for average home Figure 4 provides a general overview of User Premises Interface (UPI) pro- owner; Home Network Architectures. The vides connectivity from the DAVIC DAVIC Home Network systems are Access Network to the Home Access 10 Target evolution from simple to com- functionally divided into Home Access Network (HAN) by connecting the A1 plex networks; Networks, and Home Local Area Net- to one or more A1* interfaces. The 11 Simple to install and maintain; works. These architectures represent the UPI may be passive or active. full range of architectures to facilitate 12 Must meet local EMC and be fit for Access Termination System (ATS) detailed specifications of home network purpose requirements; provides several functions. It may be devices, topologies and technologies. an end consumer device, such as a set- 13 Must meet local safety and regula- top box, that uses all S-flows provided tory requirements. This reference model introduces many by the Delivery or HAN system. It new different concepts like: may also be a gateway device between the Delivery or HAN system and the Home Local Area Network (HLN). Examples of IWS include copper to optical repeaters, bridges, and routers. HAN HLN Interworking System (IWS) provides a ETS Point-to-point similar functionality to the HLN as the ETS UPI does to the HAN. It may act in a ATS passive or active mode to translate IWS between physical and mid-layer proto- UPI ATS cols for ETS and ATS devices on the ETS HLN. Examples of IWS include cop- ATS ETSI per to optic repeaters, bridges, and ETS routers. End Termination System (ETS) de- ATS ETS vices represent end user equipment NT UPI that may be used for a DAVIC appli- IWS cation. Examples of ETS devices are A1* A20 A20* camcorders, PCs, VCRs, television, Shared Internet appliances, and security sys- Media tems. A1 Home Access Network (HAN) is an extension of the DAVIC Delivery Sys- HAN Home Access Network ATS Access Termination System tem access network to multiple devices HLN Home LAN ETS End Trmination System within the home. The User Premises UPI User Premises Interface IWS Interworking System Interface provides separation between the HAN media at the A1* reference point. Figure 4 DAVIC Home Network Overview 162 Telektronikk 2/3.1999

164 Home Local Area Network (HLN) An Access Device is a device that con- Since a goal of the HAVi Architecture functions much like a traditional LAN nects an external access network to the is to be future-proof, interoperability is environment, with the ATS devices home network. A POTS modem, an more than a common command set. It is acting as gateways to the DAVIC ser- ISDN adapter, a cable modem, Residen- a software architecture that allows new vice provider applications. The HLN tial Gateway and DBS decoder are all devices to be integrated into the home may also support separate applications examples of Access Devices. network and to offer their services in an with no dependencies to service pro- open and seamless manner. vider resources or content. An End Device is a digital device con- nected to a network whose purpose is to The HAVi Architecture provides: provide some utility (other than network A set of software elements with their 6.3 VESA (Video Electronics service) to the end user. Examples of protocols and APIs needed to achieve Standards Association) End Devices are printers, TVs, audio interoperability (Figure 7); speakers, security sensors. The VESA Home Network committee is Device abstraction and device control formed by hardware, software, PC, dis- A Network Device is a device whose models; play and component manufacturers, cable purpose is to provide network services and telephone companies, and service An addressing scheme and lookup ser- to End Devices. Examples of Network providers. It has several goals: vice for devices and their resources; Devices include repeaters, bridges, 1 Provide an interoperability specifica- routers, brouters, and network manage- An open execution environment supp- tion which will allow the transfer of ment stations or any device that serves orting visual presentation and control information from any device to any such a purpose, such as a PC. of devices, and providing runtime other device in the home; support for third party applications; 2 Allow interoperability between differ- 6.4 HAVi Communication mechanisms for ent home networks from low to high extending the environment dynami- The HAVi Architecture is defined by bandwidth; cally through plug-and-play capa- companies like Sony, Philips, Hitachi, bilities; 3 Provide a common interface on the Sharp, Matsushita, Thomson, Toshiba, home side for Access Devices, such and Grundig. The HAVi Architecture is A versioning mechanism that pre- as the Residential Gateway; intended for implementation on con- serves interoperability as the archi- sumer electronics (CE) devices and com- tecture evolves; 4 Be able to make a transition from ana- puting devices; it provides a set of ser- logue distribution to totally digital dis- Management of isochronous data vices which facilitate interoperability and tribution; streams. the development of distributed applica- 5 Provide directory services for devices tions on home networks. HAVi is in the home. intended for, but not restricted to, CE 6.5 Residential gateway devices supporting the IEEE 1394-1995 The Bellcore concept The VESA committee feels that there and IEC 61883 interface standards. will continue to be several types of net- The concept of residential gateway has, works in the home. Today many houses The HAVi Architecture is intended for according to Bellcore, the following key have a copper wire telephone network networks based on the IEEE 1394 stan- attributes: and a coaxial cable TV network. These dard. IEEE 1394 is a powerful technol- The ability to provide efficient inter- networks will stay in the home for a long ogy that meets many of the requirements facing of one or more access network time and the new digital networks will be of home networks (see Figure 6). put in place in addition to them. The VESA Home Network architecture (Fig- ure 5) has the benefit of allowing low bandwidth low cost devices to stay on their own network and not have to handle Figure 5 VESA Home high speed data and decode complex pro- Network Architecture End Access tocols. Device Device The Backbone network spans the whole house so that devices located anywhere Access Network Access Network in the house can communicate with each Device Device other. The backbone provides sufficient Component Network quality of service for the applications and End End devices that communicate over it. End Device Device Device The Component network enables devices connected to it to communicate with each Access Network Access Network other. Examples of important Compo- Device Device Component Network nents Networks are IEEE1394, Ethernet, Powerline CEBus, and RF Wireless LAN. Telektronikk 2/3.1999 163

165 Figure 6 Example of IEEE 1394 Network IEEE 1394 interfaces with multiple home net- ligence within the RG to enable features CENELEC concept works and devices; such as remote provisioning and mainte- The working group WG5, under the nance. The RG allows network techno- Low cost, perhaps using a modular, responsibility of the CENELEC TC205 logies to be reliably terminated, with incremental expansion strategy to Home and Building Electronic Systems well-defined network provisioning, mon- accommodate additional interface, started early 1998 a new work. The itoring and loop-back capabilities. Simi- networks and devices; objective of this is to specify a standard- larly, internal home networks can be ised interface between telecommunica- Power management, with power-down more reliably installed and maintained if tion networks and home networks to sat- or standby capabilities, and either they are not extensions of outside plant isfy the need of the service providers power back-up or passive coupling networks, such as is currently the case (functionalities) and of the customers features for some critical services. for cable modems. The RG provides an (cost and usability) in the area of the important middleware interworking flexi- home automation services. Liaisons have Figure 8 shows a Residential Gateway bility point, for example interworking been established with many other groups concept. between an access network delivering (ETSI, CEN, CENELEC, TIA, DAVIC, ATM cells, and separate home networks VESA, EIBA) in order to ensure a sys- There are many functions that could be using the TCP/IP protocol on 10Base-T tem compatible to all of these future integrated in the RG (residential gate- Ethernet, and IEEE 1394 network supp- applications. The WG5 aims to get a way), but the core required function may orting isochronous digital video. first set of documents end 1998. be protocol and format conversion, intel- Device Device Figure 7 Software Elements of the HAVi architecture Control Control Application Application Application Application Device Control Application Interoperability API Communication Media manager DCM Stream Mgr Messaging Event Mgr DCM Mgr Registry 1394 DCM DCM Platform Specific API Vendor Specific Platform 164 Telektronikk 2/3.1999

166 Conventional IEEE 1394 ~ ~ ~ POTS ~ "Long-range" IEEE 1394 Optical Fiber IEEE1494 ~ PLC ~ Ethernet UPs Figure 8 Residential Gateway concept PLC: Power Line Communication UPS: Uninterruptible Power Supply It seems that the generic interface will be Reference vanced Access Networks. Deliverable able to support several kinds of home bus 11: Broadband home Network for with different protocols as well as ISDN, 1 EURESCOM Project EU-P614. residential and small business. POTS, and GSM. The capacity in terms Implementation strategies for Ad- Heidelberg, November 1998. of Mbit/s of this bus, as well as its ability to handle ATM and IP traffic, are still not clear. 7 Challenges Markus Wyss (33) received an engineer Diploma degree from the Swiss Federal Institute of Technology, in 1989, This contribution illustrates the number after which he joined the Terrestrial Microwave Radio Group of possible interfaces, transmission sys- of Swisscom Research and Development. His main activities tems, and possible architectures that may concerned various studies in the field of terrestrial radio, eg. concerning compatibility, the development of specifications, be used for Home Network applications, field trials, development of planning tools as well as studies today available. Network Operators and simulations concerning ATM over radio relay. He is a should not ignore this status and leave member of ETSI-TM4 and the ETSI BRAN project. consumer electronics companies to this market. Network operators are active email: [email protected] partners in the deployment of end-to-end solutions for telecommunications. Their choices, their strategies, like for example to sell ATM to the desktop, may deeply Frdric Pythoud (31) received the Physicist Diploma from influence the choice of the architecture, the Swiss Federal Institute of Technology of Zrich in 1992. of the transmission system, and of the He went on to perform applied research at the Biomedical interfaces. A good strategy for Home Engineering Laboratory of the Swiss Federal Institute of Networks may also increase the demand Technology of Lausanne and got his Ph.D. in 1996. He for broadband and thus directly impact then joined the xDSL group of the Swisscom Research unit on the core network. The youth of the and took part in ADSL and VDSL standardisation in ETSI Home Network market shows the huge TM6. Since 1998 he has been project leader for the Home potential for new developments in the Networking activities. future that may make the imaginary story email: [email protected] of the introduction become real. Telektronikk 2/3.1999 165

167 Achieving global consensus on the strategic broadband access network: The Full Service Access Network initiative ALAN QUAYLE AND JEFF STERN The Full Service Access Network telcos, for example it has produced the 1 Introduction (FSAN) initiative represents over half worlds first asynchronous transfer of the worlds telephony lines. They mode passive optical network (ATM Imagine how much cheaper broadband have agreed upon a common broad- PON) specification which is being networks could be if their components band access system that can be applied adopted world-wide. This paper de- could be mass produced, say in the quan- to both the business and residential scribes the current activities within tities needed for tens of millions of market sectors. The FSAN initiative FSAN, the application of FSAN com- access lines, rather than todays typical has been responsible for harmonising ponents and where FSAN can go in few thousand line trials. This is the the broadband requirements of many the future. vision behind the Full Service Access Networks (FSAN) initiative; a three year old project, that involves nineteen of the leading telecommunications network operators, supported by many of the major telecommunications equipment manufacturers. The aim is to create a common requirements specification for an access system supporting a full range of narrowband and broadband services Switch PON Head Local Cabinet Kerb Home for a market that covers nearly four Node End Node Exchange hundred million lines of telephony. VB5 UNI The aim of this paper is to communicate PON ADSL the achievements of the FSAN initiative, ATM OLT ONU NTE FTTEx describe the common access system and explain the potential application areas of VDSL ATM OLT ONU NTE FTTCab this technology. The telcos currently involved in the initiative are: Bell VDSL FTTK/ ATM OLT ONU NTE FTTB Canada, Bell South, BT, Chungwa (Taiwan) Telecom, Deutsche Telekom, FTTB/ Dutch PTT, Telecom Eireann, France ATM OLT ONU FTTH NTE Telecom, GTE, Korea Telecom, NTT, SBC, SingTel, Swisscom, Telecom Italia, Figure 1 Common network elements Telefonica, Telia, Telstra, and USWest. Within FSAN it was recognised that the needs of individual telecommunication operators differ due to different regula- tory, business, and structural environ- ment in each country. But sufficient simi- Table 1 The telecommunications operators position in the emerging larities exist in requirements for future broadband environment access networks to enable significant benefits to be achieved through adopting Strengths Weaknesses a common requirements specification. The common system is based around a Existing ubiquitous infrastructure. Existing infrastructure with ongoing Organisation, eg. sales, marketing, depreciation charges necessitates broadband fibre feeder system, ATM maintenance, able to support new re-use. PON (Passive Optical Network), and a broadband services. xDSL (Digital Subscriber Line) system, shown in Figure 1. The exact DSL sys- Reputation. Capital constraints limit investment. tem used depends upon where the optical Excellent brand awareness. POTS is devaluing so need new revenue system is terminated, eg. in the local streams. exchange, cabinet, kerb or home. Hence this broadband access system can support Market will be strictly regulated. a range of access architectures; this flexi- bility is fundamental to the consensus Opportunities Threats achieved in FSAN. [1] describes the use New revenue streams teleworking, Uncertainty in demand and regulation. of Asymmetric Digital Subscriber Line Internet and multimedia. (ADSL), one of the topology options shown in Figure 1. Lower cost of providing POTS. Competitors already providing broadband. Technology enabling broadband Low service take-up. multimedia delivery to the home. High churn. 166 Telektronikk 2/3.1999

168 2 The need for broadband Senior Management Figure 2 It is anticipated that FSAN will provide a Group Structure broadband solution to both the residential of FSAN and small/medium enterprise markets. Table 1 captures the incumbent opera- tors position in the emerging broadband environment. The existing revenue Deployment streams are under increasing pressure, Group and the nature of network usage is also changing, n.b. the rapid growth of the Internet. FSAN provides a broadband access solution that builds upon the oper- Optical Access Infrastructure ators strengths through using the exist- Network Workgroups Workgroups ing infrastructure, this also tackles some of the weaknesses by minimising capital investment. FSAN is flexible in its topol- Service Very high speed ogy through a range of re-use options, Capabilities and Digital subscriber thus managing the threats. FSAN is en- Performance Line abling operators to jointly take advantage of the broadband opportunity through Operations, minimising the risks associated with the Administration and investment. Maintenance 3 FSAN structure The FSAN initiative has been through four phases, each lasting roughly one year, the results being presented to the industry at public conferences, they are also available on the world-wide web [2]. The objectives of FSAN are to: 5 FSAN workgroups Initially the initiative focused on identi- Complete the outstanding technical fying how cost reduction could be issues by the end of 1998, and provide 5.1 Service Capabilities and achieved. Several important system a milestone upon which the manufac- Performance Workgroup components were identified, and with turers can focus their development pro- The mandate of the Service Capabilities a common specification it was believed grammes, particularly for the APON and Performance (SCP) workgroup is: that cost reduction would be possible. system; The global access product achieves cost To define and specify the capabilities reduction through two effects; competi- Sell the business opportunities from required of FSAN access network to tion in supply which reduces margins, using FSAN components in extending support the required services; and each supplier produces a greater vol- broadband to the wider market; To determine dimensioning/sensitivity ume of the common system so the learn- Share trial and deployment experi- and performance requirements of a ing curve effect has a greater impact on ences to encourage the adoption of FSAN; production costs. FSAN systems. And because FSAN is access-centric, As the initiative has progressed the focus the focus is on bearer services, ie. has moved away from bit level specifica- 4 The Deployment Group transport-related. tions to management/control plane issues and service specific aspects, eg. how fast The Deployment Group is formed from The current work areas of the group are: channel changing is done for switched the merged FTTCab and FTTH chapters. Switched Digital Broadcast (SDB), digital broadcast TV, see section 5.1. Its aims are to: which covers the following issues: Identify and develop clear business Figure 2 shows the relationship between fast channel changing (zapping) pro- opportunities for FSAN deployment the different groups within FSAN. The tocol; and sell these opportunities to the workgroups are focused on adding tech- industry; replication in the access to conserve nical content to the specification, their capacity; work will be discussed in later sections. Assemble complete packages of speci- The deployment group is developing a fications for each deployment scenario customers having access to multiple clear business opportunity statement for driven by the early mover telcos. broadcast content providers simul- FSAN deployment and communicating taneously. this to the industry. It is also responsible Real-time control of access network for ensuring any specification gaps are resources, which covers the following completed by workgroups. issues: Telektronikk 2/3.1999 167

169 service info Figure 3 Architecture for zapping ment, they can at least ensure that suit- remains in SN able customer systems are specified. Three aspects of the customer environ- IL ment need be considered: The home network; zapping The customer interface; STB Switch CPE functionality. R VBS BCC Access Resource Home network Control The basic problem that the home network needs to address is how to distribute the R - Replication (Resource) Access Network services throughout the IL - Intelligence Layer home, and provide multiple simultaneous attachment to different services, while retaining the quality of each service. The solution to this problem involves issues that range from architecture to infrastruc- ture. The basic FSAN HN architecture is shown in Figure 4; two key features are full signalling (Q.2931) versus independent of the access network. Only shown. The Access Network is de- possible new streamlined low-level the service provider and STB need to be coupled from the HN by means of an protocols; aware of the protocol. Through the bearer active NT; the HN is essentially a LAN channel control (BCC) protocol the rather than simply being an extension of resource management of network access receives the instruction for the the Access network, however, it must be termination is a key issue. channel change. a LAN that can support the required ser- Integrated POTS/voice, multi-line vice mix. This decoupling means that voice, for example: different and appropriate transmission 5.1.1 Home network aspects systems can be used in the Access and lifeline, toll quality; The work of FSAN is very much service Home environments, and that the same non-lifeline, lesser quality; driven. Since services are by definition HN can be used with a variety of Access end to end, then FSAN must ensure that Network types. The second main feature other classifications based on avail- it considers a complete end to end system is the separation of the broadband home ability. description, which includes the customer network from the existing narrowband Handling of IP capabilities and Ether- environment. This is something of a one. net/10BaseT user interfaces, with a change from the traditional telco view, focus on: where responsibility (and possibly inter- After consideration of all the existing and est) ended at the NT. Although there may developing technologies, the following multi-service, multi-terminal still be a limit to which operators can dic- solution was agreed: support; tate what occurs in the customer environ- concentrating on determining opera- tor requirements for input into stan- dards bodies. The Home Network group has been merged into this group and is de- scribed in the following sub-section. Broadband Taking the switched digital broadcast Services B-UN issue as an example, the group has de- cided not to provide a special solution for zapping between TV channels provided BroadBand BroadBand over the access system, but follow the Home Network NT VB5 architecture, as is used for all other Access Network services over the access system. Figure 3 (ADSL, VDSL, FTTx) shows the architecture for zapping. A Splitter channel change request, zapping, comes NB/BB from the STB (Set Top Box). It is not NarrowBand BroadBand terminated in the access, for many opera- Home Network NT tional and commercial reasons, eg. this avoids the need for the access network to contain customers personal information. Narrowband Narrowband Figure 4 It enables the zapping protocol to be Services Interface Home network architecture 168 Telektronikk 2/3.1999

170 PC A point to point star, architecture; Newly installed Category 5 twisted Access network Example user pair cable; connection devices NT2 NT1 50 m reach between nodes; Basic network functionality incorpo- STB rated into the NT; Physical Enhanced functionality provided by infrastructure means of a separate hub. The basic functional elements of the Figure 5 Basic functional elements home network are shown in Figure 5. The NT1 terminates the Access line system (xDSL or fibre) and may contain other functionality, in particular basic multiplexing and demultiplexing. Mul- PHY must be supported in FSAN sys- in the supply of CPE either through the tiple (customer-side) interfaces may be tems, but it will automatically inter- normal retail chain or via direct procure- presented by the NT1. The NT1 func- operate with devices that conform to the ment. The term CPE embraces PCs with tionality will be physically located in the original ATM25 specification, thereby Network Interface Cards (NICs), Net- B-NT. Full switching requires the addi- providing backward compatibility with work Computers (NCs) and Set Top tion of an NT2. This may be located in some early, existing deployments. Boxes (STBs). The Consumer Electron- a separate physical device, or could be ics industry will be invited to comment incorporated into the B-NT; however However for some specific services, and input, as necessary, to these core the degree to which this incorporation is there may be a case for terminating ATM requirements since FSANs goal is to allowed may be the subject of local regu- before the CPE (eg. in the B-NT or a sep- influence their roadmap for future prod- lation. A separate NT2 would generally arate terminal adapter) and providing ucts. Key CPE features already identified be regarded as CPE. The CPE end de- interworking to another network type. which need to be addressed include: vices can connect directly to the NT1, The only example of this alternative User friendly CPE which is truly Plug NT2 or both, depending on the imple- approach considered by FSAN is Ether- and Play; mentation. net interworking. A key issue with Ether- net presentation is where the (layer 2) User to network flows terminating in Ethernet protocol is terminated, ie. where the Access Network; The customer interface the initial routing functionality is carried User to user flows terminating beyond The general framework of FSAN is out. The preferred FSAN approach is to the Access Network; based on the fact that the transfer mode do this in the home for reasons of scal- of the access network needs to be based ability and security; this functionality Traffic shaping; on ATM to provide the controlled, full could either be incorporated in the NT or APIs / middleware; service mix. In order to retain this full as a separate, conventional small router. service capability to the user, ATM needs 10BaseT was not designed for the resi- Standardised software download; to be continued right through the home dential environment and does not have Configuration management; network to the CPE. This also avoids a desirable EMC properties. The FSAN complex interworking function in the preferred Ethernet interface type is there- Security. NT. The problem is that the home en- fore 100BaseTX. However in order to vironment is rather different from the provide compatibility with 10BaseT PC For residential services, the intention business environment with regard to the interface cards, where 100BaseTX is is that the core CPE requirements de- required transmission and EMC perform- implemented in the B-NT it must support veloped by FSAN may be added to next ance. However the ATM Forum have dual mode 10/100 Ethernet operation. generation consumer electronic equip- developed a PHY standard specifically ment such as digital satellite receivers, designed for the residential environment. digital terrestrial TV receivers, DVD CPE functionality This is based on (and backward com- players and games consoles. This would patible with) the original ATM25, but A range of appropriate broadband CPE then enable the consumer to access mul- provides: must be available, together with attrac- tiple services from a single STB. tive service offerings, to ensure that a A higher line rate mode (51.2 Mbit/s); new access network is fully utilised; lack Improved EMC properties; of CPE and services inhibited the early 5.2 Operations, Administration take-up of ISDN. An objective of the and Maintenance Physical layer OAM. FSAN members is to gain consensus The Operations, Administration and on the core requirements for Customer Maintenance (OAM) group was set up This is the chosen FSAN customer inter- Premises Equipment (CPE) and thus to look at the operational aspects of the face. It provides sufficient bandwidth to ensure compatibility and interoperability FSAN and to agree on a common set of support future HDTV services as well as with FSAN networks and services. This management functions. A framework for allowing for more extensive customer would also enable the member compa- developing a common set of require- networks. The 51 Mbit/s mode of this nies to leverage commercial advantage Telektronikk 2/3.1999 169

171 SM SM SM OSF OSF OSF if7 if8 (IF7:X) (IF7:Q3/X) SML if0(IF0:Q3) NM NM OSF OSF if1(IF1:Q3) NML EM EM EM EM OSF OSF OSF OSF if2(IF2:SNMP/Q3) DCN EML if2(IF2:SNMP/Q3) if4 if5 NEL (IF4) (IF5) Extended NE BN OLT CDN ONU NT Feeder if3 if3 if4 if6 (IF3:SNI) (IF3:SNI) (IF4) if4(IF4) (IF6:UNI) ONT Domain 2 Domain 1 Network Element Figure 6 FSAN target management architecture ments was quickly established which Another key agreement of the group was management architecture, interfaces included the following steps: to re-use as many existing standards as and functions; possible in order to make rapid progress. Understand the operational and service Platform requirements When it was clear that an architecture requirements (what are we trying to was required to progress discussions on scalability, throughput, performance, deliver); the requirements, the basic architecture operating system, security, avail- Derive the management requirements described by the Telecommunication ability; and develop a reference architecture Management Network (TMN) was Data Network Communications (how it is to be managed); adopted and extended to develop an FSAN target architecture as shown in types of data networks to be Develop specifications to ensure stan- Figure 6. supported; dardisation in the supplier industry. Test Equipment Over the first three phases of the FSAN A key decision of the group was to look initiative the group has defined require- automated testing, reduction of at the problem from an end to end per- ments on the following aspects of opera- reliance on network testing; spective which led to the group adopting tions and management: a set of high level processes from which VSDL it was possible to agree on a set of com- High level processes; operational requirements of VDSL mon operations functions. A number of Operational requirements of the equip- link (management of physical layer); high level processes are described in the ment FSAN OAM requirements document [3] ATM layer requirements which include Service Provision, Net- modular design, simple visual indi- OAM flows; work Repair, and Planning and Engineer- cations, self configuration, accurate ing. Each of these contain elements of fault diagnosis; Information Model customer handling, work management, Management requirements based on ATM Forum M4 and installation, billing and so on. ITU-T models. 170 Telektronikk 2/3.1999

