Chuong 3.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

3

THE DSL AS A MEDIUM FOR HIGH-SPEED DATA

Subscriber loops¹ which connect the customer premises to a central (or switching) of®ce (CO), were developed and deployed for voice transmission, and have been well described by many authors. [Gresh, 1969], [Manhire, 1978], [Freeman, 1981], and [AT and T, 1982] are excellent references; they are old, but then the subscriber loop is very old, and not much has changed in twenty years! A recent description of those characteristics of the loops that are appropriate for DSL appears in [Rezvani and Khalaj, 1998].

NOTE: With the advent of ®ber to the neighborhood (FTTN: see Section 1.2), subscriber loops will also be used to connect customer premises to an optical network unit (ONU) using VDSL. When describing the use of loops for generic DSL, I will refer only to CO and will differentiate between CO and ONU only when discussing VDSL speci®cally.

3.1 MAKE-UP OF A LOOP

Each subscriber loop consists of a pair of insulated copper wires of gauges ranging from 26 AWGto 19 AWG(approximately 0.4 to 0.91 mm). The insulating dielectric is mostly polyethylene, but some paper-insulated pairs still exist. A typical loop plant, as shown in Figure 3.1, consists of a multipair feeder cable emanating from the CO; this may contain up to 50 binder groups, each of which may contain 10, 25, or 50 pairs. At a feeder distribution interface (FDI) a feeder cable is then divided into several smaller (up to 50 pairs) distribution cables; these are then ®nally broken out into many individual drop-wire pairs to customer premises.

Within the cables the two wires of each pair are twisted around each other to form an unshielded (and unsheathed) twisted pair (UTP). ANSI is presently
¹See Section 3.1.3 for why it is called a loop. 21
Figure 3.1 Typical loop plant: feeder and distribution cables to customer premises.

de®ning the properties (twist length or pitch, balance, dielectric loss, etc.) of several categories of UTP: Category 3 and Category 5 in particular. Most of the installed plant is Cat-3 or lower (i.e., worse in some or all properties), but there is a small amount of Cat-5 installed, particularly from ONUs to new customer premises. The pitch of Cat-3 can vary from about 1.5 to 3 ft, and the twist is hardly discernible to the untrained eye when the outer sheath of a cable is removed. For purposes of maintaining balance, however the most important parameter is the ratio of the signal wavelength to the pitch; even at 15 MHz, which is about the highest frequency presently contemplated for use on UTPs, the wavelength/pitch ratio is about 20:1. The pitch for Cat-5 is only a few inches and is precisely varied from pair to pair within a cable; the crosstalk balance may be as much as 20 dB better than for Cat-3.

[Rezvani and Khalaj, 1998] report that in the United States most multipair cables are constructed in an attempt to make all pairs “equal”; that is, the position of any pair within the cable changes, and no two pairs stay close together for any great distance; this is intended to average the crosstalk between different pairs and to reduce the difference between the worst- and bestcase interferers (see Section 3.6). I have, however, also heard the opposite opinion: that pairs tend to maintain their position in a cross section. There may well be both types of cable out there, making the task of modeling (see Section 3.6) even more dif®cult. In other countries (e.g., Japan and Germany) two pairs are ®rst twisted as quads, which are then combined in a larger cable. The crosstalk between pairs in the same quad is much higher than average, and that between pairs in different quads is lower than average.

3.1.1 Length of the Loop

Telephone plants throughout the world vary widely in the distribution of their customers (i.e., in percentage of customers covered as a function of distance from the CO). During the development of T1.413 it was generally “agreed” that the so-called “extended carrier serving area” with a nominal 18-kft radius would include about 80% of all customers; this was consistent with Bellcore’s 1973 loop survey [AT&T, 1982], which showed 85% within 18 kft. It was probably tacitly assumed, moreover, that the remaining 20% were typically in rural areas with a lower demand for data services. As a counterexample, however, one central of®ce in San Jose, California (a modern city in Silicon Valley with highly sophisticated data-hungry residents) has approximately 64% of its customers more than 18 kft away.

3.1.2 Balance

All signals on the subscriber loop are carried in the differential mode², in which the current in one wire is balanced by an equal and opposite current in the other. Every effort is madeÐin both the manufacture of the cable and the design of the terminal equipmentÐto minimize the common-mode³component. Transmitters should be able to achieve a differential mode/common mode ratio of at least 55 dB across the used band, but because of imbalance of the two wires of any pair to “ground” (represented mainly by the other pairs), there is some differential mode-to-common mode conversion in the cable. For Cat-3, the most common type of UTP used in the United States, the output ratio is about 50 dB below 100 kHz and falls to about 35 dB at 10 MHz.⁴

3.1.3 Wire Gauge andGauge Changes

The primary parameter that controls the ability of CO equipment to perform signaling and diagnostic maintenance is the dc resistance of the loop measured between the two wires at the CO, with the wires shorted at the customer premises. In the United States, according to the revised resistance design (RRD) rules, the loop resistance is limited to 1500 .⁵Therefore, the ideal arrangement would be to adjust the gauge of the wires according to the length of the loop: the longer the loop, the larger the gauge.

