Telecom 101

Fourth Edition

Eric C. Coll M.Eng.

Teracom Training Institute

www.teracomtraining.com

1
Copyright © 2016, Teracom Training Institute, Ltd.
All rights reserved. No part of this document nor the accompanying presentation may be repro-
duced or transmitted in any form or by any means, electronic or mechanical, including but not
limited to duplication, photocopying, scanning, peer-to-peer file sharing or by any other infor-
mation storage and retrieval system, without permission in writing from the copyright holder.
Trademarks and registered trademarks referenced in this text are the property of their respec-
tive holders and are used for identification purposes only. Teracom is a registered trademark of
Teracom Technologies Inc.
Notice: The information contained in this document is provided as general background infor-
mation only. Design and implementation of a communication system requires professional ad-
vice to identify and resolve issues specific to that particular system, including but not limited to
performance and security issues. Additionally, while we have striven to be as accurate as pos-
sible, we make no representation of fitness or warranty that the information provided is 100%
accurate. The information in this document is not to be relied upon as professional advice, nor
is it to be used as the basis of a design. Users of this document agree to hold the author and
Teracom Training Institute Ltd. harmless from any liability or damages. Acceptance and use of
this document shall constitute indication of your agreement to these conditions.

eBook ISBN 9781894887786 Print ISBN 9781894887038

FOURTH EDITION. April 2016 R4
inquiries:
Teracom Training Institute
PO Box 3376
Champlain NY 12919
1-877-412-2700
www.teracomtraining.com

2
For the circus.

A little learning is a dangerous thing;

Drink deep or taste not the Pierian spring:
There shallow draughts intoxicate the brain,
And drinking largely sobers us again.
– Alexander Pope

3
Preface
This book is based on the course materials from Teracom’s instructor-led Course 101 Tele-
com, Datacom and Networking for Non-Engineering Professionals. It covers the core knowl-
edge set required in the telecom business today.
The Fourth Edition is completely updated, featuring the new-generation Optical Ethernet, IP
and MPLS-based network. More than half the book is on the IP telecom network.
It has been written for those new to telecom, those getting up to speed, those filling in gaps,
and for all those who do not have Engineering degrees specializing in telecommunications.
Our goal is to demystify jargon and buzzwords, and put in place a structured understanding of
telecom, the technologies and services, and most importantly, the underlying ideas – and how
it all fits together.
The knowledge in this book is drawn from over 35 years of experience in the telecom busi-
ness, working for telephone companies in jobs including Junior R&D Engineer, systems engi-
neer, consultant writing R&D tax credit project reports, and teaching many private onsite cour-
ses for carriers.
The style of this book, the selection of material, its ordering and pacing, and the jokes, are the
result of being the instructor at hundreds and hundreds of 2-day and 3-day seminar courses on
these topics over the past 24 years.
The result is this book, Telecom 101: the course materials for an instructor-led course that
costs US$1395 to attend, augmented with substantial additional material, available in softcover
textbook and ebook.
Telecom 101 is intended to be used as a textbook, sequentially building one concept on an-
other like an instructor-led course. It is also intended to be a valuable day-to-day reference
handbook and glossary.
Let’s get started!
EC
April 2016

4
Table of Contents
Introduction
Our Approach
How the Text is Organized
How to Use This Text
The Three Answers
Fundamentals of Telephony
History of Telecommunications
The Public Switched Telephone Network
Analog
Capacity Restrictions
Problems with Analog Transmission
Plain Ordinary Telephone Service (POTS)
Network Addresses: Telephone Numbers
SS7
Voice over IP (VoIP)
Switching
Telephone Network Architecture
Telephone Switches
Traditional PBX and Centrex
SIP, Soft Switches, Hosted PBX and IP Centrex
SIP Trunking
The Telecommunications Industry
US Telephone Companies
AT&T and Verizon
Canadian Telephone Companies
PSTN Switching Center Hierarchy
Switched Access to LD Competitors: LECs, POPs and IXCs
High-Capacity Access to Long-Distance
CLEC: Collocations and Dark Fiber
Bypass
Competitive Carrier Network Model: Rings, POPs and MANs
Digital
Analog and Digital: What do we really mean?
Continuous vs. Discrete Signals
Voice Digitization (Analog-Digital Conversion)
Voice Reconstruction (Digital→ Analog Conversion)
Voice Digitization: 64 kb/s G.711 Standard
Digital Video, H.264 and MPEG4
Digital Transmission: Binary Pulses
Transmission Systems
Channelized Time Division Multiplexing (TDM)
Multiplexers
The Digital Hierarchy: Legacy Channelized Transmission Speeds
Digital Carrier Systems:Legacy Transmission Technologies
Framing
ISDN
Statistical Time Division Multiplexing
Framing on IP Packet Networks
Coexistence and Transition from Channels to Packets
The Cloud and Service Provisioning
Anatomy of a Service
The Network “Cloud”
Inside the Network Cloud
Network Equipment
Service Provisioning Summary
Fiber Optics
Fiber Basics
Glass Fiber and Fiber Cables
Optical Wavelengths, Bands and Modes
Wave-Division Multiplexing: CWDM and DWDM
Optical Ethernet
Network Core
Metropolitan Area Network
Fiber to the Premise (FTTP) & PONs

5
DSL and Cable Modems: Last Mile on Copper
Modems
Modulation Techniques
Digital Subscriber Line (DSL)
DSLAMs
Fiber to the Neighborhood (FTTN), DSL to the Premise
DSL Standards
Broadband Carriers: FTTN & Broadband Coax to the Premise
DOCSIS and Cable Modem Standards
Wireless
Radio
Mobile Networks
Cellular Radio and 1G
Second Generation: Digital Cellular
Mobile Internet and “Data” Plans
FDMA, TDMA, CDMA and OFDM
3G Cellular: CDMA
4G Mobile Cellular: LTE
Wireless LANs
Communication Satellites
“Data” Communications
Convergence: Treat Everything Like Data
Information Theory
Data Circuit Model
DTE: Data Terminal Equipment
Data Circuits
DCE: Data Circuit-Terminating Equipment
Point-to-Point Circuits
Multidrop Circuits
LANs: Local Area Networks
WANs: Wide Area Networks
Coding, Frames and Packets
Essential Functions for Communication
Coding Quantities: Number Systems
Coding Text
“Asynchronous”: Start/Stop/Parity
Frames and MAC Addresses
Networks, Packets & IP Addresses
Packets vs. Frames
IP Packets
The OSI Layers and Protocol Stacks
Protocols and Standards
ISO OSI Reference Model
The OSI 7-Layer Model
Physical Layer: 802.3, DSL, DOCSIS
Data Link Layer: 802 MAC
Network Layer: IP and MPLS
Transport Layer: TCP and UDP
Session Layer: POP, SIP, HTTP
Presentation Layer: ASCII, Encryption, Codecs
Application Layer: SMTP, HTML, English …
Protocol Stacks
Protocol Stack in Operation: Russian Dolls
Standards Organizations
Ethernet, LANs and VLANs
LAN Basics
Ethernet and 802 Standards
LAN Cables and Categories
LAN Switches: Layer 2 Switches
VLANs
IP Networks, Routers and Addresses
Definition of Network
Simplest Example: Private Network
Routers and Customer Edge
IPv4 Address Classes
DHCP
Public and Private IPv4 Addresses
Network Address Translation

6
 TCP and UDP
IPv6
IPv6 Address Allocation and Address Types
MPLS and Carrier Networks
Introduction
Carrier Packet Network Basics
Service Level Agreements
Provider Equipment at the Customer Premise
Virtual Circuit Technologies
Packet-Switching using Virtual Circuits
Frame Relay using Virtual Circuits
ATM
MPLS
MPLS VPN Service for Business Customers
MPLS and Diff-Serv to Support Class of Service
MPLS for Integrated Access
MPLS for Traffic Aggregation
M is for Multiprotocol: Virtual Private LAN Service (VPLS)
The Internet
A Network To Survive Nuclear War
The Inter-Net Protocol
Internet Service Providers
World Wide Web
Domain Name System
Hypertext
MIME and Base-64 Encoding for Email Attachments
Internet Telephony & VSPs
Internet VPNs
T1
T1 History and Applications
T1 Circuit Components
Operation
T1 Framing
Pulses and Line Code: AMI
Synchronization: Bit-Robbing
B8ZS and 64 kb/s Clear Channels
How T1 Is Provided
Fractional T1, DACS and Cross-Connects
Subrate Data Circuits 1.2 kb/s to 56 kb/s
Voice Services and Jargon
Local Voice Services
Long Distance Voice Services
Acronyms and Abbreviations
About Teracom
About the Author
Public Seminars
Private Onsite Seminars
DVD-Video Courses
GSA Schedule
Online Courses
TCO Telecommunications Certification

7
1
Introduction
In this chapter, we discuss the approach taken in organizing the topics in this text and provide
suggestions for how to use it. The chapter is completed with the answer to all questions about
telecommunications.

1.1 Our Approach

Our approach in organizing the topics and the order we present them in can be summed up
with a simple philosophy: Start at the beginning of the story. Progress in a logical order, build-
ing one concept on another. Finish at the end of the story. Avoid jargon. Speak in plain English.
We’ve been applying this approach to telecommunications training for twenty-five years, and it
is the philosophy behind all of Teracom’s training products and services, including online cour-
ses, certifications, instructor-led training, DVD video courses, eBooks and printed textbooks.
Our objective is to fill in the gaps and build a solid base of knowledge, put a structure in place,
and show how everything fits together.
This guide is not intended to be an authoritative reference on any particular technology; speci-
fications and low-level details are best found in the official standards documents.
This is a practical guide, and we hope it will serve as a valuable resource in navigating the
world of telecommunications, and in building knowledge and understanding that lasts a life-
time.

1.2 How the Text is Organized

This book is based on Teracom’s famous three-day instructor-led core training Course 101
Telecom, Datacom and Networking for Non-Engineering Professionals, plus additional material
going beyond the Course 101 content.
The material has been organized to start at the beginning of the story with the invention of the
telephone in 1874, progress in a logical order building one concept on top of another, and fin-
ish at the end with the IP network and MPLS traffic management.
The topics can be grouped into three parts: Fundamentals of Telecommunications, Telecom-
munications Technologies, and the IP Telecommunications Network.
Part I: Fundamentals of Telecommunications
1 Introduction
2 Fundamentals of Telephony
3 Switching
4 The Telecommunications Industry
We begin with the fundamentals of telephony and the telephone network – the basis for under-
standing everything else. First is the invention of the telephone and the development of the
Public Switched Telephone Network: loops and trunks, circuit-switching, analog, the voiceband
and other key aspects of Plain Ordinary Telephone Service, including address signaling and
SS7; and fundamentals of Voice over IP.
Then we examine switching, starting with traditional telephone switches: Centrex, PBX and
PBX trunks, and how that relates to the newer ideas of softswitches, Hosted PBX and SIP
trunking. This is completed with an overview of call center equipment including IVRs and
ACDs.
This part is completed with a chapter on the telecommunications business: Local Exchange
Carriers and Inter-Exchange Carriers, ILECs and CLECs, the main players and how they inter-
connect.
Part II: Telecommunications Technologies

8
 5 Digital
6 Transmission Systems
7 The Network Cloud and Service Provisioning
8 Fiber Optics
9 DSL and Cable Modems: Last Mile on Copper
10 Wireless
The second part is devoted to telecommunications technologies: the actual methods used to
implement circuits and services. We begin with digital: what digital is, how voice and video are
digitized, and how digitized information is actually transmitted.
The next chapter is transmission systems: the high-capacity systems developed to carry many
users’ traffic. This starts with the installed base of channelized systems, the hierarchy of DS0,
DS1 and DS3 rates and an overview of T1, T3, SONET and ISDN. Then our attention turns to
the new generation packetized systems, introducing the concepts of overbooking and band-
width on demand instead of channels, how this is implemented with frames and packets, coex-
istence and transition from channels to packets.
Then we understand the “Network Cloud”, how services are actually implemented, the three
basic types of services and the equipment used for each.
Completing this part are three chapters on the technologies used to implement the network,
circuits and services. First is Fiber Optics including fundamentals of fiber, wave-division multi-
plexing, the network core, Metropolitan Area Networks and fiber to the premise.
Second is DSL and Cable Modems to implement the “last mile” on existing copper plant, cov-
ering fundamentals of modems, DSLAMs, VDSL, broadband and cable modems.
Last is Wireless, concentrating on mobile communications: cellular and mobility concepts, the
technologies TDMA, CDMA and OFDM, the generations from 1G to 4G, and the systems
GSM, UMTS, 1X and LTE. This chapter is completed with WiFi and satellite.
Part III: The IP Packet-Switched Telecom Network
11 “Data” Communications Concepts
12 Coding, Frames and Packets
13 The OSI Layers and Protocol Stacks
14 Ethernet, LANs and VLANs
15 IP Networks, Routers and Addresses
16 MPLS and Carrier Networks
17 The Internet
18 Wrapping Up
The third part of the book is dedicated to the new-generation IP telecommunications network.
We begin by understanding how convergence was achieved by treating voice and video like
data – then accordingly, cover the fundamentals of what used to be called “data communica-
tions”: DTEs, DCEs, LANs and WANs and the crucial concepts of packets and frames.
There are so many functions that need to be performed to implement phone calls, television,
web browsing, email and everything else over the IP network, a structure is necessary to be
able to identify and discuss separate issues separately. For this purpose, we use the OSI Ref-
erence Model and its layers, identifying what the layers are, examples of protocols for each
layer and how they work together in a protocol stack.
Then we begin moving up the layers. Having already mostly covered the physical connections
(Layer 1) in the chapters on fiber, DSL, cable and wireless, the next chapter is on Ethernet,
LANs and VLANs (Layer 2), including MAC addresses and MAC frames, LAN cables, Optical
Ethernet, LAN switches and how VLANs are used to separate traffic.
The next chapter is all about IP (Layer 3): how routers implement the network, routing tables,
IP addresses, subnets, IPv4 address classes, static addresses, dynamic addresses and
DHCP; public addresses, private addresses and NAT; and an overview of IP version 6.
On a real-world telecom network, a traffic management system is required. This is imple-
mented with a technique called in general virtual circuits, and in particular with MPLS. The next
chapter in the book covers the fundamentals, briefly reviews legacy technologies X.25, Frame

9
Relay and ATM, then focuses on MPLS and how it is used to implement VPNs, Class of Ser-
vice, service integration and traffic aggregation.
The last main chapter is on the Internet: its origins, what an ISP is and how an ISP connects to
the rest of the Internet via transit and peering, the web, the Domain Name System, HTML and
HTTP, SSL, MIME and base-64 encoding for email, Internet telephony and Internet VPNs vs.
business-customer “MPLS service”.
The final chapter is a summary and wrap-up, covering technology deployment from the top
down, useful reference charts listing all of the technologies, standard network designs and
ending with a look at The Future.
Appendices
Appendix A All About T1
Appendix B Legacy Voice Services and Jargon
Appendix C Acronyms and Abbreviations
Telecommunications technology is in constant change – and technologies that used to be of
prime importance are not so important today, and so have been moved from the main part of
the book into appendices. The very last part of the book provides a comprehensive list decod-
ing mainstream acronyms and abbreviations used in telecom.

1.3 How to Use This Text

There are two ways to use this text, and we suggest that you try both.
The first way is to read it sequentially, from beginning to end, ensuring that you understand the
concepts in each section before moving on to the next. This is what we do in instructor-led
seminar and video course versions of this material.
A professional instructor leading you through this material would require four to five days of in-
tense, fairly fast-paced discussion; so it might be reasonable to think that you could read this
text in a week full-time or a month part-time.
However, no one expects anyone to absorb all of this information at once. Sometimes it takes
time to really understand various concepts. Also, you will often run into the technologies dis-
cussed in text this some time after you first read it, and will need a reference and refresher.
This leads to the second way to use this text: as a day-to-day reference handbook and glos-
sary. The Table of Contents has been constructed to allow pinpoint navigation to important top-
ics, and is hyperlinked and searchable in the ebook editions.
Thousands of people in the telecom, networking, government, military, educational, financial,
insurance, health, entertainment and other sectors, including people working for Cisco, Sun,
Nortel, Alcatel, Lucent, Microsoft, the CIA, FAA and IRS, all branches of the US Armed Forces,
AT&T, MCI, Sprint, Pac Bell, Ameritech, Qwest, Verizon, Global Crossing, Transamerica Insur-
ance, Oneida tableware, the San Francisco Giants and countless others have benefited from
this knowledge and our approach to transferring it.
We hope you will too.

1.4 The Three Answers

Telecom 101 is an easy course. There are only three possible answers to any question anyone
asks:
1) Money.
2) History.
3) It’s all pretty much the same thing.
1.4.1 Answer Number 1: Money
When someone asks, “Why was it designed that way?” or “What are we really discussing?”,
the answer usually boils down to money.
The reason for this is that it is very easy to measure results in the telecommunications busi-
ness. Not like in the psychology business, where it is very difficult to measure results; in the
telecommunications business, we can measure how many bits per second are transferred from

10
one place to another – and not surprisingly, there is a strong correlation between that and how
much you have to pay for it.
When we discuss technologies like POTS, T1, T3, SONET, ISDN, IP and others, one thing that
we will try to convey in this text is that these technologies are all essentially trade-offs between
cost and performance decided by a group of people sitting around a conference table one
Tuesday morning.
T1 was an example of this. T1 was designed in the late 1950s by a group of people at Bell
Labs in Holmdel, New Jersey. The requirement was to implement “digital” communications on
the existing copper wire transmission circuits, to increase the number of phone calls carried on
a set of physical wires.
Presumably, analyses were performed, lab experiments and field trials were undertaken, and it
was discovered that if the repeaters were spaced about one mile apart, it would be possible to
transmit about 1.5 Mb/s most of the time on most of the existing wires... so this design was
chosen, and 1.5 Mb/s became a standard line speed in the industry.
The repeaters could have been spaced two miles apart, and it would have been cheaper to im-
plement, but it also would have run more slowly. This was a trade-off between cost and perfor-
mance.
More recent examples include all of the different variations of Optical Ethernet, trading off bit
rate against cost of the optical transceivers and reach.
1.4.2 Answer Number 2: History
Those who do not learn history are doomed to repeat it.
You may be interested in learning about Voice over IP (VoIP). You probably should be inter-
ested in learning about it, as all telephone calls will be VoIP in the not-too-distant future.
If you want to understand Voice over IP, there are a number of technologies that come in to
play. One is Voice. Another is IP.
Voice is digitized in the phone in a VoIP system to be carried in IP packets.
Voice digitization involves three elements:
1) Sampling the value of the analog voltage coming out of the microphone at regular intervals,
2) Quantizing the range of possible values of the sample into fixed increments, and
3) Coding the resulting quantized value into binary.
Questions that arise are: how often do samples have to be taken, what are the quantization in-
crements and what algorithm is used to represent the quantized value in binary?
It turns out that it is necessary to take samples more than twice as often as the width of the fre-
quency band of the analog voltage coming out of the microphone.
All of a sudden we’re back to the summer of 1874 when Alexander Graham Bell made some
design decisions, and a bit later when loading coils were deployed on long-distance trunks,
which directly affect the method of digitization of voice for Voice over IP.
And the quantization and coding algorithms for VoIP are the same as those used for digital
voice channels called DS0s beginning in the 1950s.
We would claim that if you don’t understand this progression of one thing on top of another,
you will never really understand where we are today, and won’t be ready to understand where
we are going tomorrow, the all-IP Network..
In this text, and in our seminars and videos, we start at the beginning, progress in a logical or-
der, and finish at the end – to build structured knowledge so that you can understand how ev-
erything fits together.
1.4.3 Answer Number 3: It’s All Pretty Much the Same Thing
We could simplify “telecommunications” by claiming that there are two kinds of traffic or infor-
mation to be communicated: information that happens in continuous streams, and information
that happens in bursts.
Video is a good example of information that happens in continuous streams: when sending
video to someone, we are constantly transmitting picture information. E-mail is a good example
of information that happens in bursts: you send e-mail to someone, and then you don’t.

11
Telecommunication service providers like Verizon, AT&T, Bell Canada, TELUS and Sprint have
two basic kinds of services: they have services that allow their customers to transmit continu-
ous streams of information; and services that allow their customers to transmit bursts of infor-
mation.
If we look one level deeper in the network, to see how these services are offered, we find that
the way that a service provider offers to its customers the possibility of communicating informa-
tion in bursts is to take a circuit that actually communicates in continuous streams, attach
boxes called MPLS routers to each end, connect a large number of customers and let them
send information whenever there’s a free spot on the circuit that communicates all the time.
Now you know everything there is to know about telecommunications. Not. We’re not going to
simplify things quite that much, but…

☞ Once you achieve spiritual nirvana in telecommunications, you will realize that all of the
services you hear about like Internet access, telephone service, MPLS service, T1,
   ISDN and the rest are all really billing plans.
There is really only one kind of transmission network, built with fiber optic transmission sys-
tems using Optical Ethernet for new installations and an installed base of SONET from days
past.
There are a few methods of providing access to these fiber transmission systems: copper
wires, radio and fiber. There are many, many ways of billing you for using some of the capacity
on the transmission system in different ways and at different times.
It’s not like there is one “fiber backbone” for voice, a different one for data, a third one for tele-
vision and a fourth one for the Internet.
It all runs over the same pieces of glass with the same light flashing on and off 10,000,000,000
times per second to represent 1s and 0s.
Those are the answers.
The rest of this book is devoted to understanding the questions...

12
2
Fundamentals of Telephony
2.1 History of Telecommunications
Telecommunications began not with telephones, but with telegraphs. Telegraph systems were
the command and control systems for railways: used to communicate information about trains
from one end of the line to the other. Railways and their telegraph systems were deployed
across North America in the first half of the 1800s, and these were the first communication net-
works.
2.1.1 Invention of the Telephone
The telephone was invented by Alexander Graham Bell between 1874 and 1876, with most of
the work done on his father’s homestead near Brantford, Ontario in the Niagara region, and
some of the work done at his winter job at a school for deaf children in Boston.
It was in Brantford, in the summer of 1874, that Bell told his father how he proposed to build a
telephone, and there in the summer of 1875 that he drew up the patent application.
Bell demonstrated the telephone apparatus over short distances of wire with the words “Mr.
Watson, come here I want you!” on March 10, 1876 in Boston, and again at the Centennial Ex-
position in Philadelphia in June 1876… but communications across distance remained elusive.
Returning to Brantford in the summer of 1876, Bell refined his apparatus and made three suc-
cessful tests of communication across distance. This is generally considered to be the first
long-distance phone call.
The article The Human Voice Transmitted by Telegraph in the September 1876 issue of Scien-
tific American magazine outlined these experiments.

Figure 1. Alexander Graham Bell

13
In August 1876, Bell successfully demonstrated speech communication across wires of the
Dominion Telegraph Company between telegraph offices in Brantford and Mount Pleasant On-
tario; then between the Bell homestead and the telegraph office in Brantford, a distance of four
miles; and on August 10, 1876 over the eight-mile telegraph line between Brantford and Paris,
Ontario with the battery 58 miles away in Toronto.
If the Dominion Telegraph Company had been able to foresee that Bell’s company (that to this
day bears his name) would eventually put them out of business, they might not have been so
cooperative in hosting the trials!
Bell patented his device in 1876. Subsequently, it became a national sport to challenge his
patent in court. There were over 600 court challenges to Bell’s patent – every one unsuccess-
ful.
The many notes and diagrams produced in Brantford in the summers of 1874 and 1875, along
with his father’s diary were used to prove Bell’s claim to the invention of the telephone.

14
 Figure 2. Chronology of the invention of the telephone in Alexander Graham Bell’s handwriting.
Figure 2 is an image of one of the many memorabilia residing in the Bell Homestead Museum
at Tutelo Heights, Brantford, Ontario, Canada. It was composed by Bell following the opening
of the Bell Memorial at Brantford 24 October 1917. Credit to telecommunications.ca for the im-
age.
Claims are made for both Boston and Philadelphia as being the place where the telephone
was invented. In 2002, the US Congress passed a resolution claiming that Italian-American

15
Antonio Meucci had in fact invented the telephone and Bell had taken his lab notes and
patented the idea.
None of these claims are consistent with the serious, repeated, detailed investigations by hun-
dreds of people alive at the time of the events leading to court decisions what was invented
where and by whom (the telephone, in Brantford Ontario Canada, by Alexander Graham Bell)
during challenges of Bell’s patent.
2.1.2 Local Phone Companies
Telephone service began with connections within cities. A company would establish a Central
or Central Office (CO) downtown, and connect subscribers to their communication service to
the CO using pairs of copper wires to carry the electrical signals representing speech.
These subscribers would alert an operator in the CO that they wanted to establish a connec-
tion by cranking a handle that caused a bell to ring at the CO, and then telling the operator the
name of the person to which they wished to be connected. The operator would use a cord to
connect the two subscribers via a large patch board. This was the first kind of telephone
switch.
Since copper is a good, but not perfect, conductor of electricity – it has some resistance to the
flow of electrons through it – the copper wires could only be a certain maximum length before it
would not be possible to hear what the other person was saying.
Thus, local phone companies providing service in a radius of a few miles around a Central Of-
fice sprung up in major cities across the continent beginning in 1878. Inter-city long-distance
communications was not technically possible yet.
2.1.3 The Bell System
In the USA, these local phone companies were either part of, owned by or licensed by the Bell
Telephone Company which became the American Bell Telephone Company in 1880.
Its Chief Operating Officer and later president, Theodore Vail, began creating the Bell System,
to be composed of regional companies offering local service, a long distance company and a
manufacturing arm providing equipment.

Figure 3. Building telecommunications networks - planting poles and stringing wires across the continent - is big
business, like railroads. In the photo, making the first call from New York to San Francisco, left to right is

16
 Theodore Vail, the man who got it done, financiers William Rockefeller (seated) and J. P. Morgan Jr. (standing)
along with network architects Samuel Trowbridge and Welles Bosworth.
The American Telephone and Telegraph Company (AT&T) was incorporated in March, 1885 as
a wholly-owned subsidiary of American Bell, with the initial business plan of providing long-dis-
tance service for the Bell System: connecting the local companies.
Building out from New York, its initial goal Chicago was reached in 1892, and San Francisco in
1915. AT&T continued as the “long-distance company” until Dec. 30, 1899, when it changed its
business model to be vertically integrated: local and long distance, by acquiring the assets of
the American Bell Telephone Company and becoming the parent company of the Bell System.
2.1.4 US Regulation and Competition
Until Bell’s patent expired in 1894, only licensees of American Bell could legally operate tele-
phone systems in the United States.
Between 1894 and 1904, over six thousand telephone companies, called independents (that
is, not part of the Bell System) went into business, and the number of telephones increased
from some 250,000 to over 3,000,000… but in many cases, there was no interconnection be-
tween the independents.
For much of its history, AT&T and the Bell System functioned as a regulated monopoly. The
idea was that the telephone system, by the nature of its technology, would operate most effi-
ciently as a monopoly providing universal service.
However, business practices prompted the United States government to sue AT&T three times
under antitrust laws: 1913, 1949 and 1974.
The 1913 suit resulted in the Kingsbury Commitment, in which among other things AT&T
agreed to connect independents to its long-distance network.
Several court decisions forced the opening of AT&T’s network from a technical point of view.
These included the Carterphone decision, which allowed customers to use their own terminal
equipment on the Bell System, and MCI’s successful suit that allowed MCI to connect to
AT&T’s network to carry long-distance calls.
A 1974 suit by the Justice Department was settled when AT&T agreed to divest itself of local
operating companies in January 1984 in exchange for loosening of regulation.
The ownership of AT&T’s local operations was transferred to one of seven holding companies,
known as the Baby Bells: US West, Pac Bell, Southwestern Bell, Bell South, Bell Atlantic,
NYNEX and Ameritech.
The remaining operations were the long lines, now called AT&T Corp., which was then forced
to compete with other companies for carrying phone calls and data services long distance.

Figure 4. The ownership of AT&T’s local operations was transferred to seven holding companies, known as the
Baby Bells.
In 1996, the federal government’s Telecommunications Act removed many of the remaining
obstacles at the federal level to wide-open competition for both local and long-distance

17
telecommunications... which meant in practice that local and long-distance operations could
merge back together.
Significant obstacles remained at the state Public Utility Commission level, which were slowly
overcome.
2.1.5 Consolidation
1996 saw the beginning of consolidation of the Baby Bells with the purchase of NYNEX by Bell
Atlantic.
SBC Communications Inc., owner of Southwestern Bell, purchased Pac Bell in 1997, SNET in
1998 and Ameritech in 1999.
In 2000, Bell Atlantic merged with GTE, owner of many independents, and baptized the result
Verizon, a focus-group-tested name from a combination of the words veritas (truth) and hori-
zon. They lead to the true horizon.
US West was purchased by Qwest, later merging with Century Tel to form CenturyLink.
Many other independent companies continued to own and/or operate local networks and re-
gional fiber backbones.
Once the LECs could be IXCs, local and long-distance operations were merged back together
by the LECs purchasing the IXCs.
In 2005, Verizon purchased MCI.
To get the valuable brand name “AT&T”, SBC implemented a reverse takeover of AT&T, pur-
chasing AT&T Corp. and changing the name of the resulting company to “AT&T”.

Figure 5. Consolidation: Verizon and AT&T

In 2006, AT&T acquired Bell South. This reconstituted most of the Bell System in two pieces:
AT&T and Verizon. The remainder - US West - ended up owned along with many indepen-
dents by Century Link.
Each company provides services in the other’s territory via fiber, collocations and wireless.
2.1.6 Other Carriers
Companies with a coaxial entry cable to the residence were historically called Community An-
tenna Television (CATV) or cable companies. Since coaxial cable supports the use of a much
broader frequency band than twisted pair, these companies are also called broadband carriers.
Cable companies offer Internet access and telephone service for residences and business us-
ing modems operating at 500 Mb/s or more on their last mile, in parallel with the cable televi-
sion programming.
In the USA, cable companies initially captured a majority share of residential Internet access.
These companies also offer fiber business services, and have progressed to full telecom ser-
vice providers and carriers.
Cellular mobile telephone service launched in Chicago in 1983.

18
VoIP and Internet telephony began commercially in 2002 and 2003, the beginning of the end
for the time-and-distance pricing model for voice communications and pure long-distance carri-
ers.
These topics are covered in detail in subsequent chapters.
2.1.7 Canadian Telegraph Companies
Telecommunications in Canada began with telegraph companies. By 1847, the Montreal Tele-
graph Company was established and providing service in the Quebec City - Windsor corridor,
with a link to Western Union in Detroit.
Telegraphs were instrumental in the construction and operation of railways. In 1886, Canadian
Pacific Railways Telegraphs came online as a competitor.
After World War I, most of Canada’s smaller railways were in serious financial difficulty. A
bailout by the federal government saw the merger of these railways into the Canadian National
Railway, and their telegraph lines became the CN Telegraph Company.

Figure 6. CN and CP railway telegraph systems were the basis for Allstream.
During the period from 1932 to 1964, these two railway telegraph companies both competed
and jointly offered services. In 1932 they provided national network services for the Canadian
Radio Broadcast Commission. In 1939, national weather service; after the Second World War
private wire services; in 1956 the first telex services in North America, and in 1964 a cross-
Canada microwave radio transmission network.
These two railway telegraph companies were fused to form CNCP Telecommunications in
1980. In 1988 Canadian Pacific bought out CN, sold 40% of the company to Rogers Communi-
cations Inc. and renamed the company Unitel.
Decision 92-12 by the Canadian Radio-Television Telecommunications Commission, the fed-
eral regulatory agency, allowed Unitel to provide competitive long-distance services. In 1993,
20% of Unitel was sold to AT&T Corporation of the United States.
Even though the regulators gave Unitel a discount on payments to the telephone companies
for using their access wires for the first five years, a number of factors including the necessity
to build and maintain a transmission network 7,200 kilometers (4,500 miles) long to connect
Victoria to St. John’s, plus the costs of the POPs and interconnection in toll centers, customer
care and billing systems, the people to run it all and the natural competitive practices by incum-
bent carriers made Unitel unprofitable.
After several years, Rogers Communications Inc. abandoned its interests in Unitel and through
a Canadian Creditors’ Protection Act bankruptcy-like proceeding, Unitel’s ownership was re-
duced to AT&T Corporation and three Canadian banks. The reorganized company became
AT&T Canada Long Distance Services Company.
Subsequently, AT&T Corporation of the USA bought out the rest of AT&T Canada through a
holding company. However, the geographic and market factors that made Unitel unprofitable
had not changed, and AT&T Canada continued to lose money. AT&T Canada sold its residen-
tial long-distance operations to Primus Telecommunications, and in 2003 went through a sec-
ond bankruptcy-like reorganization.
The resulting company was re-baptized Allstream, providing corporate and data services. All-
stream was subsequently purchased then sold by MTS of Manitoba. Allstream today is the
main facilities-based Inter-Exchange carrier competing with the telephone companies.

19
2.1.8 Canadian Telephone Companies
In 1880, the Bell Telephone Company of Canada was established in Montreal, and other com-
panies providing local service in other cities sprung up across the country. At the time, long-
distance inter-city communications was not technically possible, so these companies each pro-
vided telephone service in a local area.
In 1921, the Telephone Association of Canada was formed to promote the construction of a na-
tional network. In 1931, the Trans-Canada Telephone System, again an association of “local”
telephone companies, began the development of a national network. In 1958, a 158-station
cross-Canada microwave network was completed - the world’s longest at the time. In 1983, the
association changed its name to Telecom Canada and in 1992 to Stentor. The facilities that
made up the national network were owned and operated by member companies such as Bell
Canada and BC Tel. The Stentor Alliance was terminated effective December 31, 1999.

Figure 7. The Canadian business model mostly evolved to the same as the US: large companies owning local op-
erations in many areas, CLECs in many others and long-distance transmission facilities.
The Canadian business model mostly evolved to large holding companies owning local opera-
tions in many areas plus long-distance transmission facilities, and operating agreements be-
tween the companies for interconnect.
BCTel, AGT, Ed Tel and Quebec Telephone merged to form TELUS. Bell Canada and the four
telephone companies in the maritime provinces were reorganized as Bell Canada in metropoli-
tan areas of Ontario and Quebec and Bell Aliant elsewhere, both majority owned by BCE.
These companies both provide IPTV on VDSL or fiber, and are both expanding to provide na-
tional service via fiber, collocations and wireless.
Rogers Communications Inc. is now another major player in the telephone business. Rogers
started in the cable TV business, with the creation of Rogers Cable by Ted Rogers in 1967 in
Ontario. In 1979, Rogers acquired Canadian Cablesystems and became the largest cable
company in Canada.
In the 1980s, Rogers entered the cellular market under the Cantel brand name and later ac-
quired Microcell and its Fido brand.
After its first venture in telecom with Unitel ended, in 2004 Rogers re-entered the business, ac-
quiring Sprint Canada and Callnet, operating as a facilities-based Inter-Exchange Carrier, a re-
seller and as a Competitive Local Exchange Carrier (CLEC) with landlines in Richmond BC.
2.1.9 The Rest of the World
In many European countries, the national government operated a Post, Telephone and Tele-
graph (PTT) company that was a government-owned monopoly.
Competition for both local and long-distance voice and data communications has been intro-
duced at different rates in different countries.
In all cases, we see the progression of telecommunications service characterized by:

20
 • Monolithic organizations holding a monopoly and the mandate to provide universal service
under government ownership, control or regulation,
• Then the breakup of the monopoly to introduce competition in inter-city and long distance
communications,
• Followed by competition in providing local services.
In many parts of the world, particularly in developing areas without usable existing infrastruc-
ture, mobile wireless is a more popular method of accessing the telephone network - and inter-
net - since it is far simpler and less expensive to set up cellular radio base stations than it is to
wire or fiber neighborhoods.
In addition to services for individuals, providing high-capacity and high-availability voice and
data services for business customers like banks, distribution centers and government is a vi-
able business everywhere in the world.

2.2 The Public Switched Telephone Network

With a history lesson in place, we begin discussion of telecommunications technologies with
the basics of telephony, the Public Switched Telephone Network (PSTN) and Plain Ordinary
Telephone Service (POTS)... still in wide use, and necessary for understanding newer tech-
nologies.

Figure 8. Basic model of the Public Switched Telephone Network (PSTN).

2.2.1 Basic Model of the PSTN

We begin with a basic model and will build on it in subsequent discussions.

21
At the top of Figure 8 is a telephone and a telephone switch. The telephone is located in a
building called a Customer Premise, and the telephone switch is located in a building called a
Central Office or CO. One could refer to the telephone as Customer Premise Equipment or
CPE.
2.2.2 Loops
The telephone is connected to the telephone switch with two copper wires, often called a local
loop or a subscriber loop, or simply a loop. This is a dedicated access circuit from the cus-
tomer premise to the network.
There is usually the same arrangement at the other end, with the far-end telephone in a differ-
ent customer premise and the far-end telephone switch usually in a different central office.
Copper is a good conductor of electricity - but not perfect; it has some resistance to the flow of
electricity through it. Because of this, the signals on the loop diminish in intensity or attenuate
with distance.
The maximum resistance allowed is usually 1300 ohms, which is reached in 18,000 feet (3
miles or 5 kilometers) on standard-thickness 26-gauge cable, but could be as long as 14 miles
or 22 kilometers on thicker 19-gauge cable.
This maximum loop length of 3 miles or 5 kilometers defined the traditional serving area
around a Central Office, about 27 square miles or 75 km2.
2.2.3 Trunks and Circuit Switching
Telephone switches are connected with trunks. While subscriber loops are dedicated access
circuits, trunks are shared connections between COs.
To establish a connection between one customer premise and another, the calling party signals
the network address (the telephone number) of the called party over their loop to the network,
or more specifically, to their CO switch.
The switch makes a routing decision for the phone call then implements it by seizing an un-
used trunk circuit going in the correct direction and connecting the loop to that trunk.
The called party network address is signaled to the far-end switch, which connects the trunk to
the correct far-end loop. When the far-end customer picks up the phone, an end-to-end con-
nection is in place and maintained for the duration of the phone call.
When one end or the other hangs up, the trunk is released for someone else to use for con-
nections between those COs. This method for sharing trunks is circuit-switching, called dial-up
when telephones had rotary dials.
2.2.4 Remotes
Figure 8 was a model for the telephone network up to the end of the Second World War. With
the subsequent suburban sprawl, it was not cost-effective to build COs every five miles or eight
kilometers.
New subdivisions began to be served from remote switches or more simply, remotes, which
are low-capacity switches in small above-ground buildings or underground controlled environ-
ment vaults.
As illustrated in Figure 10, the remote provides telephone service on copper loops in the subdi-
vision and is connected back to the nearest big CO with a fiber backhaul.
The electronics and optics in the remote connect the fiber to the copper wires, or perhaps
more precisely, take information received over the fiber and transmit it to the residences over
copper loops, and vice-versa.
2.2.5 DSL and DSLAMs in the Outside Plant
In the 1990s, modem technology called Digital Subscriber Line (DSL) began to be deployed,
using the existing copper loop to connect a modem at the customer to a modem in the CO for
high-speed Internet access coexisting with telephone service on the loop.
To increase the achievable bit rate, the distance between the modems was shortened by mov-
ing the network side modem, contained in a device called a Digital Subscriber Line Access
Multiplexer (DSLAM) into the neighborhood.

22
The equipment and wiring in neighborhoods, along with transmission systems carrying trunks
is collectively referred to as the outside plant.
This remote DSLAM is usually located in a small enclosure bolted on to the side of a larger en-
closure called an Outside Plant Interface (OPI) or Serving Area Concept (SAC) box.

Figure 9. OPI with DSLAM bolted on the side

The OPI or SAC box is a wiring connection point in the neighborhood where wires in a feeder
cable from the CO are connected to wires in distribution cables running down streets.
It is also a location where a network-side DSL modem can be jumpered on to the existing cop-
per loop.
Fiber-optic and power cables connect the remote DSLAM back to the CO and from there to the
Internet.
Generically, the remote switch and remote DSLAM are fiber terminals: where the fiber ends.
2.2.6 Brownfields: DSL on Copper to the Premise
A brownfield is a neighborhood where copper wires were previously deployed.
In this case, a hundred or two customers share a fiber backhaul to the network, getting to the
fiber with DSL modems over existing copper wires for the last few hundred meters.

23
 Figure 10. Remote Switches and DSLAMs
This is increasingly being seen as a temporary measure while waiting to pull fiber to the home
in an older neighborhood.
Eventually, most customers will have their own fiber terminal... connecting to copper wires and
WiFi inside the house.
2.2.7 Greenfields: GPONs on Fiber to the Premise
In greenfields, i.e. newly-constructed neighborhoods and multi-tenant buildings, where the ca-
bling is the initial installation, fiber to the premise is routinely installed.
Gigabit Passive Optical Network (GPON) technology is usually employed, where typically 32
customers time-share a fiber connection to the network. One fiber backhaul towards the net-
work is connected in a Central Splitting Point via lenses and mirrors to 32 fibers leading to cus-
tomer premises.
Only one customer can transmit at a time, so the uplink is shared in a round-robin fashion, and
each user is reserved a fixed amount of capacity on the uplink whether they are using it or not.
In the Transmission Systems chapter, this is called channelizing or channelized multiplexing.
2.2.8 Active Ethernet to the Premise
Active Ethernet may also be deployed. In this case, the customer’s fiber terminates on its own
port on an Ethernet switch, located either in the neighborhood or at a wire center. Customers
have the possibility of transmitting upstream any time they like instead of in time slots. In the
Transmission Systems chapter, this is called bandwidth on demand or statistical multiplexing.
Statistical multiplexing is more efficient than channelizing and gives users higher upload
speeds for the same capacity backhaul... but compared to the PON of Section 2.2.7 requires

24
31 more network-side fiber transceivers, so is more expensive to install and maintain.
Active Ethernet is routinely implemented for business customers.
2.2.9 Why the Loop Still Matters
It is important to note that even though today there may be digital switching and digital trans-
mission, for traditional telephone service in established areas, the access circuit between the
customer and the network - the local loop, the “last mile” - employs analog technology dating
back to the late 1800s.
Even if residential telephone service becomes Voice over IP over fiber, or VoIP over cable mo-
dem, analog technology of the local loop from 1880 is still used!
Telephone service from the cable TV company means traditional analog telephony on the in-
side wiring, plugged into a converter that carries it as Voice over IP over cable modem outside
of the house.
DSL service, broadband from the telephone company, is delivered on the existing local loop…
by modems in the remote in the previous image.
The 64 kb/s DS0 rate for channelized digital transmission systems, covered in a subsequent
lesson, is based on the frequency band supported on the traditional analog local loop.
For these reasons, an understanding of the characteristics and limitations of the local loop is
essential knowledge.

2.3 Analog
The technique for representing information on an ordinary local loop is called analog. This term
is often thrown about with little regard for its actual meaning, so we will spend a bit of time un-
derstanding what is meant by “analog”.
2.3.1 Analog Signals
The term analog comes from the use of a microphone in the handset of the telephone. A sim-
ple type of microphone, such as those in the handset of a telephone, has a plastic housing, a
paper diaphragm and carbon particles between the two.

Figure 11. The voltage on the wires is an analog of the strength of the sound pressure waves coming out of the
speaker’s mouth.
When someone speaks, sound pressure waves come out of their mouth. The person using the
telephone holds the microphone in front of their mouth, so that the sound pressure waves push
on the paper diaphragm.
This has the effect of compressing the carbon particles in the microphone, changing its electri-
cal characteristics… the microphone’s capacitance, to be precise.
The fact that the electrical characteristics of the microphone change as the sound pressure
waves hit it can be used to make a voltage on the telephone wires change.

☞ This voltage is a direct representation or analog of the strength of the sound pressure
waves: as a pressure wave pushes on the microphone, the voltage increases; as it
   stops pushing on the microphone, the voltage reverts to where it was.
This is all that is meant by “analog”: representation. The voltage on the wires is an analog of
the strength of the sound pressure waves coming out of the speaker’s mouth. This voltage

25
could also be called an analog signal.
At the other end, a speaker is used to create sound pressure waves based on the received
analog signal. A speaker is an electromagnet glued onto a paper diaphragm.
The voltage that is the analog is applied to the electromagnet, causing the paper diaphragm to
move back and forth, creating sound pressure waves, which are hopefully a faithful reproduc-
tion of the original sound pressure waves coming from the speaker’s mouth.
2.3.2 Analog Circuits
The voltage carried on the loop is an analog signal. People then stretch this terminology and
refer to the two copper wires that form the loop as an “analog circuit”, which is not very accu-
rate.
The only thing analog in this story is the method for representing speech on copper wires using
electricity.
It would be more precise to call the loop “two copper wires that were designed to carry a volt-
age that is an analog of the strength of the sound pressure waves coming out of the speaker’s
face”.
It is possible to use digital techniques on the same copper wires.

2.4 Capacity Restrictions

Once the method of representing speech on copper wires using an analog is understood, the
next question would be: how accurate does this “analog” have to be? What degree of fidelity is
required in representing the sound pressure waves coming out of the speaker’s throat? How
faithful do the sound pressure waves reproduced at the far end have to be?
With a bit of reflection, we realize that the fundamental requirement of the telephone system is
to communicate information between people.
The first design decision made was that speech and hearing would be used as the method of
communications.
Once this highest-level design decision was made, it is necessary to understand what speech
is before the characteristics of the local loop could be designed.
2.4.1 What is Speech?
Speech is a form of sound.
A textbook would define “sound” as pressure waves in the air, meaning that at a given point,
the air pressure rapidly increases and decreases in a cyclical fashion in time.
An increase in air pressure causes the molecules to be pushed closer together or compressed;
a decrease in air pressure allows the molecules to spread out or be rarefied. Sound is the
compression and rarefaction of air molecules. Of course, water molecules, plaster, glass and
other substances can also support this phenomenon to some extent.

26
Figure 12. A representation of a speech analog generated by a stereo microphone. The upper part of the diagram
shows on the vertical axis the sound pressure, interpreted by your brain as volume; and the lower part of the dia-
gram shows on the vertical axis the frequency of the compression-rarefaction cycle, interpreted by your brain as
pitch. The horizontal axis is time.
Sound pressure waves coming out of the speaker’s face vibrate rapidly, that is, go through cy-
cles of compression and rarefaction. If this vibration occurs between 20 and 20,000 cycles per
second, the sound pressure waves are said to be audible by the human ear.
2.4.2 Do Trees Falling in the Forest Make a Sound?
Understanding that sound is cycles of varying air pressure, and knowing that if this occurs be-
tween 20 and 20,000 times per second it is audible still does not tell us how faithful the voltage
analog must be, and how faithfully the sound must be reproduced at the far end.

27
 Figure 13. Reconstructing Sounds vs. Reconstructing Thoughts
An age-old question is: if a tree falls in the forest, and no one is there to hear it, does it cause a
sound?
That depends whether you believe sound is pressure waves: air molecules being compressed
and rarefied as per the preceding textbook definition; or if you believe sound is the sensation
one gets in one’s brain when one hears the sound pressure waves.
The two choices in designing the telephone system are then to either:
a) Reproduce sound pressure waves coming out of the speaker in the far-end telephone ex-
actly as they entered the microphone in the near-end telephone; or
b) Reproduce the sensations in the listener’s brain the same as they would experience were
they speaking directly to the other person.
The difference between these two ideas is that the brain is a hugely complicated processing in-
strument, and it is possible to play different stimuli at it and get the same response.
Each choice has a dramatically different implication for the cost of implementing the system.
2.4.3 The Voiceband
What answer did Alexander Graham Bell choose? Answer (b).
Based on testing human beings’ ears, throats and brains, combined with a technical innovation
that extended the achievable transmission range, led us to transmit the information in the fre-
quency range between 300 and 3300 Hz.
Hertz (Hz) is the unit for frequency, or changes per second. The range or band of frequencies
from 300 to 3300 Hz is called the voiceband.

28
Figure 14 is an idealized representation of the voiceband, with frequency on the horizontal axis
and amplitude or intensity on the vertical axis.

Figure 14. The Voiceband

It shows that any electricity vibrating at least 300 times per second and less often than 3300
times per second will be passed (indicated by a 1).
Any electricity vibrating less often than 300 times per second will be suppressed (indicated by
a 0). Similarly, any electricity vibrating more than 3300 times per second will be suppressed.
Only electricity vibrating within the band 300-3300 Hz will be transmitted.
The suppression of energy outside this frequency band 300-3300 Hz is implemented with sim-
ple electrical circuits called filters. There is a filter in the telephone and a filter in the switch in
the CO.
2.4.4 Bandwidth
For our purposes, bandwidth means capacity. In the analog world, capacity is measured liter-
ally by the width of the frequency band supported on the physical medium by the service you
are paying for.
For the voiceband, the bandwidth is the interval between 3300 and 300 Hz, which is 3000 Hz
or 3 kHz for short.
This 3 kHz bandwidth is a standard capacity provided for ordinary telephone service by all tele-
phone companies.
2.4.5 Why Does the Voiceband Stop at 3300 Hz?
Why 3300 Hz? It would be technically possible to provide a greater bandwidth for telephone
service, resulting in crisper, clearer sound.
The two wires that make up the loop are capable of supporting electricity vibrating more often
than 3300 times per second – in fact, DSL technologies require electricity vibrating at frequen-
cies measured in the millions of times per second.
The users’ ears and brains are capable of detecting sound pressure waves vibrating more of-
ten than 3300 times per second – the human hearing range is traditionally thought to extend
up to 20,000 Hz.
So why would the capacity a user is allowed to employ be purposely limited to 3 kHz, even
though the wires are capable of more than that, and the users are capable of more than that?
The answer is, as usual, money.
This narrow frequency band was chosen based on studying people’s ears, throats and brains,
to determine the minimum capacity necessary to meet the requirements.
Returning to the question of trees falling in the forest: the sound pressure waves at the far end
are not reproduced exactly as they were at the near end; in fact, they are quite muffled and
distorted, missing most of the higher frequencies.

29
 Figure 15. The sound is reproduced just well enough so that the listener can recognize the speaker and under-
stand what the speaker is saying, thus meeting the requirement to communicate information using speech and
hearing
The sound is reproduced just well enough so that the listener can recognize the speaker and
understand what the speaker is saying, thus meeting the requirement to communicate informa-
tion using speech and hearing.
We are interested in transmitting the minimum required to meet that objective since there is a
direct relationship between the capacity a user can employ on the access circuit and the cost
of transmitting the information long-distance.
2.4.6 Problems With Voiceband Restrictions
It turns out that the voiceband is not quite enough bandwidth to be able to understand every-
thing the speaker is saying!
In particular, it is difficult to tell the difference between “S” and “F” over a telephone. This is be-
cause the frequency of sound pressure wave that distinguishes “S” from “F” is above 3300
Hz… which is not transmitted over the phone system.
Phonetic alphabets , such as one used by military forces, use words to communicate each let-
ter, for example, “S as in Sierra” and “F as in Foxtrot”

30
 Figure 16. Standard and Alternate Phonetic Alphabets
.
One could also say things like “S as in Sea”, “C as in Cue”, “A as in Are” and “E as in Eye” to
liven things up. If that doesn’t get the listener confused, there’s always “E as in Ewe” and “Y as
in You”.

2.5 Problems with Analog Transmission

2.5.1 Attenuation and Amplifiers
Aside from bandwidth restrictions, the chief impediments to transmission over analog circuits
are noise and attenuation, both of which affect the capacity of the circuit to transport informa-
tion.
Attenuation on wired circuits is caused by the physical characteristics of the copper wires that
make up the circuit. Since copper is not a perfect conductor, and has some resistance, some
of the transmitted energy is turned into heat and the signal level decreases or attenuates with
distance from the transmitter. On wireless systems, attenuation happens due to the spreading
effect of the waves.

31
 Figure 17. Attenuation and Amplifiers
In both cases, too far away from the transmitter, the signal will “disappear into the noise”, that
is, the signal level will become less than the noise level on the line, and it would be impossible
to faithfully reconstruct the speech.
Before this happens, the signal must be amplified at regular distance intervals to boost it back
up. The device that performs this function is called an amplifier. It multiplies or boosts the sig-
nal on its input by a certain factor.
The problem is noise that is added to the signal during transmission, before the signal reaches
the amplifier. The noise and signal are combined; when the signal is boosted up by an ampli-
fier, so is the noise.
This is the fundamental problem with analog transmission: the transmission system both atten-
uates the signal and adds noise to it; then to boost up the signal, the amplifier also boosts up
the noise.
An analog signal becomes noisier and noisier as it passes through each amplifier along a
transmission system.
2.5.2 Electro-Magnetic Interference
Noise comes in many forms. On copper-wire access circuits, the most problematic is caused
by radio waves, or more precisely, Electro-Magnetic Interference (EMI)
Copper wires act like antennas. When a radio wave impinges on a wire, it induces electricity
that adds to the desirable signal being carried on the wire. The source of such interfering addi-
tive noise includes television broadcasts, microwave ovens, computer chips, cellular radio
base stations, wireless LANs and other sources.
Interestingly, glass – fiber optics – does not act like an antenna and does not pick up this kind
of interference.
2.5.3 Crosstalk
Crosstalk is a specific type of EMI, the transference of energy from one wire to another via
electro-magnetic radiation. Usually, this happens when two circuits are in the same cable: a
signal placed on one circuit will create a magnetic field that passes through the other circuit
and induces current on it.
This is why you can hear other people talking on your regular wired telephone sometimes. The
annoyance factor decreases with comprehensibility.
2.5.4 Impulse Noise
Impulse noise appears like spikes of voltage on a circuit. This is caused by lightning striking
the wires, by the spark that jumps across the contacts of a switch just before it closes, and
when the brushes on an electric motor pass the unpowered portion of its armature.
In days past, this could be seen as white dots on a television screen when a drill or vacuum
cleaner is operated in close proximity.
Impulse noise is not additive noise – it hard-limits the signal to maximum, and causes a burst
of errors to happen.
The most popular way to deal with impulse noise is to format data into frames with error detec-
tion, and re-transmit a frame if there was a spike on the line that caused a burst of errors to
happen.

32
This is covered in detail in a subsequent chapter.

2.6 Plain Ordinary Telephone Service (POTS)

Basic telephone service is called POTS: Plain Ordinary Telephone Service in the business,
and sometimes referred to as dial tone or individual residence line service.

Figure 18. Plain Ordinary Telephone Service

As illustrated in Figure 18, this service consists of a rotary dial telephone connected to a line
card in a telephone switch by a two-wire copper loop.
The main components of a telephone are the microphone, speaker, hybrid converter, hook
switch, dial switch, ringer and protection.
2.6.1 Tip and Ring
The two wires that make up the loop are sometimes referred to as tip and ring. These names
come from the first kind of telephone switch, which was a board holding female jacks, to which
loops and trunks were connected.
To connect a loop to a trunk, an operator would plug one end of a patch cord into the jack for
the loop and the other end into the jack for the trunk.
The connectors on the patch cord were designed so that one of the wires was attached to the
metal tip of a male connector and the other wire was attached to a metal ring below the tip.
Plugging the connector into the jack connected the two wires that made up the loop at the
same time.
2.6.2 Twisted Pair
A problem with using two-wire circuits is that they act like antennas. Loop antennas. The
amount of energy picked up by a loop antenna is proportional to its area… and here, we are
connecting a 3 mi. / 5 km diameter loop to the telephone (!).

33
To minimize the amount of noise picked up on the wires, they are covered in plastic, then
twisted together.
Since there is plastic on the wires, they still act electrically like one big current loop, but from
an antenna point of view they appear as a series of small loops. The small loops have a
smaller area than the big loop, and so this minimizes the antenna effect of the wires.
Since there are two wires twisted together, we call them twisted pair. Twisted pair is used for
mostly all cabling, including telephone wires on poles, inside wiring and data cabling – LAN ca-
bles have four twisted pairs.
2.6.3 Line Card
The twisted-pair loop is terminated on the network side on a line card.
A line card is traditionally a small fiberglass board populated with a number of components, in-
tegrated circuits and connectors. This line card is plugged into a slot in a drawer, in a shelf, in a
rack, that is part of a traditional telephone switch.
In newer applications, the line card might be part of a gateway that converts between POTS
and Voice over IP, discussed in detail in a subsequent section.
The line card implements quite a number of functions, sometimes referred to by the acronym
BORSCHT: battery, overvoltage protection, ringing, supervision, codec, hybrid and testing.
2.6.4 Microphone and Speaker
The microphone is a kind of transducer, creating a voltage based on sound pressure waves.
The value of this voltage is a representation or analog of the strength of the sound pressure
waves coming out of the speaker’s throat.
The voltage is carried from the telephone over the loop to the line card at the near end, where
it is digitized by the codec and transported by the telephone network to be reproduced by the
far-end line card and carried by the far-end loop to the far-end telephone.
The speaker, as might be imagined, works in a manner opposite to the microphone: it uses re-
ceived voltage to create sound pressure waves that are directed into the user’s ear.
2.6.5 Balanced Signaling
Voltage is always measured as a difference between the voltage on one object and the voltage
on another.
In many cases, one object is the earth and the other is a wire, so the voltage measurement is
with respect to the ground.
This is not the case with a telephone loop. On a telephone loop, the voltage is measured be-
tween the two wires that are the loop, not between the earth and the wires.
Balanced signaling is used. This means that if the voltage on one wire with respect to ground
is some positive value, the voltage on the other wire with respect to ground will be the same
value, but negative.
Since added noise will be the same on the two wires, when measuring the voltage between the
two wires at the receiver, the signal is doubled and the noise is canceled.
2.6.6 Two-Way Simultaneous
The two wires that are the loop are used to transmit information in both directions at the same
time.
Both the telephone and the line card cause voltage analogs of sound to be placed across the
two wires of the loop. The voltages from the devices at each end are added together.
2.6.7 Hybrid Transformer
The voltage for each direction is separated by a device inside the telephone called the hybrid,
which has the two-wire loop on one side and two circuits on the other side, one for the speaker
and the other for the microphone.
A similar function is implemented on the line card, connecting the loop to the transmit and re-
ceive pins of the codec.
2.6.8 Battery

34
In addition to the voltage analog of sound, which might be thought of as an AC (or varying) sig-
nal, the line card also places a DC (or steady) voltage across the two wires that make up the
loop.
This voltage is called battery in the business, and is used to power the telephone. It is nomi-
nally -48 volts, measured from ring to tip.
2.6.9 Lightning Protection
Another item on the diagram is the protection circuit across the loop. This is to protect the tele-
phone user from being electrocuted, if lightning hits the loop or a high-voltage electrical trans-
mission wire touches the loop.
There are in fact three levels of protection: a fuse on the line card will blow if too much current
passes through it, circuitry on the demarc or demarcation point where the telephone com-
pany’s wires connect to the customer’s wires that will fall to ground if the voltage is too high,
and third, inside the telephone a circuit that will short-circuit the loop if the voltage across the
loop is too high.
2.6.10 Supervision
Two other components of the telephone, the hook switch and ringer, are used for supervision.
Supervision means regardless of to whom you wish to speak, and regardless of what you are
going to say to them, you must indicate to the other end of your loop that you want to start do-
ing all of this.
The hook switch in the telephone is normally open, so the two wires that make up the loop are
not connected, and no electricity or current is flowing around the loop.
To initiate communications, the user picks up the handset (goes off-hook), which causes the
hook switch to close, connecting the two wires together, which then allows the line voltage to
push current around in a… loop. This is why they are called loops.
This type of supervision is called loop start signaling: the two wires are connected, forming a
loop and allows current to flow in a loop.
The line card on the telephone switch detects this current and acknowledges with a dial tone
(assuming you have paid your bill).
There are variations on this theme used in other applications such as PBX switches, such as
ground start signaling, where one of the wires is plugged into the ground, so the current flows
along one wire then back through the ground; reverse battery signaling where the positive and
negative line voltage is reversed; and wink start signaling where that is done for a short interval
then returned to normal value.
For supervision in the other direction, the switch indicates it wants to initiate communications
by having the line card place a ringing signal on the loop.
This is yet another voltage, one that varies 20 times per second. It is applied to the line for two
seconds then not for four seconds in a repeating cycle.
When your phone rings, it is on-hook. This means that the hook-switch is open, so the current
pushed by this ringing signal flows through the ringer as shown in Figure 18 – originally two
brass bells with a clapper between that would move back and forth 20 times per second for
two seconds then rest for four seconds. The user acknowledges by going off-hook.
The line voltages are nominally as follows:
• On Hook: -48 Volts DC
• Ringing: -48 Volts DC, plus 100 Volts RMS @ 20 Hz
• Off-Hook: -7 to -12 Volts DC.
2.6.11 Call Progress Tones
Dial tone is a type of call progress tone. There are many others, such as busy, fast busy sig-
nals, ringback, congestion, sounder and howler tones. These are generated by the switch to
inform the user of different conditions.
Some of the call progress tones, such as dial tone and fast busy are generated by the near-
end switch. Busy signals are generated by the far-end switch.

35
2.7 Network Addresses: Telephone Numbers
Once your request to communicate is acknowledged with a dial tone, it is necessary to inform
the network where the call is to be connected.
In general, network address is the name given to the piece of information used to identify the
final destination of a connection across a network. For POTS, network addresses are of course
called telephone numbers.
2.7.1 Dialing Plan
The length of the telephone number, that is, the number of digits that have to be dialed, and
how the addresses are assigned to subscribers is called a numbering plan or dialing plan.
In days past, the North American Numbering Plan for telephone numbers was composed of
digits with specific purposes. Restrictions were placed on the values of various digits so that
dumb mechanical and analog switches could distinguish between them.
Addresses were originally of the form NBN-NNX-XXXX, where
• N is any number from 2 – 9
• B is any number from 0 – 1
• X is any number, and
• The first three digits were the area code,
• The next two were the CO code,
• The next one identified the switch in that CO,
• The last four identified the physical pair of wires.
The user had to dial anywhere from five to ten of these digits, sometimes prefaced with a 1 to
indicate the desired destination.
All of this has changed with the introduction of computer-based switches, computer control
systems for the switching, and the need for more network addresses: the last area code under
this plan was assigned in the 1990s!
Today, telephone numbers can be of the form NXX-NXX-XXXX, and the “area code” no longer
necessarily corresponds to a unique geographic area nor necessarily means long distance
charges will apply.
To provide new network addresses, area codes are split and overlaid, and users are required
in these locations to dial ten digits.
The physical destination corresponding to any particular address is now stored in a database
in a computer.
2.7.2 Address Signaling
The last main aspect of POTS is address signaling, and in particular, how the network address
of the called party is indicated or signaled from the calling party’s telephone to the CO switch.
The first kind of CO switch was a person using a switchboard and patch cords to connect loops
and trunks. In this case, the mechanism for the caller to signal the network address of the
called party was for the caller to use their voice and identify the desired called party by name.
2.7.3 Pulse Dialing
To signal numerical addresses from the telephone to the switch, a rotary dial was added to the
telephone. This dial was a metal disc with holes, connected to a dial switch inside the tele-
phone with a spring.
To indicate a digit, the caller placed a finger in the hole in the dial corresponding to that digit,
rotated the dial to a stop position, then removed their finger from the hole. As a spring rotated
the dial back to its rest position, another spring would cause the dial switch to open and close
a number of times corresponding to the desired digit.
Since the hookswitch is closed at this time, opening the dial switch would momentarily interrupt
the flow of electricity on the loop, then closing the dial switch would allow the resumption of the
current, then interrupted, then resumed, and so on.
From the line card point of view, this would appear as pulses of electrons coming down the
loop; viewed on a voltmeter, it would appear as square pulses of voltage, and so this signaling
technique is called pulse dialing.

36
 Figure 19. Rotary Dial Telephone
One question that arises is: what is the difference in function between the hook switch and the
dial switch, other than the fact that the hook switch is normally open and the dial switch is nor-
mally closed?
The answer: nothing. Both switches do the same thing: they either make or break the loop… a
Flintstones-era technology called “make-or-break” signaling. Knowing this, it should be possi-
ble to signal network addresses using the hook switch on a telephone…
The hook switch must be depressed for 45 milliseconds, then released for 55 ms, which would
be one pulse. This is repeated the number of the digit, for example, four times to indicate a “4”.
Then a pause, the inter-digit interval of 700 ms is required, then the next digit is signaled.
With some practice, it is not difficult to signal “4-1-1” using this method. Be sure to hang up be-
fore being charged for directory assistance if you try this and succeed!
There are two problems with pulse dialing: first, it is ridiculously slow – a 0 is not zero pulses,
but ten pulses, so it takes 1.7 seconds to signal a 0 including the inter-digit interval. Second,
the only device that the signaling goes to is the line card on the switch; the make-or-break-the-
loop signaling stops there.
2.7.4 DTMF: “Touch Tone”
The improvement on pulse dialing was called Touch Tone. Pulse dialing is very slow. Touch-
tone is faster.
Touch-tone is actually a registered trademark of AT&T. The generic name for this type of sig-
naling is Dual Tone Multiple Frequency or DTMF signaling. This is an address signaling mech-
anism that uses combinations of tones, i.e. single pure frequencies, to represent buttons being
pressed, and the buttons each represent a number.
On a standard telephone keypad, there are 12 buttons: 0 – 9, star (*) and octalthorpe (#). Oc-
talthorpe is commonly also called the “pound” key.
The reason this is called a dual tone signaling system is that rather than defining one tone per
button, which would require 12 tone generators in the telephone and 12 tone detectors on the
line card to represent the 12 buttons on a normal telephone keypad, the tones are instead ar-
ranged in a grid pattern, and two tones are generated to represent each button.
For example, to signal the number 4 to the line card, pressing the button marked four causes a
tone at 770 Hz and a tone at 1209 Hz to be generated.
Using two tones per button requires only 7 tones (3 + 4) instead of 12 (3 x 4), and so is
cheaper to implement: only 7 tone generators in the telephone and only 7 tone detectors in the
line card instead of 12.
Figure 20. DTMF

37
DTMF signaling is faster than pulse dialing, as the button must be depressed for a minimum of
50 ms and the inter-digit interval is 50 ms – for all buttons. A zero requires 100 ms (0.1 sec) to
signal using this method, compared to 1.7 seconds using dial pulsing.
2.7.5 In-Band Signaling
Another advantage of DTMF is that it is an in-band signaling mechanism. All of the tones are
within the voiceband: 300 - 3300 Hz. The capability put in place for voice communication is
also being used to signal control information, using tones within the frequency band used for
voice.
This allows the re-use of DTMF signaling end-to-end between customer premise equipment af-
ter the call is completed: for example, from a telephone to a voice mail system.
2.7.6 “Hidden” Buttons
Though a standard telephone keypad has 12 buttons, there are actually 16 buttons defined for
DTMF. The “hidden” four buttons are labeled A – D and share the high group frequency 1633
Hz.
These tones are used only for very special signaling situations, like Call Waiting with Caller ID.
2.7.7 Caller ID
Caller ID is another example of in-band signaling. The Caller ID is delivered to the telephone
by a 1200 b/s modem in the line card that operates in the voiceband. With standard Caller ID
service, the modem transmits ASCII code representing the date, time, calling number and pos-
sibly calling name, beginning 0.5 seconds after the first ring and ending before the second ring
happens.
During this time, the telephone is on-hook, so the called party does not hear the modem signal
being transmitted in the voiceband.
The tones corresponding to the “hidden” four buttons, A – D are used only for very special in-
band signaling situations; one example is to support caller ID with call waiting service, also
called Call Waiting ID service, where the ID of a second caller is displayed while the line is al-
ready in use with a call.
Since the Caller ID is delivered with a modem signal in the voice band, if no special measures
were taken, the called party would hear the hissing of the modem signal on the line delivering
the ID of the second caller while the first call is in progress. Plus, voice on the line might inter-
fere with the accuracy of detection of the modem signal.

38
To deal with this problem, an dual-tone CPE alerting signal of 2130 + 2750 Hz is generated by
the line card, which instructs the telephone to mute its speaker. The telephone acknowledges
with DTMF D. Then the modem signals the call waiting Caller ID and the telephone unmutes
the speaker as soon as the modem transmission is completed.
While this allows transmission of a modem signal to communicate the second Caller ID, it also
momentarily interrupts the voice communications, which can be annoying to the user.
It is an excellent example of the advantage of an out-of-band signaling system: where the con-
trol signals are not carried in the voice band, but are communicated in parallel on a separate
control circuit or channel. Having such a capability would make it unnecessary to interrupt the
voice conversation to send signals.