172 Most of this work has been completed in ence configuration of ITU-T Rec. This was left open to enable vendors to nine meetings since work began in late G.PONA (G.982) [10]. The system has innovate. The OLT is the controlling 1995 and is entirely due to the commit- been designed to support all configura- device. The ONU responds to the OLTs ment and support from both the operators tions shown in Figure 1. commands, hence the PON interface can and suppliers involved. still be manufacturer independent. There are two PON options, a sym- In the next phase the OAM group will be metric 155 Mbit/s, and an asymmetric progressing the following areas of work: 622 Mbit/s down to the customer and 5.4 Very high rate Digital 155 Mbit/s from the customer. The PON Subscriber Line Define information model details; also has a minislot capability to cost VDSL technology can deliver data at Collaborate with the Optical Access effectively support voice, STM (Syn- multi-Mbits/s over the unscreened, Network (OAN)-WG on management chronous Transfer Mode) services or twisted telephone wires originally requirements for the management dynamic capacity over the PON. The intended for bandwidths of between channel on OLT to ONU/ONT inter- PONs dimensions are a 20 km reach, 300 Hz and 3.4 kHz. This is due to face; and a maximum optical split of 32. The remarkable advances in digital signal system can operate on either 1 fibre using processing technology which allow the Support programme to standardise WDM, or 2 fibres. implementation of sophisticated digital FSAN requirements; and modulation and equalisation schemes. Explore common management require- Figure 7 shows the frame structure of the In fact, the challenges are now mainly of ments for higher layer functions. The APON. It is ATM cell based. A down- an analogue nature: channel attenuation, current sets of requirements cover the stream frame is made of 56 cells, this noise and spectrum management. More NE and EM layers of the Telecommu- includes 2 PLOAM (Physical Layer details on the use of DSL in the access nication Management Network (TMN). Operations, Administration and Mainte- network can be found in [11]. Further work is needed to extend the nance) cells. The PLOAM cells contain FSAN management requirements to the grants for the upstream slots, and It is important to note that VDSL is the Network Management (NM) and messaging for physical layer functions, required to operate on underground as Service Management (SM) layers. eg. to instruct an ONU to start ranging. well as overhead distribution cabling. The upstream cell has a 3 byte header, to This imposes some difficult requirements The FSAN requirements have been dis- cope with the multiple access nature of in order to control unwanted RF emis- cussed at a meeting of the appropriate the upstream transmission path. Because sions, particularly for overhead distribu- ITU-T work group in March 1998 and of the overhead, the upstream frame is tion cabling. VDSL will also be required a draft ITU-T document containing the made up of 53 upstream slots. The mech- to operate in the presence of Bridged FSAN requirements has been produced. anism for allocating the slots, the media Taps principally a problem for North The OAM group is currently working on access control protocol, is not defined. American telcos. the definition of an information model for interface IF1 (see Figure 4) which will be available in July 1998. In addition it is proposed to input both the require- ments and information model in to ETSI in September 1998. The aim is to have agreed recommendations by end 1999 following the normal procedures of the ITU-T and ETSI. The group is also look- Table 2 PON layers and functions ing at adoption of the FSAN require- ments as an ANSI standard but discus- Circuit layer Translation and maintenance sions are at an early stage. Path layer I.732 It has only been possible to give a high Transmission Transmission Adaptation I.732 level view of the work of the OAM group in this section. The reader is re- media convergence PON Ranging ferred to the OAM requirements docu- layer layer transmission Cell slot allocation ment [3], including past papers [4, 5, 6] for more details of the areas described Capacity allocation above. Privacy & security Frame alignment 5.3 Optical access network Burst synchronisation This group has created the worlds first ATM PON specification. It has been pre- Bit/byte synchronisation sented to and accepted by the ATMF [7], Physical E/O adaptation ITU G.PONB [8] and ETSI [9]. The specification focuses on physical layer media WDM and transmission convergence layer, layer Fibre connection shown in Table 2, following the refer- Telektronikk 2/3.1999 171

173 Downstream Frame 56 cells of 53 bytes 5.4.1 General description of VDSL PLOAM1 Cell 1 Cell 2 Cell 27 PLOAM2 Cell 28 Cell 54 Figure 8 shows a simple architectural model of VDSL. 1 PLOAM cells contain 53 grants Since the broadband and narrowband services should be able to share the same Cell 1 Cell 2 Cell 3 Cell 53 metallic distribution cable, the broadband services are introduced at frequencies 53 cells of 56 bytes well above POTS or ISDN-BRA. Figure Upstream 9 shows the VDSL signal placed well Frame above the band occupied by POTS or ISDN-BRA. 3 byte overhead, contents are programmed by the OLT Figure 7 5.4.2 The VDSL standardisation APON ATM cell process frame The introduction of VDSL transmission structure PLOAM cell, 53 BYTES systems is dependent on harmonisation of Network Operators requirements to produce a large common market. The remit of the FSAN VDSL Working Group has therefore been to identify the key requirements which are common to all telcos, and downstream these require- ments to the standards fora, notably the Narrow- and European Telecommunications Standards Broadband UNIs Institute (ETSI), the American National Standards Institute (ANSI), and more NB Existing copper re-used when recently, the VDSL Study Group within fibre supports BB services only NT the ADSL Forum. POTS or Existing OSDN-BRA TE VDSL Splitter Exchange 5.4.3 The FSAN VDSL Working filter ONU Node Group Existing copper channel Fibre fed BB or The FSAN VDSL Working Group has NB+BB service made a significant contribution to the NB NT VDSL Standards process by achieving Broadband Note POTS/ISDN NB splitter telco consensus on the key systems TE is included in the ONU requirements for VDSL. This work was Figure 8 Splitter may be published in some detail at a workshop VDSL colocated in BB NT in Atlanta in March 1997 [12]. The work concept has now been successfully downstreamed such that the current version of the ETSI Draft Technical Specification for VDSL [13] should be viewed as the primary reference source for FSAN VDSL re- quirements at this time. The ANSI standards process for VDSL is ~300 kHz 10-30 MHz less advanced than ETSI but is catching up. ANSI is adopting much of the ETSI work, with the FSAN requirements being actively progressed by the North Ameri- Signal level ISDL-BRA can members of FSAN, notably GTE and Bell Canada. ----- VDSL During 199798, the group has been addressing the complex technical issues Figure 9 of noise model and spectral compatibility Frequency for VDSL. Spectral compatibility is a separation fundamental issue for telcos to ensure of VDSL POTS that different xDSL systems can co-exist in the same cable infrastructure, maxi- Frequency 172 Telektronikk 2/3.1999

174 mising their performance, while mini- 5.4.5 VDSL spectral compatibility Telcos ultimately require that systems are mising their impact on neighbouring sys- not just spectrally compatible, but that a More work is now required by the FSAN tems due to crosstalk. Because spectral modem from one vendor will interwork telcos to define the spectral bounds and compatibility is a systems issue which is with that from another vendor. This is duplexing scheme for VDSL before ven- network dependent and which transcends primarily the task of international stan- dors have sufficient information to de- the boundaries between xDSL systems, dards. velop spectrally compatible systems. the group has remained telco-only to date. Although spectral masks for VDSL have 5.4.7 Alleviating the standards 5.4.4 VDSL noise model been agreed in ETSI, these only detail burden the major features such as notching and A major achievement of the group has The world spotlight is on xDSL as the maximum power. To ensure that systems been the definition of a noise model for technology that will provide universal are developed which are spectrally com- VDSL that was derived during a telco- broadband access as we enter the next patible with installed xDSL systems such only meeting in Paris in November 1997. millennium. This has significantly in- as ADSL, telcos must provide more The noise model is required to enable creased the pressure on xDSL technical detail to vendors on the spectral bounds telcos to benchmark vendor implementa- experts to attend meetings all over the for upstream and downstream VDSL tions and to verify that systems will work world. The FSAN initiative provides transmission, and evaluate the systems in real networks. an opportunity to alleviate some of this implications of the different duplexing pressure (and save costs) by sharing tech- techniques. At the Paris meeting, different telco nical perspectives, and co-ordinating xDSL deployment scenarios and cable attendance at meetings. This is an extremely complex problem topologies were modelled in real time to to resolve because VDSL capacity is a identify a small set of noise profiles that Following the FSAN summit in Venice complex function of topology, duplexing could be applied to all telcos. A powerful in March, 1998, the remit of the VDSL scheme and crosstalk. The final choice enabler for consensus at the meeting was Working Group has been extended to will depend on which engineering com- the availability of two independent laptop encompass general liaison on all xDSL promises telcos are willing to adopt, and computer models which could be used to issues of common interest to the telcos. this depends on the ability of the telcos check results in real time. It was found In particular, to maintain vigilance, and to model their scenarios for VDSL from modelling that all basic telco sce- cooperate to ensure that xDSL systems upstream and downstream capacity. narios could be fitted to only four noise are spectrally compatible with each profiles; a result that was not obvious other, and through technical dialogue, To resolve this problem, a common for- beforehand. These profiles have been resolve the difficult engineering issues mula and methodology for calculating accepted by ETSI [14] and have been presented by for example the Splitterless VDSL capacity in different telco scenar- provisionally adopted for inclusion in ADSL concept. ios was agreed at a meeting in Bern in the ANSI standard. February 1998. The models have been This is believed to be the first time that a verified using a single common scenario 5.5 Infrastructure and will be used by individual telcos to consensus noise model has been defined The implementation of VDSL Fibre-to- evaluate their own xDSL scenarios in so early in the standards lifecycle for an the-Cabinet has traditionally taken the order to develop a group insight into this xDSL system; demonstrating the value form shown in Figure 10. The VDSL complex problem. With group insight, it of this method of reaching consensus. electronics are housed in an above is hoped that a consensus solution will ground cabinet, and positioned close to be adopted at a meeting to be held in A key insight from the noise model work an existing copper Primary Cross-Con- London in May, 1998. was confirmation, previously highlighted nection Point (PCCP) in the network. by Nortel [14], that the spectral mask for The VDSL Access Point (VAP) is linked ADSL required modification to minimise 5.4.6 Next steps to the exchange via fibre. Copper Tie impact on VDSL capacity. This has been cables are used to link the VAP to the If, as is hoped, the group achieves con- progressed in ANSI by GTE on behalf of PCCP in order that broadband service sensus on the spectral bounds and du- the FSAN telcos. It was also noticed that can be overlaid on the existing D-Side plexing scheme for VDSL, it will have existing standards relating to the spectral copper network. achieved its primary aim of defining the masks for ISDN-BRA were not adequate spectral parameters and systems require- to prevent new systems polluting the Placing active electronics in a hostile ments for VDSL. Vendors will then have VDSL spectrum above 1 MHz. Fortun- external environment presents major enough information to confidently invest ately, measurements on installed ISDN- design challenges. The combined re- in development of FSAN compliant BRA systems (both 2B1Q and 4B3T) quirements for line capacity and the VDSL equipment. If consensus is not have confirmed that existing systems are associated power dissipation, plus the achieved, a degree of additional flexibil- VDSL friendly. But the ISDN standard is auxiliary facilities for copper, fibre and ity will be required in vendor equipment currently under review in ETSI, and vigi- power can result in a physically large to guarantee that a particular VDSL lance needs to be maintained to prevent VAP cabinet with a complex and expen- implementation will operate reliably any relaxation of the mask, which would sive cooling system. These issues have in all telco networks world-wide the open the door to spectral pollution from been the focus of in-depth discussion for overall aim of the FSAN initiative. future variants of ISDN-BRA systems. the FSAN Infrastructure Group, where telcos and suppliers are endeavouring to reach a common consensus on the design Telektronikk 2/3.1999 173

175 Existing Copper VDSL Access Point Flexibility Cabinet (VAP) Cabinet requirements for an FTTCab solution. These discussions are progressing well, but have clearly highlighted that the vari- ous telcos have significantly different requirements and constraints on the Mains Power design and use of active street furniture. Existing E-Side 230/240 Vac Generic requirements for an above Copper Tie Cables & D-Side Cables ground solution are well advanced and are expected to be issued shortly. How- ever, restrictions on certain telcos not to Fibre from Duct Network use physically large, or in some cases Exchange Footway Box Footway Box any, above ground cabinets have resulted in FSAN also proposing an underground Figure 10 Traditional above ground VAP FTTCab scenario solution. Figure 11 shows the underground con- cept that employs 8 or 16 line, sealed- for-life modular VAP enclosures which can be located in existing underground footway boxes. Each VAP module should provide suitable connections for primary power, fibre, copper tie-cables Street Pillar and and the option for battery back-up. The Power Supply 8 or 16 Line advantages of an underground system VAP Module have been identified as: Fibre Less hostile environment for tempera- Joint ture extremes. Improved electronics Tie-Cable and battery reliability; to Existing Copper Vabinet Less prone to EMC radiation damage 230 Vac and emissions; Power More aesthetically pleasing; 48 Vdc Fibre Less vulnerable to accidental or deliberate damage; JUF6 Footway box 1310 x 610 x 665 mm Savings in enclosure costs compared to cabinets. The underground system is designed Figure 11 Underground modular VAP proposal to be highly modular allowing extra modules to be added to satisfy growing demand, as demonstrated in Figure 12. The underground system is intended to be highly flexible to satisfy variations in telco requirements, and has the potential to be significantly less expensive than a cabinet solution. The proposal is attract- ing major interest from the telcos and suppliers within FSAN who are currently discussing the development issues and potential applications. 6 Next steps within FSAN The latest results in FSAN were pre- sented at Globecom98. FSAN is a tem- porary body. After the initial trials are complete, and the FSAN specification is updated to reflect the learning of the trials, and the FSAN specification [2] reflects the systems for commercial deployment, FSANs job is complete. Figure 12 Growth in demand 174 Telektronikk 2/3.1999

176 The systems are available to all, because 9 References 7 ATM Forum. Residential Broadband the specification is public, hence Architectural Framework. 1998. volumes are up and prices are down. This 1 Dickie, A, MacKensie, J. Midband : (STR-RBB-001.00.) is the FSAN vision. A solution for high speed Internet access. BT Technol. J., 16 (4), 4857, 8 ITU. ATM Passive Optical Network 1998. Specification. Geneva, 1998. (ITU 7 Conclusions G.983.) 2 The keys to FSANs success are a poli- access. 9 ETSI. Optical access networks tical will to facilitate the creation of a (OANs) for evolving services; ATM common access system specification 3 Mistry, R et al. Operations, adminis- passive optical networks (PONs) and to enable vast economies of scale, and tration and maintenance require- the transport of ATM over digital a highly flexible technology, PON and ments for a full services access net- subscriber lines. 1998. (ETSI xDSL, that can meet the disparate re- work, Issue 1. March 1998. http:// DTS/TM 03024.) quirements of the telecommunication access. operators involved in the initiative. 10 ITU. Narrowband Passive Optical 4 Mistry, R et al. Full Services Access Network Specification. Geneva, 1997. The FSAN deployment group is building Networks (FSAN) : operations, (ITU-T Rec. G.PONA (G.982).) a business opportunity statement to sell administration and maintenance. In: the benefits of FSAN systems to the Proceedings of the Full Services 11 Foster, K et al. Realising the poten- commercial people in the industry, and Access Networks Conference, Lon- tial of access networks using DSL. are accelerating the availability of FSAN don, June 1996. BT Technol. J., 16 (4), 3447, 1998. systems, through a co-ordinated approach to testing and trials. 5 Tofanelli, A et al. Management 12 Foster, K T et al. FSAN : VDSL Requirements for full services access Copper Transport System. In: IEEE The SCP groups mandate is to identify networks. In: Proceedings of IEEE VIII Workshop on Optical/Hybrid capabilities in the FSAN access network VIII International Workshop on Opti- Access Networks. Atlanta, USA, Mar to support a full set of services. cal/Hybrid Access Network, Special 1997. Section on FSAN, Atlanta, March The FSAN VDSL Working Group has 1997. 13 ETSI. Part-1 ETSI Draft Technical established a powerful discussion and Specification for VDSL. (ETSI modelling methodology for resolving 6 Mistry, R et al. Management require- DTS/TM-06003-1 V0.0.7). Jan 1998. the complex problem of xDSL spectral ments for a full services access net- compatibility, and rapidly achieving work. In: Proceedings of the Third 14 Humphrey, L D. On selection of the consensus through mutual insight. This Workshop on FSAN, Venice, March ADSL Issue 2 Downstream PSD forum is unique in the world in that it 1998. Mask. Northern Telecom Inc., 1998. contains xDSL technical experts from (ANSI T1E1.4/97-351.) the core development teams of the major telcos, working harmoniously together to resolve telco-specific transmission prob- lems, thereby accelerating standards through advance consensus. Alan Quayle (32) has two Masters degrees in electroinc systems engineering and business administration. Since The OAN group has created the worlds joining BT he has worked on a wide range of projects, first ATM PON specification. It has been from optical system design to developing VOD systems. presented to and accepted by the ATMF, Currently he is leading work identifying broadband opp- ITU and ETSI. The system has been ortunities and the access systems to exploit them. Alan designed to support the requirements of Quayle represents BT in the ATM Forum and Full Service Access Network initiative. He is a member of the IEE and all telcos. IEEE. FSAN has specified the worlds first email: [email protected] truly global broadband access system through the dedication and expertise of the people involved in the initiative. The challenge is now on how to realise Jeff Stern (54) is Development Manager, Fibre and Broad- the benefits of the consensus achieved band Networks at BT Labs. His current responsibilities within FSAN. include the development of long haul submarine, long haul terrestrial and local access fibre systems and the co-ordi- nation of development activities aimed at achieving a 8 Acknowledgements future broadband network infrastructure. Jeff Stern has played a leading role in the FSAN initiative from its outset We would like to thank the many mem- and has been Chairman of the FTTCab Chapter. bers of FSAN and colleagues within BT email: [email protected] for their help and support in creating this paper. Telektronikk 2/3.1999 175

177 Towards broadband access in Europe the view from EURESCOM UMBERTO FERRERO The access network is the portion of 1 Introduction need to understand and cope with a num- the telecommunication infrastructure ber of architectures and systems with a linking every single customer to the The Project started in late 1996 and single, comprehensive overall network closest local exchange: Traditionally involves over 50 experts from 12 Euro- perspective to assure the effective ex- dimensioned to deliver telephony, it pean telecommunication operators; the ploitation of the heavy investment now calls for a thorough upgrade to objectives are illustrated in [4]. The pro- involved. Second, the access network accommodate new midband and ject completed its activities in November evolution is moving towards real deploy- broadband services, exploiting the 1998, releasing the 14 Deliverables listed ment (as witnessed by the number of extensive range of available techno- in Table 1. market trial and roll-out programs logies. EURESCOM1) P614 Imple- announced in most European countries), mentation strategies for advanced Broadband access network deployment is and co-operative projects have to focus access networks represents the third strongly influenced by capital investment on practical issues, such as specifica- generation of access network related and regulatory boundaries, with technical tions, outside plant technologies and EURESCOM Projects, after P306 matters in the background, often under- interoperability of new and old techno- Access network evolution and prepa- estimated. logies. Third, the monitoring and contri- ration for implementation [1, 2] and bution to standardisation bodies, together P413 Optical Networking [3]. P614 focused on three main ideas devel- with the techno-economic appraisal of oped during the previous projects. First, technical implementation need to be The paper focuses on four main areas: the future access network will feature a carried out on an ongoing basis by broadband radio systems, broadband number of alternative implementations EURESCOM members, to quickly and fixed systems, technologies and instal- and several architectures, and will extend effectively react to the changing access lation techniques, and techno-eco- its domain towards longer reach allowing network scenario. nomics. for node consolidation: European PNOs The European operators declared strate- gies tend to swing between the idea of doing nothing or everything as far as broadband access network is concerned. P614 results recommend what to do (something), when to do it (with specific phasing, following the maturity of differ- Table 1 EURESCOM P614 Deliverables ent technologies), and where (in areas of well defined technical and service Title Planned issue date characteristics). The uncertainty becomes even greater with the increasing competi- D1 An analysis of the relative benefits of proposed SNI standards April 1997 tive pressure. D2 Target B-ISDN access network architectures December 1997 P614 tries to spot, and tackle, some areas that have been insufficiently taken into D3 Techno-economic analysis of major factors of B-ISDN/ATM September 1997 account so far, or at least little discussed, upgrades increasing the awareness of their poten- tial and giving answers to outstanding, D4 Opportunities for broadband radio technologies in the access January 1998 much discussed questions, or to counter- network act the easy enthusiasm on technical capabilities and trends. D5 Optical technologies for advanced access networks: early results December 1997 D6 Evolution paths towards target B-ISDN access networks July 1998 D7 Optical technologies for advanced access networks: final results September 1998 D8 Elaboration of common FTTH guidelines July 1998 D9 Contributing to standardisation bodies and other fora April 1998 D10 Techno-economic evaluation of B-ISDN access networks September 1998 architectures, scenarios and business cases 1) EURESCOM (European institute for research and strategic studies in D11 Evaluation of Broadband Home Networks for residential September 1998 telecommunications) was founded in and small business 1991 by European telcos (fixed net- work operators), as an instrument to D12 FTTH: Definition of the suitable powering architectures February 1998 perform collaborative precompetitive research and development. In Decem- D13 FTTH: Definition of access network quality and cost February 1998 ber 1997 the Institute had 23 active D14 FTTH: Definition of related service offer strategies February 1998 Shareholders from 21 European countries. 176 Telektronikk 2/3.1999

178 2 Broadband radio in geostationary orbit (GSO) which pre- of spectrum available. The channel vents them from offering a full service capacity is comparable with the DVB-T access: opportunities due to the inherent transmission delays technology and the range is several tens and limitations and consequently only limited services of kilometres. for residential users are offered. To Broadband radio technologies are be- achieve full service access, satellite tech- As for satellite broadcasting, terrestrial coming more and more close to real field nologies must migrate to low-earth orbit wireless networks redistributing satellite deployment, and are offering very pro- (LEO) which is beyond the short-term TV services have been in operation for mising applications for both incumbent perspective. some time, known as multipoint video operators and new competitors [5]. The distribution systems (MVDS). However, potentially large cost savings stimulate a Stratospheric fixed platforms consist of the MVDS systems will most likely be big hype of interest, boosted by an one or more high altitude (20 to 30 km surpassed by the local multipoint distri- aggressive marketing campaign. P614 kilometres) platforms located at fixed bution system (LMDS) technologies reviews the radio access network archi- points in the stratosphere, serving a which offer true multiservice broadband tectures and technologies giving access limited geographic area compared to capabilities. They are offering asym- to residential and business customers; the satellite technologies. HALO and metrical services with a capacity of up to a Request For Information was issued. Skystation are two such systems, with 3 Mbit/s uplink and 52 Mbit/s downlink P614 also reviewed ongoing research and operation expected to commence in 2000 with a typical range of 1 to 5 km. Gener- standardisation activities and estimated at the earliest. ally, they are related to the frequency the type of technologies available in the bands above 10 GHz. More specific, the long term. Terrestrial technologies are by far the 28 GHz and 40 GHz band are the pro- most interesting alternative among the bable candidates in most countries. No P614 highlights the truth on time for wireless technologies. They offer a widely supported CAI is currently avail- available systems and technical capabi- variety of technologies, which may be able but several standards targeting parts lities. Many systems are in their infancy arranged according to very specific of the LMDS systems exist or are under and the wide diversity of technical imple- Operator needs. way. mentation and regulatory constraints is making the successful exploitation of The wireless telecom services have up The wireless datacom technologies have such technologies more difficult. The to now been provided by point-to-point emerged as an extension of the wired need for common air interface and radio- (P-P) radio-relay technology. Capacity LANs. They were primarily intended for to-fibre interfaces has been identified as available is typically 34 Mbit/s (option- indoor use, but have evolved to cover a key enabler for a real and extensive use ally 155 Mbit/s) and symmetrical ser- outdoor services as well. For the time of broadband radio solutions; the visions vices are usually offered. Primarily, P-P being radio LANs (RLAN) or wireless elaborated are now being downstreamed in technologies target the business market LANs (WLAN) are the most adequate the appropriate bodies (see Deliverable 4). and there are no widely adopted Com- terms and the preferred frequency band mon Air Interface (CAI) standards avail- is the 2.4 GHz band. The capacity is a able. Recently several manufacturers couple of Mbit/s in Europe with an 2.1 Short term wireless have launched so-called point-to-multi- achievable range of approximately one technologies point (P-MP) technologies. They are in kilometre due to restrictions on trans- Wireless technologies have been around most aspects an extension of the P-P mitted power. Limited mobility is offered for quite a while and at the turn of the technologies and the P-MP feature is by some systems and a standard is evolv- millennium there is strong evidence in- traded for lower capacity, typically ing named IEEE 802.11. dicating that wireless solutions are about 2 Mbit/s (optionally higher), and shorter to become a viable means for providing range in order of 510 km. Preferred fre- Next Generation cellular mobile. It is broadband access to the majority of cus- quency bands are 3.5 and 10.5 GHz and also worth keeping an eye on the UMTS tomers, at least in developed countries. there are no widely adopted CAI avail- technologies, which are the evolutionary Broadband communications in this con- able. successors to the cellular mobile systems, text refers to end-user capacity in the like GSM. They may offer near broad- range 2 Mbit/s up to 155 Mbit/s in both Wireless broadcasting technologies are band capabilities and will be highly directions. developing along several different paths. standardised. In the field of terrestrial broadcasting, the There is a broad range of wireless tech- DVB-T standard is widely followed but A summary of the target deployment nologies available or predicted to be does only offer digital broadcasting at scenarios and time-scale of the different available in the short term and they are about 2025 Mbit/s per channel. Another technologies are shown in Table 2. mostly terrestrial and satellite techno- development path is represented by the logies. Satellite technologies have been multiservice multipoint distribution sys- delivering broadband broadcasting ser- tem (MMDS) type technologies. They 2.2 Going wireless: vices for some time already with a spe- emerged as a wireless extension of the main conclusions cial focus on the residential market. cable technologies and are as such often The main findings of this study may be related to the contradictory term wireless summarised as follows: Satellite technologies are very much cable. They are primarily deployed in the Terrestrial LMDS systems will be capacity limited and their main compe- spectrum below 10 GHz and typically in available within 1 to 2 years. They will titive feature is the regional or global the 2.5 GHz band and offer highly asym- be able to support asymmetrical trans- coverage. Current systems are located metrical services due to the small amount Telektronikk 2/3.1999 177

