Chuong 1.html

ADSL,VDSL,and Multicarrier Modulation . John A. C. Bingham Copyright # 2000 John Wiley & Sons, Inc. Print ISBN 0-471-29099-8 Electronic ISBN 0-471-20072-7

1

INTRODUCTION

The four principal media for transmission of high-speed data to and from a customer premises are:

1. Subscriber telephone loop [digital subscriber loop (DSL)]: the unshielded twisted pair (UTP) of copper wires used for “plain old telephone service” (POTS)

2. Coaxial cable: originally installed for unidirectional (“downstream”) transmission of television, but increasingly being used for bidirectional data transmission

3. Optical ®ber: originally used for very high-speed trunk transmission, but now being considered for either the last leg of the distribution [®ber to the home (FTTH)] or the penultimate leg [®ber to the exchange or ®ber to the neighborhood (FTTE or FTTN)]. The latter case is the only one that will concern us, because then the last leg is provided by the distribution portion of the DSL (see Section 3.1).

4. Wireless.

There is no general answer to the question of which of these is best, and the four have contended vigorously for many years for both media attention and developmental and deployment capital. In this book we are not concerned with the rival meritsÐtechnical, ®nancial, political, social, or environmentalÐof these four media¹; we will describe only the ®rst. We are concerned only with the physical layer (the lowest layer) of the OSI model; in Chapter 2 we deal with the upper part of that layerÐthe transmission convergence (TC) layerÐand in the rest of the book, with the lower partÐthe physical medium-dependent (PMD) layer. The main topic at the PMD level is multicarrier modulation (MCM)Ð in particular, discrete multitone (DMT)Ðapplied to xDSL. There are, however, many types of DSL (e.g., ISDN, HDSL, SDSL) that do not use MCM, and furthermore, MCM is used in media (particularly wireless) other than DSL; we

¹The perception of the merits seems to have depended on who put out the last set of press releases! 1
Figure 1.1 Scope of this book.

discuss these only brie¯y as an introduction to the main topic. This is illustrated in the Venn diagram of Figure 1.1; the scope of the book is less than the union but greater than the intersection.

1.1 ARRANGEMENT OF THIS BOOK

In the remainder of this chapter we describe, in sequence, the histories of DSL, MCM, and MCM applied to xDSL. In Chapter 2, by Alan Weissberger (which probably could be expanded to be a book by itself ), the TC layer is discussed. In Chapter 3 we describe the physical medium, and in Chapter 4, ways of using the medium for data. There is no completely logical order or grouping of topics thereafter. In Chapters 5, 6, and 7 the theory of MCM is discussed: the fundamentals in Chapter 5; discrete multitone (DMT), a simple version of MCM, in Chapter 6; and general MCM in Chapter 7. Chapters 8, 9, and 10 are practical, dealing with the implementation of DMT as ADSL and VDSL. Chapter 11 is the “fun” one: a discussion of some possible future improvements for A, V, and xDSL in general.

1.2 HISTORY (ONGOING) OF DATA ON THE DSL

It is dif®cult to say when the subscriber loop was ®rst used for data (telegraph; 110-bit/s voice-band modems?), but the systems that are still around are as follows.

^* Basic rate access DSL (also known as just DSL). 160-kbit/s one-pair fullduplex system. Used in the United States only for data services to provide access to the Integrated Services Digital Network (ISDN) [ANSI, 1993b], but also used in Europe for 2 Â 64 kbit/s digitized voice service. ITU Recommendation G.961 de®nes three different systems:
^* Appendix I: 2B1Q coding with echo cancellation (EC); used in North

America and much of Europe. Also standardized in North America as T1.601: see [ANSI,1993].
HISTORY (ONGOING) OF DATA ON THE DSL 3
^* Appendix II: 4B3T coding with EC; used in some European countries. ^* Appendix III: bipolar (a.k.a. AMI) coding with synchronized timedivision duplexing (TDD; a.k.a. “ping-pong”); used in Japan.

^* T1. 1.544-Mbit/s dual simplex on two pairs using AMI coding and repeaters spaced every 6 kilofeet (kft); used in North America. T1 was originally designed, and installed [Cravis and Crater, 1963] from 1962 onward for interof®ce (trunk) transmission of 24 multiplexed 64-kbit/s PCM voice channels; for that use it has now been almost completely replaced by ®ber and microwave. Since the early 1970s, however, it has also been used on the DSL, and it is by far the most severe potential source of crosstalk into ADSL.²It will be made obsolete by HDSL2, but it is very unlikely that installed systems will be replaced.

