You are on page 1of 114

Access Technologies:

DSL and Cable


Executive Briefings in
Key Technologies

James Harry Green

McGraw-Hill
New York Chicago San Francisco Lisbon London Madrid
Mexico City Milan New Delhi San Juan Seoul
Singapore Sydney Toronto
McGraw-Hill
abc
Copyright © 2002 by The McGraw-Hill Companies, Inc. All rights reserved. Manufactured
in the United States of America. Except as permitted under the United States Copyright
Act of 1976, no part of this publication may be reproduced or distributed in any form or by
any means, or stored in a database or retrieval system, without the prior written permission
of the publisher.

ISBN 0-07-138247-X

All trademarks are trademarks of their respective owners. Rather than put a trademark symbol
after every occurrence of a trademarked name, we use names in an editorial fashion only,
and to the benefit of the trademark owner, with no intention of infringement of the trademark.
Where such designations appear in this book, they have been printed with initial caps.

McGraw-Hill eBooks are available at special quantity discounts to use as premiums and sales
promotions, or for use in corporate training programs. For more information, please contact
George Hoare, Special Sales, at george_hoare@mcgraw-hill.com or (212) 904-4069.

TERMS OF USE

This is a copyrighted work and The McGraw-Hill Companies, Inc. (“McGraw-Hill”) and its
licensors reserve all rights in and to the work. Use of this work is subject to these terms.
Except as permitted under the Copyright Act of 1976 and the right to store and retrieve one
copy of the work, you may not decompile, disassemble, reverse engineer, reproduce, modify,
create derivative works based upon, transmit, distribute, disseminate, sell, publish or subli-
cense the work or any part of it without McGraw-Hill’s prior consent. You may use the work
for your own noncommercial and personal use; any other use of the work is strictly prohibited.
Your right to use the work may be terminated if you fail to comply with these terms.

THE WORK IS PROVIDED “AS IS.” McGRAW-HILL AND ITS LICENSORS MAKE
NO GUARANTEES OR WARRANTIES AS TO THE ACCURACY, ADEQUACY OR
COMPLETENESS OF OR RESULTS TO BE OBTAINED FROM USING THE WORK,
INCLUDING ANY INFORMATION THAT CAN BE ACCESSED THROUGH THE
WORK VIA HYPERLINK OR OTHERWISE, AND EXPRESSLY DISCLAIM ANY
WARRANTY, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO
IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR
PURPOSE. McGraw-Hill and its licensors do not warrant or guarantee that the functions con-
tained in the work will meet your requirements or that its operation will be uninterrupted or
error free. Neither McGraw-Hill nor its licensors shall be liable to you or anyone else for any
inaccuracy, error or omission, regardless of cause, in the work or for any damages resulting
therefrom. McGraw-Hill has no responsibility for the content of any information accessed
through the work. Under no circumstances shall McGraw-Hill and/or its licensors be liable for
any indirect, incidental, special, punitive, consequential or similar damages that result from
the use of or inability to use the work, even if any of them has been advised of the possibility
of such damages. This limitation of liability shall apply to any claim or cause whatsoever
whether such claim or cause arises in contract, tort or otherwise.

DOI 10.1036/007138247X
Contents

1 Introduction to Access Technologies 1


2 Data Transmission Fundamentals 15
3 Telephone Local Loop Characteristics 37
4 Digital Subscriber Line Technology 49
5 Cable Access Technology 61
6 Wireless Access Technology 75
7 Fiber-optic Access Technology 91
8 Application Considerations 99
Appendix 1 Glossary 109
iii
1
Introduction to Access Technologies

T elecommunications networks have made some remark-


able strides in the last few decades. Fiber-optic tech-
nology has revolutionized information transport, providing
enormous amounts of high-quality bandwidth that the older
transcontinental microwave network could not support. High-
speed routers and switches and new protocols have brought the
Internet to the masses, making it possible to obtain information
and communicate anywhere in the world for a flat monthly fee.
Despite these advances, a major bugbear remains. The high-
speed backbone network is accessed over a copper cable local
loop that was designed for a nineteenth-century network. To be
sure, copper wire quality increased significantly during the
twentieth century, and local loop technology improved, but
local access remains the chokepoint for the vast majority of
Internet users.

Copyright © 2002 by The McGraw-Hill Companies, Inc. Click here for Terms of Use.
2 Access Technologies: DSL and Cable

From the start, data networks were designed around the


voice-frequency circuit because nothing else was available.
Older telecommunications systems had a high error rate
because of noise that came from a variety of sources such as
clattering switches and relays, atmospheric conditions, and
technician activity. Since noise was an analog phenomenon
induced into analog circuits, the longer the path, the higher
the noise level, and the greater the error rate. Although a high
noise level is annoying to voice sessions, it is fatal to data, so
when fiber optics came along to replace the analog micro-
wave network that laced the country, it brought a revolution in
transmission quality. No longer was bandwidth constrained
by the channelized voice circuits in the wide area. Common
carriers and large users served by fiber optics could now
obtain an ample supply of digital capacity in whatever band-
width they chose.
The irony of the fiber-optic revolution is that, except for
users large enough to justify the expense of bringing fiber
directly to their premises, bandwidth is still constrained by the
capacity of a twisted pair of copper wires. The reason is that
the local architecture was designed long before any potential
for broadband data communications existed and the cost of
changing it is higher than the revenue potential can support.
The wide area network is concentrated into a backbone of
shared switches and circuits, fanning out to millions of dedi-
cated connections to every business and household in the
developed world. Not only are the multipliers huge, but also
these cables must pass through some expensive real estate.
Ideally, service providers would bring fiber optics to every
business and residence, but the cost of digging up the streets
is enormous and the only way it can be justified is by bring-
ing services for which subscribers will pay enough to justify
the cost. Telephone service is a given—a lifeline service that
few can do without. Beyond that are entertainment services,
most of which ride on coaxial cable today. The community
antenna television (CATV) providers have already invested
the capital to bring a broadband channel to more than 80 per-
cent of the households. Were either telephone or CATV to
start from scratch today, the billions that were invested in
Introduction to Access Technologies 3

twisted-pair wire and coax would instead be invested in fiber


optics, but replacement is an expensive proposition.
Only in the past decade has a third service emerged for
which subscribers are willing to pay: information. Before the
advent of the Internet, it was difficult to conceive that infor-
mation would become such an important commodity, but
today access to the Internet is a third revenue stream that has
important implications for the future. The issues are clear to
the service providers. Large business users require fiber optics
because no other medium can support their bandwidth require-
ments. Therefore, not only the incumbent local exchange car-
riers (ILECs, i.e., the traditional telephone companies), but
also competitive local exchange carriers (CLECs) and com-
petitive access providers (CAPs), are bringing fiber to those
users with enough bandwidth demand to justify it.
That leaves the small businesses and residences, which is
where the real multipliers are. The fact is that the copper wire
and coaxial cable plant is already in place and the providers
don’t expect enough revenue to justify replacing it. There-
fore, they are deploying methods of extending the life of their
existing plant. The ILECs are applying digital subscriber line
(DSL) technologies to their twisted-pair wire. CATV pro-
viders are converting their one-way systems, which were orig-
inally designed to deliver entertainment services, to two-way
systems. Wireless providers plan to bypass wired alternatives
altogether by using radio frequency devices. They have
already made inroads into the CATV market with direct
broadcast satellite services. Wireless providers such as
Teligent and Wavestar have made broad forays into the voice
and data market, but have encountered financial problems
that tend to dampen the enthusiasm of future investors. These
three classes of service providers are eyeing each other’s mar-
kets hungrily. Satellite providers are including local channels
to woo CATV customers. Cable companies are preparing to
offer telephone service, and local exchange carriers (LECs)
are positioning themselves to deliver video on demand
(VOD). All three offer Internet access, either by providing
their own service or as common carriers for Internet service
providers (ISPs).
4 Access Technologies: DSL and Cable

For larger businesses, Internet access has become a neces-


sity that falls just behind telephone service in importance.
These companies and agencies require full-time connections
to the Internet. In addition, most multilocation organizations
have a wide area network to tie all sites together for access to
corporate databases and e-mail. Residential subscribers and
small businesses, on the other hand, cannot justify the cost of
broadband access that most corporations enjoy, so they are
left with dial-up access. Once users become accustomed to
high-speed access at the office, however, the comparatively
interminable waiting time of dial-up access becomes so
painful that they are willing to pay for a better way.
Better alternatives are becoming available, but it is difficult
to know what to believe. Service providers’ advertisements
often make extravagant claims of access speed based on ide-
alized conditions that most users will not experience. Horror
stories of lengthy delays and inept technicians abound, many
of which are unfortunately true. The objective of this book is
to provide users with information about the various access
technologies, what to expect from them, and where they fit.
Let’s begin by looking at the default method that most users
employ, the public switched telephone network (PSTN).

The Public Switched Telephone


Network
From its inception in the nineteenth century, the PSTN has
had a simple and straightforward architecture. Subscribers
are connected by copper wire to a central switching system
located in a building known as the central office (CO). The
CO is also known as a serving wire center (SWC), so called
because all of the cables from the subscribers route to this
center. The CO houses one or more switching systems. In the
early days of the telephone these were manually operated, but
today’s switching systems use computer-driven electronic
switches. Older systems are analog, but analog technology is
obsolete and is being replaced by digital switches as eco-
nomics permit. Large metropolitan areas have multiple wire
Introduction to Access Technologies 5

centers that are connected by groups of local trunks, as Fig.


1.1 shows. A trunk is any circuit between switches. All users
share trunks and the central switching fabric within the CO.
These shared pathways are engineered to a low probability of
blockage, but calls may be blocked during unusual calling
peaks such as disasters. If a call cannot be completed because
of blockage, the network returns a fast busy signal.
This point is key to understanding access technologies: the
bandwidth restrictions in the PSTN are not in the local loop.
The switching and trunking networks limit the bandwidth to
that needed for a voice session. Except for longer telephone
loops, the local loop has far more bandwidth than a voice
session requires.
The local network connects to the long-haul network over
access trunks. The interexchange carriers (IXCs) provide
trunks from their switching systems to the LECs and inter-
connect their switches with intertoll trunks. Today, virtually
all of these trunks ride over fiber-optic facilities. For more
information on how the optical network functions, refer to the
I-book Optical Networking, available at http://shop.mcgraw-
hill.com/cgi-bin/pbg/indexebooks.html. The optical backbone
provides a high-quality facility with low error rates that is
used by voice and data alike.
Before optical networking became prevalent, those con-
structing data networks had little choice but to obtain analog
circuits from the public telephone network. These circuits
were identical to telephone circuits except that they were not
switched. Instead, they were connected directly to form dedi-
cated or private line circuits. In prefiber days the circuits were
derived over analog microwave or coaxial cable. In either
case, for data the error rate was high and the bandwidth was
constrained by the voice circuit, which is nominally 0 to 4
kHz—actually about 300 to 3300 Hz. Into this network the
Bell System began to deploy digital toll switches, converting
the analog circuits to digital circuits through a device known
as a channel bank. Gradually, as fiber optics replaced the
microwave and as competitive carriers entered the picture,
channel banks gave way to direct digital connections.
For voice connections, the all-digital network resulted in a
dramatic increase in quality. With analog circuits, noise is
Fax Local Switch

Local trunks

Y
Access trunks
Telephone

FL
SONET/SDH Backbone
Local Switch Toll Switch
Telephone Local trunks
6

AM
TE Local Switch

PBX

Local
Customer Premise Local Local Toll Transmission
Access
Equipment Loop Switching Switch Equipment
Trunks

Figure 1-1 Architecture of the Public Switched Telephone Network


Introduction to Access Technologies 7

cumulative. As amplifiers along the way boost the signal,


they also increase the noise. Digital circuits are periodically
regenerated, which keeps noise at a low level. Coupled with
the immunity to interference that is inherent in fiber optics,
digital circuits ensure voice signal quality. They also lower
the error rate for data to the point that errors may occur only
once in a trillion bits. Digital circuits also increase the band-
width by a factor of three or four compared to the maximum
data capacity of an analog circuit, but that is not enough for
broadband applications such as the Internet. The PSTN is
built on a time-division multiplexing (TDM) model. The
lowest level in this model, a DS-0, operates at 64 kbps, which
is the widest bandwidth the PSTN is designed to switch. If
more bandwidth than that is needed, say for a conference-
quality videoconference, it is necessary to bond multiple
channels together through inverse multiplexing.
The problem is further aggravated by the characteristics of
the local loop. The cable plant in the telephone network was
designed to support analog telephones. Most of the intelligi-
bility in the human voice is contained within the narrow pass
band of 0 to 4 kHz. The local telephone networks in the
world were designed at a time when either a private entity
such as the Bell System and independent telephone compa-
nies in North America or the postal telephone and telegraph
(PTT) agencies in the rest of the world owned everything
including the telephone set. These entities had the objectives
of minimizing investment and controlling maintenance cost.
Therefore, until recently with the advent of speed dial and
analog display telephones, all of the intelligence in the net-
work resided in the CO.
Even as the world has converted to digital, additional
shortcomings of the PSTN remain. For one thing, the digital
circuit hierarchy does not scale well. Voice circuit bandwidth
is 64 kbps. The next step up from a single digital channel is
T1/E1, which has 24 channels in North America and Japan
and 30 channels in most of the rest of the world, with nomi-
nal transmission speeds of 1.5 and 2.0 Mbps, respectively.
The next step up the hierarchy, T3/E3, has transmission
speeds of 45 and 34 Mbps. In the wide area, users can obtain
8 Access Technologies: DSL and Cable

fractional T1/E1 or T3/E3, but these are not generally avail-


able in the local loop. The problem is further aggravated in
North America and Japan by the fact that 64 kbps is not
available in the local loop. For reasons we will explain later,
a digital local loop in countries that use the T1 multiplexing
scheme has only 56 kbps of bandwidth.
Until recently, these constraints affected only businesses.
Residences needing access to information databases had no
choice but dial-up modems. Modems have become increas-
ingly sophisticated and quite inexpensive over the years. A
V.90 modem is theoretically capable of up to 56 kbps of full-
duplex transmission (meaning data is sent in both directions
simultaneously), but various constraints such as noisy loops
and ISP configuration limit the actual transmission speed to
something like 33.6 kbps in many cases. Dial-up modems do
an excellent job of cramming a lot of information into a
small amount of bandwidth, but the limits of the technology
have been reached. No further advances are likely.
Dial-up modems did a reasonably good job when the typical
database provider offered primarily textual information with
limited graphical content, but the World Wide Web changed
that forever. The essence of a Web page is bitmapped graphics,
which transfer slowly across a dial-up connection. The telecom-
munications industry, aware of the deficiencies of the analog
network, began to deploy digital circuits in the 1960s, and a few
years later began to develop the integrated services digital net-
work (ISDN) to provide end-to-end digital connectivity. Basic
rate ISDN (BRI) provides two digital “bearer” channels of 64
kbps each. An external signaling channel of 16 kbps completes
the circuit, which is supported on a single copper cable pair.
Primary rate ISDN (PRI) provides full T1/E1 bandwidth. ISDN
is an improvement over an analog dial channel, but it has sev-
eral drawbacks. First is the lack of general availability. Many
small LECs don’t provide ISDN, and for those that do, the cost
is often high. Furthermore, the LECs do not support the serv-
ice well. The same service representatives who handle plain old
telephone service (POTS) often can’t take orders for ISDN, and
the customer either has to get assistance in applying ISDN or
learn a lot more about the service than most want to know.
Introduction to Access Technologies 9

Perhaps a greater drawback of BRI is the fact that it is just


too slow in today’s arena. Even with both channels bonded
together to give a full-duplex bandwidth of 128 kbps, it still
seems slow to people who are accustomed to high-speed
Internet access. From the ISPs’ viewpoint, any dial-up access
has numerous drawbacks. Modems are expensive to own and
troubleshoot. Furthermore, if the customer remains connected,
as many do, ISDN or analog ports are sitting idle, perhaps wait-
ing for an occasional events such as arrival of an e-mail mes-
sage. Business customers have an even greater problem. The
need for Internet access for the small business is much the same
as for residences except for the fact that many more users need
simultaneous access. Therefore dial-up access is feasible only
for the smallest of businesses. Full-time access is essential for
nearly any business that requires an Internet connection. In
addition, businesses with branch offices almost invariably need
data connectivity for exchanging files and e-mail.
For years, the only alternative was to use dedicated circuits.
Now, broadband connections to the home are becoming
essential for many users. The emphasis has shifted from video
to Internet access. In addition, many people are interested in
working from home at least part of the time. Telecommuting
requires access to files in the office computer, and using a
remote-access server is a poor substitute for being online. The
length of time it takes to download large files over a dial-up
modem makes the remote-access server effective only for
traveling employees who do not have a broadband alternative.
In addition to providing universal access to corporate data-
bases and connection to the Internet, many call centers would
like home users to function as call center agents at least part
of the time.
The motivation for full-time access to the Internet doesn’t
rest only with the subscribers. The LECs are also interested in
getting Internet users off the dial-up network. The PSTN was
originally engineered to support an average connection hold-
ing time of about four minutes. Calls into the Internet are far
longer than that and have upset the engineering calculations
that were used for designing the PSTN. Besides the revenue
potential of providing Internet access via DSL, the LECs can
10 Access Technologies: DSL and Cable

benefit from transferring Internet connections from the dial


network to dedicated circuits.

The Cable Network


The history of cable is only about one-third as long as that of
the PSTN. The first application of cable was in Astoria,
Oregon, in 1950. From that beginning, cable expanded to the
point that it can now reach about 80 percent of the house-
holds in the United States, although fewer than that subscribe
to the service. Early cable systems received distant channels
off the air and repeated them over a coaxial network, or even
open wire transmission facilities, to subscribers who were
unable to receive television directly. That gradually gave way,
however, to the concept of cable as a medium that provides
premium entertainment channels. Figure 1.2 shows the archi-
tecture of a typical CATV system.
The CATV equivalent of the LEC central office is the
headend. Here, signals are picked off the air, downlinked
from satellites, or originated locally. The channels are
applied to modulators that link them to the cable at any com-
patible frequency. The conventional very-high-frequency
(VHF) broadcast band in North America consists of 12
channels, each of which is 6 MHz wide. These cover the fre-
quency range of 54 to 88 MHz and 174 to 216 MHz. The
ultra-high-frequency (UHF) channels 14 to 69 cover the
range of 470 to 806 MHz. Cable has an advantage over con-
ventional television in that ranges that are assigned to other
services can be used to support cable channels. Today’s
cable systems cover the range of 54 MHz to as much as 1
GHz, a bandwidth that can support more than 150 channels.
Most systems stop at 750 MHz, however, reserving the top
250 MHz for future services.
Cable was originally constructed as a one-way broadcast
medium. The transmission direction from the headend to the
subscriber is referred to as the downstream direction. (The
other direction, obviously, is upstream.) As Fig. 1.2 shows,
amplifiers are placed at intervals to overcome the losses in
Amplifiers

Splitter

Headend Trunk Cable Termination


Bridger Amplifier

Splitter

Feeder
11

Cables

Subscriber
Drops Taps

Figure 1-2 Cable Television Architecture


12 Access Technologies: DSL and Cable

the coax, which are high at frequencies approaching 1 GHz.


Amplifiers were one-way in earlier systems, but some sys-
tems were constructed with bidirectional amplifiers. In some
cases the service providers did this to support enhanced serv-
ices that required two-way communication. In other cases
their franchises required them to provide two-way communi-
cation services for the municipalities in exchange for use of
the public right-of-way. Modern two-way systems use the
frequencies below 40 MHz for upstream communications,
providing a downstream bandwidth that is far greater than
upstream.
Cable provides a broadband pipe that can potentially reach
a high percentage of the population in most developed coun-
tries, and would appear to be an ideal medium for data com-
munications. As we will discuss in Chaps. 5 and 6, however,
the shared-medium nature of cable imposes some restrictions
that are costly to overcome.