179 Table 2 Potential deployment scenarios Wireless Residential Residential Small and med. Large technologies Limited service Full service business business Now P-P * * * * * * MMDS * * * DVB-S * * * < 2 yrs RLAN * * * * DVB-T * * * P-MP * * * * * * > 2 yrs UMTS * * LEO * * * * * * * * LMDS * * * * * * * Urban Suburb Rural Urban Suburb Rural Urban Suburb Rural Urban Suburb Rural port services (typically 25 Mbit/s Operators using broadcasting type sys- 3 Broadband fixed downstream, 1 to 2 Mbit/s upstream). tems (digital satellite, terrestrial UHF They will be most useful in urban and and microwave MMDS) will also be access: interim suburban areas, although the study able to compete with established oper- and target solutions could not rule out their use in certain ators, but in highly asymmetrical ser- rural areas as well. Full coverage of vices only. The dramatic technological evolution and all potential users by terrestrial the regulatory changeout are leading to For business users broadband point-to- microwave systems will not be feasible the installation of a variety of broadband point and broadband point-to-multi- in many areas. LMDS is therefore access systems. point radio systems may be an attrac- most suitable for competitive environ- tive alternative to optical fibre access ments where full coverage is not an On the one hand, operators look at target systems due to lower cost and/or speed essential requirement for obtaining a solutions: a wider time frame scenario of implementation. significant market share. enables the investigation of fibre rich implementations, taking advantage of UMTS targeted for 2002 will offer fast ETSI BRAN and ATM Forum are work- potential benefits in terms of global net- mobile data up to 2 Mbit/s for some ing towards new broadband solutions. work rationalisation and optimisation. users and N-ISDN equivalent mobile Operators and manufacturers should The overall access network optimisation services to many users. It seems un- promote standardisation of common air potentially enables significant cost sav- likely that UMTS could be an alterna- interfaces in order to lower the costs of ings, improved service quality and even- tive to broadband wireline or wireless user radio terminals. Compatibility with tual integration of services for business access solutions. fibre based access networks should be customers, provided that specifications ensured by also defining common net- Future broadband satellites (LEOs) and standard solutions are being devel- work interfaces. Harmonisation of fre- may support broadband wireless oped. The target solutions encompass quency bands in Europe should also be access and can be a solution for cost- both ATM PON and ATM point-to-point encouraged. effective broadband access networks (Deliverable 2); one or several evolution for millions of users globally, but paths from a set of existing access net- As the mobile networks are spreading unfortunately their broadband user work architectures to the identified target globally, operators should seriously con- capacity will be negligible locally. architecture are addressed (Deliverable 6). sider the potential synergy and cost bene- Stratospheric platforms have theoreti- fits of using a common infrastructure cally more capacity and may thus be On the other hand, the compelling need (transmission to base stations, towers, able to solve some of the coverage to enter new markets and provide new etc.) for both mobile and fixed wireless problems of terrestrial wireless net- services requires the full exploitation of access networks. works. existing infrastructures, re-using twisted 178 Telektronikk 2/3.1999

180 pairs, coaxial cables and, under certain initions which are mainly theoretical, 3.2 Evolution from existing circumstances, even powerlines. The without the advantage of a feedback access networks to the broad diversity of the technical imple- experience. target solution mentation is able to match the diversity The evolution studies start with a state- of service offering and acceptance, exist- For the definition of the ATM-PON ment of their market and regulatory ing infrastructures and other local con- architecture the following two main drives, describe their starting situation, straints, enabling the investment optimi- assumptions have been adopted: ie. the existing access network, and chart sation (Deliverable 6, Deliverable 3 and The architecture is that defined by a path towards the target network through Deliverable 10). The results offer the FSAN. a small number of intermediate steps. Operators the possibility to identify the Each needs to be marked with a cost ana- relevant parameters to be considered Broadband (BB) and narrowband (NB) lysis. The evolution studies performed when planning an upgrade of its own service integration is mandatory as a are the following: access network, understanding the capa- target solution, though only broadband bility and limits of the existing initial services might be offered at a first From HFC to ATM-PON; architectures and giving full knowledge stage. From existing fibre networks to ATM- of the established target solution. PP/PON; The following main conclusions can be highlighted: From existing copper networks to 3.1 Target broadband access ATM-PP/PON; network architectures BB access networks offer a broad range of functionalities which have From existing copper to ATM-PON From the work already produced in the to be mapped in their constitutive net- through LMDS. former EURESCOM Projects, as well work elements, at a price. A compre- as from the results of the Full Services hensive set of functionalities is con- An extensive cost analysis of network Access Network (FSAN) initiative [6], it ceivable, but only one subset of it evolution from existing access networks is taken as formal input that the architec- will be included in a real deployment. to target solutions has been performed in ture towards which fixed access networks Which subset to incorporate depends Task 6 (Deliverable 10). are going to evolve is ATM-PON, and in on the Operators business case. some cases ATM on Point-to-Point links (ATM-PP). Since at this moment there is The BB access networks described 3.3 Interim solutions a variety of different access network in this document are intended for the The regulatory changes which have taken architectures, and Operators would wish mixed residential and small businesses place in the last few years, together with in some cases to reuse part of their infra- market. the emerging demand for new services structure, this Project analyses how (if ATM-PON and ATM-PP networks often developed in the Information Tech- possible), an existing architecture might offer very similar functionalities. In nology area, require a rapid and econom- evolve towards ATM-PON or ATM-PP. general, ATM-PONs make a more ical deployment of new infrastructures. efficient use of network resources, like Therefore, beside the target solutions, These two architectures have much in fibre sharing and traffic multiplexing a number of short term implementations common; for instance, the services they at distribution network interface level, allow the quick provision of new ser- are intended to support, the Customer while ATM-PP networks may be more vices. Premises Networks (CPN) to which they flexible in terms of geographic diver- can be connected, and their external sity. Several technologies are being developed interfaces. On the other hand, the aspects for this purpose, and all of them put which are linked to the use of optical Concerning BB residential access net- leverage on the possibility of bypassing splitting, like Medium Access Control works there are currently two signifi- part of the network deployment exploit- (MAC) mechanisms, are certainly differ- cant issues which are far from being ing the installed copper/coax network, ent. understood. One of them is how to power cables or radio transmission. The integrate telephony and ISDN within use of similar systems in some areas will One problem when defining these archi- an ATM stream. Another is the subject lead to an increasingly complex overall tectures is that currently they practically of BB CPNs. Both issues are being network: the integration of interim solu- do not exist. Certainly there is almost no addressed in a number of fora, but lack tions is likely to become the main chal- deployment of ATM-PONs, and while market experience. lenge to be handled at a later stage. ATM-PPs do exist, they are intended for business customers only, and they incor- In summary, when currently the deploy- Some of these solutions are likely to be porate network elements which suppliers ment of interactive BB services for resi- exploited by competitors: as an example, have produced mainly out of their own dential users is proceeding slowly, both joint ventures between local power utili- entrepreneurship, without relying on a because of a lack of a clear market and ties and foreign telecommunications solid body of backing standards. It for regulatory reasons, it is necessary to operators are very common and will ben- should be kept in mind that at the time of raise the awareness of the intricacies of efit from the powerline transmission writing even the much expected Service BB networks, which are a long way from technologies, if sufficiently mature and Node Interfaces (SNI) VB5.1 and VB5.2 the access networks in use today. economical. are not officially approved. This lack of physical existence produces network def- Telektronikk 2/3.1999 179

181 Table 3 Areas of applications for interim systems 4 Basic technologies and practical imple- Incumbent Newcomer mentation aspect Low-cost DSL Yes Yes (with unbundling) Among the six tasks of project P614, one Powerlines No Yes is devoted to enabling technologies in which major building blocks have been Radio Yes Yes investigated with main emphasis on the HFC networks Yes (if available) Yes state of the art and near term evolution of technologies as well as cost figures and standardisation issues. The contents of the first report (Deliver- able 5) are: optical cabling technologies The success of these solutions is strictly All the described interim solutions do not (hardware), optoelectronic modules, elec- related to the evolving regulation: fully support the envisaged set of ser- tronic functions, xDSL techniques, alter- vices: they cannot be considered as full native fibres for low cost cabling, power- Exploitation of radio spectrum is service options, at least with the current ing, civil work, installation techniques. strictly regulated and rather non- maturity. Therefore, there is little scope The second one (Deliverable 7) will be homogeneous across Europe; comparing them with the systems ana- focused on three major items: point-to- The copper access network unbundl- lyzed in previous sections: interim solu- point link, measurements and mainte- ing, with copper pair rental to competi- tions have their own economic viability nance of the hardware and opportunities tors, may trigger the success of low and support very specific business cases. for WDM technologies in access net- cost DSL technologies; Table 3 summarizes the possible applica- works, together with a study on feasi- tion of the technologies described in the bility of powerline communications. The use of CATV networks and local chapter; more details can be found in networks in building needs settlement Deliverable 6 and other P614 documents. Civil work and installation represent the of possible ownership problems; most important factors in the global cost Transmission over powerline has very of a new access network. Advances in complex safety and regulatory implica- this field can strongly influence the prac- tions. tical feasibility of network deployment, affecting both direct and indirect cost. Moreover, the use of new techniques allows reduction of the social and en- vironmental impact of road works, espe- cially in urban areas. FTTB 4.1 Optoelectronic modules ONU for ONU ONU MDF The optoelectronic converter (O/E) has LEX been recognised as a major stumbling FTTC/Cab block for large scale optical fibre deploy- OM MDF ment in the access network. Bearing in FTTH/A mind the above specifications, present O ONU and near term technologies will be pre- ONU sented with main focus on one fibre L transmission system since this approach T ONU leads to cost effective cabling infrastruc- ONU ture investment. Today, commercially available bi-direc- tional O/E modules are mainly based on Duct Duct/direct buried micro-optics technology: in a single Manhole package, the module contains the laser diode equipped with its monitoring Cables: Connectors: detector, a photodiode and a WDM dichroic filter. All these elements are Splices in enclosure: Splitter: mechanically assembled with lenses and MDF: main distribution frame OM: optical monitoring a fibre by using precise and time con- suming alignment techniques resulting Figure 1 Main FTTx optical cabling elements 180 Telektronikk 2/3.1999

182 in a high cost (around USD 200 for 104 Fibre ending still requires well-trained for an automatic output of utility maps, pieces). technicians and expensive tools. Innova- updating existing digital cartography. tions are foreseen to solve this problem To reduce the cost a new generation of in the near future. Other technology New construction methods alternative to O/E modules is under development in breakthroughs are related to glass or traditional digging techniques are strictly many laboratories, it is based on hybrid plastic ferrule, ferrule-less connectors, related to the need to reduce th social and integration of optoelectronic chips with multi-fibre connectors and miniaturisa- environmental impact of civil work. passive optical components (splitter, tion. WDM). The passive components are The so-called No Dig techniques allow implemented on a silicon platform which Splitters. As a basic component for PON installation of underground ducts per- contains a V-groove for fibre positioning. systems, splitter technology analysis forming small tunnels without digging To further lower the cost this hybrid reveals quite a high cost figure of about long and deep trenches which, especially module will be packaged into a surface USD 50 per end or port; in addition some in urban environments, cause many mountable case for easy soldering on extra cost should be added for field inconveniences to traffic and pedestrians. printed board. installation and maintenance. The loca- No Dig techniques, mainly used in Japan tion of splitters in the outside plant or and Italy, sometimes represent the unique This new O/E generation still requires at central office still remains an open solution to work in urban environments, many technical challenges to be over- question. where Local Authorities do not allow come, for example: expanded beam laser, opening of new trenches, in order to edge photodiode, passive alignment Optical monitoring. Several solutions avoid traffic jam and damages to artistic soldering techniques, detachable pigtail, are envisaged: dark fibre and in-service and historical areas. plastic encapsulation. Finally, as a third monitoring. In the latter case, the wave- step, monolithic integration on InP mate- length could be 1625 nm, key compo- The installation of telecommunication rial of O/E converters appears as a long nents will be embedded filters in connec- cables in sewer ducts is a possibility: term target. tors, optical switches with a high number even in this case the final aim is to of ports at low cost and WDM. reduce social and environmental impact In a short time scale, advanced devices by avoiding new civil works. Another incorporating the above mentioned inno- interesting alternative to digging is to vations are expected to generate signifi- 4.3 The ANCIT Workshop install small optical cables in shallow cant changes in cost. Beside the survey and appraisal activi- and narrow trenches along roads or pave- ties, P614 organised an international ments. workshop on access network cabling and 4.2 Optical cabling installation techniques [7]. The following There are several new techniques for technologies main points emerged: installing optical fibres. Using appropri- These technologies mainly include: ate materials and equipment it is possible 1 There is a need to develop and use new fibres, cables, connectors, fibre splicing to install directly in small underground techniques to reduce the costs related and enclosures, fibre termination mod- ducts, single fibres or fibre bundles by to the deployment of a broad-band ules, main distribution frame, splitters. blowing or by pulling. Field results and access network; A few examples will be given showing economic evaluation show the main that significant innovations are still 2 There is a need to improve the flexi- advantages offered by these techniques in needed in this area. bility of the installation techniques; comparison with traditional cable instal- lation: design and installation flexibility 3 There is a growing attention for the Fibres/cables. Cables are installed in var- and global cost saving. environmental and social impacts ious conditions: outdoors, underground related to civil work and installation. (in duct or buried), aerial, indoors; conse- quently many thermal, mechanical and 5 Broadband upgrade The second point is strictly related to the chemical constraints have to be taken ever higher constraints that Local economics into account. In urban areas and for Authorities impose on Operators and indoor cabling, the topology presents In Task 6 of the EURESCOM P614 companies working in urban environ- many bends leading to small curvature project an extensive techno-economic ment. radius for fibres and cables. assessment has been carried out, in order to assess the overall economics of broad- Starting from the network design, new One driving force of todays cable de- band access network upgrades and iden- CAD systems allow optimisation of the velopment is to reduce the installation tify economically viable implementation design itself but also help the operators costs. Compared to copper, optical cables strategies for the broadband access net- to perform quicker maintenance and up- offer tremendous potentialities in terms work technologies and strategies studied dating operations. Improvement in site of weight and diameter, then allowing in the project. Several migration alterna- investigation can avoid drawbacks due the use of cost effective laying technique tives for access network providers down- to digging or drilling works for the con- such as air blowing in ducts. town, urban, suburban and rural areas in struction of underground infrastructures. Europe have been examined, including On the subject, a New Ground Penetrat- Connectors. Field assembly capabilities the different options available for tradi- ing Radar system for the detection of appear not yet ready for true mass appli- tional telephone operators, cable opera- underground existing utilities allows a cation. Improvements in fibre preparation tors and new entrant operators. The direct link with CAD and GIS systems and positioning into plugs are needed. methodology and tool initially developed Telektronikk 2/3.1999 181

183 Fixed radio access and ultimate strategies, guidelines and technologies somehow hampered the real Mobile field deployment. Satellite PSTN (existing As a matter of fact, a number of advanced Powerlines access networks are likely to emerge, copper plant) relying on technologies under fast devel- opment. From a technical point of view the challenge for Operators is the full Broadband exploitation of a range of technologies in CATV network overlay network their own network, and effectively inter- operating legacy and new systems. Beside the technology specific issues, successful broadband access network USER deployment requires enabling and sup- porting development in mobile and fixed Figure 2 Several access techniques reaching the end user: networks convergence and broadband is access going to turn into a commodity? home network installation practices [9], favouring the customers acceptance of new services. And finally a provocative statement to by the RACE2) 2087/TITAN3) project, of the network roll-out projects in the highlight a different point of view: is and further developed in the ACTS4) 226 defined area have been examined, taking access becoming a commodity? The OPTIMUM5) project have been applied willingness to pay, revenues and com- number of alternative access systems and in the techno-economic analysis [8]. petition effects into consideration. The techniques exploiting existing infrastruc- business risks associated with the up- tures (copper pairs, CATV network, Broadband service scenarios for the next grade strategies have been quantified, power distribution), coupled to wireless ten years have been defined, based on including market introduction risk, rev- solutions, stimulates tough competition. results from international studies. The enue risk, technology risk and overall analysis of the provision of services from network evolution risk. The financial This statement stimulates a new vision the surveys covers upgrades in four net- analysis carried out with OPTIMUM has of access network evolution: the initial work area types, which have been seg- been complemented by an industrial cost competition in the transport network and mented and characterised according to assessment, in which the production cost in the business market segment, where average copper loop length in the exist- per service has been calculated for differ- the margins are shrinking, will later shift ing access network (also reflects the ent evolutionary paths. to access and residential market. Opera- density of living units in the area), avail- tors prepared to deliver a full range of ability of existing ducts and surface con- The analyses confirm that the cost of services will be in a better position in the ditions with corresponding cable deploy- increased bandwidth in the access net- access competition against players able ment type and civil works costs. Repre- work for interactive broadband delivery to offer just few services with limited sentative ranges of the characteristic is high, independent of the operators performance. parameters have been assigned to each existing network situation, area type and network area segment. broadband technology choice. Telephone operators, cable operators and new Acknowledgement Starting from the pure Installed First entrant operators are likely to face broad- This document is based on results Cost (IFC) and Life Cycle Cost (LCC) band upgrade investment levels per con- achieved in a EURESCOM project; this analysis of the selected Evolutionary nected user similar to or higher than the does not imply that it reflects the com- Paths (EPs), the business opportunities overall costs of establishing the existing mon technical position of all the access network. Moreover, the node con- EURESCOM Shareholder/Parties. The figuration and corresponding degree of author gratefully acknowledges the sup- fibre penetration has a great impact on port of EURESCOM for carrying out this the cost level. Thus, the fibre penetration work. The author wishes to express a 2) RACE: Research in advanced commu- and the location of the optical nodes in special thanks to all P614 participants, the network is a key strategic decision. nications in Europe. since this paper reflects the results of 3) TITAN: Tool for introduction scenario their work. and techno-economic evaluation of 6 Conclusions and EURESCOM P614 members are: BT, access networks. future work CNET-France Telecom, CSELT, DTAG, 4) ACTS: Advanced communications Hungarian Telecom Company, OTE, Broadband access network introduction technologies and services Portugal Telecom, Sonera Ltd., Swiss- has been extensively discussed during the 5) OPTIMUM: Optimised network archi- com, Telecom Ireland, Telefnica I+D, last ten years within all the research pro- Telenor. tectures for multimedia services. grams in Europe. The search for common 182 Telektronikk 2/3.1999

184 References 1 Warzanskyj, W, Ferrero, U. Access network evolution in Europe: a view from EURESCOM. In: Proc. ECOC 1994, Florence, Italy, 2529 Sep 1994, 135142. 2 Ims L Aa et al. Multiservice access network upgrading in Europe : a techno-economic analysis. IEEE Communications Magazine, 34 (12), 124134, 1996. 3 Ferrero, U. Broadband optical access network : cooperative work among European PNOs. In: Proc. of ECOC 1996, Oslo, 1519 Sep 1996, 3.3 3.10. 4 Ferrero, U, Mavis, M, Warzanskyj, W. The challenge of the broadband access: PNOs cooperation in Europe. In: Proc. VIII Workshop on Optical/ Hybrid Access Networks, Atlanta, 25 Mar 1997, paper 2.4. 5 Loktu, H et al. Status and future evo- lution of broadband radio access. In: Proc. NOC 98, Manchester, UK, 2325 Jun 1998, paper 111. 6 Full Services Access Networks home page: access/index.htm. 7 Proceedings of the Workshop on Access Network and Installation Techniques (ANCIT), Torino (Italy), 3031 Mar 1998. 8 Ims, L Aa et al. Economics of broad- band access network upgrade strate- gies: the European perspective. In: Proc. Globecom 98 Mini-conference on Access Networks, Sydney, Aus- tralia, 812 Nov 1998. 9 Proceedings of the Workshop on Tomorrows Residential Infrastruc- ture for Broadband And Narrowband (TRIBAN), Bern, Switzerland, 1719 Nov 1998. Umberto Ferrero (32) graduated in Telecommunication All the papers published by access net- Engineering from the Politecnico di Torino in February work related EURESCOM Projects, 1992. He joined CSELT in 1991, where he now works as together with all the P614 Deliverables Senior Engineer. He is EURESCOM P614 Project Leader and information on the workshops, are and is involved in access network architecture design and available on the Internet at: economic evaluation. He has published more than 50 papers on access networks related topics. welcome.html email: [email protected] Telektronikk 2/3.1999 183

185 Towards broadband access in Japan ATM access for Mega-Media Services KENJI OKADA The number of telephone circuits pro- as broadband access in order to satisfy telecommunication has been changed vided by NTT (Nippon Telephone and these requirements. FTTH/B (Fibre To from a family to a personal basis. While Telegraph) slightly decreased in 1998, The Home/Building) will be accomp- LANs (local area networks) have been which is the first time since NTT was lished by deploying ATM-PON sys- installed in all business offices, LANs started. Present telecommunication tems. At the initial stage, ATM-PON have also been installed at the premises networks have been constructed to systems will be deployed for business of advanced residential customers. The optimize for telephone services for applications, especially in order to introduction of LAN at home results in over a hundred years. A narrowband target the SOHO (Small Office, Home the expansion of ISDN, and finally the but a constant path is required for Office) market. 10/100 Base-T interface will be provided telephone services and especially a directly to the residential customer. The FSAN (Full Services Access Net- small transmission delay is needed for These trends confirm the broadband work) Initiative has been established analog telephony. The deployment of service demand. in order to obtain cost effective access fibreoptic cable in the access network systems based on a de facto standard. has now started from the central office to the access point for telephone ser- Specifications of the ATM-PON have 2 Trends of telecommu- been standardized in FSAN and are vices, that is near FTTH (Fibre To The expected to be determined soon in nication services Home). However, the provided service ITU-T (International Telecommunica- is still POTS (Plain Old Telephone Telecommunication service trends must tions Union Telecommunication Sec- Service). be reviewed since access networks are tor). This paper presents the trend of designed according to the telecommuni- On the other hand, there have been telecommunication services in Japan cation service demands. It is essential to remarkable increases in high-speed and describes the broadband access determine the direction of access net- digital services for computer commu- systems now under development. works and applied access systems after nication. Computer communication not only reviewing the trend of current may undoubtedly occupy the telecom telecommunication services, but also services instead of telephone services 1 Introduction anticipating the future telecommunica- and it may require broadband and tion services when shifting from POTS peaky path. Notable extensions of com- The environment surrounding telecom- (Plain Old Telephone Service) to the puter communications force the tele- munications has been changing rapidly. broadband services. The most remarkable com network to evolve. An ATM back- One of the big trends is the growth in trend in NTT is that the market for POTS bone network has already been con- data telecommunications such as Inter- has been saturated for the last ten years, structed in Japan. However, it is essen- net, whose volume obviously will exceed and moreover, the number has decreased tial to establish an end-to-end path the volume of telephone traffic. Another by one million subscribers in March without bottlenecks, not only in the trend is the explosive expansion of mo- 1998, as shown in Figure 1. During the core network but also in the access bile telecommunications. In Japan, the 100 years since NTT started telephone network. ATM-PON (Asynchronous number of mobile phones has grown services this is the first time that NTT Transfer Mode over Passive Optical beyond 40 million. There are 65 million experience that the number of POTS has Network) systems have been deployed fixed telephone lines in Japan. Voice decreased. The telephone service has dominated telecommunications up till now. There- fore the current telecommunication net- work has been optimized to the telephone service. In short, NTT have aimed at a 108 constant speed and short transmission POTS delays for the telecommunication net- 107 Internet user work. 106 ISDN On the other hand, digital telecommuni- Number of Circuits cation services such as ISDN (Integrated 105 Services Digital Network), HSD (High- Speed Digital Leased Service), FR 104 HSD (Frame Relay), OCN (Open Computer FR 103 Network, a connection-less mode OCN telecommunication service), and ATM (Metal) Megalink (ATM virtual path service) 102 OCN have shown a remarkable growth (Figure (Fiber) 1). These digital telecommunications are 10 ATM Mega link used for Internet access and telecommu- 1 nications between computers, and these 86.3 87.3 88.3 89.3 90.3 91.3 92.3 93.3 94.3 95.3 96.3 97.3 98.3 services have a great possibility to occupy the telecommunication network instead of telephone services. For refer- Figure 1 Telecommunications service trends ence, the number of Internet users are 184 Telektronikk 2/3.1999