2.8 SS7
Once the caller has signaled the desired called party’s address from the telephone to the near-
end switch, the next two functions are routing the phone call and signaling the called number
to the far-end switch.
The called number has to be forwarded to the far-end telephone switch so that it is able to con-
nect the incoming trunk to the correct far-end loop.
In the old days, this was done using Multifrequency (MF) tones similar to DTMF on the trunk
circuits.
The problem with that was again speed, especially considering that there are multiple switches
between the near-end switch and the far-end switch, and the whole number would have to be
signaled using tones from the first switch to the second, then once that was completed, from
the second to the third, then once that was completed from the third to the fourth and so on to
the far end.
Today, a control system called Signaling System 7 (SS7), also known as Common Channel
Signaling System Number 7 (CCS7 or C7) is used to do this address signaling function.
SS7 is a global standard defined by the International Telecommunication Union (ITU) Telecom-
munication Standardization Sector (ITU-T).
It defines the protocols by which network elements exchange information for call setup, routing
and control, both wireline and wireless.

39
 Figure 21. SS7
The ITU definition of SS7 allows for variants including the American National Standards Insti-
tute (ANSI) and Bellcore standards used in North America, and the European Telecommunica-
tions Standards Institute (ETSI) standard used in the rest of the world, which is called “Europe”
in the business.
2.8.1 Out-Of-Band Signaling
With SS7, signaling is out of band, that is, using digital coded messages on separate data
channels, not using tones on the voice communication channels.
In practice, SS7 is centralized computers and databases (Service Control Points, SCPs) con-
nected via the Message Transfer Part (MTP), which is data circuits and packet switches called
Signal Transfer Points (STPs), to telephone switches (Service Switching Points, SSPs).
SS7 implements an infrastructure and standard protocols for the exchange of control mes-
sages or signaling between control computers and switches. The set of call control messages
is called the ISDN User Part (ISUP).
A company’s SS7 system will exchange ISUP messages with their switches, with other compa-
nies’ SS7 systems, and with customers’ control systems. Messages to and from customer sys-
tems are usually communicated over an ISDN Primary Rate Interface (PRI) signaling channel.
2.8.2 Advanced Intelligent Network (AIN)
In a perfect world, called the Advanced Intelligent Network (AIN), all telephone call routing de-
cisions would be made by the centralized computers, the Service Control Points, and not the
switches.

40
This has large advantages for the network service provider, since it allows the rollout of fea-
tures on the one or two sets of centralized computers, rather than on the hundreds of CO
switches.
However, having the SCPs perform all call routing introduces a single point of failure into the
telephone system… proved during a nine-hour complete failure of the telephone system on the
East Coast of the United States some years ago.
2.8.3 Switch-Based Call Routing
Due to this failure mechanism, in practice, most telephone companies use a call routing com-
puter program from a supplier like Alcatel-Lucent to update CO switch-based routing tables ev-
ery ten seconds or so.
The switch uses this table to determine the call routing, rather than a table in the SCP. This al-
lows the continued functioning of the network if the call routing computer crashes.
2.8.4 SS7 In Practice
SS7 is in practice used by big telephone companies for call setup signaling, to support data-
base inquiries, and for high-end call routing features.
Call setup signaling is indicating the called number to the far-end switch, and possibly the call-
ing number for caller ID purposes.
SS7 is also used for call setup between different carriers, for example, communicating the call-
ing number and called number from the local phone company’s system to a wireless carrier
when a call is placed from a home phone to a cell phone.
An example of a database inquiry message is credit authorization for billing phone calls, such
as when you use your telephone company calling card from a payphone, or roam with your
cellphone.
2.8.5 Residential Service Application Example
High-end value-added call routing features are sometimes called AIN services. An example for
residential service is call forwarding.
When you press *72 on your phone and hear four beeps, this indicates that you are now com-
municating with the SCP, perhaps indirectly.
When you enter the number you want your phone forwarded to, an entry is made in a data-
base, and a trigger is placed on your line card. The trigger is in fact a bit set in a status register
associated with your line card in the computer called the telephone switch.
When a call is to be routed to that number using the basic switch-based routing, the fact that
the trigger on the line card is set causes the far-end CO switch to not terminate the call on that
line card, but instead to do a query on the SCP to get the routing information – which will be to
the number you forwarded your phone to.
2.8.6 Business Service Application Example
For businesses, examples include both basic 800 service and sophisticated call routing ser-
vices that change where an 800 number is terminated based on time of day, geographic loca-
tion of the caller or the call volume.
An example of the latter is an airline that has two call centers in different parts of the country,
for example, one in Utah and one in Georgia. There is a single 800 number 1-800-AIRLINE for
that airline that is valid everywhere in North America.
By default, calls are routed based on geographic location of the caller; callers in the West are
routed to the call center in Utah, and callers in the East go to Georgia.
However, the airline pays their Inter-Exchange Carrier for a service that allows them to do load
balancing: if for example the call center in Utah becomes busy and the call center in Georgia is
not, the airline can signal the network to route phone calls to Georgia, regardless of where the
caller is geographically located… and then signal the network to change the routing back to
normal a minute later.
This idea is sometimes referred to as “customer control of the network”, perhaps more accu-
rately “real-time customer control of their call routing”. It is a sophisticated service enabled by
SS7.

41
2.9 Voice over IP (VoIP)
Though there are hundreds of thousands if not millions of traditional CO circuit switches and
PBXs still in operation, new systems are based on Voice over IP (VoIP).
We begin understanding VoIP in this chapter with fundamental concepts and a top-level view
of the major components of a VoIP system.
These components may be located at the customer premise, at a carrier, at a third party, or
any combination thereof.
Understanding VoIP includes understanding IP packets, IP addresses, digitized voice, routers,
Ethernet and a number of other supporting technologies.
The numerous supporting technologies are mentioned in this fundamentals chapter, and cov-
ered in detail in subsequent chapters.
2.9.1 Packetized Voice
Figure 22 provides a very high-level block diagram view of the processes involved in communi-
cating speech in IP packets from one person to another.
Starting on the left, commands from the speaker’s brain cause combinations of lungs, di-
aphragm, vocal cords, tongue, jaw and lips to form sounds.

Figure 22. Digitized Speech Carried in IP Packets

A microphone is positioned in front of the mouth and acts as a transducer, creating a fluctuat-
ing voltage which is an analog or representation of the sound pressure waves coming out of
the speaker’s throat.
This is fed into a codec, which digitizes the voltage analog by taking samples of it 8,000 times
per second and coding the value of sample into binary 1s and 0s. Typically, the value of each
sample is represented with a byte, meaning a overall bit rate of 64 kb/s to be transmitted.
Approximately 20 ms worth of coded speech is taken as a segment and placed or encapsu-
lated in an IP packet. The IP packet begins with a header, which is control information, the
most interesting part being the IP address of the source telephone and IP address of the desti-
nation telephone.
IP packets are moved from the source to the destination over a sequence of links. The links
are connected with routers, which relay the packets from one link to the next.

42
Lower level functions such as framing and link addressing are usually performed following the
IEEE Ethernet and MAC standards. At the lowest level, the links are physcially implemented
with Category 6 LAN cables, DSL modems, Cable modems, fiber optics and radio systems.
At the destination, the bits are extracted from the IP packet and fed into a codec, which re-cre-
ates the analog voltage.
This voltage drives a speaker, which re-creates the sound pressure waves, which travel down
the ear canal to the inner ear, causing hairs on the cochlea to vibrate, triggering neural im-
pulses to the brain, making the listener think they are hearing something.
It is important to note that the voice packets are communicated directly from one telephone to
the other over the IP network. The packets do not pass through a CO telephone switch, for ex-
ample.
2.9.2 VoIP System Components
A VoIP system includes terminals, LAN infrastructure, a softswitch, voicemail server, router,
gateway, firewall and network connections.

Figure 23. Components of a VoIP System

2.9.3 VoIP Phones and Other VoIP Terminals

Terminals, including both dedicated-purpose IP telephones as well as soft phones, which are
software applications running on general-purpose computers using Windows, Android and
other operating systems.
A VoIP terminal has a keypad, a speaker and microphone, a codec to convert between analog
and digital, and a protocol stack to handle the stream of digitized speech in typically 20 ms
segments with timing variation correction, sequencing and error control.
Since it is a Voice over IP system, the terminal must also necessarily have an IP protocol stack
for communication of the segments of digitized speech in IP packets, the DHCP protocol to ob-
tain an IP address, a LAN interface and MAC address and physical wired or wireless connec-
tion to the network.
The terminal may optionally have many other features, functions and protocols such as a dis-
play screen and Internet browser incorporated.
Another type of terminal is a voicemail server, which may be called an Integrated Messaging
System if it handles multiple different types of messages like voice, email, short text mes-
sages, fax, and supports functions like voice-to-text and text-to-voice.
2.9.4 Physical Connections: Wired and Wireless LANs

43
Another component of a VoIP system is the physical connections, implemented with LAN infra-
structure, consisting of LAN cabling or wireless LANs, and LAN switches.
2.9.5 Softswitch
The softswitch, also called a SIP proxy or call manager, is an important component of a VoIP
system. Its main function is to assist in call setup. The softswitch also manages terminals, reg-
ulates admission to the VoIP system and provides terminal authentication, registration, status
and address resolution as well as call control.
2.9.6 Router
The router connects LAN segments, properly called LAN broadcast domains, to each other
within the building, and to external connections, including to the circuit-switched PSTN, the In-
ternet, VPN services and SIP trunking services that move VoIP phone calls long distance.
2.9.7 Gateway
Gateways perform format conversions. This includes both coding format and signaling format
conversions between the IP world and the circuit-switched PSTN.
2.9.8 Firewall
A firewall system is required to manage connections to other IP networks, which include:
• The Internet, which allows any communications to anywhere but with no performance guar-
antees,
• Virtual Private Networks (VPNs), which allow any communications to specific locations
(e.g. other locations of an organization), and may include performance guarantees, and
• SIP trunking, which carries VoIP to specific locations in native format, with performance
guarantees suitable for telephone calls, and may include gateway service to convert VoIP to
traditional telephony for calls terminating on the PSTN

44
3
Switching
3.1 Telephone Network Architecture

Figure 24. Telephone Network Architecture Model

The telephone network is traditionally described as being made of three parts: access, switch-
ing and transmission, sometimes called the access network, switching network and transmis-
sion network.
3.1.1 Access Network
The access network, also called the outside plant, is the equipment and cabling used to con-
nect the customer to the switching network, typically to a Central Office.
This is also referred to as the “last mile” - though, of course, the people who work in this part of
the business prefer to call it the “first mile”. It may in fact be very much shorter or longer than a
mile.
Historically, this was implemented with twisted-pair copper wires in feeder cables with a thou-
sand or more pairs leading from the CO to wiring connection points in neighborhoods called
Outside Plant Interfaces (OPIs) or Serving Area Concept (SAC) boxes.
From there, distribution cables with a hundred or more pairs run down streets, with terminals
where a drop wire to the customer premise is connected. Physical connections are made in the
terminal and OPI/SAC box to implement a metallic connection of two wires between the CO
and the customer premise. Since electrical current flows in a loop on these two wires, the pair
is also called a loop.
Neighborhoods with this infrastructure installed are now referred to as brownfields.

45
 Figure 25. Fiber to the Neighborhood OPI/SAC Box
Fiber to the Neighborhood (FTTN) then DSL to the subscriber is used to implement high-speed
internet access in the very large installed base.
A fiber is pulled from the CO to each OPI/SAC box, which may be generically referred to as an
outside plant enclosure.
Inside the enclosure, the fiber is connected to a DSLAM, which houses banks of DSL modems.
A short pair of wires is used to connect one of the DSLAM’s modems to one subscriber for
high-speed Internet.
The customer’s network access is fiber to the enclosure, then a short run of copper to the cus-
tomer premise.
The shorter the run of copper at the end, the more bits per second can be communicated.
VDSL2 technology achieves 200 Mb/s with a maximum run length of 150 meters (150 yards).
In new neighborhoods, called greenfields, fiber to the premise is deployed. For residences and
small business, a Passive Optical Network (PON) strategy may be employed, where typically
32 customers share a fiber backhaul using time sharing. Medium and large businesses might
be connected with a dedicated fiber.
3.1.2 Switching Network
The switching part of the network was traditionally organized into a five-level hierarchy, with
the Central Office at the lowest level in a hierarchy of switching centers.
A Central Office is the wire center, where all of the access wires converge and are connected
to switching equipment. This equipment is usually owned by the telephone company, but might
also be equipment owned by a competitor collocated in the CO. In the past, this switching
equipment was a circuit switch, establishing connectivity to an outgoing circuit for the duration
of a phone call. Going forward, this switching equipment is a packet switch or router, forward-
ing one packet at a time.
This equipment is called edge equipment by network engineers, as it is notionally the edge of
the telephone company’s core network. This equipment provides a data concentration function
and converts between the physical media of the access circuit and the physical media of the
connections between switching centers and the transmission network.
The COs in a city are connected to its toll center, a building at the second level in the switching
hierarchy and the interconnection point with transmission networks owned by the same tele-
phone company or by a competitor.
3.1.3 Transmission Network
The transmission network connects switching centers, providing high-capacity and high-avail-
ability connectivity between COs and between cities. This part of the network is called the net-
work core by transmission engineers. In the past, the capacity was organized into fixed 64 kb/s
channels, with switches or routers directing traffic onto the channels. Going forward, traffic on
the core is all packets, transmitted on demand.

46
3.2 Telephone Switches
Telephone switches are used to establish connections across a network for phone calls. The
purpose of a switch is to establish a connection between one input and one output.
3.2.1 Circuit Switching
In the case of a CO switch, the connection is full-time for the duration of the call, between a
loop and a trunk, or between two loops for a call local to that switch. In a toll center, this would
be trunk to trunk connections.

Figure 26. Railroad Switching Yard

To understand what a telephone switch does, it is useful to think of railroad switching yards.
Switches in a railroad yard connects tracks so that a particular spur line (analogous to a loop)
is connected to a particular trunk line for the duration of a train, then the switches are changed
so that a different spur line is connected to the same trunk for the next train.
The same principle applies to telephone switches: a route decision is made, loops are con-
nected to trunks to implement a connection for the duration of a phone call, then a different
loop is connected for the next call.
This type of switching: a full-time connection for the duration of a call, is called circuit-switch-
ing. We sometimes use the term traditional to refer to this kind of switching, as it will be slowly
but surely replaced with Voice over IP, soft switches and packet switching.
3.2.2 CO Switches
Telephone switches are computers. They are often constructed as rack-mount systems en-
closed in cabinets. Nortel switching equipment was traditionally painted brown, though is now
a light gray color. Lucent equipment was traditionally painted off-white. Others are beige.
Other than that, all telecom switching equipment looks similar: like large upright filing cabinets
with many wires connected on the back and indicator lights on the front.

47
 Figure 27. Front Bay of a DMS-100 CO Switch
Alcatel/Lucent’s CO switch product is the Class 5 Electronic Switching System (5ESS). Nortel’s
main product was the Digital Multiplex Switch model 100 (DMS-100). The DMS product line
and the servicing of its installed base was acquired by Avaya following Nortel’s bankruptcy.
These switches are capable of handling up to 100,000 loops, but are usually built up to a maxi-
mum of 60,000 loops per switch.
There are many other switch manufacturers and products.
3.2.3 Line Cards
The twisted pair loops are carried into the switch on a Main Distribution Frame. The compo-
nent of a switch that terminates a loop is called a line card.
Just as a PC can have an adapter card that allows a telephone line to be plugged into the PC,
a telephone switch has line cards to allow the connection of loops to the switch.
Individual line cards are implemented as small Printed Circuit Board (PCB) line card modules,
plugged onto a larger PCB, mounted in a drawer in a shelf in a rack as illustrated in Figure 28.

Figure 28. Line Card Drawer

There are 48 line card modules in the drawer illustrated, meaning that there can be over 1,000
line card drawers as part of this CO switch.

48
Line card drawers make up most of the footprint of a CO switch. Figure 27 illustrates only the
first row of racks of a DMS-100 switch. There are at least ten full-length rows of racks packed
with line card drawers behind it.
3.2.4 Digital Switching
All of the communication of voice information inside the switch is digital. As we will see in
Chapter 5, “Digital”, the analog voice signal on a loop is digitized at 8,000 bytes per second, or
64 kb/s on the line card.
The fundamental task of a traditional telephone circuit switch is to transfer a byte through the
switch from one input to one output, and vice versa, eight thousand times per second, for the
duration of a call.

3.3 Traditional PBX and Centrex

Traditional Centrex and PBX means systems developed and installed before VoIP was a real-
ity.
Both are based on a large rack-mount computer system called a telephone switch. The tele-
phone is connected to the switch, which establishes connections on a call-by-call basis.
There are three main differences between Centrex and PBX: location, location and location. It
depends whether the telephone switch is at the customer premise (PBX), or in the Central Of-
fice (Centrex).
3.3.1 PBX
If the switch is in the customer premise, this is called “having a PBX”, a Private Branch Ex-
change. The customer can buy, rent or lease one, and can provide their own dial tone, in-build-
ing dialing plan and in-building switching.
If the dialing plan involves assigning every telephone in the building a four-digit “extension”
number, a user enters just the four-digit number to make a call to another phone connected to
the PBX in-building.
Many features such as no-answer transfer, call pickup groups, Interactive Voice Response
menuing and call center functions like Automated Call Distributor (ACD) can be provided by
the PBX.
3.3.2 PBX Trunks
A PBX is connected to the telephone network with PBX trunks between the PBX and CO
switch. Usually one trunk is provided for approximately every ten telephones.
When a user goes off-hook, they hear a dial tone from the line card in the PBX. If they dial 9
for an “outside line”, they are assigned a PBX trunk, a circuit-switched connection is made to
the CO switch, and the user hears a second dial tone generated by the CO switch.

Figure 29. PBX and PBX Trunks

PBX trunks can be ordered as one-way incoming, one-way outgoing, or both-way. This does
not refer to the voice communications capability, which of course happens in both directions in

49
all cases; it refers to whether these trunks are used to receive phone calls, initiate outgoing
calls or both.
3.3.3 Digital Telephones: Electronic Business Sets
Both Centrex and PBXs support analog and digital telephones. Digital telephones, which are
often called Electronic Business Sets, are far more popular as they support a much richer user
interface and feature set.
“Digital telephone” means that the voice is digitized in the telephone and communicated as 1s
and 0s along with call control messages from the phone to the switch, represented by pulses
of voltage on copper wires.
For the traditional Centrex and PBX described in this section, the coding and formatting of the
digitized voice and the call control messages are not standards-based, meaning that only tele-
phones supplied with the Centrex service or PBX device will work. This is on purpose, to lock
the switch customer into buying all phones and upgrades from the PBX or Centrex vendor.
This is a source of profit for the vendor.
New-generation Voice over IP systems are much more likely - but not guaranteed - to use
standard methods of coding and call control messaging, which would allow the use of third-
party telephones.
3.3.4 PBX and PABX
The term Exchange is an older term for a circuit switch. Private means that the customer has
the switch, not the telephone company. Branch refers to the topology of PBX trunks looking
like branches off the main telephone company tree trunks.
In the beginning, a PBX was implemented with a board with jacks terminating loops and PBX
trunks, and an operator connecting loops to PBX trunks manually with a patch cord, like the
first CO switches.
Like CO switches, PBXs came to be implemented with mechanical systems then computers.
For a while, computer-based PBXs were called Private Automated Branch Exchanges
(PABXs). This term is not used much today.
3.3.5 Attendant
Even though the switch is implemented with a computer that connects PBX trunks and tele-
phones, an operator or attendant is required to route inbound calls, to connect an incoming call
to the correct telephone in-building.
Typically, all of the inbound trunks will be associated with a single telephone number valid on
the public telephone network. When a caller dials that number, the CO switch will connect the
caller to an available incoming trunk, terminating on the PBX.
In the simplest implementation, the PBX by default connects all incoming calls to an attendant
console, where the attendant answers the call and asks the caller to whom they would like to
speak.
The caller identifies the desired called party by name to the attendant using their voice, the at-
tendant looks up the corresponding extension number on a piece of paper or computer screen,
enters the extension number in the console and presses the “transfer” button.
This instructs the PBX to now connect the incoming trunk to the line card corresponding to that
extension number and start it ringing.
3.3.6 Automated Attendant
An attendant is expensive, and can route only one call at a time. A computer program running
on the PBX performing the attendant function, called an automated attendant, is much less ex-
pensive than an employee, and can handle more than one incoming call at a time.
Incoming calls are first terminated on automated attendant software on the PBX, which plays a
recorded message to the caller requesting that if the caller knows the extension number they
would like to be connected to, that the caller use in-band DTMF signaling to indicate the exten-
sion number.
Once received, the automated attendant software conveys the extension number to the PBX
switching software, which connects the incoming trunk to the appropriate line card and starts it

50
ringing.
If the caller does not supply the extension number, possibly because they have a rotary-dial
phone, or they don’t know it, or selected a preconfigured option like 0, the automated attendant
software will route the call to a human attendant.
In a low-budget implementation with no backup human attendant, the caller might be trans-
ferred to a voice mailbox and asked to leave a message.
3.3.7 IVR
The automated attendant function is usually implemented in practice as part of an Interactive
Voice Response (IVR) system running on the PBX.
An IVR provides more ways for the caller to have their call routed through the PBX to a particu-
lar telephone without knowing the extension number.
The most common implementation involves a recorded message asking the caller to signal a
number corresponding to one of a number of menu choices. The result will either be transfer to
a particular extension, or to a second menu where the process is repeated.
Speaker-independent word recognition has become reliable enough that the caller may have
the option to speak words to navigate the menu instead of using in-band DTMF signaling.
In addition to determining the extension to which to transfer the call, an IVR may be used to
have the caller enter information, for example, their account number at the called organization.
A sophisticated IVR might be integrated into the called organization’s customer care system,
allowing the caller to retrieve information without speaking to a person.
An example would be calling an airline and the caller entering their frequent-flyer number, then
having the IVR do a query on the airline’s customer care system to determine that account’s
mileage balance and communicating it to the caller using recordings of someone speaking
numbers.
3.3.8 Direct Inward Dialing (DID)
The telephone company controls the telephone numbers and charges per number, per month
to assign numbers to users.
The lowest-cost configuration with a PBX is to pay for only one telephone number for the PBX.
All inbound PBX trunks are associated with that one telephone number in a hunt group. When
a call is placed to that number, the CO switch hunts through the group of trunks to find the next
one available and connects to call to the PBX on that trunk. The caller must as a second step
indicate to an attendant or automated attendant to whom they wish to speak.
If an organization desires to have a PBX but eliminate the two-step process of first dialing a
number then dealing an attendant or automated attendant to have calls connected, the organi-
zation can pay the phone company for Direct Inward Dial (DID) service.
With DID, the hardware configuration of PBX connected to CO with PBX trunks of Figure 29
remains unchanged. The telephone company assigns a telephone number called a DID num-
ber that is valid on the public telephone network, for each of the extensions on the PBX.
When a caller dials one of those DID numbers, the call is connected from the CO to the PBX
over a PBX trunk as usual, plus the CO switch indicates to the PBX the DID number that was
called. The PBX can then look up in a table to determine the extension number associated with
that DID number and switch the call to the correct line card without any further interaction by
the caller.
This service is billed per DID number, per month by the phone company.
3.3.9 Automated Call Distribution (ACD)
The destination of a call could be an extension that identifies not a particular line card, tele-
phone and person, but instead identifies an Automated Call Distributor.
When running on the same hardware as the PBX switching function, an ACD is a computer
program that deals with situations when there are more callers than there are people or agents
to answer the calls.
The ACD is configured to have queues associated with extension numbers. The queues are
associated with specific activities, such as a particular type of caller wanting to perform a par-

51
ticular activity.
Upon being transferred to the ACD, the caller is placed in a queue and recordings are typically
played to the caller to keep them interested. Which queue the caller is placed in might be de-
termined by the number they dialed, or choices they made in an IVR before being transferred
to the ACD.
Agents are associated with queues. An agent can be dedicated to answering one call queue,
or able to answer multiple call queues. When the ACD determines an agent is available to an-
swer the next call in a queue, the caller is switched from the ACD’s recording to that agent.
3.3.10 Call Centers
Inbound call centers are places where customer service agents receive calls from customers
and access the customer’s account information via a terminal connected to a customer care
system, a data processing system.
Traditionally, this has been a customer-premise-based solution. The end-user company buys,
integrates and maintains a PBX to handle incoming calls, an IVR to get information about the
caller, and an ACD to route the call to an agent, and a customer care system to store and ma-
nipulate information about customers and orders.
The agents sit in a large “call center” room with supervisors and may have to raise a paddle to
request to go to the bathroom… and might not be allowed to if the call volume is heavy.
A sophisticated integrated system would first pass the caller through an IVR to determine their
account number and desired activity, then to the appropriate queue on an ACD, then when the
call is finally switched to an agent, send a message to the customer care system. The cus-
tomer care system would then cause the caller’s account information to appear on the agent’s
screen at the same time the agent answers the call.

Figure 30. Integrated Call Center

One of the main purposes of this integration would be to reduce the amount of time the agent
has to spend with the caller to determine who they are and what they want to do, which re-
duces the number of agents required and thus saves money.
The infrastructure can of course be outsourced, using network-based call center services,
where a third party has the IVR and ACD and handles the call queuing and call setup for
agents at a different location.
Using Voice over IP and SIP call setup, the agents can be anywhere… even working at home.
This would provide significant benefits in flexibility of staffing to handle peak call volumes to
meet call-answer-delay requirements.
A recent trend was to locate the agents in countries where there were good IP telecom ser-
vices, salaries low and employment standards less stringent than elsewhere. This was in some
cases abandoned after companies decided that low-budget customer service was a revenue-
negative idea.

52
The next step is multimedia contact centers, where there are a number of different ways that a
customer can contact the agents in the center, including speaking to the agent, e-mail, web
chat, web collaboration, click-to-talk, and click-to-see.
3.3.11 Advantages of PBX
The main advantages of a PBX system is the service pricing model and the ability to control
the hardware and features.
With Centrex service, described in following sections, the switch is at the CO, and the tele-
phone company provides telephones connected with individual lines to the CO switch.
With a PBX, the connections are trunks, not loops, with something like one trunk for every ten
telephones. PBX trunks cost more than individual lines, but not ten times as much. This means
that the monthly service cost is less.
Moreover, the cost of value-added features like call forwarding and voice mail are notionally
per PBX, not per line, which is a definite cost savings.
Plus, the customer determines which features are available, based on selection and configura-
tion of the PBX by them, not by the phone company.
Another advantage of having a PBX is not having to pay the phone company for moves and
changes. If a person moves to a new cubicle, and wants to keep their extension number, a
technician has to reconfigure the switch so that the extension number is associated with the
line card or wires going to the new cubicle, and not the old cubicle.
With a PBX, the organization can perform moves and changes with in-house staff instead of
paying the phone company to do it.
3.3.12 Disadvantages of PBX
The main disadvantages of having a PBX are capital cost, scalability, support and mainte-
nance. When an organization gets a PBX, they are going into the local telephone business in-
side their building, and must perform all the functions of a local telephone company.
This means the organization must decide which manufacturer and which model of PBX to get,
how many of them to get – one for each location is the starting point – plan for future growth
and future features and technologies, and finance the hardware.
Planning for the future is especially important considering that the connection from the PBX to
the telephone is historically not standards-based, meaning that only particular telephones
made by the same manufacturer will work with the PBX.
The implication is that once a PBX is purchased, the customer is obliged to purchase all future
telephones from that manufacturer, which may turn out to be costly.
Having a PBX means the organization must have a help desk, trouble ticket system and skilled
staff to operate, maintain and repair the telephone system and to deal with the carriers provid-
ing local and long-distance telephone service.
The organization must also decide how long they want their telephones to keep working after
the big ice storm, hurricane or earthquake knocks out all of the main power distribution for
thirty miles around the building. A minute? An hour? A week?
The latter requires a contract signed before the disaster happens for guaranteed delivery of
fuel for generators… when the city is blacked-out and everyone wants fuel, this organization is
the one that will get it.
Finally, unless the organization is willing to pay per extension per month for DID service, call-
ers have to go through a two-step process to connect a call: first dialing a phone number, then
dealing with an attendant or IVR before the call is connected.
3.3.13 Centrex
Service with the exact same look and feel to the user as having a PBX can be provided by the
telephone company. This is generically referred to as Centrex. Every telephone company has
their own brand name for this service or bundle of services.
Centrex means that the telephone service is provided by a CO switch, rather than a PBX. Typi-
cally, a part of the CO switch will be partitioned in software and dedicated to a particular cus-
tomer, making it appear to the customer as though they have their own switch, with the same
features as a PBX such as four-digit dialing and having to dial “9” for an outside line.

53
With Centrex service, the connections between the telephone company and the customer are
loops - one for each telephone. If a particular Centrex customer has many telephones, the tele-
phone company will carry the loops not on many pairs of wires, but instead on a single fiber
optic loop carrier system between the buildings.
The loop carrier system does not add any value to the service; it is simply a mechanism to
carry the information for the individual loops together on one fiber instead of on many sets of
copper wires.
In this case, as illustrated in Figure 31, the fiber terminating equipment at the customer
premise for Centrex has line cards connected to telephones by copper wires.
PBX trunks are carried the same way, on fiber, so the fiber terminating equipment at the cus-
tomer premise for PBX also has line cards connected to telephones by copper wires.
In fact, comparing the two architectures in Figure 31, the choices are identical except for loca-
tion, location and location: whether the switching is at the customer premise (PBX), or at the
CO (Centrex)
Figure 31. Traditional PBX and Centrex solutions have the same physical layout. The location of the switching is
the difference.

.
3.3.14 Advantages of Centrex
The main advantage of Centrex is that the phone company will take care of planning, purchas-
ing, installation and maintenance of the telephone switch and telephones, and provide a ser-
vice agreement specifying the availability of service and time to repair. The customer does not
need to have experts on staff to configure and maintain the telephone system.
With Centrex, there is no capital cost for the switch, though there may be for the phones.
Monthly payments with a fixed-length contract are typical.
In addition, the phone company deploys many switches in different geographical areas, facili-
tating the implementation of seamless regional and national service, and ensuring that there is
enough switching capacity for each of the customer locations.
3.3.15 Disadvantages of Centrex
The downside of Centrex is cost. Centrex is not a money-losing business at the phone com-
pany; it’s part of their bread-and-butter.
The pricing model for Centrex service is per line. The monthly service charge for dial tone is
per line. Cost for voice mail and features like call forwarding is per line. This ends up being
more expensive than service implemented with a PBX, where the pricing model is more per-
PBX.
In addition to monthly charges, another cost with Centrex is moves and changes.

54
When someone changes cubicles and wants to keep their phone number, the switch has to be
reconfigured to associate the phone number with a different line card, terminating the wires go-
ing to the new cubicle.
This costs in the neighborhood of $100 per change... every time someone changes cubicles.
3.3.16 PBX vs. Centrex
The question of PBX vs. Centrex often boils down to this question: what business is the cus-
tomer in? Do they want to devote part of their energy to providing local phone service, and
save some money by doing it themselves; or do they want to use their energy for their printing
business and pay the phone company to do phone service, knowing that the phone company
makes a profit doing it.
3.3.17 Key Systems
A key system is a low-capacity, low-budget combination of Centrex and PBX functions. A key
system terminates lines from the phone company, not PBX trunks, but, like a PBX, allows the
connection of more phones than there are lines. A 3x8 key system would support up to 8 tele-
phones in-building connected to one of 3 phone lines.
In the old days, mechanical key systems used telephones with a row of transparent buttons
across the bottom to select which line the phone was connected to.
More recent electronic key systems use Electronic Business Sets with programmable buttons
and displays, more or less identical to those used for Centrex and PBX.
In the future, all call setup will be done with SIP.

3.4 SIP, Soft Switches, Hosted PBX and IP Centrex

3.4.1 Hard Switches
In its simplest form, a switch is a device that enables communications from one point to one
other specific point, usually when there are multiple points to choose from.
A traditional Central Office telephone switch or PBX might be called a “hard” switch. The vast
majority of the floor space taken up by this kind of switch is line cards, the physical termination
for customer access circuits.
The telephone is plugged into the telephone switch, or more accurately, the loop is connected
to a line card in a drawer, in a shelf, in a rack that is part of the physically large rack-mount
computer system called a telephone switch.
The telephone switch physically moves the speech inside the switch from one line card to an-
other to implement the connection for the duration of a phone call.
3.4.2 Soft Switches
In the Voice over IP (VoIP) world, access circuits are not terminated on a switch, and the voice
communications do not flow through a switch.
VoIP telephones are connected to a LAN in the same way as desktop computers, and the digi-
tized voice is carried in IP packets directly from one telephone to another over the LAN. There-
fore, no line cards are required on a VoIP switch.
If all of the line cards are removed from a “hard” switch, what is left over is the “soft” part of the
switch or softswitch, which is the software that performs call setup functions.
Soft switches are deployed by carriers to provide network-based call setup, replacing CO
switches and toll switches – and by end-user business customers to perform in-building call
setup, replacing PBXs.
A business customer can physically implement the software that is the softswitch on a com-
puter at the customer premise, or can outsource this to a third party who implements the soft-
ware that is the softswitch on a computer at a remote location.
There are many terms used for soft switches. Most VoIP systems today support the Session
Initiation Protocol (SIP), uses the terms proxy and back-to-back user agent. Terms used by
product manufacturers for their products that might implement these and other functions in-
clude call manager, call server, VoIP switch, communication server and hosted PBX.

55
Regardless of what it is called, the main function of a softswitch is call setup, and the essential
function is to inform the telephones at each end of the call of the other telephone’s IP address,
since the voice goes in IP packets directly between the phones, not through a telephone
switch.
For both privacy and flexibility, the IP address of the called party’s telephone is usually not
published. It has to be determined before voice communications can begin.
3.4.3 SIP
During a VoIP telephone call, the telephones send IP packets containing digitized speech di-
rectly to each other. To be able to do this, the telephones must know each other’s IP address.
In a standards-based system, the SIP protocol is used to inform the telephones of each other’s
IP address.
Each phone is associated with a SIP server, which acts on behalf of or is a proxy for the tele-
phone to set up the call. Instead of a telephone number, each person has an Address of
Record, which in the SIP standard has the same format as an email address, for example,
[email protected].
For interoperability with traditional systems, the Address of Record might be translated to a for-
mat that looks like a traditional telephone number... but this has to be resolved behind the
scenes to a SIP standard format.
Everyone’s Address of Record is made visible to the public. This would be printed on business
cards and included in email signature blocks.
To make a VoIP phone call to someone, it is necessary to find out what their telephone’s IP ad-
dress is.
It is possible to find out the IP address of their SIP server by looking it up in the Domain Name
System (DNS) just like a web server... but it is not possible to determine their telephone’s IP
address. Only their SIP server knows their phone’s IP address.
When a phone is plugged in or restarted, it is assigned an IP address like any other computer.
Then the phone registers with its SIP server, that is, informs its SIP server of its current IP ad-
dress.

Figure 32. The SIP Trapezoid for Call Setup

To establish a call from telephone A to telephone B as illustrated in Figure 32, caller A asks
their SIP proxy server to initiate the phone call to B’s Address of Record.
A’s SIP server looks up the IP address of B’s SIP server in the DNS, then sends a session initi-
ation request to B’s SIP server.
Since telephone B previously registered with B’s SIP server, B’s SIP server knows the IP ad-
dress of telephone B. B’s SIP server then passes on the incoming call request to B’s tele-
phone.
If B indicates to B’s SIP server that they will take the call, B’s SIP server transmits the IP ad-
dress of telephone B to A’s SIP server, which in turn relays it to A’s telephone.
At that point, the two telephones can send IP packets containing digitized voice (called media
communication in SIP) directly from one to another, and the SIP servers are no longer in-

56
volved.
This is necessarily a simplified explanation of the SIP call setup protocol, but hopefully con-
veys the essential idea. Voice digitization, IP addresses and packets, DNS and other protocols
are covered in subsequent chapters.
3.4.4 Additional Functions
In addition to running a SIP server for call setup, the softswitch may also perform authentica-
tion, authorization and accounting functions such as generation of Call Detail Records, and po-
tentially hundreds of other call setup and processing functions, such as voice mail, integrated
messaging, call pickup groups, Interactive Voice Response (IVR) functions and Automated
Call Distributor (ACD) functions.
3.4.5 Location Independence
The softswitch can be located anywhere on the planet.
As long as IP packets containing the SIP call setup messages can be communicated from a
telephone to its SIP server with suitable maximum delay and packet loss, it is irrelevant where
the SIP server is physically located.
The SIP server could be located at the customer premise, at a telephone company building, or
at some third party data center.
The SIP messages between the telephone and its SIP server are short and simple, requiring
very low bandwidth compared to the subsequent exchange of digitized speech between the
telephones.
3.4.6 Customer Premise Softswitch
When the hardware and software implementing the softswitch is located at the customer
premise, typically purchased by the customer, it is usually called a softswitch, call manager,
unified communications system or IP phone system.
3.4.7 Centrex
When the hardware and software implementing the softswitch is located at a telephone com-
pany, its functions are provided as a service by the telephone company, and might be called IP
Centrex or Hosted VoIP by the telephone company’s marketing department.
Alternatively, the phone company might continue to the call the service Centrex to avoid con-
fusing anyone. The fact that it is implemented IP packets and SIP for new customers is just a
detail.
3.4.8 Hosted PBX
When the hardware and software implementing the softswitch is located at a third party, the
software and the hardware it runs on is provided as a service by the third party and is usually
called a Hosted PBX, similar in concept to web hosting and virtual web servers.
Of course, telephone companies may be in the business of providing Hosted PBX services in
addition to IP Centrex.

3.5 SIP Trunking

In days past, traditional business telephone systems (PBXs) were connected to the outside
world by PBX trunks from the business customer premise to the local phone company (Local
Exchange Carrier, LEC), which would then provide switched access to a long-distance com-
pany (Inter-Exchange Carrier, IXC), to connect to a far-end LEC and far-end customer to com-
plete a telephone call.
This is being replaced with IP packet-based services to communicate SIP call setup messages
and the subsequent digitized speech between the VoIP telephones in one building and VoIP
telephones in another building.
These IP packet-based services are called SIP trunking services by the marketing department.
3.5.1 PBX Trunks and Tie Lines
In medium and large installations, traditional PBX trunks were implemented with ISDN Primary
Rate Interface (PRI) technology, which carries PBX trunks digitized at the 64 kb/s DS0 rate on

57
channels.
Also in days past, tie lines were services that appeared to the customer to be dedicated lines
that directly linked the customer’s PBXs in different cities. A sophisticated system of tie lines
connecting multiple locations of an organization plus four- and five-digit dialing plans was
called a Virtual Private Network (VPN). It is to be noted that this kind of voice VPN is not the
same thing as today’s MPLS or IP packet encryption-based VPNs.
LECs, IXCs, switched access, ISDN, PRI, DS0 and channels, MPLS and IP encryption-based
VPNs are all covered in upcoming chapters.
3.5.2 VoIP Trunking
In the early days of VoIP, a business could implement VoIP on LANs in-building, but there were
no carrier services with performance guarantees suitable for moving IP packets containing dig-
itized speech between buildings.
The business had to convert VoIP phone calls to traditional telephony, that is, PBX trunks car-
ried on ISDN PRI connections to the LEC, for phone calls to their other locations, then convert
them back to VoIP at the other end.
This was in essence islands of VoIP connected with the “old” circuit-switched technology and
services.
Now, carriers offer IP connectivity for business customer VoIP systems, which can be used for
VoIP phone calls between two locations of the same business in different cities, and for VoIP
phone calls to the PSTN.
In the case of VoIP phone calls between two locations of the same business, the IP packets
containing digitized voice are carried natively between the two locations, that is, without being
converted to something else.

Figure 33. SIP Trunking

The SIP trunking carrier provides transmission characteristics including maximum delay and
packet loss suitable for voice communications, so that the reconstructed sound at the far end
is of the same quality as traditional telephony.
This IP-based SIP trunking service is generally much less expensive than the legacy ISDN PRI
+ IXC method.
3.5.3 Gateway Service
At present, there are no tariffs or agreements for terminating a VoIP phone call in IP packets
on a LEC for subsequent delivery to one of the LEC’s customers.
In the case of VoIP phone calls from a business VoIP system to the PSTN, for example, phon-
ing from work to home, it is necessary to convert the VoIP to traditional telephony to hand the
call off to a LEC.
To accomplish this, the carrier providing the SIP trunking service may also provide a gateway
service, to convert the phone call from VoIP to traditional telephony, where the voice is coded

58
as a 64 kb/s DS0 stream, then hands it off to the LEC in exactly the same way as an Inter-Ex-
change Carrier hands long-distance phone calls to the LEC.
In the future, agreements for connections to LECs in the form of IP packets instead of DS0
channels will be formalized in tariffs and this conversion will not be required.

59
4
The Telecommunications Industry
4.1 US Telephone Companies
The official parlance for a company that provides POTS in the US is Local Exchange Carrier
(LEC). These companies own the CO and the “last mile” of cabling between the CO and the
customer premise.
At the CO, they provide to the customer equal access to long distance or Inter-Exchange Carri-
ers (IXC). Many companies are both LECs and IXCs.
4.1.1 LECs, LATAs and Baby Bells
AT&T used to own the “Bell System”, with local and long-distance operations spanning the
country.
To settle an anti-trust suit, the Bell System was split into pieces in a process called divestiture:
AT&T divesting itself of its local operations in the 1980s.
To define how the physical network was to be split, the US lower 48 states were geographically
divided into 195 Local Access and Transport Areas (LATAs). The piece of what used to be the
Bell System’s local operations that found itself in a LATA became a Local Exchange Carrier, a
LEC.
The ownership of each LECs was then spun off to one of seven holding companies, known as
“Baby Bells”, Regional Bell Holding Companies (RBHCs) and Regional Bell Operating Compa-
nies (RBOCs).
The Baby Bells were US West, Pacific Telesis Group (Pac Bell), Southwestern Bell, BellSouth,
Bell Atlantic, NYNEX and Ameritech. Each of these companies owned a number of LECs.

Figure 34. The Baby Bells and Subsequent Consolidation

For example, Ameritech became the new owner of the LECs that made up the former Michigan
Bell and Illinois Bell, spray-painting “Ameritech” on the sides of all of the Michigan Bell and Illi-
nois Bell trucks.
4.1.2 Independents
Apart from the Bell System, there were somewhere around 1,600 other local carriers in the US
in 1975.

60
These companies, such as Champlain Telephone in Champlain NY are referred to as indepen-
dents since they were never part of the Bell System.
Large companies including Century Link, Frontier Communications, Verizon and Windstream
own many independents.
4.1.3 Inter-Exchange Carriers: IXCs
One of the main purposes of the breakup of the Bell System was to introduce competition in
long-distance communications.
This was initially implemented by splitting the Bell System into LECs as described in the previ-
ous section, and mandating LECs to complete phone calls within the same LATA, and equal
access to competing Inter-Exchange Carriers who would carry calls from one LATA to another.
The biggest IXC was AT&T Corp., which started out as the long lines that were left over after
the local operations were divested.
The first main competitor was Metropolitan Communications Inc. (MCI) that broke ground with
fiber-optic connections in Chicago. Sprint was number three, WorldCom number four. World-
Com purchased MCI to move up to the number two spot, and before going bankrupt, spinning
off the MCI residential long-distance unit, and changing the name of the resulting long-distance
company back to MCI.
4.1.4 Switched Access Charge
The theory was that the LEC owned the CO and the access network, also referred to as the
outside plant: the connection from the customer premise to the CO. The IXCs only have long-
distance circuits.
If a call stayed with a LATA’s boundaries, then the LEC provided access for and transported
the call.
If the call crossed a LATA boundary, three companies have to be involved: the LEC for local
access, an IXC for long-distance, and a LEC for local access at the far end.
The customer signs a contract with the IXC for long distance communications. To make a long-
distance phone call happen, LECs at each end switch in a connection from the customer to the
IXC for the duration of the phone call.
For this connection, the IXC has to pay the LEC a per-minute switched access charge which
varies from $0.0075 to $0.07 per minute in different places.
4.1.5 CLECs and ILECs
Competitive Local Exchange Carriers (CLECs) are companies that can co-locate equipment in
the CO and provide services over the local access network to customers. The LEC that built
the CO is referred to as the Incumbent Local Exchange Carrier (ILEC) to distinguish the two.
4.1.6 Resellers
Another category of carrier is a reseller. These companies lease high-capacity services from
Inter-Exchange Carriers, and buy or lease switching capability, then sign up customers and
convince them to route their calls over the high-capacity leased service: buying wholesale and
selling retail.
The profit margins for resellers have been dramatically reduced, to the point where many for-
mer resellers have either gone out of business, or been forced to purchase physical circuits
and turn into facilities-based Inter-Exchange Carriers.
4.1.7 Consolidation
A consolidation took place: Bell Atlantic purchased NYNEX then merged with GTE and re-bap-
tized the result “Verizon”; SBC purchased Pac Bell, Southwestern Bell, Ameritech and SNET
and later BellSouth.
Qwest, a fiber-backbone company, purchased US West, then merged with CenturyTel to form
CenturyLink, which subsequently purchased many independents.
Federal legislation passed in 1996 permitted wide-open competition, with all companies per-
mitted to provide all types of services.
Individual state Public Utility Commissions have a large amount of regulatory power indepen-
dent of the federal government, regulating activities within a state.

61
This means that a national carrier has to deal with paperwork and regulations from more than
51 regulatory agencies, so the possibility of the LECs operating as national IXCs took some
time.
Once the LECs could be IXCs, ownership and operation of the local and long-distance opera-
tions were merged back together by the LECs purchasing the IXCs.
SBC purchased AT&T Corp and changed its name to AT&T. Verizon purchased the next-big-
gest IXC, MCI, reconstituting most of the Bell System in two geographic landline pieces: AT&T
and Verizon. Each company provides services in the other’s territory via fiber, collocations and
especially wireless.
4.1.8 Cable TV: Broadband Carriers
Companies using a coaxial entry cable, primarily to residences were historically called Com-
munity Antenna Television (CATV) companies and more recently cable companies. With the
delivery of telephone service and Internet via broadband modems, and backbone fiber net-
works, cable companies are now full telecom service providers and carriers, for both business
and residence, and are properly referred to as broadband companies.

4.2 AT&T and Verizon

Figure 35. Verizon and AT&T ILEC Territories

Figure 35 portrays the geographic areas where either Verizon or AT&T is the Incumbent Local
Exchange Carrier (ILEC), the company that owns the Central Office and access wiring.

62
These maps reveal that the notional geographic division of the US into seven sections, each
controlled by a Baby Bell, is no longer of much use as a model.
The “eastern” Baby Bells: NYNEX and Bell Atlantic merged with GTE to form Verizon, which
dominates New York City and Boston, and also provides service in many parts of the country
including significant areas of California, Oregon and Texas among others.
AT&T, formed by the merger of Southwestern Bell, Pac Bell, Ameritech, SNET, AT&T Corp.
and BellSouth, covers more than half of the country: Texas, Oklahoma, California, Illinois,
Michigan, Connecticut, Florida, Georgia and more.
Statistics after the BellSouth merger (as of 2005 09 30 and 2006 03 01) show that AT&T ended
up a bit bigger than Verizon:
• Verizon: 49,689,000 access lines, 4,531,000 DSL lines.
• AT&T: 70,200,000 access lines; 9,400,000 DSL lines.
Wireless coverage by affiliates of these two companies is much more complete, both covering
all of the top 100 markets in the country.
It is important to also keep in mind that in addition to CenturyLink, there are hundreds of inde-
pendent ILECs that are not owned by Verizon or AT&T. And hundreds of CLECs.

4.3 Canadian Telephone Companies

In the previous millennium, there were 58 telephone companies in Canada. The largest tele-
phone companies were part of a national alliance called Stentor Canadian Network Manage-
ment, the “Stentor Alliance” or simply, “Stentor”. These companies were:
• British Columbia Telephone Company (BCTel)
• Alberta Government Telephones (AGT)
• Saskatchewan Telecommunications (SaskTel)
• Manitoba Telecom Services (MTS)
• Bell Canada
• The New Brunswick Telephone Company Limited (NBTel)
• Maritime Telephone and Telegraph Company, Limited (MT&T)
• The Island Telephone Company Limited (Island Tel)
• Newfoundland Telephone Company Limited (NewTel),
The remaining companies were called independents because they were not full members of
Stentor. There were 30 in Ontario, 16 in Quebec, one in BC and one in the North-West Territo-
ries. These companies provided both local and long-distance telecommunication services, as
well as cellular service through affiliates, under tight federal regulation.
The Stentor Alliance was terminated on December 31, 1999. In its place, a consolidation cen-
tered around two main players, Bell and TELUS, has taken place.
4.3.1 Bell Canada
Bell Canada historically provided service in Ontario and Quebec, and more recently controlled
Aliant Telecom, which was a merger of the four maritime telephone companies.
In July 2006, Aliant’s wireline operations, Bell Canada rural wireline operations plus Bell Nordiq
were combined into an income trust called Bell Aliant Regional Communications (BARC), with
about 3.4 million local access lines and over 400,000 high-speed Internet subscribers in six
provinces. At the same time, Bell Canada acquired Aliant Mobility’s wireless operations.
Bell Canada is currently composed of Bell Mobility plus the remainder of the wireline opera-
tions, concentrated in the densely-populated Quebec city to Windsor corridor.
Like TELUS in Toronto, Bell provides services to business customers in Vancouver, Calgary
and other cities where it is not the incumbent.

63
 Figure 36. Canadian Telephone Companies

4.3.2 TELUS
BCTel, AGT, Ed Tel and Quebec Tel merged to form TELUS, providing virtually all landline ser-
vice in BC and Alberta, plus service to some 443,000 customers in Quebec. Using a mixture of
POPs, collocation and subcontracts, TELUS provides service to business customers in
Toronto, Montreal and other cities where Bell Canada formerly held a monopoly.
4.3.3 Ownership
Ownership of the telephone companies is varied. BCE of Montreal, a publicly-owned holding
company, owns Bell Canada and BARC. Verizon (through its acquisition of GTE) originally
owned half of BCTel and half of Quebec Tel through a holding company in Montreal. This was
divested in 2005, sold to TELUS, also a publicly-owned holding company. SaskTel is a provin-
cial crown corporation, as were AGT and MTS until they were privatized.
4.3.4 Competitive Inter-Exchange Carriers
Competitive Inter-Exchange Carriers exist, but find their profit margins slim due to the geo-
graphical nature of the network and strong competition. The main competitor traces its roots
from CNCP Telecommunications, to Unitel, a joint venture of Rogers Cable and CNCP, to
AT&T Canada, which sold its residential long-distance voice services to Primus then changed
its name to Allstream, which was then acquired by MTS.
4.3.5 Resellers
A number of resellers including Call-Net, Lightel, fonorola, Group Telecom and others have at-
tempted to enter the business, but were not successful and were mostly consolidated under
Call-Net, a subsidiary of Sprint Canada. Sprint Canada provided both competitive long-dis-
tance services and local service to some 200,000 homes as a CLEC. Sprint Canada was ac-
quired by Rogers Communications Inc.
360networks attempted to provide bulk fiber-based services, but went bankrupt and sold most
of their assets to Bell Canada.
4.3.6 Wireless
The main wireless carriers are affiliates of the phone companies (“wireline” carriers), including
TELUS and Bell, as well as Rogers Communications Inc., Canada’s biggest cable company.

4.4 PSTN Switching Center Hierarchy

We have discussed loops and POTS at some length; now we will examine the trunk side of the
PSTN. Figure 37 illustrates the five-level hierarchy of switching centers in the Bell System.

64
 Figure 37. Five Classes of Switching Centers

4.4.1 Class 5: Central Office

At the bottom level is the End Office, End Serving Office, Serving Office, Class 5 Office, Num-
ber 5 Office or Central Office (CO).
In all cases, these words refer to the building that contains the switch to which your telephone
set is connected with a loop, or with a loop carrier system if you are connected to a remote
switch.
It is owned by the ILEC.
4.4.2 Wire Center
The CO is also called a wire center because it is the physical point where access wires con-
verge for connection to the switching network and network core. Many cables carrying many
wires and fibers arrive at a cable vault in the basement of the CO.
These are carried on risers up to a room in the center of the building where individual copper
pairs and fibers from the street are patched to individual copper pairs and fibers in horizontal
in-building cables that lead to the switches, routers or muxes and their optical transceiver ports
or line cards.
Most of the cables leaving the building lead towards customers. These cables contain the cop-
per subscriber loops and fiber access circuits.

65
Some of these cables contain fibers connecting to other COs, and fibers to the Toll Center for
interconnect with Inter-Exchange Carriers as described in the following sections.
4.4.3 Local Calls
When you place a call to your neighbor, the neighbor’s loop is usually connected into the same
CO as you, and so the call is handled within the Central Office.
If you place a call across town, your call will be routed across trunk circuits to another CO
switch and then on to the far-end loop. The Central Office is level 5 in this hierarchy, so CO
switches are called class 5 switches.
4.4.4 Class 4: Toll Center
Directly connecting the thousands of Central Offices together would not be possible due to the
enormous numbers of connections that would be required. A hierarchy was needed.
The general idea is that each metropolitan area has a building called a toll center, containing
switches to which all of the CO switches in that city are attached.
To make a phone call to another city, your call is routed from your CO to your city’s toll center,
and on the far-end city’s toll center, to the far-end CO and then the far-end loop.
This arrangement is sometimes called a tandem arrangement, and the switches in toll centers
called tandem switches or toll switches. The toll center is level 4 in the hierarchy, so toll
switches are called class 4 switches.
4.4.5 Class 1, 2 and 3 Switching Centers
Because the Bell System was so large, there are more levels in the hierarchy. Each state had
a primary center (class 3) to which all of the toll centers in that state were homed. The country
was always divided in seven sections, and all of the primary centers in each section were
homed to a sectional center (class 2). Sectional centers were connected via regional centers
(class 1).
4.4.6 High Usage Trunks
In practice, connections are installed between switching centers where traffic warrants. If there
is high traffic between two COs not homing on the same Toll Center, then a High Usage Trunk
might be installed directly between those COs.
This practice moves the actual implementation of the network from the strict hierarchical model
shown to more of a meshed network, with many different paths between switching centers.
Traditionally, these trunk circuits have been carried as reserved channels on SONET fiber-op-
tic transmission systems organized in ring patterns around town, around the region and around
the country.
Going forward, phone calls will be carried when needed in IP packets in Ethernet frames on
fiber optics.

4.5 Switched Access to LD Competitors: LECs, POPs

and IXCs
One of the main purposes of the breakup of the Bell System in the USA in 1984 was to intro-
duce competition for long-distance phone calls and data services.
In this model, the access network and the long-distance networks were owned by separate
companies who had to work together to make end-to-end connections.
4.5.1 Access Network, LECs and ILECs
The access network is also called the last mile (or the first mile if you are in that part of the
business). It is the physical cabling from the customer premise to the CO, the CO, the physical
cabling to the toll center and most of the toll center.
The company that owns this is called a Local Exchange Carrier (LEC).
With the advent of local competition, new jargon was added; these companies are now re-
ferred to as the Incumbent Local Exchange Carrier (ILEC), meaning that they were there first.

66
4.5.2 Long Distance: IXCs
The long-distance networks connect toll centers in different cities. These long lines are owned
and operated by Inter-Exchange Carriers (IXCs).
A facilities-based IXC is one that for the most part owns their own physical transmission facili-
ties, typically fiber-optic cables and equipment.
A reseller leases capacity from a facilities-based carrier to form a network. Both of these com-
panies are Inter-Exchange Carriers.
Due to the pricing structure at the time, many reseller-type IXCs sprung into business once
competition was introduced, leasing high-capacity services from IXCs, and signing up cus-
tomers who route their calls over the high-capacity leased service… buying wholesale and sell-
ing retail.
With a subsequent drop in retail prices, the profit margins for resellers were dramatically re-
duced, to the point where many went out of business, or transitioned into more facilities-based
operations.
4.5.3 Switched Access
The IXCs have equal access to the ILEC’s last mile; or, the other way around, the ILEC is re-
quired to provide their customers equal access to competitive IXCs.
The customer of the ILEC can select any IXC, and the ILEC will connect the customer to that
IXC in the toll center on a call-by-call basis. This is called switched access.
The ILEC at each end bills the IXC a per-minute switched access charge for that last mile con-
nection.
4.5.4 POP: Point of Presence
The termination of the IXC’s cabling in the toll center was called a Point of Presence (POP).
This term originated when different court decisions, regulations and agreements forced the Bell
System to provide physical space in the toll center for a competitor like MCI to terminate fibers
and house equipment.

Figure 38. Switched Access to IXC POPs

This room was MCI’s presence in the toll center; everyone would refer to it as “the MCI equip-
ment room”. From the ILEC’s point of view, it is the beginning of the IXC’s network. From the
IXC’s point of view, it is the end of their physical circuits.

67
The term POP has now moved into general usage, to mean a building where a competitive
carrier terminates at least two fiber-optic cables; a station on a regional ring.
This building is today often not the toll center, but a different building across the street or
across town connected to the toll center with fiber.
4.5.5 Equal Access and PIC Codes
Equal access means that a customer can select in advance the Inter-Exchange Carrier that will
handle that customer’s long distance, and the routing through the toll center and the POP on to
the IXC’s network is transparent to the customer.
This is implemented with an entry in a database maintained by the ILEC called the customer’s
Preferred Inter-Exchange Carrier (PIC) code.
Each IXC has a Carrier Identification Code. AT&T’s carrier code is 0288 (ATT on a telephone
keypad). MCI is 0222, Sprint is 0333, Global Crossing is 0444. There are many others.
When a customer of the ILEC changes long-distance companies, the customer’s PIC code is
changed to the carrier code of the new IXC.
When making a call from someone else’s telephone, such as a payphone at an airport, it is
possible to manually route the call through a particular IXC on a call-by-call basis by dialing
101 then the carrier code.
Dialing 1010288 connects to AT&T’s POP; 1010222 connects to MCI, 1010333 to Sprint and
so forth. Many of these companies also have 1-800 numbers that accomplish the same thing.
In 1984 in the USA, the ILEC and IXCs were strictly separate companies. Following changes in
law and regulation, today typically the holding company that owns the ILEC also owns one of
the IXCs.
This was always the case in Canada, where the phone companies were not split into separate
local and long-distance companies; but equal access to competitive IXCs was ordered in 1992.

4.6 High-Capacity Access to Long-Distance

Another part of long-distance competition is providing not just one long-distance phone call at
a time, but providing high-capacity long-distance voice and data services between locations of
large businesses and government.
One typical business case is a big insurance company with regional offices, the head office
and their hot-site backup in cities geographically dispersed. The insurance company requires
high-quality, high-capacity point-to-point services to connect these buildings.
In 1984, this meant “T1s”, or more accurately, 1.5 Mb/s DS1-rate services. Today, “high-capac-
ity” might mean 10 Mb/s or 10 Gb/s on fiber, organized as IP packets.
The carrier that the insurance company has selected has already built a network connecting
the cities together: they have leased buildings in the cities, calling them data centers or POPs,
as per the previous discussion. They have connected these POPs with redundant point-to-
point high capacity fiber-optic transmission facilities or leased services. This forms the core of
the competitive carrier’s network.
4.6.1 Dedicated Line from the ILEC In-City
The question: how to connect the “last mile”, that is, the insurance company’s building to the
POP in each city for this kind of non-switched, high-capacity service?
The simplest solution for the competitive carrier is to order a point-to-point dedicated line ser-
vice from the ILEC to connect the insurance company building to the competitor’s POP in each
city.
Essentially, the IXC is subcontracting the last mile to the ILEC.
The ILEC already has easements allowing them to install wiring along (or under) streets and
the personnel, trucks and management systems in place to do this.
The installation and recurring monthly cost of the ILEC access circuit has to be added to the
bill that the long-distance company sends to their customer. In some cases, the cost of access
this way is prohibitively high.
4.6.2 Tariffs

68
In an unregulated world, the ILEC might refuse to provide such point-to-point dedicated line
services to a competitor of the ILEC’s own long-distance services, to prevent the competitor
from providing long-distance service to the insurance company.
Fortunately for the competitor, the ILEC has filed tariffs for point-to-point “dedicated line” or
more correctly, full period services in that city.
A tariff is a legal document that specifies the service level (bit rate, delay, maximum number of
bits in error etc.), the cost for a service and the business terms under which it is provided.
Once a tariff is accepted by the ILEC’s regulator, the ILEC must provide the service to anyone
who asks, including their competitor.
If the competitive carrier has a number of customers around a particular CO (for example, the
main CO downtown in a big city), then they could order point-to-point tariffed services connect-
ing each customer to the CO, then a higher-capacity point-to-point tariffed service aggregating
the traffic between the CO and their POP.

Figure 39. Dedicated Line From The ILEC For Access

4.6.3 Advantages and Disadvantages

The advantage of this to the competitive service provider is that it does not require any invest-
ment in capital, personnel or operations systems for physical access wiring – they simply sub-
contract it.
The disadvantages having the ILEC provide a point-to-point full-period service between the
customer and the competitive carrier’s POP are:
1) The competitor has to pay tariffed rates to the ILEC, which is very costly and may make
the competitor’s pricing uncompetitive,
2) The ILEC determines the available services, technology, bit rates and quality that are
available; the competitor has no control, and
3) There might be “technical difficulties” in the connection between the ILEC and competitor
that would be avoided if the ILEC were not involved.

4.7 CLEC: Collocations and Dark Fiber

4.7.1 Unbundling
Paying the ILEC for tariffed switched access or dedicated line service to connect the last mile
from a competitive carrier’s POP to the competitive carrier’s customer started in 1984.

69
Subsequent legislation and regulatory decisions unbundled the ILEC’s physical access net-
work from the ILEC’s services provided on that network.
This enables competitive carriers to lease just the ILEC’s physical cabling to the customer in-
stead of paying for a tariffed service from the ILEC on the same cable.
The regulatory rationale is that the ILEC built the wire center and access network when they
were a monopoly. The community was obliged to pay the ILEC to build the access wiring.
Therefore, in a way of thinking, the community has a degree of ownership of the resulting
physical access wiring... and therefore, the community has the right to use it without being
obliged to also have the ILEC provide them services over it.
4.7.2 Dark Fiber and Dry Copper
If the ILEC is providing copper wires without electricity on them, i.e. not attached to a CO
switch line card, this is called a dry circuit. A fiber not attached to anything is called dark fiber.
4.7.3 Competitive Local Exchange Carrier (CLEC)
The next more sophisticated solution to connect a competitive carrier’s customer to their POP
in the same city is to rent two fibers from the ILEC: one from the CO to the customer, and one
from the CO to the POP.
The competitive carrier must also rent space inside the ILEC’s CO - at the wire center closest
to the customer, where they can locate equipment to operate the fibers.
The competitive carrier installs fiber-terminating equipment at the customer, in the ILEC’s CO
and at their POP, then connects the POP to the CO with one ILEC fiber and the CO to the cus-
tomer with another.
An organization doing this is said to be a Competitive Local Exchange Carrier (CLEC).
It should be noted that large carriers use this technique for the last mile in areas where they
are not the ILEC.
In this case, it is not really appropriate to call the large carrier “a CLEC”, as this is a very small
part of their business. It is more correct to say that the large carrier “has collocations”.
4.7.4 Collocations
In addition to being required to lease dark fiber to their competitor, the ILEC is required to build
collocation facilities in its COs.
Collocation facilities are rooms in the CO, often with separate entrance doors, where the com-
petitive carrier can locate their own equipment.
The competitive carrier thus gains access to the wire center: the termination point for the fibers
and copper wires in cables leading out to the street and ultimately to the customer premise.
In the collocation facility, the competitive carrier places or collocates network equipment like
Optical Ethernet switches and routers.
The ILEC’s dark fiber is connected to optical transceivers in the collocated equipment in the
CO at one end and at the customer premise at the other end, implementing a “last mile” con-
nection over the ILEC’s fiber.
4.7.5 Advantages
There are two key benefits of leasing dark fiber and collocations compared to tariffed services
from the ILEC for the last mile:
1) Cost: leasing a dark fiber is a tiny (!) fraction of the cost of paying for a 10 Gb/s dedicated
service from the phone company.
2) Performance: The competitive carrier is now in control of the optical technology: the optical
transceiver technology, WDM strategy, the manufacturer, the bit rates, transmission character-
istics and so forth.
4.7.6 Disadvantages
This does not happen with a phone call, as does a service from the phone company.
This requires the competitive carrier to have Engineers to design the system, select the equip-
ment and determine how it should be installed and configured. This is a never-ending process,

70
as newer, better products and technologies are constantly becoming available.
Another disadvantage is that there may not be any dark fiber available from the ILEC, and no
time frame for new cable construction in that area.
The CLEC does not have any control over the physical connection and how it is provisioned,
maintained and especially repaired.
The CLEC is relying on their competitor to provide and maintain the physical fiber, and to re-
pair the fiber and restore the service in a timely fashion after the fiber is cut.
4.7.7 Application
Collocations would be implemented when there is enough business around a particular CO to
justify it.
Collocations and dark fiber can be used to provide 10 Mb/s to 10 Gb/s Optical Ethernet ser-
vices over fiber. In theory, they can also be used to provide DSL and even POTS and T1/PRI
services on twisted pair copper wires to a customer.
For the POP to CO portion of the physical connection, sometimes called the backhaul, the
same technique of leasing a dark fiber from the ILEC between the POP and collocation may be
used.

4.8 Bypass
While collocations and dark fiber from the ILEC is a way to provide high bitrate service to a
customer, it is not the final answer.
With dark fiber, the competitive carrier is relying on their competitor to provide a mission-critical
piece of the service.
As soon as there is sufficient revenue in a particular area of a city, the competitive carrier will
install their own fiber from the POP to the collocations and to the customers, bypassing the
ILEC altogether.
This eliminates the uncertainty of relying on the competitor, as well as increasing efficiency by
eliminating the never-ending service ordering and billing interactions with the ILEC.
4.8.1 Easements
Physically installing fiber across a city requires easements from landowners, which convey the
right to place and leave cabling on the property.
The biggest landowner in a city is usually the municipal government, since they own all of
streets. Electrical companies, railroads, pipeline companies, CATV companies, regional, state
and federal governments, building owners and private bridge owners all sell easements.
An alternative way of bypassing the ILEC is to use point-to-point microwave radio between
buildings. High-capacity radio systems require a frequency license from the government to op-
erate. Other point-to-point systems can operate on unlicensed bands, eliminating the need for
a license at the cost of much higher interference and thus lower bit rates.

4.9 Competitive Carrier Network Model: Rings, POPs

and MANs
Competition today means much more than the 1984 idea of LECs providing switched access
or dedicated line access to competitive IXCs.

71
 Figure 40. Competitive Carrier Network Model
Competition today includes competitors providing various services to residences and busi-
nesses using a mix of collocation, bypass, switched access and dedicated line access.
A model for a competitive carrier’s network, depicted in Figure 41, includes POPs in cities con-
nected together to form a regional inter-city “long distance” backbone, plus one or more Metro-
politan Area Networks (MANs) built out from the POP in each city.
In-city, these MANs connect the POP to locations like the ILEC’s toll center, collocations in the
ILEC’s COs, telecom rooms in large multi-tenant office buildings and apartment buildings, and
directly to big customers.
4.9.1 Fiber Rings
To ensure high availability, redundant connections are required to provide protection against
cut lines. This means at least two fibers, in different cables, in different ducts under different
streets between each node.
It turns out that the cheapest way to achieve this objective is to connect the nodes neighbor-to-
neighbor-to-neighbor to form what looks like a ring.
This strategy means that there are two cables to each node, but does not require twice as
many cables; only one extra cable is required, so many fiber networks are built in ring patterns.
There are several strategies for implementing cut line detection and automated service
restoration on a fiber ring.
The competitive carrier’s POPs in cities are connected to form regional rings, which are inter-
connected at multiple places to implement national communications.
The competitive carrier will install or lease fiber in-city to connect their POP to the toll center,
COs and large customer buildings. These locations are connected neighbor-to-neighbor in a
ring called a MAN.

Figure 41. Collocation of Equipment at the Wire Center

72
5
Digital
5.1 Analog and Digital: What do we really mean?
The words “analog” and “digital” are often used with little regard for their actual meaning. It is
useful to review the definitions of these terms to better understand the concept of a digital
communication circuit.