^* E1. Similar to T1, but 2.048 Mbit /s for 32 voice channels, with repeaters spaced approximately every 2 km; used everywhere else in the world.
^* High-speed DSL (HDSL). 1.536-Mbit/s two-pair and 2.048 Mbit/s twoand three-pair, full-duplex systems using 2B1Q coding and echo cancellation: originally de®ned in [ANSI,1994] and [ETSI,1995], and now codi®ed as ITU Recommendation G.991.1.
^* Asymmetric DSL (ADSL). ANSI standard T1.413 [ANSI, 1995] de®nes an ADSL system to transmit downstream and upstream data rates up to 6.8 and 0.64 Mbit /s, respectively, within a radius of approximately 12 kft from the CO [known as the carrier serving area (CSA)], and 1.544 and 0.176 Mbit /s within a radius of 18 kft [the extended CSA (ECSA)]. ITU Recommendation G.992.1 de®nes a system based on T1.413 as a core, but expanded via three annexes to meet particular regional needs. G.992.2 de®nes a simpler system with a wider range of data rates and ranges (see Section 1.5 on ADSL lite) that is line compatible with G.992.1. ADSL is the main subject of this book, and T1.413 and / or G.992 should be indispensable companions while reading.
^* Very high-speed DSL (VDSL). VDSL will be used primarily in “hybrid ®ber/copper” systems to connect optical network units (ONUs) to customer premises. In ®ber to the exchange (FTTE) systems these ONUs will be in the CO, and we will call the VDSL transceivers VTU-Cs. In other systemsÐFTTN(eighborhood), FTTC(urb), and FTTB(uilding)Ð the ONUs will be outside the CO; the only difference between these will be the length of the loop from ONU to the customer premises: up to 6 kft for FTTN or 1.5 kft for FTTB systems. We will call them all FTTC(abinet) systems, and the transceivers VTU-Os. If the location (CO or outside ONU) is not important for a particular discussion we will call the “head-end” transceiver VTU-C/O. VDSL ranges vary from 1 to 6 kft, depending on the location of the ONU, and corresponding aggregate! (down plus up) data rates vary from approximately 58 to 4.6 Mbit/s. Two modes are de®ned in [Cioffi, 1998]: asymmetric with a down/up ratio of approximately 8/1, and symmetric. Three line codes have been proposed: DMT, Zipper (a variant of DMT), and CAP (a variant of QAM).
^* HDSL2. 1.536-Mbit/s one-pair full-duplex system using a mixture of frequency-division duplexing and echo cancellation, and very sophisticated trellis coding. Probably will be standardized by ANSI in 1999 and by the ITU as G.991.2.
^* SDSL. Various unstandardized one-pair full-duplex systems achieving less than 1.536-Mbit/s. The advantages over HDSL2 may include lower cost, earlier availability, and greater range.

²See Section 4.5 for a discussion of this.

The general pattern has been for each successive system to use a wider bandwidth than the preceding one, and a totally different, non-backwardcompatible modulation scheme.

1.3 HISTORY OF MULTICARRIER MODULATION

The principle of transmitting a stream of data by dividing it into several parallel streams and using each to modulate a “subcarrier” was originally applied in Collins’ Kineplex system,³described in [Doelz et al., 1957]. It has since been called by many names, and usedÐwith varying degrees of successÐin many different media:

^* FDM telephony group-band modems. [Hirosaki et al., 1986] described an orthogonally multiplexed QAM modem for the group band at 60 to 108 kHz. It used a ®xed bit loading (see Section 5.3), and its main advantage over single-carrier modems was a much reduced sensitivity to impulse noise. I do not know if there are any still deployed.

^* Telephony voice-band modems. [Keasler and Bitzer, 1980] described a modem for use on the switched telephone network (STN), and in 1983 Telebit Corporation introduced the Trailblazer modem [Fegreus, 1986], which used dynamically assigned multiple QAM. It far outperformed all single-carrier contemporaries, and for certain applications (e.g., ®le transfer using UNIX) it was ideal. It was proposed as a standard for an STN modem [Telebit, 1990] but was rejected because of its very large latency.⁴

^* Upstream cable modem. [Jacobsen et al., 1995] proposed synchronized discrete multitone (SDMT) for the 5- to 40-MHz upstream band in a hybrid ®ber coax (HFC) system. SDMT uses a combination of frequency

³ I did hear a claim that there was a system before Kineplex, but I do not remember the details. If there was such a system, I apologize to the developers for slighting them.
⁴It used 1024 subcarriers with a spacing of approximately 4 Hz.

MCM (DMT) AND DSL 5

division multiple access (FDMA) and time DMA (TDMA) and is ideally suited to both the medium and the system requirements, but it faded because of lack of commitment and a sponsor. I do not know whether it is now dead or just cryogenically preserved. The name SDMT is now used to describe another synchronized version of DMT proposed for VDSL.

^* Digital audio broadcasting. Coded orthogonal frequency-division multiplexing (COFDM)⁵is a version of MCM that uses IFFT modulation (see Section 6.1), ®xed bit loading,⁶ and sophisticated coding schemes to overcome the fades that result from multipath. It has been standardized in Europe as the Eureka system [OFDM1].

^* Digital audio radio. A version of DMT for use in the United States in the same frequency bands as the established FM stations was tested in 1994. It performed as well as could be expected in the very severe narrowband, low-power, high-noise (from the FM signal) multipath-distorted environment, but that was not good enough for widespread deployment. In-band digital radio is currently on the back burner in the United States.