186 10 G OCN service are all available as broad- band services. In Figure 3, the virtual Video path type of services are found in the 1G distribution upper part of the figure, whilst the virtual circuit type of services are found in the Picture, Journal middle of the figure. Initially, HSD was 100 M b/s 0M or Software the only broadband service. This is an Total information (bit) 60 b/s Download STM based service which provides 0M 10 M 15 The Transfer broadband leased line services at a speed POTS from 64 kbit/s to 150 Mbit/s. The CR (Cell Relay) service started as a broad- 1M b/s band service with speeds from 6 Mbit/s 6M s to 135 Mbit/s on an ATM network. How- b / 1. 5M E-mail ever, the take up has been somewhat 100 K WWW slow, possibly due to the relatively high b/s price. Therefore, the ATM-Megalink ser- 8k vice has been taken into consideration. 10 K 12 /s kb Transaction 64 kb/s This service provides an ATM virtual 28 .8 path service at a speed of 500 kbit/s 1K kb /s 135 Mbit/s. This service is growing 9.8 rapidly, as seen from Figure 1. At the end of last year a best effort type of service 100 msec 1 sec 10 sec 1 min 10 min 1 hour was offered with the ATM-Megalink, Access Time using a GFR (Guaranteed Frame Rate) function. Figure 2 Categorization of telecommunication contents 4 Access system toward FTTH At present there is a bottleneck of also shown in Figure 1. The Internet sub- 3 Telecommunication telecommunication in the access part of scribers use ISDN or OCN for their Inter- net access. HSD, FR or ATM Megalink service line-up the network, as depicted in Figure 4. are used by business customers. These NTT already has various kinds of broad- The introduction of optical transmission computer telecommunications may have band services as shown in Figure 3. systems has greatly increased the trans- a significant impact on the evolution of mission capacity in the core network. the structure of telecommunication net- The HSD service, the ATM-Megalink SDH (Synchronous Digital Hierarchy) works. service, the Cell Relay service and the systems with a few Gbit/s capacity Requirements for telecommunication net- works may be extracted from the charac- teristics of telecommunication contents. Telecommunication contents are catego- "Connection mode service" rized by the total information of contents High and the access time, as depicted in Figure Leased circuit High-speed digital leased 2. ATM Virtual Path VP The slant line indicates transmission Telephone speeds, which are derived from the total Traffic ISDN ISDN information per access time. The POTS Basic (Primary) requires 64 kbit/s of telecommunication speed and the access time from several Cell relay VC ISDN-P Frame tens of seconds to several tens of min- DDX-P relay utes. Speeds in the range from 10 kbit/s to 100 kbit/s are applied for access to Low WWW, due to the speed limitation of Speed metallic access systems. On the other 64K 1.5M 6M 50M 150 M hand, the capacity of WWW files is "Connection-less mode service" increasing year by year. Video distribu- tion signals must be transmitted in a short OCN IP (Dial up connection) OCN time interval due to the expensive server cost. Broadband access systems are needed if a short access time is required. Figure 3 Telecommunication service line-up Telektronikk 2/3.1999 185

187 who have an aerial section or duct sec- Access Network Access Network tion in the access network. FTTH is also Metallic suitable in green field areas. Fortunately, Telephone Yesterday Telephone NTT has a good availability of ducts at Coaxial cable the feeder section and distribution section Release bottleneck of Core Network Single mode fiber in business areas, and has aerial cable at the distribution section in the residential areas. Thus, it is relatively easy for NTT Optical fiber to deploy an FTTH architecture. Metallic Internet Telephone Today Internet Telephone It is important to decide on the strategy Release bottleneck of Access Network SS : Single ADS : Active Double Star for FTTH in the initial stage when the PDS : Passive Double Star demand for broadband services has Optical fiber developed. There are two approaches towards FTTH, as illustrated in Figure 5. FTTH Optical fiber FTTH One approach is a network-oriented Internet Internet Video Tomorrow Internet Internet Video approach, and the other is a service- oriented approach. Figure 4 Status of the telecommunications infrastructure In the network-oriented approach, the services provided by the optical access systems are the same as the services provided by the metallic access systems. This approach is applied by NTT in order to improve the existing infrastructure. However, optical access systems for have been deployed in the core network. plexing) technologies for the ISDN narrowband services must be realised However, metallic systems are used in metallic systems. It is advantageous for with the same cost as the metallic system most of the access network, and this HFC that cable TV companies optimize alternatives. In this approach, the ONU causes a bottleneck of telecommuni- the existing coaxial cable. However, it is (Optical Network Unit) for narrowband cation. not so advantageous for telecom opera- services is installed at the access point tors to install new coaxial cable. FTTC is (point at user side closer than feeder But the bottleneck in the core network good for the telecom operators who have point), which is near the customer has been released by introducing optical direct buried infrastructure, since it is premises (near-FTTH: n-FTTH). The systems instead of metallic systems. This very expensive to replace direct buried ONU accommodates about ten sub- is also the key issue for FTTH. If optical metallic cable with new optical cable. scribers. access networks are constructed with FTTH is good for the telecom operators reasonable cost, these will enable the provision of broadband services in a flexible and rapid manner. Especially for high speed Internet access broadband access networks will be needed. A narrowband but constant path network was required for the telephone service. However, computer communication may require a broadband and peaky path opti- FTTH FTTH mised network. It is essential to establish an end-to-end path without any bottle- N-PON(+SCM-PON) necks, neither in the core network nor in the access network. Fiber Area Near-FTTH ATM-PON There are many kinds of access systems providing broadband services, such as ADSL (Asymmetric Digital Subscriber N-PON [ Network-Oriented Approach] Loop) systems, HFC (Hybrid Fibre Co- axial) systems, FTTC (Fibre To The Service-Oriented Approach ] Curb) systems, and FTTH systems. ADSL is good for providing broadband Metallic Area Existing metallic System xDSL services fast, but limits the transmission quality, the applied distance, and the number of provided subscribers. Espe- Narrowband service Broadband service cially, there are some problems of cross talk in the Japanese network, since NTT applies TCM (Time Compression Multi- Figure 5 Approach to FTTH in NTT 186 Telektronikk 2/3.1999

188 In the case of near-FTTH, it is easy to various kinds of services, which have extend from an n-FTTH architecture to even different speed class or quality, FTTH, since there are additional fibres in through his ATM network. the optical cable from the central office ATM-PON to the access point. If a customer requires Field Trial area Through these field trials, it has been a broadband service, and an optical cable clarified that users business users in recently has been installed between the particular require ATM services at a customer premises and the access point, reasonable price level. Therefore, NTT an additional fibre is connected. If the has deployed ATM-PON systems in the customer requires both a narrowband commercial networks in order to provide Makuhari Lab. service and the CATV service, a narrow- Tokyo ATM services and utilise this experience. band PON (N-PON) and a SCM (sub- NTT was afraid to miss an opportunity to carrier multiplexing) PON (SCM-PDN) Urayasu introduce the ATM service if we were to are applied by the use of WDM (wave- Yokosuka wait for the standardized ATM-PON sys- length division multiplexing) technology. tem to become available. NTT decided to The N-PON and SCM-PDN uses the Yokosuka Lab. provide the ATM service as soon as pos- 1.3 m and 1.5 m wavelengths, respec- sible by applying the same ATM-PON tively. system as used in these field trials. This Figure 6 Field trials of ATM-PON is the so-called ATM Megalink service as systems for FTTH In the service-oriented approach, when described above. broadband services such as the ATM ser- vice are demanded, an FTTH architecture will be realized by the use of an ATM- 6 Outline of ATM service PON system. The FTTH architecture and access system includes FTTB (Fibre To The Building) in this paper. As a first step business In the present commercial network, there users are provided with broadband ser- are only a few speed classes in the STM vices through FTTH. If broadband ser- Many good features of the ATM service based HSD service. However, by apply- vices are demanded in an area dominated have been envisaged through these field ing cell-based multiplexing as in the by a metallic cable based infrastructure, trials. There are many advantages for a ATM Megalink service a huge class of xDSL technologies will be applied in network operator if an ATM network is speeds ranging from 0.5 Mbit/s to 135 order to obtain fast provisioning of ser- constructed. The network cost reduction Mbit/s is available, ie. every 1 Mbit/s vices. However, xDSL is considered to is led by simple multiplexing with non- pitch. The optimal speed may satisfy the be a temporary solution, as discussed hierarchical multiplexing and easy main- users request. Up to 44 Mbit/s (equiva- previously. It should also be noted that tenance using cell-base multiplexing. lent) services are provided through the both the operator and the user need an From the viewpoint of the customer ATM-PON systems and up to 135 Mbit/s opportunity to stimulate broadband ser- there are many attractive features, such (equivalent) services are provided vices raised from user demand in order as lower price, various classes of speed, through conventional STM-1 single star to break the chicken and egg challenge. and quality control. Of course, the basic (SS) systems, since the ATM-PON sys- concept of ATM is service integration, tem was optimized for service speeds and the network operator can provide up to 44 Mbit/s (see Figure 8). 5 Field trial of ATM-PON system and acquired understanding NTT conducted the field trial of the ATM service applying an ATM-PON system in order to investigate the demand of future broadband services and the applied system technology. Since early Urayasu Yokosuka 1996, field trials of ATM-PON systems have been conducted in Yokosuka and Urayasu city, as shown in Figure 6. 1st quarter of 1996 1st quarter of 1996 Test Period - March 199 - March 199 In these field trials, CATV, VoD and About 300 About 300 ISDN services were provided for up to Number of Subscriber 300 customers in each city (Figure 7). Cable TV These systems had NTT proprietary Service VOD (MPEG2) interfaces, but had interoperability be- ISDN tween the OLT (Optical Line Terminal) Network ATM-PON and the ONU, which were manufactured by different suppliers. This made the maintenance easy. Figure 7 Outline of the field trials Telektronikk 2/3.1999 187

189 DSU/ONU ATM-SLT ATM-SLT DSU/ONU NTT has started to deploy ATM-PON systems due to their good features, as illustrated in Figure 9. VP#1 VP#1 VP#2 The PON topology has many advantages compared to conventional SS (Single Star) topology. The first advantage is the lower system cost, due to the sharing of Access Line Network Access Line the OLT between multiple ONUs. The second advantage is the space reduction ATM (1) protected CBR VP [Dual] due to the fact that the OLT can accom- Virtual Path modate many ONUs simultaneously. For Network (2) protected CBR and UBR VPs [Extra] example, SLICs (Subscriber Line Inter- face Cards) of switching systems occupy VP cell rate (3) non-protected CBR VP [Single] much space in central offices because a 0.5Mbit/s large number of SLICs is needed for each 1.0Mbit/s 2.0Mbit/s (1) ATM-PON (0.5Mbit~44Mbit/s) customer. Optical circuits need much Access more space than metallic systems if the 134Mbit/s line conventional SS topology is applied. The (2) ATM-SS (45Mbit~135Mbit/s) third advantage is the low power con- sumption. The overall power consump- Figure 8 Outline of the ATM services tion for the PON system is less than for the SS topology. However, there is cost limitation in ATM-PON systems if pro- prietary interfaces are applied in the existing ATM-PON system. NTT decided to introduce a standardized inter- In the core network, there are three relia- gle and Dual in the ATM Megalink face in order to achieve a cost effective bility levels depending on the service service. In Extra, half of the cells which ATM-PON system. price. Customers who make much are required to be highly reliable are account of economy will choose a core transmitted using guaranteed CBR, and network without protection (Single), the rest of the cells are transmitted 7 The FSAN initiative and customers who request the best relia- through another path by using UBR Each telecom operator has different net- bility will choose completely duplicated (unspecified bit rate). Customers can work architectures, system configura- core network (Dual). CBR (Constant choose service reliability level depending tions, or deployment plans in the access Bit Rate) technology is applied for Sin- on their requirements and economy. network. This is because each of them has different geographical conditions, development strategies, or service re- quirements. However, every telecom operator targets the cost reduction of optical access system and aims at broad- Center equipment band service provisioning. SS topology In 1995 a consortium named G7, consist- ONU OLT ing of seven large operators, was estab- lished in order to work towards common system specifications. Later on, the name was changed to FSAN (Full Service Access Networks), as more companies To reduce system cost joined the initiative. The structure of To squeeze space of center equipment FSAN is shown in Figure 10. To compress electric power of center equipment Operators who have their own access networks and exhibit their requirements Center equipment to access networks can be members of PON topology FSAN. 19 operators are members as of February 1999. The aim of FSAN is cost ONU OLT reduction of optical access systems for early deployment. In the beginning of FSAN there were discussions on intro- duction strategies for optical access sys- tems for broadband services. The discus- sions led to the consensus that the most Figure 9 Features of the PON technology 188 Telektronikk 2/3.1999

190 cost-effective way to provide broadband As of Feb. 1999 services is to share common system spec- ITU-T 19 carriers ifications. During the discussion the ATM-PON system has been defined as ETSI ATM-F the common target for each operator. FSAN aims at finding cost-effective FSAN solutions by defining a multi-vendor interface in the ATM-PON system. Ser- vice requirements from operators and Eire Telia technologies from suppliers have been BT KT harmonized in order to specify the inter- NTT BellCanada KPN face between the ONU and the OLT of DTAG Chunghwa GTE the ATM-PON system. The concept of FT/CNET SwissCom SingTel FTTx is produced to overcome the diver- SBC sity in geographical circumstances and TI/CSELT Telstra BellSouth the existing infrastructure, so that the Telefonica US West ATM-PON system can be applied in any architecture, ranging from FTTH, FTTB, FTTC to FTTCab, as depicted in Figure Figure 10 The FSAN structure 11. At present the standardisation bodies are trying to decide on the standards for the systems already developed. Therefore, the systems do not always fit the require- ments of the operators, and several sys- come the severe competition. NTT plans tems have been standardized at the same 8 Commercial plan to apply the next version of the ATM- time. Moreover, it takes a long time for in NTT PON system fully compliant to FSAN the standardisation work to reach com- specifications by June 1999. The FSAN pletion. FSAN decided to study some In April 1997 NTT started offering ATM compliant ATM-PON systems are ex- unique and what was considered the best leased line services carried over ATM- pected to be more economical than the system specifications which satisfy the PON systems. In Japan, the telecommu- former systems. The interface with the requirements of the operators. The first nication market is very competitive and ATM switch, the SNI (Service Node step is to clarify the requirements from NTT needs cost-effective access systems Interface), and the operation interface are the operators, and the second step is to and new service provisioning to over- also applied in compliance with the ITU- examine the technologies from the sup- pliers which satisfy these requirements. Unified interface specifications of ATM- PON systems have been completed in a short time. After completing the unique and best solution, the operators may, NW-OpS according to these specifications, procure Q3 the system from any supplier. Suppliers Service Node can sell the system to any operator FTTx demanding requirements satisfying these Passive Optical Splitter Internet specifications. FSAN is not a standardis- ONU FTTH ation body, so the results are often input Optical Fiber to existing standardisation bodies such Leased Line as ITU-T or ETSI (see Figure 10). At ONU FTTB present, the specifications of the access Frame Relay line interface have been recommended as Cell Relay OLT G.983.1 at ITU-T. Specifications of the ONU management channel defined between NT FTTC POTS ONT/NTE and OLT have been com- Twisted Pair pleted at FSAN and are planned to be frozen as G.983.2 at ITU-T. The Q3 VoD ONU FTTCab NT interface is also discussed and planned SNI to be input to ITU-T quite soon. (VB5.2) ATM-PON VDSL FTTH:Fiber To The Home FTTC :Fiber To The Curb FTTB:Fiber To The Building FTTCab: :Fiber To The Cabinet Figure 11 The FSAN architecture Telektronikk 2/3.1999 189

191 T VB5.2 or Q3 FSAN specifications, First Version respectively. The ATM-PON system is expected to cost-effectively provide Commercial release: April, 1997 ATM-PON interface frame format: NTT proprietary broadband services over mega-stream, Service : which is the so-called Megamedia ser- - ATM megalink service (ATM leased line service) vice in NTT. In this sense the ATM- PON system is regarded as the access - ATM-PON in compliance with FSAN platform. The system evolution of NTT - SNI(VB5) in compliance with ITU-T is illustrated in Figure 12. - Q3 interface in compliance with FSAN - Function enhancement (DBR, SBR, ABR, UBR, UBR+) At present a guaranteed type of service is Next Version available in the Megalink service (see Figure 13). Commercial product delivery: June, 1999 ATM-PON interface frame format: FSAN compliant In this type of service, service speed is (Full Services Access Networks) Service : guaranteed from end to end (CBR). ATM - ATM service including ATM service (ATM leased line service) starts from the CBR type of service like - Cell relay service the existing STM-based leased line ser- - HSD (High Speed Digital Leased Line) emulation service vice. Currently, non-guaranteed cell rates - OCN (Open Computer Network) with low price may be acceptable even for a leased line service. There is a possi- bility of reducing system costs using the Figure 12 The commercial system of NTT UBR+ service (Unspecified Bit Rate with best effort type, GFR). The GFR type of service is realized within core networks first because there are plenty of multiple paths. When broadband services in the future are introduced in the SOHO [ATM mega link service] Best effort and Best effort and market, there is a possibility to provide Guaranteed type guaranteed type guaranteed type the best effort type of service in the PON section due to the big dispersion of burst Guaranteed bandwidth Guaranteed bandwidth Shared bandwidth for one customer for one customer with multiple customers data. At that time, additional cost reduc- tions can be expected. Shared bandwidth with ONU ONU ONU multiple ONU ONU ONU customers 9 Conclusion OLT OLT OLT OLT OLT OLT Occupied with Shared 150 Mbps Shared 130 Mbps The next version of the ATM-PON sys- ONU ONU ONU ONU ONU ONU Mbps at End-to-End bps at trunk section bps at End-to-End tem targets big to medium businesses. Available to use Available to use Available to use The third step is to accommodate all constant Mbps maximum Mbps maximum 30 Mbps kinds of services from high-speed ser- Bandwidth Bandwidth Bandwidth vices to narrowband services. The cost of ATM-PON systems should be as low as possible to be attractive to all users. The price target for the Megamedia service Time Time Time is around 10,000 yens (80 USD) per month for bi-directional data service of 10 Mbit/s. Considerable effort will be Figure 13 Characteristics of ATM services concentrated on this issue. Dr. Kenji Okada (48) is Supervisor of NTT Access Service Systems Laboratories. He has been working in research and development of transmission systems including high- speed coaxial transmission systems, optical trunk trans- mission systems, optical access network systems and also network system development strategies. One focus of - recent efforts is to promote optical access systems deploy- ment. email: [email protected] 190 Telektronikk 2/3.1999

192 Towards broadband access in Norway the view from Telenor LEIF AARTHUN IMS, TRYGVE JARHOLT, KJELL STORDAHL, FRODE B. NILSEN, MARKKU LHTEENOJA, BORGAR TRRE OLSEN, DAGFINN MYHRE AND STIG LKEN This paper presents Telenors strategy data and multimedia services from for a market driven development of wireline and mobile connections. One the access network infrastructure possible outcome is that Telenor even- Market towards an effective production plat- tually replaces the variety of networks Demand form serving the future service port- currently in use with a Full Service Willingness to pay folio. A future flexible and competitive Network (FSN) [2]. A key feature of User behaviour broadband access network will enable such an FSN is an improved exploita- Telenor to access the customers with tion of the current access network by the broadband products to come after an increase in the number of local ISDN, ADSL, leased lines and GSM. access points compared to the present Technology Strategy The focus is on the medium to long PSTN (Public Switched Telephone term development towards flexible Network) infrastructure. This will Technology variety Where? access capacities up to 25 Mbit/s for enable capacities higher than 2 Mbit/s Open provisioning When? fixed and mobile services (UMTS) to be offered to end-users, as well as Service integration How? in the SME, SOHO and residential providing an effective means of meet- market. ing the transport capacity requirements of the third generation mobile tele- phone system, UMTS (Universal Figure 1 The access network challenge 1 Introduction Mobile Telephone System). Telenor is the incumbent telecommuni- However, turning the last mile into a cations operator in Norway, with a cus- long term, first class asset, represents a tomer base comprising 2.5 million POTS significant challenge for Telenor and (Plain Old Telephone Service) and ISDN other access operators, well beyond the (Integrated Services Digital Network) moves mentioned above. This paper subscriptions; 1,700,000 cellular sub- addresses this challenge from a strategic scribers and 225,000 Internet subscribers, point of view, taking into account the bution are expected to become all-digi- in addition to 280,000 cable television major aspects of market, technology and tal. The exponential growth in micro- subscribers. Telenor operates in one of economics [3, 4, 5, 6, 7]. Figure 1 illus- processor power, memory size and the worlds most advanced telecommuni- trates the key issues in the development storage capacity will probably con- cation and information technology mar- of a future broadband access network. tinue over the next ten years. This kets, and is thus likely to be in the fore- We distinguish between the challenges implies that the performance of the front over the next years with regard to associated with market, technology end-user systems will continue to network evolution, including access net- and strategy, as discussed next. increase, possibly to 100 times the work development. There are several performance of the current systems. indications that Telenor is now on the threshold of a major development 1.1 Market challenges towards broadband access delivery: 1.2 Technology challenges The challenge of broadband services and the last mile is first and foremost con- The future access network architecture Telenor is currently in the midst of the nected with the uncertainty in service will most likely be different from the highest growth in ISDN subscriptions demand, willingness to pay and usage present one in at least three aspects: in the world, which among other issues patterns: implies the initial step of a future- Technology variety. Both fibre solu- proof capacity upgrade of the access Broadband service demand: How tions, satellite systems, cable-TV net- network. large a growth in demand for trans- works, radio systems and DSL (Digital mission capacity can be expected Subscriber Line) solutions over the In June this year the service Telenor within the next five to ten years? A existing copper network can be part of ADSL was launched; Telenor being very significant growth in capacity one network. Which solution or com- one of the first operators in Europe to intensive services is expected, and binations should the operator select? offer interactive broadband access on a the transmission capacity is foreseen commercial basis. Initially, the service Openness: An open architecture is to be perceived as a commodity. will be available in central regions of required in order to operate in a com- Oslo, Brum and Troms, offering Willingness to pay: Is there any incre- petitive environment with different businesses Internet access at speeds mental willingness to pay for new mul- core network operators. up to 2 Mbit/s. The service coverage timedia services? The demand level Service integration refers to an access is likely to be extended to Stavanger, will to a large extent depend on the network in which all kinds of traffic Bergen and Trondheim later this year price and the pricing schemes, in par- are mixed by means of statistical mul- [1]. ticular in the very important mass mar- tiplexing1). This is in contrast to static ket. Almost simultaneously, Telenor signed multiplexing and dedicated connec- a letter of intent on a comprehensive Usage pattern: Which consequences tions which are currently being used. collaboration with Cisco Systems, with will the development in the industry The rationale for service integration is the aim of building a test version of an have on which areas broadband ser- integrated infrastructure based on the vices are applied and how they are Internet Protocol (IP) for telephony, used? Very soon the content and distri- 1) Also called dynamic multiplexing. Telektronikk 2/3.1999 191

193 efficient and cost-effective provision- 2.1 Telenors position From its hotbird satellite position 1 ing of broadband access. The chal- West, Telenor distributes more than 35 At present both fixed network operators, lenge is to maintain QoS (Quality of analogue and 70 digital TV channels in cellular operators, cable operators, power Service) in the mixed traffic stream. the Nordic region, Central and Eastern utility companies and new entrant opera- Europe. The five satellite 1 West posi- tors are offering telecommunications ser- tion has 51 transponders in total. Telenor 1.3 Strategic challenges vices in the Norwegian market. Telenors serves the business markets in Scandi- 1998 annual revenues amounted to 22.2 Moreover, a roll-out of a broadband navia, Eastern Europe and the United billion NOK, with a work force of 19,000 access network implies a set of strategic Kingdom, as well as in Africa. The appli- employees. challenges and imperatives: cations offered include voice and data point-to-multipoint services and satellite Where, when, how and to what extent Currently Norway has the second highest Internet backbone services, including should the operator invest in access penetration of Internet hosts (including satellite intranet. network infrastructure for broadband dial-up connections) in Europe, totalling services? approximately 325,000 hosts at the end of 1998. Last year Telenor Nextel, which 2.2 Access competition Which deployment strategy should be has 70 % of this market (225,000 sub- in Norway chosen in order to balance investment scribers), was among the first Internet risk and expected market potential? Competition in the fixed network in Service Providers to start using the H.323 Norway was introduced in 1998, with Which access node configuration standard. the main service elements being: should the operator establish, and which technology is best suited? Norway is one of the two countries in Establishment of interconnect agree- the world with the highest penetration of ments since the start of 1998; GSM cellular phones (47.6 %), second 1.4 Organisation of the paper only to Finland (58.5 %) at the end of Prefix traffic since the start of 1998; The rest of the paper is organised as fol- 1998. There are two cellular operators in Carrier preselection from June 1999; lows: In chapter 2 the current market sit- Norway today, NetCom and Telenor. As Introduction of a fixed access product uation in Norway is examined, with the of April this year NetCom, which cur- in 1999. emphasis on the competitive situation rently only offers GSM, had 570,000 and the market drivers. Chapter 3 pre- subscribers, whilst Telenor had around In the access network Telenor is facing sents Telenors current network platform, 1,450,000 GSM subscribers. GSM 1800 competition from four different types of which constitutes the starting point for was introduced last year, and already actors: developing the future infrastructure. In 10 % of the mobile traffic in the larger chapter 4 the access technologies are dis- cities is flowing in the GSM 1800 net- Cable operators; cussed. Migration aspects related to the work. Telenor also offers NMT 450 Mobile operators; evolution towards the future access net- (Nordic Mobile Telephone) and NMT work are treated in chapter 5. Experi- 900, bringing the total number of cellular Power utilities; ences and lessons learned from field tri- subscribers to 1,700,000 for Telenor. Operators without infrastructure. als in Norway are presented in chapter 6. Thus in Norway in total, including GSM In chapter 7 key findings from economic and NMT, there are around 2.3 million The coaxial cable network of the largest analyses of broadband access are sum- cellular subscribers as of April this year. cable operator in Norway, Janco Multi- marised. Chapter 8 presents Telenors com, passes over 500,000 homes, mainly strategy, based on the discussions in the The telecommunications network of in the larger cities like Oslo. Janco Multi- preceding chapters. Telenor was fully digitised in 1997, com has 47 % of the cable market in twelve years after the first digital Norway. United Pan-Europe Communi- switches were installed in the network. 2 The market Telenor is now the worlds number one cations (UPC), the owner of Janco Multi- com, plans to invest 1.5 billion NOK (0.2 ISDN operator, with around 300,000 Norway has one of the worlds most billion USD) in network upgrades over ISDN basic access subscriptions by the advanced telecommunication and infor- the next years, preparing the network end of 1998. That corresponds to an mation technology markets, with among for interactive and emerging multimedia ISDN B-channel penetration of 15 %. the highest market penetration of ISDN, services. Janco Multicom already offers The growth in ISDN subscriptions in GSM (Global System for Mobile com- Internet access, with telephone services Norway is still very high. In total Telenor munication) and Internet. The Norwegian expected to be launched during the first has 2.5 million POTS and ISDN sub- market consists of 4.5 million inhabi- half of 1999. scriptions. tants, 1.8 million households and 0.2 mil- lion companies. In this chapter we will The second mobile operator in Norway, Telenor Satellite is number three of give a summary of the telecommunica- NetCom, has 30 % of the GSM market, Europes satellite operators with an tions market in Norway, including and has thus already an alternative access infrastructure consisting of satellite Telenors current position, the competi- network infrastructure. The owner of capacity and land earth stations. The tive arena, the current market trends and NetCom, NetCom Systems, has already services offered are: drivers and the emerging broadband ser- tested 42 GHz broadband radio access vice market. Mobile satellite services; solutions. Establishing such radio sys- Satellite network services; tems by the use of NetComs existing Satellite distribution services. 192 Telektronikk 2/3.1999