Figure 42. Analog Signals and Digital Signals

5.1.1 Analog Signal

“Analog” means “representation”. An “analog signal” is a signal that represents another signal.
For example, the voltage on a telephone loop represents the sound pressure waves coming
out of the speaker’s throat and so is called an analog signal.
5.1.2 Analog Circuit
The term “analog circuit” is inaccurate, but short and catchy-sounding. What is really meant is
“circuit capable of carrying an analog signal”, and usually, “circuit capable of carrying a voltage
that is an analog of the sound pressure waves coming out of a person’s throat”.
5.1.3 Digital Signal
A “digital signal” is information in a numeric format. The information could have any source.
Examples are voice, video, images, text and sensor data. The information is coded into a bi-
nary format - 1s and 0s - using some coding scheme.
5.1.4 Digital Circuit
A “digital circuit” is a circuit capable of carrying a digital signal, i.e. a circuit specifically de-
signed to carry numbers from one place to another. In many cases, a digital circuit conveys a
digital signal by coding the digital signal into 1s and 0s, then using pulses to represent the 1s
and 0s.
On a copper-wire digital circuit like a LAN cable, a pulse is changing the voltage on the wires
to a fixed amount for a fixed period of time.
On a fiber-optic digital circuit, a pulse is turning a laser on for a fixed period of time. The laser
is producing infra-red light at as close to a single pure frequency as possible.
A 1 might be represented by a pulse, and a zero by the absence of a pulse.
Another way of representing a digital signal is to use modems. In this case, some characteris-
tic of a single pure frequency called a carrier is changed or modulated in some fixed ways to
represent 1s and 0s.

73
5.1.5 Bandwidth
For the purposes of this course, bandwidth means capacity.
The capacity of an analog circuit is measured as the width of the frequency band supported on
the circuit, called its frequency bandwidth. The units of measurement are changes per second
or Hertz (Hz).
Wideband and broadband are used to describe circuits that support a large frequency band-
width, like coaxial copper cable (“coax”), which can support 3 GHz of bandwidth.
Companies that offer POTS on twisted pair call the frequency range of their basic service offer-
ing the voiceband, which is 3 kHz wide. Companies that operate broadband coax networks re-
fer to the voiceband as narrowband.
The capacity of a digital circuit is also called its bandwidth, but is not measured in Hertz, but
rather measured in bits per second (b/s). A broadband digital circuit is one that communicates
many bits per second.

5.2 Continuous vs. Discrete Signals

To understand the relationship between analog and digital signals, it is useful to consider the
difference between continuous signals and discrete signals. These are the two types of sig-
nals.
5.2.1 Continuous Signals
Continuous signals are signals that vary continuously. A continuous signal can take on any
value, or any fraction within given limits.
Examples of continuous signals are easy to find, since the world is pretty much continuous:
length, width, height and time are all continuous.
A good test for a continuous signal is to see if it could possibly take on the value between two
other values. If the answer is always “yes”, no matter which two values are chosen, then the
signal is continuous.

Figure 43. Continuous Signal

Most analog signals are continuous.
A good example of a continuous analog signal is a thermometer. This is a tube of glass with
some mercury in it. The height of mercury in the tube is a continuous signal; it can take on any
level in the tube.
As the temperature goes up, so does the level of mercury in the tube. As the temperature goes
down, so does the level of mercury in the tube.
The height of mercury in the tube is an analog of the air temperature.
5.2.2 Discrete Signals
A discrete signal is a signal that can take on only specific values.
A good example of a discrete signal is the number of people in a room. There can be 14 peo-
ple in a room, or 15 people in a room. There is no such thing as 14½ people in a room.

74
 Figure 44. Discrete Signal
Notice that the number of people in a room is discrete in value, but continuous in time. There
are always a number of people in the room.
How would we turn this into a signal that is discrete in time? We would have to, on a regular
basis, count the number of people in the room and write the answer down on a piece of paper.
This is a good test to see whether a signal is discrete or not: can it be written down on a piece
of paper? Or stored on a disk?
Digital signals are discrete.

5.3 Voice Digitization (Analog-Digital Conversion)

We look in detail at the voice digitization process to derive the coding rate – the number of bits
per second – used in standard digitization of voice.
First we look at the process that happens on the speaker’s line card: converting an analog
voice signal to digital. There are three steps: quantization, sampling and coding.
5.3.1 Quantization
Quantization is the process of changing from a signal that is continuous in value to a signal
that is discrete in value.
This is accomplished by dividing the possible range of values into a number of “bins” or levels
or steps, and assigning a number to each of these levels.
In the example of Figure 45, the range of values is divided into six levels. Then, when asked
what the value of the signal is, we say that the signal is “in level #4” rather than trying to mea-
sure its exact voltage.
Another example of quantization is sugar cubes. Instead of putting some fractional value of a
teaspoon of sugar in your coffee, your choice is “one lump or two”. The sugar has been quan-
tized into uniform lumps.

Figure 45. Quantization, Sampling and Coding

75
5.3.2 Sampling
The second step is sampling. Sampling is the process of changing the signal from being con-
tinuous in time to one that is discrete in time.
This must be done on a regular basis, like clockwork.
In the example of Figure 45, the vertical lines indicate the times at which the samples are
taken. At each of these times, the signal is sampled, that is, the value of the signal is mea-
sured and recorded.
The value of the signal that is recorded is the level it is in at that time.
5.3.3 Coding
The third step is coding. The value of the signal taken at each sample (the level number) must
be coded into 1’s and 0’s so that it can be transmitted over a digital carrier system or stored in
a computer.
The objective is to transmit the codes representing the value of each sample to the far-end line
card where the reverse process is performed: reconstructing the analog waveform from the re-
ceived codes.
The whole point in doing this is to move the analog voice signal from the near-end loop to the
far-end loop over a digital transmission system without adding any noise – and allowing voice
to be carried on the same system as data, video and any other kind of information.

5.4 Voice Reconstruction (Digital→ Analog Conversion)

At the far end, the analog is reconstructed to represent speech on the loop leading to the
phone. For a residence, this process typically happens on a line card in a switch in a CO or re-
mote. In a VoIP system, this process happens inside the far-end telephone.
At the sender, sampling, quantization and coding are performed to generate codes represent-
ing the value of sequential samples of the analog. These are transmitted to the far end, where
the reverse process takes place.
5.4.1 Reconstruction
As each code (a binary number) arrives, it is decoded to yield the level or bin number that is
the value of the sample, and the voltage on the far-end loop is changed to the value equal to
that of the center of the level.
The voltage is changed smoothly from one level to the next.
In the case of a voiceband signal, the voltage would be increased or decreased at a rate so
that the result never vibrates more than 3300 times per second, i.e. limited to the voiceband’s
3300 Hz limit.
This smoothly-changing voltage is continuous in both time and value – the reconstruction of
the original analog.
5.4.2 Quantization Error
However, it is not reconstructed exactly. Due to the fact that only the level number is transmit-
ted – not the exact position within the level – the reconstructed voltage is always set to the
center of the level.

76
 Figure 46. Quantization Error
The difference between the center of the level and where the signal actually was is a small er-
ror introduced into the reconstructed signal, and is called the quantization error. The size of the
quantization error is directly related to the size of the levels.
To make the quantization error smaller on average, more levels can be defined, to make the
levels smaller.
The telephone company uses enough levels so that a human can’t hear the quantization error
noise.
5.4.3 Aliasing Error
If samples are not taken frequently enough, then not enough information will be transmitted so
that when the “dots” are connected at the far end, the reconstructed signal is faithful to the
original signal.
This is called an aliasing error. It can be pictured by removing 2/3 of the samples in Figure 46,
then connecting the remaining dots… the result is close to a flat line and does not resemble
the original analog.
Harry Nyquist, who obtained a Ph.D. from Yale in 1917 and worked his entire career for AT&T
and Bell Labs, discovered that it is necessary to take samples more than twice as often as the
frequency bandwidth of the signal to avoid aliasing errors.
This theorem was published in his 1928 paper “Certain Topics in Telegraph Transmission The-
ory“, and is known today as the Nyquist sampling theorem. It determines the number of sam-
ples per second.

5.5 Voice Digitization: 64 kb/s G.711 Standard

There are three steps in voice digitization: quantization, sampling and coding

77
 Figure 47. 64 kb/s Voice Digitization

5.5.1 256 Quantization Levels.

The telephone system quantizes the voice signal to 256 levels. This number is chosen to re-
duce the quantization error, which would be heard as noise after the signal is reconstructed, so
that a person can’t hear it on the line. The diagram shows level numbers 127 and 128 around
zero volts.
5.5.2 8,000 Samples per Second
The second step is sampling. Since this is a voiceband signal, the frequency bandwidth is
3000 Hz, and so the sampling rate must be at least 6001 times per second, following Dr.
Nyquist’s sampling theorem.
To ensure that there are no aliasing errors, the telephone system samples more often: 8,000
samples per second.
5.5.3 8-bit Coding
The third step is coding. Traditional telephony uses 8 bits (1 byte) to code the value of each
sample.
This technique of using 8 bits per sample is sometimes called Pulse Code Modulation (PCM),
a term originally used to describe the entire voice digitization process.
5.5.4 64 kb/s G.711 Codec Standard
To determine the number of bits per second required, multiply the number of samples per sec-
ond (8,000) by the number of bits per sample (8) to get 64,000 bits per second, or 64 kb/s for
short.
This was standardized by the ITU as codec standard G.711.
5.5.5 64 kb/s DS0 Channels
This 64 kb/s rate was called Service Level Zero in the Digital Hierarchy by Bell Labs, and ab-
breviated DS0 (Digital Service Level Zero).
In the previous millennium, transmission systems were built to carry streams of voice digitized
at 64 kb/s in channels.
A channel is reserved time slots on a transmission system, a fixed number of bits per second,
an unvarying fraction of the overall capacity.
Users, each connected to a channel, take turns using the transmission system, for a specific
length of time, one after another, in a strict order that repeats over and over, 8,000 times per
second.

78
When someone talks about a channel on a digital transmission system, they usually mean a
DS0, 64 kb/s.
These channels could be aggregated into higher bit rate channels, notably DS3 at 45 Mb/s.
This is covered in Chapter 6, “Transmission Systems”.
5.5.6 64 kb/s Packetized Voice
Going forward, voice is carried in systems originally designed for data.
For efficiency, these systems do not divide the transmission capacity into fixed-size channels,
where all users take turns one after another in a strict order, but instead make it first-come,
first-served, one packet at a time.
Every destination on the network must have a network address. On a telephone network, it’s
called the phone number. On a postal network, it’s called the mailing address. On packet net-
works, it is the IP address. IP is a standard way of packet addressing that everyone has
agreed to use.
Users create IP packets by breaking their transmissions into small chunks, perhaps 1500
bytes, and pasting the IP address of the desired destination on the front.
The user then transmits the packet to a router, which relays it onward to another router, on and
on until it reaches the indicated destination.
At the receiver, the chunks of data are extracted from the packets and put together to recon-
struct the original transmission.
The IP packets contain anything… including segments of digitized voice conforming to the 64
kb/s G.711 standard.
There are more efficient coding schemes for voice, sometimes called voice compression, but
they are only used when there are bandwidth limitations. Most of the time, G.711 is employed
to avoid compatibility problems.

☞ Regardless of whether the bits are communicated in a channel or in a packet, the bot-
tom line is that a byte, representing the value of the sample, is transmitted 8,000 times
   per second to communicate digitized voice when following the near-universal G.711
standard.

5.5.7 μ-law and a-law

For advanced readers: Figure 47 shows level numbers 127 and 128 around zero volts for clar-
ity. In practice, it might be level numbers 0 and -1 around zero volts.
The figure also shows the levels all being of the same size, which is not completely accurate.
In practice, the levels will be smaller close to zero volts and become exponentially larger as the
value increases.
This technique further reduces quantization noise by taking advantage of the fact that statisti-
cally, the voltage will be around zero most of the time; so the levels are finer around zero volts.
Two standards for this progressive level size are μ-law used in North America and a-law used
in most of the rest of the world.

5.6 Digital Video, H.264 and MPEG4

5.6.1 Digital Video Cameras
A digital video camera has a lens that focuses received light on a small array of light detectors,
also called photodetectors.
The light detectors are usually arranged in groups of three: one sensitive to red, one green,
and one blue. There is one triad of detectors per pixel.
The value of each light detector is sampled at the refresh rate, often between 50 and 100 times
per second.
The value of each sample is coded into a byte. 0 corresponds to no light detected, 255 is maxi-
mum light detected, with 253 shades between the two.
The number of pixels, that is, the number triads of light detectors in the camera’s detector ar-
ray is called the picture definition. High-Definition (HD) pictures are 1280 pixels across x 720

79
down, a total of 921,600 pixels.
921,600 pixels, with three color bytes per pixel, refreshed 60 times per second is 1.3 Gigabits
per second. This is the output of the digitizer.
“A picture is worth a thousand words” is a well-known saying... but compared to 64 kb/s voice,
raw HD video is worth over 20,000 words!
This 1.3 Gb/s bit rate is lowered for storage, transmission and display using mathematical
compression techniques. Standard methods of compression are called codecs. MPEG is a
popular choice.
5.6.2 Factors Affecting Video Quality
A number of factors affect the perceived or subjective quality of the images on the far-end
user’s screen. Aside from network issues like transmission error rate and variability of delay,
the main factors are picture definition and refresh rate, the number of bits per second required
and the number of processing operations per second that must be performed to implement the
compression and decompression in real time.
The objective is to transmit high-definition images using a low number of bits per second while
achieving reconstructed picture quality people will be willing to use.
However, these factors are often in conflict: for example, high compression requires intensive
processing, and large picture size means a higher number of bits per second. It is one thing to
optimize two of the three factors; it is another to optimize all three at the same time.

Figure 48. Digital Video Tradeoffs

The reconstructed picture quality is measured by having people watch it and give their subjec-
tive rating. It’s unclear how objective measurements like signal to noise ratio relate to what a
person thinks a picture looks like.
5.6.3 Definition vs. Resolution
Definition, measured in pixels, is the correct terminology for picture size. The term “resolution”
refers to the quality of the image after reconstruction. A TV is High Definition (HD), not High
Resolution (HR).
However, calling the definition of a display its “resolution” is such a common error that the
terms are becoming interchangeable.
5.6.4 Standard Definition, Interlaced and 480i
Standard Definition (SD) in North America is 720 pixels across x 480 pixels down, refreshed 30
times per second.
Human brains have a peak in motion detection at about 15 Hz. Refreshing the picture at 25 or
30 Hz helps minimize the sensation of objects juddering, instead appearing to the brain to

80
move smoothly across the screen.
Early developers of television took advantage of the persisting glow of phosphorous on their
picture-tube displays to increase the apparent resolution of a display at no cost by doing the
refresh in two passes, every odd-numbered line then every even-numbered line.
Each half-picture is called a field. Fields are transmitted 60 times per second, leading to the
designation 60 Hz interlaced. Two fields make a frame.
SD in North America is abbreviated as 480i.
In the rest of the world, the number of lines per second is the same, but the definition is higher
at 720 x 576 since the screen is only refreshed 25 times per second. This is abbreviated 576i.
Videophones and desktop videoconferencing systems in the past supported the Common In-
terface Format (CIF) at 352 x 258 pixels.
5.6.5 High Definition, Progressive and 720p
When the screen refresh is done in one pass, it is referred to as progressive and abbreviated
with a p, for example, 480p.
The first step beyond standard definition is High Definition (HD), which is 1280 x 720 progres-
sive and referred to as 720p, and refreshed 50 or 60 times per second.
5.6.6 Full HD 1080 and 2K
Next is 1920 x 1080, interlaced or progressive: 1080i or 1080p.
This is called HD, Full HD, True HD, and in some cases 2K since there are approximately
2000 pixels horizontal definition in the consumer formats.
Advanced readers may want to note that abbreviations with a “K” are also used to describe
studio formats, which are slightly different than the consumer formats.
5.6.7 Ultra HD and 4K
More recently, the consumer format 3840x2160 has been marketed. It is referred to as Ultra
HD, Quad Full HD and 4K.
The usual application for higher definition (more pixels) is bigger screens, not more detailed
pictures on a small screen.
In the future, displays will have 4,000,000 x 3,000,000 pixels and will occupy entire living room
walls. The marketing department will probably call them “4M” displays.
5.6.8 Compression
Compression is required to store and transmit these images. Without compression, 720 x 480
at 30 Hz, with one byte each for red, green and blue is 250 Mb/s. 1280 x 720 at 60 Hz is 1.3
Gb/s. 1920x1080 is 3 Gb/s.
Compression is performed by an algorithm called a coder/decoder or codec, either on a spe-
cial-purpose integrated circuit chip, or on the shared main processor in a computer.
Video compression is lossy compression, meaning the reconstructed image is not exactly the
same as the original.
To operate in real time (at playing speed), codecs are usually implemented as highly optimized
machine code on custom-built chips containing multiple Digital Signal Processors (DSPs).
5.6.9 MPEG
Standards are required for interoperability. The Moving Picture Experts Group (MPEG) and the
ITU establish standards in this area.
MPEG-1 was for video on CDs, with the video coded at 1.15 Mb/s.
This was replaced with MPEG-2, which offers a wide range of coding and compression op-
tions, grouped in profiles. Each profile supports a certain picture size, definition, refresh rate
and image quality, and results in a different average bit rate, typically 1 to 3 Mb/s.
MPEG-2 is used as the basis for SD video stored on Digital Versatile Disks (DVDs) and trans-
mitted via cable, satellite and IPTV video services.
5.6.10 MPEG-4 and H.264

81
Part 10 of the MPEG-4 standard specifies the use of the H.264 codec standardized by the ITU.
H.264 provides the same quality of reconstructed signal as MPEG-2 for 1/3 the bit rate with
better error tolerance.
H.264 is used for coding HD video for Blu-ray DVDs, and HD channels delivered by cable,
satellite, Internet and IPTV video services. HD video is typically coded at 6 Mb/s for broadcast.
Much lower rates are used for Internet video.

5.7 Digital Transmission: Binary Pulses

“Digital” transmission means applying energy to the communication circuit, or not, for a specific
length of time, to represent a one or a zero.

Figure 49. Analog Transmission vs. Digital Transmission

This short period of energy applied to the line is called a pulse. A particular strategy for repre-
senting 1s and 0s using pulses is called a digital line code.
Most wired digital transmission systems are two-state systems – the laser is either on or off; ei-
ther there is voltage on the line, or there is not – and so are actually binary systems.
There is a very significant advantage in terms of noise performance using a binary signaling
scheme compared to using a continuous analog signal.
5.7.1 Analog: Attenuation, Added Noise and Amplifiers
Illustrated at the top of Figure 49 is a continuous analog signal on copper wires. The signal di-
minishes in value with distance away from the transmitter due to resistance of the copper to
the flow of electricity through it. This is called attenuation.
An amplifier is used to boost the signal after a given distance (usually 18,000 feet for POTS).
The amplifier simply multiplies the value on the input by a pre-determined factor like 30.
The problem is that the copper wires act like antennas. Electro-Magnetic Interference (EMI)
causes electricity to be induced on the wires, which adds to the signal.
When the amplifier boosts the signal, it also boosts the noise, resulting in difficulty in faithfully
reconstructing the original signal.
5.7.2 Digital: Pulses and Repeaters
The lower part of Figure 49 illustrates implementing digital with pulses.

82
Pulses of electricity on copper wires will also be attenuated with distance and will also have
noise added. The square corners will also be rounded off due to capacitance of the wires.
However, instead of boosting the pulse with an amplifier, a repeater is used. Repeaters do not
boost the incoming signal: repeaters are binary devices that make a decision.
If a repeater decides it detects an incoming pulse, however degraded, it regenerates a clean
copy of the signal, a new noiseless square pulse on its output, to be transmitted on the next
cable segment.
5.7.3 Repeaters on Copper Wires
Repeaters are required every mile or two on copper wire systems to be able to regenerate the
pulses while they are still detectable without errors.
For example, repeater spacing is every 6,000 feet on T1 and every 12,000 feet on High-Speed
Digital Subscriber Line (HDSL), a 1.5 Mb/s technology not related to residential DSL.
5.7.4 Repeaters on Fiber
Repeaters are required on fiber optic systems at 40 to 80 km (or more on special systems like
transoceanic cables) due to a different mechanism that degrades the pulses called dispersion.
This is the lengthening of the duration of the pulse caused by the light following different
bounce paths, called modes, inside the fiber.
Since some paths have a longer length than others, the light following those paths takes longer
to arrive at the far end, so the pulse becomes longer in time. The pulses must be regenerated
before they overlap.
5.7.5 Comfort Noise Generation
For digital transmission, voice is coded into 1s and 0s, which are represented as pulses. The
pulses can be reliably detected and regenerated as needed, allowing the communication of the
1s and 0s with a very low error rate.
This allows the eventual reconstruction of the signal at the far end with no added noise.
Sprint’s advertising tag line was “so quiet, you can hear a pin drop”.
In fact, when this was first rolled out in the Bell System beginning in the 1960s and 70s, sub-
scribers complained that it was too quiet.
People used to hearing hissing and humming on old analog trunks, didn’t like the silence, say-
ing it was difficult to tell if the call was connected.
To address this user issue, the G.711 codec standard includes Comfort Noise Generation,
where the codec adds noise to the signal so users are comfortable and don’t complain that it is
too quiet.

83
6
Transmission Systems
From about 1960 to 2000, the telecom network was constructed of high capacity transmission
systems shared amongst users by employing Channelized Time-Division Multiplexing (TDM),
also known as Synchronous Time-Division Multiplexing.
These systems are now referred to as legacy systems, meaning left over from a previous era...
but that does not mean they have disappeared. Telephone companies tend to keep existing
systems running for as long as possible.
Additionally, both GSM cellular and Passive Optical Networks implement channelized TDM, so
knowledge of channelized TDM remains part of the core knowledge set required in the telecom
business today.
In the first half of this chapter, we cover the installed base of channelized TDM transmission
systems with the DS0-based hierarchy of bit rates and the technologies T1, SONET and ISDN.
Then we begin understanding new-generation transmission systems, which share their capac-
ity amongst users with Statistical Time-Division Multiplexing, also known as bandwidth on de-
mand.
In the second half of this chapter, we cover the fundamental ideas of statistical multiplexing
and bandwidth on demand, including the critical concept of overbooking.
Subsequently, Chapter 8 and Chapters 11 through 16 cover the technologies used to imple-
ment the new-generation transmission systems, including fiber optics, fundamentals of frames
and packets, Ethernet and MAC frames, IP packets and IP addressing, and MPLS.

6.1 Channelized Time Division Multiplexing (TDM)

Synchronous or channelized TDM was first developed in 1874 – 75 by Emile Baudot, an engi-
neer at the French Telecommunications Service, for transmitting multiple telegrams on a single
telegraph circuit.
Multiplexing means combining the traffic from multiple lower-speed circuits onto a single
shared high-speed or aggregate circuit. Demultiplexing is breaking out the lower-speed circuits
from the high-speed aggregate circuit at the other end.
Time division means that the sharing of the high-speed circuit is in time.
All transmission systems work in both directions.
6.1.1 Channels
A channel is a specific, non-varying fraction of the capacity of the high-speed circuit, a con-
stant number of bits per second. Channelizing a circuit means dividing its capacity into chan-
nels; smaller pipes in a bigger pipe.
Synchronous means that one byte is transmitted for each channel in a strict order that repeats
over and over.
Since these transmission systems were designed to carry voice, one byte is transmitted for
each channel 8,000 times per second, as derived in Section 5.5 “Voice Digitization: 64 kb/s
G.711 Standard”.
As we will see later in this chapter, one byte 8,000 times per second is equal to 64 kb/s, and is
called the Digital Service Level Zero or DS0 rate.
6.1.2 Example: Time-Share Condos
A familiar example of channelized Time Division Multiplexing is time-share condos.
There is one condominium and 52 users. Each user enjoys full use of the condo for one week.
At the end of their week, they have to pack up and then the next user moves in and enjoys full

84
use of the condo for the next week, and so on in a strict rotational order.
When the rotation is completed, it repeats, year after year.
If a user pays for a week, but does not actually show up to use it, the condo is nonetheless re-
served for that user and remains empty for that week.
The users are time-sharing the condo, each getting a fixed amount of capacity. In the case of
condos, the standard amount is one condo for one week per year.

Figure 50. Time-share condos implement synchronous time-division multiplexing.

In the case of transmission systems, the standard amount is one byte 8,000 times per second,
to implement a channel size of 64 kb/s.
6.1.3 Trunk Carrier Systems
By digitizing the voice on trunks at the constant rate of 8,000 bytes per second, then inter-
spersing the bytes for multiple channels, channelized TDM became the standard method for
carrying multiple voice trunks over high-speed long-distance digital transmission circuits in the
second half of the 1900s. These were called trunk carrier systems.
Dividing transmission system capacity into 64 kb/s channels, the DS0 rate, and using a switch
to direct a user onto a channel for the duration of their communication session was a natural fit
for telephone calls digitized at a constant bit rate of 64 kb/s.
6.1.4 Inefficient for Data
However, this technology is not efficient for data communication.
A channelized system reserves capacity all the time for each channel, whether there is any-
thing to transmit or not.
Since data happens in bursts with relatively long periods of no activity between bursts, data
communications on channels is highly inefficient.
Sharing in time, but on an as-needed basis is more efficient. Users only transmit when they ac-
tually have data to communicate. If a user has nothing to transmit, another user can use that
capacity on the high-speed circuit instead. Typically, users transmit IP packets as needed.
This is called statistical multiplexing and bandwidth on demand, and is covered beginning in
Section 6.7.
Communicating voice using this technique – Voice over IP – allows the use of a single network
service for all applications, one of the long-sought goals in the telecommunications business.
In the future, there will be no channelized systems; all communications systems will be packet-
based bandwidth on demand.

6.2 Multiplexers
To implement channels, a multiplexer is attached to each end of a circuit.
On one side of the multiplexer are the users’ lower-speed access circuits, each on a separate
hardware port. On the other is the high-speed aggregate port. The multiplexer intersperses the
users’ data in a strict order to form a high-speed stream that is transmitted on the aggregate
port.
What goes in on a particular hardware port at one end comes out on the corresponding hard-
ware port at the other end.

85
Each user gets a fixed fraction of the capacity of the high-speed circuit to carry the data on
their lower-speed circuit from one building to another. This capacity is their channel.
The multiplexer implements the channels. Telephone switches and routers connected at each
end of each channel direct traffic onto the channels.
Of course, multiplexers are built into telephone switches so that both the multiplexing and cir-
cuit-switching functions are in the same product.
6.2.1 Example: T1
To understand how this kind of channelized or synchronous multiplexing is accomplished, it is
useful to consider an example technology, Trunk Carrier System 1, or T1 for short.
T1 is a technology popular from 1960 – 2000, designed to carry 24 trunks over 4 copper wires
using channelized TDM. Though fiber is now routinely used, there remain thousands of T1 cir-
cuits installed and in operation.

Figure 51. T1 Multiplexers and Circuit

A basic T1 system consists of multiplexers, Channel Service Units (CSUs) and the T1 circuit,
which is four copper wires with repeaters every mile or so.
Full details on T1 are included in Appendix A of this book… material that was in the main body
of the earliest editions of this book, now relegated to an appendix as T1 became old or
“legacy” technology.
A brief overview is provided in the following sections.
6.2.2 T1 Mux or Channel Bank
The multiplexer (one at each end) is variously referred to as a T1 multiplexer, a T1 mux, or a
channel bank. Since T1 was designed to carry 24 trunks, the mux provides 24 low-speed
ports, each running at the 64 kb/s DS0 rate.
6.2.3 Time Slots
To implement the channelized TDM, each port is allocated a fixed time slot to transmit a single
byte across the T1 circuit. This happens 8,000 times per second. The ports do this in a strict

86
rotational order, one after another. The resulting data rate is 24 x 64 kb/s = 1.536 Mb/s.
6.2.4 Framing Bits
To be able to sort out what goes where at the far end, the transmitting multiplexer sends an ex-
tra framing bit before the byte from the first port.
The receiving multiplexer uses this framing bit to identify the beginning of the byte for channel
one in the incoming bit stream, and direct that byte to low-speed output port number 1 on the
far side.
Then the next eight bits are directed to port 2, then the next to port 3 and so one until a byte for
each port has been received, then the process repeats.
Framing is covered in detail in Section 6.5.2.
6.2.5 DS1 Rate
The framing bit brings the bit rate to 1.544 Mb/s, the DS1 rate.
The entire system is two-way simultaneous.
6.2.6 CSU
The aggregate port on the multiplexer is connected to a CSU. The CSU is the circuit-terminat-
ing equipment for the T1 circuit.
This device represents binary digits on the physical wires using pulses of voltage on the cop-
per wires. It performs the same functions as a modem - but is not called a modem since it is a
digital device.
6.2.7 Repeaters
In this particular technology, repeaters are required to regenerate the voltage pulses every
6000 feet (6 kft / 1 mile / 1.6 km) along the T1 circuit.
6.2.8 Synchronization
All of the devices have to be synchronized at the bit level to know when a pulse of voltage
starts and ends.
In days past, all devices used a clock derived from the US National Bureau of Standards or
Canadian National Research Council cesium clock.
Today, clocks are derived from Global Positioning System (GPS) satellites.
6.2.9 Applications for T1
T1 was first used to carry long-distance trunks, then became an access technology for busi-
ness customers.
T1 was used to carry PBX trunks, used for ISDN PRI services, used to access Frame Relay
data services, and to implement private networks made of dedicated lines.
6.2.10 SONET TDM on Fiber
Synchronous Optical Network (SONET) technology, used for the network core from 1980-
2000, operates in the same way as T1.
SONET implements up to 129,024 DS0 channels by transmitting a byte 8,000 times per sec-
ond for each channel.
The resulting aggregate speed is measured in multiples of 45 Mb/s and is transmitted on fiber
using the Optical Carrier (OC) system.

6.3 The Digital Hierarchy: Legacy Channelized Trans-

mission Speeds
A legacy is something inherited from a previous generation. A legacy transmission system is
one that moves traffic in channels, which are bytes transmitted in fixed time slots on a high-bit-
rate system.
There is a huge installed base of legacy channelized transmission systems. In fact, a small –
but appreciated – fraction of a telephone company’s revenue is monthly recurring billing for ex-

87
pensive legacy circuits that are no longer being used... but the customer has never canceled
them.
These systems were designed to be voice trunk carrier systems, and so operate at multiples of
64 kb/s, the standard bit rate for digitized voice, referred to as Digital Service Level Zero or
DS0 for short.
To allow interoperability of systems, standardized multiples of DS0 channels were defined.
These standard multiples and the resulting line speeds are known as the digital hierarchy.
Equipment manufacturers made products operating at these standard bit rates. Telephone
companies purchased this equipment and integrated it to form networks operating at these bit
rates. Their marketing departments created products and services at these standardized bit
rates.
6.3.1 Kilo, Mega, Giga, Tera
Abbreviations are used to refer to data rates:
103 = thousand = kilo
kilobits per second (kb/s)
106 = million = Mega
Megabits per second (Mb/s)
109 = billion (US), thousand million (UK) = Giga
Gigabits per second (Gb/s)
1012 = trillion (US), billion (UK) = Tera
Terabits per second (Tb/s)

Figure 52. The Channelized Digital Hierarchy

6.3.2 DS0
Channelized digital transmission systems move the 64 kb/s DS0 rate for historical (voice) rea-
sons. Multiple DS0 channels are combined or aggregated into higher bit-rate streams for trans-
mission.
Anything below 64 kb/s is referred to as a subrate.
6.3.3 DS1 and E1

88
The first step above a DS0 is the DS1 rate. This rate is equal to 24 DS0s, or if you prefer, 24
times as fast as the DS0 rate.
Note that the mathematics does not quite work out. Multiplying 64 kb/s by 24 does not quite
equal 1.544 Mb/s. This is due in fact to some overhead added in by the T1 carrier system (the
framing bits).
There is also an E1 rate, used in Europe, which is 32 DS0s.
6.3.4 DS2
The next rate up is the DS2 rate. This rate is not interesting, and is hardly ever offered com-
mercially. It is the least common denominator between DS1 and E1, and was used as a step-
ping stone to the DS3 rate on old multiplexing systems.
6.3.5 DS3
The next rate of real interest is the DS3 rate. North American carriers’ legacy backbone trans-
mission systems operate at multiples of DS3 rates.
Both SONET fiber optic systems and point-to-point microwave radio systems were used to im-
plement n x DS3 circuits.
6.3.6 STM and SDH
In the rest of the world, transmission systems conformed to the European Synchronous Digital
Hierarchy (SDH), which moves Synchronous Transport Modules (STM).
STM is a frame size, which transmitted 8,000 times per second results in a data rate of about
155 Mb/s. This is also called an STS-3C in North America.

6.4 Digital Carrier Systems:

Legacy Transmission Technologies
This section provides an overview of widely-deployed technologies used to implement the
channelized digital hierarchy.
6.4.1 Technologies
The chart of Figure 52 does not include any mention of copper wires, fiber optics or radio: it is
not showing how these bit rates are implemented; it only lists the standard channelized system
line speeds in the industry.
An actual way of implementing these bit rates is called a technology.
Technology is two Greek words, meaning in English “knowledge of methods”. In this case, ac-
tual methods of implementing the line speeds and channelized transmission systems from the
previous section.
6.4.2 Carrier Systems
Digital Carrier System is the name given to the technologies for implementing the Digital Hier-
archy of 6.3, since they end up carrying multiple DS0-rate circuits on higher-rate DS1 and DS3
aggregate circuits.
6.4.3 T1
T1 was in the past a popular carrier system technology. T1 carries 24 DS0 channels, which is
a DS1-rate signal (1.5 Mb/s), over four copper wires.
6.4.4 T3 and Bit-Interleaved Multiplexing
There were two methods of multiplexing up to the DS3 rate (45 Mb/s): T3 multiplexing, often
also called asynchronous DS3 multiplexing is the old method.
This involves multiplexing in three stages, from DS0 to DS1, then DS1 to DS2, then finally DS2
to DS3. At the middle stage, the data is scrambled, resulting in a DS3 data stream with bits
scattered all over the place.
This is called bit-interleaved multiplexing, and means that to drop out a particular DS0 channel,
the entire DS3 has to be demultiplexed. This is not a good idea.

89
6.4.5 SONET and Byte-Interleaved Multiplexing
SONET (Synchronous Optical Network), also called synchronous DS3 multiplexing, was the
newer method. This involves byte-interleaved multiplexing right from DS0 or DS1 to DS3 rates
and beyond.
This means that it is easy to drop and insert individual channels out of the DS3. In addition, ex-
tra signaling and control for end-to-end error checking is included in overhead bits.
6.4.6 SDH
The Synchronous Digital Hierarchy (SDH) is a European standard for multiplexing that moves
multiples of 155 Mb/s, called Synchronous Transport Modules (STMs).
Where a SONET system moving multiple DS3s would be used in North America, an SDH sys-
tem moving multiple STMs is used in Europe.
6.4.7 Line Speed vs. Technology
It is important to distinguish between the line speed or data rate of a circuit and the particular
technology employed to provide a circuit with that data rate.
A common mistake is to always refer to 1.5 Mb/s as “a T1”. It ain’t necessarily so. T1 is a par-
ticular technology for providing a DS1-rate service, which is 1.5 Mb/s, using four copper wires
and a particular scheme for pulses to represent 1s and 0s.
There are other ways of moving 1.5 Mb/s, including HDSL on copper, on fiber and wireless. It
would be most accurate to refer to 1.5 Mb/s as “a DS1”.
That said, keep in mind that most people erroneously interchange “T1” and “DS1”, making
statements like “we’ve got a T1 coming into the building”.
To avoid this mistake, say “we have a 1.5 Mb/s circuit” or “we have a DS1-rate service” or “we
have a DS1 coming into the building”.
It is the rate – the line speed – not the technology that is usually of most interest.

6.5 Framing
In this section, we understand framing. Framing is extra information transmitted with the data,
allowing the demultiplexer at the far end to direct bits in the incoming aggregate to the correct
output port.
6.5.1 Synchronous Time-Division Multiplexing
To recap: in a channelized transmission system, the traffic for many users is aggregated onto a
high-bit-rate transmission system using synchronous Time Division Multiplexing (TDM).
Multiplexing means sharing. Time Division means that the sharing is done in time. Synchro-
nous means that the time-sharing is performed in a strict order in time, resulting in each user
being assigned a fixed time slot on the transmission system, called a channel.
The TDM is implemented by network equipment called multiplexers, which send a byte from
each user in a strict order, one after another, across the transmission system. This happens
8,000 times per second, and so moves 64 kb/s per channel: DS0 channels.
The stream of bytes from a particular user is interspersed with bytes from other users – other
channels – on the transmission system.
6.5.2 Framing and Transmission Frames
To allow the demultiplexer at the far end to direct the correct bits to the correct low-speed out-
put, it is necessary to also send control information.

90
 Figure 53. DS1 Frame
Since the users are sending bytes in a strict order, like a batting order at a baseball game, the
control information is minimal: it is only necessary to mark the beginning of the batting order
with framing bits.
When the far-end detects this information marking the beginning of the batting order, it then
knows that the next byte goes out to user 1, the byte following goes to user 2 and so forth… in
a strict order.
The information marking the beginning of the batting order, i.e. marking the beginning of the
frame is called the framing.
The framing, plus a byte from each user, is called a frame in the transmission business.
Frames are transmitted synchronously, 8,000 times per second.
6.5.3 DS1 Frame
The DS1 frame is the lowest level, the smallest frame, containing bytes from 24 channels plus
one bit for framing, as illustrated in Figure 53 and at the top of Figure 54.
A discussion of the actual DS1 framing bits and the framing patterns, Superframe and Ex-
tended Superframe formats is included in Appendix A.
6.5.4 STS-1 (DS3) Frames
Frames are packaged together into larger frames for high-bit-rate transmission systems. As il-
lustrated in the middle of Figure 54, the next frame size up from DS1 is the Synchronous
Transport Signal 1 (STS-1), which carries 28 DS1 frames.

Figure 54. STS-1 DS3 Frames and SONET OC192 Frames

91
The STS-1 carries a DS3, plus more framing called the transport overhead, and is commonly
called a DS3 frame. These larger DS3 frames are also transmitted 8,000 times per second.
6.5.5 SONET Optical Carrier Frames
The SONET Optical Carrier (OC) system moves multiple DS3 frames. For example, as illus-
trated at the bottom of Figure 54, a SONET OC192 system transmits 192 DS3 frames 8,000
times per second.
OC3 (3 DS3 frames), OC12 (12 DS3s) , OC48 (48 DS3s) and OC192 products were com-
monly deployed.
6.5.6 Advantages and Disadvantages of Channels
Implementation of channelized TDM results in “pipes”, that is, the capability to move a fixed
number of bits per second between A and B.
The main advantages of channelizing is that each user knows exactly what capacity they are
going to get.
The downside is that if a user has nothing to transmit, their channel is nonetheless reserved
and can not be employed by any other users. This makes the system inefficient for carrying
bursts of data.

6.6 ISDN
Integrated Services Digital Network (ISDN) is another technology for carrying DS0 channels.
Unlike T1 and SONET, which are essentially point-to-point transmission technologies, ISDN
also includes network addressing and circuit-switching: being able to specify where the DS0s
are to be terminated on a call-by-call basis.
Two flavors of ISDN are Basic Rate Interface (BRI) and Primary Rate Interface (PRI). These
are two very different technologies and must be distinguished.

Figure 55. ISDN Basic Rate Interface (BRI)

6.6.1 Basic Rate Interface (BRI)

BRI was designed for residences, running over the same twisted pair currently used for analog
POTS. This was to become the foundation for a new “basic” digital telephone service that ev-
eryone would have. This did not happen.
ISDN BRI provides two 64 kb/s DS0 channels plus a 16 kb/s signaling channel.
The DS0 channels are called bearer or “B” channels in ISDN lingo, and can be used to com-
municate voice or data.
The 16 kb/s channel is called a delta or “D” channel, and is used for signaling functions such
as call setup and release.
Since the user gets two bearer channels and one D channel, ISDN BRI is sometimes referred
to as 2B+D service.
The two DS0s can be used for two voice calls, or one voice and one data connection at 64
kb/s, or two data connections at 64 kb/s each, or two channels bonded together to form one
data connection at 128 kb/s.

92
Combining B channels to form a 128 kb/s data connection for telecommuters was one of the
applications for ISDN BRI.
The equipment needed to connect devices to an ISDN line must include the circuit terminating
function, a Network Termination Type 1 (NT-1), as well as an Terminal Adapter (TA).
These two functions usually come together in a single device, which has a jack for the phone
line on the phone company side, and a POTS jack and data equipment connector on the other
side.
6.6.2 Obsolescence of BRI
People have been talking about ISDN BRI for about 50 years. It was slow getting off the
ground and never gained much momentum. It now appears that ISDN BRI will join telegraphs
in the dustbin of history. Technologies like DSL, IP and Optical Ethernet have made it obsolete.
6.6.3 Primary Rate Interface (PRI)
ISDN PRI is not yet obsolete. PRI is a service that turns a DS1-rate access into 23 DS0 chan-
nels plus a signaling channel. PRI trunk means a DS0 that has an associated PRI signaling
channel.

Figure 56. ISDN Primary Rate Interface (PRI)

The signaling channel allows the communication of call control messages between switches.
This is the main value added feature of PRI.
The messages that can be exchanged are called the ISDN User Part (ISUP).
An outbound phone call connection request is an example of a message from the PBX to the
CO switch.
Messages from the CO switch to the PBX include
• Automatic Number Identification (ANI), which is Caller ID for PBXs and 911 systems,
• Direct Inward Dial (DID) called number identification messages, telling the PBX where to
terminate the incoming call,
and many others.
As described in Section 2.8.6, call centers use this signaling capability to dynamically change
the routing of 800 numbers to different locations depending on load.
6.6.4 PRI Physical Connection
Unlike BRI, ISDN PRI does not specify the physical connection. The standard implementation
was to run PRI on a T1 access.
PRI service can also be provisioned on a fiber-based access that carries multiple DS1s. One
signaling channel can support up to 5 DS1s.
6.6.5 T1 vs. PRI
Note that T1 can be used to carry PBX trunks from a customer’s building to the Local Ex-
change Carrier’s Central Office, or WATS lines from the Customer Premise through the LEC to
an Inter-Exchange Carrier.
This is not a “dedicated T1”; it is carrying multiple PBX trunks on a single physical access cir-
cuit; a convenience for the carrier. The T1 in itself does not change or add any value to the
story.
If ISDN PRI is ordered as a service using that T1, then PRI adds the capability to transmit and
receive control messages between the CO and the PBX via the PRI signaling channel.

93
6.7 Statistical Time Division Multiplexing
Statistical TDM is more efficient than channelized TDM.
With synchronous or channelized TDM, each user gets to use the high-speed circuit to send a
byte, in a strict rotational order like a time-share condo. This has the effect of giving each user
a fixed amount of capacity, called a channel.
This was designed for voice communications, since it is easiest to do quality voice communica-
tions if there is a constant amount of transmission capacity available for it.
If a particular device is idle, its channel, its assigned fraction of the high-speed circuit, is none-
theless reserved and cannot be used by any other devices. This makes implementation sim-
ple, but is not an efficient use of the high-speed circuit.
For data communication applications, we don’t really need – nor want – to have fixed amounts
of capacity available for transmission, since the traffic isn’t fairly constant like voice, but hap-
pens in bursts. With email or web surfing, the vast majority of the time, nothing is being trans-
mitted and occasionally, small file transfers happen.
In this case, a more efficient scheme of multiplexing called statistical TDM can be employed.
Capacity is allocated to a user when they demand it; otherwise, a different user can employ the
capacity instead. Since the term bandwidth means capacity, this is called a bandwidth on de-
mand strategy.
6.7.1 Toll Plaza Example
An example of statistical time-division multiplexing is a toll plaza.
There are a number of toll booths, with lines of traffic moving slowly through each onto a toll
highway, where traffic moves at high speed. Each line sends a car onto the highway as
needed

Figure 57. Toll Plazas are Statistical Multiplexers

.
When the highway is busy, lines send a car onto the highway when the next chance arises. If
there are no cars in the line at a particular booth, that lane will not use any capacity on the
highway, and other lanes can send cars onto the highway in those spots.
6.7.2 Overbooking / Oversubscription
Knowing that users aren’t always going to be needing capacity, the transmission circuit is over-
booked. Far more users are connected than would be with channelized multiplexing.
This technique is also called statistical multiplexing because it is necessary to know the statis-
tics of how often users will demand capacity, to know how much to overbook the transmission
circuit.
Ideally, the goal is to end up with 100% occupied transmission circuits, even though the users
are sending bursts of data. In practice, the design goal is 80% occupancy.
The example of the toll plaza in Figure 57 shows incoming lanes at 100 miles per hour (MPH),
and the outgoing lane at 10 MPH.

94
Of course, this is exaggerated – but in the direction opposite to what you might think! The “in-
coming lanes” would be the in-building LAN, which could be 1000BASE-T, running at 1000
Mb/s. The “outgoing” lane would be the WAN circuit, and could be running at 10 Mb/s… so
1000 MPH on the inputs and 10 MPH on the outputs would be closer to reality.
Clearly, it is necessary to know the statistics of how many cars per hour arrive at the toll plaza
demanding to use the outbound lanes – regardless of the fact they are traveling 1000 miles per
hour – to know how many input lanes there can be and what kind of traffic jams to expect.
Airlines also do this. Knowing that some passengers will not show up to claim their seat, air-
lines overbook flights. The objective is to end up with 100% full planes: no-one left behind, no
empty seats. It is necessary to know the historical statistics of how often people actually did
show up to claim their seat to know how much to overbook the flight.

Figure 58. Overbooking by a factor of 10

Figure 58 illustrates overbooking using statistical multiplexers at each end of a 10 Mb/s circuit.
Knowing that statistically, users are normally not going to be transmitting – normally they will
be doing nothing – ten users are connected to the 10 Mb/s circuit, they are all told they have a
10 Mb/s connection… and the network designer hopes they don’t all try to use it at the same
time.
6.7.3 The Need For Addressing
Overbooking introduces a complication, however. Channelized TDM systems are ultra-effi-
cient, requiring only minimal framing, because the channels transmit bytes in a strict rotational
order.
When the circuit is overbooked, the channels no longer transmit in a strict order. They transmit
data only when needed, and only when capacity is available. More extensive control informa-
tion is required compared to channelizing to allow the far end to determine what goes where.
The solution is to attach an address to the front of each chunk of data being sent over the sys-
tem. The equipment at the far end uses the address to determine where to send the chunk of
data.
This could be the MAC address on a frame, or the IP address on a packet. This additional con-
trol information uses a small fraction of the capacity of the high-speed circuit.
6.7.4 Statistical Multiplexing Equipment
Statistical multiplexing or bandwidth on demand is implemented with network equipment in-
cluding Ethernet switches, IP routers and MPLS Label Switching Routers.
These are the statistical multiplexers, the traffic cops, deciding which input can transmit next
on the outgoing overbooked circuit.

95
Ethernet switches, IP routers and MPLS LSRs are covered in detail in upcoming chapters.
6.7.5 Packet Networks
In this section, we have been explaining the concept of overbooking, called statistical multi-
plexing, using the simplest example: where the overbooking is done by multiplexers at each
end of one circuit.
If we then extend the concept to a whole network of high-speed circuits, what do we have? IP
packet networks.
This is the essential idea behind a packet network: the user takes a chunk of data, puts an ad-
dress on the front indicating the destination, which forms a packet, then the user sends this
packet whenever there is a free spot on the overbooked circuits that make up the network.
At each intermediate step, a device called a router examines the network address on the
packet and uses that information to decide on which overbooked circuit to forward the packet
next.
Overbooking the network circuits lowers the cost to users for a given access line speed. A net-
work where the internal circuit capacity is equal to the total of the access line capacity, called a
non-blocking network, would be prohibitively expensive.
An overbooked network, where the internal capacity is much less than the total access line
speeds is much less expensive, and gives almost the same apparent performance to the
user... based on the fact the users will normally do nothing, and only occasionally transmit a
packet at their access line speed.

6.8 Framing on IP Packet Networks

This section is included only for the purpose of comparing the method of packaging data and
framing on a legacy channelized transmission system, as illustrated in Figure 53, with the
method of packaging data and framing on new-generation statistically-multiplexed IP / Optical
Ethernet transmission systems as illustrated in Figure 59.
Ethernet, MAC frames and IP packets are introduced and briefly explained here. They are ex-
plained in detail in upcoming chapters.
6.8.1 Old vs. New
Legacy channelized transmission systems, in particular SONET and T1, transmit one byte for
each user 8,000 times per second, which results in a constant rate of 64 kb/s called a DS0
channel.
Users’ bytes are grouped into frames called Synchronous Transport Signals (STS), which are
transmitted at a constant rate of 8,000 times per second. STS-1 frames correspond to the DS3
rate.
In contrast, new transmission systems do not transmit one byte for each user bundled into STS
frames at a constant rate. Instead, a user transmits a group of bytes (an IP packet), carried in
an Ethernet frame, also called a MAC frame, only when needed.
IP packets are multiple bytes, with an IP address at the front. Like a telephone number, the IP
address indicates the final destination for a packet.
6.8.2 MAC Frames Instead of Framing Bits
IP packets are carried inside Ethernet frames, also called Media Access Control (MAC) frames
since they have a MAC address at the beginning. The MAC address identifies the destination
on the current circuit.

96
 Figure 59. New systems transmit IP packets in MAC frames as needed, instead of single bytes and framing bits
all the time
MAC framing is six bytes, a special bit pattern so that the receiver can find the start of the
frame. This is followed by the destination and source MAC addresses, indicating the sending
and destination stations on the current circuit. Then a control field indicates how many bytes
there are in the payload or information field, in which is carried typically one IP packet. This is
followed by an error detection scheme called Cyclic Redundancy Checking (CRC), imple-
mented using a Frame Check Sequence (FCS).
6.8.3 Routers
A router is used to direct the packets onto the outgoing transmission system. The users’ pack-
ets will arrive on one or more incoming physical access circuits and the router will relay or for-
ward the packets, in frames, on the outgoing transmission system.
The users only send packets to the router when they have traffic to be transmitted. The router
will forward the packets on a first-come, first-served basis if there is no prioritization imple-
mented. This implements the “bandwidth on demand”. It is the router than manages the over-
booking or oversubscription of the outgoing circuit.
6.8.4 Prioritization
If a prioritization scheme is implemented, the router will forward incoming packets not on a
first-come, first-served basis but in an order determined by the indicated priority of a packet
and the queuing algorithm that the router is implementing. For example, packets containing a
live telephone call might be prioritized over packets containing email messages.
6.8.5 MPLS
It should be noted that a traffic management system called Multiprotocol Label Switching
(MPLS) is used in a carrier’s network core. The IP packets have a label number affixed to
them, and this label number is used for routing (and possibly prioritization) instead of the IP ad-
dress. This is internal to the network and invisible to end-users.
6.8.6 Implementation with Optical Ethernet
In practice, this new generation of IP packets in MAC frames is implemented using Optical Eth-
ernet, which is Ethernet switches connected with point-to-point fibers. MAC frames are sig-
naled over the fibers by flashing a light on and off.

6.9 Coexistence and Transition from Channels to Pack-

ets
6.9.1 Old: Everything in Channels
In the previous millennium, the telecom network core was based on SONET, a technology im-
plementing channels on fiber optics.

97
Telephone calls were converted from voiceband analog on loops to 64 kb/s DS0 streams by
the line cards in the CO switch, then switched to trunks carried long distance in DS0 channels
on the SONET backbone, one trunk per channel.
Data was carried in ATM cells or IP packets interspersed in a high-speed stream by a router.
These high-speed data streams were carried point-to-point between routers in DS3 channels
on the SONET backbone.
As illustrated at the top of Figure 60, in this way, integration was achieved in the network core
by carrying everything in channels.
6.9.2 New: Everything in Packets
Going forward, the telecom network core is everything in IP packets carried on Optical Ether-
net and managed with MPLS.
A practical question for carriers is how to transition the huge installed base of analog loops,
CO switches and channelized SONET infrastructure to the new IP over Ethernet paradigm.

Figure 60. Transition From Channels to Packets

6.9.3 Gateways for Legacy Voice

As illustrated at the bottom of Figure 60, for analog loops and CO switches, a Packet Voice
Gateway is installed on the network side of the CO switch.
Like the gateway device introduced in Section Figure 57 “Toll Plazas are Statistical Multiplex-
ers”, this device converts the voice trunks from the channelized DS0 format to the packetized
IP format.
The voice packets can then be interspersed with data packets (and video packets) on the net-
work core, achieving integration via packets while preserving the CO switches and analog

98
loops for the time being
6.9.4 Packetized Voice from the Customer Premise
However, that is not the final answer, hence the label “NEWER” instead of “NEW” in Figure 60.
In the longer term, the gateway function will be moved to the customer premise and the analog
loops, CO switches and DS0 trunks will completely disappear. Traffic to and from the customer
premise will be voice, video and data interspersed in packets.
6.9.5 Packets over Non-Channelized SONET
To avoid having to immediately replace the existing network core SONET equipment with Opti-
cal Ethernet, the SONET systems might be used in a non-channelized fashion, that is, using
the entirety of a 10 Gb/s SONET OC192 link to carry MAC frames point-to-point in the same
way that a 10GBASE-x Optical Ethernet link (Section 8.5) would.
In the longer term, the SONET equipment and its optical transceivers will be replaced with Op-
tical Ethernet.

99
7
The Cloud and Service Provisioning
7.1 Anatomy of a Service
Any service provided by a telecommunication service provider is made up of three compo-
nents: access, network connection and billing agreement.
The access circuits are physical lines with circuit terminating equipment at each end. These
lines run from a user’s site to the nearest physical attachment point to the carrier’s network.
The location containing this physical attachment point is usually a Central Office (CO). It may
be in an enclosure outdoors or underground in a vault.
There are many different technologies for access circuits, including
• Plain Ordinary Telephone Service (POTS) lines
• Older-style digital data circuits at up to 56 kb/s,
• ISDN BRI digital telephone lines at 128 kb/s,
• xDSL technology at 1 – 200 Mb/s,
• Cable modem technology at 1 – 500 Mb/s or more,
• Passive Optical Networks at 1 Gb/s or more,
• T1 digital access circuits at 1.5 Mb/s,
• Cellular and point-to-point radio,
• SONET fiber-based circuits based around 45 Mb/s,
• Optical Ethernet from 1 to 100 Gb/s.
These are short circuits, hopefully less than a couple of miles long. Each type of access circuit
must have a specific type of Data Circuit-terminating Equipment (DCE), a type of customer
premise equipment, attached to the line to be able to transmit data on that circuit. Some exam-
ples are:
• Small Formfactor Pluggable (SFP) optical transceivers for fiber,
• LAN Network Interfaces: copper, fiber and wireless implementations,
• Modems for wireless, DSL, cable modem and POTS,
• Data Service Units (DSUs), for old 56 kb/s non-switched digital circuits,
• Channel Service Units (CSUs), used on T1 circuits,
• CSU/DSUs, used on switched-56 kb/s circuits,
• Optical Network Units (ONUs), Optical Network Terminals (ONTs), and Optical Line Termi-
nals (OLTs) used on fiber circuits.

100
 Figure 61. The Network Cloud
Network connections between the access circuits are made over high-capacity circuits that are
owned and managed by the network service provider. Many options as methods of connection
through their networks are offered. These can be summarized into three fundamental choices:
• Full period: connected all the time, billed as a monthly fixed charge.
• Circuit-switched: connected on demand, billed as a monthly fixed charge for the access cir-
cuit plus a per minute usage-sensitive charge.
• Bandwidth on Demand or “packet-switched”: available all the time, billed as a monthly fixed
charge for the access circuit plus in theory a usage-sensitive charge based on the amount
of data transmitted.
Note that there is a monthly charge for each access. A combination of access circuits with their
circuit-terminating equipment, method of connection, and of course, billing plan make up a ser-
vice.

7.2 The Network “Cloud”

We often draw pictures of carriers’ networks as clouds with sticks poking into them. Some peo-
ple even get carried away with jargon and start referring to network services like MPLS as the
“MPLS Cloud”.
The sticks are, of course, the access circuits discussed in previous sections. But why do we il-
lustrate carriers’ networks as clouds? To emphasize the idea that we do not necessarily know
what is inside the cloud, and frankly, most of the time, we do not care.
When you get a circuit like a “T1” from a carrier, you are not buying a circuit… you are paying
for a service. If you put signals on one access circuit, and they appear on another within a
specified delay and reproduced with a specified fidelity… that is what you are paying for. It is
none of your business how the carrier actually makes that happen.
However, when ordering and troubleshooting circuits provided by network service providers, it
is useful to know what is going on inside their network “cloud”.

7.3 Inside the Network Cloud

When ordering and troubleshooting services provided by network service providers, it is useful
to know what’s going on inside their network “cloud”.
From a network planning engineer’s point of view, the network has historically had two parts:
the core and the edge. The core consists of high-capacity fiber circuits connecting COs within
a city and/or connecting cities together. The regional rings are interconnected at multiple
places for long-distance communications.
The purpose of the ring pattern is to ensure high availability, that is, no loss of data when lines
are cut. To do this, it is necessary to provide redundant paths. The cheapest way to do this is
to connect locations neighbor-to-neighbor to form a ring, implementing two connections to ev-
ery location with only one extra circuit.
Historically, the Automatic Protection Switching, i.e. managing the redundancy, alarms and ser-
vice restoration after a cut line, has been implemented with SONET, which also implements the
optical transmission and channelizes the available capacity on the fibers.
Going forward, Optical Ethernet and Resilient Packet Rings are used to implement a core that
moves everything in IP packets and Ethernet frames.
A customer is not going to be a station on the core, and is not usually buying services at the bit
rates of the core (40 Gb/s).
Most customers order services measured in the Mb/s, with a single copper or fiber access to
the network.

101
 Figure 62. Inside The Cloud
To connect the customers’ access circuits and lower-bit-rate services to the core fiber ring,
edge equipment is provisioned at each station on the ring, and this edge equipment is con-
nected in pairs across the ring.
The edge equipment acts as a data concentrator, and as a converter between access circuit
technology (e.g. copper wires or lower-speed fiber) and the fiber-optic core technology. Cus-
tomers’ access circuits are connected to the edge equipment, which aggregates the traffic into
a stream transmitted to its opposite number. The edge equipment at the far end distributes the
traffic to the correct far-end access.
There are three basic kinds of edge equipment: multiplexer, telephone switch and router. Each
of these partitions the capacity of the connection across the ring between the users in a differ-
ent way, and so each is used to implement a different kind of network service. This is covered
on the next page.
In the future, there will be only one kind of network service: IP packets carried in Ethernet
frames, and so the switches and multiplexers will disappear and only routers will be used for
the data concentration function.

102
For business customers, the same Optical Ethernet technology used on the core will be used
on the access, so the physical conversion functions associated with edge equipment will be
largely unnecessary.
For residential customers, the existing copper entrance cable will be used for some time to
come, so the edge equipment will route packets between DSL modems and copper wires on
the low-speed (customer) side and the fiber core on the aggregate (network) side.

7.4 Network Equipment

These three types of equipment are used to aggregate low-speed services onto a high-speed
backbone. The hardware is all similar; the difference in usually more one of software and capa-
bilities.
A multiplexer is a dumb device. One has to plug in cards, wire up circuits, then sit down at a
management terminal and tell it which input goes on which channel on the output. When you
walk away, it stays that way. These are used for establishing full-period services, which are of-
ten also called dedicated lines, private lines or leased lines. Multiplexers reserve capacity on
the backbone for a user all the time.
A cross-connect is like a multiplexer, except that it has the same “high speed” circuit on both
sides. The cross-connect can move a channel in one circuit to a different channel in the other.
A switch is smarter. A switch has a processor running software capable of recognizing re-
quests from users to establish a connection, and to release it. A switch can reserve capacity
for a user on the backbone on a per-call basis. These are used for circuit-switched services
like POTS and ISDN.

103
 Figure 63. The Three Basic Types of Edge Equipment
A router doesn’t establish connections, and doesn’t reserve capacity for a particular user.
Routers are stateless devices that treat each packet they receive individually. At a network ser-
vice provider, the packets come in on low speed access circuits and go out on the next avail-
able spot on a high-speed backbone. Routers are used for bandwidth on demand services like
Internet service and commercial IP packet communication services.
The question of reserving capacity or not can be described by full-period (muxes) vs. circuit-
switching (switches) vs. bandwidth on demand (routers). This all boils down to channelized vs.
statistical multiplexing of portions of the backbone.
For advanced readers: The distinction between a router and a switch can be confusing when
considering an ATM switch. This device performs both switching and bandwidth on demand.
The easiest way to understand this is to consider that “switching” and “routing” both mean
making a route decision. Traditionally, the term “switch” also means that a connection is estab-
lished by the network equipment, and that there will be a flow of information along the same
path to a particular destination. In this case, the ATM switch is capable of establishing virtual
circuits, which are routes over which everything will travel. Conversely, routing IP packets does
not involve setting up virtual circuits, and IP packets do not have to all follow along the same
path. In MPLS, both terms are used together: virtual circuits are called Label-Switched Paths,
and to cover all bases, the network device is called a Label-Switching Router.

7.5 Service Provisioning Summary

As illustrated in Figure 64, services are provisioned, that is, put in place, with an access circuit
at each end, a network connection across the “backbone” or core, and edge equipment con-
necting the access circuits to the core.
Services are generally provided on copper or fiber access circuits running from the local car-
rier’s Central Office to the customer premise. These circuits are short - perhaps a couple of
miles long.
Since the network core runs at more than 10 Gb/s, a method of putting the lower speed service
provided to the user onto the backbone and taking it off at the other end is required.
The equipment used to perform this function is called edge equipment. The type of edge
equipment used depends on the type of network connection desired, that is, the way that ca-
pacity will be allocated to the service.
Full period connections use a multiplexer. Circuit-switched connections use a telephone switch
and packet-switched or bandwidth on demand services use a router.
The edge equipment aggregates a user’s traffic with many other users’ traffic to form a high-
speed aggregate that is transmitted over the core to corresponding edge equipment at the far
end, where the reverse process takes place with similar equipment and cabling.
In the past, the network core was channelized, using SONET OC48 and OC192, and routers
directed data packets onto dedicated channels alongside channels carrying voice trunks and
dedicated lines.

104
 Figure 64. Service Provisioning
Going forward, the network core is packetized, moving IP packets in Ethernet frames. All traf-
fic: voice, data, video, Internet traffic and anything else is placed in IP packets and inter-
spersed with other users’ packets on the core. Routers forward packets from the access cir-
cuits to the core and from the core to the access circuits.

105
8
Fiber Optics
8.1 Fiber Basics
The fundamental idea behind optical transmission is varying some characteristic of a light
beam to represent information, transmitting that light beam through a solid tube of glass that
guides the light to the far end of the tube, where the light is detected and interpreted.

Figure 65. Communication of 1s and 0s Using Pulses of Light

In most cases, the information is coded into binary digits, 1s and 0s, and light at a single fre-
quency is turned on and off to represent the 1s and 0s. Light at single frequencies – or as
close to a single frequency as possible, since nothing is perfect – is generated by a laser.
In a commonly-used technology, Optical Ethernet, “light on” represents a one and “light off”
represents a zero.
The information to be transmitted is often delivered to the optical system on copper wires, usu-
ally in the form of lines of copper called traces on a printed circuit board.
This requires an electrical - optical conversion at the transmitting end, where the electrical sig-
nal on the copper drives a switch that turns the laser on and off, and optical - electrical conver-
sion at the receiving end, where a photodetector drives a switch that turns voltage on and off.
An optical transceiver is both functions together.
8.1.1 Lamdas
Usually the light is characterized by its wavelength rather than its frequency. l is used as an
abbreviation for length. In Greek, the letter l is λ (lambda), so wavelengths are called lambdas.
There is a direct relationship between wavelength and frequency: λ = c/f; wavelength = speed
of light / frequency.
8.1.2 Pulses of Light
The burst of photons vibrating at the specified frequency emitted while the laser is on is called
a pulse.
The average intensity of the light while on is called the envelope of the pulse. This envelope is
detected by a photodetector at the far end and used to control an electrical signal, for example,
pulses of voltage on copper.

106
The rate at which the light can be turned on and off is the primary factor determining how many
bits per second can be represented on the fiber.
An exception where pulses of light are not used is fiber to the neighborhood in many cable TV
systems. On these systems, the amplitude of light is varied continuously within a broad band of
frequencies, as a direct analog of the electrical signal at the same frequencies on coaxial cop-
per wires.
8.1.3 Attenuation and Dispersion
Impairments, that is, the factors that reduce the ability to reliably detect the pulses at the far
end include attenuation and dispersion, and usually worsen with distance, so have the effect of
limiting the useful range.
Attenuation is the diminishment of signal strength with distance, caused by not-perfectly-trans-
parent glass. This is usually not the limiting factor – unless bad splices or faulty connections
severely attenuate the signal.
Dispersion causes the lengthening in time of the pulse envelope while in transit over the fiber.
If the pulse duration were to double during transit, at the far end the pulses would merge to-
gether, making it impossible to reliably detect them.
Before this happens, the pulses must be detected and regenerated by a repeater in an optical-
electrical-optical process, or with very advanced technology, the pulses are reshaped (short-
ened) optically.
There are many ways that dispersion happens: modal dispersion, chromatic dispersion, polar-
ization mode dispersion and others. This is covered in more detail in a subsequent lesson.

8.2 Glass Fiber and Fiber Cables

Glass fiber is the physical medium of choice for implementing backbone or core networks, for
three main reasons: 1) bandwidth, 2) bandwidth and 3) bandwidth. Fibers are capable of sup-
porting the transmission of huge numbers of bits per second.
A glass fiber is a physical medium for communicating information just as copper wires are.
Glass is used because it has good dimensional stability (doesn’t kink), strength, cost and
transmission properties.
On a glass fiber, binary digits are represented by flashing on and off a laser producing energy
at a frequency of around 200 x 1012 Hz, which corresponds to about 1.5 x 10-6 meters = 1.5 mi-
crometer = 1500 nanometers in wavelength.
On many systems, light-on represents 1 and light-off represents 0. The number of times per
second the laser is turned on and off determines the bit rate.
The main job of the fiber is to guide the light from one end to the other – without losing any of
it.
8.2.1 Core
A fiber consists of two different types of highly pure, solid glass mixed with specific elements
called dopants to adjust the refractive index of the glass, which is one of its transmission char-
acteristics.

107
 Figure 66. Fibers and Cables
The innermost part of the fiber is a solid tube of glass called the core. The purpose of the core
is to act as a waveguide for the light.
The core diameter is measured in millionths of a meter, called microns.
micro (μ) = 10-6; 1 μm = 10-6 m.
Fiber core sizes range from about 5 to 50 microns. A human hair is about 100 microns in diam-
eter.
8.2.2 Cladding and Coating
Around the core is the cladding, which is also glass, but with a different refractive index than
that of the core.
The difference in refractive index causes light injected into the core at certain angles to reflect
back into the core – thus constraining all of the optical energy which is the pulse to travel in the
core and exit the far end.
Around the core and cladding is a colored plastic coating to waterproof and identify the fiber.
8.2.3 Cables
A fiber optic cable contains multiple fibers, which are usually organized in bundles in colored
soft plastic tubes called the inner sheath for identification.
A sticky waterproof compound is placed in the inner sheath to repel water, which can infiltrate
glass and change its transmission characteristics.
More layers of hard plastic and metal are added to protect the fibers from water, shovels and
backhoes. A strengthening member may be present to keep the cable from being bent too
sharply during installation, causing micro-cracks or outright breaks.
Mechanical protection - armor - is added to protect against a type of signal degradation on
fibers known as backhoe fading: being cut with a mechanical shovel.
The outermost layer is called the outer sheath. Ripcords – steel and/or nylon – may be incor-
porated to allow installers to strip away sheaths without damaging the fibers.
A slippery covering can be added to make it easy to pull the cable through long runs of conduit
called ducts.
A fluorescent orange outer covering may be added to make the cable more visible and lessen
the chance of an accidental cut.
8.2.4 Redundancy

108
Cables get cut, particularly by construction crews digging up streets with backhoes for unre-
lated work.
To maintain availability of communications, two fibers on different cables following different ge-
ographical routes can be installed.
This is called redundancy or path diversity.
In some systems, the same data is transmitted on both cables at the same time, guaranteeing
no loss of data.
Other strategies implement automatic protection switching, i.e. moving traffic to a different ca-
ble after a break, which may involve some loss of data.