^* Digital TV. COFDM has also been standardized for digital video broadcasting [OFDM2].

The subtitle of [Bingham, 1990] was “An idea whose time has come” but “has come” at that time clearly should have been “is coming”, “may come”, “came and went”, or “probably will never come”, depending on what application and/ or transmission medium was being considered.

Other Forms of MCM. All of the foregoing systems used sinusoidal subcarriers, but a more general form of MCM, which uses more complex signals as “subcarriers” in order to maintain orthogonality in a distorted channel was originally proposed in [Holsinger, 1964]; it has since had many different forms, which are discussed in Chapter 7.

1.4 MCM (DMT) AND DSL

The use of DMT for ADSL was ®rst proposed in [Ciof®, 1991]. In 1992, ANSI committee T1E1.4 began work toward a standard for ADSL, de®ned a set of requirements, and scheduled a competitive test of all candidate systems. The tests were performed on laboratory prototypes in February 1993, and in March 1993 the DMT system was chosen to be the basis of the standard. I took over as editor of the standard in 1994.

Representatives of all seven regional bell operating companies (RBOCs), most European national telcos (previously, PTTs), and at least 30 telecommunications manufacturers from throughout the world participated in the drafting and revising process, and in August 1995, Issue 1 of ANSI Standard T1.413 was published. As is usual with such standards, changes were suggested at the last minute that were too late to be included in Issue 1, and work was started immediately on Issue 2. This work proceeded rather desultorily, however, because market demands had changed since the original project was de®ned. 6 ‡ Mbits/s downstream for high-quality compressed video (“video on demand”) no longer seemed economically attractive, and there was a danger that T1.413 would become a standard without an application.

⁵See the specialized bibliography in the reference section.
⁶In a broadcast mode there can be no feedback from receiver to transmitter.

Then in early 1996 access to the Internet became paramount. As [Maxwell, 1996] put it, “… simply uttering the word Internet before securities analysts doubled a company’s stock price.” ADSL was reborn with a different persona:

^* 6 ‡ Mbit/s to perhaps 50% of all households became less important than 1.5 Mbit/s to perhaps 80%.
^* ATM became a much more important transport class of data than STM.
^* Dynamic rate adaptationÐthe ability to change data rates as line conditions (mainly crosstalk) changeÐbecame important.

Work was redirected accordingly, and Issue 2 was published early in 1999. ITU Study Group 15 began work on xDSL in late 1997 and addressed the
questions of unique national and regional needs (see Appendix B.1). G.992 for
ADSL was published in 1999.

1.5 ADSL “LITE”

T1.413 was still, however, perceived by manyÐparticularly those in the computer industryÐas being too complicated, expensive, and telco-centric. This prompted demand for a “lite” modem. SG 15 took over responsibility for what was temporarily called G.lite and is now designated G.992.2. The characteristicsÐsome fairly precise, some rather vagueÐof a G.lite modem were billed as:

1. User-friendly; that is, very few options, take it out of the box, plug it in without requiring assistance from the phone company,⁷and use it.
2. Less complex; therefore, presumably, less expensive.
3. No rewiring of customer premises should be needed; existing house wiring, no matter how ancient and chaotic, should be adequate.
4. The low-pass part of the POTSsplitter (see Section 9.1) should not be needed.
5. Only transport of ATM should be supported.

⁷No “truck roll.” SOME HOUSEKEEPING DETAILS 7

6. Range should be the more important than rate; some service, albeit at only 0.7 ‡ Mbit/s downstream, should be possible out to 22 kft.
7. “Always on”; that is, an ATU-R should have a standby mode in which it would use very little power, but be readyÐwithin some small-but-stillto-be-de®ned timeÐto receive email and other unsolicited downstream transmissions.

Requirement 4 started out as the most important, but was modi®ed as work progressed.
1.6 SOME HOUSEKEEPING DETAILS
1.6.1 Units of Measurement

In most scienti®c and engineering books there would be no question that the metric system of measurement should be used exclusively. In discussing telephone systems, however, the issue is not as clear. In the United States, wire sizes and lengths are measured in American wire gauge andÐin a strange, halfhearted attempt at metri®cationÐkilofeet, and most of my experience has been in those units. Therefore, I will use them primarily and, wherever appropriate, show conversions to the metric system. I will use the compatible set of units: k , nF, mH, and MHz in all except one case: dBm/Hz is too ®rmly entrenched to be dislodged by the more convenient dBm/MHz.⁸

1.6.2 References

In order to help readers recognize references without having continually to ¯ip to the end of the book, we cite them as [Smith and Jones, 19xy] without worrying about whether we are referring to the paper or the authors. On some topics we have included block bibliographies at the end of the reference section without citation or recommendation of any particular paper.

⁸Both of them are, of course, mathematically inconsistent (x dBm/Hz does not mean 2x dBm in 2 Hz!), but mW/Hz never caught on.