194 GSM sites has the potential of substantial 2,000,000 infrastructure cost savings. The power utility companies and their regional telecommunications subsidiaries ISDN have entered the long-distance and city carrier market with significant financial strength and ambitious plans. Currently Subscribers they are targeting the business customer segment, and have taken significant mar- ket shares in some major cities. In addi- tion, power line cable modems are being POTS tested in the field. This technology may enable the power utility companies in Norway to enter the residential market, targeting Internet services. Tele2 is the major actor of the competi- tors without access infrastructure, offer- ing prefix traffic and satellite Internet access, with terrestrial twisted pair 1994 1995 1996 1997 1998 1999 2000 return. NetCom Systems has a large stake in Tele2, in addition to the ownership in Figure 2 The ISDN-BRA forecast for the residential market in Norway the mobile operator NetCom. Tele2 has seized a significant amount of customers in the business segment, and is attacking the residential market mainly with Inter- net and prefix traffic. The operator is now the second largest ISP in Norway. Transparent, non-voice ISDN 64 kbit/s enterprise), the SOHO (small office, However, Tele2 has so far not announced traffic; home office) and residential market. This any plans to install their own access net- is also supported by two different Delphi Internet; work infrastructure. surveys on future broadband service evo- Leased lines; lution, carried out by Telenor in co-oper- In summary, Telenors competitors look ation with other actors [8, 9]. The experts Frame Relay and ATM (asynchronous capable of competing in the access net- participating in the Delphi survey used transfer mode) services offered on the work, in terms of having the technologi- information about market drivers as a broadband platform. cal means, financial strength and deter- basis for their evaluation of the evolution mination. of a future broadband market. The main The current market trends indicate an drivers behind the broadband evolution evolution towards broadband services, are identified as: 2.3 Current market trends in the SME-market (small and medium and drivers ISDN was commercially introduced in the Norwegian business market in May 1994. Two years later, ISDN-BRA (Basic Rate Access) was introduced in 50.0 the residential market. At present the 2-4 Mbit/s Percentage of the residential market 45.0 ISDN market is growing very fast (see 25 Mbit/s downstream chapter 2.1). The ISDN forecast for the 40.0 and 384 kbit/s upstream residential market in Norway is shown 35.0 25 Mbit/s downstream in Figure 2. and 6 Mbit/s upstream 30.0 The current telecommunications market 25.0 is characterised by: 20.0 Data market growth; 15.0 Stagnation and levelling of voice 10.0 traffic; 5.0 A very strong growth in ISDN 64 kbit/s and Internet traffic. 0.0 2000 2005 2010 2015 Saturation Year The main drivers behind the growth in the data market are: Figure 3 Forecasts of broadband accesses Telektronikk 2/3.1999 193

195 Development of PC terminals; Price evolution of PC processor capacity; Price evolution of RAM and disk capacity; Internet development; 50 - 300 m Development of applications; Local 100 - 500 m Development of telecommuting and exchange home office solutions; Development of new technology, eg. 1 - 3 km Residential/SME digital subscriber line (DSL) modems. Figure 3 depicts the forecast results for Copper cable broadband accesses from the European Delphi survey [8]. Primary distribution point The results from the Delphi survey show that a substantial demand for broadband services can be expected in the residential and SOHO market during the next ten Fibre years. The survey shows that the house- Secondary holds incremental willingness to pay for Large distribution point additional broadband applications and business additional capacity is limited. Thus, the future interactive broadband arena, and in Figure 4 Telenors existing twisted pair network particular the residential market, is still characterised by a high degree of uncer- tainty with respect to both service take rates and willingness to pay. This imposes a particular challenge for the access network operators. 3 Existing networks Telenor has today a versatile access network platform which constitutes the starting point for developing the future infrastructure. The current access net- works of Telenor consist of: PDH/SDH-RL A twisted pair network for PSTN (pub- GSM/NMT lic switched telephone network), ISDN and leased lines services (< 2 Mbit/s); 2Mbit/s 2Mbit/s 34Mbit/s 140Mbit/s Optical fibre network for leased line Long distance SDH 155Mbit/s services (> 2 Mbit/s); SDH network ADM DXC SDH To PDH DXC SDH ADM GSM/NMT Coaxial cable networks for distribution of analogue TV-signals; TEII nx2Mbit/s TEII AXE/ 2Mbit/s Cellular networks (NMT-450, NMT- S12 AXE/ S12 CON 900 and GSM 900/1800) for mobile Regional network services; nx2Mbit/s AXE/ S12 SDH SDH SDH Satellite network, 1 West for digital ADM nx2Mbit/s ADM ADM and analogue TV-signal distribution Local network nx2Mbit/s and business communications; 2Mbit/s CON 34Mbit/s nx2Mbit/s CON ADSL (Asymmetric Digital Subscriber PDH-RL GSM/NMT SDH Line) over existing twisted pair cables. ADM To PDH Figure 4 depicts the existing twisted pair network of Telenor. Today there are Figure 5 The existing transport network of Telenor around 230 local exchanges in Telenors 194 Telektronikk 2/3.1999

196 network and approximately 3,500 remote Within each of the six groups of tech- subscriber units or systems (RSU/RSS). nologies there are several different Coaxial Digital The number of primary distribution options, resulting in a very high number cable modem points is about 50,000. Only the business of access network alternatives. This Subscriber Line customers with the highest demand are underlines the fact that we are now wit- Fibre connected by dedicated fibre optic cable nessing the end of a single access infra- in the loop to the local exchange. structure. The future multiservice access network will most likely be built on a The local exchanges are connected to the combination of systems and technologies Terrestrial Stratospheric Satellite transport network, which comprises local all supporting the vision of a full ser- radio platforms systems and regional networks as well as the long vice access network. distance network. The transport network nodes are typically connected by fibre optic transmission systems with capaci- 5 Migration ties in the range 2 Mbit/s to 2,488 Gbit/s. Figure 6 Access network alternatives Given the previously described large In the transport network both the long number of available access network distance network, the regional network architecture alternatives, the challenge and parts of the local networks have for an operator such as Telenor is to been upgraded with fibre optic SDH determine not only the target architecture (Synchronous Digital Hierarchy) ring for future broadband access delivery but structures. Wavelength division multi- also the migration towards this target plexing systems have already been opera- architecture. Thus, access network picture illustrates the fact that as mobile tional for some time on selected links of migration towards broadband is largely broadband in the range of some 100 kbit/s the transport network. Figure 5 depicts related to the existing network, the target emerges through the adoption of GPRS the existing transport network of Telenor. architecture, and the corresponding inter- (General Packet Radio Service) and mediate infrastructure changes required UMTS, even capacities higher than Approximately 5,000 business customers in order to upgrade the network to the 2 Mbit/s will be offered by the use of are connected to Telenors ATM plat- target architecture. For an operator like ADSL and fibre in combination with form, either with a Frame Relay connec- Telenor it is natural to focus on the evo- VDSL or LMDS. tion or with an ATM connection. The lution of the almost ubiquitous twisted access to the ATM network is provided pair copper access and mobile networks. on 64 kbit/s, n * 2 Mbit/s, 34 Mbit/s or The different alternative migration paths 155 Mbit/s leased line connections. The 2) FTTN: Fibre to the node. are discussed in more detail in [10]. Fig- network has today approximately 130 ure 72) shows the expected concurrent FTTC: Fibre to the curb. nodes, of which 60 are ATM switches. evolution over the coming years in terms LMDS: Local Multipoint Distribution Telenors IP network consists of around of terminal mobility and capacity. The System. 35 access nodes or POPs (Point of Pres- ence). 4 Access technologies Telenor has examined a large variety of access network architectures in order to determine the most appropriate ones for Mobility the different geographical area types and service demand profiles. The objective is to derive suitable minimum-risk strate- gies for a migration of existing network infrastructures or for deployment of a GSM completely new access network infra- GPRS UMTS structure, in view of the underlying fun- damentals of where and when to invest in order to create a positive business case for broadband services. We are facing paradigm shifts along three dimensions, as listed in section 1.2, FTTN/C that change the architectural require- PSTN ISDN ADSL VDSL/ ments of the access network. The tech- LMDS nology variety is illustrated in Figure 6, in which some of the relevant technolo- Capacity gies are sorted by transmission medium. Figure 7 Development of broadband and mobility Telektronikk 2/3.1999 195

197 EURESCOM3) project P614, Imple- BIA - Broadband Internet Access mentation Strategies for Advanced Access Networks largely reflects the BA - Broadband Access main categories of economic analysis Internet results on which access network strate- Service Transport network gies may be based [7, 11]. The methodol- Providers ADSL over copper ogy and tool initially developed by the RACE4) 2087/TITAN5) and the ACTS6) ISP 1 226 OPTIMUM7) projects, and now Standard under further development in the ACTS BAP 364 TERA8) project has been applied in Leased lines SDH/PDH/ATM the techno-economic analysis. Upgrade strategies based on a combination of ISP 2 optical fibre systems and Digital Sub- BAP: Broadband scriber Line (DSL) technologies are Access Point examined for various residential areas ATM with existing twisted pair infrastructure. VoD Figure 9 shows the line cost breakdown at 25 % take rate in the urban area under server study. Results are shown for 2 Mbit/s, 8 Mbit/s, 13 Mbit/s and 26 Mbit/s roll- outs, corresponding to different degrees Figure 8 The ADSL field trial configuration of fibre penetration in the network. A ten year (1998 2007) upgrade project with a linear penetration increase is examined. These overall investment cost analyses confirm several key cost features of the future broadband access networks: To measure statistics for the use of The cost of bandwidth in the access 6 Field trials various services; network is still very significant. The analyses confirm that a broadband Telenor has had two broadband service To test various selections of bandwidth upgrade of the access network has pilots operational since spring 1998, in to customers; costs similar to or higher than the over- which broadband Internet access using To test supply procedures and the all costs of establishing the existing ADSL technology has been tested. The administrative system; access network. trial has been run in central Oslo, with approximately 50 SMEs and residential To test operational requirements and The node configuration and corre- customers connected with capacities of procedures; sponding degree of fibre penetration up to 2 Mbit/s downstream and 448 kbit/s is a key strategic decision. To test support requirements/customer upstream. Two different ISP accesses service and procedures. High capacity upgrades and corre- have been tested: IP based access and sponding high fibre penetration sig- access based on IP over ATM. The cus- Figure 8 depicts the ADSL field trial nificantly increase the financial risk. tomers have been offered the broadband configuration. BIA (Broadband Internet Internet access service in addition to their Access) is the product sold to the end- existing ISDN or POTS service on the user by the ISP (Internet Service Pro- same twisted pair. The main objectives vider), whilst BA (Broadband Access) of the trial project have been: 3) EURESCOM: European institute for is the product sold to the ISP by Telenor To test the technical functionality of Network. The BAP (Broadband Access research and strategic studies in existing and forthcoming equipment; Point) is described in more detail in sec- telecommunications. tion 8.2. 4) RACE: Research in Advanced Commu- To test the quality and performance of access lines; nications in Europe. The BAP is a key element in the network To test the provision of various ser- architecture, as it is flexible and provides 5) TITAN: Tool for Introduction scenario vices through the same cable in terms interconnection to both the IP network, and Techno-economic evaluation of of crossed lines/interference (ADSL, ATM networks and other networks. Access Networks. ISDN, HDSL (high speed digital sub- 6) ACTS: Advanced Communications scriber line), etc.); 7 Economics Technologies and Services. To measure traffic statistics, including 7) OPTIMUM: Optimised network archi- Norwegian traffic patterns, traffic with In devising an access network strategy tectures for multimedia services. domestic, international and local pro- Telenor has examined the economics of viders; various upgrade alternatives. The techno- 8) TERA: Techno-Economic Results from economics work performed in the ACTS. 196 Telektronikk 2/3.1999

198 The analyses of the economics of broad- 2,000 band access delivery indicate that Tele- nor as an operator is faced with a set of 1,800 Cost per home connected (euro) Service specific cost strategic imperatives which need to be 1,600 reflected in order to achieve economics Basic service cost of scope and scale when rolling out a 1,400 Cable infrastructure broadband access network. 1,200 1,000 8 Telenors strategy 800 In view of the preceding chapters and 600 considerations, Telenors plans for access 400 network evolution can be formulated, both with respect to access service port- 200 folio, architecture reference model, target 0 network and deployment plan. The target 2 Mbit/s 8 Mbit/s 13 Mbit/s 26 Mbit/s network, described in chapter 8.3 is based on the architecture reference model Capacity provided described initially in chapter 8.2. In the final chapter 8.4, the deployment plan of Figure 9 Line cost breakdown @ 25 % take rate Telenor is presented. 8.1 Telenors access service portfolio One possible evolution of Telenors access service portfolio, including fixed, mobile and satellite services, is illus- trated in Table 1. The table lists different access services, indicating the underlying technology, the access speed offered and the time period during which the service is likely to be available. Table 1 Telenors access service portfolio 8.2 Architecture reference Technology Access speed Time period model Fixed PSTN 2.4 kbit/s 56 kbit/s 1986 1997 The new access network reference model is depicted in Figure 10. ISDN 144 kbit/s, 2 Mbit/s 1994 Digital (leased line) 64 kbit/s 155 Mbit/s 1986 1997 The trend towards open provisioning means that different administrative Cable modem 512 kbit/s 1998 domains will be involved, as marked by ADSL 2 Mbit/s (448 kbit/s) 1999 vertical dashed lines. The model distin- guishes between access operators, trans- Fibre + ADSL/VDSL 25 Mbit/s (asymmetric) 2002 port providers, service providers and LMDS 34 Mbit/s (8 Mbit/s) 2000 2001 platform operators. As suggested by the figure, cross-domain management is an DTTV 30 40 Mbit/s 1999 important issue for the future broadband Mobile GSM 9.6 kbit/s 1992 access network. HSCDS 28.3 kbit/s 1999 The Service Nodes (SN) of the future GPRS 115 kbit/s 2000 broadband architecture access network is shown to the left in the figure. The term UTRA (UMTS) 144 kbit/s 2002 Local Access Point (LAP) is used for the 384 kbit/s 2002 natural aggregation point covering the 2 Mbit/s 2002 customers within a geographical area of Satellite VSAT 2 Mbit/s 1976 moderate size. The Broadband Access Point (BAP) links a number of local DTH analogue --- 1987 areas on one side to one or more service DTH digital 40 Mbit/s 1997 nodes on the other side. A LAP is linked to only one BAP whereas an SN can be Turbo Internet 2 8 Mbit/s 1999 connected to several BAPs. Hence, a Telektronikk 2/3.1999 197

199 BAP defines the termination of a local determine the number of LAPs and BAPs picted in Figure 11. Basically it consists access network with the corresponding in a nationwide access network. of ADSL modems installed on the exist- interface to the service nodes denoted ing copper base, ADSL multiplexers and SNI. The model facilitates the use of new broadband access points. intermediate transport providers on both 8.3 Target network sides of the BAP. The future access network platform of The long term target architecture is illus- Telenor will be developed based on a co- trated in Figure 12, consisting of fibre The physical location of a BAP depends ordinated evolution of the fixed access optic nodes with fibre optic transmission on the customer base in the connected network and the mobile network. The systems in the primary access network LAP areas, and also the location of the data rate in the mobile network will be and VDSL modems or LMDS installed SNs. One extreme case is to co-locate the increased over the next years by the between the fibre node and the customer BAP with one of the connected LAPs. introduction of GPRS, before UMTS premises (secondary access network). The other extreme is to co-locate a BAP eventually is launched in year 2002. The fibre node structure may also serve with an SN. In practice, the optimal loca- the future UMTS base stations with the tion will be determined by a trade-off of In the initial phase the fixed network required transmission capacity. The fibre the transport costs carried by the access upgrade will mainly consist of a rapid node size is probably a more important operator and the service providers, ISDN roll-out. The medium term target choice than the technology choice itself. respectively. The same trade-off will also network for broadband services is de- Service Transport Access Transport/Access Access Private domain domain domain domain domain domain (null) (null) Management Existing PSTN, ATM, IP SN NT TE RSU BAP NT SN LAP NT SN NT BAP LAP SN NT NT LAP ? ATM ? Net element Housing Aggregation/concentration ATM assumption No protocol assumption Figure 10 An architectural reference model for the future broadband access network 198 Telektronikk 2/3.1999

200 The fibre node structure will be serving all customer groups, even if initially the SME segment is expected to be the domi- Residential/SME nating one. In the medium to long term ~ 2 Mbit/s also the SOHO and residential market ADSL will be connected by high capacity accesses to the fibre node structure. Gateway The aim is also to establish a service ISDN 50 - 300 m integrating access network, evolving PSTN ATM 100 - 500 m from an ATM based network in the ini- tial phase to an IP based network in the IP longer term. A likely evolution is that the broadband access network initially will 1 - 3 km Residential/SME be provision oriented, similar to the cur- ~ 2 Mbit/s rent narrowband network. The next step ADSL will be to implement free selection of To Copper cable service provider per session. The ulti- SN ADSL Primary mate step is to let the access network BAP distribution point become a true IP subnet interworking closely with arbitrary subnets in the core domain [12]. BAP: Broadband Access Point Fibre 8.4 Deployment plan Secondary Large distribution point Telenor has established the following business ADSL deployment plan for the future broadband access network: Figure 11 Telenors medium term target network architecture 1. Telenor will meet the market demand by targeting ISDN to the mass market in the initial phase. 2. In June 1999 a commercial ADSL ser- vice was rolled-out, providing 12 Mbit/s ADSL to the SME market. 3. Within few years a fibre node structure will be rolled out, aimed at providing capacities up to 26 Mbit/s to SMEs, SOHO and high end users. Gateway DSL 4. The high density areas are the natural ISDN targets in the initial phase, whilst the PSTN mass market is likely to follow when ATM the applications and the price level for IP Residential and broadband access equipment is right. GSM business 2 Mbit/s NMT UMTS 9 References LAP UMTS-base 1 Telenor Press Centre (1999, June 11) [on-line]. URL http://www.telenor. no/press_center/aktuell. To SN LMDS-base 2 Total Telecom (1999, June 14) [on- Fibre node line]. URL http://www.totaltele. BAP com. Residential/SME 2 - 26 Mbit/s 3 Ims, L A et al. Telenors strategy Fibre LAP towards the future broadband access network. In: Proc. BT Alliance Engi- HF neering Symposium, Oslo, Norway, 1317 June, 1999. Large business DSL Figure 12 Telenors long term target network architecture Telektronikk 2/3.1999 199

201 4 Hjelkrem, C. Forecasting Demand 7 Ims, L A et al. Economics of broad- 10 Ims, L A. Wireline broadband access for ISDN accesses when information band access network upgrade strate- networks. Telektronikk, 95 (2/3), is scarce : a case study from the Nor- gies : the European perspective. In: 7387, 1999. (This issue.) wegian market. In: Proc. Business Proc. Access Network Mini Confer- Forecasting : Telecommunications ence, Globecom 98, Sydney, Aus- 11 Ims, L A, Myhre, D, Olsen, B T. and Gas/Electric Conference, Febru- tralia, 812 Nov 1998. Costs of upgrading the residential ary 1999, Las Vegas, USA. telecommunications infrastructure to 8 Stordahl, K, Rand, L. Long term broadband. In: Proc. Globecom 98, 5 Nilsen, F B, Ims, L A. Interworking forecasts for broadband demand. Sydney, Australia, 812 Nov 1998. in the future broadband access net- Telektronikk, 95 (2/3), 3444, 1999. work a question about IP and (This issue.) 12 Nilsen, F B, Ims, L A. Interworking ATM. European Network Planning, in the future broadband access net- 1519 March 1999, Les Arc, France. 9 Istad, S, Stordahl, K. Broadband work a question about IP and demand survey in the residential and ATM. In: Proc. BT Alliance Engi- 6 Ims, L A et al. Towards a flexible SOHO market in Norway. Telektro- neering Symposium, Oslo, Norway, and competitive access network : the nikk, 95 (2/3), 4549, 1999. (This 1317 June 1999. Telenor strategy. In: Proc. Access issue.) Technologies 99 (Vision in Business), 911 March 1999, London, UK. Leif Aarthun Ims (33) is Senior Research Scientist in the Strategic Network Development group of Telenor R&D. Since 1994 he has been working on access network development in several RACE, ACTS and EURESCOM projects, and recently as project manager for long term access network development in Telenor. He is now project leader of the EURESCOM project P901. He is editor of the book Broadband Access Networks published last year and has published more than 50 papers in international journals and conferences. email: [email protected] Trygve Jarholt holds a bachelor degree in Technology, department of telecommunication. He has worked in Telenor AS for 11 years, the first years with planning data networks. From 1994 to 1996 he was manager of opera- tions for the access network in Telenor, and since 1997 he is been manager for development of new broadband solu- tions. email: [email protected] Kjell Stordahl (54) received his M.Sc. degree in statistics at the University of Oslo in 1972. He worked with Telenor Re- search Dept. for 15 years, seven of which as manager for the teletraffic field. He joined Telenor Nett in 1989 and was manager in the Planning Department until 1996. From 1997 to 1999 he was manager for Market analysis in the Market Division, Telenor Nett. He has participated in various Euro- pean projects and authored or co-authored more than 90 papers in international journals and conferences. email: [email protected] 200 Telektronikk 2/3.1999

202 Frode B. Nilsen (32) holds a Masters degree (1993) and a Doctoral degree (1998) in computer engineering from the University of Oslo. This research focused on methods for performance evaluation of high-speed networks. He is currently employed at Telenor R&D as research scientist working on networks. email: [email protected] Markku Lhteenoja (42) is Research Scientist at Telenor R&D. He is working in strategic network development with a special interest in techno-economics. Before joining Telenor he was with Telecom Finland / Sonera from 1984 till 1998 working in optical communication, broadband net- works, access networks and techno-economics. email: [email protected] Borgar T. Olsen (52) graduated with an M.Sc. from the Uni. of Oslo in 1975 and a Dr.Philos. in 1987. Olsen has worked as Scientific Asst. at Inst. of Physics, Uni. of Oslo and Research Fellow at Max-Planck-Inst. in Germany. Joined Norwegian Telecom Research in 1986, working with optical switching, and 19891992 responsible for the Component Technology group . Joined the Network Structures Group in 1992, working with studies on introduction of new technol- ogy in the access network. Project leader of ACTS AC226/OPTIMUM and AC364/TERA . email: [email protected] Dagfinn Myhre (37) is Senior Research Scientist and Head of the Strategic Network Development group at Telenor R&D, Kjeller, where he has been employed since 1986. He has been involved in several national and international projects focusing on techno-economic and strategic ana- lyses of future communication networks. email: [email protected] Stig Lken studied at the Norwegian University of Science and Technology, department of telecommunications. He has worked with Norsk Elektro Optikk AS with the develop- ment of measurement instruments based on diode laser spectroscopy. Since 1995 he has been working with Telenor Nett AS with the development of new broadband access solutions. email: [email protected] Telektronikk 2/3.1999 201