8.3 Optical Wavelengths, Bands and Modes

There are five main windows, bands, or ranges of frequencies within the light spectrum that
are exploited for transmission using fiber optics.
The light used on fiber is characterized by its wavelength, not its frequency. Light wavelengths
are measured in nanometers (nm), 10-9 m. Wavelengths on the order of 1000 nm are used for
communications.
For comparison with radio in the kilohertz to Gigahertz range, a wavelength of 1550 nm hap-
pening at the speed of light of about 2 x 108 m/s in glass, means a frequency of about 130 x
1012 Hz or 130 Terahertz in glass.

Figure 67. Wavelengths and Bands

8.3.1 Bands
The first two bands generally used in transmission systems were centered around 850 nm and
1310 nm. The first band, near 850 nm, was used almost exclusively for short-range, multimode
applications.
Single-mode fibers were first designed for use in the second window, near 1310 nm. To opti-
mize performance in this window, the fiber was designed so that a type of dispersion called
chromatic dispersion would be close to zero near the 1310 nm wavelength.
As the need for greater bandwidth and distance increased, a third window near 1550 nm called
the Conventional or C-band has been exploited for transmission. It has much lower attenua-
tion, and is within the frequencies amplified by erbium-doped fiber amplifiers.
More recently, bands above and below the C-band, called the Short or S-band and Long or L-
band have been exploited for transmission.
8.3.2 Multimode and Modal Dispersion
Dispersion is the spreading of the duration of the pulse envelope and is caused by numerous
factors.

109
As illustrated in Figure 68, if the pulse envelope lengthens too much, adjacent pulses merge
together and can not be detected at the receiving end. This limits the distance before a re-
peater is required.

Figure 68. Multimode and Modal Dispersion

One type of dispersion is modal dispersion.
A mode is a path that light can follow through the fiber’s core. Each mode is essentially a
bounce path at a different angle of incidence.
There are two general categories of optical fiber: multi-mode and single-mode. Multi-mode, the
first type of fiber to be commercialized, gets its name from the fact that numerous modes exist
simultaneously in the core.
Light in each of the different modes travels a different physical distance, so arrives at the end
of the fiber at differing times, which causes dispersion, the duration of the pulse envelope to
lengthen.
Accordingly, multi-mode fiber is not used for long distance applications. However, connector
tolerances are generous and low-cost light sources such as Light Emitting Diodes (LEDs) and
Vertical Cavity Surface-Emitting Lasers (VCSELs) can be employed.
Multi-mode fiber is usually graded index; the refractive index of the core gradually decreases
from the center of the core outward. The higher refraction at the center of the core slows the
speed of some modes, reducing modal dispersion.
8.3.3 Single-Mode Fiber
The second general type of fiber, single-mode, has a much smaller core.
A perfect waveguide, i.e. one that would allow only a single mode, in cylindrical form would
have a diameter a bit more than half a wavelength.
With wavelengths in the 1500 nm range, that would require a fiber with a core 0.8 microns in
diameter. But that does not exist yet.
Single mode means thinner cores, 5 to 9 microns (compared to multi-mode at 50 microns). In
practice, this means fewer modes than “multi-mode” fiber, and would more properly be called
“fewer-mode” fiber.
As a result, modal dispersion is greatly reduced and the possible transmission distance or
reach is greatly lengthened.
All long-distance and high-bandwidth systems use single-mode fiber.
Lasers producing wavelengths of 1300 - 1600 nm are used as the light sources for single
mode fiber.
Single-mode fiber has a step index; there is a uniform index of refraction throughout the core
and a step in the refractive index where the core and cladding meet.
This smaller core and step index both act to reduce dispersion caused by various mecha-
nisms.
8.3.4 Chromatic Dispersion
Lasers are not perfect; they do not produce energy at exactly one wavelength, but rather over
a narrow range of wavelengths.

110
Since the propagation speed of light in glass is affected by its wavelength, this imperfection
has the effect of causing some light to take longer to arrive at the far end later than other light,
causing dispersion. This type of dispersion is called chromatic dispersion.
8.3.5 Polarization-Mode Dispersion
The propagation speed of light in a fiber is also affected by the diameter of the fiber. As light is
actually two waves at right angles (horizontal and vertical) propagating forward, slightly oval-
shaped fiber causes these two waves to propagate at different speeds, causing polarization
mode dispersion.

8.4 Wave-Division Multiplexing: CWDM and DWDM

The need for speed is never-ending.
More capacity can be implemented by increasing the number of fibers in a cable, and increas-
ing the number of signals on each fiber.
Fiber counts in long-distance transmission cables increased from 6 in the 1990s to 144 and
288 fibers per cable in 2000, as the labor and right-of-way costs now far outweigh the cost of
the glass.
In greenfield access network outside plant builds, cables manufactured with factory splices and
optical connectors, pre-engineered for distance from the Central Splitting Point and far-end lot
spacing are deployed.
8.4.1 WDM
The biggest development has been to place multiple signals on each fiber. This is accom-
plished by using multiple single frequencies called carriers and referred to by their wavelength,
each pulsing on and off to represent bits. This is called wave division multiplexing.

Figure 69. Wave-Division Multiplexing (WDM)

Placing many of these carriers tightly spaced in frequency on a single fiber is referred to as
Dense Wave Division Multiplexing (DWDM). Placing as few as two carriers on a fiber is called
Coarse Wave Division Multiplexing (CWDM).
The scientific symbol for wavelength is the Greek letter lambda (λ). When someone refers to “a
lambda”, they are referring to a particular light carrier signal amongst many on a fiber.
The critical elements are the frequency bandwidth of the carriers – nothing is perfect – and
their spacing. Challenges include crosstalk: keeping wavelengths from spreading into neigh-
boring wavelengths during mixing, pulsing and transmission; and channel separation: the abil-
ity to distinguish each wavelength.
8.4.2 WDM Multiplexers
WDM multiplexers take multiple optical wavelengths and converge them into one beam of light.
At the receiving end, demultiplexers must separate the wavelengths and couple them to indi-
vidual fibers to be detected.

111
Multiplexers and demultiplexers can be either passive or active in design. Arrayed waveguide
gratings consist of an array of curved-channel waveguides with a fixed difference in path
lengths. The waveguides are connected to cavities at the input and output.
When light enters the input cavity, it is diffracted and enters the waveguide array. There, the
length difference of each waveguide causes phase delays at the output cavity, where an array
of fibers is coupled. Different wavelengths have constructive interference at different locations,
which correspond to the output fibers.
A different technology uses thin film filters; the property of each filter is such that it transmits
one wavelength while reflecting others. By cascading these devices, multiple wavelengths can
be isolated.
8.4.3 Optical Ethernet Paths
A low-cost type of WDM multiplexer is an Optical Ethernet transceiver that transmits data in
parallel over several wavelengths to achieve high bit rates.
For example, the 40GBASE-LR4 standard signals 10 Gb/s over four wavelengths in parallel to
achieve 40 Gb/s.
100GBASE-SR10 signals 10 Gb/s over ten wavelengths in parallel to achieve 100 Gb/s.
Each of the parallel wavelengths is called a path.
8.4.4 Current and Future Capacities
There is a large installed base of expensive DWDM multiplexers implementing 24 or 32 wave-
lengths around 1550 nm for very high capacity core network connections.
In the future, systems implementing 1,000 wavelengths on each fiber, each signaling at least
10 Gb/s, will result in available capacity of 10,000,000,000,000 bits per second, which is
10,000,000 Mb/s or 10 Terabits per second (Tb/s), per fiber, in the core.
In the not-too-distant future, 10 Gb/s will be a normal speed for business customer services.

8.5 Optical Ethernet

Optical Ethernet is signaling MAC frames (Section 12.5) from one device to another by flashing
a light on and off.
Light on represents a 1 and light off represents a 0.
8.5.1 Point-to-Point Connections
Normally, Optical Ethernet is implemented with point-to-point connections: from a port on one
LAN switch to a port on another LAN switch, from a port on a LAN switch to a port on a cus-
tomer edge device, or from a port on a LAN switch to a LAN interface on a computer.
8.5.2 SFP Modules and Connectors
The light flashing on and off is implemented with a laser at the transmitter and a photodetector
at the receiver. In many cases, this is the optical side of an optical-electrical interface, i.e. an
transceiver with fiber on one side and a copper-based LAN interface on the other.
These transceivers are typically implemented on Small Form-factor Pluggable (SFP) modules,
which are hot-swappable in the terminating equipment at each end.
In some cases, the SFP modules are embedded in the terminating equipment, meaning the
fibers are plugged into the terminating equipment. This allows re-use of existing fiber.
In other cases, the SFP modules are attached to fiber cables by the fiber cable manufacturer,
meaning the SFP module is plugged into the terminating equipment. This ensures the fiber and
transceiver technology are matched and the optical connection is a high-quality “factory” con-
nection.

112
 Figure 70. SFP Optical Transceivers
The SFP module format is not the subject of a standard, but rather described in industry Multi-
ple Sourcing Agreements (MSA).

Figure 71. Optical Ethernet Standards

8.5.3 IEEE Standards

There are many technologies for transceivers implemented on the SFP module. Some are pro-
prietary, many are standardized by the IEEE. In practice, the same manufacturer’s product is
used at both ends of the fiber to ensure compatibility.
Most technologies use one fiber for each direction. Some use two wavelengths for two direc-
tions on one fiber.
The 40 and 100 Gb/s technologies split the bitstream into subrates and transmit them in paral-
lel on different wavelengths called paths or lanes.
The table lists current IEEE standards. More will be published in the future.

8.6 Network Core

The network core, colloquially referred to as the backbone, provides high-capacity, high-avail-
ability connections between switching centers.
Fiber optics is used as the basis of connections between switching and routing centers since it
can support very high numbers of bits per second. Lower-speed (and lower-cost) circuits are

113
used to provide access to this core to users.
A method of organizing the bits for transmission, plus monitoring, alarming, testing and auto-
matic protection switching is required for reliable service.
8.6.1 SONET and SDH
In the past, the most popular technology for these functions in North America was a standard
called Synchronous Optical Network (SONET).
In the rest of the world, a very similar technology called Synchronous Digital Hierarchy (SDH)
is employed. There is a very large installed base of SONET and SDH systems.
8.6.2 Optical Ethernet, RPR and MPLS
For new deployments, Optical Ethernet, that is, 802 MAC frames signaled on fiber between
Layer 2 switches is deployed. Layer 2 switch is another term for a LAN switch or Ethernet
switch, since its functions correspond to Layer 2 of the OSI 7-Layer Reference Model.
A technology called Resilient Packet Ring (RPR) and/or MPLS is used instead of the protec-
tion-switching capabilities of SONET / SDH to implement the recovery from broken connec-
tions.
8.6.3 Fiber Rings
To ensure high availability, that is, the possibility of communicating even if a line is cut or
equipment fails, it is necessary to provide multiple redundant paths between each point.
Figure 72. The Network Core - Fiber Rings

114
The cheapest way to do this is to connect locations in ring patterns. This way, there are two
connections to every location, but only one extra circuit.
Rings are used to connect COs in a city together. Rings are also installed to connect cities in
regions together, and these regional rings are interconnected at multiple places for long-dis-
tance communications.
Short-cuts, i.e. connections between non-adjacent points on the ring, are implemented as traf-
fic dictates. The end result is a semi-meshed network, where some locations are directly con-
nected and others are reached via intermediate stations.

8.7 Metropolitan Area Network

Metropolitan Area Networks (MANs) are implemented by connecting Layer 2 switches with
point-to-point Optical Ethernet connections. For redundancy, locations are connected in ring
patterns, implementing two physical connections at each location for the lowest cost.
VLANs are used to separate customers’ traffic, so customers can not communicate between
each other directly via the MAN, and customers do not receive each others’ traffic.
MANs are a key part of the access network, the “last mile”, part of the infrastructure connecting
customers to a Central Office, which in turn provides connectivity to the network core.
Many different MANs would originate at a CO, connecting different types of customers. Large
business customers, for example an Internet Service Provider or data center could be on their
own MAN.

Figure 73. Metropolitan Area Network Around a CO

8.7.1 MANs to Office Buildings and Apartment Buildings

In multi-tenant office buildings, carriers implement mini-POPs, a Layer 2 switch in an equip-
ment room to which access circuits going to different tenants in the building are connected.
One MAN would connect these mini-POPs in different buildings to the CO.
In apartment buildings, Ethernet to the Suite is implemented in the same way, with a Layer 2
switch in an equipment room and fiber or copper to each suite. Another MAN would connect
these buildings to the CO.
8.7.2 MANs to Neighborhoods
In neighborhoods, a MAN would connect Layer 2 switches contained in outside plant enclo-
sures to the CO. An outside plant enclosure is a secured weatherproof cabinet located by the
side of the road. From the enclosure, dedicated fiber access circuits are pulled as spokes to
business customers.

115
In residential neighborhoods, a MAN would connect Layer 2 switches contained in outside
plant enclosures to the CO. From these switches, fiber would lead to Passive Optical Network
splitters on poles or in pedestals, where typically 32 fiber access circuits are pulled as spokes
to residences and small businesses, time-sharing the backhaul to the switch in the enclosure.
A MAN would also connect DSLAMs contained in outside plant enclosures to the CO. The
DSLAM contains network-side modems that are hard-wired to subscriber loops to implement
communications at up to 200 Mb/s over the last few hundred meters in brownfields, i.e. where
copper loops are already deployed.

8.8 Fiber to the Premise (FTTP) & PONs

Fiber optics are used for the vast majority of transmission systems connecting switching cen-
ters.
In the access network, fiber is now used routinely for new installations, both for business ser-
vices and greenfield residential builds. Deployment of fiber in existing brownfield residential ar-
eas is a longer-term project.
This is referred to as Fiber to the Home (FTTH) or more inclusively, Fiber to the Premise
(FTTP).
8.8.1 Passive Optical Network (PON)
A popular strategy for residences and small business is a Passive Optical Network (PON), the
word passive meaning that there are no powered components in the access network, only
fibers and light.
Some implementations use proprietary strategies for the optical design, others conform to the
Optical Ethernet standards like 10GBASE-PR.
Typically, an optical splitter (and combiner) is deployed in the access network in an outside
plant enclosure that might be called a Central Splitting Point.

Figure 74. Passive Optical Network

The optical splitter concentrates numerous fibers that lead to customers’ Optical Network Ter-
minals (ONT) to one fiber that is connected to the carrier’s Optical Line Terminal (OLT). Today,
a 1:32 split is common. In the future, that may rise to 1:64 and 1:128.
This is cheaper to implement than a dedicated fiber for each customer, since the 32 customers
are sharing a single OLT and a single backhaul from the splitter to the network.
Engineered cables with factory-installed optical connectors at the same spacing as the building
lots are often ordered and installed by telephone companies. This lowers the installation labor
time and cost and increases the reliability of the connectors.
Typically, two wavelengths are used, one for communications downstream and one for up-
stream.
This requires encryption of the content downstream, as it is broadcast to all ONTs, and chan-
nelized Time-Division Multiplexing on the upstream, since all ONTs are sharing a single up-
stream path to the OLT.
8.8.2 Active Ethernet
Another strategy for fiber to the premise is called Active Ethernet, or Active Optical Network
(AON), where each customer has a dedicated point-to-point fiber terminating on a dedicated

116
OLT at the service provider.
Since point-to-point dedicated connections are the normal configuration in Ethernet, presum-
ably the only reason the word “active” is included in the product name is to differentiate it from
PON products.
One variation implemented for residential customers includes proprietary equipment housing
the OLTs in the CO plus dedicated fiber from the CO to each customer.
Another variation of active Optical Ethernet is the Metropolitan Area Network (MAN) of Section
8.7, Optical Ethernet connections between switches deployed in ring topologies.
A business could be a station on a MAN, meaning there is an Ethernet switch owned by the
carrier deployed at the customer premise, plus two fiber connections.
Being part of a MAN ring in this configuration means the customer would enjoy high availability
- while paying for two optical accesses.
A business could also have a single fiber access as a spoke connecting to a station on the
MAN that is located in a CO, POP, in a collocation, in an equipment room or in an outside plant
enclosure, as illustrated at the top right of Figure 73.
In this case, the business pays only one optical access charge but would have time to restora-
tion in the case of a cut line measured in hours and days instead of milliseconds.

117
9
DSL and Cable Modems: Last Mile on Cop-
per
9.1 Modems
Modems are used to represent binary digits, that is, 1s and 0s, on copper wires for the last
mile in brownfields, i.e. neighborhoods where (brown) copper is already installed. Last mile
means the access circuit, i.e. the connection from the customer premise to the network.
Historically, this was POTS on twisted pair between the customer premise and the CO as cov-
ered in Section 2.2, but is equally applicable to cable companies with coaxial copper wire infra-
structure.
This chapter begins with general principles of modems and modulation, then covers DSL
modems for twisted pair and cable modems for coaxial cable in detail.
9.1.1 Why Bother With Modems?
One question often asked is, “Why bother with modems? Why not transmit the information ‘dig-
itally’?”
In other words, “Why not use pulses of voltage like on LAN cables, for example, +3 volts ap-
plied to the line for a short time to represent a “1” and -3 volts to represent a zero?”
The answer is that modems are required to represent 1s and 0s on circuits that are restricted
to a range of frequencies that does not include 0 Hz. This is often called a pass-band, since
there is a range of frequencies that will be passed and everything else is suppressed.
For example, with POTS, the telephone company implements the service with twisted pair ter-
minated on a line card on a telephone switch.
The line card has a simple electrical circuit on it called a filter that blocks the transmission of
energy at any frequencies outside the voiceband, that is, any frequencies less than 300 Hz or
greater than 3300 Hz. In other words, the pass-band for POTS is 300 - 3300 Hz.
To represent 1s and 0s as pulses of voltage, which would look something like a square wave
on the line (Labeled “IN” on Figure 75), would require a component at 0 Hz to be able to repre-
sent the steady-state voltage that is the top of the pulse, and many components at frequencies
higher than 3300 Hz to represent the sharp transitions of voltage.

Figure 75. Pulses on a Pass-Band Channel

But the available pass-band does not include 0 Hz – the voltage must vibrate at least 300
times per second – so there is no way to hold the voltage steady with zero changes per sec-
ond at the top of the pulse.
Similarly, the pass-band does not support the high frequencies necessary to represent the
sharp transitions at the square corners of the pulses.
What would make it through the voiceband filter would be nothing, plus small perturbations at
frequencies less than 3300 Hz when the voltage was abruptly changed from one value to the

118
other, as illustrated with the line labeled “OUT” on Figure 75.
With line noise added, it is very difficult to reliably detect the pulses, and thereby communicate
1s and 0s using pulses on this kind of circuit.
The situation for higher-frequency channels such as those employed for DSL, cable modems
and all kinds of radio is the same.

9.2 Modulation Techniques

9.2.1 Modulation of Carrier Frequencies
A design that will work on a pass-band channel is one that employs tones or carrier frequen-
cies with the pass-band.
By varying the amplitude (level or volume) of a carrier within the supported frequency pass-
band, the frequency of such a carrier, its phase, or combinations of these, signals that make it
through the pass-band are used.
The technique of representing binary digits by varying one or more characteristics of one or
more tones is called modulation, and the circuit-terminating equipment that performs this func-
tion is called a modem, a contraction of modulator and de-modulator.
The fundamental concepts are identical for DSL, cable modems, and all kinds of wireless
modems including cellular and WiFi… only the frequency pass-band changes.
9.2.2 Amplitude Shift Keying (ASK)
Amplitude Shift Keying (ASK) is the simplest technique for representing binary digits using
tones. A single carrier frequency is defined, and one volume or amplitude is used to represent
a “1” and another amplitude is used to represent a “0”.

Figure 76. Amplitude Shift Keying

These two states of the carrier are the signals that will be transmitted. Signals are also referred
to as symbols.
Note that this is not continuous amplitude modulation (AM) as is used on radio stations. At the
radio station, a carrier frequency of something like 800 kHz is selected, and the amplitude at
that frequency is varied continuously to represent the sound pressure waves coming out of the
disk jockey’s throat. ASK uses only two amplitudes.
This technique is susceptible to noise.
Most noise, like that from microwave ovens and fluorescent lights is additive; noise adds to the
signal. In the example of Figure 76, when transmitting a low amplitude to represent a 0, some-
times the noise adding to the signal will increase the amplitude to the point where the receiver
detects a high amplitude and so in error outputs a 1.
9.2.3 Frequency Shift Keying (FSK)
The design could be improved by using not one carrier frequency and varying its amplitude,
but by picking two carrier frequencies and keeping the amplitude constant instead.

119
Since noise (more or less) affects all frequencies equally, this technique, called Frequency
Shift Keying (FSK), is more robust in the presence of additive noise.

Figure 77. Frequency Shift Keying

A typical design for the voiceband would pick one frequency, for example, 1200 Hz to repre-
sent “0” and another frequency, 2200 Hz to represent a “1” (or the other way around, if you
prefer). These two frequencies are the signals that are sent over the line. The transmitter shifts
back and forth between these two signals to represent 1s and 0s.
The rate at which it is possible to shift back and forth between the frequencies and reliably de-
tect the result is limited by the width of the pass-band and the shifting technique.
Since the maximum frequency allowed on the voiceband is 3300 Hz, the maximum rate for sig-
naling, that is, shifting back and forth between the two signals, would be 3,300 times per sec-
ond.
However, by back and forth between two frequencies at a third frequency, unwanted frequen-
cies called harmonics are created, which has the effect of reducing the maximum practical sig-
naling rate to about 2,400 times per second on a voiceband circuit.
Since there is one bit communicated every time a signal is sent, this means a data rate of 1 bit
per signal x 2,400 signals per second = 2,400 bits/second = 2.4 kb/s.
9.2.4 Phase Shift Keying (PSK)
Phase shifting works better than amplitude shifting or frequency shifting.
The phase of a signal is its position with respect to a time reference. To perform phase shifting,
one carrier frequency is used, with a constant amplitude, but the position of the signal is shifted
back and forth to convey information. In effect, there is one single pure tone at one amplitude,
and a jitter added to carry the information. It is easy to detect jitters.

Figure 78. Phase Shift Keying

Figure 78 illustrates Differential Phase Shift Keying (DPSK), which is the simplest technique:
the phase of the signal is changed 180 degrees (it is shifted by half a period) to indicate a “1”,

120
and nothing to indicate a “0”.
Since the frequency is constant, the design of the receiver is simpler, and there is less har-
monic noise. Since the amplitude is constant, the technique is less susceptible to added noise.
Overall, the error performance of PSK is about twice as good as FSK.
9.2.5 Define More Signals to Communicate More Bits
Voiceband modems that can establish a data rate of more than 2.4 kb/s exist. V.91 standard
modems can achieve 53 kb/s in the voiceband.
Can this be achieved with FSK and switching back and forth between the two frequencies
53,000 times per second?
No. The maximum signaling rate on a voiceband circuit is around 2,400 signals per second.
Any more often is trying to use more bandwidth than 300-3300 Hz, which is not passed by the
filters, and the received signal will be distorted to the point where the receiver can not reliably
detect it.
The maximum signaling rate is directly related to the width of the pass-band. Once the maxi-
mum signaling rate is achieved, the key to greater data rate is not faster signaling, but more
signals.
9.2.6 Quadrature PSK (QPSK)
If four phase shifts are defined, each of 90 degrees (1/4 of a period), this yields four possible
signals that might be conveyed to the receiver.
Since there are four signals, each can be used to represent 2 bits. Writing out the numbers be-
tween 0 and 3 in binary will demonstrate this, as illustrated in Figure 79.
By making a choice of one of the four signals in particular and transmitting that signal, two bits
are communicated with one signal.
Since there are four signals, this modulation scheme is called Quadrature Phase Shift Keying
(QPSK).
In the example of the voiceband channel, we could continue signaling at 2,400 signals per sec-
ond, but now with 2 bits conveyed by each signal, the data rate is 4,800 bits/second.
On a 4G cellular (LTE) system, the signaling rate is 15,000 symbols per second per subcarrier.
QPSK, one of the allowed modulation schemes for LTE, would yield 30 kb/s per subcarrier.
As can be seen on the lower left side of Figure 79, drawing pictures of signals with different
phases using Cartesian (x,y) coordinates becomes tedious and uninformative, particularly as
the number of signals increases.

121
 Figure 79. Quadrature Phase Shift Keying (QPSK)
More often, polar coordinates are used to represent the signals. As illustrated on the lower
right of Figure 79, each signal is represented by an arrow, where the length of the arrow is the
amplitude and the rotational angle represents the phase shift.
The diagram is called a phasor diagram; the arrows are the phasors. This is the source of the
Star Trek expression “phasors on stun”.
9.2.7 Quadrature Amplitude Modulation (QAM)
Why stop at a repertoire of four signals? Why not define a million signals, using combinations
of amplitude, frequency and phase shifting, so each signal conveys log2(1 million) = 20 bits?
The answer is errors. The more signals defined, the higher the probability of making an error at
the receiver deciding which signal was transmitted, and so the effective error-free data rate de-
creases.
Quadrature Amplitude Modulation (QAM) is a technique that combines phase shifting and am-
plitude shifting to generate a repertoire of many signals that could be transmitted, to increase
the number of bits indicated by each signal.
QAM-16 uses combinations of phase and amplitude shifting to define 16 signals, evenly
spaced in a square as illustrated in Figure 80.
Since 16 = 24, each signal conveys 4 bits.

122
In the voiceband, QAM-16 at 2400 signals per second would yield a data rate of 9,600 bits per
second, a popular modem standard implemented in a fax machine.
QAM-16 and QAM-64 are modulation techniques specified in the 4G cellular LTE standard.

Figure 80. Quadrature Amplitude Modulation (QAM)

Phasor Diagram with 16 signals

9.2.8 Constraints on Achievable Bit Rate

The rate at which bits can be transmitted is proportional to the frequency bandwidth times the
signal to noise ratio.
Noise sources are in many cases external and can not be controlled.
Signal strength is limited due to crosstalk - interfering with other signals on other pairs in the
same cable.
The remaining variable is the width of the frequency band.
To allow the representation of more bits per second, a wider frequency band is required.

9.3 Digital Subscriber Line (DSL)

The need for speed is never-ending. Modulation techniques used by modems that operate
within the frequency band defined for POTS – the voiceband, 300 to 3300 Hz – hit practical
limits on the number of bits per second that can be reliably communicated.
To achieve more bits per second, a wider frequency band is required.
While it would be theoretically possible to implement a wider frequency band by increasing the
size of that defined for POTS, such a move would be unworkable due to the extremely high
number of devices with voiceband filters that would become incompatible.
Pulling new fiber cables to every home will be a long process – so the existing twisted pair
copper entry cable to residences in older neighborhoods (brownfields) must be used for some
time to come.
However, any technology deployed on existing twisted-pair copper phone lines must be back-
wards-compatible with POTS.
9.3.1 DSL: Modems Above The Voiceband

123
The solution for brownfields is the definition of a second, wide frequency band in which
modems can operate above the voiceband on the existing copper twisted pair. This is called
Digital Subscriber Line (DSL) technology, and allows broadband (high bit-rate) communica-
tions of 1s and 0s while still supporting POTS on the same line.
There are a number of DSL modulation techniques, each employing different bandwidths and
signaling schemes, with different requirements for loop characteristics and providing different
numbers of bits per second… and improving all the time.
Calling this “digital” is inaccurate. DSL does not use pulses, which is the definition of digital
transmission; instead DSL employs modulation in frequency channels above the voiceband…
a technique more associated with the terms “modem” and “analog” than digital.

Figure 81. DSL: Broadband modems operating in a wide frequency band above the voiceband on the existing
twisted pair loop.
“Broadband modems operating in a wide frequency band above the voiceband on the existing
twisted pair loop” would be more accurate... but of course, “Digital Subscriber Line” sounds
better.
9.3.2 ADSL, SDSL and XDSL
When the downstream capacity (towards the user) is larger than the upstream capacity, it is
called Asymmetric DSL (ADSL). Symmetric DSL (SDSL) has the same capacity in both direc-
tions. The term “XDSL” is used to generically refer to the idea of broadband modems on
twisted pair, regardless of the variety.

9.4 DSLAMs
This diagram illustrates the equipment used for DSL. At the customer premise, the DSL mo-
dem is connected to the twisted pair loop, which is connected to a Digital Subscriber Line Ac-
cess Multiplexer (DSLAM).
The DSLAM contains the DSL modem to which the DSL modem at the customer premise is
communicating, as well as multiplexing equipment and a fiber backhaul to the network core
and eventually to the Internet.
The DSLAM was originally located in the CO. To shorten the distance between the modems, to
be able to increase the bit rate achieved, the DSLAM is now typically deployed in an outside
plant enclosure as a type of remote fiber terminal. This is a type of Fiber to the Neighborhood
(FTTN).
9.4.1 DSL Modem Hard-Wired to Loop
DSL is markedly different than the old “dial-up” voiceband modem connections, where the dial-
up modem makes a phone call and a circuit-switched connection through the telephone switch
to a far-end modem, for the duration of the communication session.

124
 Figure 82. Fiber to the Neighborhood and DSLAM.
DSL modems on existing twisted pair DSLAM to customer..
With DSL, the customer’s modem is communicating with a modem connected to the other end
of the customer’s loop.
There is no connection through telephone switches; the DSL modems are hardwired together.
This avoids the filters on the CO switch line card and allows the use of a wider frequency band
by the modems, and hence more bits per second.
DSL service is referred to as always on: the DSL modem at the residence is always connected
to the DSL modem in the DSLAM. The connection between the modems is not broken after
each communication session like with circuit-switched or dial-up modems in the voiceband.
9.4.2 Coexistence with POTS
The telephone puts energy on the line in the POTS voiceband, and the DSL modem puts en-
ergy on the line in bands at higher frequencies. Since these are separated in frequency, the
DSL modem does not interfere with telephone service the way a voiceband modem does.
Since the DSL band is much wider than the voiceband, it is possible to communicate more bits
per second: tens of Mb/s as a standard service offering today and improving all the time.

9.5 Fiber to the Neighborhood (FTTN),

DSL to the Premise

125
 Figure 83. Fiber to the Neighborhood (FTTN)

9.5.1 Loop Length

Impairments on twisted pair copper are per foot. As the distance between the customer DSL
modem and the network DSL modem in the DSLAM increases, impairments like noise, attenu-
ation and capacitance increase, limiting the achievable bit rate.
To achieve higher bit rates, it is necessary to shorten the distance between the customer DSL
modem and the network-side DSL modem. The simplest way to do this is to locate the DSLAM
in an enclosure in the outside plant.
9.5.2 Remote DSLAMs, OPI and SAC Boxes
Remote DSLAMs are often deployed in cabinets or enclosures bolted onto existing enclosures,
which would be the Outside Plant Interface (OPI) or Subscriber Area Concept (SAC) boxes in
neighborhoods, spaced hundreds of meters apart and serving perhaps 200 customers each.
This provides access to the subscriber loop at a point close to the customer, effectively short-
ening the distance between the two modems to allow higher bit rates.
The remote DSLAM is connected to the CO with a fiber to carry the data and copper wires to
carry electricity to power the DSLAM.

9.6 DSL Standards

There are many variations of DSL, achieving ever higher bit rates on twisted pair. Very High Bit
Rate Digital Subscriber Line (VDSL) has existed in laboratories for decades (hence the nerdy
name). Recently, signal processing technology has increased, and costs have decreased, to
the point where VDSL is now routinely deployed.
The different flavors use different modulation and signal processing techniques, different fre-
quency bands and different distance limits.
9.6.1 ADSL2+
ADSL2+, standard G.922.5-2003 from the ITU, is an older high bit rate DSL technology, utiliz-
ing a frequency band of up to 2.2 MHz on the loop to implement a maximum of 12 Mb/s down-
stream and 1 Mb/s upstream with a maximum distance of 4000 feet between the modems.
ADSL2+ employs a legacy technology called ATM to aggregate the traffic on the fiber back-
haul. Since the core is now Optical Ethernet and ATM is being discontinued, a gateway or con-
verter is required to interface ADSL2+ DSLAMs to the core.
9.6.2 VDSL2

126
Newer DSL modems implement VDSL2, standard G.992.3 from the ITU. This standard uses
up to 30 MHz of bandwidth on the subscriber loop and can achieve 100 Mb/s symmetric at 500
feet.
Bonding and vectoring increase the bit rate and/or maximum range.
Optical Ethernet and VLANs are used to aggregate the traffic on the fiber backhaul.
9.6.3 VDSL2 Frequency Bands and Profiles
The frequency band for DSL modems is broken into a number of smaller bands, some for up-
loading and some for downloading.

Figure 84. VDSL Up and Down Frequency Bands

The higher the bandwidth, the more bits per second can be communicated, but also, the
shorter the usable reach.
Figure 84 illustrates the bands and their names and usage.
For comparison, ADSL2+ uses only a bit more than 2 MHz of bandwidth, barely showing up in
the D1 band in this chart.
In VDSL2, a profile means a modem that uses bandwidth up to the specified number of MHz.
9.6.4 Bonding
Bonding – using two pairs – increases the reach, i.e. the maximum distance between the two
modems.
9.6.5 Vectoring
Crosstalk – the noise created on a pair of wires by a signal on another pair of wires in the
same cable – is a major impediment to achieving very high bit rates with DSL.
Vectoring is a term used to describe very sophisticated signal processing that cancels
crosstalk. The crosstalk is determined by examining the signals on all of the pairs in a cable,
then the crosstalk on each pair is mathematically subtracted from the modem signal on that
pair.
By lowering the noise, vectoring allows the achievement of higher bit rates.

127
 Figure 85. VDSL Profiles, Speeds and Ranges

9.7 Broadband Carriers: FTTN & Broadband Coax to

the Premise
Cable TV distribution systems were originally known as Community Antenna Television (CATV)
systems.
For a city, a television signal would be received by an antenna located at a building called the
Head End, then distributed to customers in the city via coaxial cable or coax, which is two cop-
per wires, one inside the other.

128
 Figure 86. CATV FTTN and Coax to the Premise
Today, Head Ends are connected to other Head Ends in other cities with a fiber backbone for
digital content distribution.
Coax supports a much broader bandwidth than twisted pair, so these systems can be called
broadband systems and the operating companies broadband carriers.
9.7.1 Hybrid Fiber-Coax Network
This is implemented with a combination of fiber to the neighborhood then coaxial copper cable
for the last mile, and so CATV systems are also called Hybrid Fiber-Coax (HFC) systems.
An HFC network consists of a Head End, fiber to the neighborhood terminated on Optical Net-
work Units, coaxial copper feeder cables running down streets, amplifiers, splitters, taps and
drop lines into customers’ homes where a converter and television are located – along with
computers and telephones.
9.7.2 Frequency Channels
In the previous millennium, Cable TV networks carried multiple analog video signals. The am-
plitude of the video signal is an analog of the intensity of the light at a point on the screen as it
is being scanned along lines left-to-right and top-to-bottom.
The American NTSC standard scans half the screen sixty times per second, resulting in a sig-
nal about 4 MHz in bandwidth.
The bandwidth on coax is at least 450 MHz and up to 3 GHz. Many carriers currently use up to
1 GHz.
To make this wide bandwidth usable, it was divided into smaller 6-MHz frequency bands called
channels by equipment in the Head End. This technique is called Frequency Division Multi-
plexing (FDM).

129
 Figure 87. 6-MHz CATV channels
For standard-definition analog service, the Head End gathered video signals from satellites,
terrestrial antennas and local content sources and placed a video signal in each channel using
Vestigial Side Band Amplitude Modulation (VSB-AM). The video signal is combined with a sin-
gle pure sine wave at the frequency of the desired channel, called a carrier frequency, to shift
the video signal up to the channel frequency.
For those who like details: this actually creates two video signals at frequencies on either side
of the carrier. One of the copies and the carrier are suppressed, meaning one copy remains
(the vestigial copy) on one side of the carrier (the side band).
9.7.3 Fiber Serving Area
This entire group of signals, all channels together, is transmitted to the neighborhood using
analog techniques on fiber.
The fiber terminates on an Optical Network Unit, located in a cabinet on a pole or on some-
one’s front lawn, where the signals are transferred to copper coaxial feeder cables that run
down streets.
The coverage of the coax cables terminating on one ONU is called a Fiber Serving Area
(FSA), typically passing 200 – 500 homes.
Taps are installed on the feeder cables at regular intervals. A copper coaxial drop wire is in-
stalled from one of the connectors on the tap to the residence.
This has the effect of physically connecting all of the users together and to the ONU, in an ar-
chitecture similar to the original LAN bus topology. The electrical signal placed on the coax is
broadcast to everyone tapped onto the cable.
Since attenuation is more severe on coax than on twisted pair, and more severe at high fre-
quencies, amplifiers are used to boost the signal. Amplifiers are spaced typically every 660
feet, about one amplifier per block.
9.7.4 Television Converters
At a customer, the converter picks the desired frequency channel out of the entire lot, and
shifts the signal there back down to the “natural” frequency range 0 - 4 MHz so that it can be
displayed on the screen of the television. This “natural” range is called the baseband, or some-
times Channel 1.
In days past, a Video Cassette Recorder (VCR) was sometimes used as the converter device,
downshifting the desired channel to Channel 3 and the television’s tuner would downshift that
to Channel 1 and display it.
9.7.5 Modems on CATV Channels
Once a cable TV system is in place, there is no reason why it has to be used only for analog
video signals.
Modems may be attached to each end of the system, and one or more channels on the CATV
system used for communication of the modem signals.
These modems signal 1s and 0s that can be digitized video, Internet traffic and VoIP telephone
service.
“Digital” cable, for example, an HD channel, is video that has been digitized, turned into a
stream of 1s and 0s, which are transmitted from the Head End one-way to the set-top box us-
ing a modem.

130
9.7.6 Two-Way Communications Over Shared Access
For Internet access, VoIP telephone service or any other two-way communications over this in-
frastructure, modems are required for each direction.
The main obstacle is the fact that the access circuit is a multi-drop architecture: everyone on
the street is connected to and sharing the same cable… methods of rationally sharing the com-
mon communication channels are required.
One strategy would be to allocate two 6-MHz channels on the system for each user: one for a
modem for uploading and one for downloading, with corresponding modems at the Head End.
That would be a very inefficient way to allocate capacity… there are far fewer available chan-
nels than users in a fiber serving area, and 12 MHz of bandwidth is reserved for a subscriber
whether they are actually using it or not.
In practice, the users share channels. In many cases, bandwidth is allocated above the televi-
sion channels for downloading, and below the television channels for uploads from sub-
scribers. This is called a high-low split strategy and makes amplifier deployment easier.
In the downstream direction, a modem at the Head End broadcasts traffic intended for a cus-
tomer to everyone in their neighborhood. The traffic is encrypted by the Head End and de-
crypted in the device containing the customer’s cable modem. The Head End broadcasts
users’ data as needed as it arrives, or in a rotation if there are many active users.
Sharing a modem band in the upstream direction is more difficult. The users’ modems are all
tuned to transmit on the same channel, so they can’t all transmit a simple modem signal to the
Head End at the same time.

9.8 DOCSIS and Cable Modem Standards

9.8.1 DOCSIS 1: Contention-Based Channel Sharing
The first strategy for sharing, standardized as the Data over Cable System Interface Specifica-
tion (DOCSIS), defined time slots.
Customers’ downloads are encrypted and broadcast to everyone in the Fiber Serving Area.
For uploads, time slots were available for contention by all modems. Collisions can occur and
retries are used, similar to the contention-based capacity-sharing strategy for Ethernet.
This worked fine if only one person in a Fiber Serving Area has signed up for cable modem
service. The problems started when 50 people signed up, most of them teenagers running bit
torrent or other file “sharing” programs, continuously passing on search requests and starving
other users of bandwidth. These problems only get worse with the massive increase in traffic
from Netflix and youTube.
9.8.2 DOCSIS 2: Reserved Time Slots on Channels
DOCSIS 2 defined time slots allocated to specific users to guarantee performance, similar to
the TDMA strategy for 2G cellular.
9.8.3 DOCSIS 3: CDMA on Channels
To improve efficiency, DOCSIS 3 specifies the use of CDMA allowing variable rates and multi-
ple simultaneous transmissions like 3G CDMA cellular.
9.8.4 DOCSIS 3.1: OFDM
DOCSIS 3.1 specifies OFDM like 4G cellular.
9.8.5 Wider Channels
Basic cable modems move 30 Mb/s in the 6-MHz channel defined decades ago for standard-
definition analog television.
With DOCSIS 3.0 and 3.1 systems, the CDMA and OFDM technologies can use bands wider
than 6 MHz to deliver much higher bit rates to customers., heading toward Gb/s.

131
10
Wireless
10.1 Radio
When we say “wireless”, we generally mean the use of radio, which is electromagnetic waves
at frequencies measured in Gigahertz (GHz), that is, vibrating 109 or a billion times per second.
We could, in theory, be discussing electromagnetic energy vibrating on the order of 1014, hun-
dreds of trillion times per second (this is called light); but one of the problems in wireless com-
munications is obstacles.
It turns out that the higher the frequency, the longer distance it takes for energy to refract or
bend around an object.
Light does refract around objects – this is how we can tell there are planets around other suns
– but the length of the shadowed area behind the object is too long for use on a terrestrial
scale.
If we reduce the frequency of the energy, it refracts at a sharper angle and so the length of the
shadow behind an obstacle shortens.
In addition, lower-frequency energy can penetrate through objects like walls and clouds more
easily (there’s a reason why fog horns are very low frequency).
For these reasons, we tend to use energy at Gigahertz frequencies, two or three hundred thou-
sand times lower than light, and call it radio.
This chapter covers communications centered at Gigahertz frequencies, in frequency bands
with widths measured in Megahertz (MHz).
Radio is used in many different kinds of systems with different applications, including every-
thing from demagogues broadcasting angry rants on talk radio shows using analog AM, to mo-
bile cellular systems for telephone calls, messaging, youTube and anything else on the Inter-
net, trunked radio for police communications, fixed wireless to remote residences, short-range
wireless LANs, geosynchronous communication satellites, Low Earth Orbit satellites and more.

132
 Figure 88. Lower frequencies result in shorter shadows
Video broadcast, two-way voice communications and point-to-point digital microwave commu-
nications were the biggest applications for radio in the past.
Mobile voice and Internet access is a big business in the present.
In the future, wireless will be ubiquitous.
To represent information, we could take a single pure frequency (called a carrier frequency)
and vary the amplitude (volume) of the carrier frequency in a continuous fashion as an analog
of the sound coming out of the speaker’s mouth.
Or we could vary the frequency of the carrier as an analog of the sound.
These are called Amplitude Modulation (AM) and Frequency Modulation (FM) respectively.
When we wish to represent 1s and 0s, we have a more complex task. Since radio bands do
not include zero Hertz, sometimes called DC, pulses can not be used to represent 1s and 0s
as on copper wires.
Instead, it is necessary to use techniques similar to those used in telephone line modems to
represent the 1s and 0s, such as shifting back and forth between specific amplitudes, frequen-
cies or phases, or combinations thereof.

10.2 Mobile Networks

This section covers the basic components and operation of a mobile communication network.
Mobile network is the term given to distributed radio systems designed so that many users,
who may or may not be moving around, can share a radio band and communicate amongst
themselves and to other people or computers on the Internet or the PSTN.
10.2.1 0G: The Mobile Phone System

133
The first kind of radio systems connected to the PSTN were called MPS: the Mobile Phone
System.
These employed radios in automobiles fitted with big whip antennas to communicate to base
stations in large metropolitan areas.
The caller had to call a “mobile operator” and ask for a particular “mobile number”, and would
(maybe) be patched through.
The geographical areas where service was available – the coverage – was very limited.
There was very little capacity – not many people could use the system at the same time.
And ironically, it did not support mobility: once the call was connected… if the person with the
mobile radio drove too far away from the base station, the call would be dropped.
10.2.2 Mobility
The definition of mobility is having the ability to start a communication session using a terminal
communicating with a particular antenna, then move away from the antenna and not lose com-
munications, but rather be handed off to another antenna.
10.2.3 Base Station, Cell, Airlink and Handset
The cellphone, called a mobile, terminal or handset, is connected to the network via an airlink
to a base station.

Figure 89. Components of a Base Station

A base station includes the Base Station Transceiver (BST), also called a Base Transceiver
Station (BTS), in an enclosure at ground level. This rack-mount equipment produces energy vi-
brating at radio frequencies in the form of electricity.
It is connected with thick coaxial cables to antennas that convert the electricity to electromag-
netic waves vibrating at the same frequencies, which for some reason then propagate away at
the speed of light.
The area on the ground covered by the base station is a cell.
10.2.4 Mobile Switch
The base stations are connected to a Mobile Telephone Switching Office (MTSO). This is a
building with routers, traditional telephone switches, connections to networks: the Internet,
IXCs, LECs, Content Delivery Networks, as well as to other MTSOs for mobile-mobile calls.

134
In the MTSO, routers direct Internet traffic to an ISP, or to a specific content provider. The
routers direct phone calls to a mobile switch.

Figure 90. Mobile Network Components

A mobile switch is a telephone switch with additional capability to keep track of users moving
between base stations.
In the mobile switch is a database called the Location Register (LR). The Location Register is
used to keep track of users and where they are (or last were) via radio control channels, Elec-
tronic Serial Numbers (ESNs) and Subscriber Information Module (SIM) cards.
10.2.5 Backhaul
The connection from the base station back to the mobile switch to connect to the network is
called the backhaul.
In some cases, the base station will be connected to the mobile switch with fiber, particularly if
the operator is an affiliate of the phone company, the ILEC.
In other cases, a point-to-point microwave link will be used to connect one base station to an-
other that has a connection to the mobile switch. This method of backhaul is often used by
their competitors.
The mobile switch is connected to the Public Switched Telephone Network (PSTN), to allow
calls to landlines.
Separate connections are made to Internet Exchange (IX) buildings for connections to Tier-1
Internet Service Providers, peer ISPs and networks and to content delivery networks.
10.2.6 Registration and Paging
On power-up and periodically thereafter, the handset registers with the switch, which duly
records the handset’s location, or more accurately, the base station the handset is using, in the
LR.

135
When there is an incoming call, the mobile switch will page the handset from the base station it
last registered on.
If the handset does not answer the page, the network will resend the page from all base sta-
tions in the area or in some cases, all of the base stations on the network.
Once the handset answers the page and the user presses the “talk” or call button, voice com-
munications take place over the radio airlink to the base station, then over the backhaul to the
mobile switch.
For a mobile to mobile call, the communications will be routed to a base station. For mobile to
landline, the call will be routed to the PSTN.
10.2.7 Handoff
If a user moves during a call, at some point, the user will be handed off from one base station
to another.
This means that the network will switch to using a different base station to communicate to the
handset, and, depending on the technology employed, may involve changing the radio fre-
quency of the handset.
The handoff implements mobility – the ability to maintain communications while traveling.

10.3 Cellular Radio and 1G

This section covers:
• The requirements on the communication system: mobility, coverage and capacity;
• The idea of a cellular radio system, and how it is used to meet the coverage requirement;
• How frequency-division multiplexing was used to meet the capacity requirement in the first
generation of “cellular”, called AMPS in North America;
• The implications of a handoff to implement mobility
• The limitations of the first generation and room for improvement.
10.3.1 1G: The Advanced Mobile Phone System
To meet the requirements of coverage, capacity and mobility, cellular radio systems were de-
ployed.
The first generation of cellular, the improvement on MPS, was called the Advanced Mobile
Phone System (AMPS).
Radio frequency bands or spectrum was allocated for this service by the federal government
around 800 MHz in North America, and 900 MHz in many other countries.
In North America, a block equal to half of the spectrum was given to an affiliate of the incum-
bent telephone company in each market (which in an Orwellian piece of jargon is called the
wireline cellular) and half was given to a competitor.
An operator that was given a block of frequencies would need a real estate department to find
locations where they could construct the base stations, which included, in the first generation,
large ugly towers to support the antennas.
10.3.2 Cells
An operator would divide their block of frequencies into seven groups of frequencies, then at a
base station, tune the base station transceiver to use one of the seven groups of frequencies.
The radio coverage area around the tower would be something like 3 miles or 5 kilometers in
radius. On the ground, this is the cell.

136
 Figure 91. Radio coverage in cells, each using a different group of frequencies.
Figure 91 illustrates a cell centered on Menlo Park, CA in Silicon Valley.
Then the operator would find another location six miles or ten kilometers away, and build a
second base station, using a second group of frequencies… for example, in Fremont across
the Bay.
This pattern could be continued to build seven base stations, using all seven groups within the
block of spectrum available.
At that point, the operator would have coverage in the geographical area illustrated, all the way
over to Cupertino CA where Apple is headquartered, and all of their spectrum would be used.

137
 Figure 92. Frequency re-use. Using the same frequencies in two cells spaced miles apart.

10.3.3 Frequency Re-Use

Then, the operator could find another geographic location, for example, near Woodside CA,
where Neil Young has a 1500 acre ranch in the middle of some of the nicest real estate on the
planet, and convince Neil to let them build an eighth tower and base station, where they could,
in this example, re-use frequency group #7.
Since the second tower is in Woodside, more than 20 miles away from Cupertino, and the ra-
dios have a relatively short range, we can re-use the same frequency group, and the base sta-
tions will not interfere with each other.
This is the idea behind a cellular radio system: being able to re-use the same frequency
groups over and over again in different geographic locations, to meet the coverage require-
ment.

138
 Figure 93. Frequency re-use in cells to meet the coverage objective

10.3.4 Analog on Radio Channels

To meet the capacity requirement, in AMPS, the group of frequencies used in a cell was di-
vided into 45 sets of 30 kHz radio channels, and each user in a cell is assigned a set of radio
channels.
Voice or modem signals were represented on the radio channel using continuous frequency
modulation of the channel frequency.
This analog technique resulted in middling- to poor-quality voice communications… and there
was no encryption or coding of the voice, allowing eavesdropping with relatively simple equip-
ment.
AMPS was also not so good for data communications using a modem over the analog radio
channel.
10.3.5 AMPS Handoffs
When a user drove too far away from a base station, they had to be handed off to another
base station.
Two things have to happen: since each cell employs different frequencies, the cellphone has to
change which frequency it’s operating at; and the network has to hand off the phone call from
one base station to another.
This takes about 0.2 seconds, during which the communications will be interrupted or muted.
This was a bit annoying for voice; it caused users’ modems to disconnect.
Several strategies were attempted to keep the modems from hanging up, but in practice, it was
necessary to “pull over” to send a mobile fax.
Unfortunately, even if you are stationary, you can still change cells: if someone drives in the far
side of your cell, you might be bumped to the next (overlapping) cell to make room for them.
10.3.6 AMPS Capacity
Another problem with AMPS was low capacity meaning a high per-channel cost to the carrier.
There are 45 sets of frequencies per cell, so in theory, there could be 45 users per cell. In
practice, it’s 40, because it is necessary to keep some channels free for people driving into the

139
cell.
40 users, 3 miles radius… about 1.5 users per square mile… doesn’t work so well in Silicon
Valley, Manhattan or just about anywhere else.
Sectorization, using antennas with shaped beams to create triangular-shaped cells (called sec-
tors to confuse people) can be employed to improve capacity; but not enough for the immense
popularity of mobile communications. We had to move on to better technologies.

10.4 Second Generation: Digital Cellular

In this section, we will:
• Review the second generation of cellular: GSM and PCS
• How the second generation was digital
• What “digital radio” means and how it is implemented
• Mobile phone calls
• Mobile Internet access.
In an upcoming section, we will cover the details of the different spectrum-sharing technologies
that were employed for the second generation: TDMA, GSM and CDMA.
10.4.1 PCS and GSM

Figure 94. 2G
The second generation of cellular technology employed lower power, smaller cells and imple-
mented digital communications.
It was in some cases referred to as Personal Communication Services (PCS), and in many
places, the Global System for Mobile communications (GSM).
The advantage of implementing digital communications is better sound quality, better signaling
and control capability, and mobile access to the Internet and other networks.
Second-generation cellular (2G) was initially deployed in North America on frequency bands
centered around 1.9 GHz, whereas AMPS was deployed on frequency bands centered around
800 MHz.
Handsets were dual-mode, meaning they could support both AMPS at 800 MHz and PCS at
1.9 GHz.
Several different technologies were deployed by different carriers for spectrum-sharing for the
second generation.
These included techniques called CDMA or Code Division Multiple Access and TDMA or Time
Division Multiple Access (TDMA) in North America.
In the rest of the world, a 2G TDMA scheme called GSM, the Global System for Mobile com-
munications was widely deployed.
The fundamentals of TDMA and CDMA are covered in an upcoming section.

140
10.4.2 Digital Cellular Radio
Putting aside for the moment the discussion of TDMA vs. CDMA, we’ll first understand how
digital cellular radio works.
The cellphone contains a microphone, which creates a voltage that is an analog of the strength
of the sound pressure waves at the microphone.
This analog signal is fed into a codec or vocoder inside the terminal that digitizes the analog
waveform.
Additional complex digital signal processing may performed.
The result is 1s and 0s representing the digitized speech.
Then a modem that operates at radio frequencies is used to represent those 1s and 0s using a
modulation technique such as Quadrature Amplitude Modulation within the radio band.
This modem waveform is broadcast into space by the antenna on the cell phone.
At the base station, an antenna detects the radio waves and feeds them into a modem in the
Base Station Transceiver that interprets it and produces 1s and 0s.
Complex signal processing is performed on those 1s and 0s to extract the original digitized
speech.
This digitized speech is then backhauled or transmitted back to the mobile network’s switch via
a Base Station Controller, where it can be routed to the PSTN for a mobile-to-wireline call, or
routed to another base station for a mobile-to-mobile call.

Figure 95. Digital Mobile Voice Communications

The speech is digitized to somewhere between 9 and 13 kb/s for transmission over the airlink.
This is not a standard technique, but can be used since the communications are on a private
wireless network.
For mobile to wireline calls, the speech has to be converted to the 64 kb/s DS0 rate for inter-
connection between the mobile network and PSTN wireline network

141
10.5 Mobile Internet and “Data” Plans
The great thing about digital cellular is that we can take advantage of its inherent capability to
move bits to communicate not just digitized speech, but 1s and 0s that are representing e-mail
messages, web pages, video, music or literally anything else.
In this section, we’ll understand how the system designed to carry digitized speech using
modems can be employed to carry anything coded into 1s and 0s.

Figure 96. “Data”: Mobile Internet

In a quaint, old-fashioned use of terminology, all traffic apart from telephone calls, Short Mes-
sage Service (SMS) text messages and network control messages has been referred to as
“data” in cellular billing plans.
This is done so the carrier can bill customers twice: once for voice minutes, and a second time
for a “data plan”.
“Internet traffic” would be a more accurate term than “data”, as this category includes for exam-
ple Skype VoIP telephone calls, Skype videoconferencing, youTube videos, Netflix videos,
Google maps data, app downloads and updates, web pages and facebook along with count-
less other applications, all communicating to another device over the Internet.
10.5.1 Cellphone as a Tethered Modem
This diagram provides an addition to the previous block diagram showing how the modem in
the cellphone and the modem in the base station can be used to move data from a computer
to the Internet via the cellphone.
When using a laptop connected to a cellphone with a cable as illustrated, the cellphone is said
to be acting as a tethered modem, i.e. physically tied to the computer.
10.5.1.1 USB Cable Tethering
One way to implement this is to connect a computer to a data port on a cellphone with a USB
cable.

142
The computer then sees the cellphone like an external modem, very much as if one had
plugged a landline modem into the computer.
10.5.1.2 Wireless Tethering with Bluetooth
One could also connect the computer to the cellphone with a short-range Bluetooth wireless
link, which is completely separate from the cellular radio.
10.5.1.3 WiFi Bridging
Another option is to activate the feature in a smartphone that turns the smartphone into a Wire-
less LAN (WiFi) access point.
The computer then connects to the smartphone’s WiFi access point just as it would connect to
any other WiFi access point at Starbucks or at home, and the smartphone internally bridges
the WiFi connection to its cellular Internet data connection.
10.5.2 Packet Relay to the Internet
Essentially, the microphone, speaker, screen and keyboard in the phone are ignored; the com-
puter connects via USB cable, Bluetooth or WiFi to the cellular modem in the cellphone and
uses that along with the radio, antenna and battery to communicate data from the computer to
the cellular network base station.
The cellphone has to tell the base station that this is a “data” call, so that the received 1s and
0s are not fed to the cell phone company’s voice switch for onward forwarding to the PSTN or
to another cell phone, but rather to a router for onward forwarding to the Internet.
The router relays packets to local content servers, servers on the Internet, or perhaps servers
reached over a managed IP/MPLS network for video connections or secure corporate connec-
tions.
10.5.3 Dongles
A very similar story can be implemented with what the marketing department might call a
“stick” – a modem, radio and antenna built into a small dongle that plugs into a USB port on
the computer.
The dongle implements the same capability as a cellphone, but without the speaker, micro-
phone, codec, battery, keyboard and screen.
10.5.4 Cellphone as the Terminal
Of course, all cellphones are also computers.
The keypad on the cellphone can be used as an input device and the screen on the cellphone
used as the display.
The keypad could be the regular telephone keypad, requiring the user to press the 2 button
three times to select the character “c”, for example.
The keypad could also be a qwerty-type keypad included with the phone like on a classic
Blackberry, or of course, implemented on a touch screen with an underlying graphic image of
keys as on smartphones.
10.5.5 “Data” Billing Plans
One must be very careful with billing plans when using a cellphone for Internet access.
There may be one set of rates for using the cellphone as a tethered modem and a different set
of rates for using a browser integrated in the cellphone.
If a user does not add a data plan to their account before using the phone to access the Inter-
net, they will be charged the “default” or “casual use” rate, which can be astronomically high. In
Cozumel Mexico, in 2016, a carrier was offering “data roaming” at $5 per MB.
That’s $5,000 per Gigabyte. $6000 to download an HD movie via torrent.
$125,000 to download the data on a single-layer Blu-ray DVD.
It is not unusual to hear of people watching youTube and Netflix on a smartphone while roam-
ing without a “data” plan, then getting a bill of $20,000 from their home carrier, who is collecting
the funds and paying the Mexican roaming partner.

143
10.5.6 Fluid Layout, Responsive Design & “Mobile” Pages
When using a small cellphone screen as the display, special measures may needed due to the
small number of pixels on the phone’s screen.

Figure 97. Web pages for mobile devices

To address that issue, service providers set up navigation servers that provide the user with
links to web pages that have reduced-bandwidth content, for example, text-only pages without
fancy graphics that would be hard to display.
Use of these pages is optional. Should a user with a small screen navigate to a site like
gsa.gov, they may be delivered a web page very rich in graphics and a complicated layout.
In this case, the browser in the cellphone will have to reduce the richness of the content to be
able to display it on a small screen, or the user will have to do a lot of horizontal and vertical
scrolling to see all of each web page.
Alternatively, the web server could detect that the page request is coming from a device with a
small screen and serve up a reduced-richness page. This implemented by amazon.com and is
the goal behind fluid layout and responsive design.
Or… the user could be invited to surf the “mobile” site, i.e. the user is instructed to request
pages from an address specifically set up to serve to mobile terminals, with content already re-
duced in graphics and complexity. Examples include m.google.com and barnesandno-
ble.com/mobile/.
In addition to small screen size, slow connections would be another reason to reduce the rich-
ness of content delivered to a cellphone.
This was a significant issue with 2G technologies like GSM and CDMAOne that supported
maximum throughput of 9.6 kb/s or 14.4 kb/s, and often very much less.
For third-generation cellular, described in subsequent sections, the High-Speed Packet Access
(HSPA) variant of UMTS and the 1XEV-DO variant of CDMA2000 support data transfer rates
measured in Mb/s at present and measured in tens of Mb/s in proposed revisions.

144
4G LTE claims data transfer measured in the hundreds of Mb/s, eliminating this area of con-
cern.

10.6 FDMA, TDMA, CDMA and OFDM

Cellphones transmit and receive signals over shared radio bands.
To separate users so that they do not interfere with one another, nor hear each other’s conver-
sations, service providers use one of four radio band or spectrum sharing methods: Fre-
quency-Division Multiple Access (FDMA), Time-Division Multiple Access (TDMA), Code-Divi-
sion Multiple Access (CDMA) and Orthogonal Frequency-Division Multiplexing (OFDM) and its
multiple-access version.
In this section, we’ll explain how FDMA, TDMA, CDMA and OFDM work, and in this section
how they were deployed for first and second generation with names like AMPS, TDMA (IS-
136), GSM and 2G CDMA (IS-95).
In subsequent sections, we’ll take a closer look at CDMA for third generation (1X and UMTS),
then 4G LTE which uses OFDM.
10.6.1 FDMA
Frequency-Division Multiple Access (FDMA) is a spectrum-sharing method where a block of
spectrum is divided into small frequency bands called channels, which are organized into
groups.
Notionally, there was one group per cell.
Users are assigned a set of channels when their call begins. There are channels for voice
each way and for control signals each way.
To communicate voice, a carrier frequency is centered in a channel and its frequency is varied
or modulated continuously in proportion to the voltage coming out of the microphone – which in
turn is an analog or direct representation of the strength of the sound pressure waves coming
out of the talker’s mouth. This is referred to as analog radio.
The same idea is used for FM radio.

Figure 98. FDMA - AMPS

In the mobile system, if the user does not move, and there are not a lot of other users moving
around, the user will stay on those radio channels for the duration of the call.
This makes eavesdropping easy, as all that is required is an FM radio scanner that can be
tuned to the cellphone frequency – there is no encryption or coding of the voice.
If the user moves, or others do, the mobile switch may hand them off to another cell.
This means that both the base station and the radio frequency channels may change during
the call, since each cell uses different groups of channels.

145
During a handoff, the end-to-end communications path will be interrupted or muted for a short
period of time while the base station and frequency channel change is made.
This interruption is heard as a “click” during a voice call; it causes modems to disconnect.
FDMA was the method used in first-generation “analog” systems, including AMPS, NMT, and
TACS used in various countries.
10.6.1.1 AMPS
In North America, AMPS was deployed in radio bands 25 MHz wide at 800 MHz (824-849 MHz
uplink, 869-895 MHz downlink).
The radio band is divided into 30 kHz channels, with groups of channels allocated to base sta-
tion transceivers front-ended with antennas mounted on towers, providing radio coverage in an
area around the tower: the cell.
The groups of channels are allocated so that they can be re-used geographically far enough
away so that they do not interfere.
Organizing the channels into 7 groups is referred to as N=7, and allows coverage of arbitrarily
large areas using a honeycomb pattern for the cells as illustrated in Figure 93.
The capacity of the system is limited by the number of channels in the group of frequencies,
and is relatively low… 1.5 users per square mile in the previous calculation.
It is possible to sectorize a cell to achieve higher capacity, that is, more users per square mile.
Sectorization means using directional antennas with 120 degree or 60 degree beamwidths, in-
stead of an omnidirectional antenna with a 360 degree beam.
Sectorization implements a number of pie-wedge-shaped cells emanating from a single tower.
Typical plans for AMPS were to use 7 groups with 3 sectors for “rural” areas and 4 groups with
6 sectors for “urban” areas.
However, the resulting capacity is still too low, and data communications was very difficult us-
ing dial-up modems over the analog radio system.
10.6.2 TDMA
One strategy used for spectrum sharing for 2G was Time-Division Multiple Access (TDMA),
where a radio channel is shared in time between a number of users, hence the term Time Divi-
sion.

Figure 99. TDMA is time-sharing of radio channels via fixed time slots. The North American version IS-136 is illus-
trated. GSM has 16 time slots on 200 kHz channels.
Users transmit and receive modem signals one after another in a strict order in time slots on a
radio channel.

146
TDMA is called digital radio, since the modems integrated in the handset and base station
transceiver end up moving 1s and 0s, which could be coded digitized speech, text messages,
web pages, control and signaling or anything else.
In North America, the IS-136 standard for TDMA was deployed.
10.6.2.2 GSM
In most of the world, a form of TDMA called the Global System for Mobile Communications
(GSM) was deployed for 2G.
In GSM, the channels are 200 kHz wide with 16 time slots, meaning seven users per 200 kHz:
7 time slots each way for voice or data and 1 each way for control.
This was widely deployed, and became the most popular spectrum-sharing technology.
10.6.2.3 IDEN
The Integrated Digital Enhanced Network (IDEN) from Motorola is an overlay on TDMA that al-
lows group walkie-talkie functions similar to trunked radio systems used on construction sites
and at sporting events.
10.6.2.4 Inefficiency of TDMA
TDMA provided an improvement in capacity, by having users time-share radio channels; but it
is not an efficient way to share.
Users are assigned specific time slots and these time slots are reserved for them, whether
they have anything to transmit or not.
This is inefficient for telephone calls, where one speaks only half the time, and highly inefficient
for data communications like web browsing, where one has nothing to communicate most of
the time.
10.6.3 CDMA
Another spectrum-sharing strategy deployed for 2G was Code-Division Multiple Access
(CDMA). CDMA is completely different than FDMA and TDMA.
In a CDMA system, the available spectrum is divided into relatively wide frequency bands
called carriers in the business. For 2G, these carriers were 1.25 MHz wide.
The carrier is not divided into radio channels, and the users do not time-share radio channels.
Instead, all users transmit at the same time, spreading energy across the width of the same
carrier, in the same geographic area.
Each user is assigned a code, which is a binary number typically 64 bits long. If the user wants
to transmit a “0”, they send their code. If they want to transmit a “1”, they send the mathemati-
cal complement of their code.
The codes are arranged so that when codes from multiple cellphones are received at the base
station at the same time, the base station can determine which cellphones transmitted and
which didn’t.
This is analogous to being at a cocktail party where everyone is speaking at the same time in
the same space, but each pair of people are speaking unique languages - and you only under-
stand your partner’s language.
You can understand this by trying to match sounds you hear to your vocabulary. Only your
partner’s words make sense; everything else sounds like noise to you.

147
 Figure 100. CDMA implements “spread spectrum” by transmitting a 64-bit code to represent 0s and its comple-
ment to represent 1s, causing the energy to be spread over a wider band. At the receiver, transmissions from mul-
tiple handsets are added together. The receiver detects which codes have been transmitted by complex signal
processing.

10.6.3.5 Spread Spectrum

Since frequency bandwidth is proportional to the bit rate times the signal to noise ratio, sending
a 64-bit-long code instead of a single 1 or 0 has the effect of transmitting at a higher bit rate,
which spreads the energy of the transmission across a wider frequency band than non-coded
transmission does, and so CDMA is also referred to as a type of spread spectrum radio.
10.6.3.6 Spectral Efficiency
CDMA systems employ variable-bit-rate codecs and statistical multiplexing, which means that
if no noises are coming out of the user’s face, nothing is transmitted.
The capacity of the system is designed based on the statistics of how often noise does come
out of a person’s face during a phone call; speaking is treated as bursts of sound.
CDMA is the most spectrally-efficient method of those discussed here, meaning that it allows
the most number of phone calls per Hertz of radio band than any other.
Since the CDMA system was designed to treat speech like bursts of data, it is also inherently
efficient for data communications.
Qualcomm, a technology company out of San Diego, patented certain methods of power con-
trol and synchronization necessary to make mobile, multi-user, multi-base-station CDMA work,
and sells chips or licenses that implement these patented techniques.
CDMA for second generation cellular followed the IS-95 standard, deployed by Verizon and
Sprint in the USA and Bell and TELUS in Canada.
CDMA is the spectrum-sharing technique underlying most third- generation cellular, including
both the 1X and UMTS variations.
3G is covered in detail in the next section.
10.6.4 OFDM
The spectrum-sharing method for the fourth generation is Orthogonal Frequency-Division Mul-
tiplexing and its Multiple Access variation.
While it is a frequency-division technique like FDMA, there are significant changes compared
to the 1G analog FDMA.

148
In an OFDM system, hundreds or thousands of subcarriers are defined within the main carrier.
A subcarrier is a single frequency that will be modulated, spreading energy in a small band
around the subcarrier.
So far, this is the same as FDMA and its channels.
OFDM diverges from FDMA in two significant ways:
First, the subcarriers are not modulated continuously like “analog” FM radio; instead a modem
signal representing 1s and 0s is at each subcarrier frequency in the handset and at the base
station.
Second, users are assigned more than one subcarrier, allowing communication of high bit
rates by transmitting multiple streams in parallel.