Wireless Access
Wireless has long played a vital role in telecommunications,
primarily in the long-haul transmission network and in
mobile applications. Satellite services once were—and for
some countries still are—a way of bringing telecommunica-
tions services to locales that lack heavy enough demand to
justify undersea cable. Point-to-point terrestrial microwave
was once the primary means of intercontinental telecommu-
nications service, and still is a convenient way of bypassing
obstructions that are expensive to cross with cable. In the
access market, however, wireless has been impaired by sev-
eral factors. First is the fact that broadband demand didn’t
really develop in the local network until fiber optics was
available and the quality of fiber is so much higher that is
always preferred if it is economically feasible.
The most limiting aspect of wireless has been the lack of
available frequencies. Microwave got its start from radar
technologies that were first employed in World War II and
then converted to commercial telecommunications service
Introduction to Access Technologies 13

after the war. The lowest microwave bands were developed


first because the technology was easy to use and the trans-
mission characteristics are superior to those of the higher
bands, where rain, fog, and obstructions can limit the usabil-
ity of the medium. Microwave frequencies are a finite, lim-
ited resource, and competition for the available wireless
spectrum is intense.
In recent years, the low-band microwave spectrum center-
ing around 2.4 GHz has been released for telecommunica-
tions services. The common carrier personal communications
service (PCS) occupies two sub-bands, with unlicensed spec-
trum between the two bands. One of the objectives of splitting
the band in this manner is to enable transceivers to hop easily
from the licensed PCS band to the unlicensed spectrum that is
available for private wireless use. A single instrument can
serve users in both bands, even enabling users to move from
a private system to the public network without interrupting a
session. This same spectrum plus others are available for pub-
lic and private wireless use, including access services, as we
will discuss in later chapters.
Wireless appears at first glance to be an ideal replacement
for the telephone local loop. So far, however, most wireless
local loop (WLL) application has been in developing coun-
tries. In the developed nations wired services are in place,
and there has been little impetus to replace them.
2
Data Transmission Fundamentals

T he subject of access technologies revolves primarily


about data transmission. The traditional networks
work fine for voice, for which fixed-bandwidth circuits are
perfectly adequate, but most data applications need some-
thing more. The purpose of this chapter is to discuss the
nature of data transmission for those who may not be famil-
iar with the details. Those who are may choose to move to the
next chapter.
Bandwidth requirements for data depend on the nature of
the application. Some applications are bursty—that is, they
need a substantial amount of bandwidth for a short period of
time and then need next to nothing. Internet access is a good
case in point. When the user downloads a graphic file, it typi-
cally puts a heavy instantaneous demand on the network. After
the file is received, the user may browse it on a local personal
computer (PC) and put no demands on the network until the
15

Copyright © 2002 by The McGraw-Hill Companies, Inc. Click here for Terms of Use.
16 Access Technologies: DSL and Cable

next download. Other applications—video, for example—


have a steady flow. The video coder-decoder (codec) is pump-
ing information in both directions simultaneously.
This leads to a second characteristic of data flow: whether
the application is symmetrical or asymmetric. An asymmet-
ric flow has more data flowing in one direction than the
other. Internet browsing is asymmetric. An upstream infor-
mation request takes a relatively small amount of bandwidth
compared to the response the network delivers. E-mail is
largely symmetrical in that information flows in both direc-
tions in about the same amount. In between are a variety of
other applications. For example, voice applied to a data net-
work is somewhat bursty, and flows in only one direction at
a time. The ping-pong nature of voice sessions results
because one person speaks while the other listens.
Another characteristic of a data session is the amount of
Y
delay it can tolerate. Response time in a data network is the
FL
elapsed time from pressing the enter key following an infor-
mation request until the response arrives. Most data applica-
tions are tolerant of delay within limits. In practice, the user
AM

is most likely to lack tolerance for slow response, but some


data protocols time out if the response is delayed excessively.
Furthermore, some applications, such as voice and video, are
TE

real-time. When delay exceeds certain bounds, the session is


unusable because the participants tend to talk over one another
and miss the short listening responses that indicate that the
message is understood. When voice and video are applied to
data networks, quality must be controlled tightly or the ses-
sion becomes unusable. These characteristics are discussed
more fully in the I-book Voice and Video over IP, available at
http://shop.mcgraw-hill.com/cgi-bin/pbg/indexebooks.html.

Modulation Methods
Data is passed across a network by having its signal encoded
on the transmission medium through a process known as mod-
ulation. Baseband modulation applies the signal directly to the
physical facility. A telephone system, for example, applies the
Data Transmission Fundamentals 17

signal directly to copper wires using amplitude modulation.


By contrast, a CATV system divides the bandwidth of the
medium into frequency segments and modulates the signal
onto a carrier frequency. The carrier itself carries no intelli-
gence; it is merely a frequency within the pass band of the
transmission medium that can be modulated to carry informa-
tion. In a CATV system, multiple carriers or channels are used,
a process that is known as frequency-division multiplexing
(FDM) of a broadband medium. Technically, the International
Telecommunications Union (ITU, www.itu.int) defines broad-
band as any bandwidth greater than primary rate interface
(PRI), but the term is widely used in conjunction with DSL
bandwidths that are less than T1/E1.
Three different parameters of an electrical wave can be
modulated: frequency, amplitude, and phase. Early modems
used frequency-shift keying (FSK), in which one frequency
was assigned to a 0 and the other frequency to a 1. High-
capacity modems use a complex combination of phase and
amplitude modulation. Figure 2.1 shows how these modula-
tion schemes can convey information at a higher rate. For
example, 2 binary 1 quaternary (2B1Q) coding, shown on the
left side of the figure, assigns a 2-bit code to each of four
voltage levels. Phase-shift keying (PSK) assigns a 2-bit code
to each of four different phases. Since there are four combi-
nations of 2 bits, i.e., 00, 01, 10, and 11, any 8-bit binary
number can be encoded by sending a combination of four
phase shifts. These changes in state are known in the indus-
try as symbols, or sometimes by the old telegraph term baud.
Amplitude shifts may be combined with phase shifts to
form a modulation method known as quadrature amplitude
modulation (QAM), as shown in the other two diagrams.
QAM is a combination of amplitude and phase shift. With
QAM, two carriers are phase-shifted 90˚ with respect each
other, a condition described as quadrature. Each carrier is
amplitude-modulated with half the data. For example, a 16-
QAM signal could have eight phase and two amplitude points,
or four phase and four amplitude points, to detect. The eight-
level QAM signal constellation in Fig. 2.1 shows how bits are
transmitted three at a time. Complex QAM modulation
10 111 000
10 00
11 110 001
18

01 Carriers 101 010


11 01
00
100 011
2B1Q
4-Level QAM
8-Level QAM
Figure 2-1 Modulation Constellations
Data Transmission Fundamentals 19

schemes require circuits that are relatively free of noise and


distortion. In addition, many modems employ forward error
correction (FEC). FEC systems transmit redundant informa-
tion that the receiver uses to determine bits that are in error
and to change these bits to what they should have been. In the
final analysis, the error-correcting protocol employed by the
end devices ensures that the signal is received free of errors.

T1/E1 Carrier
The basic digital facility over which virtually all voice and
data signals pass is known as T1/E1 multiplex or carrier.
Data signals can be connected directly to a digital circuit
through a device known as a channel service unit (CSU).
Analog signals are digitized through a process known as
pulse code modulation (PCM). Nyquist’s law states that if the
amplitude of an analog signal is sampled at twice the highest
frequency it contains, the samples can be used to reconstruct
the original signal with a reasonable degree of fidelity. Since
the highest frequency in the audio pass band is 4 kHz, this
means that a telephone signal is sampled 8000 times per sec-
ond. Each sample is scaled into an 8-bit word through a
process called quantizing. This signal, which is known as a
DS-0 in North America, comprises a 64-kbps signal, which
is derived from the product of 8000 samples per second × 8
bits per sample. In North America, 24 DS-0s are time-divi-
sion multiplexed into a T1 frame. The E1 frame in Europe
consists of 32 circuits, of which two are used for signaling
and 30 for information. Although the T1/E1 signal is chan-
nelized into 24 or 30 channels for voice, it can be obtained
unchannelized for wideband data. In ISDN terms, a
T1/E1signal is known as a PRI.
PCM is not a new process. It was developed in England in
1938, but it wasn’t practical with vacuum tube technology
because of the problems of power drain and need for floor
space. A decade later, the invention of the transistor brought
PCM to the realm of practicality, but not until the 1960s did
20 Access Technologies: DSL and Cable

digital carrier begin to replace the older analog multiplex


then in use. North American and European digital standards
developed somewhat independently and without any require-
ment for interoperability. Transoceanic circuits were analog,
and nothing was on the horizon to suggest that a lack of inter-
operability would be a problem. Fiber optics changed that.
The DS hierarchy in North America was programmed to go
as high as DS-4, which is 274 Mbps—the highest rate that
coaxial cable was expected to support. When fiber optics
came on the scene in the late 1970s, the industry bypassed
the DS-4 rate without a backward glance. Numerous manu-
facturers were producing fiber-optic multiplexing equip-
ment, but each had its own hierarchy. The lack of uniformity
and interoperability quickly became a problem and led to the
development of the synchronous optical network (SONET)
protocol, known as synchronous digital hierarchy (SDH) in
Europe. Table 2.1 shows the most common SONET/SDH
standards. One benefit of SONET/SDH is that the North
American and European standards meet at OC-3.
The entire Internet backbone is composed of circuits rid-
ing on high-speed fiber-optic connections. The fiber-optic
carriers provide bandwidth in OC-x increments to Internet
providers and IXCs. This bandwidth is time-division multi-
plexed and completely synchronous. By this we mean that a
signal can be extracted from a higher-bandwidth signal if its

Table 2.1 SONET/SDH Hierarchy


SONET Bit Rate SONET SDH SDH
Signal (Mbps) Capacity Signal Capacity

OC-1 51.840 28 DS-1s STM-0 21 E1s


or 1 DS-3
OC-3 155.520 84 DS-1s STM-1 63 E1s or 1 E4
or 3 DS-3s
OC-12 622.080 336 DS-1s STM-4 252 E1s or 4 E4s
or 12 DS-3s
OC-48 2488.320 1344 DS-1s STM-16 1008 E1s or 16E4s
or 48 DS-3s
OC-192 9953.280 5376 DS-1s
or 192 DS-3s STM-64 4032 E1s or 64 E4s
Data Transmission Fundamentals 21

position in the bit stream is known. This synchronicity was


not the case with the older multiplexing methods.
T carrier uses a bipolar line-coding scheme in which alter-
nate ones digits are poled in opposite directions. Each sym-
bol on the line represents 1 bit. ISDN doubles the carrying
capacity per symbol by using 2B1Q modulation. Four differ-
ent voltage levels (one quad or Q) are imposed on the cable
pair with each level representing 2 bits (2B). Figure 2.2 illus-
trates the difference between bipolar and 2B1Q line coding.
This coding scheme uses the entire bandwidth of the cable
pair down to 0 Hz or direct current (DC), which makes it
0
1

Superimposed
Sine Wave
0
0

10
1

00
0

01
1

11
1

Level 2 +

Level 1 +

Level 1 -

Level 2 -
+

-
Unipolar

Bipolar

2B1Q

Figure 2-2 Bipolar and 2B1Q Coding


22 Access Technologies: DSL and Cable

impossible to support voice and data simultaneously. As we


will discuss in Chap. 4, some types of DSLs separate voice
from data with filters, making it possible to retain the analog
voice channel below the data. In the case of ISDN, the sig-
nals are being sent simultaneously in both directions over a
single cable pair. ISDN modems employ echo canceling to
prevent the transmitted signal from being reflected back to
the receiver.
When a T1 signal is applied to a line, it uses a coding
scheme known as alternate mark inversion (AMI). In an AMI
signal, every other ones bit is inverted, as shown in Fig. 2.2.
This coding scheme accomplishes two results. First, in looking
at a sine wave superimposed on Fig. 2.2, it can be seen that the
frequency of the sine wave is half the frequency of a corre-
sponding unipolar signal. This allows greater range between
regenerators. Second, AMI serves as a means of error detec-
tion. If a line hit adds or deletes a ones pulse, it results in two
adjacent ones pulses poled in the same direction, a condition
known as a bipolar violation. Terminating devices such as the
office repeater on a trunk or a CSU on a loop can display bipo-
lar violations as a means of detecting trouble.
T1 carrier is designed as a trunk carrier and does not fit
well into the distribution network. It uses separate pairs for
sending and receiving, and these must be sufficiently sep-
arated, either by sending and receiving in separate cables
or by the use of D screen cables, which separate the trans-
mit and receive directions with shielding. T carrier is
designed to fit the physical structure of loaded trunk
cables. These have inductors spaced at 6000-ft (1830-m)
intervals with end sections of 3000 ft (915 m). To fit into
this spacing, the first regenerator is placed at 3000 ft, with
additional regenerators at every 6000 ft and a final 3000-ft
end section. If the end section doesn’t fit the spacing, it is
built out electrically.
A further factor that must be understood about the North
American digital hierarchy is the effects of its signaling
protocol. When the DS hierarchy was developed, much of
the signaling between central offices was carried over the
Data Transmission Fundamentals 23

talking channel. This concept is called in-band signaling.


Analog circuits use tones and DC to convey signals.
Digital circuits use a protocol that robs the least significant
bit of every 6th and 12th frame for signaling the on-hook
or off-hook status of the trunk. This bit-robbed signaling is
of no consequence to voice or analog data, but it obvious-
ly cannot be tolerated for digital data applied directly to
the line. Therefore, only 7 of the 8 bits can be used for dig-
ital data, resulting in bandwidth of 56 kbps. An option
known as extended super frame (ESF) is available in most
North American and Japanese localities to provide a clear
channel with 64 kbps of bandwidth. This is generally avail-
able only in T1 or greater bandwidths.
T1/E1, and its derivative, 56/64 kbps, are widely used
for access as well as point-to-point circuits. The problem
with T1/E1 access has been a lack of scalability. The
ILECs provide either a single 64/56-kbps circuit or a full
T1/E1. Most ILECs do not provide a fractional T1 in the
local loop. The reason is that the architecture for fraction-
al T1 is identical with that for a full T1. Since the rate is
based, at least to some degree, on the cost of providing the
service, the ILECs have no incentive to provide fractional
T1. As we will discuss the next chapter, however, DSL
changes that situation.

Data Protocols
Data devices communicate with one another by using pro-
tocols, which are sets of programmed instructions that a
processor can execute. Protocols are used to set up sessions
between devices, determining such factors as which end con-
trols the session, what participants are authorized, what
transmission speed and code set will be used, and myriad
other functions. Protocols handle other such functions as
addressing, error detection and correction, and recovery from
network failures. Before devices can communicate, they
24 Access Technologies: DSL and Cable

must confirm that they are prepared to use the same protocols
and resolve how they will apply any optional features. Data
protocols are classified as connection-oriented or connection-
less. In a connection-oriented protocol the path is established
at session setup and remains for the duration of the session.
The telephone network is a prime example. A connectionless
protocol launches data packets into the network, where they
are routed to the destination as an independent unit. Internet
Protocol (IP) is a connectionless protocol.
As with most computer programs, protocols are built in
modules or layers. Each module has a specific function and
has clearly identified application programming interfaces
(APIs) at the boundary so developers can write to and use the
functions of the protocol. Table 2.2 shows the layers and their
functions in the International Standards Organization (ISO)
Open Systems Interconnect (OSI) model. OSI is not intended
to be a complete protocol. Instead, it is a conceptual model
with descriptions of the functions that fit within each layer so
developers can build on its structure.
Table 2.2 shows a brief summary of the functions of each
layer in OSI. Each end of the session uses the same layer
definitions. The sending end of the session passes information
to the receiving end by appending records to the information
block. The receiving end receives the instructions in the block,
executes them, and strips the extra records until only the infor-

Table 2.2 The ISO Open Systems Interconnect Model


Layer Number Layer Name Unit Function
7 Application Packet User application support
6 Presentation Packet Data syntax and encryption
5 Session Packet Session setup and
termination
4 Transport Packet End-to-end transport and
error correction
3 Network Packet Packet routing across the
network
2 Datalink Frame Frame transfer and error
correction
1 Physical Bit Transfer of bits from point to
point
Data Transmission Fundamentals 25

mation block remains free of errors and with all information


correctly sequenced.
Data networks can operate without identifying each of the
layers exactly as defined. For example, TCP/IP, which is the
protocol that supports the Internet, uses layers 1, 2, and 3 of
OSI. IP is the network or layer 3 protocol, but its definition
doesn’t conform exactly to OSI’s layer 3, and Transmission
Control Protocol (TCP), which is the transport protocol, com-
bines many of the layer 4 and 5 functions. The OSI model is
customarily shown with layer 1 on the bottom because it is the
base on which the other layers are built. Two devices can com-
municate using only layer 1 simply by connecting their serial
ports together, but this connection has no ability to signal, cor-
rect errors, send data to multiple addresses, or perform numer-
ous other functions that higher-level protocols accomplish.

LAN Protocols
The operation of local area network (LAN) protocols illus-
trates in practical terms how the OSI model functions. This
will also aid in understanding the operation of 802.11b wire-
less, discussed in Chap. 7. The IEEE 802 Committee, which
developed the LAN protocols, designed the protocols to
operate within the first two layers of the OSI model. The
most popular LAN protocol, Ethernet, is used almost univer-
sally as a LAN protocol to connect to the Internet. Figure 2.3
shows the layered structure of the Ethernet protocol stack.
Ethernet drivers are available to connect to any of the popu-
lar transmission media. Early implementations used RG-8
and RG-58 coaxial cable, but unshielded twisted-pair wire
(UTP) is by far the most common medium. Drivers are also
available for fiber optics and wireless systems.
Ethernet’s link layer is divided into two portions; media
access control (MAC) and logical link control (LLC). The
MAC layer is responsible for controlling access to the medi-
um—which, in the case of Ethernet, is based on contention.
A station with information to transmit listens to the network
to see if it is idle. When it determines that no other station is
sending, it launches an Ethernet frame, which is shown in
26 Access Technologies: DSL and Cable

Higher Layer
Protocols

Logical
Link
Control
LLC
Data Link Layer Media
Access
Control
MAC

Physical
Link
Signaling
PLS
Physical Layer
Attachment Unit Interface

Y Physical Medium Attachment


FL
Medium-Dependent Interface
Medium
(Wire, coax, fiber,
radio, etc.)
AM

Figure 2-3 Ethernet Protocol Stack


TE

Fig. 2.6. If two stations transmit simultaneously, their frames


collide and are mutilated. Any station on the network hearing
the collision sends a jamming signal. This signals the stations
to cease transmitting and back off for a random length of
time before attempting to transmit again. The reason for the
random time is so that stations don’t immediately attempt to
transmit and collide again. As Fig. 2.4 shows, each Ethernet
frame has a 4-byte cyclical redundancy check (CRC) block
immediately following the data block. The transmitting MAC
calculates CRC using a complex algorithm. The receiving
MAC uses the same algorithm to calculate CRC and match-
es it to the received CRC. If the CRCs match, the block is
accepted and acknowledged; if they don’t match, the receiv-
ing end notifies the sending end to retransmit.
The LLC communicates with the higher-order protocols.
It funnels data streams to and from the network layer if one
Start of Frame
Delimiter
27

Destination Source
Preamble Length Data Pad CRC
(1) Address Address
(7) (2) (0-1500) (0-46) (4)
(2 or 6) (2 or 6)

Figure 2-4 Ethernet Frame


28 Access Technologies: DSL and Cable

exists. A network layer is not required in a LAN. If all com-


munication is carried over a closed network such as a single
Ethernet segment, link layer communication is all that is
needed. Segments can also be interconnected at the link
layer. A bridge is a two-element device that can be used to
divide Ethernet networks into multiple segments. The bridge
listens to and learns MAC addresses on each segment and
keeps traffic confined to its home segment unless it is des-
tined for an address on the other segment. An Ethernet switch
is a multiport bridge that connects segments together long
enough to pass a frame between them. If each station has its
own port on the Ethernet switch, collisions are eliminated.

Routing
Since a bridge is a two-port device, it lacks the ability to
make routing decisions. Some bridges may deliberately or
inadvertently develop more than one path between segments.
This condition is prohibited by the Ethernet protocol, and is
prevented by using a protocol known as spanning tree. More
frequently, a routing protocol such as IP is used and the net-
work is connected through routers. The TCP/IP protocol falls
under the blanket of packet switching. The U.S. Department
of Defense initially conceived packet switching as a security
measure. If a message is broken into small packets and sent
through the core network over random patterns, it is difficult
to reassemble the message except at the access point. If the
access circuit is kept short, it is easy to secure. Furthermore,
if the core network has plenty of bandwidth, packets from
other sessions can be interleaved to use the excess capacity.
The problem with a connectionless network is that packets
can arrive with errors or out of sequence, and some method
must be used to preserve the integrity. Packet reassembly is
one function of data protocols such as TCP.
IP operates on top of the Ethernet LLC at layer 3 in the
OSI model, and introduces a second layer of addressing. The
MAC address is always the means by which a station can be
identified. It is permanently burned into the network inter-
Data Transmission Fundamentals 29

face card (NIC) as a unique 48-bit address. The IP address,


on the other hand, can easily change. If the station moves
from one segment to another, for example, the IP address
changes, perhaps by as little as one digit. In addition, IP
addresses are often assigned only as long as the computer is
active. Most networks use a protocol known as Dynamic
Host Configuration Protocol (DHCP) to assign IP addresses
to station as they boot up. A protocol known as Address
Resolution Protocol (ARP) links IP addresses with the cor-
responding MAC addresses.
Once a router is connected to the network, packets can flow
anywhere in the world. Routers understand the IP addresses
and know what to do with any address that is handed to them,
but users rarely communicate by means of IP addresses. More
likely, they use a uniform resource locator (URL), which has
the form of user_name@domain_name.aaa, where aaa is a
suffix such as com, gov, edu, and so forth. The router must
have a way of translating the URL into an IP address. It does
this by accessing a server running a protocol known as domain
name service (DNS).
The process of setting up a session is handled by a fast
interchange of messages among devices on the network.
Although devices could theoretically communicate using
nothing more than IP, it wouldn’t be satisfactory to the appli-
cation. IP is a connectionless datagram protocol. A datagram
is a single unacknowledged packet that flows from node to
node on the network until it finally arrives at the destination.
IP is an unreliable protocol, which means that data delivery
is not guaranteed. If a router doesn’t deliver in time or if it
encounters congestion and can’t deliver the packet, it simply
discards it. Furthermore, IP doesn’t check for errors. In data
communications, a packet with a single bit error is worse
than no packet at all. This need for purity, incidentally, is true
only for data packets and not for other types of packets such
as voice.
To turn IP into a reliable network, we need another proto-
col, and this is where TCP comes into the picture. TCP sets
up a logical path with its peer at the other end of the connec-
tion. The two ends carry on a dialog in which they transfer IP
30 Access Technologies: DSL and Cable

packets across the network complete with error checking,


sequencing, and receipt acknowledgment. Instead of
acknowledging packets one at a time, TCP acknowledges an
entire sequence of packets with a single message. TCP has an
interesting method of controlling congestion. With each
acknowledgment, the receiving end returns a window num-
ber, which indicates the number of packets it is prepared to
receive in the next transmission. In the absence of conges-
tion, it opens the window wide. If it is receiving more pack-
ets than its buffers can handle, instead of dumping packets
(which it is permitted to do), it closes the window, which
informs the sending end to throttle back.
As mentioned earlier, some applications, such as voice
over IP, do not need packet acknowledgment and error check-
ing. Therefore, instead of running under TCP, they run under
a much simpler protocol known as User-Datagram Protocol
(UDP). UDP greatly reduces the number of overhead (nonin-
formation) bits that are transferred across the network. Other
protocols also use UDP. For example, Simple Network
Management Protocol (SNMP), which is used to report
alarms and accept orders, also runs under UDP.
Applications can talk directly to TCP/UDP. For example,
Fig. 2.5 shows the TCP/IP protocol stack, including some of
the standard application protocols such File Transfer
Protocol (FTP) and Simple Mail Transfer Protocol (SMTP).