Such an ideal cannot be achieved in practice, however, because, as shown in Figure 3.1, different pairs (all necessarily of the same gauge) in a large feeder cable emerging from a CO might eventually go to premises at widely varying distances. Therefore, a common practice is to start out from the CO with feeder cables containing many ®ne-gauge pairs, and increase the gauge at an FDI as the distance from the CO increases. At least one gauge change, therefore, may occur within the feeder/distribution cables and must be considered in any mathematical analysis (see Section 3.5).

² Originally, differential-mode current was called metallic circuit current to distinguish it from the common-mode current, which used a ground (i.e., nonmetallic) return.
³Originally called longitudinal mode.
⁴More on this in Sections 3.5 and 9.3.2.
⁵Eighteen kilofeet of 26 AWGhas a resistance of 1500 .

3.1.4 Bridge Taps

Bridge taps are open-circuited lengths of UTP that are connected across the pair under consideration. They can be the result of many different installation, maintenance, and house wiring practices:

^* Party lines. In the early days of telephony it was common for several customers to share the same pair. Then when more cables were installed and privacy became more affordable, the drops to the other premises were just disconnected, leaving the unterminated pairs (open-circuit stubs) still connected to the used loop. A simple con®guration is shown in Figure 3.2(a); a more complicated and less common one, with a bridge tap on a bridge tap, is shown in Figure 3.2(b).

^* Extension of the distribution cable beyond the drop to the customer premises. According to [AT&T, 1982], “the cable pair serving the customer usually [my emphasis] extends past the customer to the point at which the particular cable run ends.” These are sometimes called tappedin drops.

^* Repairs. If a pair breaks somewhere inside a cable, the repairer may simply splice in another pair without disconnecting the broken sections. It can be seen from Figure 3.3 that this leaves two bridge taps connected to the loop in use.

Figure 3.2 Bridge taps: (a) simple; (b) bridged. Figure 3.3 Two bridge taps caused by a repair.

^* Extra capacity. To allow for future service to any one of several potential customer premises, it is common practice to splice one pair in a feeder cable to one pair in each of several distribution cables. The unused pair(s) then form bridge tap(s).

^* Multiple telephone outlets within customer premises. The most common in-house wiring con®guration is a tree with its base at the service entrance. All branches that are either unterminated or teminated in onhook telephones constitute short bridge taps that may be signi®cant at VDSL frequencies. This is discussed in more detail in Chapter 10.

3.1.5 Loading Coils

A loop is often thought of as having a bandwidth of only 4 kHz, but that limitation is imposed by multiplexing equipment on the network side of the CO; it is not inherent in the loop itself. Because the switched telephone network (STN), which interconnects COs, originally used frequency-division multiplexing based on multiple 4-kHz bands,⁶the signals within the STN must be bandlimited to something less than 4 kHz. Therefore, if the subscriber loop is to be used only for access to the STN, there is no need for a bandwidth greater than 4 kHz.

At low frequencies UTP acts like a distributed RC circuit, and its response droops across the 4-kHz voice band (by as much as 12 dB on long loops). That droop reduced the capacity of early telegraphy systems and degraded the voice quality, so Heaviside⁷proposed that lumped inductors be added in series at regular intervals along the loop. A common con®guration in the United States is 88-mH coils inserted every 6000 ft; a 26-AWGloop so loaded would be designated 26H88. These loading coils ideally convert a droopy RC network into a maximally-¯at low-pass ®lter with a cutoff around 3.0 kHz. In the process of improving the voice-band response, however, loading coils greatly degrade the response beyond 4 kHz, so they must be removed (or perhaps just shorted out) to allow any wider-band service to operate on the loop.⁸Removing the coils may be a signi®cant part of the cost of providing DSL service.

3.1.6 The Drop Wire

When a pair ®nally emerges from a distribution cable, it is connected to the customer premises by a drop wire. The term refers to the “drop” from a pole, which often occurs even if the distribution cable is underground. Drop wires may be copper, steel, or a mixture. They may be ¯at or twisted, and their balance is usually much worse than that of the UTP part of the loop. This may result in pickup of radio-frequency (RF) noise (see Section 3.7.1). The characteristic impedance of drop wires is typically higher than that of UTP, and the result of the impedance mismatch on the attenuation may be signi®cant on the shorter loops and at the higher frequencies used for VDSL. They were ignored in the de®nition of the test loops for ADSL, but are considered for VDSL.

⁶ And has now largely converted to digital systems based on 8-kHz sampling.
⁷See [Riezenman, 1984] for an interesting story about the invention of loading coils.
⁸It is ironic that loading coils, which were originally added to increase the capacity of loops, must now be removed to increase it further!

3.2 LADDER MODEL OF AN UNSHIELDED TWISTED PAIR⁹

NOTE: This section contains much more detail than most readers will want, but I included it all because such a level of detail will be needed for the study of crosstalk cancellation, and because I would like this to be a comprehensive description of UTP used for xDSL.

A UTP comprises distributed inductance and resistance in series, and distributed capacitance and conductance in shunt. All four primary parameters are cited per unit length (kft in the United States; km elsewhere). A homogeneous section of unspeci®ed “unit” length and a cross section of a pair are shown in Figures 3.4(a) and (b).

Figure 3.4 (a) Lumped model of unit length; (b) cross section of UTP. ⁹If the model is used only for the differential mode, it is equally valid for both UTP and ¯at pairs; nevertheless, for simplicity we refer only to UTP from here on.
The capacitance per unit length is given by
C