Figure 101. OFDM subcarriers. In the most sophisticated implementations, a modem with 64 signals is run on
each subcarrier. In 4G LTE, multiple subcarriers are assigned dynamically to terminals for massive parallel down-
loads.
Essentially, this is taking the idea of a modem and its modulation and implementing it hundreds
or thousands of times in narrower frequency bands within the larger frequency band, then as-
signing multiple “modems” to each user to employ in parallel.
Other aspects of OFDM – which are of interest mainly to mathematicians and engineers with
graduate degrees in Digital Signal Processing – include the fact that the combined output
waveform from all of the subcarrier modems is calculated in a single mathematical step called
an Inverse Discrete Fourier Transform, and that the waveform is transmitted at the same rate
as the subcarrier spacing.
OFDM is also used for 802.11 wireless LANs and DSL.
We examine its use in 4G cellular, called LTE, in Section 10.8.

10.7 3G Cellular: CDMA

The third generation of cellular is usually referred to as 3G.
The main objectives of the third generation were to improve capacity, the number of simultane-
ous users, and to increase the number of bits per second that can be transmitted over the air-
link, for mobile wireless high-speed Internet access and video.
In this section, we’ll cover 3G mobile cellular radio technologies:
• How the quest for an international standard to resolve the I-95 CDMA vs. GSM TDMA in-
compatibility led to a Frankenstein standard called IMT-2000,
• How IMT-2000 included five different incompatible variations for implementing 3G,
• How two of them were of most interest: IMT-MC, also known as 1X, and IMT-DS, also
known as UMTS, both employing CDMA technology,

149
 • The tragic-comic attempts to deploy UMTS without reliance on Qualcomm or on the United
States government’s GPS,
• The data-optimized variations of the two, 1XEV-DO and HSPA respectively, and
• The capitulation of the 1X camp to the UMTS camp’s plan for 4G, and how that led to the
widespread deployment of HSPA.
10.7.1 IMT-2000
To try to avoid a repeat of the 2G CDMA vs. TDMA dichotomy, in 2000, a standards committee
attempted to define a world standard for 3G called IMT-2000.
They failed.
The result was a “standard” describing five incompatible implementation variations. Like many
other technologies, we ended up with one solution for “North America” and a different solution
for “Europe”.
To support higher bit rates over the airlink, more frequency bandwidth is required.
Out of the five variations in IMT-2000, the two serious ones both specified CDMA as the
method for spectrum-sharing – but disagreed on the width of the radio bands and how many
bands there should be.
10.7.2 1X or CDMA2000: IMT-MC
Service providers using CDMA for 2G, primarily North American and certain Asian countries,
favored a strategy that was basically a software upgrade from 2G, employing existing 1.25
MHz radio carriers and allowing multiple carriers.
This is called IMT-MC or multi-carrier CDMA.
Qualcomm’s brand name for this was CDMA2000.
The service provider could purchase licenses for as many bands as desired, and the bands
could be variable sizes to meet different countries’ radio licensing plans, providing a flexible
and scalable capacity.

Figure 102. Verizon deployed 1X 3G technology

A single 1.25 MHz carrier version of this referred to as 1X was widely deployed.
10.7.3 UMTS or W-CDMA: IMT-DS
Service providers using GSM TDMA for second generation, primarily cellular carriers outside
North America, favored the deployment of CDMA in a 5 MHz wide band.
This was called IMT-DS, Direct Spread, Wideband CDMA (W-CDMA) and Universal Mobile
Telephone Service (UMTS).
An incentive for GSM operators was they would be able to re-use some control infrastructure
from the second-generation TDMA GSM systems.
However, practical functioning of a multi-user, multi-base station, mobile CDMA network re-
quires among other things constant control of the power on the cellphones, so that the re-
ceived power at the base station is the same from all the phones; and compensation for time
delay differences from signals from the same phone received at different base stations.
In the 1X systems, this was accomplished using techniques patented by Qualcomm (and pay-
ing Qualcomm a royalty for every cell phone and every base station transceiver), and the

150
United States government’s Global Positioning System respectively.
European operators, with their UMTS, did not favor the notion of paying an American company
royalties, and did not favor building a network dependent on the American government’s GPS.
Since UMTS required mathematical calculations across a 5 MHz band, compared to 1X’s 1.25
MHz band, at the time, the processor in the phone required to perform such calculations drew
so much current from the battery that the battery heated up to the point that people burned
their hands on the phones during trials.
The GSM/UMTS Europeans embarked on a seven-year-long odyssey attempting to circum-
vent Qualcomm patents, and avoid using GPS.
After a number of strategies failed, a Euro-GPS called Galileo was created for UMTS; the first
satellite was launched December 28 2005.
This delayed deployment of UMTS until 2007.
1X was deployed and working years earlier.
The tipping point between 2G and 3G in the GSM/UMTS camp was reached in the summer of
2007, when more new activations on these carriers’ networks were 3G CDMA (UMTS) instead
of 2G TDMA (GSM).
The 2G TDMA technology GSM still had far more users, but like 1G analog, GSM will eventu-
ally disappear.
10.7.4 Data-Optimized Carriers: HSPA and EV-DO
For Internet access and watching video on cellphones, variations of the coding schemes opti-
mizing for the statistical characteristics of “data” were developed and deployed by both camps.
In both cases, these were deployed on carriers (the 1.25 or 5 MHz bands) apart from those
used for telephone calls.
Accessing these data carriers required either a “stick”, the USB dongle described in an earlier
section, or dual radios in a phone, one tuned to the traditional carrier for telephone calls and a
second tuned to the data-optimized carrier for watching video.
The 1X camp developed a variation called 1X Evolution Data-Optimized (1XEV-DO), allocating
a carrier for data communications and promising 2.4 Mb/s over the airlink in the first incarna-
tion.
Proposals for future revisions of EV-DO promised to support more than 70 Mb/s over the air-
link.
In the UMTS camp, the variation was called High Speed Packet Access (HSPA), referring to
improvements in the UMTS downlink, often called High Speed Downlink Packet Access (HS-
DPA) and in the uplink, High Speed Uplink Packet Access (HSUPA) and also Enhanced Dedi-
cated Channel (E-DCH).
Revisions of HSDPA promised download rates of 14.4 Mb/s then 42 Mb/s.
10.7.5 The End of the Standards War
Market forces finally pushed the two camps together.
The fact that there were far more 2G GSM users on the planet meant that for one thing, hand-
set manufacturers produced 2G GSM phones before 2G CDMA phones.
GSM phones were less expensive and had better features.
This trend was continuing into 3G, where UMTS phones would have the same advantage over
1X phones.
Another fact was that Steve Jobs of Apple only permitted carriers operating TDMA systems to
have the iPhone, then only permitted carriers with HSPA systems to have the iPhone 3G.
Finally, the 1X camp threw in the towel and decided to go with the UMTS camp’s proposal for
the fourth generation to level the playing field.
As soon as that decision was made, then the deployment of 1XEV-DO was more or less
capped, and 1X carriers began deploying HSPA instead.
And the fact is, as soon as carriers that were in the 1X camp, like Verizon in the US and Bell
and TELUS in Canada deployed HSPA, Steve Jobs allowed the iPhone on their networks.

151
As the iPhone was at the time the most popular phone, this was a major incentive for the 1X
camp.
It appears that one of the legacies of Steve Jobs will not just be the iPhone, but a key part in
ending the spectrum-sharing standards war.

10.8 4G Mobile Cellular: LTE

10.8.1 Universal Terrestrial Radio Access Network Long-Term Evolution
After more than 20 years, it appears that an almost universally-accepted standard for mobile
radio is being implemented, bringing the GSM/UMTS vs. CDMA/1X standards war to an end.
Carriers from both of the factions supported the GSM/UMTS Third Generation Partnership
Project (3GPP) release 8, known as Universal Terrestrial Radio Access Network (UTRAN)
Long Term Evolution (LTE), winning out over alternatives including 802.16 WiMax.
LTE promises co-existence with other standards, allowing in theory handoffs between cells
supporting LTE and cells supporting UMTS, GSM/GPRS, 2G CDMA, 1X or 1XEV-DO.
LTE’s spectrum-sharing method, called Orthogonal Frequency Division Multiplexing (OFDM), is
different than that of previous generations, providing flexible and efficient use of different car-
rier bandwidths along with tolerance to noise and multipath interference.
Bit rates are on the order of 100 Mb/s when a 20 MHz carrier is employed. As with all claims
for wireless bit rates, these are peak burst rates under ideal conditions with one user per cell.
Multiple-Input, Multiple-Output (MIMO) antenna designs can increase the bitrate using spatial
multiplexing, which is basically gluing several transceivers and antennas together.
10.8.2 OFDM
The modulation and spectrum-sharing scheme for LTE is OFDM, which is different than FDMA,
TDMA and CDMA.
It is interesting to note that OFDM is also used for 802.11 wireless LANs (WiFi), 802.16 Wire-
less Broadband MANs (WiMax) and for DSL modems on copper wires.
It’s not really necessary to understand the details of how the modulation actually works… but
since OFDM is used for 4G, WiFi, DSL and cable modems, it’s worth knowing the basic idea:
hundreds of modems working in “parallel” within the available frequency band, and the radio
waveform resulting from all of them is calculated in a single step.
For those interested in a longer explanation: the basic idea with OFDM is the definition of hun-
dreds or thousands of subcarriers within the main carrier.
A subcarrier is a single frequency which will be modulated like any other modem carrier signal,
spreading energy in a small band around the center.
Essentially, this is taking the idea of a modem and its modulation and implementing it hundreds
or thousands of times in narrower frequency bands within the larger frequency band.
The incoming bit stream is then divided up in some way to use as the input bits to modulate
each of the subcarriers.

152
 Figure 103. OFDM
In the simplest implementation, illustrated in Figure 103, the incoming bits are used to turn
subcarriers on or off, splitting the incoming bit stream at a rate of 1 bit per subcarrier and im-
plementing binary Amplitude Shift Keying, with one of the amplitudes zero.
In the most complex implementation, the incoming bit stream would be allocated at a rate of 6
bits per subcarrier, used to implement QAM-64 on each of the subcarriers.
The outputs of the subcarrier modulations are all added together to produce a transmittable
waveform.
To allow multiple users at the same time, Orthogonal Frequency-Division Multiple Access
(OFDMA) is implemented… essentially adaptively assigning specific subcarriers to particular
users.
The beautiful part of OFDM (at least to Engineers) is that the modulation of each subcarrier
and adding them all together is calculated in a single step with a digital signal processing oper-
ation called an Inverse Discrete Fourier Transform. At the receiver, a Discrete Fourier Trans-
form performs the reverse process to yield the original bit stream.
This calculation is performed, and a waveform transmitted at the same frequency as the spac-
ing of the subcarriers, which has the result of making the harmonics of all of the subcarriers
cancel each other out at the receiver, hence the term orthogonal.
In the LTE standard, the subcarriers are spaced 15 kHz apart, and the output is calculated
15,000 times per second.
Prior to modulation, Forward Error Correction is implemented, adding redundancy to the bit
stream so correct decisions can be made based on maximum likelihood in the presence of im-
pairments like noise and fading.
The bit stream is also shuffled or interleaved, re-arranging the order of the bits in time so that
burst errors are no longer sequential errors.
In the uplink, LTE uses a pre-coded version of OFDM called Single Carrier Frequency Division
Multiple Access (SC-FDMA) to avoid needing power amplifiers, which would increase handset
cost and shorten battery life.
10.8.3 3GPP Standards Committees

153
The 3GPP Technical Report 25.913 contains the detailed requirements specification for LTE.
The system architecture, in Technical Specifications 36.300 and 36.401, is simplified to two
principal network elements: evolved Network Base stations (eNBs) and Evolved Packet Cores
(EPCs). eNBs communicate with EPCs, with each other and with user equipment.
The ITU defined as 4G supporting at least 1 Gb/s downloading. LTE does not meet that crite-
ria, and so in a strict standards committee environment, LTE would be called a 3G technology.
An updated version, 3GPP release 10, called LTE-Advanced, is expected to be submitted to
the IMT-Advanced standards committee, which would cause those standards committee mem-
bers to declare LTE-Advanced to be a 4G standard.
Everyone else will refer to LTE as 4G from the start.
10.8.4 Qualcomm Patents
One of the reasons for the 3G standards war was the requirement to pay American company
Qualcomm royalties on patents for several techniques necessary for a mobile CDMA system.
LTE is not CDMA, so those royalties are avoided… but it turns out that Qualcomm filed or has
purchased many patents that underpin LTE.
Additionally, since LTE phones will have to be backwards-compatible with 3G CDMA networks,
Qualcomm sees “no impact” on patent royalty revenue for the first ten years of LTE develop-
ment according to COO Sanjay Jha.

10.9 Wireless LANs

In this section, we provide an overview of the 802.11 wireless LAN standards, often called
WiFi.
We will concentrate on understanding the variations of 802.11, the frequency bands they oper-
ate in, bit rates to be expected and practical issues.
Since 802.11 is wireless LANs, there are a number of associated topics: LAN frames, also
called MAC frames, MAC addresses, LAN switches, IP addresses, routers and network ad-
dress translation.
Those topics are covered in other Chapters, particularly Chapter 12, Chapter 14 and Chapter
15. Here, we concentrate on the radio aspects.
10.9.1 System Components
Perhaps the most widespread broadband wireless data communication technology today is
WiFi: 802.11 standard wireless LAN technology.
This is essentially Ethernet LANs over the air, at data rates measured in the tens to hundreds
of Mb/s and ranges measured in hundreds of feet.
A typical set-up employs a radio base station, called an Access Point (AP) and wireless
adapter cards plugged in to or built into computers, printers, cameras and other devices.
The Access Point connects to the wired LAN in-building via an integrated Ethernet switch,
which may provide wired connections to other computers and network-enabled printers as il-
lustrated at the top of Figure 104. Ethernet LANs and Ethernet switches are covered in Chap-
ter 14 “Ethernet, LANs and VLANs”.
Most Access Points also include an edge router, sometimes called the Customer Edge (CE)
function, which allows it to connect to an ISP and authenticate, and includes features such as
Network Address Translation, firewall functions, port forwarding, DHCP server and DHCP
client.
These topics are covered in Chapter 15 “IP Networks, Routers and Addresses”.

154
 Figure 104. WiFi

10.9.2 Unlicensed Radio Bands

All WiFi systems operate in unlicensed radio bands, also called Industrial, Scientific and Medi-
cal (ISM) bands in North America.
“Unlicensed” means that it is not necessary to obtain a license from the national government to
emit electromagnetic energy at these frequencies.
At other frequencies, it would be necessary to obtain a license, which usually involves proving
that you will not interfere with anyone else at the requested frequencies in a specific geo-
graphic area.
10.9.3 802.11 Standards
There are a number of standards for WiFi, published by the 802.11 working group of the Insti-
tute of Electrical and Electronic Engineers (IEEE): 802.11a, 802.11b, 802.11g, 802.11n,
802.11ac and surely more to come.
802.11b and 802.11g operate in the 2.4 GHz unlicensed radio band, offering a maximum of 11
and 54 Mb/s respectively.
However, there is a lot of interference in the 2.4 GHz band, since many other devices take ad-
vantage of the fact that it is unlicensed.
Cordless phones, baby monitors and Bluetooth all operate in this band.
It is possible to listen to 802.11 data transmissions on an analog 2.4 GHz cordless phone.
Microwave ovens operate at 2.4 GHz… and if one were to place an 802.11 access point on top
of a microwave oven, one would have no wireless LAN while the oven was on because of the
interference.
Another issue is that the communications are half-duplex, or more accurately, alternating: only
one device can transmit at a time, so they must alternate.
This dramatically reduces the effective throughput compared to a wired system where there is
two-way simultaneous or “full-duplex” communications.
802.11a operates in a 5 GHz ISM band, supporting a maximum of 54 Mb/s.
This band is relatively free of interference, but the higher frequency also means shorter range
and poorer penetration through walls. In practice, line-of-sight between the access point and
the terminal are necessary to achieve 54 Mb/s.
802.11n is a newer standard.
It uses the 2.4 and/or 5 GHz ISM bands, matching the frequency plan of existing 802.11 de-
vices as well as optimizing for power to noise ratio between the bands.
802.11n also supports 20 MHz and/or 40 MHz channel support – using more of the wireless
spectrum when available to enhance performance, and allows spatial multiplexing modes for

155
simultaneous transmission using 1 to 4 antennas, allowing very high data rates.
In theory, 802.11n will implement 300 Mb/s with a single antenna… but that would be on the
moon, where there are no atoms between the transmitter and receiver and no interference.
802.11ac purports to achieve 500 Mb/s.
As soon as there is anything between the transmitter and receiver – like water molecules, plas-
ter, concrete and so forth, and/or interference, the power-to-noise ratio and thus bit rate drops.
10.9.4 VoIP over Wireless LANs
Most of the time, 802.11 wireless LANs are used to access the Internet for email, web,
youTube and the like, and cellular is used for phone calls.
Phone calls over a device using a “free” wireless LAN connection instead of a paid cellular
connection will be a growth area.
As soon as a smartphone supports WiFi, and uses an app like Google Hangouts, which pro-
vides free phone calls to the PSTN, there is no need to pay for cellular airtime on the smart-
phone to make a phone call whenever the phone is connected to WiFi.
10.9.5 Wireless Security
A major concern with wireless LANs is security.
Network security that can be enacted is Media Access Control (MAC) address filtering: setting
the base station to only accept connections from specific, predefined wireless LAN cards.
This protects the network connection from access by unauthorized users – but does not pro-
tect legitimate users’ transmissions from eavesdropping.
If someone can get physically close enough to receive signals, there is no way to prevent them
from eavesdropping on communications, which can include intercepting and re-using user-
names and passwords and intercepting and “wikileaking” sensitive information.
This is particularly troublesome in coffee shops, airports and anywhere else the communica-
tions are not encrypted, an “open” hotspot.
In 2010, a plugin for the Firefox browser was made available that allowed someone sitting in
such a coffee shop to eavesdrop on everyone else’s communications – and with one click, to
re-use other people’s credentials to log in to servers.
This means that secure encryption of communications over the airlink is now mandatory, not
optional.
If it can be ensured that the users always, without fail, implement client-server encryption
(sometimes called Transport Layer Security… though anyone who has taken the OSI Layers
course will know it is Presentation Layer security), by using VPN software for connecting to
work, typing https:// for all web surfing, and using encrypted email communications, then there
is no need for encryption of the airlink.
However, users can not be relied upon, so encryption of the communications on the airlink be-
tween the access point and terminal must be implemented whenever possible.
Wired Equivalent Privacy (WEP) was the first encryption algorithm for wireless LANs; but its
use is not recommended as there are software tools available that can crack it in a matter of
minutes.
WiFi Protected Access (WPA) or the 802.11i WPA2 should be implemented on the airlink
whenever possible.

10.10 Communication Satellites

In this last section of this chapter, we will take a quick overview of communication satellites,
understanding the basic principles and the advantages and disadvantages of the two main
strategies: Geosynchronous Earth Orbit and Low Earth Orbit.
10.10.1 Transponders
Communication satellites are orbital platforms that carry multiple base station transceivers with
antennas pointed towards the surface of the earth.

156
Instead of base station transceiver, the term transmitter/responder or transponder is used in
the satellite business.
In two-way systems, radio signals are transmitted from the earth to the transponder, which re-
sponds with radio signals directed back down to the surface at a different frequency to avoid
interfering with the surface transmitter.
There are two basic choices for the orbits of communication satellites: geosynchronous orbit or
low earth orbit.
10.10.2 Geosynchronous Orbit
Geosynchronous satellites are parked 22,300 miles (35,680 km) above the surface of the earth
above the equator.
At that radius, the orbital speed is the same as the rotational speed of the earth, and hence the
satellite appears to stay in the same spot in the sky.
This is geosynchronous or geostationary... depending on your point of view.
Geosynchronous communication satellites are operated by the International Telecommunica-
tions Satellite Organization (Intelsat), the International Marine Satellite Organization (In-
marsat), numerous private companies, government and military.

Figure 105. Geosynchronous Satellites

Each country has the right to spots in geosynchronous earth orbit above their country based
on international agreements. The countries can use these spots, or lease them. Occasionally,
there are disagreements as to who has the right to a spot and two countries will launch a satel-
lite and park it in close proximity to the other, causing interference.
The three main advantages of geosynchronous satellites are broadcast, broadcast and broad-
cast: a transponder on a geosynchronous platform is 22,300 miles up in the sky – and if de-
sired, can provide radio coverage to a third of the earth’s surface.
The main disadvantage is path delay: the radio waves have to travel 22,300 miles up and
22,300 miles down.
At the speed of light, that takes about 1/4 second each way.
For interactive communications between two points on the earth via a geosynchronous satel-
lite, that means up and down for the inquiry, and up and down for the response, a total of just

157
under one second delay.
If the two locations can not see the same satellite, then an intermediate ground station must be
used, meaning a path delay of about two seconds.
This wreaks havoc with the protocol people use to decide who gets to talk next during a phone
call, and the extended delay causes users to hear echoes that are normally suppressed.
No one likes using use geosynchronous satellites for phone calls; trans-oceanic fiber optic ca-
bles are preferred because they are much shorter, meaning that the path delay is negligible;
about 25 milliseconds from New York to Paris on fiber, for example.
One-way communications is the natural application for geosynchronous satellites.
Television is the biggest market.
Radio-frequency modems communicate video that has been digitized, coded using MPEG-2 or
H.264 (MPEG-4 Part 10) and encrypted, from a Digital Broadcast Center of a satellite TV com-
pany up to the transponder, which repeats the modem signal back to the earth.
Access to the Internet is also implemented on geosynchronous satellites for those who do not
have DSL, Cable or WiMax terrestrial links available.
The “upload” path from the customer to the Internet is often over modems over a regular
phone line, and the download path is via satellite.
This makes the customer premise electronics cheaper and cuts the delay in half.
Two-way satellite communications is also available.
Another service based on geosynchronous satellites is Very Small Aperture Terminal (VSAT),
which means “small dishes” in plain English.
Applications for this two-way wide-area data communication service include emergency
backup communications capabilities and nationwide VPN services.
10.10.3 Low Earth Orbit
The path delay problem of geosynchronous platforms can be fixed by bringing the satellite
closer in.
These types of communication satellites are called Low Earth Orbit (LEO).
These can be used for voice communications because the path delay is reduced to an accept-
able level.

Figure 106. Low Earth Orbit Satellites

158
This introduces two different problems: the satellites do not stay in the same position in the sky
to an earthbound observer, and the coverage or footprint of the satellite is reduced.
Multiple satellites to ensure coverage and a switching system for doing handoffs from one
satellite to another as they move are required.
The handoff problem is similar to that of cellular, where the base stations are stationary and
the users move around, except that with LEO satellites, the users are mostly stationary and the
base stations move around.
Motorola’s Iridium project was one example of this. Iridium planned to launch 77 satellites (Irid-
ium is element number 77 in the periodic table of the elements), but went live after deploying
only 66 satellites.
Gaps in coverage, poor in-building penetration and difficult data communication over the ana-
log radio system led to poor user response and Iridium was a financial failure.
Motorola announced that they would “de-orbit” the satellites at a loss of five billion dollars. At
the last minute, an “entrepreneur” purchased Iridium for 25 million dollars. It cost five billion
1998 dollars to build.
Other LEO companies include Orbcomm and Globalstar. Orbcomm was a joint venture be-
tween Teleglobe and Orbital Sciences Corporation, intended to provide two-way data commu-
nications and the capability to track trucks on highways and tanks on battlefields.
Globalstar is a consortium of telecommunications companies operating a constellation of 48
low-earth-orbit (LEO) satellites.
Globalstar phones incorporated both the satellite radio and a cellphone, and could operate on
local cellular and/or satellite.
Globalstar sold high-capacity inter-city links wholesale to regional and local telecom service
providers around the world.
All three LEO systems have experienced severe financial difficulties.

159
11
“Data” Communications
Since the Holy Grail of convergence has been achieved by treating voice and video like data,
we begin understanding the new-generation telecom network with a chapter on what used to
be called data communications, historically a different topic than voice or video communica-
tions.

11.1 Convergence: Treat Everything Like Data

Telephone calls were historically carried on channelized systems as a natural evolution from
physical trunk circuits, business data was carried on packetized systems for efficiency, and
television programs were delivered in frequency channels on cable TV.
Convergence means carrying all traffic: telephone calls, business data, Internet traffic, music,
television, video on demand and everything else on the same network.
It means moving away from separate networks for telephone, television and Internet, with sep-
arate entry cables, separate equipment, separate services and separate bills for each.
Convergence, also called service integration, means many services delivered over one net-
work, one access circuit, with one bill.
On a network, there are two main methods of dividing sharing the available capacity amongst
users. One method is dividing the capacity into fixed fractions called channels and assigning
channels to users for the duration of their communication session – whether they actually have
anything to transmit or not. This is called channelization and channelized multiplexing.
Another method of is allowing users to employ transmission capacity on demand, that is, only
when they actually have something to transmit. This is called packetization, statistical multi-
plexing and bandwidth on demand.
To implement convergence, integrating everything on a single network, the two choices would
then be to either:
• treat everything like voice and carry it in channels, or
• treat everything like data and carry it in packets.
11.1.1 Convergence via ISDN
Beginning in the 1960s, Integrated Services Digital Network (ISDN) was viewed as a solution
for convergence using channelized multiplexing for all applications: carry everything like voice.
In theory, users would communicate data by establishing circuit-switched connections end-to-
end for the duration of a data communication session, then disconnect.
For many reasons, ISDN did not gain momentum and the technologies developed for access
to the ISDN, such as Basic Rate Interface, are now obsolete. This was a first attempt at “con-
vergence”.
11.1.2 Convergence via ATM
In the 1980s, significant effort was expended in product and service development of a technol-
ogy called Asynchronous Transfer Mode (ATM), which would carry everything: voice, video
and data over a single statistically-multiplexed network in packets called cells... treating every-
thing like data.
ATM was deployed to manage data traffic on carriers’ backbones, and some very large organi-
zations deployed ATM, but it was so complicated and expensive that it remained a technology
mostly limited to the core of the network. ATM was not deployed on the telephone network to
carry POTS trunking. ATM is now an obsolete legacy technology.

160
11.1.3 Convergence via IP
The third attempt, carrying everything in IP packets plus a traffic management technology
called MPLS in the network core, has proved to be the charm, and the Holy Grail of integration
of all services on a single network appears to have achieved.
With IP and MPLS, we are treating voice and video like data, carrying everything in data pack-
ets and providing bandwidth to users on demand.

11.2 Information Theory

Data communications as we know it today started with the publication of a set of papers by
Claude Shannon in the Bell System Technical Journal in July and October 1948, entitled “A
Mathematical Theory of Communication”.
This was the beginning of what is known today as information theory.
The core premise of information theory is the desire to communicate information across a
physical medium.
To do this, the information is coded in binary digits or binits: 1s and 0s.
The code rate, i.e. the number of binary digits required to represent the information can never
be less than the information content of the signal measured in bits.1
The physical medium, being real, has a maximum capacity, a maximum rate at which informa-
tion can be transmitted across it. The information transfer rate in bits/second can never be
faster than the channel capacity.
The capacity is proportional to the frequency bandwidth times the signal-to-noise ratio.
In this course, we explore the practical aspects of data communications, and leave the study of
information theory as a homework assignment for the interested reader.

11.3 Data Circuit Model

To start our journey through the world of what used to be called data communications, we’ll es-
tablish a model for discussing data circuits and communications, then look at examples to put
the model in context and preview what we will cover in the remainder of the book.

Figure 107. ITU model for a circuit at the lowest level.

A model or paradigm widely used to discuss some of the more fundamental ideas of data com-
munications comes from an organization that used to be known as the CCITT (le Comité Con-
sultatif International de Téléphone et de Télégraphe). The name was changed to the Telecom-
munications Standards Sector of the International Telecommunications Union, and is abbrevi-
ated ITU-T, or simply ITU.
This is an international treaty organization, which has general meetings, committees, study
groups and working groups and that historically operated on a four-year study cycle. Each
member country sends delegates, who study and sometimes decide on international standards
for telecommunications and data communications.
The ITU model has three components: Data Terminal Equipment (DTE), the devices between
which one wishes to communicate information; the physical medium over which the informa-

161
tion will be communicated; and the Data Circuit-terminating Equipment (DCE), the devices
placed at each end of the physical medium.
To communicate information, it must be coded into binary digits (1s and 0s), which are repre-
sented on the physical medium. The data circuit could be a guided system such as wires or
optical fiber, or non-guided like radio.
The DTE is the source of the 1s and 0s. The DCE represents them on the physical medium.
There are specific types of DCE for each physical medium.

11.4 DTE: Data Terminal Equipment

The word terminal comes from the Latin word terminus, and is used to indicate that these are
the devices at the ends of the circuit. There are many kinds of devices that can be Data Termi-
nal Equipment.

Figure 108. Data Terminal Equipment (DTE)

In the beginning, there were two categories of terminal equipment: dumb terminals and hosts.
Today, computer-to-computer client-server communications is most common today.
Dumb terminals have been called display terminals, ASCII terminals, Video Display Terminals
(VDTs) and even Cathode Ray Tube (CRT) displays.
These are all names for the same thing: a device that has a keyboard and a screen, and does
nothing but transmit binary codes corresponding to keystrokes and display characters corre-
sponding to received keystroke codes on the screen. ASCII, the American Standard Code for
Information Interchange, is a standard method of coding keystrokes into binary digits.
Dumb terminals as such are now a rare breed. Today, a Personal Computer (PC) is often used
as a dumb terminal by running terminal emulation software on the PC. Emulation means taking
on the attributes of something else. Popular software for dumb terminal emulation includes tel-
net, allowing remote login to a computer running UNIX such as a web server; SSH, a secure
(encrypted) version of telnet, and Remote Desktop Connection from Microsoft for remote login
to a desktop computer running Windows.
An intelligent terminal, also called a host, server, client or simply computer is one that has a
keyboard and a screen, plus a processor and memory. In fact, that is the definition of a com-
puter: input, output, processing and memory. An intelligent terminal can do all of the functions
of a dumb terminal, plus local data processing and local data storage. This functionality lends
itself to client-server computing: where intelligent terminals (clients) access centralized re-
sources (servers) and much of the processing - like creation of a graphics screen - is per-
formed in the terminal, not on the server.
The most popular software for turning a PC into a client for client-server computing over a LAN
is Microsoft Windows. The most popular software for turning a PC into a client for client-server
communication over the Internet is a web browser.
Point of sale terminals usually have other functions incorporated, such as magnetic stripe
readers and chip card readers. Examples of point of sale terminals are the “cash registers” in
department stores, and the credit authorization terminals at gasoline stations.
Many other devices use data circuits, and can be considered as DTE. Examples include print-
ers, bar-code readers, and FedEx and AVIS Rent-A-Car portable data terminals.

162
11.5 Data Circuits
A circuit is made of a physical medium on which the information coded into binary digits is to
be represented. There are many different types of circuits, and many different media available.
The binary digits are usually communicated over the media using bursts of energy: electrons,
photons, electro-magnetic radiation. Guided transmission systems are those that use a guide
for the energy, like a wire or a glass fiber. Non-guided systems are those that transmit through
the air, such as wireless LANs.
11.5.1 Analog Data Circuits
Data circuits are often categorized as analog or digital. The word analog comes from the
method by which voice is represented on an ordinary telephone line: a voltage is used to rep-
resent, or be an analog of, the voice. The voltage is an analog signal, and the circuit is popu-
larly (and inaccurately) called an analog circuit.
Data can be communicated on these “analog” circuits using modems as the circuit-terminating
equipment. The modems essentially produce a waveform to represent 1s and 0s that is like an
analog signal, which can be carried over a circuit designed to move analog signals.
11.5.2 2-Wire and 4-Wire Circuits
In days past, dial-up circuits – regular two-wire phone lines – were used for data communica-
tion. Expensive dedicated point-to-point two-wire circuits could also be employed. Conditioning
could be applied to these dedicated lines, improving their transmission characteristics and thus
the number of bits per second communicated.
It is also possible to order dedicated point-to-point two-wire circuits from the phone company
that are not telephone lines, they are bare copper wires, not restricted to the voiceband as cov-
ered in Section 2.4.3. These are called unloaded or dry circuits.
Two of those circuits, one for each direction, were used for expensive point-to-point 4-wire data
circuits in the previous millenium.
11.5.3 Broadband Analog
Cable TV systems and radio employ wider or broader frequency bands than pairs of copper
wires, and so are sometimes referred to a broadband circuits. Just as with telephone lines,
modems are used to represent 1s and 0s on cable TV and radio channels.
11.5.4 Analog on Fiber
Hybrid Fiber-Coax cable TV systems, covered in Section 9.7, employ fiber for distributing sig-
nals to neighborhoods. On these fibers, techniques involving varying the intensity of the light
on the fiber continuously as an analog of modem signals or analog television pictures are often
used. This is “analog” on fiber.
11.5.5 Digital
“Digital” communications involves the use of circuits that were not designed to carry analogs of
signals, but instead were designed from the beginning to communicate 1s and 0s using pulses
of energy.
On twisted-pair copper wire digital circuits, like LAN cables, a “pulse” is raising the line voltage
to some fixed value like 3 volts for a fixed period of time, then returning it to zero. A graph of
voltage vs. time would look like a square wave.
On fiber systems like Optical Ethernet, a “pulse” is turning a laser on for a fixed period of time,
then turning it off. The laser produces a pulse of light. A graph of light intensity vs. time would
look like a square wave.
In both cases, on a “digital” circuit, the signaling is binary: either a pulse is happening, or not.

11.6 DCE: Data Circuit-Terminating Equipment

Data Circuit-terminating Equipment (DCE) is the equipment that is placed at each end of the
data circuit. It is this equipment that represents binary digits on the physical medium.

163
Just as we categorize the way we represent information on circuits into “analog” and “digital”,
so we classify circuit-terminating equipment as analog or digital.

Figure 109. Data Circuit-terminating Equipment

Modulation is the name given to the technique of representing binary digits using a signal that
resembles a voice signal, used on circuits designed for voice. This is covered in detail in Sec-
tion 9.1.
There are many different modulation techniques and standard types of modems. The device
that performs the modulation at the transmitter is a modulator. The device that interprets this
modulation at the receiver is a demodulator. Modem is a contraction of the words modulator
and demodulator, a device that can perform both functions so it can communicate in both di-
rections.
When people say “digital transmission”, what they are supposed to be meaning is that the in-
formation is represented using a sequence of pulses. This is not called modulation, this is
called a digital line code, so the equipment that does these pulses is not called a modem.
There’s no such thing as a “digital modem”. That’s an oxymoron. Digital is digital; modulation is
the analog technique.
Unfortunately, there is not one single generic name for digital data-circuit terminating equip-
ment. A popular type is a LAN Network Interface. Antother is a device containing an optical
transceiver, including Optical Network Units (ONUs), Optical Line Terminals (OLTs), Optical
Network Terminals (ONTs), and Small Form-factor Pluggable (SFP) terminations.
Older copper-wire devices included Channel Service Unit (CSU), Data Service Unit or Digital
Service Unit (DSU), CSU/DSUs and ISDN Basic Rate Network Termination Type 1s (NT1s).
Regardless of what they are called, all of these pieces of hardware perform the same function:
representing 1s and 0s on the physical medium that makes up the circuit.

11.7 Point-to-Point Circuits

Now, we begin to look at some circuit configurations. The simplest configuration is when there
are only two devices to be connected. The term point-to-point is reasonably self-explanatory:
the circuit goes from “A” to “B” and nowhere else. There are two different strategies for com-
municating bits on point-to-point circuits: serial and parallel.

Figure 110. Parallel

The requirement is usually to communicate groups of bits, for example, groups of eight bits.
One choice would be to connect eight circuits, and represent one of the bits on each circuit,

164
then tell the other box to look at the circuits because the data is valid. This is called parallel be-
cause the wires are literally in parallel in a cable.
The other choice would be to connect up one single circuit, and represent the bits one after an-
other in a sequence in time on the single circuit. This is called serial, though a mathematician
would prefer to call it sequential.
Inside a computer, data is grouped into bytes, which are grouped into files. To communicate
these groups of bits over a serial line, a function traditionally called a serial port is required to
represent the bits one after another in a sequence in time on the single circuit. Another serial
port function is required at the far end to look at the line at the appropriate times to receive the
bits.

Figure 111. Serial

Today, the Universal Serial Bus (USB) or the LAN interface on a terminal is used to perform
this function for data circuits. The Serial Advanced Technology Attachment (SATA) standard is
used to connect drives to motherboards inside terminals.
In days past, this was the serial port on a PC, driven by an inexpensive chip called a Universal
Asynchronous Receiver / Transmitter (UART), which supported bit rates up to about 100 kb/s.
Virtually all communication circuits are serial as manufacturing costs are lower with only one
line driver and one line receiver required.
But... to achieve very high bit rates, parallel communications reappears with the splitting of the
bit stream into a number of lower-speed streams communicated in parallel.
For example, Gigabit Ethernet on copper LAN cables is four 250 Mb/s parallel streams on four
pairs of wires. 40 and 100 Gb/s Optical Ethernet on fiber splits the bit stream into paths carried
on multiple wavelengths in parallel. 4G LTE cellular can split the bit stream into 100 or more
lower-rate streams communicated on individual frequency channels called subcarriers.

11.8 Multidrop Circuits

A multidrop circuit is a single circuit with multiple derivations, or viewed the other way around,
multiple stations connected on a single circuit.

165
 Figure 112. Multidrop Circuit
Examples of multidrop circuits include Wireless LANs (WiFi), Cable TV distribution and IBM
mainframes.
One implementation is an unbalanced mode, where there is a primary station, or controller,
which controls the link, communicating to secondary stations or controlees. The secondary sta-
tions are computers, and in turn have Human-Machine Interface (HMI) devices attached. Ex-
amples of HMI devices are television displays, telephones, PCs and dumb terminals.
Other implementations are balanced configurations, where there are no controllers and con-
trolees: all stations are equal.
WiFi is a balanced configuration, where the stations alternate: first the primary station (the Ac-
cess Point) transmits and all of the stations hear its transmission, then there is a quiet period,
then a secondary station can transmit and all of the stations hear its transmission.
On Cable TV systems, for downstream communications (to the residence), your information is
broadcast to everyone in your neighborhood, with an address indicating for whom it is in-
tended. For upstream, the secondary stations (residences) can either contend for the right to
transmit, or transmission time slots or subcarriers can be reserved for specific users. This is
covered in more detail in Section 9.7.
The unbalanced architecture was used in large IBM mainframe computer installations. The
Front End Processor or Communications Controller is the primary station, and Remote Termi-
nal Controllers are the secondary stations. These type of circuits and the data that they carry
are called legacy systems and legacy traffic... meaning that they are leftovers from previous
eras.
A strategy for controlling which station can communicate on the shared link is required. Polling
techniques were implemented for mainframes. The primary polls the secondaries and gives
them permission to transmit by selecting them.
These systems have a problem with scalability… the number of secondary stations cannot be
arbitrarily extended due to the control overhead. For in-building wiring, we tend not to use this
architecture much anymore.

11.9 LANs: Local Area Networks

Local Area Networks began as the technology used to connect computers within buildings, to
exchange data, share resources and information. The technology subsequently migrated to
both the network core and access circuits in the form of Optical Ethernet.
LANs are covered in detail in Chapter 15.

Figure 113. Original LAN Design

166
Originally, LANs were multidrop data circuits: each computer was physically connected to a
common central bus cable.
A LAN adapter or Network Interface Card (NIC) was plugged into the motherboard of a com-
puter to implement the physical connector. Today the LAN function is usually integrated in a
computer – wired or wireless – and so the term LAN interface is used instead of NIC.
LANs changed to multipoint circuits, where each device had its own wires connected to a cen-
tral hub, which evolved into a switch.
LAN switches come in all different sizes… four ports, eight ports, sixteen ports, 192 ports.
Eight is a popular size. If one took eight PCs with LAN interfaces and wired each to a switch,
this would form a LAN.
Any PC in this group then has the possibility of communicating information to any or all other
PCs in the group. For this reason, the group of eight PCs in this example is said to form a
broadcast domain.
Several control functions are required to deal with the fact that the PCs are connected in this
way: it is necessary to transmit an address (called the Media Access Control or MAC address)
along with data to indicate for whom it is intended, as all stations in the broadcast domain
might receive the data; and an access control mechanism is required to determine which sta-
tion can transmit, as only one station can transmit at a time.
A LAN switch is a wiring hub with a processor in it. The processor examines the destination
address and directs the transmission to the correct station. Note that this is not like a tele-
phone switch, which makes a connection for the entire communication session. Here, the
transmission is “switched” to the correct destination one block of data at a time.
The original brand name for the LAN technology used today was Ethernet. This term is now
used to refer to technology that follows the 802 series of standards that are almost but not
quite exactly the same as Ethernet.
Ethernet subsequently migrated from in-building connections to the telecommunications net-
work core and access circuits. Extremely high horsepower LAN switches in different locations
connected with point-to-point fiber called Optical Ethernet is now the basis for the converged
network core and access circuits.
Please refer to Section 8.5 for Optical Ethernet and Chapter 15 for complete discussion of
LANs.

11.10 WANs: Wide Area Networks

The definition of a WAN is connecting LANs in different locations together with data circuits.
For reasons of availability, we often connect the sites with redundant paths, so that there is
more than one way to get from “A” to “C”. For this reason, we need equipment at each site that
is capable of making a route decision: which route to take to get to the desired destination.

167
 Figure 114. Wide Area Network
Router would be a good name for a device that can make route decisions.
The router needs information to use to make the routing decisions. The most popular strategy
is to assign network addresses to all of the computers. When one wishes to send a block of
data to another computer, one puts the destination’s network address on the front of the block,
forming a packet, which is sent to the router. The router uses the destination network address
as the basis of making a route decision.
There are a few choices for addressing, routing and end-to-end error checking protocols. The
TCP/IP suite of protocols, which includes the Transmission Control Protocol (TCP) and the In-
ternet Protocol (IP) are by far the most popular.
There are many choices for the telecommunications services used to connect the sites.
The most popular, flexible and cost-effective services for connecting the buildings are band-
width on demand services, in particular IP packet-based services managed with MPLS.
In days past, Frame Relay and ATM were used for this purpose, as well as dedicated lines like
dark fiber and old-fashioned T1s, and circuit-switched services like ISDN and dial-up modems.
Most of the rest of this book is devoted to exploring and explaining all of the different compo-
nents illustrated in Figure 114.
1 This was the original definition of the word “bit”, a measure of information content in a signal. This original
meaning has been lost in the mists of time, and “bit” is now used to mean binary digit: 1 or 0.

168
12
Coding, Frames and Packets
In this chapter, we explore essential functions required to communicate information on circuits.
As in the previous chapter, many of these concepts were associated with what used to be
called “data communications”, and are now used for everything: voice and video, as well as
business data services and Internet traffic.
To get started, we begin in this chapter with coding, error control, framing and addressing. This
leads to the critical concepts of frames, packets and how packets and frames and their ad-
dresses are related.
The next chapter, “The OSI Layers and Protocol Stacks” completes the story with the full list of
functions required to communicate.

12.1 Essential Functions for Communication

12.1.1 Coding
Coding means representing data in binary digits, i.e. 1s and 0s. Different techniques for coding
are employed, depending on the nature of the data.
When the data to be conveyed is a quantity, it is coded using a numbering system: binary. For
ease of use, a related system called hexadecimal is often employed.
When the data to be conveyed is a character in the English language (and selected others),
usually originally indicated by pressing a button on a keyboard, US-ASCII and variations are
used. To be able to represent other characters, including Asian scripts, unicode is slowly be-
coming dominant.
When the data is a continuous signal, such as an analog of speech, music or video, then the
coding is performed via a sampling, quantization and coding process in a codec as covered in
Chapter 5.
12.1.2 Bits
The smallest unit in communications is a logic level, either a “1” or a “0”. In a computer, all in-
formation is coded into 1s and 0s, and these are called bits.
Actually, 1s and 0s in a computer is not the original meaning of “bit”! The original use of this
word comes from information theory, the body of theoretical knowledge that is the base for
much of what we are discussing. The term bit was first used in information theory as the unit of
measurement for how much information there is.
Just as fluid ounces or milliliters are used as the unit of measurement for how much water is in
a glass, bits were the unit of measurement of how much information is in a message. This in-
formation is coded into binary digits (1s and 0s) using a coding technique. The binary digits are
then communicated over a data circuit made of a physical medium.
Because of this, when discussing data communications, we sometimes use the term binary
digit instead of bit to discuss logic levels
However, all this has been lost in the mists of time. You can safely ignore it. These days, the
term “bit” is used to mean everything that is a 1 or a 0.

169
 Figure 115. Bits and Bytes

12.1.3 Bytes
Bits are organized into groups of eight, called bytes. Historically, bytes have had various sizes:
6, 7, 8, 12 bits. Today, eight bits per byte is more or less standard.
Half a byte is called a nibble. (Really).
To refer to individual bits within a byte, they are given numbers: bit 1, bit 2, ..., bit 8. Often, due
to serial transmission, there is ambiguity as to which bit is bit 1 and which is bit 8… which end
is which.
To avoid this problem, the terms Least Significant Bit (LSB) and Most Significant Bit (MSB) are
used. The LSB is the one which has the least numeric value, the bit that changes the most of-
ten when counting.
12.1.4 Error Control
All communications is subject to errors during transmission. Normally, methods for error control
are implemented to deal with this. Error control consists of error detection and error correction.
Error detection methods include parity checking and the more reliable Cyclic Redundancy
Check, both of which involve adding redundancy (extra bits) to the transmitted data so the re-
ceiver can determine if an error happened. Error correction is implemented by retransmitting
errored data.
Forward Error Correction (FEC) means that a great deal of redundancy is added to the trans-
mitted data so that the receiver (the forward end) can determine if an error occurred, and cor-
rect the error without a retransmission.
12.1.5 Framing
Whether pressing buttons on a keyboard or downloading a web page, data happens in bursts;
there are times when there is data to be transmitted and times when there is not data to be
transmitted.
To indicate to the receiver the start and end of a group of data, markers or delimiters are
placed before and after the data. This is called framing.
12.1.6 Addressing
A single circuit may have more than one terminal connected. In the example of the multi-drop
circuit and wireless LAN from the previous chapter, many devices were physically connected,
and all devices received all transmissions.
In this case, an address is required to indicate for whom on the circuit the data is intended –
which station should react to the data, since all of them will hear it. This is called the link ad-
dress, and in the IEEE LAN standards, called the Media Access Control or MAC address.
A network is composed of many independent circuits. Routers connect these circuits together
and perform a relay function, moving data from one circuit to another.
To provide a means of determining to which circuit to relay the data, to eventually get it to the
correct circuit, devices are assigned a network address. The most popular standard method of
assigning network addresses is IP.
The link address, along with framing and error detection is contained in a frame. The network
address is contained in a packet.
Understanding how these are related, and how the link address changes as the data moves
from one circuit to the next in a network is a critical part of this chapter, and a critical part in un-
derstanding any kind of communications.

170
12.2 Coding Quantities: Number Systems
The purpose of this section is to cover the fundamentals of binary numbers, and explain hexa-
decimal, used to represent LAN interface MAC addresses and IPv6 addresses.
To explain binary and hexadecimal, we begin with decimal.
All of these are number systems, which are the coding step to represent quantities. Humans
currently use the decimal number system; computers and communication systems use the bi-
nary number system.
12.2.1 Decimal
Decimal is a number system based on tens… presumably because most people have ten fin-
gers. There are ten symbols in the decimal number system: 0, 1, 2, 3, 4, 5, 6, 7, 8 and 9.
Quantities are represented as powers of ten.
When expressing quantities in the decimal number system, we use a shorthand notation to in-
dicate how many of which powers of ten are needed to make up the quantity.
For example, when we write the number “1967”, what we mean is
1 x 103 + 9 x 102 + 6 x 101 + 7 x 100. This could also be written as
(1 x 1000) + (9 x 100) + (6 x 10) + (7 x 1).
The digits 1, 9, 6, and 7 indicate, for the appropriate power of 10, how many of that power of
ten go into making up the quantity 1967.
In other words, the numbers 1, 9, 6 and 7 are placeholders in the shorthand notation, indicat-
ing how many of the powers of ten in that place go in to making up the quantity.

Figure 116. Decimal Numbering System Example

12.2.2 Binary
Communication systems and computers use binary numbers to represent control information
and data, because computers only have two fingers: on and off.
Computers are built using transistors, and the way that a transistor is used in a digital com-
puter is like a switch: open or closed. All of the processing part of your computer is built out of
tiny switches. All of the live memory, the RAM, is built out of tiny switches.
To represent the state of the RAM, i.e. what number is in there, it is most efficient to use a
numbering system with two states… binary.
The binary number system is the same as the decimal number system, except it is based on
two instead of ten. You could even call it base 2 arithmetic if you wanted to. There are two
symbols in the binary number system: 0 and 1. Quantities are represented as powers of two.
When expressing quantities in the binary number system, we use a shorthand notation to indi-
cate how many of which powers of two are needed to make up the quantity.
For example, when we write the number “11001001”, what we mean is
1 x 27 + 1 x 26 + 0 x 25 + 0 x 24 + 1 x 23 + 0 x 22 + 0 x 21 + 1 x 20.
This could also be written as
(1 x 128) + (1 x 64) + (0 x 32) + (0 x 16) + (1 x 8) + (0 x 4)+ (0 x 2) + (1 x 1).

171
 Figure 117. Binary Numbering System Example

The binary digits 1, 1, 0, 0, 1, 0, 0 and 1 indicate, for the appropriate power of 2, how many of
that power of two are in the quantity.
In other words, the numbers 1, 1, 0, 0, 1, 0, 0 and 1 are placeholders in the shorthand notation,
indicating how many of the powers of two in that place go in to making up the quantity.
Compare this to decimal numbers, and it becomes apparent that the concept of binary and
decimal are the same – only the base is different. Decimal is based on ten and binary is based
on two.
12.2.3 Hexadecimal
The hexadecimal number system is the same as the decimal and binary number systems, ex-
cept it is based on sixteen instead of ten or two. Hexadecimal could even be called base-16
arithmetic.
There are sixteen symbols in the hexadecimal number system:
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, A, B, C, D, E, F.
Note that the letters A, B, C, D, E and F are used as symbols to represent the quantities 10,
11, 12, 13, 14 and 15 respectively, since symbols can have only one character.
Quantities are represented in hexadecimal as powers of sixteen. Just as in decimal and bi-
nary, when expressing quantities in hexadecimal, we use a shorthand notation. This indicates
how many of which powers of sixteen are needed to make up the quantity.
For example, when we write the number “7C9H”, what we mean is
7 x 162 + 12 x 161 + 9 x 160.
This could also be written as
(7 x 256) + (12 x 16) + (9 x 1).

172
 Figure 118. Hexadecimal Numbering System Example
The hexadecimal symbols 7, C and 9 indicate, for the appropriate power of 16, how many of
that power of sixteen are in the quantity.
In other words, the numbers 7, C and 9 are placeholders in the shorthand notation, indicating
how many of the powers of sixteen in that place go in to making up the quantity.
Compare this to decimal numbers, and it will become clear that the concept of binary, decimal
and hexadecimal are all the same – only the base is different. Hexadecimal is based on six-
teen; decimal is based on ten and binary is based on two.
12.2.4 Common Use for Hexadecimal
Why would we bother with a numbering system based on 16s? Not many people have sixteen
fingers…
Hexadecimal (or hex for short) is perhaps most often used in practice as a short form for bi-
nary numbers.

Figure 119. Hexadecimal as a short form for binary numbers

Long binary numbers written on screens or on paper are cumbersome and largely incompre-
hensible to humans. How successful would attempting to communicate 011111001001 to an-
other person be? How easily would they remember this number?
A method more useful for human comprehension would be to organize bits in groups and rep-
resent values of the groups with symbols.
Since the world has more or less standardized on the octet, i.e. groups of 8 bits, a numbering
system that is efficient for groups of 8 bits would be indicated. However, defining symbols for
all values that could be represented by 8 bits would require 28 = 256 symbols (like Chinese),
too many to be easily comprehended unless learned from birth.
Using decimal to represent the value of a byte is possible, and is used in IP version 4’s dotted-
decimal notation, but is awkward and difficult. It requires a computer program to do the conver-
sion in one direction.
The solution is to divide a byte in half; into two groups of 4 bits.

173
To represent any pattern four bits long requires 24 = 16 symbols. Therefore, a byte can be rep-
resented with two of these symbols, instead of eight bits. Since 161 = 24, conversion from bi-
nary to hex is simple: the binary number is segmented into groups of four bits, and each set of
four bits is converted individually to their hex equivalent.
Consider the binary representation of 1993D: 011111001001.
For readability, instead of commas we use semicolons: 0111:1100:1001.
The hexadecimal equivalent of this number is 7C9H.
The hex version uses only 1/4 the symbols, and so is easier to write and to pronounce.

12.3 Coding Text

12.3.1 ASCII
Figure 120 depicts the American Standard Code for Information Interchange (ASCII, pro-
nounced ASK-EE). This is a standard method of representing keystrokes, i.e. text, using binary
digits. The official standard for ASCII is ANSI X3.4-1986, which defines a 7-bit code.
For example, the ASCII code for a capital “A” is, at the top of the row 100 and at the left of the
column 0001 to make 1000001.
There are a number of codes other than the alphabet and punctuation. These are control
codes, for example BEL (7) would tell the receiver to ring its bell, CR (13) is carriage return, FF
(12) is form feed and so forth.
ASCII was primarily defined for teletypes, i.e. printed English characters, which (including both
uppercase and lowercase) can easily be supported within a limit of 128 codes, and so speci-
fied codes that were 7 bits long.
This saved bandwidth compared to 8-bit codes, and allowed the appendage of a parity bit for
error detection to make up an 8-bit byte.
However, as computers standardized on octets, more languages with different characters were
required to be supported, and parity checking was abandoned, sets of 8-bit codes were re-
quired.

174
 Figure 120. ASCII Code Set
IBM used an 8-bit code set called the Extended Binary Coded Decimal Interchange Code
(EBCDIC) code set on mainframes. With respect to this chart, an EBCDIC chart would be up-
side down, backwards, twice as big and none of the characters would be the same. Other than
that, EBCDIC is exactly the same thing as ASCII - it is a standard way of coding keystrokes
into bytes.
Various ad-hoc extended ASCII code sets were defined, particularly in Microsoft’s Disk Operat-
ing System (DOS), adding an extra bit, doubling the size of the table. The characters in the ex-
tended ASCII table are Greek letters, box drawing characters, és and so forth. However, there
were dozens of variations to choose from, so ¬ sometimes came out as î, if for example a
computer and printer did not agree on what extended ASCII variation is in use.
The International Organization for Standardization (ISO) and their friends at the International
Electrotechnical Commission (IEC) eventually published a standard code set ISO/IEC 8859-1
(Part 1 of ISO/IEC 8859), an 8-bit code set for the Latin alphabet. The original definition of this
code set did not include the control codes, and was not used in practice, but did form the basis
for two 8-bit code sets that are now in wide use: ISO-8859-1 and Windows-1252.
ISO-8859-1, defined in RFC1345, is a superset of the original ISO/IEC 8859-1 to include con-
trol codes and is at present widely used for plain text web pages and email. It is the default en-
coding for MIME type “text”.
Windows-1252 is almost the same, but substitutes a number of printable special characters
like left double quotes where ISO-8859-1 has control codes. These special characters are ren-

175
dered as a question mark or hollow box when displayed on a web page by a browser, since
they are undefined characters in ISO-8859-1, which the browser uses by default.
12.3.2 Unicode
Unicode and its Unicode Transformation Format (UTF) may end up becoming universal stan-
dard codes for character sets.
Unicode defines a codespace of 1,114,112 codes in the range 0 to 10FFFFH and methods of
representing them, called transformation formats. The most popular is UTF-8, which allows
one to four bytes to represent a character.
It is normal to reference a Unicode code by writing “U+” followed by its value in hex. Often,
double-byte or four hex characters are used, for example U+005A for Z and U+548C for 和.
In HTML, characters may be expressed as &# followed by the decimal value of the code and a
semicolon. For example, Z would be rendered as Z. Or, characters may be expressed as
&#x followed by the hex value of the code and a semicolon, for example, Z.
The characters allowed in URLs (web addresses) may be represented as % followed by the
hex value of the code (for example, %5A).

12.4 “Asynchronous”: Start/Stop/Parity

Formatting and packaging characters to be transmitted one at a time used to be called asyn-
chronous.
As this method was used mostly for connecting modems through serial ports and dumb termi-
nals, it has diminished greatly in importance and the details of start, stop and parity bits are no
longer part of the required knowledge in the telecom business.
Feel free to skip to Section 12.5 “Frames and MAC Addresses”.
12.4.1 Asynchronous Communications
It would have been more precise to refer to this as “a type of asynchronous communications”,
because it is possible to transmit many things asynchronously.
From the Greek, asynchronous means “not timed”. The word asynchronous was associated
with transmitting characters one at a time, because pressing a key is not a timed event. The
time between keystrokes is random.
More generally, another way of thinking of the meaning of asynchronous is from the communi-
cation circuit’s point of view: statistically speaking, if someone is doing asynchronous commu-
nications, most of the time, they are doing nothing. Every once in a while they send some infor-
mation, then go back to doing nothing.
In this older method of communications, characters were coded into seven-bit patterns using
the US-ASCII codes of Figure 120.
12.4.2 Framing: Start and Stop Bits
Communications is for the most part serial, that is, there is one data circuit established, and
when there is a group of bits to communicate, the bits are transmitted one after another in a
sequence in time on the circuit.
When communicating asynchronously, normally nothing is happening, and data can be ran-
domly transmitted. Because of this, there must be a mechanism for the receiver to detect the
beginning of a serial group of bits, capture the group and re-package it into a parallel form at
the receiver. The general term for this is framing.
For transmitting keystrokes, the framing is performed on each byte. First, a start bit is transmit-
ted to warn the receiver that a byte is coming, then the character code, followed by a stop bit.
To understand what a start bit is, first, consider that most serial communication systems are
also binary communication systems: they have only two levels. This is not a tri-state system
that has three levels – high, low and disconnected. This is a binary system that has only two
levels – high and low.

176
 Figure 121. Start/Stop/Parity Format
Since the circuit is never disconnected, it is always communicating one or the other. One of
these levels must be chosen as the idle condition, i.e. the condition that is presented to the re-
ceiver when nothing is happening.
Understanding this, deciding what a “start bit” should be is simple. It is NOT the state chosen
to be the idle condition. This idle to not-idle transition warns the receiver that a byte is coming.
A stop bit is the idle condition to guarantee an idle to not-idle transition when the next start bit
happens.
12.4.3 Parity Checking
To perform error detection at the receiver, a single extra bit called the parity bit is appended to
the 7-bit ASCII code and sent over the communication circuit. Just one bit is used, to imple-
ment a simple technique while minimizing the overhead in extra bits.
This is one reason why ASCII was designed as a 7-bit code: 7 bits of data and one bit of error
detection to make up an 8-bit byte.
There are two parity rules: even parity and odd parity. The transmitter and receiver must de-
cide in advance and agree upon which rule they will use, then stay with that rule.
Under the even parity rule, the extra bit is set at the transmitter so that the total of all of the bits
including the parity bit is even. The receiver checks to see if the total is even. If not, it knows
an error happened in transmission, and can flag a parity error. If the total is even then there
were no errors.
The odd parity rule works in the same manner.
Unfortunately, this even/odd scheme does not work if there are two bits in error, or four, or
six… and errors happen in bursts.
The probability that if there is one bit in error, that the bit beside it is also in error is between 20
and 50% depending on the physical medium being used. For this reason, parity checking is al-
most useless.
Many systems ended up using no parity, and instead of using one bit for parity, use it for data.
This is often represented as a code “8N1” that has to be typed into a setup screen on software.
Using what was the parity bit for data instead extends the ASCII code to 8 bits as discussed in
Section 12.3.1.

12.5 Frames and MAC Addresses

177
Formatting and packaging data one character at a time is not efficient. Percentage-wise, the
overhead for framing and error control is very high.
It is more efficient to code characters, quantities or analogs into bytes, then group many bytes
into a block and transmit the block as a package.

Figure 122. A Frame

Both the framing and the error detection is then performed on a per-block basis rather than on
a per-byte basis, making this more efficient and more reliable.
To deal with multiple stations on a circuit, all of whom will receive the transmission, a link ad-
dress is prepended to the block, indicating which station should react to the transmission.
A block of data with framing, error detection, a link address and other control information is
called a frame.
Receipt of a frame can be acknowledged by the receiver. If an error is detected, the frame can
be re-transmitted.
Successful transmission of a frame results in communication of the payload from one com-
puter to another that are on the same physical circuit, or as we will see in Chapter 14, in the
same broadcast domain.
12.5.1 Data Link Protocol
The method for formatting frames, sending and acknowledging them is called a data link proto-
col. The mother of all data link protocols is the High Level Data Link Control protocol (HDLC).
The most popular is the combination of IEEE standards 802.2 and 802.3, used on LANs.
Early designs for transmission of blocks of data by IBM for mainframes called this synchronous
communications. This term should not be used because it means many other things as well.
“Layer 2”, “MAC Layer”, “Frame transfer” or “broadcasting frames” would be more accurate.
12.5.2 Framing
The framing is a special bit pattern one or more bytes long identifying the beginning and, in
some cases, end of the frame.
12.5.3 Address
An address is usually included at the beginning of the frame. The address on the frame is used
to indicate which device on the same circuit the frame is intended for.
Recall a multi-drop circuit is a single circuit with multiple derivations, and when the primary sta-
tion sends out data, every secondary station receives it. The frame address is used to indicate
for whom the data is intended; which station should react to the data, since every station on
the circuit will get it. The frame address in a LAN application is the LAN interface Media Ac-
cess Control (MAC) address of the intended receiver.
12.5.4 Control Field
A control field follows the address. In 802.2, this contains the length of the payload. In HDLC,
the control field can identify the type of frame: Information frame, Supervisory frame or Unnum-
bered frame. An information frame is used for data transfer. Sequence numbers, acknowledg-
ments and a poll/final indicator are placed in the control field for the information frame. The
other two types of frames are used for data link control, and have codes indicating which ac-
tion to take.
12.5.5 Payload

178
The data field or information field or payload follows the control field. This field contains the
block of data that is being communicated.
The data field can be any length. In practice, it is a fixed length on a particular circuit or particu-
lar virtual circuit. The length is determined by hardware capability, circuit noise level and how
the data is formatted on cabling systems passing data to this one. Common implementations
have data field sizes ranging from 48 bytes to 8,192 bytes. LANs tend to use frames about
1500 bytes long.
12.5.6 CRC: Cyclic Redundancy Checking
An error detection scheme called Cyclic Redundancy Checking (CRC) is implemented using a
Frame Check Sequence (FCS) appended to the block. The receiver uses this extra information
to determine if any errors have occurred anywhere in the frame during transmission. With the
right choice of FCS, this method is very reliable.

Figure 123. Cyclic Redundancy Checking

The Frame Check Sequence is the redundancy, and the process is called a cyclic check be-
cause, amongst other reasons, the mathematical operation used to both generate the FCS at
the transmitter and check for errors at the receiver employs polynomial codes whose basis ele-
ments are cyclic permutations of each other.
The FCS is chosen so that the entire frame is exactly divisible by some predetermined number.
If the division at the receiver results in a remainder, this indicates that an error occurred some-
where in the frame.
The explanation of how the FCS is chosen requires a mathematical proof, which is presented
for those interested.
Feel free to skip to Section 12.6 “Networks, Packets & IP Addresses”.
Using modulo-2 arithmetic, we can think of the data as a large number k bits long. For exam-
ple, in a 512-byte chunk of data, k = 512 x 8 = 4096. To this number, we would like to append
an n-bit-long FCS.
If we define:
T as the (k + n)-bit long frame to be transmitted,
M as the k-bit long message,
F as the n-bit FCS,
P as the predetermined divisor, (n + 1) bits long,
We could write T = 2nM + F. (1)
We want T/P to have no remainder.
Dividing 2nM by P, we get 2nM/P = Q + R/P, a quotient and remainder. (2)
This remainder R is used as the FCS.
To check to see if this satisfies the condition, substitute into (2) into (1):
T/P = (2nM + F)/P = (Q + R/P) + R/P.
Since any number added to itself modulo 2 equals zero, we have
T/P = Q, with no remainder, which satisfies the requirement.
Note that P is used both to generate the FCS at the transmitter and to check for errors at the
receiver. Common choices for P are in North America and Europe respectively, CRC-16: P(X)
= X16 + X15 + X2 + 1, CRC-ITU: P(X) = X16 + X12 + X5 + 1, which generate 16-bit FCS with X=2.
Other 12-bit (telex/twx) and 32-bit (LAN) versions are also used.
This technique detects the following:
• All single-bit errors,
• All double-bit errors,

179
 • Any odd number of errors,
• Any burst error (n-1) bits long or less, and
• Most larger burst errors.

12.6 Networks, Packets & IP Addresses

The word “network” comes from fishing nets, where there are many ropes tied together to form
a mesh, and one could trace many different possible routes between any two knots or nodes
on the net-work.
In the communications business, the word network is used in the same way. Networks are im-
plemented by high-capacity circuits connected by network equipment. In the case of packet
networks, the network equipment is a router.

Figure 124. Networks and Packets

In most cases, the router will be connected to more than two circuits, and so there will be multi-
ple possible routes across the network.
The definition of a network is having to make a route decision: which route to take to get to the
destination.
The network equipment relays or forwards the data in hops from one circuit to a different circuit
to another circuit to reach the far end.
If there are no one-of-many route decisions being made, and the data is not relayed from one
circuit to another, for example, if the data is broadcast to every station, then this is not, strictly
speaking, a network.
Access circuits are connected from the end-user location to the network equipment. Users
transmit data end-to-end across the network by transmitting it over their access circuit to the
network equipment, whence it is relayed or forwarded from one piece of network equipment to
another until the far end is reached, then delivered over a far-end access circuit.
Data to be transferred over a network is formatted into packets, which are sometimes also
called datagrams or Network Protocol Data Units. We will avoid getting caught up in jargon
and use the term packet.
Packets are blocks of data with network control information. The most interesting type of net-
work control information is the network address, which indicates the final destination of the
packet.
Network equipment, such as routers, look at the network address to decide which route to take
to get to that destination. The router implements the decision by forwarding the packet to the
next hop, in other words, transmitting the data to the next router along the route.

180
The most popular standard protocol for formatting packets with network addresses is the Inter-
net Protocol (IP).

12.7 Packets vs. Frames

It is important to understand how packets and frames are related.
Packets are for networks. A packet is a block of user data, such as a piece of an e-mail mes-
sage, with a network address on the front. The network address is the final destination.
Network equipment like routers receive a packet on an incoming circuit, examine the network
address, use it to make a route decision, then implements the decision by forwarding the
packet on a different circuit to the next router.
A frame is a lower-level idea. Frames are used to communicate between stations on the same
circuit. The circuit may have multiple stations physically connected onto it, like a wireless LAN,
or may have only two stations like a point-to-point LAN cable.
A frame has framing to mark the beginning and end, sender and receiver addresses to indicate
the stations on the circuit, control information, a payload and an error detection mechanism.
The frame is transmitted on the circuit, and all stations on the circuit receive it. If an error is de-
tected at a receiving station, the frame is discarded and might have to be retransmitted some-
how.
If no errors are detected, the end result is that the payload is communicated to the correct sta-
tion on the same circuit with no errors.

Figure 125. Packets are carried as the payload in frames

The payload of a frame is a packet.
The main purpose of packets is to append a network address to your data. The network ad-
dress is used by network equipment to make route decisions: to relay the packet from one cir-
cuit to a different circuit.
To actually transmit a packet on a circuit within the network, the packet is inserted as the pay-
load in a frame, then the frame is broadcast on the circuit.
12.7.1 Link Address vs. Network Address
Notice that there are two addresses: the network address and the link address. The network
address on the packet is the final destination, and so does not change. The most popular stan-
dard for network addresses is IP.
The link address on the frame indicates the destination on the current circuit, and so is
changed as the data is forwarded from one circuit to another. The most popular standard for
link addresses comes from the 802 series of standards from the IEEE, which refers to them as
MAC addresses.

12.8 IP Packets
The Internet Protocol (IP) is part of the TCP/IP suite of protocols developed by the US military,
now used on the Internet and carrier networks. IP is a network protocol, defining the network
packet format and network addressing scheme.

181
IP was originally published as military standard MIL-STD-1777, then updated in Request for
Comments RFC 0791 Internet Protocol. IP version 4 (IPv4) is currently in use. IP version 6
(IPv6) is coming next.