Asynchronous Transfer Mode


Asynchronous transfer mode (ATM) was initially intended as
a broadband ISDN protocol to replace circuit switching in
the carrier backbone network. In the 1990s it was welcomed
with unbounded enthusiasm by developers who planned to
run it to the desktop so data could flow seamlessly from
device to device without media conversion. ATM delivers
bandwidth on demand and provides service quality that the
application and the network negotiate. The universal protocol
vision has faded as ATM has proved to be complex and
expensive, particularly in face of Ethernet, which has proved
Data Transmission Fundamentals 31

Application Services

File Transfer Protocol (FTP)

Simple Mail Transfer Protocol (SMTP)

TELNET

Hypertext Transport Protocol (HTTP)

Transmission User
Control Defined
Protocol Protocol
(TCP) (UDP)

Internet Protocol
(IP)

Address Reverse
Resolution Address
ARP RARP
Protocol Resolution
Protocol

Datalink

Physical

Figure 2-5 TCP/IP Protocol Stack


32 Access Technologies: DSL and Cable

so durable. Gigabit Ethernet delivers the speed that develop-


ers intended for ATM at a lower cost. Nevertheless, ATM has
developed a role for itself in the local loop. It is the most pop-
ular protocol for multiplexing services on DSL lines, and it
is also used in wireless and fiber.
ATM is a connection-oriented protocol that divides the
data stream into short cells. The cell payload is 48 bytes with
a 5-byte header that steers the cells over a virtual path to the
destination. There are two principal reasons for using short
fixed-length cells instead of the long variable-length packets
that TCP/IP employs. The first reason is that cells can be
switched in hardware, which is faster than routing. The fixed
cell length makes it easier to maintain a high degree of uti-
lization. The second reason is that it is easier to control laten-
cy and jitter, which are critical quality of service (QoS) ele-
ments in real-time applications such as voice and video. If a
cell is lost in transmission, it has a minimal impact on quali-
ty because the amount of voice or video information con-
tained in one cell is negligible.
An ATM header carries a virtual path indicator (VPI) and
a virtual channel indicator (VCI). These correspond to
ATM’s two types of circuits: virtual paths and virtual chan-
nels. Virtual channels are groups of channels between ATM
devices, and virtual paths are groups of virtual channels. A
connection between end points can be provisioned as a per-
manent virtual circuit (PVC) or by switching and signaling to
establish a switched virtual circuit (SVC). A PVC is analo-
gous to a private line. The carrier sets it up and it remains set
up until disconnected by service order. An SVC is set up by
the application and is charged like a telephone call.
Like all modern protocols, ATM is a layered protocol.
User applications communicate with the ATM adaptation
layer (AAL). The AAL is divided into two sublayers, the
segmentation and reassembly (SAR) and the convergence.
The SAR segments outbound traffic and reassembles
it inbound. The convergence sublayer protocols are differ-
ent for the various types of information such as voice,
video, and data. The AAL supports five different classes of
traffic:
Data Transmission Fundamentals 33

1. Class A traffic is constant-bit-rate, connection-oriented


real-time traffic such as voice and video.
2. Class B traffic is variable-bit-rate, connection-oriented
real-time traffic such as packet video.
3. Class C traffic is variable-bit-rate, connection-oriented
traffic such as bursty data.
4. Class D traffic is variable-bit-rate, connectionless traffic
such as datagram services.
5. Class X allows user-defined traffic and timing relation-
ships.

The ATM adaptation layer recognizes that different class-


es of traffic have different bit rates, and provides for four
classes: constant bit rate (CBR), variable bit rate (VBR),
available bit rate (ABR), and unspecified bit rate (UBR).
CBR is the highest class of service, providing the ATM
equivalent of a T1/E1 or T3/E3 dedicated line. It is connec-
tion-oriented and designed for time-sensitive applications
such as voice and video. VBR is also connection-oriented,
and is designed for any application, such as LAN intercon-
nection and frame relay, that requires a variable amount of
bandwidth. ABR is offered as a discounted service to take
advantage of the bandwidth that is left over after CBR and
VBR traffic have been accommodated.
The industry has not settled the issue of whether ATM or
IP should be used as the information transport protocol in the
local loop. As we will discuss in Chap. 9, the passive optical
network (PON) has advocates for both protocols, and the
same is true of DSL. IP has the advantages of being less
complex and having a broad range of applications and soft-
ware drivers to support it. ATM is more complex and expen-
sive, but it has QoS built into the design, and many of the
ILECs use it as the backbone of their DSL networks.

Frame Relay
Toward the end of the 1980s, a new protocol known as frame
relay began getting attention. Unlike fixed circuits, frame relay
34 Access Technologies: DSL and Cable

operates on the principle of access circuits from multiple cus-


tomers that are statistically multiplexed into the carrier’s back-
bone. The service provider can oversubscribe the bandwidth
because users aren’t all transmitting simultaneously.
Oversubscription means that less bandwidth is available in the
network than the potential demand. We’ll have more to say
about frame relay in Chap. 10, but note that each subscriber
node on the network requires an access circuit. The access cir-
cuit is typically T1/E1 or 56/64 kbps and extends from the sub-
scriber’s premises to the carrier’s nearest switch node. These
access circuits are distance- and bandwidth-sensitive—that is,
the faster the circuit and the longer the span, the more it costs.
As we mentioned earlier, fractional T1/E1 is generally
unavailable in the access network. Moreover, the carriers’
switch nodes are usually in large metropolitan areas. In small-
er cities, carriers often have a rating point of presence (POP)
and backhaul the traffic at their expense to the switch node.
Traffic is generally channelized time-division multiplexing
(TDM) and fed through a device such as a digital cross-con-
nect system (DCS) so it can be concentrated into a wideband
circuit to the frame relay switch. A DCS is a digital switch in
which the connections are established in software and remain
connected until rearranged by the carrier’s provisioning
process. Carriers have considerable motivation to avoid chan-
nelized circuits, which, because of the bursty nature of most
data applications, are wasteful of bandwidth. Therefore, frame
relay is a major market for improved access service.
The architecture of frame relay is very much like that of a
predecessor protocol known as X.25 packet switching. The
principles of X.25 are similar to those of frame relay with
some notable exceptions. First, X.25 users often have analog
access circuits. These are dialed or dedicated, but the nature
of X.25 is such that its applications are rarely graphical
because the bandwidth is insufficient. However, X.25 estab-
lishes the principle of packet switching and virtual circuits. A
virtual circuit is one that is implemented over shared band-
width as opposed to being dedicated to the session. Users
deliver their media to the edge of the network, where the car-
rier may packetize it.
Data Transmission Fundamentals 35

X.25 is an edge-oriented protocol. The core of the network


is up to the carrier. It decides how much bandwidth to provide
between nodes and what routing protocols to use within the net-
work. Routing is a matter of determining how packets flow
from one node to another. The route can be fixed, i.e., set up at
the start of the session and left for the duration of the session.
The problem with fixed routes is that a circuit failure or node
failure disrupts the session and congestion slows the session
down. More effective is the ability to reconfigure the session
within the core network. The service provider can use a routing
protocol to enable nodes to exchange information about their
loads and circuit status. Circuit failures and congestion can be
made almost transparent to the user application. X.25 checks
for errors at every node. A packet is not transferred to the next
node until it has been received error-free. This process was log-
ical for circuits that had a high error rate, but when the error
rate is as low as it is with fiber optics in the backbone, node-by-
node error checking is unnecessary. Frame relay leaves it up to
the end user to check for errors.
A frame relay network has three billing elements: the access
circuit, a PVC between end points, and the port into the net-
work. The PVC is defined with a specific amount of band-
width. It represents the minimum bandwidth the carrier agrees
to transport. Frames above that bandwidth are carried if the
backbone has sufficient capacity. If not, frames can be dropped.
Frame relay is a convenient and flexible protocol, but the
access circuit is a problem for both carrier and the customer.
Only two choices are available: 56/64 kbps or full T1/E1. The
carriers pass the access cost along to the user, but they bear
the cost of transporting the data to their switch node. Newer
access technologies such as DSL are advantageous because
they can concentrate the traffic into a broadband circuit
much closer to the customers. Since DSL terminates at the
wire center, frame relay frames can be routed over a cable
pair to the wire center, where they are concentrated onto the
LEC’s backbone and routed over a shared network to the
IXC. The result is a substantial reduction in the cost of
access. Whether this cost is passed along to the customer in
the form of lower rates is up to the IXC.
3
Telephone Local Loop Characteristics

A broad understanding of how the PSTN, and particu-


larly the local loop, functions is a prerequisite to
understanding access technologies. As mentioned in Chap. 1,
the PSTN consists of a network of switches that interconnect
circuits to form end-to-end connections that are exclusive to
the parties for the duration of a session. When one of the
users hangs up, the connection drops and the circuits are
returned to a pool of inactive circuits until they are seized by
the next session. All circuits and apparatus in the PSTN are
shared except for the local loop, which is the cable that
extends from the central office to the subscriber.
The telephone network was initially constructed with the
objective of line sharing. Party lines were once the norm,
particularly in rural areas where the cost of running a line to
each house was prohibitively expensive. When party lines
were prevalent, the cable plant was constructed so that the
37

Copyright © 2002 by The McGraw-Hill Companies, Inc. Click here for Terms of Use.
38 Access Technologies: DSL and Cable

same pair might appear on cable legs running down differ-


ent streets, improving the chances of finding a vacant pair to
serve a subscriber. That architecture, which is known as
multiple plant, often resulted in bridged tap, a situation in
which dead wire is connected across an active cable pair.
Furthermore, the subscriber’s service is connected to a dis-
tribution cable pair by means of drop wire. Drop wire is
bridged across a pair, but the cable extending beyond is nor-
mally not trimmed off, also resulting in unterminated bridged
tap. The party line has gradually surrendered to the individ-
ual line, which relieves subscribers of the contention that
once characterized the telephone system. Some of the ves-
tiges of multiple plant remain, however, and have an adverse
impact on high-frequency transmission.
At the heart of the local telephone network is the wire
center, which is the building in which copper cable and fiber
extending to the subscriber terminate. Copper cable starts
at a cross-connection point called a main disturbing frame
(MDF). The MDF includes protectors that prevent injury to
personnel and damage to equipment resulting from lightning
strikes or crosses with electrical power. The cable is com-
posed of 50-pair groups that are known as binder groups or
cable complements. These are formed into large cables of
up to 3600 pairs. The wire gauge varies from 26 AWG (0.4
mm) to 19 AWG (0.9 mm). Fine gauge (26 AWG) is used
nearest the central office so that more pairs will fit in a 4-in
(10.2-cm) conduit. Each cable pair is constructed with pre-
cise twists and is twisted around its neighboring pairs. The
manufacturing objective is to expose each wire in the cable
to approximately the same amount of potential interference
from signals both within and outside the cable. When cable
pairs are carefully balanced in manufacturing, they are capa-
ble of carrying bandwidths well outside the voice frequency
range. For short distances, as in the EIA/TIA 568B struc-
tured cabling standards, precisely manufactured category 7
twisted-pair wire can carry bandwidths of 600 MHz for dis-
tances of up to 100 m.
Wire centers in most communities have existed in the same
location for decades. Wire centers are located to achieve a rea-
Telephone Local Loop Characteristics 39

sonable balance between cost and transmission quality. The


cable connecting the central office to subscribers is known as
feeder cable. At distribution points the cable is connected to
the distribution cable, which terminates at the subscriber’s
premises. A serving area interface is a large cross-connect
point where feeder pairs are connected to distribution pairs.
Figure 3.1 shows the general structure of cable distribution.
Collectively, the cables, conduits, and poles are known as out-
side plant.
Cables are constructed with a layer of shielding material
surrounding the copper pairs. An outer sheath—typically of
polyethylene—surrounds the shielding, keeping the pairs
reasonably free of moisture. Many cables are kept under air
pressure or filled with a gel substance as a further defense
against moisture. If cable pairs do get wet, much of their
protection against interfering signals is lost. While cable
manufactured in the last three or four decades is normally
impervious to water and other elements, older cable is not.
Older cables are built with lead sheath and paper insulation
on the pairs. Splices deteriorate in time, and as a result the

Serving
Area Subscriber
Interface Drops
Distribution
Cable

Feeder Cables

Central Office

Serving
Area
Interface
Distribution
Cable

Subscriber
Drops

Figure 3-1 Telephone Cable Plant


40 Access Technologies: DSL and Cable

cable is incapable of carrying high-bandwidth signals. Wire


is carefully twisted to reduce crosstalk between pairs. Cable
pairs are formed into 50-pair complements.
Copper cable provides a DC path from the subscriber to
the CO. The CO feeds −48 V of DC toward the subscriber.
When the telephone is inactive or on hook, no current flows
in the line. When the subscriber lifts the handset off the hook,
current begins to flow. This change in current signals the
central office to return a dial tone, which signals the user to
begin dialing.
Some large businesses are served over fiber optics, but the
default method of connecting subscribers to the serving CO
is over a copper cable pair. At the heart of the CO is a switch-
ing system. Since most systems today are digital, an analog
signal originating at the subscriber’s premises must be digi-
tized. This is done at the line interface in the switch. ISDN
loops are an exception. With ISDN, digitizing is done in the
telephone set. Many large companies have private branch
exchanges (PBXs) that also connect to the CO over digital
loops. Unless these loops are fiber optics, the digital signal is
connected to the CO over some form of DSL technology.
Central offices are connected to one another by means of
interoffice trunks. These trunks are digital circuits with 64
kbps of bandwidth. It is important to understand that both
switched and dedicated circuits are identical in all respects
except that dedicated circuits are permanently connected
rather than switched. This process is called provisioning in
the telephone industry, and it involves connecting circuits by
means of jumper wire or logical connections through a DCS.

Telephone Electrical Concepts


To understand why access technologies behave as they do, it
is necessary to have some appreciation of electrical terms.
Any circuit, such as a twisted-wire cable pair or a coaxial
cable, has inherent electrical characteristics such as resistance
and impedance. Direct current flows in a predictable manner
in response to Ohm’s law, which states that the amount of
Telephone Local Loop Characteristics 41

current flow is directly proportional to the voltage or electri-


cal pressure and inversely proportional to the resistance. In
mathematical terms this is denoted as I = E/R, where I is cur-
rent, E is voltage, and R is resistance. When an alternating
current (AC) signal from a human voice or a high-speed
modem is applied to the circuit, the phenomenon of imped-
ance arises. If we turn Ohm’s law around to express it as R =
E/I, we can see that resistance is the ratio of voltage to cur-
rent. In alternating current terms we express resistance as
impedance, and denote it with the letter Z.
AC flows are affected by circuit elements of capacitance
and inductance that do not affect DC. Without going into
detail, suffice it to say that any transmission line has capac-
itive and inductive qualities that are distributed along its
length. Capacitance and inductance impose opposite effects
with respect to frequency. As shown in Fig. 3.2, the higher
the frequency, the greater the resistance to AC an inductor
imposes. (It’s actually called reactance, but for our purposes
we can think of it as resistance). Capacitance, on the other
hand, offers less resistance as frequency increases. Since
capacitance is distributed along the length of the line between
the two wires of a pair, the higher the frequency, the more the
capacitance takes on the electrical effects of a short circuit,
which is a direct connection between the two sides of a cable
pair. If a transmission line is infinitely long, it begins to
assume a characteristic impedance, which is the ratio of volt-
age to current along the line. This characteristic impedance
is stated in ohms (Ω), which is the same unit for DC resist-
ance or AC impedance. For example, category 5 unshielded
twisted-pair wire used for LANs has a characteristic imped-
ance of 100 Ω.
Transmission lines are never infinitely long, but if they are
terminated in a circuit that has the same characteristic imped-
ance, the line assumes uniform electrical characteristics. A
line with perfect impedance doesn’t exist outside the labora-
tory. In the real world, many things happen to alter the imped-
ance of lines. In twisted-pair cable, gauge changes, irregular
splices, and moisture are examples of things that can cause
impedance changes. The effects of some of these may be
42 Access Technologies: DSL and Cable

C
nc

ap
ta
uc

ac
d

ita
In

nc
e
Frequency

Reactance
Figure 3-2 Reactive Effects of Capacitance and Inductance

moderate, but an unterminated cable causes a severe irregu-


larity, the effects of which are difficult to predict. When an
electrical wave reaches the end of an unterminated line, some
of the energy is reflected back to the source.
Although the architecture and transmission media of a
CATV network are completely different from those of the
PSTN, the same impedance considerations apply to coaxial
cable as well. Kinks and dents can cause impedance irregu-
larities in the trunk cable, but the main issue is ensuring that
all subscribers are connected to the distribution coax in a
nondisruptive manner. The network must be constructed so
that nothing subscribers do on their premises affects the
integrity of the signal on the backbone. In other words, each
household must be electrically isolated from the network.
This is achieved by using high-impedance taps that allow the
video signal to pass to the subscriber, but that isolate each
subscriber from the rest of the network.
As a signal is applied to copper wire, it is attenuated over
distance. The amount of attenuation depends on the length of
the cable, its gauge, and several other factors of lesser impor-
tance such as temperature. A cable pair looks electrically like
a capacitor across the cable pair. Capacitors pass higher fre-
quencies more readily than lower frequencies As a result, the
higher the frequency, the greater the tendency of the cable to
Telephone Local Loop Characteristics 43

short-circuit the signal, so the greater the attenuation. This


makes little difference to a voice circuit up to a point, but the
loss of high frequencies results in a loss in intelligibility. To
compensate, ILECs connect load coils in series with cable
pairs when the cable length exceeds more than 18,000 ft
(5,500 m) from the central office. A load coil is an inductor
that has high attenuation to frequencies above the voice band.
High-frequency services such as DSL cannot operate through
load coils, so they must be removed if the cable is to support
high-speed services.
As discussed earlier, bridged tap results when an untermi-
nated length of wire is connected in parallel with the talking
path. It also exists when a length of cable extends beyond the
termination point. Most types of DSL cannot operate with
more than a small amount of bridged tap. Removing it is not
a trivial matter because it may involve opening splices to cut
off the offending wire. Furthermore, the ILECs’ records are
not always reliable enough to show exactly where it exists.
Before DSL can be provided, it is often necessary for the
ILEC to remove load coils and bridged tap, an operation that
we will refer to as conditioning.