203 The challenge of civil works GIOVANNI CIOCHETTO Introduction Another interesting alternative to digging Nevertheless, when working with this is represented by Micro Cabling System, kind of documentation it is often neces- Civil works and installation are respon- which allows installation of small optical sary to distinguish between planning sible for more than half of the total cost cables in shallow and narrow trenches and executive drawings; the latter may of the network and have a significant along roads or pavements. sometimes be reconstructed. impact on environmental and social activities. The use of new techniques As far as cable installation is concerned A direct on-site survey is therefore nec- allows a reduction of the global costs interesting results can be obtained by essary in order to locate the exact posi- related to the deployment of the access using air blowing techniques. One of tion of the existing underground utilities, network and an improvement of the flex- the most interesting points is represented in particular the most hazardous ones (eg. ibility of the installation techniques, by the new techniques for optical fibres gas, power, high density TLC cables) and reducing significantly the environmental installation. Using appropriate materials to produce updated maps. and social impact related to civil work and equipment it is possible to install sin- and installation. gle fibres or fibre bundles directly into The most popular method for locating small underground ducts by blowing or underground utilities is definitely Ground This paper presents an overview of cur- by pulling. Penetrating Radar (GPR), due to its rent practice and emerging techniques. rapidity of execution, good quality results, and capacity to supply 2D images Improvement in site investigation using Ground Penetrating Radar of the subsoil. Ground Penetrating Radar systems can An exhaustive knowledge on the subsoil avoid drawbacks of digging or drilling An electromagnetic wave is transmitted right from the first design phases of the works for the construction of under- by a planar antenna into the ground and construction of underground network ground infrastructures. the returned scattered radiation is re- infrastructures is essential, both to limit ceived by another antenna and then possible damage to existing utilities and to New construction methods alternative to processed to extract the information reduce the number of failures especially if traditional digging techniques are also relevant to the buried objects. In general new trenchless techniques are used. presented. The so-called No Dig tech- any dielectric discontinuity is detected. niques allow installation of underground Targets can be classified according to The information needed can be obtained ducts performing small tunnels without their geometry: planar interfaces, long partially by consulting existing documen- digging long and deep trenches which, and thin objects, localised spherical or tation on any work (eg. laying of utilities, especially in urban environments, cause cuboidal objects. etc.) carried out in the area in question, many inconveniences to cars and pedes- or through documentation made available trians. The time-domain impulse radar system by local authorities and other companies has become the most widely used com- (eg. gas, water, power, etc.). mercial system for this scope and is available commercially. Manufacturers usually offer a range of antennas to suit the desired probing range. Depth of less than 3 metres can be investigated using wide-band antennas with a centre fre- quency of 200500 MHz. In order to reach greater depths it is necessary to use lower frequency antennas, with the draw- back of reducing the resolution and the precision of the probing. Most antennas have relatively small foot- prints which means that rapid and wide- area surveying can only be achieved with multichannel radar systems. These systems use more than one antenna mounted in a fixed scheme; it allows the acquisition of a large amount of data in a relatively short time which facilitates the final interpretation of the probing results. These systems are mainly used for identi- fying utilities and obstacles, although they have also recently been applied for lithological recognition of the subsoil. More recently, a new conception of radar systems have been developed to improve Figure 1 On site acquisition system equipped with a multi-antenna array the detection of underground utility and (Courtesy of IDS Ingegneria dei Sistemi S.p.A., Italy) to automate the production of the final 202 Telektronikk 2/3.1999

204 utility map. Multi-antenna and multi- channel arrays, powerful post-processing tools, direct connection with CAD and GIS systems are the main features of these equipment. The array architecture (see Figure 1) allows performance of a three-dimen- (1) (2) sional survey of the ground (see Figure 2) acquiring more channels, that is more information, at the same time. The post processing software helps the operator to interpret the radar diagrams by producing groups of parallel radar (vertical) sec- tions (see Figure 3), where targets are represented in a deformed geometry and Figure 2 Comparison between: planimetric radar (horizontal) view of (1) 2D (single antenna) survey slices of subsoil (see Figure 4), where (2) 3D (array of antennas) survey the real target geometry has been recon- structed. Finally, the post-processing software may provide a link with a CAD station which allows direct transfer on a digital map of the information relevant to the position and depth of the detected under- ground utilities, that means considerable time saving and a more functional use of the data. No dig techniques The use of no dig techniques in the field of telecommunications was for a long time limited to the execution of railway-, road- and motorway crossings as the only possible solution for the construction of network infrastructures across such obstacles. Starting in the late 1980s, the use of no dig techniques has grown rapidly and has in some cases become a valid alternative to traditional digging techniques for the construction of longitudinal TLC net- Figure 3 Example of multiple radar section representation work infrastructures along roads. The great advantage compared to digging techniques, is represented by the reduc- tion of social costs, such as traffic delay, hazard for pedestrians, noise, pollution, and road damage, which are normally related to road works. In fact, in using no dig techniques the ducts are installed underground without digging long and deep trenches, but by performing small tunnels at a fixed depth in which the duct is subsequently pulled. In this way little damage is caused to the road surface, that means less inconve- nience to traffic and a considerable cost saving on road repairs. Figure 4 Example of planimetric view of a pipe Telektronikk 2/3.1999 203

205 Most, but not all, guided boring ma- chines use a fluid-assisted drill head which is pushed through the ground at the end of a string of drill pipes. The drill head is usually angled, so that constant rotation of the drill string produces a Drill pipe straight bore, whereas keeping the head in one position causes the line to deviate. A sonde or beacon is usually built into Guided boring Existing services the head or fixed close to it and signals machine Reception emitted by this are picked up and traced Bore path pit by a receiver on the surface, so allowing the direction, depth, and other parameters to be monitored. Hard-wire guidance sys- Figure 5 General scheme of the directional drilling technique: drilling the pilot hole tems are also used, with the cable run- ning through the drill string, particularly in cases where the bore path cannot readily be traced on the surface (across rivers, for example) or where the depth of the bore is too great for accurate loca- tion by the radio-frequency methods. There are also location systems which use magnetometers. Pipe coil A bentonite/water mix is often used as the drilling fluid or mud, which carries the debris in suspension and may be fil- tered through a recirculation system. On completion of the pilot bore, the thixo- Guided boring tropic mud stabilises the hole ready for machine back-reaming. The service pipe or duct, Product pipe generally polyethylene or steel, is drawn Reamer fitted in behind the reamer as the original bore to drill string is enlarged. Figure 6 General scheme of the directional drilling technique: backreaming and pulling In the case of larger machines, much of in the product pipe the work is done by the rotation of the drill string and the torque of the unit is as vital as the axial thrust and pull-back. As with smaller rigs, it is normal practice to drill a smaller pilot hole and then back-ream to the required diameter while On the other hand, no dig techniques are gradually curved and the direction of the pulling in the conduit behind the reamer, strictly related to the nature of the soil, drilling head can be adjusted at any stage using a drilling fluid to assist the cutting so that in some circumstances and with during the bore to steer around obstacles operation and to lubricate and cool the some machines it is impossible to per- or under highways, rivers or railways. cutting head. The fluid may also power a form the drilling successfully. In any Drilling can be carried out between pre- down-hole mud motor for cutting rock case it is necessary to make preventive excavated launch and reception pits, or and other hard formations, in which case investigation of the subsoil to know the from the surface by setting the machine higher fluid flow rates are necessary. location of existing utilities or buried to drill into the ground at a narrow angle. obstacles in order to avoid them during The use of new small machines allows Some systems are designed for dry oper- the drilling operation. the drilling operation to be performed ation without the use of water drilling directly between existing chambers or fluids. These are simpler to operate, Different techniques allow installation manholes, thus considerably reducing create less mess and do not require as underground ducts without digging; in the overall dimension of the working site much on-site equipment, but there may the following paragraphs a brief descrip- and the consequent impact on traffic. be restrictions on the sizes that can be tion of these techniques is reported. installed and on the ground conditions Installation of the product pipe or duct that the machines can cope with. is usually a two-stage operation. A pilot Directional drilling hole is first drilled along the required An increasingly common feature is the Guided boring and directional drilling path (see Figure 5) and the bore is then use of percussive action to complement techniques are used for the trenchless back-reamed to a larger diameter to axial force and rotation. This can be installation of new pipelines, ducts and accommodate the product pipe (see Fig- achieved either with a percussive cables. The drill path may be straight or ure 6). hammer at the bore-head, or by generat- 204 Telektronikk 2/3.1999

206 ing the percussion at the machine on the Although hydraulically driven percussive tools, soil displacement hammers, impact surface and transmitting it along the drill moles are available, most are powered by hammers, percussive moles or pneumatic string. Either way, this can significantly compressed air. A potential drawback of moles, depending on the term used by the improve the ability of guided boring air-driven moles is contamination of the manufacturer and the region of the world machines to punch through difficult product pipe by lubricating oil in the where the equipment is being used. ground or hard inclusions. Dry direc- exhaust, although there are methods for tional drilling machines, due to the small overcoming this. Hydraulic moles require Because impact moling is generally size of the rig and the absence of drilling two hoses (flow and return) and tend to unsteered, the technique is most suitable fluid, are probably the best solution for have greater mechanical complexity. for short bores (up to 50 metres). A no digging in urban environments straight bore can often be maintained The basic mechanism of impact moling more easily at large diameters. Diameters A typical mid-range, surface-launched is the reciprocating action of the pneu- range from about 45 to 200 mm depend- guided boring machine would, depending matically or hydraulically powered ing on the pipe or cable being installed. on ground conditions, generally be hammer within the cylindrical steel body. capable of installing pipes of approxi- The piston is driven forward and, on mately 250500 mm diameter over dis- striking the forward end of the unit, Pipejacking and tances of between 100 and 350 metres. imparts its kinetic energy to the body microtunnelling which is driven forward. The energy of The largest directional drilling rigs are the piston for the return stroke is regu- Pipejacking and microtunnelling are used primarily for long or large diameter lated so as to reposition it for the next essentially from the same family of crossings under rivers, estuaries, major forward stroke, rather than reversing the pipeline installation techniques, used for highways and long sections. At the other unit out of the bore (unless required to installations from about 150 mm dia- end of the scale compact rigs for use in do so). meter upwards. A pipejack is defined restricted spaces can install pipes of up as a system of directly installing pipes to about 160 mm over distances of up to Repeated impacts of the hammer piston behind a shield machine by hydraulic 100 metres. This again depends on advance the whole unit through the jacking from a drive shaft, such that the ground conditions. Some include the ground. As the forward movement takes pipes form a continuous string in the facility to reduce the track spacing for place, the soil in front of the mole is ground (see Figure 8). The pipes, which passage through narrow openings. forced aside and compacted by the coni- are specially designed to withstand the cal or stepped nose to form the walls of jacking forces likely to be encountered Dry directional drilling systems using a the bore. The power of the unit is also during installation, form the final cone shaped reamer with tungsten-car- often used to pull the product pipe, cable pipeline once the excavation operation is bide cutting teeth connected directly to or cable duct through the bore at the completed. the drilling rods can perform the installa- same time as the impact mole advances. tion of small diameter pipes, ducts or Within this description, microtunnelling cables (up to about 65 mm diameter). Impact moling tools are known by sev- is specifically defined as being a steer- eral other names including earth piercing able remote-controlled shield for install- Dry installation of pipe diameters of up to 250 mm require a pneumatically powered reaming hammer on the drilling head. Impact moling Impact moling, or earth piercing as it is commonly known in North America, is defined as the creation of a bore by the use of a tool which comprises a percus- sive hammer within a suitable cylindrical casing, generally torpedo shaped (see Figure 7). The hammer may be hydraulic or pneumatic. The term is usually asso- ciated with non-steered or limited steer- ing devices without rigid attachment to the launch pit, relying for forward move- Launch pit or ment upon the internal hammer action to chamber Impact mole overcome the frictional resistance of the ground. During operation the soil is dis- placed, not removed. An unsupported New PE pipe bore may be formed in suitable ground, or a pipe may be drawn or pushed in Reception pit immediately behind the impact moling or chamber tool. Cables may also be pulled in. Figure 7 Pipe installation by impact moling Telektronikk 2/3.1999 205

207 Micro Cabling System The purpose of the Micro Cabling Sys- tem, an installation system developed by Siemens, is to dramatically reduce instal- lation costs and installation time. This is achieved by means of a simple but rugged design of a fibre optic cable Launch pit which is laid in a groove dug into the asphalt of roads or pavements. As a Jacking Microtunnelling Reception result the cable is required to meet exact- Hydraulic jacks pipes machine pit ing demands to crush resistance and, in particular, temperature resistance which is needed when sealing the cable in the groove with hot bitumen. Alignment and The optical fibres are preferably enclosed jacking frame Flush-fitting collars in a copper tube filled with a suitable fill- ing compound and surrounded by a PE Figure 8 Installing pipes by microtunnelling jacket. There are currently two cable types in use: type 1 with up to 60 optical fibres and an outside diameter (over PE jacket) of about 7 mm, and type 2 with up to 144 fibres and an outside diameter of about 9 mm. The groove depth is usually 8 to 10 cm, ing a pipejack with internal diameter the width about 1 to 1.2 cm. These smaller than that permissible for man- dimensions are guide values but can eas- entry. Microtunnellers often use a laser ily be adapted to suit other road condi- guidance system to maintain the line and tions, eg. with a thin asphalt cover. level of the installation; though, as with larger pipejacking installations, both The cable itself runs off the reel lightly laser guidance and normal survey tech- braked and is laid in the groove previ- niques can also be utilised. ously cleaned with hot air. A retaining strip is then run in to fix the cable in Systems are available for the installation place inside the groove. of both main pipelines and branch con- nections. The retaining strip is covered by free- running, highly water-repellent filling Both pipejacking and microtunnelling are materials. The groove and groove edges well suited to situations where a pipeline are sprayed with primer (bonding agent) has to conform to a rigid line and level and then sealed with hot bitumen (see criterion. The guidance and control sys- Figure 9). tems allow accurate installation within close limits of the target. Where the cable makes a sharp change in direction or closures have to be installed, Most microtunnelling drives are straight holes are made by core drilling. The between shafts, although specialised sys- holes thus formed are such that they are tems are available for curved drives. able to accept the circular closure or can Where line-of-sight is not possible be- be used to change the direction of the tween the drive shaft and the micro- cable without exceeding the smallest tunnelling machine because of the cur- permissible bend radius. vature of the tunnel, specific alignment systems (eg. gyroscopic devices, com- For jointing or branching the Micro binations of electromagnetic induction Cable there are suitable closures avail- and liquid pressure difference) may be able which are able to accommodate up used as an alternative to the usual laser to 60 splices or 144 splices. These acces- equipment. Figure 9 Street cross section sories are installed level with the surface of the road or pavement and can be pro- vided with a heavy-duty (vehicle-bear- 206 Telektronikk 2/3.1999

208 ing) cover. Other suitable accessories fibres. A small premium is incurred, in will automatically adjust their speed to such as eg. underground containers and terms of product cost, for all but small match the speed of the master machine. terminating accessories for the end points fibre counts. This eliminates the need for re-coiling are also available. fibre and substantially reduces installa- It is also possible to pull the fibre optic tion time. unit through the tube distances up to Fibre installation 1000 m with a pulling force sufficiently low as not to cause any damage to the Conclusion Blown Fibre consists of a small fibre fibre optics and not impair its long term bundle or ribbon composed of a number Two focal points must be taken into properties. of standard fibres held in a resin matrix account for the future deployment of the and a tube assembly composed of a num- TLC networks: a growing attention for In order to use this technique, special ber of polyethylene tubes within a pro- the environmental impact of civil works tubes are manufactured with pull cords tective sheath. Compressed air is applied and the need for reducing their costs, pre-installed. The pull cords were spe- to the Blown Fibre tube and the fibres are which now represent more than half of cially developed for the application hav- fed into the airflow by a special blowing the total cost of the network. The deploy- ing very low friction characteristics and a head. The Blown Fibre principle relies ment and the use of new techniques for braided construction so that joints on the viscous drag of air over the fibres the construction of TLC network infras- between the fibre unit and the cord could that produces a distributed installation tructures and for the installation of cables be easily achieved. force on the fibres themselves. and fibres therefore represent essential actions to take up the challenge for a sus- Also appropriate pulling machines with Blowing distances of 1000 m were easily tainable development. a continuous speed and tension read out achievable and in certain routes up to must be used, which are combined pull- 1500 m was possible. ing and coiling machines capable of pull- Acknowledgement ing surplus fibre through a length of tube Blown Fibre has several advantages over and re-coiling it into a pan for onward The author would like to mention the conventional optical fibre cables. It pulling. By use of this machine it is pos- great effort that Telecom Italia is doing allows greater installation flexibility and sible to install long continuous lengths of for the development of new installation significant cost deferral. Routes can be fibre; however, one disadvantage is that techniques and for the reduction of social easily reconfigured using push fit con- the re-coiling operation substantially and environmental impact of civil works. nectors and fibre bundles or ribbons can increases the time needed to install the be recovered and replaced with new fibre unit. Special thanks to Mr. Feliciano Esposto, specification or higher fibre count ver- manager of the outside plant engineering sions. Moreover, by installing fibre only To overcome this problem and speed up at the Telecom Italias headquarters in as it is needed, the upfront investment the installation of the fibre slave ma- Rome and to his collaborators Mr. can be minimised whilst still enabling the chines have been developed. These can Andrea Berardo and Mr. Paolo Trom- rapid deployment of further fibre capac- be positioned at an intermediate location betti: The work performed together in the ity. so that fibre could exit the duct at this last years has been essential for writing point and re-enter an adjacent length of this paper. Mechanical strain on fibres during in- tube at zero tension. The slave machines stallation and subsequent use is negli- gible, and long splice free lengths are possible by using range extending blow- ing techniques. The disadvantages are represented by higher stores cost than subducted con- ventional cables for all but small fibre counts and lower potential fibre density than conventional cables. Labour costs could be reduced by using higher fibre count bundles or ribbon and by employing simultaneous installation of more than one bundle. Fibre density Giovanni Ciochetto (34) received his degree in Electronic may also be improved by using higher Engineering from Turin Politecnico in 1991. In 1992 he fibre count bundles and by using a higher joined CSELT, the research centre of Telecom Italia group, tube count. where he is now a senior researcher. He is involved in out- side plant construction as project leader responsible for development of low cost installation techniques. Since 1997 In summary, it can be seen that Blown he has been responsible for Question 12/6 (Trenchless Fibre offers the flexibility to only install techniques for the construction of underground infrastruc- fibre when it is required (thereby re- tures for TLC cable installation) in ITU-T Study Group 6. ducing up-front investment) as well as enabling rapid installation of additional email: [email protected] Telektronikk 2/3.1999 207

209 Measurements and maintenance of the future access networks ANGELANTONIO GNAZZO 1 Introduction of the general FTTx networks: this paper In the FSAN group [3] the definition of reports an analysis of the work per- Optical Maintenance has been stated as: In a liberalised telecommunications formed. Optical Maintenance (OM) covers all market, where the customer can choose the means to guarantee the performances the service provider, the downtime of of the optical carrier (fibres, splices, networks represents a key aspect for the 2 Access network and connectors, passive components), and quality of service for a given operator. maintenance definition specifically to detect and locate any fault To maintain networks with minimal in the OAN (Optical Access Network). downtime, companies need a variety of The access network is commonly defined tools and practices. By combining a mix as the collection of equipment and infras- Starting from these two definitions, of appropriate tools and practices, com- tructures necessary to perform the con- EURESCOM P614 defines Hardware panies are able to minimise downtime, nection between the customer equipment Maintenance as: Hardware Mainte- correct it quickly when it does occur, and and the core network. nance (HM) covers all the means to obtain cost savings in the process. Tools guarantee the performances of the physi- are represented by monitoring and fault Figure 1 [1] shows a variety of ways to cal carrier (fibres, twisted copper pairs, localisation systems. build broadband access networks based coaxial, splices, connectors, passive on fibre deployment. The basic FTTx components, enclosures), and specifically Moreover the importance of operation architectures under study in most coun- to detect and locate any fault in the AN and maintenance aspects has been in- tries are FTTEx, FTTCab, FTTB and (Access Network). creasing in the last few years: the com- FTTH. The most salient feature of the plexity of the network requires an up- FTTx network architecture is its hybrid dating of all infrastructure records in nature: different network segments may 3 Preventive mainte- order to keep track of all changes in the use a different bearer medium. In a nance implementation infrastructure. This would be possible if FTTCab or FTTB network, the optic the NOs (Network Operators) had a cen- fibre feeder and distribution segments are Today acceptance tests are done in two tralised record of all fibre plant with geo- connected to the customer premises by a phases: acceptance of the components graphic information, technical character- drop section based on twisted copper pair (fibre, splitters, filters, etc.) and accept- istics and a well defined program of tests. or coaxial cable. ance of the infrastructure once the com- In this way an investment in network ponents are installed. maintenance tools and technical solutions ITU-T [2] give the following definition could be a real investment, not just an of maintenance: Maintenance involves The first phase is normally done in the expense. the whole of operations required for laboratory, on a representative sample of setting up and maintaining, within pre- the components. Its objective is to verify EURESCOM Project P614 worked on scribed limits, any element entering into if the components characteristics match these items giving some ideas to how the setting-up of a connection. the specification presented by the supplier. implement the maintenance/monitoring This procedure should be seen as a test of supplier control. In order to reduce costs, the responsibility for performance quality can be transferred to the supplier, but for the time being we consider occasional component testing necessary to ensure Service PON Head Local Cabinet Kerb Home quality, even if we trust the supplier. Node End Node Exchange VB5 UNI After the acceptance tests for com- PON VDSL or ADSL ponents have been approved, two ATM OLT ONU NTE FTTEx basic tasks are required to implement an effective preventive maintenance scheme VDSL ATM OLT ONU NTE FTTCab before corrective maintenance: VDSL FTTK/ Network qualification/acceptance; ATM OLT ONU NTE FTTB Network monitoring. FTTB/ ATM OLT ONU FTTH NTE Network qualification intends to estab- lish that the actual plant matches the ADSL - Asymmetric Digital Subscriber NTE - Network Termination Equipment specifications given by the network plan- Line/Loop OLT - Optical Line Termination ner. Absolute value measurements are FTTB - Fibre To The Building ONU - Optical Network Unit necessary to check that each of the net- FTTCab - Fibre To The Cabinet PON - Passive Optical Network work elements conforms to specified FTTEx - Fibre To The Exchange VDSL - Very high-speed Digital performances. Qualification of the physi- FTTH - Fibre To The Home Subscriber Line/Loop FTTK - Fibre To The Kerb cal carrier should be performed before any service runs over it and may be even before any actual equipment and systems Figure 1 FTTx concept (from FSAN) are installed at its ends. 208 Telektronikk 2/3.1999

210 To proceed with network qualification, physical carrier and comparison to a ref- These systems could be integrated into tests may be required to ensure that erence measurement established on the the network management system, in cabling components have been correctly basis of the qualified network. If the link order to allow the correlation of the installed and have not been damaged dur- is already qualified, no absolute measure- transmission alarms and the fibre tests. ing transport or installation. This phase is ment is necessary and only relative ones valid both for fibre based and copper are required. The decision to invest in automatic test based media. systems should be evaluated taking into account the adaptation of the optical A modem connection (GSM or PSTN 4 PON monitoring infrastructure necessary to install the test networks) between the measurement system and the required changes in the Regarding the PON (Passive Optical Net- equipment installed in the access points organisation of the operation and mainte- work) infrastructure, we focused our and the centralised data base should be nance procedures of the operator. The attention, as an example, on an archi- considered. In this way we are able to first one is related to the MDF (Main tecture with 16 branches, and one or keep a record of the processes, not on Distribution Frame) where the changes two splitters. The location of the splitters the local level but as a part of the global adopted are a matter of technical and depends on the FTTx solution and on the infrastructure information. The most economic consideration. The second location of the ONUs. important aspect is the organisation of relies on the maintenance personnel who this information, in order to keep track of should benefit from the new automatic The most favourable solution, from the the technology history to support future test procedure. viewpoint of maintenance efficiency, is investment decisions. The decision to to have all splitters in the central office. record and to organise this information To implement the monitoring of a PON, In this particular implementation, the will influence the quality of the infras- three methods can be envisaged: PON becomes equivalent to simple tructure, and as a consequence the quality point-to-point transmission for mainte- Use of dedicated branched fibre moni- of service provided by the operator. nance purposes but at the expense of toring (configuration 1); increased network fibre cost. For the optical infrastructure acceptance Dark fibre (configuration 2); measurements should include: Optical monitoring should consider pro- On-line monitoring (configuration 3). Link attenuation; cesses that identify problems such as: The dedicated branched fibre monitoring Return loss of splices, connectors and Slow and uniform degradation of the solution shown in Figure 2 can be imple- other components. attenuation of a fibre; mented using one of the splitter ports at Slow increase of attenuation at a dis- each splitting stage for the monitoring Other parameters to be determined, for crete point and localisation of new function and using dedicated fibres to analogue transmission, are chromatic and events in a link (events such as attenu- bring the signals back to the monitoring polarisation dispersion. ation points, reflective points, etc.). head. It should be noted that with this method we are able to monitor the sec- Regarding the twisted copper pairs for In this way we are able to separate: tions of the PON but different fibre xDSL lines, the parameters to be deter- return paths are required. Data signals mined include: Link fault from equipment fault; can also be used as monitoring signal for Attenuation (the main factor limiting Link degradation that does not impact path loss measurements, while for fault cable length); on the service; location an OTDR working at 1625 nm could be used from the fibres connected Crosstalk (unwanted coupling between Immediate fault localisation if optical pairs); monitoring is performed with OTDR. Insulation vs. other conductors and Another important issue is the creation of ground (low insulation increases cross- an automatic process to record the infor- talk); mation. If we do not create this process, Characteristic line impedance. we risk losing information. Fibre optic automatic testing systems are available For coaxial cables loss and return loss on the market today. These systems are (unwanted reflected signal) are of interest. based on a centralised database con- Data nected to several test probes (OTDR signal It is generally accepted nowadays that platforms) which periodically test the the infrastructure is stable and reliable. fibre links via optical fibre switches. Consequently, preventive maintenance takes on a low priority and most inter- Automatic testing could guarantee an Monitoring ventions in the optical infrastructure are effective maintenance routine, doing the system corrective. tests according to programs specified by the operator, producing updated informa- Network monitoring, performed with tion about the network, measured under Figure 2 Dedicated branched fibre automatic systems, can be considered the same conditions as the last measure- monitoring: one of the splitter ports at as a part of preventive maintenance. It ment was done and guaranteeing a proper each splitting stage is looped back over involves periodic measurement of the and automatic record of the information. a dedicated fibre Telektronikk 2/3.1999 209