Figure 126. IP Version 4 Packet

12.8.1 IP Packet Header

Included in the IP packet specification is the header, which is the “network control information”,
added on to the front of the block of data.
Figure 126 illustrates the elements in the IP version 4 (IPv4) header:
• The version of IP being used (version 4 is popular, IPv6 next);
• How long the header is;
• What service the packet is being used for;
• How long the packet is;
• Do not Fragment (DF) field, used to control whether the packet may be segmented to be
carried in a number of (smaller) frames or not;
• More Fragments (MF) field, used to indicate whether the packet has been chopped into
segments, and if this is the last segment or not;
• Time to Live field, which starts with some initial value and is decremented every time the
packet passes through a router. The packet is discarded when this reaches zero, to prevent
packets from endlessly circulating;
• Which transport protocol is using this packet service;
• A checksum: like a CRC, but faster to calculate, to protect the header information (does not
cover the payload in the packet);
• The most interesting part of the IP header: the IP address of the sender and of the destina-
tion. This is the main purpose of IP.
The Identification, Don’t Fragment (DF), More Fragments (MF) and Fragment Offset fields are
used when the packet does not fit inside a frame and has to be broken into parts or segments
or fragments and carried in separate frames.
When that occurs, a copy of the IP header is repeated at the beginning of each fragment – dra-
matically increasing overhead – and typically if one of the fragments is lost, all of the fragments
are discarded.
12.8.2 Connectionless Network Service
Using IP implements a connectionless network service, which means that at the network level,
there is no communication with the receiver.
Packets are sent on the network to fend for themselves. There are no guarantees that the
packets will be delivered.

182
In fact, in an IP network, there are no guarantees that a packet will be transmitted, when that
might happen, and how often that might happen.
There is no error check on the payload, only on the data in the header.
12.8.3 Relationship to TCP
Since there is no guarantee an IP packet will be received, and no error checking of the payload
being carried in the packet, a higher-level protocol is required to perform end-to-end error
checking.
The most popular higher-level protocol for end-to-end error checking is the Transmission Con-
trol Protocol (TCP). The abbreviation TCP/IP is sometimes used, but should be avoided, as it
inaccurately suggests that it is necessary to use TCP and IP together, or that TCP is part of the
network.
IP is the packet and address format for the network. TCP is a protocol for communicating be-
tween the sender and receiver to ensure error-free delivery of messages. TCP is used for file
transfers including web pages and email, retransmitting message segments that are lost or er-
rored.
The User Datagram Protocol (UDP) is an alternative to TCP. It implements error detection, but
does not retransmit lost or errored message segments. UDP is used for streaming applications
like voice and video.

183
13
The OSI Layers and Protocol Stacks
13.1 Protocols and Standards
We use the term protocol in the data communications business the same way it is used in the
diplomacy business: it is a plan for how two different systems will interact.
In diplomacy, protocol officers get together in advance and hammer out the plan: it says who is
going to greet whom at the bottom of the steps of the aircraft, what color the carpet is going to
be, what music the band will be playing, are you allowed to be sitting down while the president
of the United States is in the room… the plan on how two countries will interact.
To communicate, it is necessary to have a set of conventions that specifies how the systems
are going to communicate. This is the definition of a protocol. Mutual adherence to an agreed
protocol or set of protocols makes communication possible.
13.1.1 Functions To Be Performed
Quite a number of areas and functions must be covered in a communication protocol.
Taking e-mail as an example, first, it is necessary to agree what the format of the message will
be.
How will the message be coded into 1s and 0s? Will it then be encrypted? There had better
have an agreed plan for that, or not much communications will be happening.
Most communications today is client-server… and e-mail is an easy example. When checking
Outlook-type email, it is necessary to log on to the mail server with a username and password
and be authenticated… so part of the protocol has to be how to transmit usernames and pass-
words to the server.
One could imagine the mind-numbing complexities created if it is desired that the password not
be transmitted as clear text, but encrypted as a measure against eavesdropping… how to
transmit the decryption key for the password without encrypting it?
Once authenticated, then it is necessary to transport the message from the server to the client,
and there are a number of things that have to be figured out.
Segmentation and reassembly are usually required, breaking up the message into manage-
able pieces for transmission and putting it back together at the receiver… in the correct order.
The segment of the message has to be encapsulated in control information. An example of
control information for a segment of data is a network address.
Once a packet with a network address is created and transmitted to a router, how are routers
going to make routing decisions based on those network addresses? And how is the route de-
cision-making kept up to date as new links are added, others are removed or become busy?
Probably the most important aspect is error control: sending data with errors and not knowing
about it is probably worse than not sending any data at all. Sometimes error control is per-
formed on each link. Sometimes not. It ends up being necessary to check errors end-to-end
between the sender and recipient.
How is flow control implemented: when one system can’t process information as fast as the
other, and has to have a way of temporarily interrupting the flow of data.
How is access control implemented – when there is more than one station on the link, which
gets to transmit next?
At the bit level there are things that have to be specified: what physical medium to use, and
how to represent the bits on the physical medium. How will conversions between different me-
dia and different bit rates be implemented? All this and more has to be part of the plan.

184
 Figure 127. A protocol is a plan

13.1.2 Monolithic vs. Structured Protocols

There are two basic choices for the plan: monolithic or structured protocols.
A monolithic protocol would embody all of the required functions in a single plan. The problem
with this approach is that it becomes unwieldy when all possible variations are included in the
single package, and makes maintenance impossible.
A structured approach, where the totality of functions is divided into easy pieces, then individ-
ual protocols covering each of the pieces are developed is more workable.
This allows mix and match of protocols for different functions on the systems: for example, ev-
erything could be the same on all systems, except that the access circuit at one end will be
wireless, the network is implemented with fiber, and the access circuit at the other end is DSL
on copper wires.
Other than that, everything from frame and packet format and addressing to the message for-
mat and coding is the same across the entire system.
13.1.3 Open Systems and Standards
In an open system, the protocols are published: all information necessary to implement com-
munications is available to the public.
It would be possible for an individual to develop a set of open, structured protocols for commu-
nications. This would only be useful if everyone, or at least a critical mass of users, agreed to
use that particular set of protocols.
We are always interested in implementing standard protocols.
A protocol is a plan.
A standard is when everyone agrees on a particular plan.

13.2 ISO OSI Reference Model

One approach to implementing structured data communications protocols is the OSI Refer-
ence Model. In 1983, the International Organization for Standardization (ISO) adopted a “Basic
Reference Model for Open Systems Interconnection”.
The purpose of this model was to “provide a common basis for the coordination of standards
development for the purpose of systems interconnection, while allowing existing standards to
be placed into perspective within the overall reference model”.

185
A key point is that this is a model for discussing protocols and standards. It does not specify
how to actually perform a function, but instead describes what functions must be performed,
and organizes these functions into manageable groups or layers.
13.2.1 Layers
A layer is a subset of the totality of functions that must be implemented to interwork diverse
systems. Protocols are established for each layer.
The physical connection between the systems is specified by one layer, and implemented in
hardware and signaling using electricity, radio waves or light to communicate 1s and 0s be-
tween the systems.
All of other functions, all of the other layers, are implemented in software. A particular software
package may implement one or more layers.
13.2.2 Separability of the Layers
The choice of functions included in each layer was made so that the layers were separable:
the functions performed by one layer are independent of the functions performed by another
layer.
This allows systems to choose a protocol for a particular layer without having to take into con-
sideration the choices made for other layers. For example, the choice of email message format
is independent of the choice made for network packet format.
Dividing functions into separate layers is also useful for understanding the different functions
that must be performed and how they are implemented, being able to discuss separate issues
separately and not get things confused.

Figure 128. Layers in a stack

13.2.3 Protocol Stacks

In the OSI Model, a piece of software implementing a layer following some protocol will per-
form its task, then call on a lower-layer piece of software to perform some utility function for it.
The lower-layer piece of software takes the data it is handed, performs its function following
another protocol, then asks a yet lower layer to perform some utility function for it in turn.
For example, the top layer might generate an email message, then ask the next layer down to
encrypt the message. The lower layer takes the message and encrypts it, then might ask the
next layer down to transport the encrypted output reliably to the far-end destination.
This process repeats until the bottom layer is reached, which transmits bits one at a time on an
outgoing circuit.
Since the layers and the protocols implementing them work in this chain-like fashion, they are
often depicted as sitting on top of each other in a stack like a layer cake, and the collection is
called a protocol stack.

13.3 The OSI 7-Layer Model

The OSI Reference Model is referred to as a 7-layer model because the total set of functions
required to interwork diverse systems was defined and then broken up into seven groups or

186
layers, arranged in a hierarchy. Each layer has a name and a number. The numbering starts at
the bottom:
1: Physical Layer The physical layer provides a raw bit stream service. It moves 1s and 0s
between the systems. This is all it does, but it has to do this completely. The physical layer in-
cludes the mechanical, electrical, functional and procedural specifications for moving binary
digits over a physical medium.
2: Data Link Layer The data link layer manages communications on a single circuit, a single
link. There may be several stations connected to the circuit as is the case with a wireless LAN,
or there may be just two stations on the link, as is the case with a LAN patch cable. The data
link layer performs access control, flow control and error detection on the link, transmitting
frames on the physical medium. This allows communications of blocks of data from one device
to another that are on the same circuit.
3: Network Layer The definition of a network is multiple data links connected by network
equipment. Instead of broadcasting data to all stations on all of the links, data is relayed from
one link to the next to eventually be delivered to the correct link to which the desired station is
attached. A router moves packets from one link to another, essentially a forwarding function.
Knowing which link to forward the data on is the routing part of the story. All of these functions
are the network layer.
The first three layers working together form a communication network, giving the user the abil-
ity to send data to a destination on a different circuit.
4. Transport Layer The transport layer implements two major functions. One is reliability. The
other is network connection sharing.
Some network protocols, IP for example, do not provide guaranteed delivery of packets. The
transport layer communicates between the source and destination across the network to verify
that each segment of a message is successfully received, and in the case of file transfers, re-
transmits lost segments.
The second function performed by the transport layer is to identify the software application the
data is intended for at the far end. There may be many apps running on the far-end computer.
The port number in the transport layer header indicates which app the segment of data is for.
This allows multiple applications to use the same network connection, for example, an email
program and a browser can both receive packets over a single shared network connection.
The port number indicates whether an incoming packet is for the email application or the
browser application.
These four layers working together provide a transport service, moving data reliably from an
application on one system to an app on another system. This is also called a socket in the IP /
UNIX worlds .

187
 Figure 129. The OSI 7-Layer Model
The remaining three layers are called the upper layers:
5. Session Layer The session layer manages sessions between applications, including initia-
tion, maintenance and termination of information transfer sessions. Usually this is visible to the
user by having to log on with a password in the case of client-server sessions.
6. Presentation Layer The presentation layer is very important: this is the coding step, repre-
senting the message to be communicated in 1s and 0s. ASCII is an example of a presentation
layer protocol. Compression and encryption also fit into the presentation layer – they are meth-
ods of coding messages into 1s and 0s, as are codecs for voice and video digitization.
7. Application Layer Sitting on top of all of this is the application layer. The application layer
defines the format of the messages that will be exchanged, and usually implements a Human-
Machine Interface.
Using the application layer is a person.
The person interacts with the system via the Human-Machine Interface implemented by the
application layer, that lets the person create a message.
In turn, the application layer would ask the presentation layer to code it, and then that would
ask the session layer to open a session with the far-end piece of software, and in turn ask the
transport layer to move it reliably to a particular application on the far-end system.
The transport layer would ask the network layer to move it to the far-end computer, perhaps on
a best-efforts basis, then the network layer will move a packet to the next hop, the next router,
by putting the packet in a frame and transmitting the frame one bit at a time on a physical con-
nection like a LAN cable or wireless frequency.
At the far end, the network-layer packet is received in a link-layer frame over a physical-layer
connection. The content of the packet is extracted and passed to the transport layer, which
would perform error recovery if necessary then pass it to the correct computer program.
That computer program would pass the data to its presentation layer to decode it, then to the
top layer, the application layer to display it as a message to a human.

188
13.4 Physical Layer: 802.3, DSL, DOCSIS
Layer 1, the physical layer, provides a raw bit stream service to higher layers.
The physical layer includes the mechanical, electrical, functional and procedural specifications
for moving binary digits over a physical medium.
The mechanical specification includes which type of physical medium will be used. This could
include copper wires - shielded cables, twisted pair, or coaxial cable; it could include space (ra-
dio), or optical fiber. The connectors or antennas are also specified.
The electrical specification dictates how binary digits will be represented on the physical
medium - the modulation technique or digital line code.
The functional specification indicates how many individual wires or circuits will be used to
make up a single communication channel, and the function of each circuit.
The procedural specification specifies the relationship between the circuits: are some for data,
some for control; is there a circuit that has to operate first, one second and so forth.
There are many different physical layer protocols.
Any kind of modem implements a physical layer protocol by specifying the physical medium
and how bits are represented on it. This includes DSL modems that operate over twisted pairs
of copper wires, DOCSIS cable modems operating in channels on a Hybrid Fiber-Coax sys-
tem, all kinds of “digital” wireless systems where modems communicate 1s and 0s over radio
channels, and of course, old-fashioned dial-up modems.
LANs include a physical layer protocol – the LAN interfaces that provide the familiar LAN jack
implement signaling using pulses of voltage on twisted pairs in Category 5, 5e and 6 cables
following the 803.2 standard.

Figure 130. A Physical Layer protocol specifies the physical medium and how bits are to be represented on it.
Optical Ethernet employs optical transceivers to signal using pulses of light on fiber to commu-
nicate bits between devices.
The older SONET includes a physical layer protocol, specifying how the laser is turned on and
off to signal anywhere from 500 Mb/s to 10 Gb/s.
Repeaters, amplifiers, pulse shapers, DWDM frequency spacing and anything related are part
of a physical layer protocol.
ISDN Basic Rate Interface (BRI) - digital telephone lines - include a physical layer, specifying
pulses on the loop.
The old T1 technology implemented a physical layer protocol, moving 1.5 Mb/s over four cop-
per wires using the AMI line code, CSUs and DSUs.
The old serial port standard RS-232 is a physical layer protocol.
The list goes on and on.

13.5 Data Link Layer: 802 MAC

Layer 2, the data link layer, is concerned with communications between devices on the same
physical circuit, or more accurately, between devices in the same broadcast domain.
There may be just two stations, as in a point-to-point link; or there may be many stations, as in
a wireless LAN, but in both cases, the stations are directly connected on the same physical cir-
cuit, and are in the same broadcast domain.
Most implementations of a data link protocol do only error detection, and discard data that has
errors in it. Some implementations do both error detection and error correction.
13.5.1 LANs, Frames and Layer 2 Switches

189
The data link protocol encapsulates segments of data to be transferred into frames: adds a link
address and control information in a header, a frame check sequence in a trailer, and framing
around the whole lot.

Figure 131. LANs implement Layer 2 and Layer 1

The frame is then broadcast on the physical circuit, and any stations in the same broadcast do-
main might receive it.
At each receiver, errors are detected by the data link software performing a Cyclic Redundancy
Check (CRC). If it passes, the contents or payload from the frame is extracted and passed up
to the next higher layer.
If it fails the CRC, the frame is discarded and will have to be re-transmitted somehow. This is
usually implemented by a higher layer protocol like TCP, but can also be done by the data link
protocol, as in fax machines.
The dominant method of implementing the Data Link Layer is with LANs. A LAN includes Layer
1 (physical cabling and LAN interfaces) and Layer 2 (frames and link addresses).
LANs move frames between computers on the same physical circuit, or connected to the same
LAN switch. The most accurate way of saying this is LANs move frames between devices in
the same broadcast domain.
Since LANs are implementing Layer 2, LAN switches, particularly the very high capacity ones
in carrier networks, are also referred to as Layer 2 switches. This is covered in detail in Section
14.4.
13.5.2 MAC Frames and MAC Addresses
The frame format for LANs is IEEE standard 802 Media Access Control (MAC) service plus
802.2 Logical Link Control (LLC), hence the use of the term MAC frame and MAC address in
conjunction with LANs.
IEEE standard 802.3 defines particular implementations of the MAC service on twisted pair,
coaxial cable, radio and fiber, specifying the framing, timing, method of representing bits and
other Layer 1 functions. 802.3 is commonly referred to as Ethernet.
13.5.3 Other Data Link Protocols
The ISO High-Level Data Link Control Protocol (HDLC) is the mother of all data link protocols.
The ITU Link Access Procedures (LAP-) for public communication networks, the ANSI Ad-
vanced Data Communication Control Procedures (ADCCP), IBM’s Synchronous Data Link
Control Protocol (SDLC) are all data link protocols derived from or similar to HDLC. The legacy
Frame Relay network service from carriers is usually discussed as a Layer 2 protocol.

190
13.6 Network Layer: IP and MPLS
The data link layer handles communications between devices on the same physical circuit.
What if there is not a single physical circuit, but 86 of them, and it is not desired that data be
broadcast to all stations on all 86 circuits, but rather routed or switched and delivered to a par-
ticular destination? This is the definition of a network, and Layer 3 of the OSI model.
A network is made up of many network devices like switches or routers connected with high-
speed data links. Access circuits are provided to the network equipment to allow users to send
data into the network. The network equipment moves data from one circuit to another, es-
sentially a relay function.
Networks always have two points of view: from the user’s point of view, how does the user in-
dicate to the network where the data is to be delivered? This information usually takes the form
of a network address.
From the network’s point of view, if it receives data to be sent to a particular network address,
how does it actually decide which route to take to reach that destination address?
13.6.1 Packet-Switched Networks
The most widely-deployed type of network used to be a circuit-switched network, the traditional
Public Switched Telephone Network (PSTN). To place a call, the caller tells the network the ad-
dress of the person to whom they wish to connect - their telephone number - then a route is
chosen and then trunk circuits are switched in and reserved to form an end-to-end path for the
duration of the call.

Figure 132. Networks are made of high-capacity links connected by routers or switches. Access circuits connect
the users to the network.
This is now replaced with a packet-switched network, where trunks are not reserved for the du-
ration of a communication session, but rather voice, video, Internet traffic or anything else is
segmented and placed in packets that are transmitted into the network on an as-needed basis,
and relayed from one router to the next, interspersed with many other users’ packets until it is
delivered to the far end. This is also called bandwidth on demand.
Every destination on the network is assigned a network address. To transmit data to a destina-
tion, the address of the desired destination is placed in the header at the beginning of the
packet, and each router uses the destination address to determine the next hop.
The router implements the routing by taking the packet from an incoming circuit (or more pre-
cisely, the incoming broadcast domain), and transmitting it out on a different circuit. Routers
perform essentially a relay function. Knowing which circuit to move the packet to is the routing
part of the story. This whole process is called packet switching.

191
The most popular protocol for assigning network addresses and formatting packets is IP, the
Internet Protocol, developed by the Department of Defense and is now maintained by the Inter-
net Engineering Task Force (IETF). It is, of course, the protocol for network addresses and
packet format on the Internet, and also used for networks not directly connected to the Inter-
net… and has become the only standard for packets worth learning about.
13.6.2 Routing Table Updates
The devices that perform the routing of packets must have a way of making route decisions. In
general, they use routing tables and look up the routing for each packet based on the destina-
tion address.
The routing tables are kept updated using protocols like Open Shortest Path First (OSPF) and
Border Gateway Protocol (BGP). There are other proprietary protocols also available for this
purpose.
13.6.3 MPLS
It should be noted that IP networks generally have Multiprotocol Label Switching (MPLS) im-
plemented to allow the management of traffic and transmission characteristics. MPLS is an im-
plementation of virtual circuits, where a path for packets is pre-determined and programmed
into the routers by equipment at a Network Operations Center. In this case, MPLS replaces
OSPF and IP addresses for routing.
MPLS is covered in detail in Chapter 16.

13.7 Transport Layer: TCP and UDP

13.7.1 Reliability
With IP, there is no guarantee that a packet will be transmitted, when that might happen, or
how often it might happen. There is no guarantee that if a packet is transmitted, it will be re-
ceived.
In fact, in IP, there is no way for the transmitter to know whether it was received or not. Nada.
So, what if some packets are, indeed, not delivered? This means segments of the message
are missing... so a protocol is required to deal with the problem.
Dealing with missing data at the receiver is one of the two functions performed by Layer 4, the
transport layer.
Before a segment of a message is put in an IP packet by the sending device, it is passed to its
Layer 4 software, which puts a sequence number and error check on the message segment,
then passes it to the IP software.
At the far end, its Layer 4 software checks the sequence number and error check on the mes-
sage segment. If a message segment fails the error check, or is missing, there are two basic
strategies: retransmit the missing segment, or interpolate it.

Figure 133. The Transport Layer implements error checking end-to-end between the sender and receiver
The most popular transport protocol is the Transmission Control Protocol (TCP), which pro-
vides sequence numbers, error checking and retransmission of data that is received with er-
rors or not received at all.

192
The TCP software at the sender puts a sequence number and error check on the segment,
and the TCP software at the receiver normally returns a message to the sender acknowledging
successful receipt. If the sender’s TCP software does not receive this acknowledgment, it au-
tomatically retransmits the segment. The result is 100% error-free communication.
This is for file transfers, including email messages and web pages.
For live, streaming communications, like Voice over IP and video over IP, there is no time to
perform retransmission of bad data, so a different transport protocol, the User Datagram Proto-
col (UDP) is used instead.
UDP implements error checking, but not retransmission. Instead the receiver might interpolate
the missing data – fill in the gaps – using prior and subsequent data values to guess what the
missing one was.
13.7.2 Port Numbers
Another important function of the transport layer is to identify the application that is sending the
message and the application it is intended for on the far end.

Figure 134. The port number identifies the sending and receiving app
There is usually more than one application using an Internet connection on a computer or a
phone; for example, email and browser both running.
When a packet arrives at the computer, how does the computer know whether this packet is
for the email application or for the browser?
Every application is given a number called a port number. The first two bytes of the Layer 4
header are a field where the port number of the source application is populated, and the port
number of the destination application is populated in the next two bytes of the header.
This information at the beginning of the layer 4 header is used by the far end computer to de-
termine where to direct the data – which application this data is for – on the far-end computer.
The near- and far-end computers are called terminals, endpoints or hosts.
The IP address of the host concatenated with the port number of the application is called a
socket in UNIX and IP. It is called the transport service in the OSI model.
This allows segments of messages to be moved reliably from a particular application on one
host to a particular application on a different host.

13.8 Session Layer: POP, SIP, HTTP

The remaining layers are referred to as the upper layers.
The first stop on the upper layers is Layer 5, the session layer.
Once we can get the data to the destination, the next question would be, are we allowed to
send data to that computer? This is the function of the session layer.
The session layer manages communications sessions between applications, including initia-
tion, maintenance, sometimes restoration and certainly termination of information transfer ses-
sions.
13.8.1 Password Authentication

193
Establishing a session is often implemented by “logging on” to a remote system with a user-
name and password. An agreement on how the password chosen by the user is transmitted to
the far-end computer during account creation.
One can imagine the complexity when it is desired that the password be encrypted before
transmission, so that it can not be intercepted and re-used... how to transmit the decryption
key for the password without encrypting it?
13.8.2 Authentication Servers
Another area of development in session establishment is authentication servers.
Without an authentication server, there are two basic choices for remembering user names
and passwords on servers:
1) Use different usernames and passwords for every server you access. The question is, how
does one remember all these user names and passwords? Perhaps recording them in a file
called “user names and passwords.doc” in your My Documents folder? That does not sound
very secure!
2) Use the same username and password on every server. This exposes you to a serious se-
curity risk: that your username and password will be stolen from one of those servers by an in-
truder or a technician, then re-used to log in to your accounts on other servers.
An authentication service, like Google Accounts or Log in with Facebook, allows you to only
have one username and password, which allows you to access many services. You log in the
authentication server and it provides credentials to the server you are logging in to, without re-
vealing the username and password to the third party.
If a particular service, like the control panel for your web-based email wants extra protection, it
will ask you to log in again – but the username and password you type in are not validated on
the email server, they are passed to the authentication server for verification.
This way, you don’t have to store a user name and password on every server. Just one.
Google, Facebook and many other companies implement forms of authentication services.
13.8.3 Cookies
After you log on to a server, it would be nice if it remembered what you were doing last time
you logged on… restore your previous session. A method for session restoration used on the
Web is cookies.
When your browser uses HTTP to request a file from a web server, the web server replies with
the file – but first an instruction to your browser to store a cookie for the server’s domain using
the Set Cookie instruction.
The cookie is one or more name-value pairs and the server’s domain name in plain text, saved
in a small file in a folder on your hard drive.
These name-value pairs could be your username and password for the server, to be used to
log you on transparently later on. In this case, the cookie might be userid=yaright; pass-
word=fuggedaboudit; domain=forgetit.com.
Every subsequent time you request a file from that domain, your browser automatically sup-
plies the name-value pair as part of the HTTP file transfer request.
The example above has the problem of storing your password in a plain text file on your com-
puter, and giving it out to pretty much anything that asks for it.
Additionally, since the session information is stored on the client, the designer of the system
would have to account for obvious issues for users with more than one client computer.
To avoid these problems, Google sets a cookie that is one name-value pair, an encrypted code
that identifies the user, and the cookie information for their applications is stored on their
servers.
A problem is privacy. One well-known Internet web page banner advertising company was
caught defining their cookie in such a way that every client computer returned ALL of its cook-
ies to them.
They were accused of using this trick to – unethically – gather data for data-mining to deter-
mine where you had surfed.
They were literally stealing cookies from children (!).

194
13.8.4 Client-Server Sessions
An example of a standard session layer protocol is the Post Office Protocol (POP), an agree-
ment on how your computer logs on to a mail server to check for new e-mail messages then
downloads them.

Figure 135. Client-Server sessions are usually established by the client logging on to the server with a username
and password.
When setting up a POP-type email account, such as in Outlook, it’s necessary to start the
client software, then configure your user name and password and the name of the POP server.
Then, when you click “send and receive”, your POP client attempts to log on to that POP
server using that user name and password.
If it is successful the server indicates how many messages there are and then it uses the file
transfer protocol to download the email messages –which are data files – one at a time, from
the server to the client.
If the transfer is interrupted in the middle of a message, the next time the POP client runs, the
transfer resumes from the beginning of the message – so POP implements session state and
session restoration as well.
One could argue that the Hypertext Transfer Protocol (HTTP) is a session-layer protocol; this
is the protocol used to initiate a download from a web server by a browser.
The session only lasts for the transfer of all of the files referenced in one web page, and there
is no authentication – but an example of a session establishment nonetheless.
13.8.5 Peer-Peer Sessions
In standards-based VoIP systems, the Session Initiation Protocol (SIP) is used to establish
VoIP phone calls.

Figure 136. SIP uses proxy servers for peer-to-peer or “client-client” VoIP phone call session setup
In this case, the result will be a session between two telephones, where the two telephones
are peers, meaning they are equals. Instead of a client-server session, a VoIP phone call is a
client-client or peer-to-peer session.
To set up a phone call, it is necessary to communicate the IP address and communication port
numbers used by a phone to the other phone. The two phones subsequently transmit IP pack-
ets from one phone to the other.

195
SIP includes proxy servers that act as intermediaries between the caller and called party so
that the caller only finds out the called party’s IP address if the called party wants to take the
call.

13.9 Presentation Layer: ASCII, Encryption, Codecs

Layer 6, the presentation layer is very important: this layer is responsible for coding applica-
tion-layer messages into 1s and 0s.
Examples of presentation layer protocols include keystroke-coding protocols like ASCII and
Unicode, transformation protocols like MIME, data compression, encryption and codecs.
13.9.1 Character Coding
The American Standard Code for Information Interchange (ASCII) was primarily defined for
teletypes, i.e. printed English characters, which can easily be supported within a limit of 128
codes, and so specified codes that were 7 bits long.
As computers standardized on octets, more languages with different characters were required
to be supported, and parity checking was abandoned, sets of 8-bit codes were required. ASCII
formed the basis of two 8-bit code sets that are now in wide use: ISO-8859-1 and Windows-
1252.
Unicode and its Unicode Transformation Format (UTF) may end up becoming universal stan-
dard codes for character sets. Unicode defines a huge superset of characters, including Kanji
characters used in Asian languages, and methods of representing them, called transformation
formats. The most popular is UTF-8, typically using two bytes per character.
13.9.2 E-Mail Coding
Another example of a presentation layer protocol is MIME: The Multipurpose Internet Mail Ex-
tensions. This is a protocol for transforming or transcoding messages consisting of 8-bit bytes
(like an image, spreadsheet or computer program) into messages consisting of 6- or 7-bit
bytes that look like ASCII text for backwards-compatibility with email systems based on 1970s-
era UNIX computer technology that only supported 7-bit ASCII coding.
A MIME header is placed at the beginning of the transformed message so that the far-end can
apply the reverse transformation to re-create the message in its original 8-bit bytes.
This is transforming images into what look like giant plain text messages to email them, and is
no longer necessary, but is still almost universally implemented for backwards-compatibility
with old systems.
13.9.3 Codecs
Codecs for voice and video digitization are presentation layer protocols.
The G.711 standard from the ITU specifies voice coding at 64 kb/s – carried in IP packets or
DS0 channels. There are many other voice coding standards, most of which use fewer bits per
second.
The H.264 standard, specified in Part 10 of the MPEG-4 standard for video coding is used for
HD video.
All of these are protocols for coding messages into 1s and 0s.
13.9.4 Data Compression
Another example of a presentation layer protocol is data compression like WinZip, implement-
ing the ITU standard V.42bis.
This is again a method of representing information in 1s and 0s – just using fewer 1s and 0s to
represent our information.
13.9.5 Symmetric Encryption: Private Key
Encryption is a presentation layer protocol. There are two basic methods of encryption: sym-
metric key encryption and asymmetric key encryption.
Symmetric key encryption is also called private key or secret key encryption, as there is a sin-
gle key (a binary number) that both encrypts and decrypts the file, so the key is kept private.

196
Popular methods include the Advanced Encryption Standard (AES) using the Rijndael (“rhine-
doll”) algorithm.
13.9.6 Asymmetric Encryption: Public Key Encryption and Digital Signa-
tures
The other type of encryption is called asymmetric key encryption. What this means is that there
is a key pair; and what key A encrypts, key B can decrypt… and what key B encrypts, key A
can decrypt.
This is used in two different ways: for secure communications, called public key encryption,
and for authentication, called a digital signature.
For secure communications, a key pair is generated. One of the keys is made public, available
on a public key server, and the other key is kept private.
To communicate securely, the sender creates a message then uses the receiver’s public key to
encrypt it and transmits the encrypted message. The receiver uses the private key to decrypt
the message.
This avoids the problem inherent with symmetric or private key encryption for communications,
which requires transmitting the key, exposing it to potential eavesdropping. With public key en-
cryption, the decryption key is never transmitted.

Figure 137. Encryption is a presentation layer protocol

Algorithms developed by Rivest, Shamir, and Adelman (RSA) for public key encryption are
widely used. These are so good, if you use 1024-bit long keys, it’s estimated that it takes the
National Security Agency hours to read your e-mail.
They don’t like this idea, and had software that implements RSA declared a weapon so it is
controlled. Spooky. Servers implementing free public RSA must be located outside the USA.
On the other hand, people who want to kill you for religious reasons use it to hide their commu-
nications.
13.9.7 Example of Separability of Layers
Encryption is a very good illustration of the independence and separability of layers in the OSI
model.

197
Higher and lower layers know nothing of the encryption process. Higher layers have simply se-
lected a secure communications option, and know nothing of the details of how this is accom-
plished.
They pass messages to the presentation layer, which performs the encryption on the transmit-
ting side, then decryption on the receiving side before passing the message up to the applica-
tion layer at the receiver.
Lower layers know nothing of the encryption process – they are just tasked with moving 1s and
0s just like any other data.
13.9.8 Example of Peer Protocol
Encryption is also a very good illustration of the idea of peer protocols: having the same plan
on both systems at each layer.
If the sender uses one protocol for encryption and the receiver is using a different protocol,
there is not going to be any communication ...

13.10 Application Layer: SMTP, HTML, English …

The application layer, Layer 7, is the highest layer in the protocol stack, and includes the speci-
fication of the messages to be exchanged and often, Human-Machine Interfaces (HMIs).
The application layer transmits and receives messages. The application layer protocol is an
agreement on the format of the message and how it is going to be exchanged.
13.10.1 Email
An easy example of an application layer protocol is e-mail, and an easy example of that is the
Simple Mail Transfer Protocol (SMTP), a standard for exchanging email between computers.
SMTP specifies the three upper layers; it describes the Layer 5 functions of how mail transfer
sessions are initiated and terminated, how pieces of the file will be sent and acknowledged,
and for coding Layer 6 and message format Layer 7 references RFC 5322 “Internet Message
Format”.
RFC 5322 specifies US-ASCII coding for Presentation Layer 6, and finally, to get to the part of
interest here, the specification of the structure and format of a mail message, which is Layer 7:
RFC 5322 Internet Message Format tells you a message is made of lines. It tells you a line is a
sequence of characters coded into US-ASCII that ends with 13 (Carriage Return) 10 (Line
Feed). It tells you how the lines are grouped: header and body, separated by a blank line.

198
 Figure 138. SMTP Header and HMI
RFC 5322 defines the structure and content of the header. It tells you the format of a header
line is field-name:field-body. It tells you what the header field-names are. It tells you what the
allowed field-body values are. It tells you how to format the time, and the allowed values. It
tells you how to format an address.
It doesn’t tell you what message to write in the body.
13.10.2 More Application Layer Examples
English is an application layer protocol: its syntax rules define the format of messages and its
vocabulary is the allowed content of messages.
HTML is also an application layer protocol: it specifies the structure, syntax and vocabulary of
messages colloquially referred to as web pages.
File transfers could be considered as being in the application layer, though some might argue
that file transfers and file systems are actually all three of the upper layers... but it’s not a very
interesting argument. The File Transfer Protocol (FTP) is an example of this type of protocol.
Remote operations: remote monitoring and control of devices from a central station are a class
of application protocols, and represent a growing market segment, especially in the WAN man-
agement arena. An example is the Simple Network Management Protocol (SNMP).
This is also a messaging protocol, allowing the transfer of status inquiry and response mes-
sages and alarm messages between network elements like routers and a central monitoring
station running software like HP Openview.

13.11 Protocol Stacks

It is necessary to choose particular protocols to implement each layer. The protocols are usu-
ally drawn one on top of another, so the collection of protocols to perform all necessary func-

199
tions is called a protocol stack.
The same protocol is required at each layer. This is called a peer protocol.
The peers communicate, even if their communications are encapsulated inside other protocols’
data units to be carried to the other system.
Figure 139 provides a visual summary of the material discussed in the previous pages, and is
used to illustrate how information travels down through the protocol stack on the left, through
the network equipment in the center, and back up the protocol stack on the right.
13.11.1 Example: Web Surfing
The protocol stack when surfing the web is: application-layer messages formatted following
HTML, coded into 1s and 0s using ASCII at the presentation layer, retrieved from a server us-
ing HTTP at the session layer, communicated reliably between server and browser using TCP
at the transport layer, in network layer IP packets, in link layer MAC frames, on a LAN cable for
the last three feet.
13.11.2 Voice over IP
The protocol stack for a VoIP telephone call is: application-layer messages formatted using
English, coded by the presentation layer into 1s and 0s using the G.711 codec, a session set
up between the telephones using SIP, communicated using best efforts using UDP at the
transport layer, in network layer IP packets, in link layer MAC frames, on a LAN cables for the
last three feet.

Figure 139. Protocol stacks and peer protocols

13.12 Protocol Stack in Operation: Russian Dolls

To actually communicate between systems, protocols covering all seven layers, stacked on top
of each other, have to be implemented on both the sending system and the receiving system...
and it must be the same protocols at each layer implemented on each system.

200
Figure 140. Each layer’s output, called a Protocol Data Unit (PDU), is encapsulated inside the next layer’s PDU,
like Russian Matryoshka dolls.

13.12.1 Communications Flow

Communications begins with a person having information - a thought - to communicate, start-
ing at the top of Figure 140. To communicate this information to another person, the user cre-
ates a message and enters it into the system via software implementing the application layer
protocol.
The application layer protocol encapsulates the message with application layer control infor-
mation such as which person the message is to, then hands the result, called the application
layer Protocol Data Unit (PDU) to the next layer down, the presentation layer.
The presentation layer will code the message into 1s and 0s, and add a header with control in-
formation such as an indication of which coding scheme has been used, and pass the result to
the session layer.

201
The session layer might put some audit information on the front, like which client it came from
and any session authentication information that’s necessary, then give it to the transport layer.
The transport layer is responsible for identifying the source and destination applications, and
for end-to-end error checking, so it will take what it gets from the session layer, and put the
source port number and destination port number, a sequence number and error check on it,
and give this transport layer protocol data unit to the network layer to transmit to the far-end
host.
The network layer will take the incoming transport layer PDU and put that into a network layer
PDU – called a packet – with the network address of the final destination on the front of the
packet.
The packet goes into the data link layer PDU – called a frame – with the MAC address of the
destination on this particular circuit on the front of the frame, for transmission via a physical
port.
The frame is then transmitted one bit at a time over the physical layer: one bit at a time over
the LAN cable, over an airlink, over a fiber.
The physical layer on the next system receives the bits and passes them up to the data link
layer protocol software, which performs an error check on the frame, looks at the MAC address
on the frame, and compares it to the MAC address hard-coded into its LAN interface, and if
they are a match, indicating this is the desired receiver, it extracts the payload from the frame
(which is a packet) and passes it up to the network layer.
The network layer software will look at the address on the packet and use that as the basis of
making a route decision. If it is going to route the packet somewhere, the way it implements
the route decision is to take the packet and put it back in a frame and change the destination
MAC address, (because now it’s going to a different destination on a different broadcast do-
main), recalculate the frame check sequence and then transmit it out on a different circuit or
different broadcast domain.
Eventually the packet will arrive at the far-end network layer software, which will see that the
destination IP address on the packet is the same as its own IP address, so that will extract the
data from the packet and give it to the transport layer on the far-end computer.
The transport layer will check the error check that its peer (on the originating computer) put on
the information, and if it fails the error check, discards the received segment.
If it passes, then the transport layer extracts the payload from the transport protocol data unit,
and passes this to the correct software application on the far-end computer indicated by the
destination port number in the layer 4 header.
The received codes are passed to the presentation layer on the far end, which will decode
what it receives and pass the result to the application layer, which will recreate the original
message and display it to the person at the far end via a Human-Machine Interface.
13.12.2 Segmentation at Each Layer
At each stage, the protocol might segment the data unit it receives from a higher layer and
transmit a number of smaller data units to its peer protocol on the opposite system, which re-
assembles them back into the original size to hand back up to the higher layer.
13.12.3 Nested Headers: Matryoshka dolls
By passing segments of data to a lower layer, which performs its function, adds a header and
passes the result to a yet lower layer, the protocol data units of each layer end up being nested
one inside another inside another like Russian Babushka dolls, properly called Matryoshka
dolls.
The innermost, smallest doll is a segment of the application-layer message.
At the bottom of the protocol stack, all of the headers added by the layers are in place, one af-
ter another as illustrated in Figure 140. The result is a lot of overhead – all those headers – but
also the ability to make the best choice for protocols at many different levels independently.

13.13 Standards Organizations

Since standards are such a good idea, we write lots of them!

202
Many different organizations with different perspectives and agendas have become involved.
Out of the resulting myriad choices, particular protocols become standards in the actual sense
of the word through popularity - the choices most popular in the market, sometimes referred to
as the “thundering herd”.
13.13.1 ISO
The International Organization for Standardization (ISO) defined the OSI Reference Model that
we examined in detail.
It’s important to keep in mind that the OSI Reference Model does not tell us how to do all of
these functions - it tells us what we have to do, and gives us a structured way of discussing
what we have to do so we can discuss separate issues separately, and not get things jumbled
up.
In addition to the reference model, ISO does publish particular protocols, such as the data link
protocol HDLC. These OSI-published protocols enjoy varying degrees of actual industry use:
slim to none. This is a side issue to the OSI Reference Model and the concept of open sys-
tems.
13.13.2 DOD and IETF
There are a number of standards organizations for communications. The US Department of
Defense (DOD) published specifications for a suite of protocols including the Internet Protocol
(IP) and Transmission Control Protocol (TCP).
These are now maintained by the Internet Society, through the Internet Advisory Board (IAB)
and the clique called the Internet Engineering Task Force (IETF) that publishes Internet stan-
dards called Request for Comments (RFCs).
13.13.3 ITU and Bellcore
Lest we forget! The telephone network is the world’s biggest network; it made the Internet look
tiny in comparison in the previous millennium. Eventually, the Internet and the telephone net-
work will be the same thing. In the meantime, there are standards specific to the telephone
network.
The Comité Consultatif International de Téléphone et de Télégraphe (CCITT), now officially
called the Telecommunications Standards Sector of the International Telecommunications
Union (ITU-T) is an international treaty organization, with strong European telephone company
influences. This organization publishes many standards, including the V. series of modem
standards, the X. series of data network access standards, and the I. series of digital tele-
phone network standards.
The former Bellcore (Bell Communications Research), now called Telcordia and originally part
of Bell Labs, publishes standards for the North American public telephone network.
13.13.4 TIA and IEEE
Industry organizations include the Telecommunications Industries Association (TIA), which is a
subgroup of the Electronic Industries Association (EIA), which publishes the old RS-232 stan-
dard for modem cables connections, and the newer TIA-568 standard for building wiring.
The Institute of Electrical and Electronic Engineers (IEEE) publishes standards for how to build
LANs on TIA-568 cabling, the 802 series of LAN standards. “Ethernet” is 802.3 and 802.2 to-
gether.
13.13.5 ANSI
In addition to these organizations with specific areas of interest, there are national organiza-
tions such as the American National Standards Institute (ANSI) that try to coordinate standards
at a national level.
Sometimes, in an attempt to coordinate two similar but not identical competing standards from
different groups, and come up with a unified standard, ANSI ends up creating a third standard
that then competes with the two existing “standards”.

203
14
Ethernet, LANs and VLANs
LANs were commercialized in 1979 by DEC (now HP), Xerox and Intel with a product called
Ethernet. LANs became popular for connecting computers, printers and file servers in-building
during the 1980s.
One of the original requirements for a LAN was to connect computers in an office to a shared
laser printer.
This requirement generalized to connectivity between devices for the sharing of all kinds of re-
sources, including hardware resources: hard disks and surveillance cameras; information re-
sources: centralized databases; software resources: network address configuration programs,
communication resources: WAN circuits, amongst countless other examples.
Ethernet, and its many subsequent updates in the IEEE 802 standards, is now almost univer-
sally used on fiber, twisted-pair copper, and over the air to implement the Layer 2 links that
make up the telecom network.
In doing so, Ethernet achieved one of the long-sought goals in the telecommunications busi-
ness: the same technology used in the network core, on the network access circuit, and at the
customer premise.

14.1 LAN Basics

14.1.1 Bus Topology
The original design for a LAN used a bus topology and coaxial cable. Topology is the way the
system looks viewed from the top, its layout.
Bus comes from electrical power distribution systems: a power distribution bus is a thick bar
conducting electricity, for example, the bar in a circuit breaker panel to which all of the circuit
breakers are attached.
In LAN terminology, this term was borrowed to mean a cable running down an office building
hallway connecting a floor full of PCs. As illustrated in the graphic, all of the PCs, or worksta-
tions, terminals, devices or simply stations were connected to the bus.
14.1.2 Broadcast Domain
The bus implements a multi-drop circuit: anything any station transmits is received by all of the
other stations.
For this reason, this group of stations is said to form a broadcast domain: any station has the
possibility of communicating directly with any other station in the broadcast domain without the
need of other equipment or protocols.
This has obvious implications for both security and performance.
The bus has been replaced with an Ethernet switch, also called a LAN switch and Layer 2
switch. Ethernet switches are covered in detail in Section 14.4.

204
 Figure 141. Broadcast Domain

14.1.3 Balanced Configuration

LANs are balanced configurations with combined stations, which means there are no con-
trollers - all stations are equal on the LAN.
Multidrop distribution networks like cable TV are unbalanced configurations, with controllers
and controlees.
14.1.4 Collision Domain
Figure 141 also illustrates a collision domain, meaning there is the possibility that the transmis-
sions of two stations will collide – two stations transmitting on the circuit at the same time.
The electricity or light that each station would put on the circuit adds together, so that other sta-
tions would not be able to reliably interpret what they detected, and nothing would be commu-
nicated.
14.1.5 CSMA-CD Access Control
Since anything any station transmits is received by all of the other stations, only one can trans-
mit at a time.
An access control mechanism, a protocol for determining if a particular station may begin
transmitting is required.
The designers of Ethernet chose Carrier-Sensing Multiple Access with Collision Detection
(CSMA-CD), a contention-based access control protocol, where stations contend for the use of
the physical connection on a first-come, first-served basis.
In essence, the CSMA-CD protocol is:
1) Listen for a carrier signal, to hear if any other station is transmitting.
2) If you hear another station transmitting, wait.
3) When there is silence, transmit a frame containing the intended receiver’s MAC address,
your MAC address, a block of data and error checking.
4) Listen while you transmit.
5) If you hear something different happening, this means another station started transmitting,
so stop transmitting and return to Step 1.
14.1.6 MAC Address
As illustrated in Figure 141, anything any station transmits is received by all stations: this is a
multidrop circuit, a broadcast domain.
Since all stations receive everything, when transmitting data, a station has to indicate the ad-
dress of the desired receiver on the link: an indication of which station on this link should react

205
to the data; in other words, indicating for which station the data is intended.
The software that implements the addressing is described in the standards documents as be-
ing part of the Media Access Control sublayer, and so the link addresses are called MAC ad-
dresses.
Every LAN interface is given a hard-coded 48-bit MAC address by its manufacturer. The first
three bytes of the address identify the manufacturer and the last three bytes are a serial num-
ber.
14.1.7 Communication of MAC Frames
The mechanism for communicating to another station on a LAN is to transmit a frame with the
MAC address of the intended receiver in this broadcast domain in the destination address field
of the frame.
The frame is transmitted, all stations receive it, perform the CRC error check (which protects
the address), then compare the destination address on the frame to their own MAC address.
If the MAC address on the frame is not the same as a station’s MAC address, it is supposed to
ignore the frame.
If they are the same, then the station knows it is the intended receiver and processes the
frame, extracting the payload from the frame and passing it up to the next higher-level soft-
ware.

14.2 Ethernet and 802 Standards

The LAN was invented at the Xerox Palo Alto Research Center (PARC) in Silicon Valley in Cal-
ifornia – along with the mouse and the windows graphical user interface.
And people say Xerox never does anything original!
LAN technology was commercialized in 1979 by a consortium of three companies: Digital
Equipment Corporation (DEC), now part of HP, Xerox and Intel.
It was branded Ethernet, presumably by the marketing department, evoking the idea of com-
municating via the Ether, the fabric of space itself.
However, instead of the Ether, thick yellow coaxial cables were used.
14.2.1 IEEE 802 Standards
The Institute of Electrical and Electronic Engineers (IEEE) subsequently formed standards
group 802 in February of 1980, and developed a number of standards for LANs.
Standards 802 and 802.1 described the overall architecture and MAC addressing scheme.
802.2 described Logical Link Control, the protocol for exchanging frames between stations.
802.3 described physical coax cabling, signals and framing in a way that was almost identical
to Ethernet.
14.2.2 Ethernet vs. 802.3
Other than a difference in the address format, the two were identical. Both Ethernet and 802.3
used the bus topology, provided 10 Mb/s communications and used the CSMA-CD access
control protocol.
Eventually, the market adopted the “open” 802.3 standard and Ethernet failed as a commercial
product.
We now use the term Ethernet to refer to the 802.3 standard, which copied the design of Eth-
ernet then stole its market. The people who invented Ethernet must spin in their graves each
time this happens.
14.2.3 Token Ring
The IEEE also developed standards for other LAN architectures, notably 802.5 which was
IBM’s Token Ring product.
This involved passing frames neighbor-to-neighbor around a ring, and passing a token or per-
mission to originate a new frame around the ring. Token Ring is now obsolete legacy technol-
ogy.

206
14.2.4 Baseband LAN
Ethernet is a baseband system: a station uses the entire capacity of the bus when transmitting.
The CSMA-CD access control mechanism is used to decide if a station can transmit on the
bus at any given time.
IBM attempted to commercialize an in-building communication system very much like modems
on Cable TV, where there was a wide frequency bandwidth and multiple channels. IBM called
this a broadband LAN.
IBM’s product no longer exists – there is no such thing as a broadband LAN, and all LAN tech-
nologies are “baseband” LANs, hence the designation BASE in the 802.3 standard.
14.2.5 10BASE-5
In the initial design, to connect a station, a transceiver was physically attached to the coaxial
cable bus and a short tail circuit run from the transceiver to the station’s Ethernet card.
Stations communicated by broadcasting frames with the MAC address of the source and de-
sired destination at the beginning of the frame. Anything a station transmits is received by all
other stations.
The original design is referred to as 10BASE-5, since it provides 10 Mb/s, implements a single
baseband channel on the bus, and the maximum length of a cable segment is 500 m.

Figure 142. Connecting to the Bus

14.2.6 10BASE-2
The first improvement to this design was to reduce the cost of the bus cable and transceiver. A
thinner coaxial cable was specified, and the transceiver function moved to the adapter card in-
side the PC instead of being a separate device.
This was referred to as a Thinwire Ethernet or 10BASE-2, because the maximum cable seg-
ment length is 185 m with the thinner cable.
It is sometimes still used to run Ethernet over existing in-building coaxial cable TV wiring.
14.2.7 10BASE-T
The next improvement, 10BASE-T, implemented Ethernet using point-to-point twisted pair ca-
bles connected to a passive hub to replace the bus. The maximum length of twisted pair cable
is 100 m.
14.2.8 100BASE-T
100BASE-T is 10BASE-T ten times faster, on Category 5 unshielded twisted pair, employing
two of the pairs for data with a 3-volt, 3-level Manchester line code. The other two pairs are ei-
ther unused, or sometimes used to deliver power to terminal devices.
Cable categories are covered in the next section.

207
14.2.9 1000BASE-T
1000BASE-T is Gigabit Ethernet, implementing two-way simultaneous transmission over all
four pairs of a LAN cable in parallel.
The bit stream is divided into four and 250 Mb/s is transmitted over each pair, using bandwidth
of approximately 100 MHz on each of the pairs.
For those who like details: the line coding is 5-level Pulse Amplitude Modulation (PAM) with 4-
dimensional 8-state Trellis Forward Error Correction encoding, pulse shaping and signal equal-
ization, Near-End Cross Talk (NEXT) cancellation and echo cancellation using digital signal
processing. Available in quantity 1 for less than $10!
In theory, the next step is 10 Gb/s on Category 6 copper cables.
14.2.10 Optical Ethernet
Optical Ethernet, that is, signaling MAC frames point-to-point by flashing a light on and off on a
fiber, begins with 1000BASE-SX and 1000BASE-LX Gigabit Ethernet over fiber, the SX being
short wavelength (850 nm) and the LX being long wavelength (1550 nm) with a specified range
of 5 km.
Optical Ethernet is covered in Section 8.5.
At time of press, the high end for Optical Ethernet is 100GBASE-ER4, 100 Gb/s Extended
Range, signaling the bits on four wavelengths in parallel with a range of up to 40 km. This will
increase in the future.

14.3 LAN Cables and Categories

LANs for the most part run over cables inside buildings.
The term cable is often used to mean “bundles of wires”. Connectors or terminations may also
be included as a package.
14.3.1 Unshielded Twisted Pair (UTP)
Historically, copper wires have been used for two-wire telephone access circuits, called loops.
Pairs of copper wires are also used for LAN cables.
Copper is used because it is a good conductor of electricity, inexpensive, pliable, corrosion-re-
sistant, and easy to extrude into long, thin wires.
The two wires are twisted together to reduce pickup of noise, and so are often referred to as a
twisted pair.
The wire may be solid or braided, the latter being more expensive to manufacture but better re-
sistant to breakage.
14.3.2 Shielding
A shield may be placed around individual pairs, and/or around the entire bundle of wires in a
cable.
The shield is a metal foil or mesh that prevents both the ingress and egress of electro-mag-
netic energy, which causes interference on copper wires.
Unshielded Twisted Pair (UTP) is often used, as adding shielding to reduce noise also reduces
the frequency response.
14.3.3 TIA-568 LAN Cable Categories
The most widely-followed standard for LAN cables is TIA-568, published by the Electronic In-
dustries Association and its Telecommunications Industry Association sub-group.
This standard defines categories of twisted-pair cabling that support different line speeds.
Telecommunications Systems Bulletin TSB-67 adds the requirements and methods for field
testing installed cable systems. Taken together, these are the authority how to design and in-
stall a structured cabling system.
• TIA-568 Category 1 cable is existing telephone cabling, also called Rusty Twisted Pair
(RTP).

208
 • Category 2 cable was 25-pair multiconductor cables for old key telephone systems that had
buttons to press to access different lines.
• Category 3 cable was for 10 Mb/s Ethernet on twisted pair, 10BASE-T.
• Category 4 cable was specified for 16 Mb/s token ring.
• Category 5 cabling was for The Future at up to 1000 Mb/s.
Categories 1 through 5 are no longer installed.

Figure 143. Category 5e LAN cable. Bulk cable is terminated on keystone connectors that snap into the back of a
cover plate. Patch cables, illustrated on the left, can be made by crimping RJ-45 connectors on bulk cable..
Category 5 (Cat 5) cable was supposed to support Gigabit Ethernet, but in practice turned out
to be missing the specification of some required transmission characteristics.
Enhanced Category 5 (Cat 5e) was subsequently specified to guarantee the operation of Giga-
bit Ethernet on twisted pair, 1000BASE-T.
Whether a cable can be certified as conforming to a standard is often dependent on the con-
sistency and placement of twists during manufacturing.
Category 6 cable is specified to support 10 Gb/s on twisted pair.
At 1 Gb/s, it becomes necessary to specify the frequency bandwidth supported on the twisted
pair, along with all of the other transmission characteristics, to enable signaling at these line
speeds.
In theory, Category 7 supports 100 Gb/s on twisted pair. This is in the same league as current
mainstream fiber-optic transmission systems, so one could probably expect it will be a while
before there is any significant deployment of Cat 7 copper wires.
14.3.4 TIA-568A vs. TIA-568B
There are two specifications for which wires in the cable go to which pins on the connectors:
TIA-568A and TIA-568B. There is no difference between the two in terms of performance – but
it is necessary to pick one of the two configurations and use it consistently on every jack, every
patch panel, every patch cord and every connector.
Figure 143 illustrates TIA-568B, which is the most popular choice. Holding a male Category 5e
connector in front of you with the retainer clip facing you and the metal contacts on the top, pin
1 is on the left. The wires are color-coded in a standard way, using white, orange, green, blue
and brown.
The TIA-568B connections are:
Pin 1 – white/orange
Pin 2 – orange
Pin 3 – white/green
Pin 4 – blue
Pin 5 – white / blue
Pin 6 – green

209
 Pin 7 – white / brown
Pin 8 – brown
This pinout must be used consistently, as the design of both the connector and the cable and
their performance measured in transmission characteristics such as crosstalk, insertion loss,
echo and other metrics are based on particular signals being on particular wires.
14.3.5 Maximum Cable Length and Cabling Architecture
All categories specify cables with four pairs (eight wires) and a maximum length of 100 meters.
This means the maximum run length of the cables – including runs through risers, poles, con-
duits – is 100 m (330 feet).
To be conservative, devices would be connected to a switch located in a wiring closet within a
radius of perhaps 200 feet.
These wiring closet switches could be connected to centralized Ethernet switches on each
floor and/or connected to a router in the communications room, possibly using fiber.
14.3.6 Difference Between Categories
The difference between the categories rests in guaranteed transmission characteristics of the
cable, including specifications for Near-End Crosstalk (NEXT), Attenuation to Crosstalk Ratio
(ACR), supported frequency bandwidth, all of which affect the maximum possible information
transfer rate, and hence what kind of devices can be successfully attached to each end of the
cable.
One of the main factors in getting a cable certified to meet the TIA-568 category is quality con-
trol, particularly in the consistency of the twisting and placement of the pairs.
Two pairs will be twisted at a particular number of twists per inch, but offset by half a period to
minimize crosstalk between the pairs. The other two pairs will be twisted at a different rate that
is not a multiple of the other, and similarly with the twists exactly not lined up.
How well and how consistently this is accomplished during the manufacturing process deter-
mines how successful the manufacturer will be in having the cable certified as meeting the
standard.
14.3.7 Which Category To Use
When determining which category of cable to use, life cycle and labor cost are determining fac-
tors.
For a patch cable connecting a DSL or Cable Modem to a device inside a residence, where we
have an expectation that the line speed will not exceed 100 Mb/s in the foreseeable future,
then Cat 5 patch cables may be used.
For an extra ten cents, a Cat 5e patch cable would allow the continued use of the cable were
the line speed to increase above 100 Mb/s, as it inevitably will at some time in the future.
Since the labor cost is usually far greater than the cable, it is strongly recommended to install
cabling inside walls with capacity greater than immediate needs, and twice as many cables as
what the conventional wisdom dictates.
Two Category 6 cables to each work area would be the Cadillac solution.
Two Category 5e cables to each work area would be well positioned for the future.
One Category 5 cable to each work area would probably be viewed as a mistake ten years
down the road.

14.4 LAN Switches: Layer 2 Switches

Since the invention of Ethernet and its standardization, the technology has evolved and im-
proved in several different ways. One main improvement is the Ethernet switch, also called a
LAN switch or Layer 2 (L2) switch.

210
 Figure 144. LAN Switch
This device replaces bus cables and hubs, providing dramatic improvements in performance
plus the possibility of implementing improved traffic management and security through Virtual
LAN (VLAN) technology.
14.4.1 Hardware
In concrete terms, a Layer 2 switch is a small dedicated-purpose computer with anywhere from
two to hundreds of LAN hardware ports, an internal bus, memory and software performing the
switching function and possibly additional VLAN-related functions.
Each hardware port is an Ethernet jack, and should support 1000, 100 and 10 Mb/s full-duplex.
14.4.2 Purpose and Operation
The essential function of a Layer 2 switch is to receive frames from devices, examine the desti-
nation MAC address on the frame, determine which hardware port(s) this corresponds to, and
relay the frame to the computer connected to the indicated hardware port(s).
To determine the MAC address of the computer connected to a particular hardware port, the
processor reads the sender MAC address on frames and stores this information in what might
be called a MAC table.
14.4.3 Buffers
Since there is typically only one processor and one internal bus connecting these hardware
ports inside the switch, small amounts of memory called buffers are provided for each port to
allow stations to send frames simultaneously.
The frames are stored in the buffer then relayed to the appropriate port(s) by the switch’s pro-
cessor, normally on a first-come, first-served basis, or in a prioritized order in the case of an
expensive switch implementing prioritization protocols.
14.4.4 Frame Forwarding
In normal operation, the processor relays a frame from one port to one other, and it does this in
a lightning-fast manner, since it only reads the destination MAC address, does a lookup in the
MAC table then forwards the frame to the indicated port. It does not receive the whole frame
and perform an error check; the destination computer performs error recovery.
14.4.5 Broadcast Domain Defined by Switch
In exceptional circumstances, the Layer 2 switch broadcasts the frame to all the hardware
ports.
This happens when there is no entry in the MAC table for the destination MAC address. It also
happens when the content of the destination MAC address field explicitly instructs the switch to

211
broadcast the frame, when the sending computer is running the Address Resolution Protocol,
for example, attempting to discover the MAC address of the computer that owns a particular IP
address.
Because there is the possibility the switch will send a copy of the frame to all of the hardware
ports, all of the computers connected to a Layer 2 switch are in a broadcast domain: any sta-
tion has the possibility of communicating directly with any other in the broadcast domain with-
out the need of other equipment or protocols.

14.5 VLANs
14.5.1 Broadcast Domains Defined in Software
VLANs are essentially a software trick, implemented by the switch, to define broadcast do-
mains in software, for the purpose of traffic management.
A basic LAN switch does not implement VLANs. All of the devices physically connected to the
basic LAN switch form a broadcast domain; there is no way of preventing one device from
communicating with another.
A more sophisticated switch supporting VLANs allows an administrator to identify specific
hardware ports as belonging to a particular VLAN group, identified by a 12-bit number.
In the simple example illustrated in Figure 145, the ports for the IP phones producing voice
packets are defined to be in VLAN 1 and the ports for the desktop computers are defined to be
in VLAN 2.
Once this is set up, the processor will only forward frames between hardware ports that are in
the same VLAN, and if it is necessary to make a copy of a frame to send to “all” ports, a copy
is only sent to the ports in the same VLAN and not to any others.
14.5.2 Routing Between VLANs
In the example illustrated in Figure 145, the port labeled “uplink” is the connection leading to
the rest of the network, i.e. to a router. This port is defined by the administrator to belong to
both VLAN 1 and VLAN 2.
Communication from a device in VLAN 1 to a device in VLAN 2 can implemented by transmit-
ting a packet in a frame from the device in VLAN 1 to a router via the uplink port, whereupon
the router could transmit the packet in a frame back to VLAN 2.

Figure 145. VLANs implemented by a Layer 2 switch

The purpose of this architecture is to prevent direct communications between devices in differ-
ent VLANs, and allow communications only through an external router... where rules can be
entered by an administrator specifying if the communication between the VLANs is allowed or
denied.
14.5.3 Header Tag

212
To indicate to the device on the uplink port which VLAN a frame is originating from, an extra
Tag Header conforming to the 802.1Q standard is added to the frame immediately following
the address fields, and the VLAN ID is populated in the Tag Header.
Tagging the frame with the VLAN ID allows the definition of VLAN groups that span multiple
physical switches.
Additionally, the Tag Header includes the Tag Protocol Identifier identifying the frame as a
tagged frame, and following the 802.1p protocol, can optionally carry a three-bit number indi-
cating the priority of the frame for Quality of Service mechanisms.
14.5.4 Traffic Management and Network Security
VLANs are a powerful low-level tool for traffic management and network security. It allows the
grouping of devices into separate broadcast domains so that devices in one VLAN can not
communicate to devices in a different VLAN, a measure against attacks launched from in-
fected Windows computers against a VoIP system, for example.
It is also an essential tool used to separate customers of a carrier who are using a shared facil-
ity. By putting each customer’s hardware ports in a unique VLAN, traffic from different cus-
tomers will be interspersed on a shared circuit, but the customers can not communicate to
each other nor receive copies of other customers’ traffic.

213
15
IP Networks, Routers and Addresses
This chapter could equally be called “Layer 3”.
In this chapter, we cover networking, which is Layer 3 of the OSI model, including routers,
packets and network addresses. We’ll understand how the network is built by connecting cir-
cuits with routers, and trace the flow of a packet from end to end.
The standard method of formatting packets and assigning network addresses is of course IP,
formerly known as the Internet Protocol, so this chapter could also be called “All about IP”.

15.1 Definition of Network

The definition of a network includes the requirement to make route decisions: not broadcasting
the data to every station, but instead forwarding, switching or routing the data from one circuit
to the next to the next to eventually deliver it to a particular destination.
A packet network is constructed of point-to-point circuits connecting routers in different loca-
tions. The routers physically move packets from one circuit to a different circuit, a forwarding or
relay function.
Knowing which circuit to relay a packet to is the routing part of the story.
Packet networks incorporate two main ideas: packet switching and bandwidth on demand.
Packet switching means forwarding IP packets from one router to another, plus IP addresses,
routing tables and algorithms to determine which circuit to forward them to.
Bandwidth on demand means giving many devices access to a circuit, and giving each the
possibility of transmitting a packet, but not reserving capacity for any particular device. If a
device does not have anything to transmit, another can use the available capacity.
Since the term bandwidth is used to mean transmission capacity, this is called a capacity on
demand or bandwidth on demand strategy.

15.2 Simplest Example: Private Network

We begin with the simplest example of a network illustrated in Figure 146: three locations con-
nected in a ring with point-to-point circuits.
This is called a private network, since there are connections only between those three loca-
tions, and the circuits that implement the connections are not shared with others.

214
 Figure 146. Simplest Example of a Network
The inter-building or Wide Area Network (WAN) circuits could be implemented with point-to-
point fibers, point-to-point radio links or full-period “dedicated line” services like T1 from a car-
rier.
This is useful as the simplest framework for understanding how circuits, routers, routing, IP
packets and IP addresses, MAC addresses and MAC frames, copper and fiber work together
to implement a packet network.
In the next chapter, “MPLS and Carrier Networks”, the story is made more realistic – and more
complex – replacing the dedicated lines with packet-switched bandwidth-on-demand services
from a carrier.
15.2.1 Broadcast Domain at Each Location
In the example of Figure 146, at each location there are a number of terminals or devices –
VoIP telephones, desktops, servers, plus a router, all connected to a LAN switch.
All of the devices connected to the LAN switch in building A are in the same broadcast domain.
This allows the communication of MAC frames between the devices in building A and the
router in building A.
15.2.2 Edge Router at Each Location
The router in building A has three Ethernet jacks for terminating circuits. One of them is con-
nected to the LAN switch in building A. The other two are connected to circuits leading to the
routers in other buildings.
By internally moving packets from one jack to another, the router moves packets between
buildings.
In Windows, the building A router is called the default gateway by building A devices. In any
device’s routing table, the IP address of Router A is listed as the default route.
If a device wants to send a packet outside its broadcast domain, its only hope is to send the
packet on its default route – to that gateway – for onward forwarding.
That gateway, more generally called the edge router for building A, is the only device in build-
ing A that connects to other broadcast domains.
15.2.3 Subnet Assigned to Broadcast Domain
Each broadcast domain, i.e. each building, is assigned a unique range or block of IP ad-
dresses, called a subnet.

215
The terminals and the router in each broadcast domain are assigned an IP address within the
subnet for that broadcast domain.
15.2.4 Default Gateway
The terminals are informed of the IP address of the router in their subnet, which is their default
gateway to other broadcast domains.
Informing the devices of the IP address of their edge router allows the devices to communicate
packets to it for onward forwarding.
15.2.5 Subnet Mask

Figure 147. Subnet Mask and Addresses. This is much easier to understand in binary than decimal.
The devices are also informed what subnet they are in and how big it is. This is accomplished
via a number called the subnet mask that identifies which part of their address identifies the
subnet, and which part of their address identifies their machine within the subnet.
The beginning of the address is the subnet ID, and will be common to all machines in the sub-
net. The end of the address is the machine or host ID, so will be different for each machine in
the subnet.
The subnet mask identifies which bits are the subnet ID with 1s in the positions that are the
subnet ID, and 0s in the positions that are the host ID.
15.2.6 Packet Creation
To communicate VoIP from the telephone in building A to the telephone in building C, tele-
phone A first has to find out the IP address of telephone C, usually using the SIP protocol as
described in Section 3.4.
Once the conversation starts, telephone A creates IP packets addressed to telephone C con-
taining snippets of digitized speech.
15.2.7 Packet Transmission from the Source
To send a packet from telephone A, there are only two choices: send the packet directly to
telephone C, or if that is not possible, send the packet to its default gateway, router A, for on-
ward forwarding.
To determine which of these two possibilities to use, telephone A first determines if it can send
the packet directly to telephone C.

216
By definition, that would require telephone A and C to be in the same broadcast domain. Since
each broadcast domain has been assigned a subnet, by definition, that requires telephone A
and C in the same subnet.
Telephone A can determine the answer by applying the subnet mask using the logical AND op-
eration to its own address, and to the address of telephone C, then comparing the result.
If they are the same, the two telephones are in the same subnet, and thus in the same broad-
cast domain, and so telephone A can transmit the packet directly to telephone C.
In this example, the result will not be equal, allowing telephone A to determine that telephone
C is in a different subnet, which means in a different broadcast domain, and so by definition,
telephone A knows it can not communicate the packet directly to telephone C.
It must instead send the packet to the router in building A (which is in the same broadcast do-
main) for onward forwarding.
Once telephone A has decided the destination is router A, it transmits the packet to router A by
putting the packet in a MAC frame with destination MAC address that of router A, then repre-
senting the bits that make up the frame one at a time on the copper wire LAN cable plugged in
to the phone by putting electrical voltage pulses on the wires.
15.2.8 IP to MAC Address Resolution Protocol (ARP)
Since telephone A has been informed of router A’s IP address, it can determine router A’s MAC
address by asking the router what it is using the Address Resolution Protocol (ARP).
Telephone A transmits a packet addressed to router A in a frame with all 1s as the MAC ad-
dress, an instruction that a copy of the frame should be sent to all devices in the broadcast do-
main.
After router A replies to telephone A with its MAC address as the source address in the frame
header, telephone A can address frames to router A.
15.2.9 Packet Routing
Upon receiving the MAC frame and extracting the packet from it, router A will physically for-
ward or relay the packet from the LAN in building A to a circuit that can get to building C.
Determining where the packet should be relayed is the routing.
Networks are built with redundant connectivity for service availability reasons: more than one
way to get from A to C The router in building A is connected to two circuits that lead to building
C.
Router A must decide which circuit to forward the packet on.
Router A has a routing table, which has entries relating ranges of IP addresses (subnets) to
the IP address of a device that can forward a packet there, and the cost.
Cost is usually measured by number of hops, i.e. circuits to traverse.
The routing table is populated by entries manually typed in by a technician, by the routers com-
municating with each other in the background, or by a central control system in a Network Op-
erations Center.
In this case, the routing table will have two entries for subnet C, which contains telephone C:
1) All of the devices in building C are reachable by going to router C, and the cost is one
hop, and
2) All of the devices in building C are reachable by going to router B, and the cost is two
hops.
The router picks the least-cost route, and forwards the packet to router C.
15.2.10 Overbooking & Bandwidth on Demand
Beside the question of routing is a different discussion: performance.
In the example of Figure 148, each device has the possibility of transmitting packets to the
LAN switch then to the router at 1000 Mb/s, and onward to other buildings at 10 Mb/s… but
none of those bits per second are reserved for any particular device, either on the LAN or the
WAN.
Statistically speaking, most of the time, telephone A does not transmit anything. Occasionally,
it will transmit a packet in a frame to the LAN switch then router A over the LAN at 1000 Mb/s.