Crosstalk Considerations
The local loop was designed and constructed to support
voice-frequency signals. The use of one pair to serve one
subscriber does not begin to use the bandwidth capacity of a
cable pair. For short distances, plain copper wire is cost-
effective, but for longer subscriber loops or for large concen-
trations of loops into a single subscriber, multiplexing can
increase the capacity. Therefore, the ILECs sometimes use
multiplex equipment in the local loop. Several factors insti-
gated this. One was the development of digital loop carrier
(DLC), which allows the ILECs to extend subscriber lines to
the outer reaches of the wire center using a system similar
to T1/E1. The DLC is housed in an environmentally con-
trolled enclosure and fed from the CO with fiber optics. This
is a cost-effective alternative to using coarser cable gauges
44 Access Technologies: DSL and Cable

and range extension. The latter amounts to electronic ampli-


fication of the voice signal and boosting the DC signaling
range of the switch. The second factor was demand for direct
IXC access. Divestiture took the ILECs out of the long-haul
toll business and gave this function to IXCs. When the ILEC
switches the call from the subscriber to the IXC, The IXC
pays time-sensitive access charges. To avoid access charges,
large users bypass the ILEC with dedicated lines. The most
economical way of providing toll bypass is with T1/E1.
As discussed in Chap. 2, T1/E1 was designed as a short-
haul interoffice trunk carrier. Interoffice trunks in a metro-
politan network were originally provisioned on copper cable
that was usually loaded. T1/E1 was designed so that regener-
ators would fit in manholes, which are spaced at 6000-ft
(1830-m) intervals with 3000-ft (915-m) end sections. An
office repeater feeds power to midspan regenerators. Much
of this trunking network has been replaced by fiber optics.
When the demand began to develop for T1/E1 services in
the local loop, the logical approach was to use T1/E1 carrier
or DLC on a cable system that was designed for voice fre-
quencies. Digital signals contain high-frequency components
that are subject to crosstalk, which is the reception of any
unwanted signal that is induced from another source.
Consider the diagram in Fig. 3.3. As a T1 signal is injected
into the transmit pair of a circuit, it is at a high level com-
pared to the signal on the receive pair, which has been atten-
uated by the transmission loss from the previous regenerator.
Unless precautions are taken, an excessive amount of the
transmit signal radiates from the source and is coupled into
the receive path of its own circuit or others in the cable. This
path is known as near-end crosstalk (NEXT). Crosstalk is
also possible at the far end of the circuit, but, as we have
seen, the receive signal level is attenuated compared to the
transmit level, so the coupling from the receive pair into
the transmit pair is a minor concern.
The best way to eliminate NEXT is by separation. If sep-
arate transmit and receive cables are available, this is the
preferred method. Another method is by using so-called D
screen cable, in which transmit and receive complements are
High Signal Level
Low Level Signal

Near End
Transmitter
Cable Pairs
Far End
Receiver
45

Far End
Transmitter

Near End Far End


Near End Crosstalk
Receiver Crosstalk

Figure 3-3 Crosstalk


46 Access Technologies: DSL and Cable

enclosed in separate shielding. Crosstalk may also be reduced


by physical separation of the complements carrying T carrier.
In a large cable with multiple binder groups, cable pairs used
for other services can provide the separation, but generally
only one T1/E1 service can exist in each binder group.

Echo Considerations
Another impairment that must be considered is echo. Echo
occurs whenever a four-wire circuit is converted to a two-wire
circuit. Trunks are almost always four-wire because they are
provisioned on multiplex equipment that inherently has sepa-
rate transmit and receive paths. Local loops were designed
as two-wire to conserve cable. Many older switches were also
two-wire, but all digital switches are four-wire, meaning the
Y
transmit and receive directions of a session are separated. A
FL
device known as a hybrid does the four-wire-to-two-wire con-
version. A hybrid works on the principle of balancing the
impedances of the two-wire and four-wire loops. As shown in
AM

Fig. 3.4, the four-wire transmit and receive sides of the hybrid
feed into two of its ports. The two-wire cable pair and a bal-
ancing network feed into the other two ports. If the balancing
TE

network is a perfect impedance match for the cable pair, then


the four-wire legs are completely isolated. The balance is
never perfect, however, because cable pairs have widely vary-
ing characteristics. Therefore, a balancing network is selected
as a compromise that is less than perfect. Some of energy
from the receive side of the hybrid feeds into the transmit side
shown in Fig. 3.4 as the echo path.
Echo is present in every connection, but it is noticeable
only if there is a significant time delay before the talker’s
voice returns. A short time delay of less than about 45 ms is
heard as sidetone, the faint feedback that makes the tele-
phone sound alive when you talk into it. Echo is not a prob-
lem in the local exchange network unless an application such
as voice over IP is being used. Where echo is a problem, it
is controlled by means of echo cancelers. An echo canceler
compares the transmitted voice with the signal on the return
Telephone Local Loop Characteristics 47

Echo Path
4-wire Receive
Pair

Balancing 2-wire Pair


Network

4-wire Transmit Hybrid


Pair
Figure 3-4 Hybrid Circuit

path, and where they are identical the echo signal is canceled
out. Echo cancelers are built into the IXC’s network, and are
not needed in the local exchange.
T carrier is excellent in the trunking network where
crosstalk can be controlled by use of dual cables or shielded
cables. It is not an ideal medium for the subscriber loop, how-
ever, because of crosstalk limitations. The line-coding method
of 1 bit per symbol change is not as efficient as newer line-
coding methods that are used in ISDN and other services.
Therefore, T carrier is generally replaced in the local loop by
one of several DSL technologies, which are the subject of the
next chapter.
4
Digital Subscriber Line Access Technology

D SL is the prime prospect the ILECs have for capital-


izing on the information age. The millions of miles
of cable that are in the ground and strung from poles are func-
tionally obsolete, yet they are an asset that has enabled the
ILECs to preserve their monopoly in the face of incipient
competition. POTS has about run its course as a dynamic
growth business. Nearly every household in the developed
world has at least one line now, and cell phones are fulfilling
much of the demand for additional residential lines. Growth
in the telephone industry follows the trend of business and
residential development to a large degree—steady, reasonably
predictable, and only marginally profitable.
Although the revenues from local service are enormous
(estimated to be in excess of $70 billion per year in the
United States alone), the service is regulated. While profits
are all but guaranteed, regulation prevents the ILECs from
49

Copyright © 2002 by The McGraw-Hill Companies, Inc. Click here for Terms of Use.
50 Access Technologies: DSL and Cable

achieving the profitability of more exciting industries. About


all the ILECs can count on to increase their revenue stream
is additional telephone services such as their custom signal-
ing services, which include caller ID and other profitable
services. While these command high margins, consumers
find them easy to drop during downturns in the economy.
DSL, however, is a new and profitable service that enables
the ILECs to enter new markets. The copper cable to support
it is in place, it is inexpensive, and it does the job well. The
bandwidth restrictions that limit modem speed are in the
central office and trunking network, not in the local loop.
From an architectural standpoint, DSL consists of a spe-
cialized modem at the subscriber’s premises, feeding into a
DSL access multiplexer (DSLAM) in the central office. Some
types of DSL ride on top of a POTS line, while others use the
cable pair down to a DC level, which precludes the use of a
POTS line. Figure 4.1 shows the layout of DSL service. The
DSLAM demultiplexes the signal from the subscriber, split-
ting the service so the telephone service connects to the
central office while the data portion feeds into the ILEC’s
backbone network, where it is routed to the service provider.
DSL is not an ILEC monopoly service. ILECs typically did
not lease bare cable pairs until the Telecommunications Act of
1996 required them to make unbundled network elements
(UNEs) available at any feasible interconnection point. The
interface to the cable pair in the central office is such a point,
which makes it feasible for service providers other than the
ILECs to provide DSL service. To obtain access, a service
provider must register as a CLEC. The CLEC has two alter-
natives for physical connection to the cable pair. If space is
available, it can lease space from the ILEC, install its own
DSLAM in the ILEC’s central office, and connect directly to
cable pairs. The other alternative is to run a copper cable
extension from the CO to a nearby building. Since DSL is dis-
tance-limited, this has the effect of reducing the range from
the central office to the DSL subscriber. Several companies,
such as Covad and Rhythms, provide DSL service, but with
the downturn in the economy in 2000–2001, competitive DSL
providers have had difficulty achieving profitability.
Voice Switch

Local Loop
PC
Bay Networks
51

LEC Frame DSL Router


Relay or ATM

DSLAM
ISP
LEC Central Office
Telephone

DSLAM= DSL Access Multiplexer Customer Premises

Figure 4-1 DSL Configuration


52 Access Technologies: DSL and Cable

An attractive aspect of DSL is the dual use of cable pairs


for voice and data service. When competing carriers first
began to emerge, the ILECs refused to permit them to provide
DSL over existing telephone lines. That meant that either
competitors leased a single-purpose cable pair or the cus-
tomer purchased an additional telephone line. The DSL carri-
ers appealed to the Federal Communications Commission
(FCC), and in its 1999 line sharing order, the FCC required
the ILECs to permit competitive DSL carriers to provide
service over existing telephone lines.

DSL Standards
One problem with DSL has been a glut of standards. The serv-
ice comes in an alphabet soup of options going by the names
of HDSL, SDSL, IDSL, ADSL, RADSL, and VDSL, often
collectively described by the term xDSL. Each of these has
different characteristics and fits a different niche in the mar-
ketplace. To add to the confusion, two different modulation
methods—discrete multitone (DMT) and carrierless ampli-
tude phase (CAP)—are used. The muddle isn’t quite as severe
as it seems at first. High-speed DSL (HDSL), and to some
degree single-pair DSL (SDSL), are used by the carriers them-
selves to provision point-to-point T1/E1, so the subscriber is
insulated from these. The industry itself drops the first letter in
its promotions, so users don’t have to know what variety of
DSL they are getting. Table 4.1 lists the types of DSL and their
approximate transmission ranges.
The inability of ILECs to respond to service requests has
hampered DSL’s growth. Many users have complained about
long delays in receiving services they have ordered and then
difficulties in getting the service to work. Technicians must
sometimes make multiple trips to get the PC and modem prop-
erly configured. To resolve these issues and to bring some
order to the chaos of multiple standards, the industry formed
the Universal ADSL Working Group in 1998 to develop a sin-
gle standard. The carriers’ principal objective was to make it
easy for subscribers to purchase self-configuring modems and
Table 4.1 DSL Types
Type of Upstream Downstream Range in
DSL Acronym Bandwidth Bandwidth ft (m)
Asymmetric ADSL 16–640 kbps 1.5–9 Mbps 18,000 (5,500)
53

High-bit-rate HDSL 1.544 or 2.048 Mbps 1.544 or 2.048 Mbps 12,000 (4,000)
ISDN IDSL 144 kbps 144 kbps 18,000 (5,500)
Single-pair SDSL 1.544 or 2.048 Mbps 1.544 or 2.048 Mbps 12,000 (4,000)
Splitterless G.lite 16–640 kbps 1.5–6 Mbps 18,000 (5,500)
Very-high-bit-rate VDSL 1.5–2.3 Mbps 13–53 Mbps 1,000–4,500 (330–1,500)
54 Access Technologies: DSL and Cable

set up the service themselves. In the industry vernacular,


“truck rolls” (i.e., dispatching a technician to the subscriber’s
premises) were hampering profitability. The result, a protocol
known as G.lite, is acceptable within limits, but it has draw-
backs that prevent its becoming a universal solution.
Today users can purchase modems that operate on either
DSL or cable and, theoretically at least, can set the service up
without a technician, but the results have been mixed. Some
of the difficulty lies in the nature of the ILECs’ plant. As dis-
cussed in Chap. 3, bridged tap is a major problem because
the cable records don’t always show whether it exists, or, in
some cases, where it is. Loading is not as great a problem
because load coils are generally placed on loops that are
more than 18,000 ft (5,500 m) long, and these loops are too
long for most varieties of DSL.
The solution to the loop length limit is, to some degree,
resolvable. For years, ILECs have used DLC to serve sub-
scribers that are located outside the range boundaries of the
central office. These are often known as GR-303 remotes
after the BellCore (now Telcordia) standard for interfacing
DLC to central office switches. Where the T1/E1s support-
ing DLCs are provisioned over fiber from the central office,
it is not difficult to equip the copper cable extension for
DSL. One alternative is simply to install a DSLAM in the
enclosure housing the DLC equipment and separate voice
and data at that point. Specialized DLCs are available that
are equipped with combination line cards and with the
DSLAM function built into the system. They do not, how-
ever, work with legacy DLCs. Enclosures in which both
POTS and DSL are collocated are often called neighbor-
hood gateways. Part of the motivation for using neighbor-
hood gateways is the ability to use very-high-bit-rate DSL
(VDSL), which is capable of handling video and high-speed
data signals within a limited range of 1000 to 2000 ft (330
to 660 m).
Other architectures are also in use. Some CLECs offer
voice over DSL (VoDSL), in which case the customer’s voice
and data signals are fed into an integrated access device
(IAD) at the subscriber end. At the central office end, the sig-
Digital Subscriber Line Access Technology 55

nal may feed into a media gateway, which in effect is a front


end that separates the voice and data signals. The alternatives
are many and varied, and more are yet to come.

DSL Technology
DSL lives in an environment that was originally designed for
voice transmissions. As discussed in Chap. 3, the principal
limiting factor is crosstalk. Since cable plant is called on to
carry high-speed signals that were not intended for copper
wire, care must be used to prevent interference from various
types of line signals. DSL can be categorized by two primary
criteria: the line coding and modulation methods.

Coding Methods
HDSL and ISDN DSL (IDSL) use the entire bandwidth of
the cable pair by applying a DC signal directly to the cable
pair. These types of DSL cannot coexist with a POTS line.
Other types of DSL separate voice from data with filters
and do permit the use of a POTS line. High-speed modems
encode information in symbols instead of a raw bit stream.
By using complex coding schemes, a single symbol can be
made to represent some number of consecutive bits.
As we have discussed, T1/E1 uses a simple coding scheme
that encodes only 1 bit per symbol. ISDN improves on that
by using 2B1Q line coding. Each pair of bits represents one
quaternary signal. Higher modem speeds are achieved with
QAM. With QAM, two carriers, each having the same fre-
quency, are phase-shifted 90˚ with respect each other. One
signal is called the I signal and the other the Q. Each carrier
is amplitude-modulated with half the data. The two signals
are combined at the source and transmitted to the receiver,
where the signals are separated and demodulated to produce
the original data stream.
Early QAM modem systems used eight phases and two
amplitude levels to transmit 4 bits over each symbol. These
56 Access Technologies: DSL and Cable

are known as 8-QAM. An interesting form of error correction


can be obtained by increasing this to 16-QAM. Here, 12
phase changes are used together with two amplitude levels.
Amplitude changes occur on only four of the phase changes.
The result is that both a phase change and an amplitude
change are valid in some symbols but not in others. Trellis
coding defines the way in which signal transitions are allowed
to occur. Signal transitions that do not follow the expected
pattern are detected as errors. By using finer phase and ampli-
tude changes, QAM can be extended to 6 bits per symbol (64-
QAM) and 8 bits per symbol (256-QAM).

Modulation Methods
The second issue is the modulation method. DSL uses either
DMT or CAP. DMT separates the spectrum above the voice
Y
band into 256 narrow channels called bins. Each bin is 4 kHz
FL
wide. Data is modulated onto each channel using QAM modu-
lation with a 4-kHz symbol rate, resulting in up to 60 kbps per
AM

bin. A simple splitter separates voice from the data signal. The
modems detect which channels may be impaired and spread the
data to unimpaired channels. The upstream direction can use
bins 6 to 38, which is about 25 to 163 kHz. Bins 33 to 255 (142
TE

to 1.1 MHz) are used downstream. This method of modulation


is inherently rate-adaptive because it uses all the available chan-
nels and ignores those that have poor transmission quality. For
example, if an AM broadcast station is interfering with a band
of frequencies, these bins are skipped. DMT is the American
National Standards Institute (ANSI) standard T1.413. The
European Telecommunications Standards Institute (ETSI) and
ITU also recognize the standard.
CAP is closely related to QAM. The data rate is divided
in two and modulated onto two orthogonal carriers phase-
shifted 90˚ to one another. CAP systems use a combination
of amplitude and phase states to encode the data. The two
channels are combined, fed into a digital-to-analog converter,
and transmitted. A simple filter separates voice and data.
Hybrids and echo cancelers are not needed with CAP systems
because the two directions of transmission are separate. The
Digital Subscriber Line Access Technology 57

modems use adaptive equalization to overcome, to some


degree, the effects of bridged tap, gauge changes, and splices.
CAP is a proprietary protocol, but despite its lack of standard
status, it has a large following.

Spectral Compatibility
The degree to which DSL signals can coexist in the same
cable is a matter of concern to the ILECs. The limiting fac-
tor is usually near-end crosstalk. NEXT can be controlled by
limiting the nonvoice signals that are assigned to particular
groups of pairs. The NEXT problem is most severe nearest
the central office because the cable density is highest there.
DSL systems therefore use the lowest frequencies for the
upstream direction to compensate for the fact that the higher
the frequency, the greater the coupling from one pair to
another. To limit interference, each DSL standard includes a
power spectral density (PSD) mask that describes the PSD by
frequency in the upstream and downstream directions.
The degree of crosstalk that one DSL type imposes on
another defines their spectral compatibility. The degree of
compatibility depends on such issues as the spectrum the
DSL type uses and its placement in the cable.

Types of DSL
Each of the DSL types listed in Table 4.1 has a particular
function. This is usually not a matter of concern to the sub-
scriber because the service provider determines which type
will be used. This section briefly discusses the characteristics
and applications of each type.

HDSL
T1/E1, which is designed as a trunk carrier, doesn’t fit well
into exchange cables. Not only is repeater spacing critical, but
also, crosstalk considerations limit the number of systems that
can be installed in a 50-pair cable complement. HDSL is
designed to overcome these limitations. The ILECs have used
58 Access Technologies: DSL and Cable

HDSL for years to provide T1/E1 service over two twisted-


pair wires up to 12,000 ft (4,000 m) long. This enables them
to avoid the regenerators that T1/E1 requires, which are
expensive and difficult to install on many loops. Terminating
equipment is installed both in the central office and at the sub-
scriber’s premises. Unlike other DSL versions, HDSL does
not support regular telephone service on a cable pair.

SDSL
SDSL (also known as symmetrical DSL) serves the same
market as HDSL, with two important functional differences.
Instead of dual cable pairs, it provides T1/E1 service on a sin-
gle cable pair and it also derives a POTS line under the data
signal. It transmits and receives in the same band of frequen-
cies using an echo-canceling protocol, making it susceptible
to NEXT. SDSL uses 2B1Q line coding that is adopted from
basic rate ISDN.

IDSL
IDSL provides 128 kbps of bandwidth using the same proto-
col as BRI. The main difference is that IDSL terminates in a
router instead of an ISDN switch port. This precludes it from
carrying a voice signal. This technique has limited application
because its limited bandwidth cannot compete effectively
with other DSL technologies.

VDSL
VDSL was developed as a means of providing video on
demand in fiber-to-the-curb implementations. Fiber is used to
bring the video signal to an access node with twisted-pair wire
supporting the last span. VDSL supports data rates as high as
52 Mbps, which is enough to carry a DS3 over short spans of
copper wire. Rates of 52 Mbps can be supported over a 1000-
ft (330-m) range, but the rate drops to about 13 Mbps beyond
4500 ft (1500 m). The technology is asymmetric, with upstream
speeds of 1.5 to 2.3 Mbps. It can be overlaid over POTS lines.

ADSL
Asymmetric DSL (ADSL) can carry as much as 1.5 Mbps
downstream over a range of up to 18,000 ft (5,500 m). Data
Digital Subscriber Line Access Technology 59

travels much more slowly in the upstream direction—in the


64- to 640-kbps range. For residential Web surfing, ADSL
is an excellent alternative because the nature of the appli-
cation is asymmetric. Many business applications require
symmetrical bandwidth, however, so ADSL may not be
adaptable in many cases. ADSL is most attractive for casual
users, while SDSL is the most effective for businesses and
telecommuters.
Rate-adaptive DSL (RADSL) is a variation of ADSL. The
DMT modulation method is inherently rate-adaptive, but
CAP is not. When the industry refers to RADSL, it is usually
with respect to a version of CAP in which the modems auto-
matically optimize the line to the highest effective transmis-
sion rate. Rate adaptation compensates for changes that occur
because of weather changes and aging of components. If the
line quality degrades, the modems step down to a lower rate
to preserve transmission integrity.
Standard ADSL requires a splitter at the customer’s prem-
ises. The splitter not only separates the voice and data sig-
nals, it also terminates the line with a constant impedance.
The telephone wire in households is not a reliable termina-
tion, because wire quality varies greatly, unterminated wire
runs are common, and telephone sets being unplugged or
going off hook cause sharp impedance changes. Also, the
telephone sets themselves are of varying quality, and, since
they are connected in parallel across the line, the quantity of
sets on the line affects the impedance. The splitter isolates
the line from these variations.
The Universal ADSL Working Group developed G.lite, an
ADSL variation, to eliminate the need for a splitter. The
objective was to enable subscribers to install the service with-
out the need for a technician visit. G.lite uses the DMT mod-
ulation method and is ANSI standard T1.413. G.lite uses 96
upstream bins, which limits the throughput to about 400 kHz.
The downstream direction may support up to 1.5 Mbps,
depending on the length of the loop to the central office. ATM
is used as the transport medium. G.lite is designed to retrain
itself rapidly in the face of sudden impedance changes, as
when a phone is lifted off-hook.
60 Access Technologies: DSL and Cable

DSLAM
The DSLAM separates voice-frequency signals from data at
the CO end. The DSLAM aggregates data traffic from several
loops into the backbone. In the case of ILEC-provided DSL,
the backbone is usually ATM. Competitive access providers
may use IP. ISPs and other network providers connect to the
network with PVCs.
A multiservice DSLAM can handle traffic from SDSL,
IDSL, and ADSL. It may also support video access in addi-
tion to Internet access. Multiservice capability may also
include multiple service types such as frame relay, ATM,
TDM, and IP running to the subscriber from a single plat-
form. Frame relay is provisioned to the subscriber using
TDM or ATM.