211 services are running, excluding dark Test Signal fibres where only a fibre running in the transmitter same cable is tested. One of the schemes that could be used is depicted in Figure 4. The monitoring could be done out of Test Signal band at 1625 nm, using reference reflec- receiver tor (selective or not) to overcome the low dynamic of the OTDR. It will be neces- sary to install monitoring wavelength fil- Figure 3 Dark fibre monitoring fibre loop ters on each branch even before any real ONU is installed, so that the reference measurement is ONU connection inde- pendent. The superimposition of traces is unavoidable, but branch length trimming devices can be used to enable distinction to the monitoring system. A filter to stop One possible dark fibre method is to use between different branches. To imple- this wavelength must be used in front of two additional fibres for each fibre link ment this scheme, connectors with the transmitter and receiver equipment. It (see Figure 3) one upstream, the other embedded filters, WDMs and optical must be noted that reverse measurement downstream and loop them at the user selectors are key components [4]. from the far end to the central office is end with splices. The result is one fibre possible. loop passing through all fibre ends with Figure 5 shows the cost comparison for no splitting and the 1550 nm wavelength the various considered PON monitor The dark fibre testing method is a direct can be used for monitoring. systems. A 16 branches PON and 10 km application of the following assumptions: length is assumed as a reference for the Due to the high cost of fibres, this solu- Most network failures affect not only cost assessment. The following three tion could be implemented if the branch one fibre but the whole cable, so that configurations are used: lengths are short or if the splitter is lo- most of the time monitoring of one cated in the cabinet. 1 Dedicated branched fibre monitoring: fibre is sufficient; one of the splitter ports at each splitt- Only a small portion of failures is due On-line techniques refer to the monitor- ing stage; to passive devices. ing of every transmission fibre, while the 2 Dark fibre monitoring: fibre loop; 3 Generic Optical Maintenance scheme, based on Reference Reflections and/or Optical Selectors. Central Office Outside Plant Assumptions as follows have been con- sidered: Splitting Splitting ratio n1 # ratio n2 Configuration 1 uses two splitter Optical s stages 1 by 4 (one in central office and # Terminal F the other in the field) so that the total # number of ONUs for this configuration # is 9. Monitoring is done with data sig- To database T Rm F nals and fault location with OTDR at OS ONU 1625 nm using filters to stop this m s wavelength. OTDR To other PONs m Configuration 2 uses only one splitting stage in central office. A possible way to implement Control the reference reflections Configuration 3 uses one or two splitt- ing stages. F # = WDM WDM s ONU Considering configuration 1, the cost per OS = Optical Selector m m = Monitoring wavelength ONU using the OTDR is of course higher C s = Wavelength used by the transmission system than using the power meter. The use of T R = Reference Reflection (not selective) the dark fibre is not convenient due to the Rm = Reference Reflection (selective on m) R high cost of the fibre. The cheaper solu- T = Branch length trimming device tion is represented by the generic optical F = Filter to stop m R and F (or WDM and F) maintenance scheme based on reference C = Connector for maintenance access could be integrated in the ONU reflections and/or optical selectors. The most favourable solution, from the view- point of maintenance efficiency, is to Figure 4 Generic Optical Maintenance scheme, based on Reference Reflections have all splitters in the central office. In and/or Optical Selectors this particular implementation, the PON 210 Telektronikk 2/3.1999

212 becomes equivalent to simple point-to- 5000 point transmission for maintenance pur- poses but at the expense of increased net- work fibre cost if compared with the Config. 1 (Power Meter) splitter in the outside plant. 4000 Config. 1 (OTDR) Config. 2 (Dark fibre) Config. 3 (Splitter in outside plant) Cost per ONU (EURO) 5 Drop section Config. 3 (Two splitters) 3000 Config. 3 (Splitter in Central Office) monitoring The final section of the access network could be supported by copper pairs (or 2000 coaxial cable) or optical fibre. As the ONUs terminate the distribution section and perform a kind of medium adaptation with active interference with the trans- 1000 ported signal from one section to the other, it is difficult to implement test mechanisms allowing testing of the transmission medium directly from the 0 exchange. In this case it is necessary to Figure 5 Cost assessment for the considered PON monitor systems install a test probe in the ONU. This installation is costly and its impact is related to the number of users con- nected to one ONU. The decision to install this OAM (Operation Administra- tion and Maintenance) facility can be jus- quality of the infrastructure, and as a To monitor PON, it is possible to use tified by the reduction in the OAM costs consequence, the quality of service information from the equipment sys- due to the fact that we do not need to provided by the operator. tems. send technicians to the field to identify Three solutions for optical monitoring The most favourable solution, from the the possible fault. Equipment should are envisaged for ODN: dedicated viewpoint of maintenance efficiency, have a maintenance channel to allow the branched fibre monitoring, monitoring is to have all splitters in the central communication between the management over dark fibres and on line monitoring. office. In this particular implementa- test system and the test head installed in tion, the PON becomes equivalent the ONU. The first solution uses one of the to simple point-to-point transmission splitter ports at each splitting stage; for maintenance purposes but at the Future networks will be based on sys- continuity testing and monitoring expense of increased network fibre tems that need an active NTE (Network changes in attenuation could be done cost. Termination Equipment), so first we with the same data signal. Fault loca- should integrate OAM functions in this tion could be performed using reverse Regarding the drop section, as future element. With the OAM functions imple- OTDR at 1625 nm. networks will be based on systems that mented from the OLT to the NTE we need an active NTE, we should inte- The use of 1625 nm as test wavelength could certify the working conditions of grate maintenance functions in this implies the ODN achromaticity up our network and save time. component. With the OAM functions to this value: we would propose to implemented from the OLT to the support the standardisation bodies to NTE we could certify the working 6 Conclusion extend the present value (1580 nm) up conditions of our network and save to 1650 nm. time. The conclusions the group achieved Measurements using dark fibre with a regarding the measurements and mainte- fibre ring (two fibres each cable) are nance of the future access networks are possible at 1550 nm where the fibre Acknowledgement listed below. has its attenuation minimum. This document is based on results Acceptance of components, network On line monitoring could be the most achieved in a EURESCOM Project; qualification and network monitoring efficient solution: in this case we are this does not imply that it reflects the are the three key phases before correc- limited to the use of an out of band common technical position of all the tive maintenance. wavelength (16251650 nm), and EURESCOM Shareholders/Parties. It is very important to have a central- additional key components are re- The author gratefully acknowledges the ised database that collects the measure- quired such as embedded filters in con- support of EURESCOM for carrying out ments in order to keep a record of the nectors, low cost optical switches with this work, and wishes to express a special infrastructure. The decision to record high number of ports and WDMs. thanks to all Task 4 of P614 participants, the information and the organisation of since this paper reflects the results of this information will influence the their work. Telektronikk 2/3.1999 211

213 References FTTCab Fibre To The Cabinet FTTEx Fibre To The Exchange 1 Stern, J et al. Full services access network requirements specification. FTTH Fibre To The Home In: Proc. VIII International Work- GSM Global System for Mobile shop on Optical/Hybrid Access Communication Network, Atlanta, Georgia, USA, 27 Mar 1997. HM Hardware Maintenance ITU-T International Telecommu- 2 ITU. Maintenance Philosophy for nications Union Telecommunication Networks. Telecommunications Geneva, 1992. (ITU-T Rec. M.20.) MDF Main Distribution Frame 3 Zylbersztejn, A et al. Low cost com- NO Network Operator ponents for full services access net- works. In: Proc. VIII International NTE Network Termination Workshop on Optical/Hybrid Access Equipment Network, Atlanta, Georgia, USA, OAM Operation Administration 27 Mar 1997. and Maintenance 4 Tomita, N et al. Future targets and OAN Optical Access Network technologies for an optical fibre line ODN Optical Distribution Net- remote testing system. In: Proc. work NOC97, Photonic Networks, Optical Technology and Infrastructure, OLT Optical Line Termination Antwerp, 1719 Jun 1997. Amster- OM Optical Maintenance dam, IOS Press, 1997, 209217. ONU Optical Network Unit Abbreviations OTDR Optical Time Domain Reflectometer AN Access Network PON Passive Optical Network EURESCOM EUropean institute for PSTN Public Switched Telephone RESearch and strategic Network studies in teleCOMmuni- cations WDM Wavelength Division Mul- tiplexer FSAN Full Services Access Net- works xDSL (Family of) Digital Sub- scriber Line FTTB Fibre To The Building Angelantonio Gnazzo (36) received his degree in physics from the Universit degli studi di Torino. In 1988 he joined CSELT and became involved in the design and realisation by MCVD of special optical fibres, such as dispersion shift- ed, dispersion flattened, polarisation maintained and rare earth doped fibres. From 1994 to 1996 he moved his activ- ity into the fields of integrated optics and Bragg grating devices. Since 1996 he has been working in the plant topology and maintenance department of CSELT. email: [email protected] 212 Telektronikk 2/3.1999

214 Towards a practical implementation of DSL preparing for new technology and new regimes PER KLEPSLAND 1 Introduction connection from one intermediate cross The normal lead diameters in the access connect frame to another (e). Such op- network are 0.4 mm and 0.6 mm. In the Telenor has maintained and developed tional cables (e) should not be used to trunk network and for long stretches in the access network for the support of carry ADSL systems. In Oslo a bridged the access network the lead diameter is telephone and data services. The present tap may in some cases be present usually 0.9 mm. Old cables may also situation calls for new solutions: new close to the customer. have a lead diameter of 0.5 or 0.8 mm in transmission technologies promise en- the access network, some sea cables had hanced feasibility and set new conditions In large parts of Norway there are areas 0.7 mm leads and in the trunk network for service providers and other network with a very low population density. In 1.2 mm was used. operators request for copper access. some valleys or narrow coastlines there may be in the order of ten subscribers per In the fifties and sixties the plastic insu- In this article the old and new regimes of kilometre. This will result in a structure lated cables were introduced. The first the access network are discussed. To a with many cable splices and end cross cables had massive polyethylene insu- great extent the new regimes will set the connect frames along the route (f). lated star quad wires in a layer structure. conditions for the practical implementa- Cables with a lead diameter of 0.9 mm In the late sixties the polyethylene was tion of design rules of how to develop the are used where the distance between cus- foamed and the cable kernel was filled access network. tomers and the telephone switching with vaseline. equipment exceeds 5 7 km. This article also describes transmission measurements set up in Telenors access 3.1 Kernel structure of network to characterise the quality and 3 Typical cables in Tele- modern cables performance of installed cables. An auto- nors access network The modern cable design was established matic measuring system was designed to in the very late seventies. The basic perform autonomous transmission mea- The present cables in Telenors access design is a ten-pair group where all surements in the access network. Para- network are a result of the investments of twisted pairs have different lay lengths meters like far end crosstalk loss, near several decades, and some of the cables and a random position within the ten-pair end crosstalk loss, insertion loss, im- may be dated to the period just after or group. Larger cables are composed of pedance and dc measurements were even before World War II. During this one or more fifty-pair groups, alterna- collected from approximately 500 sites. period of time the construction and lay- tively one or more hundred-pair groups. Such measurements are essential in order out of cables have changed significantly. The cable kernel is filled with vaseline to to establish planning rules for the provi- The first period is characterised by paper minimise the effects of intruded water. sion of different transmission systems on insulated star quad cables constructed in The lead diameters are 0.4, 0.6 and 0.9 copper cables in the access network, both layers and encapsulated in a lead mantel. mm. The overall sheath usually consists existing HDSL- and IDSL-systems and The mantel was later made from alu- of a foil screen and massive polyethylene. future systems like ADSL and VDSL. minium. 2 Typical structure of Tele- nors access network Figure 1 depicts some characteristic lay- out of Telenors local loop plant. In vil- lages and towns the topology of the cop- per cable network is a typical star struc- (a) (f) ture. From the switching centre (local exchange or remote concentrator) the (c) Main size of the main cables ((a) in Figure 1) (c) cross (e) are usually in the range of 100 to 2000 connect twisted pairs. The greater cables are spliced into cables (b) with fewer twisted (a) (c) pairs before they are terminated in main (d) cross connect frames. The main cross (spare) connect frames serve several end cross connect frames by direct cables (c). In (a) some cases cables from two or more end (b) (d) (c) cross connect frames may be spliced into Main cross connect frame one cable (d) on their way to the main cross connect frame. Cable joints are Intermediate cross connect frame (b) more frequent in old installations. Inter- End cross connect frame mediate cross connect frames are nor- mally not installed today. In old environ- Cable joint ments there may be one - or even two intermediate cross connect frames in cas- cade, and occasionally there may be a Figure 1 Typical topology of the local loop plant Telektronikk 2/3.1999 213

215 3.2 Rigid cables for harsh In sparsely populated areas where the 4 New technologies electrical environment distance to the telephone exchange may be far beyond 5 km, cables with a lead The present access network was devel- In Norway the soil is in large areas char- diameter of 0.9 or more are used. For sea oped from the requirements for support acterised by a high resistivity. In one cables the lead diameter is often 0.9 mm. of the plain old telephone service. This cable design developed in the seventies resulted in exchanges with associated the outer cable construction is made by access networks with a radius of 4 to 7 helical wrapped solid aluminium wires 3.4 Cable types in the present km. New technologies, often denoted enclosed by semiconductive polyethylene access network digital subscriber line and abbreviated sheath. This type of cable is mainly in- The most typical types of cables installed DSL [1], with the potential of greater stalled in the proximity to power plants, in the access network are listed in Table 2. capacity will be introduced in the access power lines and electrified railways. network. However, the practical digital The actual cable type installed depends transmission systems will often set 3.3 Typical design rules on preferences made at the time of instal- stricter limits for the obtainable line of copper plant lation (availability, cost, mechanical and range than that for the analogue network. electrical properties). The route from the The lead diameter is related to the num- main cross connect up to some end cross ber of pairs in the cable. 0.4 mm pairs connect frame may as well be installed 4.1 ISDN are only present in cables of at least 100 over a long time period, for instance Telenor experiences a great demand for pairs. 0.5 mm pairs are only present in some decades. And also in the time to ISDN services in Norway. The actual cables of 1000 pairs, and for a lead dia- follow some part of the distance may penetration for the ISDN basic access is meter of at least 0.9 mm the cables rarely have been refurbished due to damage or about 20 per cent, and within a few years exceed 100 pairs. conflicting constructional work. As a it may become the leading carrier ser- result the present route may consist of vice. The incitement to use 0.4 mm cables was a large number of cable types, and this primarily the investment cost associated complexity makes the planning in the The capacity of ISDN basic access is two with 0.6 mm cables. When cables with access network difficult and time con- B-channels of 64 kbit/s and one D-chan- 0.4 mm (and 0.5 mm) pairs allowed for suming. nel of 16 kbit/s. Several transmission the transport of analogue telephone, these options exist for ISDN basic access, but cables were normally used close to the Some convenient simplifications may the normal system [2] makes use of the exchange. The residual installation was result in a list of practical cable types as line code 2B1Q1) and echo cancellation usually made using cables with a 0.6 mm shown in Table 1. technique. This results in a modulation lead diameter. speed of 80 kbaud with a baseband of up to approximately 40 kHz. The ISDN primary rate consists of 31 B- channels (plus one channel for framing) of 64 kbit/s making up a total of 2048 kbit/s. This is a common bit capac- ity in both the PDH2) and the SDH3) hierarchy. The transport may be done Table 1 Simplified cable types for practical planning in the access network by traditional and proprietary digital line systems (HDB3 Cables with 0.4 mm leads line systems) or standardised HDSL sys- Paper insulated cables with 0.4 mm leads (twisted pairs) tems. The modulation speed for HDB34) systems [3] is 2048 kbaud and the result- Plastic insulated cables with 0.4 mm leads ing frequency range extends to approxi- mately 1 MHz. Paper insulated cables with 0.5 mm leads (twisted pairs) Cables where maximum 40 % of the length is with 0.4 mm leads Cables where maximum 20 % of the length is with 0.4 mm leads Cables with 0.6 mm leads Paper insulated cables with 0.6 mm leads (star quads) 1) 2B1Q (two binary one quarternary) Plastic insulated cables (non-filled) with 0.6 mm leads Line code comprising four levels. Plastic insulated cables with 0.6 mm leads 2) Plesiosynchronous digital hierarchy. Paper insulated cables with 0.8 mm leads (star quads) 3) Synchronous digital hierarchy. Paper insulated cables with 0.9 mm leads (star quads) 4) HDB3 (high density binary, 3) Three- Plastic insulated cables with 0.9 mm leads level line code with granted synchro- nization. 214 Telektronikk 2/3.1999

216 Table 2 Technical properties of cables installed in Telenors access network (For explanatory notes see page 216) Dia- Lead Capaci- Numb. Units Individ. Position Numb. Numb. Type Overall Overall Number Installa- meter insu- tance of twist in sub- of sub- of of shield sheath of tion (mm) lation nF/km units lengths group groups groups cable pairs period 1 0.6 FPE + 45 10 pair Yes random 5 / 10 10 buried, foil + PE 2, 5, 10, 20, 30, 50, 1979 ... vaseline ducted1) leader, Al 70, 100, 150, 200, wires *) 1) 300, 500, 700, 1000 2 0.6 FPE + 45 10 pair Yes random 10 (1) aerial 6) foil + PE 2, 5, 10, 20, 1979 ... vaseline leader 30, 50, 70, 100 3 0.4 FPE + 45 10 pair Yes random 5 / 10 20 buried, foil + PE 100, 150, 200, 1979 ... vaseline ducted 2) leader, Al 300, 500, 700, wires *) 2) 1000, 1500, 2000 4 0.6 FPE + 45 5 quad Yes fixed 5 / 10 10 buried, foil + PE 2, 6, 10, 20, 30, 50, 196879 vaseline ducted 1) leader, Al 70, 100, 150, 200, wires *) 1) 300, 500, 700, 1000 5 0.4 FPE + 45 5 quad Yes fixed 5 / 10 20 buried, foil + PE 100, 150, 200, 196879 vaseline ducted 2) leader, Al 300, 500, 700, wires *) 2) 1000, 1500, 2000 6 0.6 PE 45 5 quad Yes fixed 10 (1) aerial 6) foil + PVC 2, 6, 10, 20, 30, 196579 leader 50, 70, 100 7 0.6 Paper 37/ 250 quad max 8 layer (1) buried, Pb or Al, jute or 10, 20, 30, (40), 50, ... 80 45 ducted steel PVC 5) (60), 70, (80), 100, *) 3) 4) (140), 150, 200, 300, 400, 500 8 0.4 Paper 37/ 50/ pair max 8 layer 18 (1) buried, Pb or Al, jute or 100, 150, 200, 195780 45 100 ducted steel *) 3) PVC 5) 300, 500, 700, 1000, 1500, 1800 9 0.5 Paper 50 100 pair max 8 layer 10 (1) buried, Pb, jute or 1000 195068 ducted steel *) 3) PVC 5) 10 0.6 PE + 45 10 pair Yes random 5 / 10 5 sea Steel PE 10, 30, 50, 1979 ... vaseline wires 100, 200, 500 11 0.9 PE + 45 10 pair Yes random 5 / 10 (4) sea Steel PE 10, 30, 50 1984 vaseline wires (, 100, 200) 12 0.9 FPE+ 45 10 pair Yes random 5 (1) aerial 6) foil + PE 2, 5, 10, 1979 ... vaseline leader 20, 30, 50 13 0.6 PE 37 20 quad max 8 layer (1) aerial 6) foil + PVC 2, 6, 10, 195967 leader 20, 30, 40 14 0.6 PE + 45 250 quad max 8 layer (1) sea Steel jute 10, 30, 50, 197284 vaseline wires 100, 150, 200, 300, 500 15 0.9 PE + 37 37 quad max 8 layer (1) sea Steel jute 8, 14, 24, 196984 vaseline wires 38, 54, 74 16 0.9 PE 37 19 quad max 8 layer (1) aerial 6) foil + PVC 2, 8, 14, 24, 38 196675 leader 17 0.8 Paper 37/ 200 quad max 8 layer (1) buried, Pb or Al, jute or 10, 20, 30, (40), 195575 45 ducted steel PVC 5) 50, (60), 70, (80), *) 3) 100, (140), 150, 200, 300, 400 18 0.9 FPE+ 45 10 pair Yes random 10 (1) buried, foil + PE 2, 5, 10, 20, 30, 1984 vaseline ducted leader 50, 70, 100 19 0.7 PE 33/ 50 quad max 8 layer (1) sea Steel PE 6, 10, 20, 196675 60 wires 30, 50, 100 Telektronikk 2/3.1999 215

217 Notes to Table 2 tions. The dominant and upper part of the *) frequency band is only used in the direc- Optional, depending on installation. tion requiring the highest capacity. 1) The option of helically wrapped aluminium wires and conductive PE-jacket may also support aerial and simple sea cable applications. (Good EMC perfor- If one should add an ADSL modem con- mance.) nection to an already existing one to make up a composite transfer with equal 2) The option of helically wrapped aluminium wires and conductive PE-jacket may capacity in the two directions, this will also support simple sea cable applications. (Good EMC performance.) most probably vanish; because the two 3) The shield consists of 2 (optionally 4) helically wrapped steel tapes. ADSL modems will now be limited by NEXT. For short distances this arrange- 4) For simple sea cable application helically wrapped steel wires are added. Note ment of the two ADSL modems may 3) applies. work, but the line range is dramatically 5) reduced. Cables with a PVC sheath do not have steel tapes. 6) The overall sheath has been extruded onto the supporting steel rope and the 4.4 VDSL cable core making the characteristic shape of an 8. VDSL modems [8] will support symmet- Early appearance in the table implies that a large volume of the actual cable type is ric or asymmetric information transfer installed. with high capacity. It should be possible Cables with pair numbers of 40, 60, 80 or 140 were installed before 1970. with duplex capacity in the range up to 50 Mbit/s. However, the line range will Abbreviations: decrease to some hundred metres for the PE = polyethylene or polypropylene, FPE = foamed PE, PVC = polyvinylchloride, higher capacities. Although some Cu = copper, Al = aluminium-sheathed, Pb = lead-sheathed modems have been demonstrated, the Comments to Individual twist lengths: Where stated max 8, there will be a maxi- market demand belongs to the future. mum of four different twist lengths within one layer. The next layer will have a new set of four individual twist lengths. 5 Old and new regimes in the access network The technology for broadband access to be deployed in the access network devel- ops rapidly and the relevant regimes are changing a lot. The history related to the 4.2 HDSL small portions, and the ADSL system access network is complex. As telegra- tries to get the best performance possible Principles for HDSL systems have been phy and later telephone were introduced by allocating the information to be trans- recommended [4, 5] for instance by there were local telephone companies ferred into the most prosperous frequen- ETSI. Several options exist, but the nor- developing the local loop plant. Even as cy slots. mal technique is echo cancellation using late as the fifties and sixties there were the 2B1Q line code. There are solutions still private telephone companies in some Although the DMT technique will add for three pairs, two pairs and even for of the big cities in Norway. This multi- performance to the ADSL modem itself, one pair. This results in a base band sig- company situation also had an impact on the existence of such ADSL modems nal with a frequency domain with an the installation procedures and the types makes the overall planning of the access upper limit in the 200 to 600 kHz range, of cable being installed. In the late sixties network more difficult. The noise from depending on the number of pairs used. Telenor (previously Telegrafverket, an ADSL modem is not quite determinis- later Televerket) gained control of the tic but depends to some degree upon the access network in Norway and was in 4.3 ADSL noise environment surrounding the actual fact in a monopoly situation. twisted pair in the cable. ADSL [6, 7] is characterised by un- balanced information transfer in the two During the last fifteen years the regula- The line range of the transmission sys- directions. The capacity from network tion has again relaxed the monopoly situ- tems described previously (ISDN, HDSL to customers is normally in the range up ation, first by giving companies other and HDB3 systems) are all limited by to 6144 kbit/s. In addition there is some than Televerket permission to install and NEXT (near end crosstalk). For ADSL duplex capacity up to 640 kbit/s. For maintain PBXs, later by allowing net- modems the effect of NEXT may be some applications the bit capacity may work operators for mobile services and avoided by an unbalanced use of avail- be limited to lower values set by the net- providers of value-added services in able frequency range in the two direc- work operator. general. The latest issue is related to the access of the access network. To what The ADSL systems are not so easy to extent should other network operators handle in the planning process. Solutions have access to the copper pairs running often utilise the DMT5) technique. The 5) Discrete multi-tone is a technique util- from Telenors exchanges up to private actual frequency domain from almost dc ising up to 256 discrete carriers for and business customers? to approximately 1 MHz is divided into digital modulation. 216 Telektronikk 2/3.1999