217
The router will relay the packet to a jack that has a 10 Mb/s dedicated line to another router,
and transmit it at 10 Mb/s to the other router. Occasionally, a different device will transmit a
packet to router A.

Figure 148. Overbooking LAN and WAN circuits

If many packets arrive at router A on the 1000 Mb/s LAN cable from the LAN switch, the router
may have to temporarily store or buffer the packets before being able to forward them on a
WAN circuit at 10 Mb/s.
The same problem exists (to a much lower degree) at the LAN switch, where there are four de-
vices with 1000 Mb/s connections sharing a single 1000 Mb/s connection to the router.
If the overload persists, the oldest packets in the buffer, still waiting to be transmitted, get over-
written by the newest incoming packets. When a packet is overwritten in a buffer, it disap-
pears. These would show up in any measurement of packet delivery rate or dropped packets.

15.3 Routers and Customer Edge

In this section, we take a closer look at the router for Building A of Figure 146 and how it is
configured to both implement the network and control network traffic.
15.3.1 Customer Edge Device
The router in Figure 149 is the gateway between the LAN and the WAN. It is connected to in-
building Layer 2 switches on one side, and to carrier circuits on the other side.

218
 Figure 149. Customer Edge Router
From the point of view of a carrier that might be providing the point-to-point links, it defines the
edge of the customer’s in-building network, and so is called the Customer Edge (CE) device by
carriers.
This device has also in the past been called the premise router, the customer premise router,
and is called the “default gateway” by Windows computers.
It can be implemented as a $20 stand-alone device or included in the same device that houses
a DSL or Cable modem or fiber termination for home or small office use. Industrial-strength
versions costing thousands of dollars are of course also available from companies like Cisco
for larger offices.
15.3.2 Router Connects Broadcast Domains
In the configuration illustrated, the CE router belongs to four broadcast domains: the two
VLAN-defined broadcast domains on the upper LAN switch, the hardware-defined broadcast
domain on the lower LAN switch, and the WAN circuit is a fourth broadcast domain.
Without a router, these four broadcast domains are like individual standalone circuits. The
router implements the network by implementing the possibility of communications between the
broadcast domains.
15.3.3 Routing
The router examines the destination address field in the Layer 3 header (Network Layer
header) on a packet, and uses this value along with information in its routing table to determine
where to forward the packet.
The routing table essentially lists ranges of addresses (subnets) against the address of a de-
vice that can relay a packet to any address in that subnet, and at what cost.
The “answer”, result of the route calculation is the address of the next hop, in other words, the
device to which the packet should be forwarded to get to the destination address.
The next hop address is resolved to a broadcast domain, then to a hardware interface, then
the packet is physically forwarded in a frame.
15.3.4 Denying Communications
In addition to implementing the network by implementing the possibility of forwarding packets
between broadcast domains, a router also acts as a point of control, denying communications.

219
This is part of basic network security practice. The objective is to compartmentalize the net-
work, allowing communications only between machines and/or applications when there is a le-
gitimate and desirable reason to do so.
In practice, this is implemented by denying all communications by default, then permitting com-
munications between specified machines and/or between specific applications.
15.3.5 Packet Filtering
Permitting communications to specific machines is implemented with rules based on source
and destination network addresses in the Layer 3 header, and is called packet forwarding.
When denied, it is called packet filtering.
15.3.6 Port Filtering
Permitting communications to specific applications is implemented with rules based on the
source and destination port number in the Layer 4 header.
The port number is essentially an identification of the computer program running on a machine.
This is referred to as port forwarding and filtering.
15.3.7 Firewall
Note that packet or port filtering alone is not a firewall.
Packet or port filtering is a low-level traffic management tool that is the first stage in a firewall.
Firewalls bring in the beginning of a message - contained in a number of packets - and exam-
ine the content of the packets to determine the application being carried in the packet, and ap-
ply permit / deny rules based on that.
The technique used is called Stateful Packet Inspection (SPI).
A packet or port filter bases its permit/deny decision only on the address or port number on the
packet. It does not look inside the packet to see what the content is.
Hence, a properly-configured packet or port filter restricts communication of packets to desti-
nations that have a legitimate and desired use – but allows all communications, including at-
tacks, to reach those destinations.
If the traffic does not come from a trusted source, it is necessary to examine the content of
packets permitted through a packet filter to make a final permit or deny decision.
A firewall has both the packet filter and SPI functions.

15.4 IPv4 Address Classes

To send information from a machine in one broadcast domain to a machine in a different
broadcast domain, it is necessary to have a router relay the information from one broadcast
domain to another.
15.4.1 Packets and Network Addresses
The mechanism for this is to give each machine a unique network address. The information to
be communicated is placed in a packet and the network address of the desired destination is
populated in a field in the packet header.
The packet is then broadcast by the sending machine on its broadcast domain, which includes
the router.
The router receives the packet and uses the contents of the destination address field as an in-
put to making a route decision.
The decision is implemented by the router transmitting the packet in a frame on a different out-
going broadcast domain.
IP version 4 is a standard method of formatting packets and network addresses, specifying 32-
bit-long network addresses called IPv4 addresses or simply IP addresses.
This is the current network address standard for the Internet.
15.4.2 Historical Network Classes
In the beginning, the Internet was the Inter-net, a protocol for addressing machines that were
on different pre-existing networks that used other packet and network addressing schemes.

220
A method of assigning IP addresses to machines on these pre-existing networks was neces-
sary.
To make routing tables efficient, it was desirable to associate a contiguous range or block of
addresses with a pre-existing network. The block would ideally have as many addresses as
there were machines in that network.
The developers decided to standardize on three typical sizes of networks, which they called
classes of networks: big, medium and small, or Class A size networks, Class B size networks
and Class C size networks respectively, and so three standard sizes of blocks of addresses
that would be assigned: Class A blocks, Class B blocks and Class C blocks.

Figure 150. IPv4 Address Classes

Today, the inter-net addressing protocol has taken over, and the “other” packet and network
addressing schemes have disappeared, so the terminology of A-sized networks, B-size net-
works and C-size networks being connected by the inter-net is dated.
The term “address class” is still used; today, it might best be interpreted to mean “block of IP
addresses”.
15.4.3 Class A, B and C
A Class A address space is a contiguous block of 16,777,216 IP addresses. Since this was
such a large block size, there weren’t very many of them available: 128 Class A blocks.
A Class B address space is a contiguous block of 65,536 addresses. Since it is a smaller block
size, more were available: 16384.
A Class C address space is a contiguous block of 256 IP addresses. 2,097,152 Class Cs were
available.
IPv4 addresses are 32-bit binary numbers. Knowing that 28 = 256, 216 = 65,536 and 224 =
16,777,216, one can see that the classes or blocks were defined so that they lined up with the
byte boundaries in the address space.
15.4.4 Network ID and Host ID
A Class A block of addresses begins with a 0. The first byte of the address of all the machines
would be the same, as it identifies the block of addresses.
In keeping with the original idea of the inter-net, this part of the address space is called the
“network ID”, though it would be more appropriate to call it the “block ID” or “address range ID”
today.
The remainder of the address space, the last three bytes or 24 bits was called the “host ID”,
i.e. machine number, can be used to sequentially number 224 or 16,777,216 machines.
A Class B block begins with 10. The first two bytes are the “network” ID and so would be the
same for all machines. The last two bytes or 16 bits can be used to sequentially number 216 =

221
65,536 machines.
A Class C block begins with 110. The first three bytes of the address are the same for all ma-
chines and the last byte or 8 bits are used to number 28 or 256 machines.
Addresses beginning with 111 were originally reserved for “escape to extended addressing
mode”, then divided into two parts:
15.4.5 Class D and E
Addresses beginning with 1110 are multicast addresses, sometimes referred to as Class D.
The division of address space between “network” ID and “host” ID is not defined.
Addresses beginning with 1111 remain reserved for some unknown use, and are sometimes
called Class E.
15.4.6 Classless Inter-Domain Routing
For more flexibility, blocks of IP addresses are no longer restricted to one of the three sizes.
Blocks of addresses called subnets, that are of any power of 2 size less than Class A can be
defined.
The subnet ID is the first part of the address, which will be the same for all devices in the sub-
net.
The subnet, and its size, is identified with a starting IP address followed by /n, with n indicating
the number of bits at the beginning of the address that are the subnet ID.
This evolution of address assignment from one of three classes to arbitrarily-sized blocks was
called Classless Inter-Domain Routing (CIDR).
15.4.7 Dotted-Decimal Notation
IPv4 addresses are 32-bit binary numbers – but writing 32-bit binary numbers on pieces of pa-
per or computer screens, or speaking them between people is unwieldy if not impossible.
Hexadecimal, a numbering system based on 16s, is a good short form for binary numbers, as
it is simple to convert to binary and segments the address into groups of four bits.
Unfortunately, those that came up with IPv4 decided to use decimal as a short form for binary
numbers, and came up with an awkward notation called dotted-decimal, where the 32 bits are
divided into four groups of 8, then the groups of 8 are converted independently to decimal,
yielding addresses written like 232.155.166.1.
Using decimal as a short form for binary numbers is awkward, since it is difficult to convert be-
tween decimal and binary.

15.5 DHCP
IP addresses may be static or dynamic. Static means that the address assigned to a machine
generally does not change.
Dynamic means that an IP address is assigned to a computer on demand, for a fixed lease pe-
riod. The computer may be assigned a different address each time it demands one.
Addresses are assigned to a computer using the Dynamic Host Configuration Protocol
(DHCP).
15.5.1 Dynamic Addresses for Clients
Dynamic addresses are acceptable for a machine running client software, since the way things
are organized is that the client initiates communications with a server, and includes its return
address (the source IP address) in every packet sent to the server.
15.5.2 Static Addresses and DNS for Servers
To communicate to a server, it is necessary to find out the numeric IP address of the server be-
fore the client can communicate to it. That is often accomplished through the Domain Name
System (DNS), essentially tables where the IP address of a server can be looked up.
To avoid having to frequently update those tables, servers are generally assigned static ad-
dresses.
15.5.3 DHCP Client – Server Communications

222
The system administrator provisions a DHCP server, configured to assign IP addresses within
a specific block (within a subnet) to clients. Computers are loaded with DHCP client software.
Communications between the DHCP client and server are effectively application-layer mes-
sages, coded into ASCII and carried in UDP protocol data units, which are carried in IP pack-
ets, which are carried in MAC frames.
The desired recipient of the messages is indicated as being the DHCP on a machine by popu-
lating in the UDP header destination port = 67 for messages to the server and destination port
= 68 for messages to the client.
The messages are “broadcast”, which means that the destination IP address is all 1s and des-
tination MAC address is all 1s. The actual addresses are used for source MAC and IP ad-
dresses, except that the client uses “0” as its IP address, since of course the whole point of the
exercise is to get an IP address.

Figure 151. DHCP Client and Server

15.5.4 DHCP Message Exchange

Each computer will run a DHCP client when it starts, generating a DHCP Discover message.
Any DHCP server that receives it, and there may be more than one, will respond with a DHCP
Offer message, with an offered IP address and a lease time.
The client will answer with a DHCP Request message to confirm its selection of an offered ad-
dress, then the server will complete the cycle with a DHCP ACK that usually includes other
configuration information such as the IP address of the default gateway (CE router), the IP ad-
dress of one or more DNS servers, the lease time, and the subnet mask, which indicates what
bits in the address are the host or machine ID.
There are several variations on the basic process, all of which are enumerated in the relevant
standards document, RFC 2131.
15.5.5 Lease Expiry
On expiry of the lease time, the DHCP client must begin the discover process anew.
If a DHCP client runs while still holding a valid lease, it will request to be assigned the same IP
address.
If there are many clients constantly running DHCP (an ISP’s customers, for example), then it is
likely that a different IP address will be offered by the server each time a computer runs its
DHCP client.

223
The lease time may be configured by the system administrator to any value. This function
might be useful to help manage situations where there are more clients than addresses.
If the DHCP server reaches the end of its configured range of addresses, it attempts to re-as-
sign previously-assigned addresses to new requestors, beginning with those previously-as-
signed addresses for which the lease has expired. Before re-assigning the address to a differ-
ent machine, the server might optionally ping the address to determine if it is still in use.
15.5.6 DHCP to Assign Static Addresses
Even though it is the “dynamic” host configuration protocol, DHCP is also used to assign static
addresses to machines.
This is accomplished with a table in the server, configured by the system administrator, which
relates MAC addresses to IP addresses.
Whenever a computer with a MAC address contained in the table asks for an IP address, it will
always be assigned the IP address specified in the table.
This allows the assignment of static addresses to computers from a centralized management
system (the DHCP server), conveyance of other information like default gateway and netmask,
and eliminates the need for any human involvement (and its associated errors) in configuring
computers.
In Windows, you can see the IP address currently assigned to a computer, as well as its LAN
card MAC address by opening the Network Connections folder and viewing the detailed “sta-
tus” of the LAN card.
If under “properties” of the TCP/IP protocol the choice “obtain a network address automatically”
is selected, the DHCP client is run at startup and when the adapter is disabled then re-en-
abled.

15.6 Public and Private IPv4 Addresses

15.6.1 Public Addresses
Generally speaking, to obtain an IP address that is valid on the public IP network (the Internet),
it is necessary to rent it from an Internet Service Provider (ISP).
15.6.2 Regional Internet Registries
The ISP is either in turn renting addresses from an upstream ISP, or renting addresses from its
Regional Internet Registry (RIR), which, in turn is allocated addresses by the top-level Internet
Assigned Numbers Authority (IANA). The RIR that rents blocks of addresses to North Ameri-
cans is the American Registry for Internet Numbers (ARIN).
There are no more public IPv4 addresses left to be allocated by IANA to the five Regional In-
ternet Registries.
The Regional Internet Registries have some blocks of IPv4 addresses still available, but the
supply is dwindling, and so the policies for being able to rent blocks of addresses are stringent.

Figure 152. Internet Address Authorities

An ISP has to already be efficiently using 212 = 4096 addresses rented from an upstream ISP
before ARIN will consider allocating them their own block.
An end-user has to demonstrate a need for a block of addresses, such as multi-homing where
they have more than one ISP for availability reasons, and must prove they will use a minimum

224
block of 4096 addresses efficiently.
The cost for a block of 4096 addresses from ARIN is $2,250 per year, or about 50 cents per
address per year.
ISPs resell these addresses as dynamic addresses bundled with end-user Internet access ser-
vice. Web hosting providers resell these addresses as static addresses, bundled with a hosting
plan.
Providers also rent “additional” addresses at costs like $2 per month per address… a markup
of 2400%, and a very lucrative business.
15.6.3 Unassigned or Private Addresses
However, the Internet Society didn’t give all of the IP addresses away. RFC 1918, “Address Al-
location for Private Internets” defines three contiguous blocks of IPv4 address space that are
not used, and not valid, on the public IP network (the Internet).
These addresses are officially called unassigned addresses and usually referred to as private
IP addresses. Sometimes they are called non-routable addresses, though this is not very accu-
rate; routers can route them, just not on the Internet.
Using private addresses in-building allows the use of IP and all of its associated protocols and
services for in-building communications without having to pay anyone for a block of rented ad-
dresses.

Figure 153. IPv4 Private Addresses

While it would be theoretically possible to use any IP addresses on a private network not con-
nected to the Internet, it is recommended to use addresses in the ranges defined in RFC 1918.
Presumably, network equipment would be configured by default to know these addresses are
not valid on the Internet, and so would be better suited to handle them if and when the private
network is connected to the Internet.
But – it is necessary to have a legitimate public IP address to be able to receive anything from
the Internet. A popular solution is to use private addresses in-building, pay for one public IP
address for external communications, and connect the two worlds with a Network Address
Translator (NAT).

15.7 Network Address Translation

In the previous section, we covered private IP addresses, and why these were preferable to
use on an in-building network. We also noted that if any of the users on the private network
want to receive packets from the Internet, a public IP address is required.
To enable Internet communications for all users in-building without having to rent a public IP
address for every user, a Network Address Translator (NAT) may be used.

225
 Figure 154. Network Address Translation

15.7.1 Network Address Translator

A Network Address Translator is a software program running on the Customer Edge device.
It has a DHCP server, connected to the in-building network and configured to assign private
addresses to the machines in-building, and a DHCP client, connected to the ISP, which obtains
a public IP address from the ISP.
15.7.2 Outbound
When a computer on the private side initiates communications with a server, it populates the
source IP address field in the packet header with its private address and the destination IP ad-
dress field with the public IP address of the server.
The packet is then transmitted in a MAC frame to the computer’s “default gateway”, which is
the Customer Edge router, where the NAT function is performed.
The NAT changes the source IP address from the private IP address of the sender to the pub-
lic IP address of the NAT, i.e. the CE router, then transmits the packet in a frame on the public
network (the Internet).
15.7.3 Inbound
The Internet server uses the source address in the packet it receives as the destination ad-
dress to answer back to the client. Therefore, it will send the response back addressed to the
NAT.
When the NAT receives the packet, it changes the destination IP address on the packet from
the Internet to the private IP address of the appropriate computer, then transmits the packet in
a MAC frame to the computer.
One question that arises is: how does the NAT know what computer on the private network a
packet received from the Internet is intended for?
It turns out that the NAT uses the Layer 4 header to keep track of things. The Layer 4 header
(TCP or UDP) begins with two octets that are called the “source port” then two octets for the
“destination port”.

226
These fields are used to indicate which application on a computer the message is being sent
from and to.
The NAT selects an arbitrary “fake” port number to identify a computer on the private network,
and records this port number against the private address in a table.
When a packet is transmitted to the Internet, the NAT records the actual source port number
then changes the source port value to the “fake” port number.
When the reply from the server is received from the Internet, it has the “fake” port number in
the destination port field of the Layer 4 header. The NAT uses this to look up the correct private
IP address and correct port number and enter those values in the destination address and des-
tination port number fields, thus relaying the incoming packet to the correct computer on the
private network.
15.7.4 Advantages of NAT
NAT provides a number of advantages:
1. A NAT allows multiple computers in-building to share a single Internet address and Internet
connection.
2. A NAT provide a truly “always-on” connection to the Internet. Services like DSL and Cable
modem described as “always on” are always connected at the Physical Layer. They do not
provide “always on” at the Network Layer, since DHCP must be run every time the attached
device restarts to get a public IP address.
When a NAT is inserted, it runs DHCP to get the public IP address; so if the NAT is not pow-
ered off, the site will always have a public IP address assigned, and thus a connection to the
Internet always ready for immediate use.
3. A NAT shields machines from attacks from the Internet. Since a private IP address is not
reachable from the Internet, there is no way for a machine on the Internet to initiate communi-
cations to a machine on the private network. The only device exposed to the Internet is the
NAT.
Normally, the NAT is not running on a computer running Windows, so attackers have a greatly
diminished chance of finding a vulnerability to exploit compared to connecting a computer run-
ning Windows naked onto the Internet.
15.7.5 Implementation
Devices that perform this function are available in industrial-strength versions from companies
like Cisco. Hardware devices to do this are also available for about $20 from companies like
Linksys for use on a DSL or cable modem connection. They often include both an Ethernet
switch and an 802.11 wireless LAN access point for the private network side.
Most ISPs now provide the CE router with NAT function integrated in a device that includes the
DSL or Cable modem, or for the lucky few, the fiber terminal supplied by the ISP.

15.8 TCP and UDP

Implementing IP yields an unreliable, connectionless network service. There are no guaran-
tees that a packet will be transmitted, when that might happen, nor how often that might hap-
pen. There is no guarantee that a packet will be received.
In fact, in IP, there is no way for the device to which a packet has purportedly been transmitted
to report back whether it received the packet or not. The packets have to fend for themselves
on the network and may be corrupted or discarded at intermediate nodes.
To make communication over an IP network reliable, users must run transport-layer protocols
end to end. The most popular is the Transmission Control Protocol (TCP), which performs re-
transmits for file transfers and assures integrity. Another choice is the User Datagram Protocol
(UDP), which is used for “best efforts” transmission of individual packets, and does not do re-
transmits.
Before a segment of data is passed to IP, TCP adds a header to the segment with an error
check and sequence number, and starts a timer at the sender.
The receiver’s TCP checks the error check and sequence number. If the data is corrupted, the
receiver discards the data, and after a period of time, the timer started by the sender’s TCP will

227
expire, and the sender will automatically retransmit the segment.

Figure 155. TCP Protocol Data Unit

If the data is received without error, the receiver’s TCP sends an acknowledgment back to the
sender’s TCP and the sender stops retransmitting.
TCP turns the underlying unreliable connectionless IP network into a reliable transport service
for use by upper layers. TCP is for file transfers – when it absolutely, definitely has to get there,
and if it doesn’t, the missing piece(s) will be retransmitted.
This is not very useful for voice and video. During a live, streaming communication session,
there is no time to retransmit missing pieces. For these applications, a different transport-layer
protocol, the User Datagram Protocol (UDP), is employed. It is similar to TCP, but does not im-
plement retransmission of missing or errored data. It provides best-efforts transport.

15.9 IPv6
The main limitation of IPv4 was a shortage of network addresses. Though having a 32-bit ad-
dress space, yielding 232 or 4.3 billion addresses, the assignment of large blocks of addresses,
particularly Class A, caused a rapid exhaustion of available addresses.
Additionally, mechanisms for security and traffic management were not provided or not well
supported by IPv4, requiring the development of additional protocols, headers and overhead to
perform these functions.
RFC 2460 Internet Protocol, Version 6 (IPv6) emerged from a pack of contenders to be the
eventual replacement for IPv4. The improvements that IPv6 offers over IPv4 are expanded ad-
dressing capabilities, header simplification, improved support for extensions and options, sup-
port for traffic management and support for data integrity and data security.
15.9.1 Expanded Addressing Capabilities
The main improvement is expansion of the address field from 32 bits to 128 bits, expanding
the available address space to 2128 = 3 x 1038 addresses
(340,282,366,920,938,463,463,374,607,431,768,211,456 to be exact) … enough to allocate a
block of 560 trillion trillion addresses to every person on earth.
15.9.2 Header Simplification
Some IPv4 header fields have been dropped or made optional, to reduce the bandwidth cost of
the IP header, and to reduce the number of operations – and thus time – required to forward a
packet.
15.9.3 Improved Support for Extensions and Options
A flexible mechanism for adding to the IP header with variable-length extension headers has
been implemented. This allows optional implementation of error detection, source authentica-
tion and encryption as standardized services at the network layer that would be available to all
applications.
15.9.4 Support for Traffic Management

228
Fields in the header allow identification of priority, and identification of a packet as belonging to
a flow of packets, that is, a sequence of packets originating from the same source and going to
the same destination and intended to receive the same forwarding treatment. Currently, this
would be an MPLS label number. These capabilities can be used to implement traffic manage-
ment and prioritization as Quality of Service mechanisms.
15.9.5 IPv6 Packet Format
The IPv6 header is 40 octets long, and includes the following fields:
Version field: 4 bits, indicating the version of IP. This would be “6”.
Traffic Class: 8 bits, indicating a “priority” or precedence for this packet. This field can be popu-
lated by the originator of the packet, or by subsequent network equipment. This could be used
to support differentiated Classes of Service for different applications.
Flow Label: 20 bits that can be used to identify the packet as belonging to a group or class of
packets which should receive the same forwarding treatment on the network. This would typi-
cally be an MPLS label.
Payload Length: 16 bits, containing length of the payload immediately following this header,
which includes any optional extension headers.
Next Header: 8 bits identifying the type of header following. In the simplest implementation, the
IP packet will be encapsulating a transport layer protocol data unit, such as that output by TCP.
In that case, the header immediately following the IP header would be the TCP header.
Hop Limit: 8 bits, populated by the source with a number between 1 and 255. This number is
decremented by each device that forwards the packet. When it reaches zero, the packet is dis-
carded. This prevents endless forwarding of packets in loops.
Source Address: 128 bits, identifying the originator of the packet.
Destination Address: 128 bits, usually identifying the final destination on the network. If a Rout-
ing extension header is present, the destination address field will contain the address of the
next router through which the packet must travel.

Figure 156. IPv6 Packet

The IPv6 header could also be followed by one or more IPv6 extension headers, which can in-
clude a Hop-by-Hop Options header, Routing header, Fragment(ation) header, Destination Op-
tions header, Authentication header and/or an Encapsulating Security Payload header.

15.10 IPv6 Address Allocation and Address Types

IPv6 addresses identify interfaces. An interface is typically an integrated circuit driving a wired
or wireless LAN connection on a device.
The notation /n is used to mean the first n bits in the address.
15.10.1 Internet Registry Identification
The first 12 bits of the address identifies the Regional Internet Registry.

229
In North America, ARIN’s policy is that the first 32 bits of the address identifies a block allo-
cated to a Local Internet Registry, most of the time a big ISP.
15.10.2 Sites and Global Routing Prefix
The first 48 bits of the address is called the Global Routing Prefix and identifies a site. Most of
the time, this will be an ISP’s data center, though it might be a university campus or large orga-
nization’s building.
15.10.3 Interface ID
The last 64 bits of the address is called the Interface ID, and could be the updated version of a
MAC address called EUI-64, or a random number for privacy reasons.
It identifies the integrated circuit running a LAN connection – wired or wireless – on a device.
For consumer equipment with one LAN connection like a PC or smartphone, it effectively iden-
tifies the device.

Figure 157. IP version 6 Address Structure

15.10.4 Subnet ID
Between the 48-bit Global Routing Prefix, which essentially identifies buildings or campuses,
and the 64-bit Interface ID, which identifies LAN connections is 16 bits called the Subnet ID.
The Subnet ID can be used to implement a hierarchy of addresses assigned to end-users
and/or subnets at a particular end-user.
Residential users generally do not have multiple subnets, so in the case of an ISP’s site, this
16-bit field can be used to assign one subnet, that is, one /64 block to 65,536 customers per
site.
In this case, all IP addresses at the residence (the end-site) would have the same first 64 bits,
and the last 64 bits would be IDs of interfaces at the residence. Every light switch, light bulb,
every electrical socket, both slots in your toaster … everything will have an IP address in the
future.
Large government and corporate end-sites would normally have more devices and multiple
subnets (broadcast domains) to be compartmentalized for network security reasons, so they
might be assigned multiple subnets, for example a /56 block from the ISP’s site.
In this case, the end-user in the office building would employ the lower 8 bits of the subnet field
to identify up to 256 subnets (broadcast domains) at their end-site. On each of these, the first
64 bits of the IP address would be the same for all devices on the subnet (in the broadcast do-
main), and the last 64 bits are the Interface ID. The ISP could service up to 256 of this kind of
customer from one /48 site block at the ISP’s data center.

230
Customers of an ISP that have sites bigger than a large corporation or government building in-
clude… smaller ISPs. For this type of customer, a /48 block would allow the downstream ISP
to resell the /64 block residential and /56 block corporate / government scenarios just de-
scribed.
15.10.5 Subnet Prefix
The first 64 bits of the address are called the Subnet Prefix, and identifies a broadcast domain.
15.10.6 IPv6 Address Types
Three main IPv6 address types are defined: unicast, anycast and multicast.
A unicast address identifies a single interface. A global unicast address is basically a valid In-
ternet address – that may or may not be directly reachable from the Internet, for security rea-
sons.
Both anycast and multicast addresses identify a set of interfaces. A packet addressed to an
anycast address is delivered to the nearest interface in the set, while a packet addressed to a
multicast address is delivered to all of the interfaces in the set.
The address 0 is called the unspecified address, used as the source address of an interface in
the process of acquiring an address using DHCP, for example. The address 1 is called the
loopback address, and is used by an interface to reference itself.
Addresses beginning with 1111:1101 (FDH) are called unique local addresses, used in the
same way as IPv4 private addresses. These addresses can be routed on a private network,
but are not valid on the public internet.
Addresses beginning with 1111:1110:10 (FE8H - FEFH) are link-local unicast addresses. They
end with a 64-bit interface ID, and are valid only on a single broadcast domain, for functions
like neighbor discovery. Routers are not allowed to forward packets addressed to these ad-
dresses to a different broadcast domain.
Addresses beginning with 1111:1111 (FFH) are multicast addresses. All other addresses are
global unicast addresses, i.e. addresses for the public Internet. Anycast addresses are taken
from unicast address space.
One other type of address worth noting is the IPv4-mapped IPv6 address. This begins with 80
zeros, then 16 ones (FFFFH), followed by a 32-bit IPv4 address. This is a method for transition
from IPv4 to IPv6 and may end up being the way that a “legacy” IPv4 addressing scheme is
accommodated on an IPv6 network.
RFC4291 “IPv6 Addressing Architecture” is the authoritative reference for the discussion in this
section.

231
16
MPLS and Carrier Networks
16.1 Introduction
Packet networks incorporate two ideas: packet switching and bandwidth on demand. In this
chapter, we examine how these principles are implemented by common carriers, i.e. organiza-
tions that build networks and carry many users’ packets over common facilities.
16.1.1 Overbooking
Packet switching, also called packet forwarding and routing, means relaying user data in pack-
ets from one circuit to a different circuit, or to be exactly precise, from one broadcast domain to
a different broadcast domain.
Routers physically move packets from one circuit to another, using network addresses to de-
termine which circuit to relay them to.
Bandwidth on demand means giving many devices access to a circuit and giving each the
possibility of transmitting. If a device does not have anything to transmit, another device can
use the available capacity.
This allows the implementation of overbooking or oversubscription, where the total of the in-
coming line speeds is greater than the outgoing line speed.
The appropriate level of overbooking can be calculated based on the historical demand statis-
tics, how often the devices actually transmit data – regardless of what access line speed they
have – and so overbooking is also called statistical time-division multiplexing.
16.1.2 Congestion, Contention and Packet Loss
When the demand exceeds the available capacity – more packets being sent in to a router
than can be sent out – the network is said to experience congestion. At a router, packets are
stored in temporary memory called buffers while waiting to be transmitted.
Under heavy load, the buffers can fill up. Then, in the case where a new packet arrives before
the oldest one in the buffer can be transmitted, the new packet over-writes the oldest one in
the router’s buffer memory.
When a packet is over-written, it disappears. This is also called a dropped packet, non-deliv-
ered packet and packet loss.
For applications like email and web pages, packet loss is typically not a problem; the TCP soft-
ware re-transmits the missing data in a new packet from the source. The user might only no-
tice the page taking longer to load.
For live telephone calls and television programs carried in IP packets, there is no time to re-
transmit missing data, so packet non-delivery and excessive delay can result in poor voice
quality on phone calls, and serious pixilation or block-averaging distortions on video.
16.1.3 MPLS Traffic Management System
To assure suitable performance for delay- and packet-loss-sensitive applications, a traffic man-
agement system called MPLS is used by the network operator.
Performance is usually defined as packet delivery percentage, maximum delay and maximum
variability in delay.
Specific performance thresholds are called Classes of Service (CoS), and are part of the ser-
vice contract between the carrier and their customer called a Service Level Agreement (SLA).
One Class of Service could be defined for delay- and packet-loss-sensitive applications like
voice and video, with guaranteed high packet delivery rate and low delay.

232
Another CoS could be defined for delay- and packet-loss-tolerant applications like web pages
and email, with lower guaranteed packet delivery rate and longer delays.
If there is congestion at a router, packets with the higher CoS are transmitted at the expense of
packets with a lower CoS that are delayed or dropped.
In addition to managing performance guarantees, traffic management is also required for net-
work load balancing and recovery from equipment failure and cut lines.

16.2 Carrier Packet Network Basics

A carrier builds a packet network by obtaining expensive, high-capacity routers and placing
them in buildings in different cities.
These buildings might be called switching centers, toll centers or Central Offices when the car-
rier is also the local telephone company; other times the buildings might be called POPs or
possibly data centers.
The carrier then connects these routers at their own expense. This forms the network core or
backbone.

Figure 158. Carrier Packet Network

The connections are implemented with capacity on a fiber, either owned by the carrier or
leased from a third party.
Redundant connections are made to ensure high availability: a minimum of two connections
are required at each location to protect against cut lines.
The cheapest way to implement two connections at each location is to connect neighbor-to-
neighbor to form a ring. In practice, additional “shortcuts” will be implemented where traffic
warrants.
16.2.1 Provider Edge (PE) and Customer Edge (CE)
This network core is front-ended with carrier edge equipment, often called the Provider Edge
(PE).
The Provider Edge router is connected to the Customer Edge (CE) router with a physical dedi-
cated access circuit.
The Customer Edge router is customer premise equipment situated between the customer cir-
cuits and the access circuit. It performs functions that can include acting as a point of control
for traffic as part of network security, a data concentration function and conversions between
frame formats, physical mediums and line speeds.
The Provider Edge is equipment owned by the carrier, performing similar functions: a point of
control for network security, data concentration, media and format conversions. Some of the

233
PE functions may be implemented in provider equipment deployed at the carrier premise. This
is discussed in an upcoming section.
16.2.2 Access
The access circuit is generally a dedicated physical connection.
Current choices for access circuits include copper-wire DSL and cable modem technologies at
speeds measured in the tens to hundreds of Mb/s, and fiber systems moving 10 Mb/s to 40
Gb/s. These line speeds will continue to increase in the future.
In days past, copper-wire technology including 56 kb/s digital service and 1.5 Mb/s T1 were
popular choices. Higher-end services might have used SONET on fiber at rates like 150 Mb/s
(OC3) or 500 Mb/s (OC12).
16.2.3 Advantages of Packet Networks
There are a number of significant advantages to the use of packet network services from carri-
ers instead of dedicated lines or circuit-switching.
First, there is no circuit set-up delay as with a dial-up modem; with a packet-switched service,
the possibility of communicating to the network is maintained constantly over the access cir-
cuit, so communications can begin anytime without delay simply by transmitting a packet from
the Customer Edge to the Provider Edge router. The packet is delivered to the far end CE
sometime later.
Second, users can send packets addressed to many different destinations interspersed on a
single access circuit, thus communicating ‘simultaneously’ to many destinations with only one
access circuit at each location.
Since there is a monthly charge per access circuit or port in this business, this is a large cost
advantage compared to dedicated or circuit-switched services, which require a separate ac-
cess circuit for each simultaneous connection.
The third advantage is cost.
For all types of services, there is a flat rate per month for the access. Then, for dedicated lines,
there is a mileage charge per month. Circuit-switched phone calls are billed per minute.
Packet services are (in theory) billed per packet. Since the users normally will be doing noth-
ing, it will be cheaper to pay per packet than pay all the time – which is a “dedicated line” actu-
ally is.
Whether or not a packet service is billed per packet, or flat rate, or flat rate up to a bandwidth
cap then per packet for overage is a business decision on the part of the carrier.
In any case, the cost for packet network service is less than the cost of dedicated line or cir-
cuit-switched services, because the network circuits are overbooked and so the cost to the car-
rier is lower.

16.3 Service Level Agreements

A Service Level Agreement (SLA) is a contract between the customer and the service provider.
It is the core technical specification of the relationship between the customer and the service
provider… specifying what the customer is going to get when transmitting packets over the
carrier’s network.
The service provider agrees to provide specified transmission characteristics, on condition that
the customer stays within a specified traffic profile.
A particular set of guaranteed transmission characteristics is often referred to as a Class of
Service (CoS). It would typically include specification of the minimum packet delivery rate, the
maximum end-to-end delay and the maximum variability in delay.
16.3.1 Traffic Profile
A traffic profile is the specification of the bandwidth – the number of bits per second that the
user will transmit over time. It is the “statistics” in statistical multiplexing.
The traffic profile would specify the maximum average bit rate measured over a specified pe-
riod of time, the maximum peak or burst rate, how much data the user is allowed to transmit at
the burst rate, and how often bursts are allowed to occur.

234
16.3.2 Contract
The contract is: as long as the customer traffic is less than the restrictions of the traffic profile,
the service provider is obliged to provide the specified Class of Service, i.e. meet the specified
transmission characteristics.
If the provider does not meet the specification, then in many cases, the customer can get a
partial rebate of service cost for the month.
16.3.3 Business Decisions
Whether the service provider requires the customer to specify a traffic profile, and the details of
traffic profiles are a business decision on the part of the carrier.
Some carriers, perhaps to distinguish themselves from the competition, do not enforce traffic
profiles. Or perhaps more accurately, they only support one traffic profile, which is “transmis-
sion at full access line speed 24/7”.
If the customer has a physical access to the provider’s network at 10 Mb/s, then they are al-
lowed to transmit 10 Mb/s all the time – and still require the carrier to meet the Class of Service
guarantees.
Other carriers only sell SLAs with traffic profiles; in other words, they will only guarantee a
Class of Service if the customer guarantees they will restrict their traffic below set limits.
Historically, this has been the case because of limited resources – because the carrier network
is heavily overbooked, and the carrier must restrict the incoming traffic to be able to meet the
CoS guarantees.
Their network is in no way capable of supporting all customers transmitting at full line speed
24/7 at the same time. The carrier’s network was purposely designed to support lower traffic
profiles, to reduce the cost, giving the customers high apparent network speeds at a lower
cost.
An easy example of this today is residential internet access. To reduce the cost to customers,
the access network in a neighborhood is heavily overbooked.
When a customer signs up for Internet access at 25 Mb/s for $60/month, they are not paying
for a traffic profile that allows them to transmit and receive 25 Mb/s 24 hours per day, 7 days a
week, 365 days per year.
They are paying for a residential user traffic profile, which is “receive at 25 Mb/s in short bursts
once in a while”.
Two-way 25 Mb/s Internet access guaranteed to work at full speed 24 hours per day, 7 days a
week, 365 days per year is available – but that costs $500 per month not $60 per month.
16.3.4 Enforcement: Out of Profile Traffic
On the enforcement side, traffic is metered, and traffic exceeding an agreed profile is said to
be out of profile.
The question is then what to do with out of profile traffic. If the network is heavily overbooked,
customers might need to be forced to conform to their agreed traffic profile if the network is go-
ing to meet the Class of Service standards for all customers… and not have to give refunds to
everyone every month.
In this case, out-of-profile traffic might be:
• assigned to a lower Class of Service, and so becoming more likely to be dropped by some
equipment downstream,
• billed at a higher rate,
• temporarily delayed to bring short-term overages into compliance with the traffic profile
(called traffic shaping), or
• discarded at the input to the network (traffic policing).
For business customers, traffic profiles were required in the past for the same reason: the net-
work was overbooked.
Today, if there is practically infinite bandwidth underlying all parts of the networks between
business customer locations, then there is no technical need for traffic profiles.

235
There is, however, a business reason why a carrier would continue requiring traffic profiles af-
ter the technical need for them has diminished… which is of course revenue.
Out-of-profile traffic could be, and is, billed as “overage”. From a service provider business
point of view, the service provider is keeping the cost of service offerings low for everyone, and
charging extra to those that transmit more than everyone else.
From a customer point of view, this might be seen as a hidden cost that is only discovered af-
ter a contract is signed.
16.3.5 Abusive Applications: Bit Torrent
Residential customers that run the bit torrent file “sharing” application transmit and receive far,
far more traffic than regular users, and usually face traffic policing and/or charges for overage.
Some go as far as to claim that net neutrality is required, and in their case, they believe net
neutrality means they should not be subject to traffic caps, traffic policing (throttling), or extra
charges, instead pay the same as their neighbors – even though they are using far more band-
width.
It must be noted that the activity these customers are usually undertaking is reception and dis-
tribution of intellectual property for which they have not paid the copyright holder.
From a technical point of view, these users are essentially asking their neighbors to subsidize
what the copyright holder would likely consider to be criminal activity.

16.4 Provider Equipment at the Customer Premise

The Provider Edge (PE) is equipment owned by the carrier, performing a number of functions:
a point of control for network security, data concentration, media and format conversions.
As we will see in this chapter, one of the “format conversions” the PE may implement is affixing
and removing virtual circuit IDs – called labels in MPLS. It turns out that the best place to per-
form this function is at the customer premise – so it is not unusual to see at least part of the
Provider Edge function in equipment deployed at the customer premise.
The PE equipment at the customer premise can also be the traffic meter and traffic policing de-
vice described in the previous section.

Figure 159. Provider Equipment at the Customer Premise

This PE equipment at the customer premise is the ingress device, called a Label Edge Router
in MPLS and the upcoming sections.
Having equipment at the customer premise also allows the service provider to use it as a re-
mote test head, so the provider can perform service level assurance: monitoring, tests and
troubleshooting between a centralized test system and the equipment at the customer
premise.

236
16.5 Virtual Circuit Technologies
Traffic management on IP networks is implemented with virtual circuits internal to the network,
invisible to users, with a protocol called Multi-Protocol Label Switching. MPLS is a tool en-
abling the centralized control of routing and prioritization of different kinds of network traffic
such as telephone calls vs. file transfers.
The remainder of this chapter explains the basic principles of virtual circuits, briefly covers
legacy technologies, then MPLS and how MPLS is used for traffic management, VPNs, service
integration and traffic aggregation.
In an IP network, the route decision for a packet arriving at an IP router is calculated using a
relatively complicated algorithm that takes into account the destination IP address, the cost of
different routes to the destination and other factors.
This algorithm is run for each packet, on every router in the chain. This takes a relatively long
time, increasing network delay. More importantly, it makes practically impossible the control of
routes, prioritization and resulting traffic characteristics since each router operates indepen-
dently.

Figure 160. Virtual Circuits

The idea behind virtual circuit technology is to not run the IP routing algorithm on each packet
at every router, but instead defining classes of traffic, pre-determining the end-to-end route for
all traffic in a class, and programming the route for the class into the routers from a software
application in a Network Operations Center.
16.5.1 Traffic Classes
A class of traffic has the same source, same destination and should experience the same
transmission characteristics, such as maximum delay and loss.
To establish communications, classes of traffic are defined, then a route is determined for each
class. The route is generically called a virtual circuit: it is the path that all of the traffic belong-
ing to the class will follow – if any such traffic is ever transmitted in the future.
16.5.2 Traffic Class ID & Virtual Circuit ID
A number is used to refer to both the traffic class and the route. This number is generically
called a class number and a virtual circuit ID. There is specific jargon for each technology.
To implement the virtual circuit in the network, the routing table in each piece of network equip-
ment along the virtual circuit is populated with an entry that specifies the next hop for the class.
This completes the setup.
16.5.3 Ingress Device: Packet Classification
Later, when traffic (e.g. a packet) is actually presented to the network for transmission, a piece
of equipment called the ingress device analyzes the traffic to determine what class it belongs
to: it classifies the incoming traffic. Once an answer is reached, the ingress device stamps the
class number on the packet.

237
In many cases, the ingress device is part of the Provider Edge (PE), meaning that the entire
story of virtual circuits and traffic classes is internal to a carrier’s network and invisible to the
customer.
The best place to at least initially classify traffic is at the customer premise, as this allows use
of local information not available in the carrier network, and allows subsequent aggregation of
multiple traffic classes on a single physical access.
For this reason, the traffic classification function is often performed by provider (carrier) equip-
ment located at the customer premise.
In other cases, the ingress device is part of the Customer Edge, meaning that the customer is
responsible for purchasing this complicated equipment, managing it and coordinating traffic
classes and virtual circuit IDs with the carrier at both ends.
16.5.4 Forwarding Based on Class Number
Once the traffic is classified, the carrier network equipment does not use the IP packet ad-
dress, but instead uses the class number to look up the next hop and possibly relative priority
in its routing table.
This reduces the routing decision at each router from a complex algorithm to a simple table
lookup, reducing the delay through each router and load on the router’s processor.
More importantly, it provides a carrier with a mechanism for managing flows of packets end-to-
end, implementing load balancing and swift service restoration after a fault, by managing the
next hop entry for each class in each routing table from a centralized Network Operations Cen-
ter (NOC).
16.5.5 Differentiated Services
Implementing multiple traffic classes that might all go from the same place to the same place,
but each associated with a different priority allows the implementation of Differentiated Ser-
vices, i.e. multiple Classes of Service on a packet network, so that packets with different con-
tent will experience different transmission characteristics. An alternate method, particularly
suited to IPv6, is one virtual circuit code and a separate traffic class code.
16.5.6 SVCs and PVCs
In some technologies, there are additional buzzwords describing two flavors of virtual circuits:
Permanent Virtual Circuits (PVCs) and Switched Virtual Circuits (SVCs).
Switched Virtual Circuits are set up in a manner similar to making a phone call: your network
equipment asks the network to establish a connection to some destination, the network sets it
up without further human intervention, you communicate, then the connection is released when
you have finished communicating.
The difference between an SVC and a phone call is that full-time capacity is not reserved in
the network for an SVC… it is just a path, a route, a possibility. With a phone call, 64 kb/s are
reserved in the network during your communication session, whether you’re using them or not.
Permanent Virtual Circuits are exactly the same as Switched Virtual Circuits, except that they
are set up and never released.
The set-up process for virtual circuits is in many cases a manual operation, performed by a
technician sitting at a control console that commands the network routing equipment.
For this reason, in practice all virtual circuits tend to be Permanent Virtual Circuits… set up and
left set up.

16.6 Packet-Switching using Virtual Circuits

We now turn to examining actual protocols and technologies to implement virtual circuits on
carrier packet networks. We begin in this section with a brief overview of the first technology,
X.25, to pave the way for understanding Frame Relay (still in use), the “fading star” ATM and fi-
nally the current technology, MPLS. These four technologies are all essentially the same thing:
virtual circuits – with different jargon and buzzwords.
We’ll use X.25 to establish a graphical method that illustrates the protocol stacks on each de-
vice, and how packets travel in frames over physical circuits from one system to another, within

238
the framework of the OSI 7-layer reference model.

☞
If you are not interested in the precursor / legacy technologies, feel free to skip ahead to
   Section 16.9, “MPLS”.
X.25 was a widely-accepted standard protocol for packet networks, standardized by the CCITT
(now ITU) in 1976 and deployed by all telephone companies for business and government
data communications. Some precursors of what we know today as the Internet, for example,
CompuServe, ran on X.25 networks.

Figure 161. X.25

X.25 typically offered at most 56 kb/s access speed, and less throughput. It was not scalable to
higher line speeds, and did not have any QoS mechanisms to provide service level guaran-
tees. It is now obsolete.
16.6.1 X.25 Network Structure and Operation
At the left of Figure 161 is the customer equipment, a terminal or computer (called Data Termi-
nal Equipment (DTE) by the ITU), plus another function called a Packet Assembler-Disassem-
bler (PAD). The far-end customer equipment is at the right of the diagram. In the middle is an
X.25 network operated by the phone company, the military or another private company.
This X.25 network is composed of routers (called packet switches) in cities connected long-dis-
tance. This is called the core and regional rings in other chapters.
The Customer Edge is the PAD. On one side, terminals are connected by data circuits. On the
other side, it runs the X.25 protocol stack to communicate with the packet switch. The PAD as-
sembles keystrokes or file records received from the terminal into packets, carried in frames,
signaled over an access circuit to the packet switch.
First, this packet communication capability is used for a control function. The terminal commu-
nicates via the PAD to network control equipment, requesting that a connection to a particular
far-end terminal be set up.
The network sets up a virtual circuit to that terminal, populating routing tables all along the line,
and returns the virtual circuit ID. Then the packet communication function is used by the near
end to communicate to a far-end PAD by populating the far end’s virtual circuit ID in the packet
header and transmitting it to the packet switch.
The packet switch receives the frame one bit at a time, and once it has the entire frame, layer
2 software on the packet switch performs an error check, verifies its link address is on the
frame, does any resequencing or error recovery necessary at the frame level, and when all of
this is complete, extracts the packet from the frame and gives it to a second piece of software,
the layer 3 network software running on the X.25 packet switch.
The layer 3 network software on the packet switch performs error recovery or resequencing at
the packet level, then uses the virtual circuit ID to look up in its routing table where to send the

239
packet.
This software passes the packet back to the layer 2 software on the packet switch, along with
the link address of the next hop. The layer 2 software on the packet switch revises the frame
address, recalculates the frame error check value then transmits the frame on the appropriate
physical output to the next packet switch.
This repeats until the packet is delivered in a frame to the far-end PAD, which extracts the data
from the packet and passes it to the far-end application.
16.6.2 Reliable Network Service: Guaranteed Delivery
X.25 implements an error-recovery mechanism, retransmitting missing or errored data on indi-
vidual links, to guarantee delivery of user data (supporting dumb terminals). There are no tim-
ing or delay guarantees.
X.25 in effect implements a single Class of Service that might be called guaranteed data deliv-
ery. This is also referred to as a reliable network service.
16.6.3 Connection-Oriented vs. Connectionless Network Service
Connection-oriented communications means in general that there is communication with the
far end before a file transfer begins; the sender gets an acknowledgment that the receiver is
online and ready to accept data. Since a virtual circuit is set up before communications begins,
X.25 implements connection-oriented communications at the network level. X.25 was reliable,
connection-oriented packet communications from the Phone Company.
Contrast that with unreliable, connectionless network service. Unreliable means that there is
no guarantee from the network that a packet will be delivered, and no acknowledgment of
transmission of a packet is provided by the network.
Connectionless means that there is no communication with the far end before a file transfer
begins. To use an unreliable, connectionless network service, the user must perform the relia-
bility and connection functions. The Postal Service is an example of an unreliable, connection-
less network service. The Internet is another.
Business, government and the military liked X.25, because it was cheaper and much more
flexible than dedicated lines, allowing communication to many locations over one access cir-
cuit.
However, X.25 was not scaled to higher throughputs to support LAN-LAN communications,
and did not support any kind of class of service other than “data”. It was necessary to deploy
other packet network and virtual circuit technologies that supported higher line speeds and
more sophisticated classes of service.

16.7 Frame Relay using Virtual Circuits

Frame Relay was the standard mainstream solution for business and government data wide-
area networking popular in the 1990s and 2000s.
Carriers have put a cap on new deployment of Frame Relay, but there are many organizations
still using it, pending a migration to IP and Optical Ethernet. In this section, we review Frame
Relay, its jargon and buzzwords before moving on towards discussing its replacement, MPLS.

240
 Figure 162. Frame Relay

16.7.1 Elimination of a Layer of Software

Frame Relay is a bandwidth on demand service that was faster than X.25. It could provide
communications at rates of up to 1.5 Mb/s (and from some carriers, 45 Mb/s) and so was suit-
able for client-server wide area networking.
Frame Relay is faster than X.25 mainly because of the elimination of a layer of software.
In a packet-switched network, information on the packet header is used for routing decisions.
At every packet switch, layer 2 software has to receive the frame, perform the error detection,
examine the frame address, do any error recovery necessary, then extract the packet from the
frame and pass it to a second software program, the layer 3 routing software.
This software receives the packet, does any network-level error recovery, then looks at the net-
work address on the packet, uses that to make a route decision, then passes the packet back
to the layer 2 software along with an indication of the outbound link. The layer 2 software
changes the link address in the frame and recalculates the frame check sequence before
sending it off.
This would be like sending a parcel via UPS and putting the address inside the box. Any time
a UPS employee had to make a routing decision, they would have to open the box, find the ad-
dress, look at it, make a decision, tape the box shut and then throw it in the appropriate pile.
Wouldn’t it be faster to put the address on the outside of the box?
This passing of the packet between two sets of software and the duplication of functions takes
a relatively long time, reducing the end-to-end throughput.
The main idea behind Frame Relay is to have the network equipment use information in the
layer 2 header (the frame header) to make routing decisions. Then, it is not necessary to have
a second (layer 3) software program on the network equipment, and not necessary to pass a
packet from one piece of software to another then get it back on every piece of network equip-
ment.
In this way, a layer of software is eliminated, speeding up routing, and changing the service
from being a packet-switched service (layer 3 + layer 2) to a frame-relay service (layer 2 only).

241
This is implemented by defining virtual circuits and associating them with a virtual circuit ID
called a Data Link Connection Identifier (DLCI), and populating this in the frame address field.
A control system configures routing tables in the network elements with the routing for each
DLCI. When a frame arrives at the network element, the DLCI on the frame is used to look up
the next hop.
On a private network, the DLCI can be the same across the network. On a public network, the
DLCI will change from link to link, and so the end-to-end connection is a virtual circuit made up
of a sequence of DLCIs.
16.7.2 Unreliable Service
Another reason why Frame Relay is faster is because it is does not provide reliable network
service... the delivery guarantee and error recovery protocols of X.25 are replaced with mere
error detection.
If a frame is corrupted, or if the network gets busy, the network discards the frame. The net-
work does not retransmit the frame (like X.25 does). This also eliminates overhead and redun-
dancy, improving throughput.
16.7.3 Network Structure and Operation
Frame Relay was designed for LAN to LAN client-server data communications between loca-
tions of a business or government.
Permanent Virtual Circuits identified with a sequence of DLCIs will established between every
two of the locations.
To communicate from one customer location to another, the customer must have edge equip-
ment that relates DLCIs to destinations, and packages the customer data into the same frame
format used by the service provider.
The customer-premise equipment that performs this function is called a Frame Relay Access
Device (FRAD). Typically, IP subnets, that is, blocks of IP addresses, will be assigned to each
customer location. The routing table in the FRAD is then populated with which DLCI to use to
get to a given IP subnet.
Since there is a possibility that frames will be discarded during transit, users must run an end-
to-end error-checking and retransmission protocol to implement reliability. TCP is normally
used.
16.7.4 No Guarantees for Voice
Frame Relay provides no guarantees as to end-to-end delay in the delivery of frames, and no
guarantees as to the maximum variability of delay, called jitter.
This means that while it is possible to communicate digitized speech in packets in frames over
a Frame Relay network, it is not possible to guarantee the quality of the reconstructed speech
in the case of a live telephone call.
A 20-ms-long segment of digitized speech arriving 300 ms after the previous segment usually
has the same effect as not arriving at all: reconstructed speech that has parts of syllables
missing and noticeable clicking noises.
A technology that can guarantee transmission characteristics such as delay and jitter is re-
quired to be able to guarantee the quality of delay- and loss-sensitive applications like tele-
phone calls and live television.

16.8 ATM
For a long time, Asynchronous Transfer Mode (ATM) was thought to be the answer to all of the
requirements for guaranteeing different transmission characteristics for voice, video and data
interspersed on the same circuit.
Unfortunately, ATM became very complicated and very expensive and is no longer used for
new deployments.
This section provides an overview of ATM and its jargon before moving on to its replacement,
MPLS in the next section.

242
 Figure 163. ATM
ATM is similar to X.25 and Frame Relay, transferring cells (instead of packets or frames) of in-
formation over virtual circuits. The difference is that ATM was supposed to be able to guaran-
tee suitable transmission characteristics for any type of traffic: telephone calls, television, busi-
ness data, web pages, e-mail or anything else, to achieve the goal of integration or conver-
gence, everything on one network.
16.8.1 Future-Proof Technology (Not)
Some even were claiming that ATM was future-proof, supporting switched virtual circuits for
flexibility, scalable to arbitrarily high line speeds and supporting any type of traffic. There would
be no technology after ATM.
Unfortunately, this did not happen. If it had, people would be discussing “Voice over ATM” in-
stead of “Voice over IP”. ATM was used on carrier networks to achieve integration of all data
services, but it was never deployed on the PSTN to carry delay-sensitive telephone calls.
The establishment and management of Switched Virtual Circuits in ATM is so complicated that
it was rarely implemented – most often, manually-configured Permanent Virtual Circuits were
used. ATM became so cumbersome and expensive that it is headed for the dustbin of history,
replaced by MPLS and Differentiated Services.
16.8.2 ATM Cells
ATM packages data into 53-byte packets (called cells to confuse the innocent), consisting of
48 bytes of data and a 5-byte header. Three of the bytes in the header are a virtual circuit ID
used to route the cell.
16.8.3 Service Classes
ATM implemented Quality of Service (QoS), allowing the specification of service classes for
virtual circuits. This allowed the integration of many services over the physical circuits that
make up a network (e.g. 10 Gb/s OC192) by establishing ATM virtual circuits between numer-
ous types of network and edge equipment across the physical circuits and assigning a service
class to each.
On an ATM virtual circuit, the user agrees to a traffic profile: the number of bits per second
steady-state, the maximum short-term burst rate, how long it can last and how often it can hap-
pen.
In exchange, the network guarantees transmission characteristics like delay, variability in de-
lay, number of errored bits and so forth.
ATM Service Classes were a set of standardized choices for traffic profile and transmission
characteristics:

243
 • Continuous Bit Rate for constant bit rate traffic with fixed timing, typically for full-period ser-
vice emulation.
• Variable Bit Rate - Real Time for variable bit rate traffic with fixed timing; for example,
phone calls and television. This was never deployed in practice on the PSTN, and here ATM
failed to meet the objective of service integration and convergence.
• Variable Bit Rate - Non-Real Time for variable bit rate traffic with no timing relationship be-
tween data samples, but requiring guaranteed average bandwidth. This was used to inte-
grate traffic for all data services, including Frame Relay and IP, on the core.
• Available Bit Rate (ABR): “best efforts” service, where flow control is used to increase and
decrease the capacity allowed to the user based on available network capacity. Designed to
transport LAN-LAN communications, which opportunistically use as much bandwidth as is
available from the network.
• Unspecified Bit Rate (UBR): no guarantees. The user is free to send any amount of data up
to a specified maximum while the network makes no guarantees at all on the cell loss rate,
delay, or delay variation that might be experienced.
When setting up a virtual circuit, ATM switches implement an algorithm called Connection Ad-
mission Control (CAC) to determine if it is possible to deliver a requested service class.
Using link parameters and end-to-end connection metrics, the switch determines whether ac-
cepting “just one more” connection would impact its ability to meet Service Class guarantees
for existing virtual circuits.
The network can enforce a traffic profile by traffic policing. The ATM switch will meter an in-
coming stream to confirm it is respecting the agreed traffic profile. A switch can either discard
out-of-profile cells or tag them by setting a Cell Loss Priority bit in the cell header, marking the
cell to be first to be discarded should there be congestion in the network.
ATM switches from companies like Nortel, Lucent and Cisco that implemented CAC and traffic
policing cost $800,000 or more in 1995 dollars each, plus yearly license fees and upgrades.

16.9 MPLS
IP is firmly established as the standard protocol for networking… but in itself, does not have
any way of implementing performance guarantees measured by characteristics like packet de-
livery rate and delay.
16.9.1 MPLS vs. TCP
TCP can deal with non-delivered packets, implementing communication between the source
and destination for delivery confirmations and retransmission of non-delivered data.
But TCP only retransmits lost data; it does not influence the packet delivery rate or end-to-end
delay, both of which are critical for telephone calls and live video over IP as well as business
data services.
To control packet delivery rate and delay, a traffic management system is required to manage
and prioritize flows of IP packets.
Multi-Protocol Label Switching (MPLS) is used for this purpose, providing network operators
with IP packet traffic management using virtual circuits.
MPLS concepts are the same as other virtual circuit technologies X.25, Frame Relay and ATM
covered in the preceding sections, and the general concepts of Section 16.5 “Virtual Circuit
Technologies”… but the jargon is changed.
16.9.2 Forwarding Equivalence Class
For “traffic class”, MPLS uses the term Forwarding Equivalence Class (FEC) to mean a group
of packets that are forwarded over the same path with the same forwarding treatment.
16.9.3 Labels
Instead of “virtual circuit ID”, labels are used to identify a FEC.
In IPv4, the label is typically contained in an MPLS Shim Header, which is four bytes of extra
overhead prepended to the IPv4 packet. Twenty bits are used for the label, three bits for exper-
imental functions, one bit to indicate “last label” in a stack, and eight bits for time to live.

244
In IPv6, the label can reside in the “Flow Label” field in the packet header defined for this pur-
pose.
A packet can have multiple labels, organized on a last-in, first-out basis, called the label stack.
This allows a hierarchy of FECs, and aggregation of traffic by type (e.g. telephone calls, televi-
sion, web pages, bit torrent) so all of the instances of a single type of traffic can be managed
as a single entity in the core. The processing is always based on the top label, regardless of
whether any others might be “below” it.

Figure 164. MPLS

16.9.4 Label-Switched Path

Instead of “virtual circuit”, Label Switched Path is the term used to describe a sequence of
routers that all work on a particular packet’s label at the same depth in the label stack.
This is similar to the notion of the route associated with a virtual circuit. LSPs can be internal to
a network. Many LSPs could end at the same egress router, to be forwarded on a common
outgoing LSP… or vice-versa.
LSP route selection is the definition of the actual path through equipment and over circuits that
an LSP will follow. In theory, this could be done on a hop-by-hop basis, with each LSR choos-
ing the route to the next hop, much the same as IP routing.
In practice, particularly commercial (phone company) networks, the route for an LSP is defined
by a system at a Network Operations Center (NOC), to facilitate network management func-
tions like load balancing and service restoration after a network fault.
16.9.5 Label Edge Routers
The ingress router, part of the Provider Edge function, as illustrated on the left of Figure 164 is
called a Label Edge Router (LER). The LER analyzes an incoming IP packet, determines what
Forwarding Equivalence Class it belongs to and labels the IP packet accordingly.
Since this analysis is only performed once at the ingress, the classification decision can take
into account factors not available to IP routing, such as the source port or VLAN ID, and can
be as complex as desired.
Subsequent devices – the LSRs – use only the label affixed by the LER to the IP packet for
forwarding decisions, not the IP address.
At the destination side of the MPLS network, a Label Edge Router performs an egress router
function, mainly removing the last label and its header from the packet before forwarding the
packet to the user destination.
16.9.6 IP User-Network Interface
MPLS is a traffic management system for IP packets implemented internally to the carrier net-
work, invisible to customers.
The interface between the user and the network is IP packets. MPLS labels are added to the
packet by carrier equipment at the entry to the carrier network, and removed by carrier equip-

245
ment at the exit from the carrier network, before the IP packet is delivered to the user. The user
never sees MPLS labels.
16.9.7 Label-Switching Routers
The routing devices internal to an MPLS network are Label Switching Routers (LSRs). These
devices use the value of the topmost label on a packet to look up the forwarding and possibly
prioritization instructions for the packet, then forward the packet.
Making the routing decision a table lookup rather than a complicated algorithm, minimizes de-
lay through the LSR and facilitates control of routing via an external system populating the con-
tents of the table.
In the LSR, the Incoming Label Map is the “lookup table”, indexed by label number. The Next
Hop Label Forwarding Entry is an entry in the Incoming Label Map that contains information on
forwarding a labeled packet: the next hop, what operation to perform on the label stack, and
can contain other information needed to properly forward the packet. There can be more than
one entry for a given label value.
The essential function of an LSR is label swapping. The LSR examines the label at the top of
the stack, and does a table lookup in the Incoming Label Map to get the Next Hop Label For-
warding Entry, then uses that information to encode a new label on the packet and forward it
on the appropriate outgoing link with the appropriate relative priority.
The labeled packet can be forwarded to the next LSR or LER over a data link running any kind
of layer 2 protocol, typically Ethernet.

16.10 MPLS VPN Service for Business Customers

The remainder of this chapter examines the practical uses to which MPLS is put by carriers.
One large part of carrier revenue is providing high-quality point-to-point communications be-
tween specific locations of a business, government or other organization. Large banks, to give
one example, have budgets measured in the tens of millions of dollars per year for this type of
service.
16.10.1 Private Network Service
In the 1970s, this type of service would be a private network service, consisting of multiple
dedicated point-to-point lines between different locations of the bank. Later, the “dedicated
lines” would be implemented as dedicated channels on a channelized TDM system.
In both cases, the bank’s communications is private: other customers of the carrier could nei-
ther see the bank’s traffic nor the bank’s sites, nor can the bank communicate to other cus-
tomers of the carrier over these services.
And in both cases, the carrier can sell a Service Level Agreement to the bank, guaranteeing
transmission quality and service availability for a price.
16.10.2 Virtual Private Network (VPN)
A Virtual Private Network (VPN) means that the private network service is not, in fact, imple-
mented with dedicated point-to-point connections – it just appears that way to the customer.
In reality, there are many users’ IP packets interspersed on the circuits, hidden from each other
using encryption and/or MPLS LSPs.
The term VPN is used to describe at least three different things in telecom. Two in current use
are Internet VPNs and MPLS VPNs.
16.10.3 Internet VPNs
Internet VPNs are secure point-to-point communications across the Internet. What appears to
the user to be a point-to-point dedicated line is implemented by encrypting packets on the
sending computer, transmitting them over the Internet to a particular pre-selected and authenti-
cated destination, then decrypting them at the far-end receiving computer.
The encryption and exchange of keys is specified in a set of protocols referred to as IPsec.
IPsec implements a secure VPN over the Internet – but there are guarantees of transmission
quality since no single carrier controls all of the Internet circuits over which the packets travel
end-to-end.

246
A popular application for Internet VPNs is working from home, accessing servers at work over
the Internet. Internet VPNs are covered in Section 17.9.
16.10.4 MPLS VPN
MPLS VPNs are a different story. In the case of a VPN service provided by a carrier to a bank,
the bank’s traffic is not sent over the Internet between branches, it is sent over the circuits of
the carrier that is selling the service to the bank.
That carrier does control all the circuits over which the bank’s traffic travels end-to-end.
The carrier uses its MPLS traffic management system to define label-switched paths between
the bank’s buildings, and associate a Class of Service with each LSP. The LSP acts like a tun-
nel, carrying the customer traffic end-to-end.
This allows the marketing and sales departments to sell banks and government reliable IP
packet communication services, backed up with the Service Level Agreement they require.
Multiple such point-to-point IP packet communication paths connected with routers at each
bank building effectively implements a private network: the bank can only communicate be-
tween the locations where the LSPs are set up, and the traffic moving over these LSPs is not
visible to any other customers of the carrier. Plus, the carrier can guarantee transmission qual-
ity by prioritizing traffic on the LSP.
In this case (and any other case) the bank would encrypt their traffic before giving it to the car-
rier. The general rule in the security business is “if it is not encrypted, it has been released to
the public”.

Figure 165. MPLS VPNs. A carrier defines MPLS Label-Switched Paths between customer locations in pairs. A
Class of Service is associated with each LSP to implement performance guaranteed to the customer.
This service will eventually replace all existing carrier “business customer” data services like
“dedicated T1s” and Frame Relay.

16.11 MPLS and Diff-Serv to Support Class of Service

Differentiated Services (DS) or Diff-Serv is an IP-based solution for prioritization, providing dif-
ferent transmission characteristics or Class of Service (CoS) for different types of traffic.
Diff-Serv provides a mechanism to classify packets as to the Class of Service (CoS) they
should experience, specification of transmission characteristics like delay and packet loss,

247
then give classified packets appropriate forwarding treatment in terms of prioritization at each
hop in a DS-compliant network.
16.11.1 DS Codepoints
Packets are classified at the ingress or boundary of a network supporting DS, associating the
packet with a DS codepoint, which is jargon to mean “Class of Service”, and at the packet
level, “relative priority”.
In a DS router, each DS class is associated with a Per-Hop Behavior (PHB), defining the for-
warding behavior, i.e. transmission characteristics desired for that class.