VoDSL
DSL started out to keep voice and data separated by means
of splitters and filters, but several new protocols and prod-
ucts have been developed or are under development to merge
voice and data at the circuit level. The products compress
the voice signal down to a fraction of the 64-kbps bandwidth
of a conventional voice channel. Voice can be compressed to
8 kbps with only a small loss of intelligibility. The quality is
equivalent to a solid cell phone connection. The protocols
add some overhead, so the number of channels is less than a
straight multiple of the compressor output, but the voice-
carrying capacity is substantial. Note, however, that faxes
and modems require an uncompressed channel.
The amount of bandwidth the voice channel occupies
depends on the protocol used. ATM is the most common pro-
tocol. With 5 bytes of header per cell, the overhead is approx-
imately 10 percent. Equipment vendors claim from 16 to 20
voice channels per DSL line, but of course that depends
on the actual data rate the line can sustain and also on the
amount of bandwidth needed for data.
5
Cable Access Technology

C able television facilities pass more than 80 percent of


the homes in the United States, and about 65 percent
of these subscribe to the service. Cable was originally devel-
oped as an entertainment medium, but having a broadband
pipe into the majority of residences makes it an obvious can-
didate as an information medium as well. In fact, in the
1960s the Bell System made a few forays into the cable busi-
ness for just that reason. These were, however, thwarted by
regulatory authorities that opposed adding cable to the Bell
companies’ telephone monopoly. Many of these restrictions
are disappearing, however, as cable companies prepare to
offer telephone service and telephone companies investigate
DSL as a means of delivering video-on-demand. Despite the
fact that technology may make it possible to cross between
the traditional media, there is often a wide gap between what
is possible and what is feasible. Differences in telephone and
61

Copyright © 2002 by The McGraw-Hill Companies, Inc. Click here for Terms of Use.
62 Access Technologies: DSL and Cable

video signals still leave the conventional architectures more


suitable for their traditional markets.
One of the impediments to expanding cable for other than
video service has been the one-way nature of the conven-
tional system. At the time cable systems were constructed,
entertainment was inherently one-way. That is still the case to
a large degree because television is not interactive, but games
and Internet access are. Two-way cable opens the potential
for additional revenues for the system operators, leading
most providers to upgrade their systems. In some cases, fran-
chising authorities compel cable operators to enter the two-
way business. In exchange for permission to use the public
right-of-way, some municipalities require cable operators to
provide two-way networks for public purposes.
Although cable has plenty of bandwidth for voice and
data, it has some inherent drawbacks. At the time the early
two-way systems went into operation, the industry lacked
standards, so systems were constructed using proprietary
protocols. Only in the last few years, with the development of
the Data over Cable System Interface Specifications (DOC-
SIS), has a standard existed. Before that the industry relied
on proprietary equipment such as the LANcity (now a Nortel
company) transceiver that provided access to the physical
cable medium. Even now, DOCSIS is not fully accepted as an
international standard. The Institute of Electrical and
Electronic Engineers (IEEE) organized the 802.14 commit-
tee to develop a standard, but DOCSIS was completed first,
leading many to question whether 802.14 has missed the
boat. Some cable operators still maintain proprietary equip-
ment, which requires matching modems to communicate.
DOCSIS, which is discussed in more detail later, is today the
best protocol the cable industry has. Cable Labs (www.cable-
labs.com), the industry association that developed DOCSIS,
tests and certifies modems as compatible with the protocol,
thereby enabling users to obtain their modems from sources
other than the cable provider.
A major issue in many jurisdictions is that of open access
to the medium. The municipalities that regulate cable gener-
ally prefer that their constituents have free access to any ISP,
Cable Access Technology 63

but at this point, cable providers have been successful in


warding off requirements to open their networks.
Consequently, most cable users have access only to the cable
provider’s captive ISP. The city of Portland, Oregon, attempt-
ed to require AT&T Broadband to provide open access to all
ISPs, but AT&T defeated the order in court. The open access
issue is still boiling on the back burner, and undoubtedly will
resurface as the industry matures. One consequence of the
lack of open access was the widespread disruption that
occurred in December, 2001. AT&T’s provider
Excite@home disconnected the service following its bank-
ruptcy necessitating changes to every AT&T subscriber.

Conventional Cable Architecture


A cable system is constructed on a completely different
architecture than the telephone system. Where the telephone
companies extend one or more cable pairs from the central
office directly to each subscriber, cable companies connect
their subscribers in parallel across a cable that delivers all
channels to all subscribers. For those who choose not to sub-
scribe to premium channels, the cable company either filters
such channels out or scrambles them so that a device provid-
ed by the cable company is needed to descramble them.
The architecture of a cable system is shown in Fig. 1.2.
The headend consists of off-the-air receiving apparatuses for
picking up local channels, satellite receivers for receiving
premium channels, modulators to stack the video signals in
the appropriate frequency slots, and equipment to receive
upstream transmissions and connect them to the appropriate
service. The headend feeds trunk cables with various enter-
tainment and information channels. The trunk cable is a high
grade of coaxial cable with diameters of 1.9 to 2.5 cm (0.75
to 1 in) and amplifiers placed at intervals of about 500 m.
Automatic gain control in the amplifiers compensates for
changes in transmission loss with temperature variations.
The system feeds power to amplifiers over the coaxial center
conductor, with main power feeds located approximately
64 Access Technologies: DSL and Cable

every 1.5 km. To continue essential services during power


outages and amplifier failures, the cable operator provides
redundant amplifiers and backup battery supplies.
Bridger amplifiers separate the feeder or distribution cable
from the trunk cable. Multiple feeders are coupled with split-
ters, which match the impedance of the cables. Feeder cable is
smaller and less expensive and has higher loss than trunk
cable. Subscriber drops connect to the feeder cable through
taps, which are passive devices that isolate the feeder cable
from the drop. The tap must have enough isolation so that dis-
turbances at the subscriber’s premise do not affect other users.
The FCC assigns 6 MHz of bandwidth to each TV channel
(8 MHz is assigned in Europe). Only 4 MHz is used for trans-
mission; the remaining bandwidth is used for guard bands to
prevent interference between adjacent channels. Cable is
inherently an analog frequency-division medium. The earliest
cable systems were intended strictly for off-the-air pickup and
retransmission. Channels 2 to 6 of the broadcast spectrum in
North America cover the range of 54 to 88 MHz. Channel 7
starts at 174 MHz, with channel 13 assigned 210 to 216 MHz.
UHF channels 14 to 69 cover the frequency range of 470 to
806 MHz. Set-top converters and cable-ready television sets
can receive as many as 100 channels, which is approximately
the capacity of a high-bandwidth cable system.
Cable has a major advantage over regular broadcast chan-
nels. Households on the fringe of a broadcast coverage area
can easily experience interference on adjacent channels
despite the guard bands. The FCC, accordingly, normally does
not assign adjacent channels in the same metropolitan area.
Cable is under no such restrictions because all channels are
transmitted at the same power level so adjacent channel inter-
ference is not a problem. The headend equipment can pick sig-
nals off the air and assign them to any channel. The audio and
video signals are stripped from the channel carrier, reapplied
to channel modulators, and inserted into the trunk cable.
The earliest cable systems were constructed to transmit
only the 12 VHF channels. Any UHF channels in the area
were applied to a vacant VHF channel on the cable.
Gradually, however, cable systems were expanded to the
Cable Access Technology 65

point that the typical system today has a bandwidth of 1 GHz,


of which the top 250 MHz is reserved for future applications,
leaving room for about 100 channels.
An analog cable system could amplify signals to extend
the range almost indefinitely, except for the problems of
noise and distortion, which limit the serving radius of 80-
channel CATV to about 8 km from the headend. As with any
analog medium, when the signal is amplified any accumulat-
ed noise is amplified along with the desired signal. The
measure of quality in a cable system is the signal-to-noise
ratio (SNR). The SNR should be at least 35 dB to provide a
satisfactory signal. The video portion of the signal is ampli-
tude-modulated. The lowest level in the picture portion of the
signal is black, so any noise spiking above the black level
shows as snow on the screen. Cable is engineered to keep
noise within prescribed boundaries, but the more media that
are carried on a cable, the more susceptible it is to noise.
Each subscriber is a potential noise source. Since hundreds
of subscriber drops are bridged in parallel across the cable,
these can induce unwanted signals. Amateur radio and any
apparatus with a high electrical noise level can introduce
noise into the cable. Cable systems must therefore be care-
fully constructed to prevent interference with other services.
Signals leaking into the cable from the outside would distort
the picture and signals radiating from the cable could inter-
fere with other services.

Two-Way Systems
Two-way cable systems are constructed on the same princi-
ples as one-way systems, but with bidirectional amplifiers.
Filters split the signal into the high band for downstream
transmission and the low band for upstream transmission.
The ratio of downstream to upstream bandwidth is heavily
weighted in the downstream direction. The 5- to 40-MHz
range is used for upstream bandwidth, with a guard band
about 15 MHz wide to separate the two directions of trans-
mission. The upstream direction shares frequencies with
66 Access Technologies: DSL and Cable

high-powered short-wave transmitters operating at frequen-


cies of 5 to 30 MHz, so interference from these sources is a
potential problem. Cable and amplifiers must be adequately
shielded to prevent interference.
Headend equipment is more complex in a two-way than in
a one-way CATV system. The cable system is constructed as
a shared medium, which means that some method is required
to regulate access. The downstream direction is simple to
regulate. Communicating devices are assigned addresses and
programmed to respond only to their address, much in the
same way as Ethernet operates. Upstream is another matter.
User devices access the upstream cable by contention, token
passing, or being polled from the headend. Some systems use
a transponder at the user end to receive and execute orders
from the headend. For example, a polling message might
instruct the transponder to read utility meters and forward the
Y
reading over the upstream channel. One solution to the two-
FL
way cable problem is using a telephony return. One of the 6-
MHz TV channels is devoted to downstream, while the
upstream direction uses a conventional modem on the tele-
AM

phone line. Although this method works for some services, it


has obvious disadvantages for any service requiring always-
on access.
TE

Hybrid Fiber-Coax Architecture


The traditional cable architecture shown in Fig. 1.2 is being
displaced as cables are converted to two-way transmission.
One problem with traditional cables is that amplifiers are
connected in series, leaving the service vulnerable to ampli-
fier failures. Failure of a trunk amplifier can disrupt service
to an entire neighborhood. Redundant amplifiers are avail-
able, but a more effective method is to bypass the trunk cable
with fiber optics, which doesn’t require amplifiers. The typ-
ical cable network today uses fiber to a neighborhood node,
where the signal is converted from optical to electrical. This
architecture is known as hybrid fiber-coax (HFC). Fiber has
more than enough bandwidth, and has the additional advan-
Cable Access Technology 67

tage of eliminating a string of amplifiers, which increases


reliability. Furthermore, dense wave-division multiplexing
(DWDM) can expand the capacity of the fiber. In concept,
DWDM is much like frequency-division multiplexing, which
is used on coax. Signals are applied to different “colors,”
with multiple wavelengths applied to the fiber simultaneous-
ly. The signals aren’t truly colors, because the wavelengths
used in fiber are below the visible spectrum, but the multi-
ple-color concept is a useful way of understanding DWDM.
Data signals may be transported upstream and down-
stream as digital signals in the fiber portion of the network
and combined with the video signal in the coaxial portion.
The result is a medium that is inherently shared for data with
all the subscribers in the neighborhood coaxial section,
which typically numbers from 500 to 2000 subscribers.

Cable Access Technology


In many ways, cable is an ideal Internet access technology in
the downstream direction. With around 750 MHz of band-
width, the cable operator can easily sacrifice an entertain-
ment channel to enable subscribers to receive data transmis-
sions addressed to them. The addressing scheme is standard
IP, and, using QAM modulation, one 6-MHz television chan-
nel can carry 27 Mbps of downstream data. The asymmetric
nature of Internet access fits well into the cable transmission
scheme with its limited upstream and abundant downstream
bandwidth. On the surface, it would appear that cable is a
superior method to DSL, which rarely exceeds 1.5 Mbps
downstream and is usually much less. Cable would be prefer-
able to DSL except for three limiting factors:

1. Cable is inherently a shared medium, which means that


the bandwidth available to a subscriber is a direct func-
tion of the number and activity of other subscribers on the
same coaxial segment. By contrast, DSL is a dedicated
connection to the DSLAM, and is shared only in the serv-
ice provider’s access circuit into the backbone network.
68 Access Technologies: DSL and Cable

2. Access to upstream bandwidth must be regulated by some


sharing protocol. The upstream bandwidth must divided
among users or users must contend with one another for
access. DSL provides a direct channel that is exclusive to
the subscriber.
3. Cable is primarily intended as a residential service. It may
not be available to some businesses, and even if it is, its
asymmetric nature may preclude its use for many business
applications.

A data network using a shared medium must have a


method of regulating access. Token passing is one method.
The token is a software mark that circulates past all stations
in turn. A station can transmit only when it possesses the
token. Then it can transmit a frame, after which it relinquish-
es the token to the next station in line. Token passing is used
successfully in the IEEE 802.4 token bus protocol, but it is
too complex for a cable network serving hundreds of inde-
pendent stations that may come online and drop off at ran-
dom. If a station drops off just at the time it possesses the
token, the other stations must regenerate the token, and sta-
tions must at all times know the identity of immediately adja-
cent stations to regulate token flow.
A more practical method is a protocol known as Aloha,
developed by the University of Hawaii for data communica-
tions between the islands over radio. When a station has data
to send, it transmits it. If two stations transmit simultaneous-
ly, their transmissions collide, and the data is mutilated. The
transmitting station, receiving no acknowledgement of
receipt, sends again until the message is acknowledged. The
problem with Aloha is similar to the problem in a cable net-
work. Since the stations can’t hear transmissions of all the
other stations, they have no way of waiting until the channel
is clear before transmitting. The busier the network, there-
fore, the greater the probability of collision.
Ethernet evolved from the Aloha protocol and is the pre-
dominant protocol for resolving network access, but it works
well because all stations on the network can hear transmis-
sion from the other stations. When a station is transmitting
Cable Access Technology 69

just at the time other stations have data to send, the stations
wait until the network is idle.
The Aloha situation is similar to that in a cable network.
Stations can hear the transmissions of upstream stations, but
they cannot hear downstream stations. If stations simply lis-
ten to the channel and wait for an idle moment to transmit,
the upstream stations grab a disproportionate share of the
bandwidth and squeeze out the downstream stations. This
can be resolved by using a protocol known as slotted Aloha.
Stations are granted a time slot within which they can trans-
mit. The headend is responsible for ensuring that bandwidth
is evenly distributed by assigning transmission timeslots to
active stations.
Proprietary methods were used for years on cable net-
works, but the result was that users could obtain only the
types of modems that the cable operator certified for its net-
work. Proprietary equipment tends to be more expensive than
standard equipment, and also limits the growth of the tech-
nology, which led the cable industry to develop a protocol
that all could employ.

DOCSIS
DOCSIS was developed by CableLabs, Inc., a consortium of
equipment manufacturers that collaborated on creating a
standard for data transmission over cable. In Europe, DOC-
SIS is known as a Euro-DOCSIS, and is derived from the
U.S. version. The European cable community adopted the
standard in early 2000. DOCSIS consists of several compo-
nents. The cable modem (CM) connects the subscriber’s PC
to the cable network. At the headend is the cable modem ter-
mination system (CMTS) and a variety of specialized
servers. This equipment operates with the subscriber’s
modem as either a bridge or router. In between is the HFC
plant, which forms the radio frequency (RF) link between the
CMTS and the CM. DOCSIS specifies the RF physical layer
with respect to modulation methods and symbol rates. It
specifies modem initialization procedures, security, and data
70 Access Technologies: DSL and Cable

control. At the subscriber’s premises a splitter separates the


data and television portions of the signal.
Cable companies can use DOCSIS standards to facilitate
delivery of Internet and other information services over their
CATV networks. The standard relieves users of the need to
purchase a proprietary modem to operate over cable.
Although DOCSIS is designed for cable, with certain modi-
fications it can also be used in wireless multipoint multi-
channel distribution service (MMDS) and local multipoint
distribution service (LMDS).
The headend contains several specialized servers in addi-
tion to the CMTS server. These include a DHCP server,
which downloads IP addresses when CMs become active. A
user authorization server controls access to the system and
may provide accounting. A time-of-day server enables the
CM to synchronize its time with the rest of the system. The
DNS correlates URLs to IP addresses. A Trivial File Transfer
Protocol (TFTP) server is provided to facilitate file transfer.
In addition, a service provider will undoubtedly provide an e-
mail server. Many cable operators cache popular Web pages
in servers close to the user to reduce the amount of band-
width required in the Internet backbone.
Security is a potential problem with a shared medium
because any station with a protocol analyzer on a shared
coaxial leg can see the transmissions of the upstream sta-
tions. If that station happens to be closest to the fiber node,
it could see all upstream packets and intercept confidential
information such as passwords. To prevent this, the DOCSIS
protocol includes encryption and key exchange.
A cable modem is effectively a frequency-agile RF trans-
ceiver that is tuned to upstream and downstream channels.
Downstream, the cable operator typically allocates a single
6-MHz channel for data. The cable modems are all tuned to
the same channel and use the IP address to pick their packets
from the stream. The CM usually connects to a PC with
Ethernet. Some universal serial bus (USB) modems are avail-
able, but these have the disadvantage of allowing only one
PC to connect to the modem, whereas Ethernet can connect
multiple PCs through a hub or switch.
Cable Access Technology 71

When a CM is first turned on, it begins scanning the cable


for a downstream data channel. From this it learns the fre-
quency of the upstream channel. It broadcasts its presence to
the CMTS, which validates its right to use the service,
obtains an IP address from the DHCP server, and sends the
IP address to the CM. With this initial handshake completed,
the CM is prepared to communicate over the cable. Cable
modems may also use a telephone line for the return path, a
method known as a telephony return interface (TRI) system,
but it is not suitable for always-on access. First, it ties up a
telephone line, adding to the expense, and second, the tele-
phone line is limited to a maximum of 33.3 kbps upstream,
which is much slower than the CM.
DOCSIS uses a frame format similar to Ethernet in both
the upstream and downstream directions. The LLC is stan-
dard IEEE 802.2 protocol. The rest of the data link layer has
two other sublayers in addition to the LLC: link security and
MAC. The link security sublayer has three sets of require-
ments: baseline privacy interface (BPI), security system
interface (SSI), and removable security module interface
(RSMI). The BPI encrypts data traffic between the user’s
modem and the CMTS. The MAC layer includes collision
detection and retransmission, error detection and recovery,
and procedures for registering modems. It also performs
ranging, which enables the CMTS to evaluate the time delay
to each cable modem and allocate the upstream time slots
accordingly. Table 5.1 may help to put the architecture in per-
spective with respect to the OSI model.
Since cable is an analog medium, cable modems commu-
nicate with the headend over modulated analog channels. The
downstream channel, which is standardized as ITU J.83, is
either 64-QAM or 256-QAM. The downstream payload of
64-QAM is approximately 27 Mbps. Eight amplitude levels
are used to modulate the carriers and 6 bits of data are trans-
mitted at a time. The payload of 256-QAM is 39 Mbps using
16 amplitude levels and transmitting 8 bits of data at a time.
The bandwidth of the RF signal is 180 kHz to 6.4 MHz, with
the data rate varying from a low of 320 kbps to a high of 20.5
Mbps depending on the modulation method and the RF
72 Access Technologies: DSL and Cable

Table 5.1 DOCSIS Compared to OSI


OSI DOCSIS
Higher layers Applications DOCSIS control messages
Transport TCP/UDP
Network IP
Data link IEEE 802.2
Link security (BPI, SSI, RSMI)
Media access control
Physical Upstream Downstream
TDMA MPEG-2
5–40 MHz 54–850 MHz
QPSK/16-QAM 64/256-QAM

bandwidth used. All users on the coaxial portion share the


aggregate bandwidth, so the actual throughput any user will
experience is much lower and varies with the amount of
activity. The upstream direction uses either quadrature phase-
shift keying (QPSK) or 16-QAM. Although 16-QAM has
twice the data rate of QPSK, the latter is more tolerant of
interference. The upstream direction is much lower in speed
than the downstream channel, having a typical bandwidth of
300 kbps to 1 Mbps.
Downstream data is encapsulated in MPEG-2 frames.
MPEG stands for the Motion Picture Experts Group, which
developed a standard method of transmitting compressed
digital video. MPEG-1 was developed to compress video
onto audio CDs using a low-bit-rate process that resulted in
a picture with about the same resolution as VHS tape.
MPEG-2 provides studio-quality video, including support for
high-definition TV (HDTV). It allows multiple channels to
be multiplexed into a single data stream.
DOCSIS specifies Reed-Solomon FEC as a means of
improving error performance. FEC adds redundant bits to the
bit stream and sends them along with the information bits. If
errors occur, the decoder attempts to correct them before the
bit stream is presented to the application, with the final error
check done at the receiving apparatus. The Reed-Solomon
code used in video transmits 204 bytes per frame, of which
Cable Access Technology 73

188 bytes are MPEG header and information and the remain-
der are parity bits to correct errors. The purpose of forward
error correction is to reduce the number of frames that must
be retransmitted because errors occurred. Each frame starts
with a packet identifier (PID). Standard PIDs are assigned
for video, audio, clock, and other data such as the program
guide. MPEG-2 can multiplex multiple audio and video pro-
grams together on the same bit stream.
One feature of DOCSIS is the provision of various class-
es of service. Cable providers can offer priority to customers
who are willing to pay, just as the airlines offer first-class,
business class, and tourist seats. Although some compromis-
es are required to use cable as an access medium, it is a reli-
able system that provides sufficient bandwidth to meet near-
ly all residential requirements. From the cable operator’s
standpoint, it is a profitable service. The main problem is that
as demand grows, response time slows. The only solution is
to segment the coaxial portion of the network, which is an
expensive proposition and not always practical.
6
Wireless Access Technology

M any companies are relying on wireless to loosen


the ILECs’ stranglehold on the local loop.
Humpty Dumpty could have been talking about the WLL
when he said that words mean exactly what he wanted them
to mean. By WLL some people mean cellular, others mean
one of the wireless local distribution systems, and some
writers even categorize cordless telephones as WLL. We
will reject that definition because cordless telephones can’t
connect to the wide area voice or data network directly
without going through a POTS line. The other technologies
qualify, though, and in this chapter we will briefly discuss
the following:

• Cellular modems: Specialized modems connected to an


ordinary cell phone. The speed is limited and the usage
is metered.
75

Copyright © 2002 by The McGraw-Hill Companies, Inc. Click here for Terms of Use.
76 Access Technologies: DSL and Cable

• Next-generation cellular: Also known as 3G and 2.5G cel-


lular. New hand-held devices overcome the speed limita-
tions of conventional cellular data.
• Cellular digital packet data (CDPD): Makes use of idle
cellular channel capacity. TCP/IP connections at up to
19.2 kbps can be supported.
• IEEE 802.11b wireless: Wireless LAN protocol used for
in-building applications and for fixed point-to-point con-
nections at up to 54 Mbps.
• Multipoint multifrequency distribution system (MMDS):
Delivers video bandwidth from a central transmitter to
multiple receivers in the 2.5-GHz band.
• Local multipoint distribution system (LMDS): Uses the
25-GHz and higher spectrum to deliver point-to-point or
point-to-multipoint broadband services.
• Satellite services: Both equatorial and low-orbiting satellite
Y
services provide bandwidths ranging from voice circuit
FL
to video.