218 Set-up designed for transmission measurements This set-up was used by Telenor for Before measurements the twisted pairs quencies scattered over the frequency autonomous transmission measure- to be measured were taken out of service range from 800 Hz to 8 MHz. The result ments in the period from 1984 to 1992. and connected to the measuring equip- from one site could be as much as The measuring equipment was trans- ment. The autonomous program, written 40,000 single measurements. The mea- ported and installed in a car as the mea- in Basic, ran for some hours. When the surements were usually done without surements were performed at different program stopped, the measuring equip- interactions from an operator. The sites in the access network. ment was disconnected and the twisted equipment used will support measure- pairs were put into service again. All mea- ments in the frequency domain up to surements were stored on tape. Measure- 13 MHz. ments were performed at 15 (or 19) fre- 8888.88 Local Remote DMM Matrix Matrix 2x50 2x50 IEEE-488 IEEE-488 balanced balanced pairs pairs Tape 8888.88 # 8888.88 A Cable under test B VVM GEN Referanse Sync Sync V.24 PC DIV 10 MHz to 50 KHz PLL 50 kHz to 1 MHz V.24 V.24 BUS Modem Modem BUS Equipment Equipment PC Hewlett Packard HP 85A, with IEEE-488 modem Modem, Alcatel DCB 19200 MK2 and V.24 interfaces VVM LF impedance analyser, DIV Frequency divider, 10 MHz to 50 kHz Hewlett Packard LF 4192A (custom design) DMM Digital multimeter, Hewlett Packard HP-3438A PLL AGC amplifier and PLL frequency synthesis, 50 kHz to 1 MHz (custom design) GEN Function generator, Hewlett Packard HP-3325A tape Tape cartridge recorder, Tandberg Data a/s TDC 3000/3025 BUS IEEE-488 extender, Hewlett Packard HP-37201A matrix Balanced matrix, Creative Engineering a/s (Oslo) CE-1006-1 The two balanced matrices are set by The local IEEE-488 instrument bus is the 50 kHz signal is amplified and used commands over the IEEE-488 bus. Each extended to the remote end by modems in a frequency synthesis to restore a matrix can provide two independent bal- connected to two spare balanced pairs 1 MHz reference signal for the signal anced connections from any of the 50 in the cable under test. The internal fre- generator. This is necessary when the balanced ports to one or more of the ten quency reference in the LF impedance remote signal generator is used as a instrument ports. Some instrument ports analyser (VVM) is divided down to 50 kHz source for the LF impedance analyser. have integrated balancing transformers. and transmitted to the remote end. Here Telektronikk 2/3.1999 217

219 A substantial demand [9, 10] for access 6 Systematic registration impairments should be clearly marked in to copper pairs in the access network is the records to avoid this pair being pro- anticipated. However, this is dependent of installed infrastructure vided for some service requiring better on the benefits and drawbacks related to performance. A cable without any record When serving new regimes it will be copper access compared to other techni- has no substantial value. even more important to have a good cal solutions providing transmission knowledge of the cable infrastructure capacity [10, 11, 12]. And it is indeed a matter of economics. The Norwegian already installed in order to enable a 7 Transmission proper engineering of the development government is also engaged in the legal and maintenance of the local loop plant measurements on and practical aspects of local loop unbundling [13]. and to support broadband services. subscriber cables It is important to keep records of the In 19931994 the Eurescom project P306 This article focuses on the technical cable capabilities; not only the registra- [14] focused on the transmission quality aspects assuming that the fee for copper tion of a pair from one distinct point to and performance of the access network. access should be based on cost. The criti- another, but also records giving the char- Most measurements reported were previ- cal point is to what extent the impair- acteristics about the transmission para- ously collected by Telenor [15, 16, 17]. ments in the copper cables will increase meters throughout the route. All cable Unfortunately the NEXT performance the maintenance cost when new compa- types along the route should be recorded presented in 1985 was approximately nies introduce new transmission systems at the time of installation, at the time of 20 dB too poor due to an error in the into the subscriber cables, and who refurbishment and whenever a cable has measuring procedure. The set-up was should be responsible for the compatibil- been damaged and repaired. Polite later re-calibrated and the old NEXT ity between the different transmission records should keep trace of and relevant results proved to be still valid, but systems provided? Obviously the net- information on all pairs, over-voltage improved by approximately 20 dB. work operator should guarantee for the protectors, cables, splices, branches, dis- compatibility, but in that case the net- tributors, connecting material and elec- The first automatic transmission mea- work operator must provide strict techni- tronic equipment. A pair with substantial surements in Telenors access network cal limits in order to guarantee the neces- took place in February 1984 in the city sary quality of service. of Drammen. The main part of the set-up was installed in a car and consisted of a calculator, a voltage vector meter, cus- tom design matrix, modem equipment and a digital multimeter. See brief details of the equipment in separate frame or Table 3 Measuring methods and conditions for the relevant parameters search more comprehensive information [14, 18]. Measurement Near end Far end Derived parameters In total, measurements were collected method / condition condition from more than 500 sites in Telenors access network. A brief summary of the Loop resistance between leads short circuit loop resistance transmission measuring methods and the km derived parameters are given in Table 3, Leakage resistance each lead to ground open circuit and some of the measurements are pre- sented in the following. Short circuit each lead to any lead open circuit 7.1 Attenuation constant Insertion loss 12 dB pad 12 dB pad attenuation constant The cable attenuation constant may be 12 dB pad attenuation constant obtained from three different kinds of Insertion loss in loop looping phase constant measurements. The results in Figure 2 12 dB pad phase velocity come from impedance measurements (remote end terminated with short circuit Differential insertion attenuation constant and open end respectively) for frequen- Impedance loss with the termi- short circuit / phase constant cies below 1 MHz. At higher frequencies measurement nated line connected open circuit phase velocity the attenuation constant is calculated in serial or as shunt characteristic impedance from measurement of insertion loss from one end to the other. NEXT 120 ohm 120 ohm NEXT For each set-up the mean value of the ELFEXT6) 120 ohm 120 ohm FEXT, ELFEXT attenuation constant of the twisted pairs (normalized to 1 km) has been calculated. The mean value of the attenuation constant from set-ups 6) Equivalent level far end crosstalk with cables of 0.4 mm leads only, is recorded in the solid red line. The lead 218 Telektronikk 2/3.1999

220 resistance has been carefully checked to 100 eliminate the set-ups with faulty informa- dB/km tion about the lead diameter or the length of cable. The other two red and dashed lines give the upper and lower limit defined by the standard deviation. At frequencies below 40 kHz there are 59 set-ups included for 0.4 mm cables. At 0,4 mm higher frequencies some of the set-ups 10 with long lengths drop out as the high insertion loss ruins the accuracy. 0,84 +/- 0,2 0,6 mm Similar calculations for 0.6 mm cables are also included in Figure 2. At frequen- cies below 80 kHz there are 166 set-ups represented by the solid blue line. The other two blue and dashed lines give the 1 upper and lower limit defined by the z z z z z z z z z z z z z z z standard deviation. kH kH kH H H H H H kH kH kH kH kH kH kH M M M M 0 0 0 0 5 4 80 10 20 40 80 5 1 2 4 8 16 35 70 1, 3, The results in Figure 2 gives some indi- Figure 2 Attenuation constant for cables with 0.4 mm and 0.6 mm leads cations of the present state of attenuation constant in an access network where the cables have been installed over a long time period. 7.2 Phase velocity In Figure 3 the phase velocity is calcu- lated from impedance measurements. For each set-up the mean value of the phase velocity is calculated. Based on these mean values the mean value and the stan- dard deviation of the phase velocity have been achieved for cables with 0.4 mm leads and 0.6 mm leads respectively. The mean value of phase velocity is repre- sented in red for 0.4 mm leads and in blue for 0.6 mm leads. The two solid lines represent the mean value as the dotted lines represent the upper and lower limit defined by the standard devi- ation. For frequencies below 40 kHz the 250 curves for 0.4 mm and 0.6 mm leads m/s result from 58 and 167 set-ups respec- tively. 200 0.6 mm 7.3 Near end crosstalk attenuation (NEXT) 150 Measurements of NEXT have been col- lected from almost 500 set-ups. For some 100 of the set-ups there may be less than 50 0.4 mm measurements for each frequency while more comprehensive set-ups may contain more than 1000. If, for instance, one pair 50 was transmitting and the crosstalk was measured on the other, measurements were not done with these two pairs inter- 0 changing as these two measurements are z z z z z z z z z z z z z z z kH kH H H H H H kH kH kH kH kH kH kH kH supposed to be quite equal. M M M M 0 0 0 0 5 4 80 10 20 40 80 5 1 2 4 8 35 70 16 1. 3. Figure 3 Phase velocity for cables with 0.4 mm and 0.6 mm leads Telektronikk 2/3.1999 219

221 80 All measurements have been collected dB and presented in plots showing the fre- quency dependence of selected per- 70 50% centiles, see Figure 4. For instance, the 5 % percentile separates the better 95 % of measurement from the worse 5 %. As 20% there are more than 220,000 measure- 60 ments for each frequency, the percentiles 10% 0,1% 1% range from 50 % down to 0.001 % the last one separating only four measure- 50 ments from the others. 0,01% Some interesting details can be observed from Figure 4. The first observation is 40 the slope of the different percentiles at 0,001% frequencies above 160 kHz. The slope seems to be 12 13 dB/decade rather 30 than the traditional 15 dB/decade. At frequencies below 80 kHz the per- 20 centiles tend to be more or less indepen- dent of frequency. This may originate z z z z z z z z z z z z z z z from reduced balance performance kH kH kH H H H H H kH kH kH kH kH kH kH M M M M 0 0 0 0 80 10 20 40 80 2 4 5 1 2 4 8 16 35 70 against earth in the measuring set-up at the lower frequencies and insufficient Figure 4 NEXT for all cables measured grounding. Another observation is that all percentiles in the 0.5 % to 5 % range have greater near end crosstalk attenuation at 160 kHz than at 80 kHz. These percentiles are the most interesting ones for network perfor- mance dimensioning. Figure 4 indicates that the cables have statistical better crosstalk performance at 160 kHz than at 80 kHz! Figure 5 shows the same phenomenon 80 even better. The NEXT values are here dB representative for plastic insulated cables with lead diameters of 0.6 mm. The most 50% interesting percentiles are as much as 4 to 70 6 dB better at 160 kHz than for 80 kHz. For other frequencies the trend in Figure 20% 5 is quite similar to that of Figure 4. 60 1% 10% 0.1% Paper insulated cables with lead diame- ters of 0.6 mm show some tendency to 50 have a few percentiles at approximately the same value for these two frequencies, 0.01% and cables with 0.4 mm all have mono- 40 tonous percentiles in this frequency range, so this phenomenon is closely related to plastic insulated cables with lead diameters of 0.6 mm. 30 7.4 NEXT related to quads 20 The different set-ups have been carefully categorised depending on certain para- z z z z z z z Hz z z z z z z z kH kH kH H H H H kH kH kH kH kH kH kH meters. For instance, set-ups containing a M M M M 0 0 0 0 80 10 20 40 80 2 4 5 1 2 4 8 16 35 70 dominant part of cables with quads form such a category, and in this category there Figure 5 NEXT for plastic insulated cables with 0.6 mm leads are 5,619 pair combinations within the 220 Telektronikk 2/3.1999

222 quads and 218,002 pair combinations 80 where the two pairs will not fit into a dB quad. 70 Figures 6 and 7 illustrate the crosstalk performance related to quads. The first figure shows percentiles of NEXT within quads and the second figure shows 60 NEXT for the remaining pair combina- 50% tions. For frequencies above 40 kHz the crosstalk performance is 10 to 15 dB 20% 50 10% better outside the quad than within. 1% 0.1% In Figure 7 it is also observed that the 40 percentiles are sloping merely at 10 dB/ decade instead of the traditional 15 dB/ decade. 30 7.5 NEXT for nearby pairs For most of the set-ups it has been pos- 20 sible to evaluate if a pair combination z z z z z z z z z z z z z z z incorporates two pairs that are located kH kH kH H H H H H kH kH kH kH kH kH kH M M M M 0 0 0 0 quite close to each other7) or not. NEXT 80 10 20 40 80 2 4 5 1 2 4 8 16 35 70 values for pair combinations residing in the same area are shown in Figure 8, and Figure 6 NEXT within quads for cables with quad structure in Figure 9 the complementary measure- ments (NEXT between pairs residing in different areas) are shown. By comparing Figures 8 and 9 the NEXT values be- tween different areas exhibit a perfor- mance that is approximately 8 to 10 dB better than within the same area at least for frequencies above 80 kHz. This is a typical trend recognised for each type of cable. 7.6 NEXT for paper insulated 80 cable dB NEXT for paper insulated cables with leads of 0.4 mm is depicted in Figure 10. 70 50% These cables are characterised by twisted pairs organised in layers. It can be seen that the performance for frequencies 20% 60 above 80 kHz is comparable to the gen- 10% eral trend in Figure 4. At lower frequen- 0.1% 1% cies the performance seems to be better; however, the number of measurements is 50 limited. 0.01% 40 7) Close relation or same area: Pair 0.001% combination characterised by either - two pairs belonging to the same ten- 30 pair group; - one pair in one quad and one pair in the same quad or one of the follow- 20 ing/preceding two quads in the same z z z z z z z z z z z z z z z kH kH kH H H H H H kH kH kH kH kH kH kH M M M M 0 layer; 0 0 0 80 10 20 40 80 2 4 5 1 2 4 8 16 35 70 - two pairs separated by no more than three pairs in the same layer. Figure 7 NEXT between pairs residing in different quads Telektronikk 2/3.1999 221

223 80 dB NEXT for paper insulated cables with 70 0.6 mm leads is shown in Figure 11. These cables are constructed with quads 50% organised in layers. Comparing with Fig- 60 ure 4 the paper insulated quad cables tend to have 1 to 3 dB worse NEXT for 20% 0.1% most frequencies. Similarly the NEXT 1% 50 10% within quads for paper insulated cables (no figure given) tends to be 2 to 5 dB worse than the NEXT within quads for 0.01% all quad cables as given in Figure 6. 40 7.7 Equivalent level far end crosstalk (ELFEXT) 30 The equivalent level far end crosstalk has been measured and presented in Figure 12. At lower frequencies there are more 20 than 230,000 measurements at each fre- quency. Some of the smaller percentiles z z z z z z z z z z z z z z z kH kH kH H H H H H kH kH kH kH kH kH kH M M M M 0 0 0 0 80 10 20 40 80 2 4 5 disappear at high frequencies as the 1 2 4 8 16 35 70 insertion loss of some of the measured Figure 8 NEXT between neighbouring pairs cables becomes too large and ruins the value of these particular measurements. 80 8 Planning rules dB 50% Design rules should be established for all 70 transmission systems to be transferred in the copper cables. Examples of actual 20% digital transmission systems are shown in 10% Table 4. The list should be expanded to 60 0.1% 1% accommodate for ADSL and other DSL- systems as soon as practical transmission equipment is ready for deployment. 50 0.01% As the analogue telephone service still has the highest demand, the provision for 0.001% it should also be granted at an acceptable 40 quality of service. In Telenors access network all analogue telephone lines are immediately terminated in an analogue- 30 to-digital converter. As long as the ISDN basic access may be supported at accept- able quality, the analogue service will 20 most probably also be provided with an acceptable level of quality. This may not z z z z z z z z z z z z z z z kH kH kH H H H H H kH kH kH kH kH kH kH necessarily be the case if the trunk net- M M M M 0 0 0 0 80 10 20 40 80 2 4 5 1 2 4 8 16 35 70 work is composed of analogue lines (trunks). Figure 9 NEXT between pairs other than neighbouring pairs Some analogue one-plus-one carrier sys- tems still exist in Telenors access net- work, and there may be some interfer- Table 4 Digital transmission systems on copper cables ence problems for lines exceeding 3 km. However, these carrier systems may be 2B1Q: IDSL, ISDN basic access 160 kbit/s succeeded by digital IDSL-systems or 2B1Q: modem, with 320 kbit/s and 576 kbit/s avoided by upgrading the infrastructure of the copper cable network. 2B1Q: HDSL with (one,) two or three pairs BiPh: modems with 128 kbit/s to 320 kbit/s HDB3: systems with 704 kbit/s and 2 048 kbit/s 222 Telektronikk 2/3.1999

224 8.1 Principles for establishing 80 design rules dB Telenors design rules for allowable line 50% ranges are based on some millions of 70 crosstalk measurements in Telenors access network. The measurements are 20% (almost) evenly distributed in the fre- 60 10% quency range from 800 Hz to 8 MHz. The measurements are sorted out and 1% interpolated to support calculations of allowable line ranges for the different 50 transmission systems in Table 4 within each of the cable categories defined in Table 1. The estimated line ranges are 40 calculated based on various numbers of disturbers from the same type of trans- mission system. Telenor established the 30 first planning rules [19] for digital trans- mission systems on balanced copper cables in 1989 based on the collected transmission measurements. 20 z z z z z z z z z z z z z z z kH kH kH H H H H H kH kH kH kH kH kH kH M M M M 0 0 0 0 80 10 20 40 80 8.2 Line range formula 2 4 5 1 2 4 8 16 35 70 The formula to calculate the line range Figure 10 NEXT for paper insulated cables with 0.4 mm leads is as follow: anext (f, Pconf, , n) 10 log10(n) ad(f, l) l s/N margin > 0 The estimated NEXT value (anext) is determined by a gamma distribution characterised by the frequency, the confi- dence level, the parameter (calculated from the quotient between the mean value and the standard deviation) and the number of disturbers (n). 80 The term 10 log10(n) relates to the equiv- dB alent number of disturbers. 70 The term related to attenuation, ad(f, l) 50% l , is calculated from the sum of the mean value and twice the standard devia- tion of the attenuation at the relevant fre- 60 20% quency. The value of is close to 1.0 in 1% 10% most cases. 0.1% 50 The signal-to-noise ratio and the margin are specific for the actual transmission system. 40 0.01% 8.3 The number of disturbers For a particular transmission system the 30 presence of other types of equipment may effect the allowable line ranges. A rough estimate of the equivalent number 20 of disturbers should be established. z z z z z z z z z z z z z z z kH kH kH H H H H H kH kH kH kH kH kH kH M M M M 0 0 0 0 80 10 20 40 80 2 4 5 For each transmission system equivalent 1 2 4 8 16 35 70 values for the impact of the other types of transmission systems are presented. Figure 11 NEXT for paper insulated cables with 0.6 mm leads and quad structure Telektronikk 2/3.1999 223

225 70 8.5 Practical aspects related to local loop unbundling dB The provision of open access to copper 60 50% cables is often described as local loop unbundling. The effects of local loop unbundling may have a great impact on 20% the design rules for provision of compati- 50 ble transmission systems in the access 10% network. 40 One example may well illustrate this. 1% One network operator provides an HDSL transport to his customers and also pro- 0,1% vides for local loop unbundling. A com- 30 0,01% petitive service provider terminates his 0,001% leased lines with HDSL systems with a different line code and with an output 20 level that is at least 10 dB higher than that of the network operator. How will this impact the maintenance cost for the 10 network operator? and what about compatibility between different transport z z z z z z z z z z z z z z z services? kH kH kH H H H H H kH kH kH kH kH kH kH M M M M 0 0 0 0 80 10 20 40 80 2 4 5 1 2 4 8 16 35 70 Even if the line system on the leased line Figure 12 ELFEXT, equivalent far end crosstalk for all cables should be defined by standards given by ETSI or any other standardisation body, the situation is still different to that of the network operator not supporting local loop unbundling. As the network opera- tor provides a transmission system for his own network he will also have the possi- bility to make special agreements with his equipment supplier to adjust the sup- For each type of disturbing system the One example illustrates this effect. The ported equipment, for instance to request amount should be multiplied with the cable attenuation decreases as the lead for more strict filtering or more exact impact value and added to give one diameter is increased. However, for some output level of the transmitting circuits. single equivalent number of all dis- transmission systems the calculated line However, the network operator will nor- turbers. If the number of equivalent dis- ranges may even be less on 0.6 mm cable mally not be in a position to influence the turbers exceeds ten, the actual number than on 0.4 mm cable. The main reason equipment provided for the competitive has little impact on the line ranges. The for this effect is that the 0.6 mm cables service provider as long as the agreed estimated number of disturbers should may be severely affected by near end standards are followed. The impact of accommodate for the future situations. crosstalk, particularly between the two this situation is not a lot of decibels or pairs in a star quad of a paper insulated kilometres but may well result in an esti- In practical situations the number of cable. mated line range being reduced by a few neighbouring pairs that contribute with hundred metres. crosstalk noise is also limited. The For this reason the presented line ranges crosstalk coupling is usually greatest for also have optional values where NEXT adjacent pairs. As the distance between values within the star quads have been 9 Concluding remarks pairs increases, the crosstalk coupling excluded. These results may be useful if tends to decrease and the incremental a strategy to use only one of the two pairs The nature of the cable measurements impact becomes quite moderate, compare in the star quads may be established for presented in this article shows that there Figures 8 and 9. this particular service. is a need for complex, but simplified, planning rules to aid network planners These examples clearly illustrate the in how to deploy the access network for 8.4 Practical results obtained value of keeping polite records; not only existing systems and for new DSL-tech- from the formula of each pair, but also of the cable types nology. The formula may be used to calculate used throughout the cabling route. practical line ranges for the actual trans- The nature of the copper cables will also mission system. The allowable line impact the evolution of new regimes in ranges are given as a function of the the access network and define to what an equivalent number of disturbers. The extent local loop unbundling will be line ranges are primarily limited by near practical or not. end crosstalk. 224 Telektronikk 2/3.1999

226 References 11 Full local loop unbundling unlikely 17 Klepsland, P. Evaluating the quality in UK. London, Point Topic Ltd, of twisted pair cables in the sub- 1 ITU-T. Transmission systems and 1999,02,25. [online] URL: scriber network. Telektronikk, 81, media. Overview of digital sub- 129134, 1985. scriber line (DSL) recommendations. .html Geneva, ITU, 1998. (Draft G.995.1.) 18 Klepsland, P, Langland, K. Doku- 12 Oftels five options for local loop mentasjon av utstyr og programmer 2 ITU-T. Digital transmission system access. London, Point Topic Ltd, ved mling p abonnentkabler. on metallic local lines for ISDN basic 1999,02,25. [online] URL: Kjeller, Telenor FoU, 1984. (TF R rate access. Geneva, ITU, 1993. 2/1984.) (G.961.) .html 19 Planleggingsregler for dimensjoner- 3 ITU-T. Physical/electrical character- 13 St. meld. nr. 24 (19981999). Om ing av digitale transmisjonssystemer istics of hierarchical digital inter- enkelte regulatoriske sprsml i tele- p symmetriske kabler. Oslo, faces. Geneva, ITU, 1991. (G.703, sektoren. Oslo dep. (http://odin.dep. Telenor, 1989. (Art nr 03-0297.) clause 6.) no/repub/9899/stmld/24/) 4 ETSI. Transmission and multiplexing 14 EURESCOM. Access network evolu- (TM); High bitrate digital subscriber tion and preparation for implementa- line (HDSL) transmission system on tion, Deliverable 2, Feasibility of metallic local lines; HDSL core spec- enhanced copper technologies. Hei- ification and applications for 2048 delberg, Eurescom, 1994. (Eurescom kbit/s based access digital sections. P.306, Task 5.2.) Sophia Antipolis, ETSI, 1998. (TS 101 135, Ver. 1.4.1.) 15 Klepsland, P. Bakgrunn for planleg- gingsregler i abonnentnettet. Trans- 5 ITU-T. Transmission systems and misjonsmlinger p kabel. Kjeller, media. High bitrate digital sub- Telenor FoU, 1987. (TF R 41/87.) scriber line (HDSL) transmission system on metallic local lines. 16 Klepsland, P. Evaluating the quality Geneva, 1998. (Draft G.991.1.) of twisted pair cables in the sub- scriber network. ITU Telecommuni- 6 ITU-T. Transmission systems and cation Journal, 52, 621626, 1985. media. Asymmetrical digital sub- scriber line (ADSL) transceivers. Geneva, 1998. (Draft G.992.1.) 7 ETSI. Transmission and Multiplexing (TM); Access transmission systems on metallic access cables; Asymmet- rical Digital Subscriber Line (ADSL) European specific requirements [ANSI T1.413 1998, modified]. Sophia Antipolis, ETSI, 1998. (TS 101 388 Ver. 1.2.1.) 8 ETSI. Transmission and Multiplexing (TM); Access transmission systems on metallic access cables; Very high speed Digital Subscriber Line (VDSL); Part 1: Functional require- ments. Sophia Antipolis, 1998. (TS 101 270-1 Ver. 1.1.1.) Per Klepsland has been with Telenor since 1969 and is 9 Lewin, D, Kee, R. Local loop now Senior Research Scientist at Telenor R&D, Kjeller. unbundling in Norway. London, He is mainly engaged in the transmission and quality OVUM, 1997. (OVUM CC583) aspects of copper cables and related transmission sys- tems. He also participates in Cenelec TC 215 in standard- 10 Dall, S. Hurtigt Internet bliver meget ization of installation procedures and specification of langsommere. Ingeniren, uge 7 generic cabling systems. 1999. ( email: [email protected] 0799/adsl01.html) Telektronikk 2/3.1999 225

227 Calculation of cable parameters PER KLEPSLAND 1 Introduction 2 Properties of a two-port network During the 1980s Telenor developed a A two-port network may be characterised by a matrix representation. In Figure 1 the measuring set-up and relevant methods currents and voltages are depicted. The two-port network may not necessarily be recip- to measure important transmission para- rocal or symmetrical. The two-port network is characterised by the matrix A, repre- meters of balanced copper cables within sented with the elements a11, a12, a21 and a22, the image impedance, Z1, on port one, . the access network. Most measuring the image impedance, Z2, on port two and the image propagation functions, g and g, equipment lacks the n