Wireless has much in common with cable in that the medi-


AM

um is shared. As with cable, sharing is no problem in the down-


stream direction. All that is needed is to program the receiving
device to respond to its address and ignore the others. The
TE

upstream direction is a problem because multiple contenders


are vying for access to the same bandwidth. Upstream sharing
takes place by one of three multiple access methods: frequency
division, time division, or code division.
From a high-level standpoint, FDM is ultimately used in
every wireless technology. The available spectrum is divided
into frequency segments and allocated to the various services.
The radio division of ITU regulates bandwidth assignments
through international treaties, but the countries have consid-
erable latitude, particularly in microwave bands where radio
waves can be confined within a country’s borders. An excep-
tion to this is satellite broadcasting, which cannot be confined
to national boundaries. Microwave frequencies used for satel-
lite can be focused narrowly and reused many times, with ter-
restrial and satellite services often sharing the same frequency
spectrum. Within the assigned frequency segments, FDM is
Wireless Access Technology 77

often used to segregate channels. The earliest mobile and cel-


lular services used FDM with analog modulation.
The second method of sharing is time-division multiple
access (TDMA), which is used in digital cellular and satel-
lite uplinks. On wired media, Ethernet in LANs and DOC-
SIS in cable systems use this method. Each device receives
a share of the bandwidth. The key is to allocate bandwidth
so every station gets an equitable share. When stations can
hear one another as they do with Ethernet, they simply wait
for an idle moment. When they cannot hear one another, as
is the case with satellites, cable, and cellular, access is regu-
lated by some other means such as assigning time slots from
a master station.
Code-division multiple access (CDMA), also known as
direct sequence spread spectrum (DSSS), is somewhat more
difficult to grasp. Instead of waiting for an available time
slot, stations simply launch their signals across a wide band
of frequencies. A code embedded in the transmission enables
the receiver to pick the desired signal from the jumble. A use-
ful analogy is to visualize a crowded room filled with people
speaking many different languages. Each listener is able to
focus on his or her native language and pick out that one
conversation from a cacophony of voices.
Wireless has a lot to recommend it. In fact, for some appli-
cations it is the only alternative. Today, cruise ships and air-
lines offer their passengers telephone and Internet access.
Emergency vehicles and mobile services, even down to the
much-maligned cell phone in the restaurant, all rely on radio
waves for access to international voice and data networks.
The industry provides a variety of wireless alternatives, some
of which provide access options that can’t be matched by
wired services.

Categories of Wireless Service


Wireless service can be characterized as broadband or nar-
rowband. Broadband alternatives generally provide down-
stream bandwidth on 6-MHz television channels and provide
78 Access Technologies: DSL and Cable

a wireless service that is similar to wired service on cable. Some


even use a version of DOCSIS protocol. Narrowband services
provide bandwidths ranging from voice-grade services to the
equivalent of T1/E1.
Wireless can further be characterized as licensed or unli-
censed. Unlicensed systems can operate in specified fre-
quency bands, but the users must resolve any interference
problems because the FCC or other regulatory authorities
make no guarantee of an interference-free channel. Licensed
systems are more likely to be free of interference, but the per-
mit process requires interference studies and a sometimes
lengthy licensing process.
A third factor classifying radio is whether the application
is fixed or mobile. Fixed radio access substitutes a radio link
for wire; otherwise the wired and wireless services produce
an equivalent result. The data output from the wireless link
feeds a LAN for distribution inside the building.
The frequency band further classifies wireless applica-
tions. The higher the frequency, the more spectrum that is
available and the more bandwidth that an application can
occupy. Although more bandwidth is available at higher fre-
quencies, coverage becomes more of an issue. The highest
microwave bands in use today are on the order of 40 GHz and
may have a range as short as 2 mi or so. The lower bands are
heavily used and therefore more congested. The higher fre-
quencies are more readily adaptable to frequency reuse than
the lower bands. Cellular, which operates in the 800-MHz
band, is based on the principle of reuse. Distance and terrain
are used to separate cell sites using the same frequencies.
With the use of directional antennas, different sectors can be
served by transmitters on the same frequency without mutual
interference. At the shorter wavelengths, signals can operate
on the same frequency by cross-polarization, i.e., one signal
is polled vertically and the other horizontally.
All services and the bands we have been discussing share
a common characteristic: they operate only within the line
of sight. They may penetrate walls, but communication is
not reliable. The higher the frequency, the shorter the wave-
length and the more metal framework and masonry walls
Wireless Access Technology 79

attenuate waves. Microwave frequencies above about 10


GHz are subject to rain fading. In areas of heavy rainfall,
such as the Gulf Coast of the United States, raindrop size is
a significant fraction of a wavelength, so that rainfall may
depolarize or absorb microwave signals. Attenuation from
rain and fog can be overcome with higher power, which can
be achieved by increasing use of transmit output power,
receiver sensitivity, or antenna gain. These effects are pre-
dictable and a properly designed system should operate reli-
ably over narrow ranges.
Radio signals are also attenuated by multipath fading.
This phenomenon occurs when a signal is reflected. The
reflected path takes a longer route to the receiving antenna
and arrives at the antenna slightly out of phase with the
direct path, resulting in a reduction of the received signal
level. Under normal conditions and with a properly designed
system, multipath fading does not occur, but a sudden fog
bank can introduce unpredictable reflections. Because of the
transient nature of multipath reflections, not much can be
done about them except to tolerate the outage. Commercial
microwave systems have a variety of diversity methods that
improve the path reliability, but these are generally not fea-
sible for access systems.
Microwave radio is an excellent broadband downstream
information system. Just as a cable system can devote a
6-MHz channel to downstream information, a satellite or
terrestrial microwave can do the same because of the one-
to-many nature of the channel. Each receiver is tuned to
recognize its address and pick its information out of the
data stream. Upstream, however, is another matter because
of its many-to-one characteristics. No license is needed to
receive data, but in many cases the upstream direction
must be licensed, which adds delays and some complexity.
Using a telephone return may solve the upstream problem,
but this has the same disadvantages we discussed in the
Chapter 5.
With this background in mind, we will discuss some of the
methods of providing local access service over wireless. As
with the other services, this discussion will emphasize data.
80 Access Technologies: DSL and Cable

Cellular Dial-up
Within limits, cellular dial-up can be an effective access
option. The first limitation is low data throughput, which
makes it unsatisfactory for Web surfing. Ordinary commer-
cial modems may not work well with cellular, particularly at
high speeds. The modulation methods of these modems are
complex, and they do not gracefully handle interference,
fades, and signal dropout. Handoff between cells causes a
momentary interruption, and may cause the connection to
drop. Furthermore, many modems are designed to operate
only after they recognize dial tone, which cellular does not
provide. Therefore, special cellular modems that can adjust
speed to the signal conditions are required.
For occasional use such as mobile file transfer, cellular is a
satisfactory alternative, but it is expensive for short messages.
Most cellular operators levy a one-minute per-call minimum
charge, which makes cellular a poor choice for short transac-
tions that may last only a few seconds. Also, the setup time is
long compared to other alternatives. Cellular has the advan-
tage of good coverage. In general, its applications are similar
to those of the PSTN: it is acceptable for facsimile and file
transfers, but poor for short, bursty messages.

Next-Generation Cellular
The cellular industry is touting so-called third-generation
cellular as the solution for Web surfing from a hand-held
mobile device. The first cellular generation was analog fre-
quency-division cellular, some of which is still in operation.
As the industry ran out of capacity in the cellular bands,
which are between 824 and 894 MHz, second-generation
digital cellular was introduced to make more efficient use of
the spectrum. Most of the world elected to use the global sys-
tem for mobile communications (GSM) modulation method.
In the United States, the FCC elected to let the market deter-
mine the most effective method. As a result, two methods are
used, neither of which is compatible with GSM. AT&T uses
Wireless Access Technology 81

TDMA to apply three digital channels in the same spectrum


used by one analog channel. Several other carriers have
adopted CDMA. Since these methods use voice compression
to make better use of the spectrum, they are not suitable for
high-speed data communications.
The third generation, which goes by the acronym of 3G
cellular, promises data transmission at as much as 2 Mbps,
which makes it competitive with the best wired alternatives.
The only problem is that 3G cellular will have a difficult time
fulfilling its promise. Several technical hurdles remain to be
overcome before 3G is practical. One problem is power
drain. The circuitry needed to sustain higher data rates gives
off heat, which means high battery drain—more than today’s
batteries can provide. One solution is to reduce the data rate,
perhaps by a factor of four or more. This will limit some of
the expected applications while still providing a respectable
speed for Web surfing. Furthermore, the small screens and
keyboards of hand-held devices put additional limits on the
applications. This is causing manufacturers to scale back
their plans to something called 2.5G cellular, which has data
rates in the order of 384 kbps. Although this is far from the
promise of 3G, it is still considerably more than dial-up land-
line services.
A major issue with both 2.5G and 3G wireless is spectrum.
The existing PCS spectrum is insufficient to meet the bandwidth
requirements, which has led the industry to ask the FCC to
allocate bands in the 1750- to 1850-MHz and the 2500- to 2690-
MHz ranges. This spectrum is currently occupied by fixed wire-
less and the Department of Defense. Whatever the outcome of
this request, it is clear that next-generation wireless has a long
way to go before it is a mainstream access choice.
Another technology getting a lot of attention is Bluetooth,
which is intended to enable devices to communicate with one
another over short ranges using unlicensed 2.4-GHz frequen-
cies. Bluetooth is intended to link devices such as personal
digital assistants (PDAs) with master devices over a wireless
link. Bluetooth is not intended as a public communications
medium, so it does not fit into our definition of an access
technology. It is worthy of attention by potential customers if
82 Access Technologies: DSL and Cable

for no other reason than the fact that it occupies the same fre-
quencies as 802.11b wireless and that coexistence may be
a problem.

CDPD
CDPD is a good alternative for obtaining wireless coverage in
metropolitan areas and along major highways because it offers
the same coverage area as cellular if the service provider has
elected to equip its cell sites for the service. CDPD is a packet-
switched data service that rides on top of cellular and uses idle
analog channel time. It can be added to existing cell sites at a
moderate cost. Carriers charge by the packet or kilobyte
instead of by connect time, and the long call setup time and
minimum connect time charges are eliminated. This makes
CDPD good for short, bursty messages such as point of sale,
dispatch, package tracking, telemetry, and e-mail. CDPD is
available in most metropolitan areas.
CDPD operates at 19.2 kbps using a TCP/IP type of pro-
tocol, which raises the problem of IP addressing because the
subnet is mobile. For CDPD to be entirely effective for some
applications, a laptop user should be able to disconnect from
the LAN, travel to another location while remaining in con-
tact with the network through a wireless connection, recon-
nect to the LAN at the distant location, and become part of
the network again. The process is possible today, but the user
needs to understand how to do it. Therefore, it is not yet fea-
sible for the true mobile laptop or PDA application.
To implement CDPD, carriers install mobile database
stations (MDBSs), which retrieve packets from the wireless
network, and a mobile data intermediate system (MDIS),
which routes them. Frames are picked up by the MDBS and
handed off to the MDIS. Mobile stations use a protocol called
Digital Sense Multiple Access with Collision Detection
(DSMA/CD) for access to the network. The access method is
similar to Ethernet’s Carrier Sense Multiple Access with
Collision Detection (CSMA/CD). A station wishing to trans-
mit listens to the outbound channel to determine if a carrier is
Wireless Access Technology 83

present. If not, it transmits a packet. If so, it waits and attempts


a short time later.
While TCP/IP is an excellent protocol for wired services,
it is not optimized for mobile use, particularly where a per-
packet charge applies. Frequent acknowledgment packets
are returned by TCP/IP, and, although these are short, unless
the charge is byte-oriented instead of packet-oriented, as
much as one-third of the cost of a session may be taken up
with acknowledgment packets. The hazards of mobile and
portable communications make dropped and out-of-sequence
packets likely.
The main advantage of CDPD is the coverage it can offer.
The main population areas of the country are well covered
for cellular, and if the carrier elects to overbuild the network
with CDPD, data coverage can be equivalent. CDPD is good
for short, bursty applications, but for lengthy file transfers,
dial-up application over regular cellular may be less costly.
The major drawback of CDPD is its narrow bandwidth,
which makes it a poor choice for Web surfing.

The 802.11b Wireless Protocol


For portable and mobile operations, wireless is the only way to
get freedom of movement. The 802.11b protocol, also known
as “Wi-Fi” (wireless fidelity) is effectively Ethernet over a
radio link. The protocol is a bit different than Ethernet and the
throughput is not as high, but the principle is the same.
Transceivers can be placed in locations such as offices, hotels,
airports, and warehouses, and users can link up with hand-held
and laptop devices. Tiny wireless modems fit in a PC card
slot and enable users to roam the coverage area. The service
can be private or offered by a public service provider, which
enables travelers to establish accounts to log on to the Internet.
The 802.11b specification is similar to 802.3 with a wire-
less physical layer that uses DSSS. As with 802.3, the data
link layer is subdivided into the MAC and LLC sublayers. The
access control protocol is Carrier Sense Multiple Access with
Collision Avoidance (CSMA/CA). The standard CSMA/CD
84 Access Technologies: DSL and Cable

protocol that Ethernet uses is not feasible with wireless


because the portable receiver cannot hear the transmissions
of all the other units. Collision avoidance requires the sta-
tions to broadcast a request-to-send (RTS) packet containing
the sending and receiving addresses and the length of the
information packet. If the receiving station is in a position to
accept a packet, it broadcasts a clear-to-send (CTS) packet.
Stations on the network that receive the RTS or CTS pack-
ets set internal timers that prevent them from sending during
the expected duration, thereby reducing the probability of
collisions.
Wireless LANs operate on unlicensed industrial, scien-
tific, and medical (ISM) band frequencies in the ranges of
2.4 to 2.4835 GHz and 5.1 to 5.825 GHz. The FCC requires
such devices to use spread spectrum modulation with a
maximum of 1 watt of power. Most of the wireless LAN
products on the market use the 2.4-GHz band, but the 5.1-
GHz frequencies offer the potential of faster speeds. Since
the frequencies are unlicensed, interference is always a pos-
sibility, but spread spectrum can distinguish the wanted
signal. In the DSSS method used in 802.11b, the band is
divided into 14 overlapping 22-MHz channels.
The protocol operates over a 2.4-GHz carrier wave that
is modulated with different techniques depending on speed.
The protocol is designed to sustain the highest throughput
possible, which is 11 Mbps under ideal conditions but may
step down to as low as 1 Mbps. Since the output power is
limited to 1 watt, as the mobile device moves away from the
wireless access point, the data rate drops. Binary phase-shift
keying is used at 1 Mbps, encoding 1 bit for each phase
shift. At 2 Mbps, QPSK is used to encode 2 bits of informa-
tion per symbol. The top speed of 11 Mbps is achieved with a
unique QPSK coding scheme that encodes 8 bits per symbol.
The 802.11b protocol promises to be one of the chief meth-
ods of obtaining wireless communication. Service providers
are beginning to offer Internet access service in public loca-
tions such as airports and hotels. ISPs use 802.11b wireless
bridges to connect to their subscribers. Inside office build-
ings, 802.11b substitutes for wired LANs, which is a princi-
pal objective of the protocol.
Wireless Access Technology 85

The main limitation of 802.11b wireless is transmission


speed, which is much slower than 100-Mbps Ethernet. Typical
speeds do not exceed 1 or 2 Mbps, which is still plenty for
most access applications. The IEEE is working on a new pro-
tocol, 802.11g, which is intended to double the speed to 22
Mbps. In addition, IEEE 802.11a is being developed to use the
5-GHz band. This protocol operates at 54 Mbps. Both Wi-Fi
and Bluetooth operate in the same frequency range, which
raises compatibility issues that have yet to be resolved.

LMDS
In 1998 the FCC auctioned 1.3 GHz of spectrum in the 28-
to 31-GHz range for LMDS. In each geographical area, WLL
providers bid for the right to use frequencies known as the A
block, with 1150 MHz of bandwidth, and the B block, with
150 MHz of bandwidth. LMDS is intended for such services
as multichannel video, interactive gaming, streaming video,
telephone service, and Internet access. LMDS is a point-to-
multipoint service. The service provider locates a hub in the
center of a serving area that may serve several thousand
homes. Subscribers are equipped with small rooftop anten-
nas and transceivers and feed data into the pipeline using an
ATM-like protocol or IP. Multiple hubs are linked with fiber
optics. The bandwidths available to customers range from
one DS-1 to as much as OC-3. Figure 6.1 shows the archi-
tecture of a typical LMDS network. In Canada a similar serv-
ice is called local multipoint communication systems, and
similar services are available in many European countries.
LMDS is intended to allow service providers to bypass the
copper local loop with an economical service that can be
deployed rapidly. There is no need to dig up the streets to
place cable, and the scalable nature of the service enables
providers to meet customer demand in a few days after the
hub is installed. A major objective is to defer the amount of
unused investment. With copper wire plant, the ILECs must
build initially for what they forecast will be the ultimate serv-
ice demand because of the high cost of adding more. This
Y
PSTN

FL
86

AM
OC3 to 12

ATM Base Station ATM Core Switch


TE
Internet

Figure 6-1 Local Multipoint Distribution Service Architecture


Wireless Access Technology 87

results in unused investment sitting idle until demand devel-


ops. With LMDS, the major investment is the hub. The cus-
tomer premises equipment can be installed only as needed.
The base station connects to the wide area network over a
fiber-optic link. The structure of the base station is up to the
service provider. It could include a telephone switch, high-
speed routers for Internet access, backbone pipes to other serv-
ice providers such as frame relay, and so on. The microwave
signal is fed into antennas with narrowly focused beams to
transmit the signal in multiple sectors of, perhaps, 30 degrees
beam width. Upstream access is typically allocated with
TDMA using PSK or QAM modulation. The customer prem-
ises configurations will include outdoor microwave equipment
connected to an internal distribution network that could be a
LAN, T1/E1, or whatever is required by the service the cus-
tomer subscribes to.
A key issue with LMDS is availability, i.e., the percentage
of time the service is available for use. Availability figures
are quoted as a percentage of total uptime. An availability
figure of 99.9 percent would equate to 9 hours of downtime.
This is calculated from the following:
365 days × 24 h per day = 8760 h per year
× 0.999
= 8751 h of uptime
= 9 h of downtime

The service provider must take into account factors that


cause the signal to fade. With factors such as transmitter
output power, receiver sensitivity, and antenna gain held
constant, availability will be a function of the distance from
the customer’s site to the hub. The longer the signal path, the
lower the availability. The maximum distance a subscriber
can be located from a cell site while still achieving accept-
able service reliability is referred to as the link budget. A
link of 8.5 mi (14 km) may be achievable in some climates,
while in heavy rainfall areas the link budget may drop to
1.5 mi (2.5 km).
88 Access Technologies: DSL and Cable

MMDS
MMDS was originally intended as a wireless replacement
for cable. For TV signals with 6 MHz of bandwidth, 33
channels can fit into the 200 MHz of spectrum that is allo-
cated for MMDS use between 2.5 and 2.7 GHz. The archi-
tecture of MMDS is similar to that of LMDS, consisting of
headend equipment similar to that used in a CATV system
and receiving equipment at subscriber locations. For video
reception, a set-top converter demodulates the incoming sig-
nal to the frequency of a conventional television channel.
The MMDS signal is transmitted from an omnidirectional
antenna. Repeaters may be used to extend the range or to fill
dead areas caused by shadows in the coverage area.
Although MMDS was initially intended for one-way
video, it is now authorized for two-way service, making it
applicable to Internet access as well. Typical service offer-
ings provide downstream transmission rates of 1 Mbps or
higher, scalable up to 10 Mbps, and upstream speeds up to
512 kbps, which makes MMDS competitive with DSL and
cable access. The MMDS spectrum is shared with instruc-
tional television fixed service (ITFS), which is intended for
distance-learning video. The 6-MHz video channels can be
modulated with data signals using the same concepts, and in
many cases the same hardware, as cable modems.

Satellite Service
One of the major attractions of satellite service has always
been the elimination of the local loop. Satellite is ideal for
broadcast applications where it is desirable to uplink a signal
from one location and downlink it to a vast area. This is the
nature of the direct broadcast satellite services that have been
competing with cable for the past few years. These services
are acceptable for Internet use as well, except that their one-
way nature means that a telephone return is required, with its
attendant disadvantages.
Wireless Access Technology 89

Two-way satellite services are readily available, and are


feasible for multiple-site operations. This service, known as
very-small-aperture terminal (VSAT), uses a small receiver
about the size of a PC at each remote site. These terminals
are supervised by a central hub having uplinks to a geosyn-
chronous satellite, which orbits the earth at an altitude of
22,238 mi (35,580 km). At this altitude the satellite appears
stationary with respect to a point on the earth’s surface. The
customer connects to the VSAT hub over a dedicated terres-
trial circuit. The hub sends the data signal to the satellite,
which converts it to the downlink signal and broadcasts it
over the coverage area. The VSAT terminals copy messages
with their address and ignore the rest. VSAT operators gen-
erally use TDMA to regulate uplink access. VSAT is typically
used for applications that have multiple sites and reasonably
slow data transmission requirements, such as department
stores, service stations, and the like.
The principal drawback of satellite services is the time
delay. The signal travels from earth to the satellite at the
speed of light, but at the distances involved, the round-trip
delay of a quarter of a second is still slow for some services.
Many data protocols rely on receiving acknowledgements
within an expected interval. If the acknowledgement is not
received in time, the sender assumes the packet has been lost
and retransmits. One solution to this is to use a delay com-
pensator, which returns the acknowledgement to the sending
station and communicates with the receiving station over the
satellite link using a separate protocol. A better solution,
many satellite operators believe, is to use a lower orbit to
reduce the delay.
Low-earth-orbit satellite (LEOS) is as yet an unproven
service. The technology has been proven, but the economic
realities have yet to be worked out. The first LEOS system
was Motorola’s Iridium satellite network, which failed to
fulfill its promise and fell into bankruptcy. A competing
network, Teledesic, is not yet in operation. These services
are worth following because they may have application in
the future.
7
Fiber-optic Access Technology

I f you could see a few decades into the future, what might
the access network look like? No one is certain at this
point because there are so many conflicting opinions. It
seems clear, however, that the twisted-pair wire and coaxial
cable of the past must give way to a medium with greater
bandwidth. Based on what is now known, fiber optics will be
that medium. Today, copper wire, coaxial, and wireless alter-
natives all rely on fiber trunking to bring information to
neighborhood nodes. There, depending on the service
provider, the information jumps on a short twisted-pair wire
run, coax, or a wireless link of some kind. The fiber stops at
the neighborhood node because carrying it all the way to the
subscriber is too costly.
As long as the infrastructure can be installed before streets
and sidewalks are installed, the cost of fiber is supportable,
but fiber still has to fit into the existing service provider’s
91

Copyright © 2002 by The McGraw-Hill Companies, Inc. Click here for Terms of Use.
92 Access Technologies: DSL and Cable

architecture. It doesn’t pay to design a new fiber architecture


that will be installed going forward into new developments
and gradually retrofit it into existing developments without
having an accepted architecture, and that issue is not settled.
The issues are political and economic, not technical. The
technology for an all-fiber network exists today. The question
is whether the service revenues are high enough to justify the
cost. It’s clear that the revenues from the three legs of the
stool—voice, video, and Internet—are not of themselves suf-
ficient to support a fiber network to residences. The technol-
ogy to combine the three on the same fiber can easily be
developed, but political considerations intervene. None of
the three service provider classes is willing to concede the
market to the others. The logical conclusion would be to
develop a local information distribution utility that owns the
backbone and carries anyone’s information for a fee. In other
words, service provision would be separated from ownership
of the physical facility.
The CLECs have been arguing this point for the past few
years. The ILECs own the physical facility now, and also pro-
vide telephone and leased-line services. Their motivation to
assist the CLECs to share the physical facility is less enthu-
siastic than support of their own services. Furthermore, the
process is unbelievably complex. While separating the own-
ership of the physical facility from the provision of sub-
scriber services is technically possible, it would require some
major changes that the ILECs would undoubtedly oppose.
For larger businesses, fiber is readily available in most
metropolitan areas today. ILECs and competitive access
providers have constructed fiber networks to serve concen-
trations of population. Fiber is available in most major office
buildings, and more is being installed daily.

Overview of Fiber-optic Technology


To appreciate the issues involved in applying fiber in the loop,
a basic understanding of the technology is needed. This sec-
tion provides a brief overview. For more detailed information
Fiber-optic Access Technology 93

on fiber optics, the I-book Optical Networking is available at


http://shop.mcgraw-hill.com/cgi-bin/pbg/indexebooks.html.
Fiber is a hair-thin waveguide made of ultra-pure glass.
The core of the fiber is purer than the outer cladding, and
guides light pulses along the way. When pulses strike the
cladding, they are refracted so that they follow the path
defined by the core. Fiber comes in two varieties: single-
mode and multimode. Multimode fiber has a relatively wide
core of 50 or 62.5 µm (millionths of a meter). As a light pulse
passes down the fiber, it takes multiple paths, which leads to
rounding of the pulse, a condition known as dispersion.
Because of dispersion, multimode fiber has higher loss than
single-mode fiber, which has a core of 9 to 10 µm. The core
of single–mode fiber is so small that the light pulse is not
dispersed by reflections. Multimode fiber is largely confined
to building and campus applications. Metropolitan and wide
area networks use single-mode fiber almost exclusively.
On the transmitting end, electrical pulses are fed into a
tiny laser, which injects light pulses into the fiber. The light
pulses travel through the core to the distant end, where an
avalanche photodiode converts them back to an electrical sig-
nal. The distances fiber can span are a function of numerous
factors such as transmitter output power, receiver sensitivity,
fiber loss, and miscellaneous losses that result from couplers
and splices. Fiber can connect most metropolitan areas with-
out regeneration, although nonregenerative light amplifiers
may be needed to boost the signal.
Fiber operates below the visible spectrum. Three trans-
mission windows are typically available: 850, 1300, and
1550 nm. A single fiber can carry light pulses simultaneous-
ly at these three different wavelengths by using a technology
known as wave-division multiplexing (WDM). Long-haul
fibers often employ dense wave-division multiplexing
(DWDM), in which 40 or more wavelengths operate simulta-
neously. DWDM can also be used in metropolitan areas, but
generally it is less expensive to install multiple fibers than to
invest in DWDM equipment.
Fiber is deployed in one of three topologies. Point-to-point
is a common configuration for fiber originating in a location
94 Access Technologies: DSL and Cable

such as a central office and radiating out directly to the users.


Fiber installed in this configuration is vulnerable to failure,
and, because of its enormous carrying capacity, a failure may
disrupt many important services. Service providers may offer
self-healing service, switching to a protection fiber with only
a few milliseconds of delay. To gain the maximum protection
from this and other self-healing services, full route diversity
should be reviewed. If any portion of the main fiber path is
carried over the same cable route as the protection fiber, the
degree of protection is less than complete. Even if the fibers
run in separate cable sheaths, a single event such as a fire or
dig-up is likely to kill both paths.
Greater protection is secured with a self-healing ring con-
figuration. SONET multiplexers are generally capable of
both add-drop and self-healing protection. The fact that the
carrier has deployed its fiber in a metropolitan ring does not
necessarily mean that the ring is self-healing. The carrier
may provide point-to-point service without self-healing
capability. The same precautions mentioned in the preceding
paragraph about single points of failure must be reviewed. A
self-healing ring is ineffective protection if both directions of
transmission are in the same cable sheath or the same conduit
run. Also, be on alert for spurs, which are unprotected fiber
segments that branch off the protected ring.
The third configuration is a branching tree, which has a
topology similar to that of a cable network. This architecture is
used in the PON. As a general rule, fiber is the only solution
when bandwidths greater than T1/E1 are required. Multiple
T1/E1s can be used on copper, but the point is reached where it
is more economical to provide service over fiber.
The primary drawback of fiber is the expense of demulti-
plexing it. When a service provider brings fiber into a build-
ing, the minimum bandwidth is usually OC-3, which is capa-
ble of supporting 84 DS-1s. That is a lot of bandwidth for
most users, with the exception of ISPs, but it is only a frac-
tion of what a fiber pair is capable of carrying. The multi-
plexing equipment to break this down to T1/E1 is expensive
and can be economically justified only where the service
provider can support multiple customers. Therefore, fiber
Fiber-optic Access Technology 95

has been supplied primarily to carriers, large end users, and


large office complexes. As discussed later, the PON may
change that situation.
From a user’s point of view, an important issue is whether
to obtain fiber bandwidth or dark fiber (fiber without the elec-
tronics to illuminate and multiplex it). Many common carriers,
as a matter of policy, do not provide dark fiber for variety of
reasons. For one thing, the fiber is rarely located where the
user wants it without extensive cutting, splicing, and rearrang-
ing, the result of which may be small isolated fiber segments
that have no commercial use. More important to most carriers
is that fiber bandwidth can be multiplexed to carry enormous
quantities of information and multiplexing is far more prof-
itable than leasing dark fiber. Some nontraditional providers,
however, are willing to provide dark fiber, particularly if they
have more than they expect to sell service on in the foreseeable
future. Companies that have access to the public right-of-way,
such as power companies, gas and water utilities, CATV com-
panies, and even some municipalities, may have idle capacity
and be willing to lease dark fiber.
From a user’s standpoint, multiplexing dark fiber is inex-
pensive. Gigabit Ethernet, for example, provides a lot of
capacity for a wide variety of applications. Gigabit switches
are designed to interface fiber. The bandwidth is sufficient and
voice and video can be supported without undue concern for
QoS issues. If 1 Gbit is not enough, 10-Gbit Ethernet is on the
near horizon. Fiber optics is the default transmission medium
for gigabit and 10-Gbit Ethernet. Many industry observers
believe that Ethernet will eventually take the place of SONET,
largely because it is simpler and less expensive to deploy.

The PON
The PON is a logical physical structure for single-pipe infor-
mation services in the local loop. Conceptually, the service is
not much different than cable except for the physical medi-
um, which provides significantly more bandwidth. Figure 7.1
shows the PON architecture. It is typically installed with a
96 Access Technologies: DSL and Cable

tree topology, but it can also be a star or ring. The network is


passive in that power is not required except at the originating
and terminating ends. The fiber-optic trunk network starts at
a central location, which may be the same as or completely
distinct from the ILEC wire center. Fiber is not limited to the
loop length that constrains copper wire, so the center of the
PON network could be anywhere. The network is capable of
extending up to 12 mi. A single fiber is used for both
upstream and downstream transmissions. The data stream is
reflected into the various legs of the fiber through passive
splitters. Each group of 32 or more subscribers shares the
bandwidth of one fiber. A PON can deliver up to 622 Mbps
downstream to the users and up to 155 Mbps upstream. The
PON can also serve as a trunk between a larger system, such
as a CATV system, and a neighborhood, building, or home.
Since nothing installed in the middle is specific to the bit
Y
rate, the PON can carry any kind of data, analog or digital.
FL
The fiber originates in an optical line termination (OLT)
at the service provider’s headend. It terminates at the sub-
scriber’s end in an optical network unit (ONU) that performs
AM

the optical-to-electrical conversion, likely connecting to an


Ethernet LAN. The ONUs listen to the downstream address-
es, copying the packets addressed to them and discarding the
TE

rest. In the upstream direction, ONUs are assigned time slots


and are permitted to transmit in turn. The current architecture
has no way of using vacant time slots, but the bandwidth is
high enough that it doesn’t matter.
PONs can use either Ethernet or ATM as their protocol. To
distinguish between the two, they are known respectively as
EPONs or APONs. The choice between the two rests on
which protocol the service provider furnishes. As with DSL,
ATM has an advantage in that the ILEC as service provider
normally uses ATM in its backbone. Ethernet, however, is a
simpler protocol that supports full-length frames in contrast
to the 48-byte payload of an ATM cell.
PONs have not been widely developed yet. The ITU has
assigned G.983.1 to the protocol. The Full Service Access
Network Initiative (www.fsanet.net) is a group of telecom-
Fiber-optic Access Technology 97

ONU

ONU

Optical ONU
Feeder Passive
Line ONU
Fibers Splitters
Termination ONU

ONU
Central Office

ONU = Optical
Network Unit
ONU
ONU

Figure 7-1 Passive Optical Network

munications companies and equipment suppliers that are


working on PON standards that have been presented to sev-
eral standardization bodies. SBC Corporation has announced
the use of PONs in its Operation Pronto, which is intended to
bring broadband service to about 80 percent of its sub-
scribers.
8
Application Considerations

T he choice of access technology depends to a large


degree on what is available. For Internet access, the only
universally available choice is dial-up, which is a poor alterna-
tive for full-time access. Even ISDN with its 128 kbps of
bandwidth is insufficient for many users, and, if it is a measured
service, the cost may be prohibitive. Only the smallest busi-
nesses will find dial-up access satisfactory, not only because of
a lack of bandwidth, but also the need for multiple accounts, the
difficulty of distributing e-mail, and numerous lesser problems.
For a business with multiple employees who must have access
to e-mail, a full-time connection is essential. At the other end of
the scale are large businesses that directly engage in electronic
commerce. For these, T1/E1 and higher multiples are the only
alternative. In between are residences and countless smaller
businesses that cannot justify the cost of T1/E1. The choice
may boil down to the question of what is available. Cable
99

Copyright © 2002 by The McGraw-Hill Companies, Inc. Click here for Terms of Use.
100 Access Technologies: DSL and Cable

access is available only in localities in which the cable operator


has upgraded the facility to two-way and provides access serv-
ice over the cable. This excludes many, if not most, small com-
munities and rural areas, at least in the near term.
DSL also has availability problems. First is the question of
loop length and bridged tap, which probably precludes DSL for
around 40 percent of telephone subscribers. In many cases the
lack of availability isn’t technical, but is simply due to the fact
that no one offers the service. A certain critical mass of sub-
scribers is needed to make it economical to install the DSLAMs
and connecting data networks. Many rural communities have no
fiber-optic link to the outside world, which may mean that back-
bone capacity isn’t available. SBC Corporation, which serves at
least a third of subscribers in the United States, has announced
its Project Pronto, which is designed to bring DSL to a least 80
percent of its subscribers. This still leaves a substantial percent-
age of users and exchanges for which DSL is available but that
still cannot get the service. Furthermore, many smaller com-
munities simply do not have enough potential subscribers to
make it worthwhile for the LECs to provide DSL. Internet
access is not defined as a lifeline service and DSL is generally
not regulated, so market conditions dictate its availability.
This leaves wireless as an alternative. Terrestrial wireless
depends on having a line of sight, which precludes it for a sig-
nificant number of potential subscribers. Furthermore, it is a
shared medium and therefore is subject to the same kind of
bandwidth limitations that cable experiences. Satellite technol-
ogy circumvents line-of-sight requirements, but radio uplinks
are expensive. Many satellite services use the telephone line for
the upstream direction. This is acceptable for Internet service
for residences, but it requires a second telephone line, which is
expensive. A telephone uplink for businesses is unsatisfactory
because of bandwidth limitations for e-mail attachments.

The Oversubscription Issue


All telecommunications service providers oversubscribe
their bandwidth. By oversubscription, we mean that the total
Application Considerations 101

potential demand is more than the available capacity. The


principle is built into every telecommunications network
because the service provider can predict with some degree of
confidence that users will not all require service simultane-
ously. Occasionally, abnormalities such as storms, national
emergencies, earthquakes, and critical news events cause
traffic to exceed the peaks against which the network was
designed. In such a case, every network has some method of
protecting itself from collapse while still providing service to
crucial subscribers.
The Internet, in particular, has numerous chokepoints, and it
is impossible for the end user to determine where the band-
width restriction actually is. A dial-up user knows that about 53
kbps downstream and 33.6 kbps upstream is the maximum a
modem can transfer, so the other chokepoints in the network
are probably irrelevant. The objective of subscribers with
always-on access is to eliminate the access restriction or at least
to provide predictable performance. DSL can deliver pre-
dictable performance, but cable cannot because of the way it is
designed. The only way a service provider can reduce the over-
subscription ratio is by reducing the number of households
served by a coaxial segment, which is expensive.

Voice and Video Service


Most of the applications we have discussed are asymmetric
in nature, so the limited upstream bandwidth of cable, wire-
less, and ADSL present no problems. Once voice and video
are added to the equation, however, a symmetrical channel is
required. Furthermore, these services, which can be lumped
under the label voice over IP (VoIP), have other characteris-
tics that impose entirely different requirements on the net-
work than data. First, VoIP operates in real time. A packet
stream leaving the transmitter must not be delayed on its
way to the receiver or intelligibility is lost. Anyone who has
talked over a satellite circuit knows the delay can be discon-
certing unless the parties to the conversation change their
normal mode of talking. The nature of the IP protocol is that
102 Access Technologies: DSL and Cable

packets do not arrive evenly spaced. This variation in spac-


ing between packets is called jitter. If allowed to remain,
jitter distorts the voice and further affects intelligibility.
Buffering packets and releasing them in a steady stream can
cure jitter, but this adds to delay, so it is effective only with-
in limits. These factors are discussed in more detail in the
McGraw-Hill I-book Voice and Video over IP, available at
http://shop.mcgraw-hill.com/cgi-bin/pbg/indexebooks.html.
Another difference is that error correction is unnecessary
with voice. Bit errors that would render a data file unusable
are of no consequence in voice and video, and furthermore,
real-time services cannot tolerate the time delay involved in
retransmission. Therefore, VoIP runs under UDP instead of
TCP. Since data cannot tolerate errors, this makes it imprac-
tical to combine it with voice at the source.
Since the ILECs own the cable plant, they have little inter-
est in deviating from their traditional architecture. The
CLECs, however, must lease cable pairs, so they have con-
siderable interest in increasing the carrying capacity through
the use of VoDSL. Because of the nature of voice sessions,
one of the symmetrical DSL options—generally SDSL—is
used. As we have seen, the loop length is limited to about
10,000 ft (3,000 m) from the central office, which means that
VoDSL is not a universal alternative. It has an advantage over
cable, however, in that the upstream direction is not a shared
medium. Moreover, if the line protocol is ATM all the way to
the subscriber, the delays inherent in IP are avoided.
Voice over cable is feasible and is part of the serving plan
of large cable service providers. The latest version of DOC-
SIS supports voice over cable networks, but the service has
not yet been widely used.

Branch Offices and Telecommuters


The old model of a fixed office where everyone reports to
work in a central building is largely a thing of the past.
Today’s office has multiple locations, and many pressures
Application Considerations 103

weigh in favor of enabling workers to access the corporate


network from home or in a neighborhood office. Crowded
highways, air pollution standards, and personal preference
make it desirable to enable people to work away from the
office at least part of the time. Telecommuting solves staffing
problems for companies that need call center agents during
calling peaks, and is an ideal application for workers who
have small children, disabilities, or other reasons for working
less than full time.
Remote office alternatives have been limited until recent-
ly. For corporate branch offices the choices have been expen-
sive and not truly suitable. A 64-kbps fixed or frame relay
access circuit is insufficient for remote applications that need
wide bandwidth on demand. T1/E1 is plenty of bandwidth,
but it is expensive with nothing in between. Now, new access
technologies, principally DSL, can provide a satisfactory
alternative under the right conditions. The conditions have to
do with the suitability of the network between the remote and
central locations. QoS standards are not yet fully developed.
When they are, it will probably be years before the Internet
can provide reliable transport. Other alternatives may be fea-
sible if plenty of bandwidth is available.

T1/E1 Access
If a full point-to-point T1/E1 can be justified to a branch office,
the same bandwidth is available in both the access circuit and
the backbone and the facility has plenty of stability to handle
both voice and data. T1/E1 is an ideal medium if the bandwidth
requirements justify the cost. This is often the case with branch
offices, but it is rarely justified for telecommuting.

DSL Access
If the branch or home office is within range, DSL can be an
ideal medium. Scalability problems that are typical of T1/E1
service are resolved, and the service provider’s architecture
104 Access Technologies: DSL and Cable

may solve the QoS issue. As we have seen, most ILECs use
ATM in the access circuit and in their backbone. If a single
carrier provides the service end-to-end, the quality should be
controllable within tight limits and could be enough to sup-
port both voice and data. Many PBX manufacturers provide
a proprietary IAD that works with their digital telephone.
Security is tight, and the telecommuter or branch office
worker should have facilities equivalent to a station collocat-
ed with the PBX.
Quality is less predictable when more than one service
provider is involved. In a metropolitan area with two ILECs,
if each one has an ATM network with a network-to-network
interface, the service may be equivalent to the ATM network
of a single provider. When the network extends between two
metropolitan areas, it may be necessary to connect them over
an IXC’s IP or ATM backbone. Some IXCs provide service
level agreements (SLAs) that may be high enough in quality
to support a branch office or telecommuting application. Be
aware, however, that many IXCs quote their SLAs as an aver-
age over 30 days. To obtain satisfactory quality of service,
you must know the worst-case delay and jitter figures to
determine whether quality will be satisfactory to support the
application.

Cable and Wireless


Cable and wireless can also be used for telecommuting
access, with certain provisos. First is the question of whether
the service provider permits any but its own services to use
its facility. If it does not, it may not be possible to control the
path sufficiently to obtain satisfactory QoS. In addition, the
nature of the medium, with its asymmetric data flow and
shared bandwidth, may render it unsatisfactory for voice
communication unless the service provider has a method of
prioritizing voice packets.
Branch offices can often use the PSTN for all voice com-
munication, leaving the private network to handle only data.
This is not the case with most home office applications,
Application Considerations 105

however. Telecommuters usually prefer to have real-time


access to both voice and data, receiving telephone calls and
e-mail as if they were in the office. DSL is clearly the best
access medium for this application.

Dial-up access
For some access applications, dial-up is the only feasible
alternative. Despite the bandwidth limitations of the PSTN, it
is almost universally available, although not necessarily at a
low cost. Many ISPs provide dial access to the Internet as
well as e-mail. A premium price may be charged, and access
is available almost anywhere, but it is almost precluded as a
telecommuting medium because of lack of bandwidth. Some
PBX manufacturers provide devices to enable the use of a
digital telephone over an analog loop, but these are not high-
ly effective for full-time telecommuting.
If bandwidth limitations prevent the use of dial-up, then
ISDN may be an alternative. Basic rate ISDN suffers from
the same problems as DSL, however, in that it has an 18,000-
ft (5,500 m) limitation. Furthermore, unless the channels are
bonded, the bandwidth may not be enough of an improve-
ment over analog to justify the extra cost. BRI is often
metered, even in locations where flat-rate analog is available,
and an always-on connection to the Internet or headquarters
can be prohibitively expensive. Finally, BRI suffers from a
general lack of availability in the United States. Many
European countries have much better ISDN availability, but
many central offices in the United States have not been
equipped for BRI. Some multioffice wire centers have ISDN
available in one switching system, but a number change may
be required to switch from analog to ISDN service.
One promising service, always-on dynamic ISDN
(AO/DI), uses the D channel for a full-time connection to the
Internet. If more bandwidth is needed for Web surfing or to
transfer an e-mail file attachment, the service brings a B
channel online long enough to provide the necessary band-
width, after which it disconnects. The theory of AO/DI is
106 Access Technologies: DSL and Cable

appealing, but the practice is much less so because few LECs


are equipped to offer the service.

Data Network Access


DSL can be an effective network access method for both
frame relay and point-to-point data services. The primary
consideration is whether the providers have a service offering
that takes advantage of DSL as the access method. Nothing
technical prevents the data network provider from connecting
over the ILEC’s network or that of a competitive DSL
provider.
Security becomes a concern when the network goes out-
side the bounds of the service provider’s closed network and
connects to the public Internet. A network that can be
Y
reached over the Internet is technically known as a virtual
FL
private network (VPN). The access circuit for a VPN can be
DSL, cable, dedicated, or wireless. Any of these should work
well for typical VPN applications, which are e-mail and file
AM

transfer. Time-sensitive applications such as voice require


tight control over latency. Before using shared media tech-
nology or any Internet connectivity for a voice VPN, QoS
TE

must be carefully considered.


For a straight data VPN, any of the access technologies
should be satisfactory. Security considerations, which
include authorization, authentication, encryption, and fire-
walling, must be taken into account. These functions are nor-
mally included in VPN access devices.

Summary and Conclusions


The principal conclusion that most users will reach after
studying the access arena is that the perfect solution does not
exist. Every technology discussed in this volume has its
drawbacks, and it is important that prospective subscribers
understand them. The first factor affecting the choice is like-
Application Considerations 107

ly a simple question of what is available. A significant per-


centage of users will go no further than a POTS line with a
V.90 modem—not because it is ideal, but just because it is
available and affordable.
For many users, dial-up is not an acceptable option
because of the nature of the application. Data wide area net-
works will require some form of dedicated access. If DSL is
not available or if it has insufficient bandwidth, then the
choice turns to T1/E1 or a 56/64-kbps access loop. The latter
will normally be of interest only for business applications.
Dedicated 56/64-kbps connections lack the necessary band-
width and T1/E1 is prohibitively expensive. Exceptions to
these general comments will occur, but they are rare.
Except for telecommuting, most residential data applica-
tions are confined to Internet access. If DSL and cable are
both available, and if the prices are competitive, the deciding
factor is performance. It is impossible to generalize which
will provide the best performance. The answer is that it
depends on a variety of factors including distance, number of
users sharing the medium, time of day, and the ISP’s service
quality. The best way to evaluate the service is to ask ques-
tions of other users. Unfortunately, what they tell you today
may not be valid tomorrow.
Appendix 1: Glossary

AAL: ATM adaptation layer.


ABR: available bit rate.
AC: alternating current.
ADSL: see asymmetric digital subscriber line.
alternate mark inversion (AMI): the T carrier line-coding sys-
tem that inverts the polarity of alternate ones bits.
always-on dynamic ISDN (AO/DI): an ISDN service that keeps
the D channel actively connected to a service provider. B chan-
nels are called in as needed.
AMI: see alternate mark inversion.
ANSI: American National Standards Institute.
AO/DI: see always-on dynamic ISDN.
API: application program interface.
ARP: Address Resolution Protocol.

109

Copyright © 2002 by The McGraw-Hill Companies, Inc. Click here for Terms of Use.
110 Access Technologies: DSL and Cable

asymmetric digital subscriber line (ADSL): a technology for


multiplexing a high-speed data or compressed video signal above
the voice channel in a subscriber loop.
asynchronous transfer mode (ATM): a broadband connection-
oriented switching service that carries data, voice, and video
information in fixed-length 48-octet cells with a 5-octet header.
ATM: see asynchronous transfer mode.
AWG: American wire gauge.
bandwidth: the range of frequencies a communications channel
is capable of carrying without excessive attenuation.
baseband: a form of modulation in which data signals are pulsed
directly on the transmission medium without frequency division.
basic rate interface (BRI): the basic ISDN service consisting of
two 64-kbps information or bearer channels and one 16-kbps data
or signaling channel.
baud: the number of data signal elements or symbols per second
a data channel is capable of carrying.
B channel: the 64-kbps “bearer” channel that is the basic build-
ing block of ISDN. The B channel is used for voice and circuit
switched or packet switched data.
bin: frequency segments into which a discreet multitone DSL line
is divided.
binder group: a 50-pair group of cable pairs that is bound
together with a colored binder.
BRI: see basic rate interface.
bridge: circuitry used to interconnect networks at the MAC layer.
bridged tap: any section of a cable pair that is not on the direct
electrical path between the central office and the user’s premises,
but is bridged onto the path.
broadband: a term used to describe high-bandwidth transmission
of data signals. Technically, any signal greater than primary rate
(T1/E1).
cable modem termination system (CMTS): a system of inter-
facing the cable network with a data network.
CAP: carrierless amplitude phase; competitive access provider.
Carrier Sense Multiple Access with Collision Detection
(CSMA/CD): a system used in contention networks where the
network interface unit listens for the presence of a carrier before
Appendix 1: Glossary 111

attempting to send and detects the presence of a collision by mon-


itoring for a distorted pulse.
CATV: community antenna television.
CBR: constant bit rate.
CDPD: cellular digital packet data.
central office (CO): a switching center that terminates and inter-
connects lines and trunks from users.
channel service unit (CSU): an apparatus that terminates a T1
line providing various interfacing, maintenance, and testing
functions.
CLEC: see competitive local exchange carrier.
CM: cable modem.
CMTS: see cable modem termination system.
CO: see central office.
competitive local exchange carrier (CLEC): a company offer-
ing local service in competition with an incumbent local
exchange carrier.
complement: a group of 50 cable pairs (25 pairs in small cable
sizes) that are bound together and identified as a unit.
CRC: cyclical redundancy checking.
cross-connect: a wired connection between two or more elements
of a telecommunications circuit.
crosstalk: the unwanted coupling of a signal from one transmis-
sion path into another.
CSMA/CA: Carrier Sense Multiple Access with Collision
Avoidance.
CSMA/CD: see Carrier Sense Multiple Access with Collision
Detection.
CSU: see channel service unit.
CTS: clear-to-send.
Data over Cable Interface Specification (DOCSIS): an industry
specification for providing data and voice communications over
CATV.
DC: direct current.
D channel: the ISDN 16-kbps data channel that is used for out-
of-band signaling functions such as call setup.
DCS: digital cross-connect system.
112 Access Technologies: DSL and Cable

dedicated circuit: a communications channel assigned for the


exclusive use of an organization.
dense wave-division multiplexing (DWDM): the process of mul-
tiplexing fiber optics with multiple wavelengths—40 or more
with today’s technology.
DHCP: see Dynamic Host Configuration Protocol.
digital loop carrier (DLC): A multichannel digital device that
enables several subscribers to share a single facility in the local loop.
digital subscriber line access multiplexer (DSLAM): a device
in the central office that splits voice and data signals and connects
voice to the PSTN and data to a high-speed backbone.
dispersion: the rounding and overlapping of a light pulse that
occurs at different wavelengths because of reflected rays or the
different refractive index of the core material.
DLC: see digital loop carrier.
DMT: discrete multitone.
DNS: see domain name service.
DOCSIS: see Data over Cable System Interface Specifications.
domain name service (DNS): a service that translates host
names to IP addresses.
downstream: data flowing from the headend to the subscriber.
DSL: digital subscriber line.
DSLAM: see digital subscriber line access multiplexer.
DSMA/CD: Digital Sense Multiple Access with Collision
Detection.
DSSS: direct sequence spread spectrum.
DWDM: see dense wave-division multiplexing.
Dynamic Host Configuration Protocol (DHCP): a protocol that
allocates IP addresses to network clients at startup.
echo canceler: an electronic device that processes the echo signal
and cancels it out to prevent annoyance to the talker.
ESF: see extended super frame.
ETSI: European Telecommunications Standards Institute.
extended super frame (ESF): a T1 carrier framing format that
provides 64-kbps clear channel capability, error checking, 16-state
signaling, and other data transmission features.
FCC: Federal Communications Commission.
Appendix 1: Glossary 113

FDM: see frequency-division multiplexing.


FDMA: frequency-division multiple access.
FEC: forward error correction.
firewall: a device that protects the connection between a network
and an untrusted connecting network such as Internet. The fire-
wall blocks unwanted traffic from entering the network and
allows only authorized traffic to leave.
frame relay: a data communication service that transports
frames of information across a network to one or more points.
Cost is based on three elements: committed information rate,
access circuit, and port speed.
frequency-division multiplexing (FDM): dividing the bandwidth
of a transmission medium by separating frequency segments.
FSK: frequency-shift keying.
FTP: File Transfer Protocol.
Gbps: gigabits per second.
G.lite: a modified type of ADSL that enables users to connect to
a DSL line without a splitter.
HDSL: see high-speed digital subscriber line.
HFC: see hybrid fiber-coax.
high-speed digital subscriber line (HDSL): a protocol for deliv-
ering T1/E1 over two pairs of wire for up to 12,000 ft (4,000 m).
hybrid: a multiwinding coil or electronic circuit used in a four-
wire terminating set or switching system line circuit to separate
the four-wire and two-wire paths.
hybrid fiber-coax (HFC): a cable television transmission method
that uses a combination of fiber optics to a neighborhood node
and coaxial cable to the subscribers.
IAD: integrated access device.
IDSL: see ISDN digital subscriber line.
ILEC: see incumbent local exchange carrier.
impedance: the ratio of voltage to current in an alternating cur-
rent electrical circuit.
in-band signaling: telephone signaling in which the signals are
carried in the talking path.
incumbent local exchange carrier (ILEC): the traditional tele-
phone company that serves a particular franchised area.
114 Access Technologies: DSL and Cable

integrated services digital network (ISDN): a set of standards


promulgated by ITU-T to prescribe standard interfaces to a
switched digital network.
interexchange carrier (IXC): a common carrier that provides
long-distance service between local access transport areas.
IP: Internet Protocol.
ISDL: see ISDN digital subscriber line.
ISDN: see integrated services digital network.
ISDN digital subscriber line (IDSL): a DSL protocol that uses a
basic rate ISDN signal. Provides symmetrical speeds of 144 kbps.
ISM: industrial, scientific, and medical.
ISO: International Standards Organization.
ISP: Internet service provider.
ITFS: instructional television fixed service.
ITU: International Telecommunications Union.
IXC: see interexchange carrier.
jitter: variation in arrival intervals of a stream of packets.
kbps: kilobits per second.
kHz: kilohertz.
LAN: local area network.
LEC: see local exchange carrier.
LEOS: see low-earth-orbit satellite.
link budget: the amount of loss that can be tolerated between the
two ends of a link while still maintaining the required service
quality.
LLC: logical link control.
LMDS: see local multipoint distribution service.
local exchange carrier (LEC): any local exchange telephone
company that serves a particular area. See also competitive local
exchange carrier; incumbent local exchange carrier.
local multipoint distribution service (LMDS): a microwave-
based service that serves multiple voice and data users from a
central base station hub.
low-earth-orbit satellite (LEOS): a global personal communica-
tions service technology that uses a constellation of satellites
orbiting earth at a few hundred miles up for communications with
hand-held units.
Appendix 1: Glossary 115

MAC: media access control.


main distributing frame (MDF): the cable rack used to termi-
nate all distribution and trunk cables in a central office.
Mbps: megabits per second.
MCU: multipoint control unit.
MDBS: mobile database station.
MDF: see main distributing frame.
MDIS: mobile data intermediate station.
MHz: megahertz.
MMDS: multipoint multichannel distribution service.
modulation: the process by which some characteristic of a carrier
signal, such as frequency, amplitude, or phase, is varied by a low-
frequency information signal.
MPEG: Motion Picture Experts Group.
multiple plant: cable plant in which cable pairs are connected in
parallel.
near-end crosstalk (NEXT): the amount of signal received at the
near end of a circuit when a transmit signal is applied at the same
end of the link.
neighborhood gateway: an enclosure that houses telephone
equipment such as DSL or digital line carrier in a neighborhood
to shorten the length of the wire run.
NEXT: see near-end crosstalk.
NIC: network interface card.
Open Systems Interconnect (OSI): a seven-layer data communi-
cations protocol model that specifies standard interfaces that all
vendors can adapt to their own designs.
PBX: private branch exchange.
PCM: see pulse code modulation.
PCS: personal communication service.
PDA: personal digital assistant.
PDU: protocol data unit.
permanent virtual circuit (PVC): in a data network a PVC is
defined in software. The circuit functions as if a hardware path is
in place, but the path is shared with other users.
plain old telephone service (POTS): the standard analog dial-up
service provided by the PSTN.
116 Access Technologies: DSL and Cable

PON: passive optical network.


POP: point of presence.
POTS: see plain old telephone service.
PRI: see primary rate interface.
primary rate interface (PRI): in North America, a 1.544-Mbps
information-carrying channel that furnishes ISDN services to end
users and consists of 23 bearer channels and one signaling chan-
nel. In Europe, a 2.048-Mbps channel consisting of 30 bearer and
two signaling channels.
private line: see dedicated circuit.
protocol: the conventions used in a network for establishing com-
munications compatibility between terminals and for maintaining
the line discipline while they are connected to the network.
provisioning: the process of assembling all of the elements that
make up a circuit.
Y
PSD: power spectral density.
FL
PSK: phase-shift keying.
PSTN: see public switched telephone network.
PTT: postal telephone and telegraph.
AM

public switched telephone network (PSTN): a generic term for


the interconnected networks of operating telephone companies.
pulse code modulation (PCM): a digital modulation method that
TE

encodes a voice signal into an 8-bit digital word representing the


amplitude of each pulse.
PVC: see permanent virtual circuit.
QAM: see quadrature amplitude modulation.
QoS: quality of service.
QPSK: see quadrature phase-shift keying.
quadrature amplitude modulation (QAM): a method of modu-
lating digital signals on a carrier using both amplitude and phase
modulation.
quadrature phase-shift keying (QPSK): a method of modulat-
ing digital signals on a carrier encoding two digital bits on each
of four phase states.
RADSL: rate-adaptive digital subscriber line.
ranging: the process by which a cable modem evaluates the time
delay in transmitting to the headend.
Appendix 1: Glossary 117

RF: radio frequency.


RTS: request-to-send.
SAR: segmentation and reassembly.
SDH: synchronous digital hierarchy.
SDSL: see single-pair digital subscriber line.
serving wire center: the ILEC wire center that serves a particular
subscriber.
short: a circuit impairment that exists when two conductors of
the same pair are connected at an unintended point.
sidetone: the sound of a talker’s voice audible in the handset of
the telephone instrument.
single-pair digital subscriber line (SDSL): a protocol similar to
HDSL except that it uses only one cable pair. It also provides a
POTS line under the data.
SLA: service level agreement.
SNMP: Simple Network Management Protocol.
SNR: signal-to-noise ratio.
SONET: synchronous optical network.
splitter: a device that connects and splits two different media
types. For example, a DSL splitter separates the data and telephone
signals. Also, a device used to connect branches of a cable system.
spread spectrum: a radio modulation method that transmits a
signal over a broad range of frequencies (direct sequence method)
or rapidly jumps from one frequency to another (frequency hop-
ping). Provides excellent security and resists interference.
SVC: switched virtual circuit.
SWC: serving wire center.
symbol: see baud.
tap: a device for connecting a subscriber’s premises to a coaxial
CATV line. The tap isolates the subscriber from other services on
the coax.
TCP/IP: Transmission Control Protocol/Internet Protocol.
TDM: see time-division multiplexing.
TDMA: time-division multiple access.
time-division multiplexing (TDM): a method of combining sev-
eral communications channels by dividing a channel into time
increments and assigning each channel to a time slot. Multiple
118 Access Technologies: DSL and Cable

channels are interleaved when each channel is assigned the entire


bandwidth of the backbone channel for a short period of time.
time slot: in a TDM system, a unit of time sufficient for the
transmission of 8 bits that is assigned to each byte. This unit of
time recurs at the same instant in a transmission frame.
TRI: telephony return interface.
trunk: a communications channel between two switching systems
equipped with terminating and signaling equipment.
UDP: User Datagram Protocol.
UNE: unbundled network element.
upstream: data flow direction from the subscriber to the headend.
URL: uniform resource locater.
USB: universal serial bus.
UTP: unshielded twisted pair.
VBR: variable bit rate.
VC: virtual channel.
VCI: virtual channel identifier.
VDSL: see very-high-bit-rate digital subscriber line.
very-high-bit-rate digital subscriber line (VDSL): a DSL type
that delivers 13 to 52 Mbps downstream and 1.5 to 2.3 Mbps
upstream over a single copper twisted pair over a short range.
video on demand (VOD): the delivery of video services to cus-
tomers in response to their specific request. VOD is contrasted to
conventional cable television, where all channels are delivered
over the medium.
VOD: see video on demand.
VoDSL: voice over DSL.
VoIP: voice over IP.
VPI: virtual path identifier.
VPN: virtual private network.
VSAT: very-small-aperture terminal.
WDM: wave-division multiplexing.
WLL: wireless local loop.
2 binary 1 quaternary (2B1Q): a line-coding technique in which
2 bits are mapped to one quaternary symbol.

You might also like