☞
This only becomes meaningful when there is congestion: contention for available pro-
   cessing and transmission resources.
Applying PHB criteria to DS classes assigns relative priorities to packets passing through a DS
router when contention occurs. The result is the ability to implement externally-observable CoS
in terms of bandwidth, delay, jitter and dropped packets.
16.11.2 Assured Forwarding and Expedited Forwarding
RFC2597 Assured Forwarding and RFC2598 Expedited Forwarding contain suggestions for
actual values for the DS codepoints, but appear to be largely academic exercises, defining
dozens of Classes of Service.
In practice, a carrier might implement three priority levels:
1. Telephone calls
2. Television programs and
3. Internet traffic.
A more sophisticated implementation might have eight priority levels, and so eight Classes of
Service, in order from highest to lowest:
1. Network control messages
2. Live telephone calls
3. Live streaming television programs
4. Live Internet web surfing
5. Video and music download
6. Email and other Internet traffic
7. Filler material like news headlines
8. Bit torrent peer-to-peer intellectual-property-stealing file-sharing traffic.
Since the classification is performed only at the input to a DS domain, the complex decision-
making process – deciding what QoS a packet should receive – is performed once. Subse-
quently, each DS router has a simpler decision-making process, based on actual traffic and
pre-assigned PHBs to determine relative priorities.

☞   This is very similar to MPLS labeling at the ingress to the network.
Considered separately, the MPLS label identifies the routing, and the DS codepoint identifies
the priority for a packet.
These two ideas can be combined by implementing multiple LSPs, all going from the same
place and to the same place, but each associated with a different DS priority level. Then, the
label on the packet is used by the router internal to the network to determine both the routing
and prioritization of the packet.

248
 Figure 166. Differentiated Services
An alternate implementation is to keep the MPLS label and Diff-Serv codepoint separate. In
this case, the 6-bit DS codepoint is populated in the Type of Service field in the IPv4 packet
header, or in the Traffic Class field in the IPv6 packet header, and the network routers would
process the label and codepoint separately.

16.12 MPLS for Integrated Access

MPLS labels and traffic classification can be used to combine all of the types of communica-
tions of a business or organization onto a single access circuit. This idea is sometimes called
convergence, though service integration is a more accurate term. It results in a large cost sav-
ings compared to one access circuit for each type of communications.
At each location, a typical organization would have the requirement to communicate
• Telephone calls to/from the PSTN,
• Telephone calls to/from other locations of the organization,
• Data to/from other locations of the organization, and
• Data, video and possibly voice to/from the Internet.

Figure 167. Before: Separate Access Circuits

As illustrated in Figure 167, in days past, the organization would have had four physical access
circuits and services – along with four bills:
• ISDN PRI over T1 to a LEC for telephone calls to/from the PSTN,

249
 • Tie lines or a voice VPN with a custom dialing plan from an IXC for telephone calls to/from
other locations of the organization,
• Dedicated T1s from an IXC for data to/from other locations of the organization, and
• DSL, Cable or T1 access from an ISP for data, video and possibly voice to/from the Inter-
net.
16.12.1 SIP Trunking, VPN and Internet on One Access
As illustrated in Figure 168, moving to an all-IP environment, these four circuits can be re-
placed with one bill for one 10 Mb/s to 10 Gb/s Optical Ethernet access circuit with three traffic
classes, each identified with their own label number.
The three traffic classes / labels would be:
• A traffic class for telephone calls. This might be called a “SIP trunking service” by the mar-
keting department. This virtual circuit will carry VoIP phone calls to/from the carrier for com-
munication either in native IP format to/from other locations, or conversion to/from traditional
telephony for phone calls to/from the PSTN.
• A traffic class for data. This might be called a “VPN service” by the marketing department.
This virtual circuit carries file transfers, client-server database communications and the like
securely to/from other locations of the organization.
• A traffic class for Internet traffic. This virtual circuit carries anything in IP packets to/from
the Internet.

Figure 168. After: Integrated Access - One Access Circuit, Separate Labels
All of this traffic is IP packets interspersed over the single access circuit. The way the traffic is
distinguished is by classifying it on a piece of carrier equipment at the customer premise, tradi-
tionally called an Integrated Access Device (IAD), which in this case classifies the packet then
stamps the appropriate label on each packet.
At the other end of the access circuit, the carrier uses the label to route the traffic onward and
to prioritize it to assure the appropriate service level.
The result is all of the organization’s traffic carried over a single access circuit, using a single
technology. This is the Holy Grail of the telecommunications business, called convergence or
service integration, having significant advantages in cost and flexibility.

16.13 MPLS for Traffic Aggregation

MPLS labels can be stacked. In other words, virtual circuits can be carried over other virtual
circuits… or in MPLS lingo, LSPs can be carried over other LSPs.
This is implemented to aggregate traffic so that the same kind of traffic can be managed as a
single entity. This happens both on integrated access circuits and in the network core.
16.13.1 Label Stacking

250
Consider an example of a bank with an MPLS VPN for data between major offices in New
York, DC and San Francisco. Each of these bank locations has a specific IP subnet, a unique
block of IP addresses. To emphasize the fact that these communications do not go over the In-
ternet, IP addresses in the private address space are used in the example.

Figure 169. Aggregating VPNs for Management - Label Stacking

For communication of data between the bank building in New York and the bank building in
DC, a LSP will be established by the carrier between the two bank locations, and associated
with a label number.
When the LER in New York, the ingress device, determines that an outbound IP packet con-
tains data for the bank location in DC by looking at the destination IP address and finding it to
be in the DC subnet, it will label the packet with the appropriate MPLS label for “Customer
VPN Data New York-DC”, the first label on Figure 170.
To transmit this over the integrated access, the LER will add a second label identifying it as be-
longing to traffic class “all VPN data on integrated access”.

Figure 170. Label-Stacking Protocol Headers:

MAC, MPLS Label, MPLS Label, IP Address
Other packets, like customer VPN data New York – San Francisco would also have the label
“all VPN data on integrated access” added.
Then, by referring to the topmost (leftmost in Figure 170), second label, all VPN data packets
on the integrated access can be managed by a control system as a single entity.
The packet at this point, with its destination IP address, the two labels, carried in a MAC frame
is illustrated in Figure 170.

251
The carrier does the same thing on the core.
When the packet arrives at the network end of the customer’s access circuit in New York, the
second label saying “all VPN data on integrated access” is removed as it is no longer meaning-
ful, and a new second label saying “all VPN data on core New York – DC” is added. Other
packets, labeled as VPN data from other customers, would also have this second label added.
The result of all having the same second label is all VPN data traffic New York - DC on the
core can be managed as a single entity: a single icon on a monitoring console at the NOC,
configured as a single Class of Service and single route in MPLS LSRs by the carrier.
At the other end, the core LSR in DC would remove the second label and replace it with a new
second label indicating data on the integrated customer access in DC.
The LER in DC, the egress device, would remove both labels and pass the packet to the cus-
tomer edge router at the bank building in DC for forwarding to the machine with private IP ad-
dress 192.168.102.8 in this example.

16.14 M is for Multiprotocol: Virtual Private LAN Ser-

vice (VPLS)
The “M” in MPLS stands for “Multiprotocol”.
In this chapter, we have referred to packets and the forwarding of labeled packets.
While this is the most common use of MPLS, forwarding of labeled frames is also possible.
This can be called a “layer 2 service”, as opposed to MPLS VPNs, which are layer 3 services.
Carrier Virtual Private LAN Service (VPLS) moves MAC frames point-to-point using MPLS la-
bels. From the customer point of view, the carrier appears to be a giant, nationwide LAN
switch, moving MAC frames between customer locations.
Very large network operators order this service as part of a carrier redundancy strategy. When
the customer pays different carriers to supply redundant connections between critical loca-
tions, the customer does not want any of the carriers involved in assigning blocks of IP ad-
dresses and routing IP packets, since the carriers support different blocks of addresses. The
customer manages the IP addresses, and has the carriers provide Layer 2 and Layer 1 ser-
vices.
This can be implemented with Ethernet over MPLS (EoMPLS), where the customer’s MAC
frame has MPLS labels pasted on the front of the frame.
LSRs do not look at anything except the label. Therefore, once a MAC frame has a label
pasted on the front, it is a block of data treated the same way as a labeled packet by LSRs in
the MPLS network… encapsulated in a MAC frame for forwarding over a physical circuit.
A more efficient implementation would insert the label value in the frame address field to avoid
duplicate framing with a labeled MAC frame carried in a MAC frame. That was the idea behind
Frame Relay, Section 16.7.

252
17
The Internet
17.1 A Network To Survive Nuclear War
The Internet has its roots in anarchy. It’s like some sort of fungus, spreading across the planet.
There are theories that it will become self-aware one day. The humans will try to unplug it, and
the network will retaliate by nuking the humans and hunting the survivors.

Figure 171. A Network to Survive a Nuclear War

One of our favorite urban myths is that The Internet started off as a research project funded by
the US Department of Defense (DOD) Defense Advanced Research Projects Agency (DARPA)
to develop network protocols that would be capable of surviving a nuclear war.
Whether that’s true or not (probably not), it’s a useful way of understanding the difference be-
tween IP and other strategies.
This research project began before the ISO Reference Model was established and before X.25
packet networks.
The philosophy for this work was markedly different than commercial datacom packet services.
In contrast to the telephone companies’ X.25, which provides solid, reliable connection-ori-
ented service, the DARPAnet design was based on the premise that communications might
take place over a number of intervening networks not controlled by the sender or receiver, and
that the links might be totally unreliable; circuits would fail, and loads upon remaining links
might become very high.
17.1.1 Unreliable, Connectionless Network Service
To be able to survive a nuclear war, and allow the exchange of data files, the core idea of the
DARPAnet was that it would provide a connectionless service, allowing the transfer of packets
from one computer to another to another across multiple links and intervening sub-networks,
few of which are under the control of the sender or receiver.
Since the intervening networks are not under control of any single organization, the ideas of
virtual circuits, fixed routes, reserved bandwidth and prioritization of the carrier services in
Chapter 16 can not be implemented by a user or any one carrier on all of the links end-to-end.
Therefore, no virtual circuits would be used, and no prioritization would be specified. Routing
decisions would be made in a decentralized way, routers communicating with their peers to de-
termine routes rather than being programmed by some carrier’s Network Operation Center.
Protocols for exchanging router table update messages between routers were developed. With
parallel and path-diverse redundant links, data could be re-routed around trouble spots. Parts
of the network could be terminated, and the rest would continue functioning.

253
17.1.2 Al Gore Invents the Internet
The Internet started out as federally-funded data links connecting universities and research in-
stitutes like UC Berkeley, UCLA and MIT as well as defense establishments and defense con-
tractors.
The links between these locations were paid for by the government, and the universities and
research institutes. One of the government institutes was the National Science Foundation
(NSF), which paid for some of the most expensive links.
Al Gore was apparently instrumental in getting funding for the NSF approved, and in a moment
of weakness publicly stated that he had “invented the Internet”, words he quickly regretted.
The DARPA net was renamed the ARPA net, and then the Internet.
Today, the Internet is no longer just a national security project, but has gone global, and will
soon replace the telephone and television networks.
What started off as a small group of technically advantaged researchers and computer buffs
has turned into a network accessed by billions around the planet.
17.1.3 Who Pays for the Internet?
The government no longer pays for the links.
Telephone companies, broadband carriers, fiber backbone companies and other commercial
organizations provide the network connectivity at their expense, then bill for access to it and
packet delivery over it. Bigger companies bill smaller companies and users in a giant pyramid
scheme.
The end users end up paying for the Internet through monthly fees to their immediate service
provider for access and packet delivery.
17.1.4 Primitive Beginnings
The initial implementation of the DARPAnet was based on UNIX computer communications de-
veloped by UNIX computer programmers.
The DARPAnet had a Human-Machine Interface (HMI) developed by the computer program-
mers, which was more or less useless to the general public. It took very specialized knowledge
– and access – to use the network to ask people in Toledo what the weather was like there.
Luckily, this has been fixed with the Web and browsers, giving us a point-and-click, tap or
swipe Graphical User Interface (GUI).
Some things are taking longer to fix. In the beginning, UNIX only handled seven-bit bytes: 128
ASCII characters or a seven-bit number. The Internet email system, using unix commands,
could only transfer 7-bit bytes.
To be able to transfer an image made of eight-bit bytes (called “binary”), users or Internet email
programs had to transform the image into a file of 7-bit groups which could be coded as “text”,
then transform it back into 8-bit bytes at the receiver. This still happens to this day, but is auto-
mated.
The Internet was designed for data communications. Short, 7-bit ASCII text messages. Text
messages are not sensitive to delay and non-delivered packets can be retransmitted.
There were no mechanisms in the design to implement a guaranteed Class of Service... which
is necessary for packet loss and delay-sensitive applications like phone calls and streaming
video.
The addition of MPLS as a traffic management system internal to the network and invisible to
users provides this mechanism.

17.2 The Inter-Net Protocol

The “Internet” was originally the “Inter-net”, connecting diverse existing local packet data net-
works at universities and research institutes like UC Berkeley, UCLA and MIT with data cir-
cuits.

254
 Figure 172. Need for an Inter-Net Protocol

17.2.1 Gateways
The devices that connected the local networks to the data circuits were called External Gate-
ways.
In general, gateway means protocol converter. These devices were called gateways because
they converted between different packet and network address protocols used at the different
universities and research institutes.
Today, these devices are called routers, and they do not convert packet formats or network ad-
dresses. The IP packet format and addressing scheme took over the world.
17.2.2 IP: Common Packet Format and Address Scheme
To send a message to a user on a network, it is necessary to know their network address. If
the destination is on a different network, there may be no method for finding out what their ad-
dress is, nor any way of actually transmitting a packet with an address in a different network’s
format from the source computer.
The Internet Protocol (IP) was the solution to this problem: network addressing and packet for-
mat that would be used by all parties.
To send a message to another user, the message is segmented into chunks and the IP ad-
dress of the destination is added to form packets. These IP packets were then transmitted from
one network to another, originally carried inside whatever packets were being used at each in-
stitution.
Subsequently the IP took over and all of the institutions adopted IP as the native packet for-
mat.
17.2.3 Connectionless
Each IP packet is treated by an IP router as being completely independent from any other.
Packets might follow different routes across various different networks and experience different
delivery delays depending on changing congestion conditions.
17.2.4 Unreliable
There were and are no guarantees on the Internet. Data can be corrupted, copied, or thrown
away. Bombs might fall. Routers and links might turn into fused glass. The network does not
provide any information on the status of a packet in transit.
17.2.5 Need for TCP
Strong end-to-end error checking implemented by the users is required to check to see if data
arrived, and if not, retransmit it.
This end-to-end error checking is implemented by users running the Transmission Control Pro-
tocol (TCP), which employs sequence numbers, error checks, source timers and positive ac-
knowledgments to implement reliability on the unreliable IP network.

255
17.2.6 Routing and ASes
A group of interconnected routers controlled by the same organization is called an Autono-
mous System (AS).
Routers require tables with values identifying what address to go to in order to get to a device
with an address within a particular range or subnet.
The values in the routing tables can be entered manually and/or learned from other routers.
The Routing Information Protocol (RIP) was first used to exchange routing tables between ad-
jacent routers in an AS every 60 seconds.
RIP was replaced with Open Shortest Path First (OSPF), which exchanges update messages
between all routers in an AS, but only when a change occurs.
The Border Gateway Protocol (BGP) is used to define routing between different Autonomous
Systems.

17.3 Internet Service Providers

17.3.1 The Internet is a Business
The Internet is a vast, loosely-regulated collection of interconnected Autonomous Systems.
The connections between ASs are not specified by a central authority or world government,
but are implemented on a case-by-case basis by the operators of an AS for business reasons.
The Internet is not free. It is not a public utility. It is a business.
17.3.2 ISPs
Internet access providers, called Internet Service Providers (ISPs) provide the capability to ac-
cess and transmit and receive IP packets over the Internet.
ISPs are for the most part business units of facilities-based carriers, i.e. telephone companies
and cable companies.
These ISPs install a physical access circuit to the customer premise and provide the Customer
Edge (CE) device.

Figure 173. Internet Service Providers

The ISP also re quires security and access control equipment to manage customer traffic, and
routers to connect customers to networks. An ISP is an Autonomous System.
17.3.3 Interconnect, Peering and Transit
ISPs will connect to competitors and content providers like Google to exchange traffic terminat-
ing on each other’s network (called peering), and will connect to larger organizations who will
assure delivery of packets to other destinations (transit).

256
The networks are physically connected at Internet Exchange (IX) centers such as Equinix
Chicago at 350 E Cermak. These are buildings with equipment implementing network intercon-
nection operated by a neutral third party. The ASs are responsible for paying for connectivity to
the IX.
Peering is settlement-free, i.e. no money is exchanged. Packets are exchanged and forwarded
on a best-efforts basis.
Transit, assuring the delivery of packets, is a commercial service. Larger ISPs charge smaller
ISPs for transit services.
The largest networks are sometimes called Tier-1 service providers. “Tier-1” is not an officially-
defined term. A “Tier-1 network” might best be thought of as one operated by a large facilities-
based carrier that has a presence in most or all IXs on the planet.
Virtually all networks employ a mix of peering and transit agreements to connect to other net-
works. The exact nature of such connections is non-disclosed confidential business informa-
tion.
The ISPs build the access network and peering or transit connections to other networks, then
charge the users for access. It’s a pyramid scheme. The end users end up paying for all.
In addition to access services, the ISP provides a DHCP server to lend you an IP address valid
for use on the Internet, a Domain Name Server, and auxiliary services like e-mail servers and
web hosting.
17.3.4 Resellers
In the Flintstones era, when dial-up Internet access was first available, telcos were a bit slow to
react, so for a while, companies like Netcom, MindSpring, Portal, Pipeline, iStar and others
had their day in the sun.
These organizations were resellers, leasing circuits from a carrier and reselling them to users
under per-minute or per-month billing plans.
The carriers eventually began competing with resellers, who mostly went out of business, sell-
ing their customers to the carriers. For example, Netcom is now part of Earthlink, which is ma-
jority owned by Sprint.
For the most part, it is business units of the companies that own the cables coming into your
home – the LEC and the cable TV company – along with wireless carriers that are the ISPs to-
day.
Current reseller-type ISPs require a LEC or cable company to provide and install the physical
and network connection. They are essentially buying large blocks of carrier services at volume
discount rates, and reselling the carrier service at retail prices.
Deciding whether to use a reseller-type ISP involves evaluating the level and rapidity of cus-
tomer service as well as Class of Service guarantees versus any cost savings compared to
buying directly from the carrier.

17.4 World Wide Web

17.4.1 Clients and Web Servers
The World Wide Web is client-server computing over the Internet. Web servers store files
called web pages that contain text, formatting instructions and instructions to transfer other
files containing graphics, audio and video that are also stored on the web server.
The user needs client software called a browser to ask for the web page file, download it, and
perform local processing to create a graphical display.

257
 Figure 174. The Web: Client-Server Computing: fishing net, spider web... many links tied together.
HyperText Markup Language (HTML) is the main standard for formatting web page files.
The browser client does not communicate with the web server for more than the download of a
one file at a time. The user or the browser implements navigating to other pages, page refresh
and session restoration.
17.4.2 Hyperlinks and URLs
Some of the text on the display will be underlined, indicating links. Associated with the text that
is underlined is a Uniform Resource Locator (URL), which identifies the file to download, along
with the name of the server it lives on and the protocol to use to get it.
Tapping or clicking the underlined text causes the browser to look in the page code for the ref-
erenced URL and go there to download the indicated files and display a new page.
These protocols have changed the Human-Machine Interfaces to the Internet from difficult key-
stroke-based and command-line interfaces to a point-and-click Graphical User Interface (GUI).
The unpleasant details of network addresses, file names and file transfers are mostly hidden
from users.
The default file that is transferred from a server if no file name is specified is called the home
page or index page, and has by default the name index.html. There is no difference between
an index page and any other page.

17.5 Domain Name System

Web browsing is client-server communications, which is set up so that it is always the client
browser that initiates communications with the web server. The server listens for requests like
GET and POST from the client.
For the client to be able to communicate with the server, the client must know the server’s IP
address. This implies that the browser must be able to find out the numerical IP address of a
heretofore unknown server.
The server IP addresses will be stored in a data table on a server somewhere... the remaining
question is how will the servers be identified, to be able to look up their IP address in the ta-
ble?
The solution is a human-readable and human-rememberable identifier for the server called a
domain name.
The Domain Name System (DNS) is the structure allowing a client to determine a server’s IP
address for the first time.
17.5.1 Domain Zone Files
A Domain Name Server has a table with a zone file for every domain.

258
 Figure 175. Domain Zone File entries
The domain zone file is a text file of records with the numeric IP address for the domain’s web
server, its email server, subdomains and others.
The Start of Authority (SOA) record indicates what domain this zone file is for. Address (A)
records match a domain name to an IP address. There are many other kinds of records includ-
ing Mail Exchange (MX) and text (TXT)records for application-specific information like the
server’s public key.
17.5.2 Name Resolution
When a person wants to access a server, they use an application-level program – typically a
browser – that provides the user interface and contains a web client, a DNS client and a re-
solver.
The user can learn the domain name of the desired server by reading it in an advertisement,
then typing it in the address bar of the browser. The user might also learn the domain name of
the server from a search engine like Google. The domain name is a major component of an
URL.
The IP address corresponding to that domain name is determined by a small program in the
browser called a resolver. The resolver sends a request to a Domain Name Server inquiring
what the binary IP address corresponding to the domain name is.
The numeric IP address of the server is returned as the response to the query, whereupon the
browser can start sending packets to the server at that IP address, with a message requesting
transfer of the desired file.

Figure 176. Domain Name Resolution

If the Domain Name Server does not have an entry for a domain, it can ask another Name
Server either up or down in the hierarchy for resolution. This is called recursion.

17.6 Hypertext
17.6.1 HTML
The language used in the files transferred for the World Wide Web is called HyperText Markup
Language (HTML).
As illustrated in the example of Figure 177, an HTML file is plain text, and includes both the
text to display and embedded formatting commands, similar to old word processing systems.
The browser interprets the file to produce the page displayed on the screen.
HTML files contain hypertext references (HREFs), which specify the name and location of
other files, in a standard format called a Uniform Resource Locator (URL) or Uniform Resource

259
Identifier (URI).
The URL describes the protocol which should be used to transfer this other file, the network
address, and the file name and type.
Along with the page text, HTML files can also contain directives to download and display im-
age files, video, sound and other media.
The HTML from Figure 177 when interpreted by the browser yields the page displayed to the
user in Figure 178. The HTML file has a separate entry for each of the images shown, and
each one has an associated HREF causing the browser to download and display a different
page when the image is clicked in the browser.

Figure 177. HTML For Web Page

17.6.2 HTTP
The Hypertext Transfer Protocol (HTTP) includes commands that a client browser can send to
a web server. The most popular command is GET.
For example, “GET /online-courses-previews.htm HTTP/1.1” is an actual command sent to the
server of the page in Figure 178. Presumably, this command was generated by the browser
when the user clicked text saying “free previews” that had that URL in an HTML HREF tag.
The server would return an HTTP response message that included the indicated file. Other
HTTP response messages include the familiar 404 Not Found and 500 Internal Server Error.
There are many others.
HTML is now ubiquitous, used for defining the screens in many Graphical User Interfaces
(GUIs).

260
 Figure 178. Web page

17.6.3 SSL: Secure Socket Layer and HTTPS:

Typing HTTPS:// at the beginning of an URL in the address box of the browser – or clicking on
an HREF with this attribute – tells the browser to use the Secure Socket Layer (SSL) protocol,
which causes the contents of the packets to be encrypted on the client and decrypted on the
server.
This would prevent an eavesdropper from understanding the message being transmitted,
which might contain sensitive information such as a credit card number or password.

17.7 MIME and Base-64 Encoding for Email Attach-

ments
The basic email sending program used on most mail servers, sendmail, was developed by
UNIX computer programmers for transfers of text files using the UNIX to UNIX Copy Protocol
(UUCP), which historically only supported 7-bit printable ASCII characters.
This allows the transmission of plain text in emails.
However, mostly anything else, like images and spreadsheets are coded using all eight bits for
data in each byte.
17.7.1 Binary, Text and uuencode
To send an image in an email, it was necessary for the sender to re-package the 8-bit bytes of
the image (which people called binary) into at most seven-bit groups that could be displayed
as printable characters (called text).
The result would then be emailed just like plain text, and the receiver had to perform the re-
verse packaging to re-create the original image file.
In the beginning, users did this using a utility program called UUENCODE.

261
17.7.2 MIME
Now this function is automated: the sender’s e-mail program automatically transforms any “bi-
nary” elements of the message into “text” and tags the result with header information the re-
ceiver requires to perform the reverse transformation.
This is described in detail in RFC 2045: Multipurpose Internet Mail Extensions (MIME). MIME
defines a number of header fields that are used to describe the content of a message.
The Content-Type header field specifies the nature of the data in the body of an entity by giv-
ing media type and subtype identifiers.
17.7.3 Quoted-Printable
Two transforms are often used: quoted-printable encoding, and base-64 encoding. Quoted-
Printable Content-Transfer-Encoding is intended to represent data that largely consists of
octets that correspond to printable characters in US-ASCII. Octets with decimal values of 33
through 60 inclusive, and 62 through 126, inclusive, may be represented as the US-ASCII
characters that correspond to those octets. Any octet, except a CR or LF may be represented
by an “=” followed by a two digit hexadecimal representation of the octet’s value.
17.7.4 Base64 Encoding
The base64 transformation is premised on the fact that 3 x 8 = 4 x 6. As illustrated in Figure
179, three octets of the image file are re-packaged into four groups of six bits each.
Since 26 = 64, any combination of 1s and 0s in each of the groups can be represented by one
of 64 printable ASCII characters, hence the name of the transformation.
The result can then be transmitted as if it were plain text. In fact, as if it were a telex message
from the 1970s. Telex was a service that transmitted short text messages, the improvement on
telegrams.

Figure 179. Base64 Transformation

The tragic part is that it is no longer necessary. The operating system that the mail server is
running on and the mail server both use 8-bit bytes! But to be backwards-compatible with
1970s-era computer technology, email programs like still Outlook automatically apply this
transformation.
The result in practice is that the image file expands in size by 33%.
Following is an example of a calendar event invitation after being base64 encoded and with
message headers added:
--f46d044787ffe35c5304c1b9fae7
Content-Type: application/ics; name=”invite.ics”
Content-Disposition: attachment; filename=”invite.ics”
Content-Transfer-Encoding: base64
QkVHSU46VkNBTEVOREFSDQpQUk9ESUQ6LS8vR29vZ2xlIEluYy8vR29vZ2xlIENhbGVuZG
FyIDcwLjkwNTQvL0VODQpWRVJTSU9OOjIuMA0KQ0FMU0NBTEU6R1JFR09SSUFODQpN
RVRIT0Q6UkVQTFkNCkJFR0lOOlZFVkVOVA0KRFRTVEFSVDtWQUxVRT1EQVRFOjIwMTI
wNjE2DQpEVEVORDtWQUxVRT1EQVRFOjIwMTIwNjE4DQpEVFNUQU1QOjIwMTIwNjA1VD
EzNTM0OVoNCk9SR0FOSVpFUjptYWlsdG86ZXJpY0B0ZXJhY29tdHJhaW5pbmcuY29tDQp
VSUQ6MDQwMDAwMDA4MjAwRTAwMDc0QzVCNzEwMUE4MkUwMDgwMDAwMDAwMDM

262
wQjU1MjBERjQ0MkNEMDEwMDAwMDAwMDAwMDAwMDANCiAwMTAwMDAwMDA1NEUy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-
fip+KlxuXG5cbg0KTEFTVC1NT0RJRklFRDoyMDEyMDYwNVQxMzUzNDhaDQpMT0NBVElP
TjoNClNFUVVFTkNFOjANClNUQVRVUzpDT05GSVJNRUQNClNVTU1BUlk6SXNsYW5kDQp
UUkFOU1A6VFJBTlNQQVJFTlQNCkJFR0lOOlZBTEFSTQ0KQUNUSU9OOkRJU1BMQVkNC
kRFU0NSSVBUSU9OOlRoaXMgaXMgYW4gZXZlbnQgcmVtaW5kZXINClRSSUdHRVI7VkFM
VUU9REFURS1USU1FOjIwMTIwNjE1VDE2NDUwMFoNCkVORDpWQUxBUk0NCkVORDpW
RVZFTlQNCkVORDpWQ0FMRU5EQVINCg==
--f46d044787ffe35c5304c1b9fae7--
Running that “text” through a base64 decoder yields the application-level calendar invite mes-
sage interpreted as 8-bit bytes:
BEGIN:VCALENDAR
PRODID:-//Google Inc//Google Calendar 70.9054//EN
VERSION:2.0
CALSCALE:GREGORIAN
METHOD:REPLY
BEGIN:VEVENT
DTSTART;VALUE=DATE:20220616
DTEND;VALUE=DATE:20220618
DTSTAMP:20120605T135349Z
ORGANIZER:mailto:[email protected]
UID:040000008200E00074C5B7101A82E0080000000030B5520D-
F442CD01000000000000000
01000000054E223AC47E23943B3D633C2D1951BB9
ATTENDEE;CUTYPE=INDIVIDUAL;ROLE=REQ-PARTICIPANT;PARTSTAT=AC-
CEPTED;CN=Eric V
anasse;X-NUM-GUESTS=0:mailto:[email protected]
CREATED:20120605T122103Z
DESCRIPTION:When: Saturday\, June 16\, 2022 12:00 AM to Monday\, June 18\,
2022 12:00 AM (UTC-05:00) Eastern Time (US & Canada).\n\nNote: The GMT offset above
does not reflect daylight saving time adjustments.\n\n*~*~*~*~*~*~
*~*~*~*\n\n\n
LAST-MODIFIED:20120605T135348Z
LOCATION:
SEQUENCE:0
STATUS:CONFIRMED
SUMMARY:Island
TRANSP:TRANSPARENT
BEGIN:VALARM
ACTION:DISPLAY
DESCRIPTION:This is an event reminder
TRIGGER;VALUE=DATE-TIME:20120615T164500Z
END:VALARM
END:VEVENT
END:VCALENDAR

17.8 Internet Telephony & VSPs

Virtually all carriers convert telephone calls to Voice over IP (VoIP) for transmission over their
network, converting it back to POTS at the far end.

263
This section covers VoIP phone calls where at least one end has an Internet connection in-
stead of a POTS line.
17.8.1 VoIP Service Provider (VSP)
With “net to phone” services, packetized voice travels over the Internet to a gateway, which
converts VoIP to regular telephony and connects it an Inter-Exchange Carrier or directly to a
Local Exchange Carrier, eventually connecting the VoIP call to someone with a regular phone
line over the LEC’s access network.
This type of telephone service provider could be called a VoIP Service Provider (VSP). Vonage
is currently the most visible player, charging a monthly fee. Google provides the service for
free in their messaging app.

Figure 180. VSPs and Internet-PSTN Calling

In the example illustrated in Figure 180, the user arranges and pays for high-speed Internet ac-
cess. The VSP provides a softswitch for call setup and management, a media server to sup-
port web-based voicemail, a gateway to convert between VoIP and traditional channelized DS0
telephony, and connections to LECs. The LECs connect the call through their access network
to the “far end” telephone.
VSPs provide a softswitch to relate PSTN telephone numbers to IP addresses. For an inbound
call to the VSP customer, the softswitch allows the originating gateway to find out the IP ad-
dress of the customer assigned a particular PSTN phone number and establish a VoIP phone
call.
For an outbound phone call from the VSP customer, the softswitch sets up the connection from
the customer to a gateway that coverts VoIP to channelized telephony for connections out-
bound to the PSTN.
17.8.2 Internet - PSTN Connection
At present, the industry standard for connecting telephone calls between carriers is 64 kb/s
DS0 channels and ISDN signaling, not VoIP.
To terminate a phone call on a user with a POTS line, an IXC must have physical connections
to their LEC at their toll center. The IXC pays the LEC per month for each DS0 connection, and
per minute during each phone call, to compensate the LEC for the use of their wires.
At present, a VSP has to connect the same way, hence the need for a gateway converting the
VoIP to 64 kb/s DS0 and SIP to ISDN call setup.
As illustrated in Figure 180, a VSP could immediately open up shop by purchasing gateways,
installing them in toll centers, signing interconnection agreements with LECs and connecting to
the LEC via DS0s like any other IXC.
However, it forces the VSP to purchase and deploy gateways and pay for expensive DS0 inter-
connections and switched access charges to terminate their customers’ phone calls.
The VSPs would prefer that the interconnection be IP, not DS0, and the LEC should accept na-
tive VoIP calls over the Internet and terminate the call on their POTS customer, preferably for

264
free.
Eventually, there will be industry standards and formal agreements (called tariffs) for exchang-
ing VoIP packets and using SIP call setup.
This may take some time, as the extensive existing reliable, proven infrastructure and industry
standard practices to support this must be completely renewed for VoIP.
Elements that have to be agreed on are standards for voice coding; standards for call signal-
ing, setup and termination; agreements on what network addresses are reachable; standards
for call quality and how it will be measured; standards defining when a call begins and ends;
standard methods of rating and billing calls… to name a few.
17.8.3 Adapters
Only a small portion of residential customers will have a computer with a headset for making
native VoIP phone calls. The solution for the VSP is to supply an adapter to connect the cus-
tomer’s POTS phones to the customer’s Internet connection.
This adapter is a gateway, converting between POTS at the customer premise to VoIP for the
Internet. It includes computer functions: LAN and IP network protocols, analog-digital conver-
sion, speech coding and VoIP protocols, plus provides jacks to plug in regular telephones.
17.8.4 Cost Savings
VSP phone service is lower cost than POTS for several reasons:
a) The switched access charges added by the LEC to any long-distance phone call are
avoided – at the originating end in the example illustrated.
b) This telephone service occurs over a data service (the Internet connection), not over a
POTS line, so avoids regulation and fees imposed on POTS.
c) There are no “long distance” charges to locations the VSP serves.

17.9 Internet VPNs

The term Virtual Private Network (VPN) is used to mean at least three different things in the
telecom business.
An expensive VPN connecting locations of a bank or government or other large organization is
usually implemented by a carrier using MPLS.
As described in Section 16.10.4, MPLS VPNs are implemented with label-switched paths be-
tween the organization’s locations over the carrier’s carefully-controlled network, using virtual
circuits for the dedicated lines. That kind of VPN service is backed up by a Service Level
Agreement.
In this section, we cover the use of the term in conjunction with Internet communications.
17.9.1 Virtual Point-to-Point Connections
The “private” in private network means that it appears to the user that they have dedicated
point-to-point connections between their locations – other users can neither see their traffic nor
their sites, nor can the user see sites not their own.
“Virtual” means that there are not, in fact, dedicated point-to-point connections; there are many
users’ packets interspersed on the network, hidden from each other. It just appears to the user
there are dedicated lines.
17.9.2 IPsec and Tunnels
An Internet VPN uses customer-premise VPN equipment connected in pairs over the Internet
to implement secure communications over the unreliable Internet.
In this case, the VPN hardware authenticates each other over the Internet using public key en-
cryption, then exchanges a private key for bulk encryption of subsequent transmitted informa-
tion. Standard protocols for these functions are a set of five RFCs referred to collectively as
IPsec.

265
 Figure 181. Internet VPN using IPsec
As illustrated in Figure 181, any IP packets leaving the near-end secure network and destined
for the far-end secure network are intercepted by the near-end VPN hardware, encrypted, then
packaged or encapsulated inside another IP packet that is addressed to the far-end VPN hard-
ware and transmitted across the non-secure IP network.
The far-end VPN hardware receives these packets and decrypts the contents, extracting the
original packet and routing it on its way in the far-end private system. This encrypt-transmit-de-
crypt process is referred to as tunneling, and the point-to-point communications called a tun-
nel.
17.9.3 Hardware
The IPsec protocols can be implemented on special-purpose stand-alone hardware (“VPN
hardware”) to serve multiple users, to increase speed, and implement a higher level of security
compared to running the protocols on a shared Windows-based computer. This would be a
typical choice at a customer’s building.
For people working from home or on the road, the IPsec protocols are of course implemented
in software that runs on a PC or laptop. In this case, the PC is the “VPN hardware”.
This is slightly less secure than dedicated-purpose hardware, as the PC runs on the Windows
operating system that may have security weaknesses, the IPsec software shares memory with
other applications, which might be exploited to defeat the IPsec software, and the security is
dependent on the individual user to ensure proper configuration.

266
Appendix A
T1
This appendix provides detailed information on the carrier system technology called T1. These
discussions used to be a principal part of telecommunications courses, but are now relegated
to the back of the book, as T1 is a copper-wire technology running at 1.5 Mb/s.
That said, there are thousands of T1 circuits installed and in use, and some readers of this ref-
erence book have picked it up precisely to learn about T1 because they have been tasked with
supporting it, or auditing an existing installation.
Power companies, the military and government still have T1s in place, along with T1s at big or-
ganizations that are no longer being used, but have been forgotten and are still being paid for
each month... a small but appreciated part of any phone company’s revenue.
Another remaining use of this chapter on T1 is a detailed explanation of synchronous Time-Di-
vision Multiplexing, framing and channels.
Since the principles of operation of legacy SONET fiber-optic transmission systems are the
same as T1, learning about T1 is also learning about SONET.
Chapter 5, “Digital” should be read before this one.

☞
If you do not need to know about T1 or channelized time-division multiplexing, feel free
   to skip this Appendix.

A.1 T1 History and Applications

The T1 Carrier System was designed in 1958 by Bell Labs. The main requirement for this sys-
tem was to increase the circuit density on existing copper wiring between Central Offices… to
increase the density of phone calls and thus revenue on existing copper long-distance circuits.
In the old days, this might have been called a pair gain system, because it increases the num-
ber of circuits actually carried on each pair of wires.
T1 carries 24 digitized voice signals on a single set of copper wires. It was originally designed
and deployed by AT&T for use by AT&T for long-distance voice calls.
T1 was popular in the previous millennium as an access circuit installed from the customer
premise to the service provider. It implemented “high speed” data access at 1.5 Mb/s or 24
DS0 channels to carriers’ voice and data services for business, government and other organi-
zations
T1 was used for
• “dedicated T1s”, a DS1-rate (1.5 Mb/s) point-to-point connection across a carrier’s network
all the time,
• PBX trunks and ISDN PRI service, carrying circuit-switched connections to a LEC and/or
IXC for the duration of a communication session, and
• Frame Relay, VPN and Internet data service connections at 1.5 Mb/s.
Today, T1 is not used much at all. Optical Ethernet has taken its place.

A.2 T1 Circuit Components

As illustrated in Figure 182, a basic T1 system consists of multiplexers, Channel Service Units
(CSUs) and the T1 circuit, which is four copper wires with repeaters every mile or so.
The multiplexer (one at each end) is variously referred to as a T1 multiplexer, a mux, or a
channel bank. Since T1 was designed to carry 24 trunks over 4 copper wires, the mux has 24

267
hardware ports, each one running at 64 kb/s.
The multiplexer’s aggregate or high-speed output connects to a CSU.
The CSU is the interface device connecting the T1 multiplexer and the actual T1 circuit. This
device is the one that represents binary digits on the physical T1 circuit. It performs the same
functions as a modem – but since it is a digital device, it is not called a modem (Section 11.6).
The T1 circuit is four copper wires, two for each direction.
Binary digits are represented on these copper wires using pulses of voltage following a line
code called AMI, covered in Section A.5.
Repeaters are spaced every 6 kft (1 mile / 1.6 km) along the T1 circuit.

Figure 182. T1 Circuit Components

The data rate on the T1 circuit is 1.544 Mb/s, which is a DS1-rate signal. 1.544 Mb/s includes
extra bits added in by the multiplexers for framing.

A.3 Operation
A T1 system works in a strict, ordered rotation, where each user transmits one byte at a time.
First, the user attached to hardware port 1 on the multiplexer gets to use the outgoing high-
speed aggregate circuit to transmit a byte to the corresponding port 1 output at the other end.
Then, port 2 sends a byte, then port 3, and so on down the line in strict order, until port 24
sends a byte.
Then, one extra bit called the framing bit is transmitted, marking the end of the batting order,
and the process repeats: port 1 sends a byte, port 2 sends a byte and so on. The process re-
peats 8,000 times per second.
The bytes from each port are interspersed or interleaved on the T1, and so it is called a byte-
interleaved system.
At the far end, the high-speed aggregate circuit is plugged into a demultiplexer, which directs
each byte to the correct output hardware port one at a time.
The entire system is two-way simultaneous: both directions at the same time. When we say
“input”, we should really say “input and output”… but it is easier to discuss it one direction at a
time.
The end result is to communicate 24 DS0s, that is, 24 64 kb/s channels in both directions at
the same time over four copper wires.

268
A.4 T1 Framing
To ensure that each user ends up with a fixed fraction of the capacity of the high-speed circuit,
a channel, the users transmit one byte at a time, one after another in a strict order as illus-
trated at the top of Figure 183.
After each cycle, a framing bit is transmitted. The framing bit is used at the far end to locate
each user’s bytes in the incoming bit stream, to direct each byte to the correct output.
A byte from each port plus a framing bit makes up a T1 frame.
A T1 frame is 24 channels x 1 byte/channel x 8 bits/byte + 1 framing bit = 193 bits long.
Frames are transmitted 8,000 times per second.

A.4.1 Superframe Format

Every 193rd bit coming down the line is a framing bit, sent to mark the beginning of the frame…
but what good is sending one framing bit per frame?
After all, a single bit could be either a zero or a one; it looks just like all the other bits… is it not
necessary to have a unique pattern of bits that can be recognized at the far end instead of one
bit?
Yes, a pattern of framing bits is required; but instead of putting the pattern as a clump at the
beginning of the frame, which would increase the overhead, the designers were clever, and
considered a number of frames in a row to be a larger unit called a superframe.
This way, there is not just one framing bit, there is a group of bits. But they are not in a clump:
they are spread out, one framing bit at the beginning of each of the frames in the larger group
of frames.
The original design took groups of 12 T1 frames together to make a superframe. Since there
are 12 frames, there are 12 framing bits (spread out, one at the beginning of each of the 12
frames).

Figure 183. Framing and Superframe Format

Now that there are 12 framing bits, we use a framing pattern defined by Bell Labs, 12 bits long,
and insert it where the framing bits occur, one bit from the pattern every 193rd bit in the data
stream.
At the far end, when first powered up, the receiving multiplexer brings in 12 frames’ worth of
data and looks every 193rd bit to see if it sees the framing pattern.
Chances are 1:193 that it is not looking in the right place, so it shifts over one bit and looks
again, and keeps shifting over until it sees that specific pattern sitting in the incoming data
stream every 193rd bit.

269
The receiving multiplexer has now found the framing bits, and frame synchronization is
achieved.
The receiving multiplexer now knows which byte goes to which output, and at this moment, all
of the data on the outputs becomes valid.
This process usually happens once, when the circuit is turned up.
This idea is referred to in general as framing, and this technique is called Superframe Format
or D4 format in particular after a type of AT&T equipment.

A.4.2 ESF
An improvement called Extended Superframe Format (ESF) or D5 format was made to the
original design, for more efficient use of the framing bits.
The rationale behind Extended Superframe Format is that it was not really necessary to have
one bit per frame = 8 kb/s for framing, so only some of those framing bit positions would actu-
ally be used for framing, and the rest could be used for other functions, like error checking and
reporting.
ESF groups T1 frames in groups of 24, and uses 2 kb/s for framing, 2 kb/s to perform a CRC-6
(Section 12.5.6) on each frame and provide a 4 kb/s free data channel between the multiplex-
ers.
This data channel is often referred to as the Facility Data Link (FDL).
It is used to report the results of the CRC check and other performance parameters down the
line. AT&T Technical Publication 54016 and ANSI T1.403 are standards for use of the FDL.

A.5 Pulses and Line Code: AMI

Here, we examine how the binary digits actually represented on the copper wires. Even though
this is a “digital” system, we still need a method of representing the binary digits (1s and 0s) on
the physical medium, which in the case of T1 are pairs of copper wires.
Digital transmission means applying energy to the communication circuit, or not, for a pre-de-
termined length of time to represent 1s or 0s.
The burst of energy is called a pulse, and the strategy for representing 1s and 0s using pulses
employed by a particular technology is called its digital line code.
The line code used for T1 circuits is called Alternate Mark Inversion (AMI).

Figure 184. Alternate Mark Inversion Line Code

1s are represented as pulses, that is, charging the line to 3 volts for a short period. 0s are rep-
resented by doing nothing for a short period of time.
Also, the pulses must alternate in polarity: +3 volts, -3 volts, +3 volts, ...
The CSUs are the devices that perform this function.
Mark is an obsolete term from telegraphy meaning an active condition on the line. Thus we
have 1s represented as pulses (marks), zeros as nothings (spaces), and the marks alternate in
polarity +, -, +, -,... Alternate Mark Inversion.
The pulses alternate in polarity so that the average value on the line is zero, and more impor-
tantly, giving a simple error detection mechanism.
Interested observers will note that the pulses occupy only half of each bit time slot. This is
called a 50% duty cycle return-to-zero line code. This was the choice made for the T1 system
in 1958.

270
A.5.1 Repeaters
A pulse is energy applied to a circuit by the transmitter for a pre-determined length of time. At
the receiver, we wish to make a simple decision: whether a pulse is happening or not.
However, as the energy which is the pulse travels over the physical medium between transmit-
ter and receiver, it will be degraded due to the imperfect nature of the physical medium.
On a T1 system, a pulse is voltage carried on copper wires… and the voltage is attenuated by
the resistance of the copper wires. This is exactly the same problem encountered when dis-
cussing analog techniques and maximum loop lengths in Section 5.7.1.
The shape or envelope of the pulse will also be distorted, with the corners rounded by the ca-
pacitance of the wires.
If the distance between the transmitter and receiver is such that the pulse will become so badly
degraded that it is not possible to make a reliable decision whether a pulse is happening or
not, it is necessary to regenerate the pulse at intermediate points using a repeater.

Figure 185. Repeaters Regenerate Pulses

Repeaters are binary devices that make a decision. If they decide they detect a pulse on the
input, they regenerate a new noiseless square pulse on the output to send down the next cable
segment.
Repeaters are required every 6,000 feet on a T1 circuit.
As discussed in Section 5.7.1, it is possible to boost analog signals using an amplifier, to be
able to transmit information more than the maximum loop length; however, the amplifiers boost
both the signal and the noise, making analog transmission noisy and distance-limited.
As illustrated in Figure 185, the advantage of using pulses is that the repeaters do not boost
the incoming signal, they regenerate it, essentially discarding the noise at each intermediate
point and only transmitting the signal.
Using this idea, information can be coded into 1s and 0s, which are in turn represented as
pulses on the line and transmitted long distances via regeneration of the pulses at regular in-
tervals.
The result is communicating the information without adding in any noise; so quiet, you could
hear a pin drop.

A.6 Synchronization: Bit-Robbing

Framing is for synchronization at the byte level, so the demultiplexer can determine what goes
where.
Synchronization of the start and end of pulses, at the bit level, is also required.
When the sending CSU charges up the line to 3 volts to indicate a “1”, it is necessary that the
receiving repeater is looking at the line at that exact same time to decide if there is a voltage
there or not.
We do not want the far end looking at the line after the sender has finished doing a pulse on
the line. Both ends must be in synchronization.
In the 1980s and 1990s, transmission systems were synchronized by clocks derived from a
master clock operated by the Federal Government. We were all marching to the Federal beat
of the drum.

271
Later, carriers started using the Global Positioning System (GPS) satellites to derive their own
master clocks.
In 1958, there were not any central network clocks nor GPS satellites. Timing was kept by re-
synchronizing on the rising edges of the pulses that are the AMI line code.
Every time a pulse happened, its rising edge was used to pull wayward devices back into
alignment if they had drifted a bit.
Since the system uses pulses for synchronization, there is a requirement to send a certain
number of pulses down the line; and since pulses are caused by sending 1s, this boils down to
a requirement to send a certain number of 1s down the line.
This is referred to as the Ones Density Rule. A simplified version of this rule is that there must
be at least one 1 per byte to keep synchronization.
The designers of the T1 system in 1958 came up with an inelegant solution to this require-
ment: they made the multiplexers always set the least significant bit of most of the channels to
a 1.
This was called bit robbing... the network appropriated one of the eight bits in every byte for
network clocking purposes.
Except in frames 6 and 12 of the Superframe; these positions are reserved for supervision sig-
naling for voice trunks.
The interested voice communications reader may want to note that these bit positions usually
hold digitized versions of the E&M signaling leads from the analog trunks that the T1 carrier
system replaced. These bits are referred to as the A and B signaling bits respectively.
If the system is carrying digitized voices, which was its original intent, the effect is to add in a
bit more quantization noise to the voice signal, because half the time the received signal is in
error by one level.
A human being cannot hear this happening on a voice call. Since T1 was designed for voice
only, it was deemed at the time that this robbing of one of the bits for network synchronization
purposes was acceptable.

A.6.1 56 kb/s for Data

The problem with this bit-robbing scheme to synchronize all of the equipment at the bit level is
that when we try to use this system for communicating data, we find that the least significant
bit of every byte is always set to 1... and so cannot be used.
We have only 7 bits per byte useful for data communications. Bytes are always communicated
at the rate of 8,000 per second.
7 bits per byte 8,000 times per second yields 56,000 bits per second useful data bandwidth per
channel, or 56 kb/s.
This is why 56 kb/s was historically a popular data circuit speed from service providers.

☞
This has nothing to do with “56K” modems. “56K” modems and 56 kb/s digital data cir-
   cuits are different things.

A.7 B8ZS and 64 kb/s Clear Channels

Bit robbing was fine as long as the requirement for T1 was overwhelmingly voice. Then appli-
cations for T1 became as much data as anything else, and so a design modification was
needed to eliminate the robbing of the least significant bit.
The only byte that really causes trouble with the Ones Density Rule is the byte that has 8 zeros
in it (00000000). All of the other bytes have at least one 1 in them.
The design modification was to stop setting the LSB of every byte to a 1 and let the bytes pass
through the T1 system unmolested, except that when a byte with 8 zeros was to be transmit-
ted, instead of sending nothing for 8 bit periods, as the AMI code would require, a special line
code normally never seen would be substituted instead.
For example, the “special” line code to represent eight zeros in a row for the case when the
previous pulse was negative is illustrated in Figure 186. If the previous pulse was positive, this

272
special line code is inverted.
This code is special, because it causes pulses of the same polarity to occur one after another:
in effect, we have created an exception to the alternating rule to mean eight zeros in a row.
When this special code is received at the far end, it is interpreted to mean a byte with 8 zeros
in it. This technique is known as Bipolar Eight Zero Substitution (B8ZS).
The result is that the LSB in each byte is not molested by the transmission system, and so the
user can employ this bit to transmit data, resulting in being able to employ all eight bits of each
byte for data transmission.
This capability was referred to as clear channels in the business. At eight thousand per sec-
ond, the result is 64,000 bits per second or 64 kb/s per channel for data communication.

Figure 186. Bipolar Eight-Zero Substitution

This is 12.5% better performance than the original bit-robbed scheme described on the previ-
ous page, where one could only use 7 bits per byte for data communications.
The special code causes pulses of the same polarity in a row, which is normally considered a
Bipolar Violation (BPV) by the system, since this violates the Alternate Mark Inversion rule.
This use of a normally illegal code means that all of the repeaters and CSUs have to be up-
graded in software to know about this special code and not think it a BPV error.
The installed base of T1 circuits did not historically support B8ZS and clear channels, so it was
imperative to specify “64 kb/s clear channels” when ordering T1 circuits... otherwise you could
get 56 kb/s bit-robbed circuits, since this was the original plan and only “new” (1980+) facilities
supported the new plan.

A.8 How T1 Is Provided

In practice, T1 was an access technology, used only for the last mile or two: copper wires used
to get a DS1-rate service from the customer premise to the fiber systems. The information is
transported on fiber long-distance.

273
 Figure 187. T1 Provisioning
As illustrated in Figure 187, T1 is a 4-wire copper circuit running from the local phone com-
pany’s building (usually a Central Office) to the customer premise.
At the customer side, the wires are terminated on a Channel Service Unit (CSU), which pro-
vides an interface for the customer’s multiplexer.
At the Central Office, the wires are terminated on an Office Channel Unit, which performs the
same functions. If obtaining service from an Inter-Exchange Carrier for a long-distance circuit,
the service will be carried through the local phone company’s CO to the IXC’s Point of Pres-
ence.
A 1/0 multiplexer, with DS1 on one side and DS0s on the other, can be located at the customer
premise. The T1 carries the DS1 to the CO. Fiber backbone transmission systems carry multi-
ples of DS3-rate signals, not DS1s, and so at the CO, the information on the T1 will be com-
bined with many other DS1-rate streams to form a DS3-rate stream by a 3/1 multiplexer.
These DS3-rate streams are then moved long distance over fiber.
At the far end, the reverse process takes place with similar equipment and cabling. If the other
end is in Europe, the signal may be delivered as an E1 over the CEPT-1 carrier system.

A.8.1 HDSL
For advanced readers: T1 as such was not actually used for many T1 services (!) An issue
with T1 is that it requires repeaters: the first one at 3,000 feet, and every 6,000 feet thereafter.

274
Repeaters are expensive to install and maintain. Variations on T1 that do not require repeaters
up to 12,000 feet were developed.
These technologies are called High-Speed Digital Subscriber Line (HDSL). [not related to resi-
dential DSL].
When someone said, “we have a T1” from here to there, this might have been wholly inaccu-
rate. They had HDSL access and SONET transport, and no T1 technology at all.
It would be more accurate to say, “we have a full-period DS1-rate service” from here to there.
But “T1” is short and catchy-sounding…

A.9 Fractional T1, DACS and Cross-Connects

If the customer does not require the capacity of a full T1, but rather only a few of the DS0
channels, a service generically called fractional T1 was available.
From the customer premise to the CO, the equipment and cabling is identical to full T1 service.
The only difference is that the customer’s T1 multiplexer can only put data on the channels
that the customer purchased. The rest are not connected, and the carrier will ignore any data
received on the extra channels.

Figure 188. Fractional T1

The carrier requires additional equipment to provide fractional T1. Since these customers are
only using some of the channels in their DS1-rate signal, it is necessary to drop out the DS0

275
channels that are being used, and insert them into a DS1 with other customers, to make up a
DS1 with all channels used.
This piece of equipment is called a Cross-Connect, a Digital Cross-Connect System (DCS) or
Digital Access and Cross-Connect System (DACS).
It takes fractional DS1s from a number of customers to make up a full DS1. This DS1 then is
multiplexed into a DS3 and carried over the backbone just as for full T1 service in the previous
section.
At the far end, the same equipment is required, and the reverse process happens, picking out
the right DS0s and sending them over a T1 to the far end customer premise.
This is a good illustration of difference between “T1” and “DS1”. T1 is the physical layer proto-
col for physically cabling together the OCU and CSU. DS1 is the rate of signal that it carries.
The DS1 can be split into individual DS0s. The T1 is four wires.

A.10 Subrate Data Circuits 1.2 kb/s to 56 kb/s

If the customer did not require even fractional T1 service, but needed only a subrate data cir-
cuit at 9.6 kb/s or 56 kb/s, services were available for that.
From a conceptual point of view, the equipment is much the same. The fundamental difference
is that the service is provided from a T1 multiplexer in the carrier’s building instead of a T1
multiplexer on the customer premise.
Since it is usually a far distance from the Central Office to the Customer Premise, a line exten-
der system is required to extend this single channel from the CO to the customer.
Carriers typically installed a 4-wire circuit from the CO to the CP, and put line extender devices
at each end. At each end, a standard interface like EIA-232, EIA-422 or V.35 is presented to
the network equipment.

A.10.1 CSUs, DSUs and CSU/DSUs

There are numerous terms in use for T1-related circuit-terminating devices. The most common
names are Data Service Unit (DSU), Digital Service Unit (DSU), and Channel Service Unit
(CSU).
Other names for this device include Digital Terminating Unit (DTU), Terminal Interface Unit
(TIU), Terminal Interface Equipment (TIE), and Network Channel Terminating Equipment
(NCTE).
They all perform the same function.
Switched services require a combined CSU/DSU, which can perform signaling functions as
well as line extension.

276
 Figure 189. Subrate Data Services
The terms CSU, DSU and CSU/DSU are often interchanged. Most people in the business
called all three devices a “CSU/DSU” without knowing the official definitions:
• A CSU was the circuit-terminating equipment for a T1.
• A DSU was a line extender for extending single channels from the CO to the customer.
This is called a tail circuit, since the DSUs are on the slow side of the multiplexer.
• A CSU/DSU is a DSU that can also signal control information to the network, used for
switched 56 kb/s services, which were called Dataphone Digital Services (DDS).

277
Appendix B
Voice Services and Jargon
B.1 Local Voice Services
This appendix provides a high-level overview of popular local and long-distance voice services,
and explains some of the associated jargon.

B.1.1 POTS and Party Lines

The local telephone company provides a number of basic services including the familiar indi-
vidual line, called Plain Ordinary Telephone Service (POTS) in the business. The telephone
company also offers Party Line service (legally, technically still available most everywhere),
where more than one user shares the same loop.

B.1.2 CLASS Services

In addition to POTS, enhanced features on basic service such as Call Waiting, Call Forward-
ing, Call Display and so forth are available. These enhanced services are known generically as
Custom Local Area Signaling System (CLASS) services.

B.1.3 Local Measured Service

POTS is often a flat-rate service. The telephone companies would like to (and have been able
to in some areas) change this to Local Measured Service (LMS) where all phone calls are paid
for on a per-use basis. They claim that this would be a revenue-neutral transition. The philoso-
phy is to get the users in the habit of paying for services on a per-use basis... so that the intro-
duction of other services, which are paid for on a per-use basis, is facilitated.

B.1.4 Public Coin Telephone Service

Both an affiliate of the local phone company and other private companies can own the coin
telephones. Pay phones and long distance are both deregulated, so one must be careful when
placing a long-distance call from a payphone or hotel. It is prudent to first dial an access num-
ber to reach your carrier before giving out your credit/calling card information. If you dial 0 from
a pay phone, you might get a nasty surprise when the bill arrives.

B.1.5 Directory Services

Another service provided by the local telephone company is directory information. Note that is
no longer free, as it used to be, since there is no guarantee that you will make the revenue-
bearing call over the local company’s long-distance network.

B.1.6 Business Services

Business services provided by the local telephone company include
• Business lines (same as individual lines, but at higher cost)
• Off-Premise Extensions (OPX): an extension cord running from your Private Branch Ex-
change (PBX) telephone switch to another location across the street, across town or on the
other side of the region.
• Private Line service: an always-connected or full period line running between two points.
Various different combinations of powering and ringing can be ordered. The direct taxi line at
the supermarket is an example.

B.1.7 Access

278
One of the most important functions after dial tones: the local phone company provides the ac-
cess to long-distance carriers.
Long-distance phone calls are digitized on the CO switch or remote owned by the LEC, then
multiplexed together into DS3-rate streams to be placed on a SONET fiber connection to a toll
switch owned by an Inter-Exchange Carrier.

B.2 Long Distance Voice Services

Long distance service is offered by Inter-Exchange Carriers (IXCs), resellers and in some
cases by the same company that owns LECs. The LEC usually owns the loop: the access
wiring from the customer premise to the central office, as well as the connection from the CO
to the toll center, the jumping-off point for long distance calls. Long-distance connections can
be made by any company that has circuits – real or leased – between those toll centers.

B.2.1 Operator Services

Operator services are required to offer collect calling, third-party billing or person-to-person
calling. The telephone companies and IXCs offer operator services. Many resellers do not.
Some of these functions can be performed by a computer – a voice recognition and response
system – instead of a person.

B.2.2 Foreign Exchange

In addition to basic long-distance dial-up voice connections, other types of circuits and configu-
rations are available. An example of an older service is a Foreign Exchange circuit (FX), which
runs from the Central Office closest to the head office of a business to a CO in another city.
Customers in the far city can call a local number, and be routed (without knowing) over this
dedicated line to the head office of the business. This gives the customer in effect a 7- or 10-
digit toll-free number to the head office, as the business pays the dedicated line charge.

B.2.3 OPX: Off-Premise Extension

OPX circuits are offered under a long-distance package just as they are offered for local ser-
vice. Dedicated lines, or more accurately, full-period services, can be installed to connect two
points together over any distance.

B.2.4 Tie Line

Tie line is a term for dedicated voice lines that connect two Private Branch Exchanges (PBXs)
in different cities together.

B.2.5 Private Networks

Private networks can be constructed by connecting PBXs in different cities together with dedi-
cated lines. Private lines and private networks are full-period services, billed on a 7/24 basis,
i.e. 7 days a week, 24 hours a day.

B.2.6 WATS
Current services include WATS, which stands for Wide Area Telephone Service and is volume
discounts on outgoing long distance calls.
800 service used to be called INWATS, and is the same thing as WATS except that the called
party pays... and has a volume discount. Of course, 888, 877 and 866 are also used for these
toll-free calls.

B.2.7 AIN Services

Services and features like usage-sensitive billing, time-of-day routing changes, remote call for-
warding and so forth are all implemented in software, and are evidence of the change of the
telephone network from being hardware-defined (dedicated lines) to a Software-Defined Net-
work.

279
These kinds of services are sometimes referred to as Advanced Intelligent Network (AIN) ser-
vices.
The Signaling System Seven (SS7) control network, which is made up of computers and data-
bases connected to telephone switches in COs and toll centers provides the infrastructure
which allows the deployment of these services.
A trigger can be associated with a particular telephone number, telling the switching system to
perform a database lookup via the SS7 network to determine the routing for that particular call.

B.2.8 Virtual Private Voice Networks

An example of the application of AIN features is Virtual Private Networks (VPNs), which have
replaced private networks for voice communications.
A VPN is volume discounts for calls between locations together with an abbreviated dialing
plan: to reach someone in the west coast office, a user in the east coast office would only have
to dial the 4-digit extension. The network would look up and add the required digits or routing
information.
The big advantage of Virtual Private Networks is that they are implemented not with expensive
dedicated lines, but with services defined in software and billed on a usage-sensitive basis,
and so are more flexible and cheaper.

☞ Caveat: VPN means abbreviated dialing when used in context of voice services, and
secure packet data communications when used in context of IP networks. Even though
   these two ideas have the same name, they are completely different concepts.

280
Appendix C
Acronyms and Abbreviations
10BASE-2 Thinwire Ethernet LAN
10BASE-5 Ethernet LAN
10BASE-T 10 Mb/s Baseband Ethernet LAN on Twisted Pair
100BASE-T 100 Mb/s Ethernet LAN on Twisted Pair
1000BASE-T 1 Gb/s Ethernet LAN on Twisted Pair
1xEV-DO 1X Evolution, Data-Optimized
3GPP Third Generation Partnership Project
ABR Available Bit Rate
ACD Automated Call Distributor
ACK Acknowledgment
ACR Attenuation to Crosstalk Ratio
ADCCP Advanced Data Communication Control Procedures
ADSL Asymmetric Digital Subscriber Line
AES Advanced Encryption Standard
AIN Advanced Intelligent Network
AM Amplitude Modulation
ANI Automatic Number Identification
ANSI American National Standards Institute
AON Active Optical Network
AP Access Point
ARIN American Registry for Internet Numbers
ARP Address Resolution Protocol
AS Autonomous System
ASK Amplitude Shift Keying
ASN Autonomous System Number
ATM Asynchronous Transfer Mode
b/s bits per second
BGP Border Gateway Protocol
BPV Bipolar Violation
BRI ISDN Basic Rate Interface
BST Base Station Transceiver
BTS Base Transceiver Subsystem
CAC Connection Admission Control
CAT Category
CATV Cable TV or Community Antenna Television
CCITT Comité Consultatif International de Téléphone et de Télégraphe
CD Collision Detection
CD Compact Disc
CDMA Code Division Multiple Access
CIDR Classless Inter-Domain Routing
CIF Common Interface Format
CLEC Competitive Local Exchange Carrier
CO Central Office
CoS Class of Service
CP Customer Premise
CPE Customer Premise Equipment
CRC Cyclic Redundancy Check
CSMA-CD Carrier Sensing Multiple Access with Collision Detection
CSU Channel Service Unit
CTNS Certified Telecommunication Network Specialist
CWDM Coarse Wave-Division Multiplexing
DACS Digital Access and Cross-Connect System
DARPA Defense Advanced Research Projects Agency
DCE Data Circuit-terminating Equipment
DCS Digital Cross-connect System
DF Don’t Fragment
DHCP Dynamic Host Configuration Protocol
DLCI Data Link Connection Identifier
DMS Digital Multiplex Switch
DNS Domain Name System
DOCSIS Data over Cable System Interface Specification
DoIP Data over IP

281
DPSK Differential Phase Shift Keying
DS Differentiated Services
DS0 Digital Service Level 0: 64 kb/s
DS0A Subrate multiplexing scheme “A”
DS0B Subrate multiplexing scheme “B”
DS1 Digital Service Level 1: 1.5 Mb/s
DS2 Digital Service Level 2: 6.3 Mb/s
DS3 Digital Service Level 3: 45 Mb/s
DSL Digital Subscriber Line
DSLAM Digital Subscriber Line Access Multiplexer
DTE Data Terminal Equipment
DTMF Dual Tone Multiple Frequency
DTU Digital Terminal Unit
DVD Digital Versatile Disk
DWDM Dense Wave Division Multiplexing
EBCDIC Extended Binary Coded Decimal Interchange Code
EIA Electronic Industries Association
EMI Electro-Magnetic Interference
eNB Enhanced Network Base Station
EoMPLS Ethernet over MPLS
EPC Evolved Packet Core
ESF Extended Superframe Format
ESN Electronic Serial Number
ETSI European Telecommunications Standards Institute
FCS Frame Check Sequence
FDL Facility Data Link
FDM Frequency Division Multiplexing
FDMA Frequency Division Multiple Access
FEC Forward Error Correction
FEC Forwarding Equivalence Class
FM Frequency Modulation
FRAD Frame Relay Access Device
FSA Fiber Serving Area
FSK Frequency Shift Keying
FTP File Transfer Protocol
FTTH Fiber to the Home
FTTN Fiber to the Neighborhood / Node
FTTP Fiber to the Premise
FX Foreign Exchange
G Giga = 109 = Billion (US), Thousand Million (UK)
GB Gigabyte = 1030 bytes
Gb/s Gigabit per second
GHz Gigahertz
GPON Gigabit Passive Optical Network
GPRS General Packet Radio System
GPS Global Positioning System
GSM Global System for Mobile Communications
GUI Graphical User Interface
HD High Definition
HDLC High-level Data Link Control protocol
HDSL High-Speed Digital Subscriber Line
HFC Hybrid Fiber-Coax
HMI Human Machine Interface
HSDPA High Speed Downlink Packet Access
HSPA High Speed Packet Access
HSUPA High Speed Uplink Packet Access
HTTP Hypertext Transport Protocol
HTTPS Secure Hypertext Transport Protocol
Hz Hertz = cycles per second
IAB Internet Advisory Board
IAD Integrated Access Device
IANA Internet Assigned Numbers Authority
IC Integrated Circuit
IDEN Integrated Digital Enhanced Network
IEC International Electrotechnical Commission
IEEE Institute of Electrical and Electronic Engineers
IETF Internet Engineering Task Force
ILEC Incumbent Local Exchange Carrier
IMT International Mobile Telecommunications
IMT-2000 International Mobile Telecommunications 2000 (3G)
IMT-DS IMT-Direct Spread (UMTS, W-CDMA)

282
IMT-MC IMT-Multicarrier (CDMA2000, 1X)
INWATS Incoming Wide Area Telephone Service
IP Internet Protocol
IP-PSTN IP Packet-Switched Telecommunications Network
IPsec IP Security
IPTV Television over IP
IPv4 IP version 4
IPv6 IP version 6
ISDN Integrated Services Digital Network
ISO International Organization for Standardization
ISP Internet Service Provider
ISUP ISDN User Part
ITU International Telecommunications Union
IVR Interactive Voice Response System
IX Internet Exchange
IXC Inter Exchange Carrier
JPEG Joint Photographic Experts Group
k kilo = 103 = thousand
K 210 = 1024
kb kilobit = 1,000 bits
KB Kilobyte = 210 bytes = 1024 bytes
kft kilofeet = 1000 feet
λ lambda (wavelength)
L2 Layer 2
L3 Layer 3
LAN Local Area Network
LATA Local Access and Transport Area
LD Long Distance
LEC Local Exchange Carrier
LED Light-Emitting Diode
LEO Low Earth Orbit
LER Label Edge Router
LLC Logical Link Control
LMS Local Measured Service
LSB Least Significant Bit
LSP Label-Switched Path
LSR Label-Switching Router
LTE Universal Terrestrial Radio Access Network Long Term Evolution
LX Long Wavelength
LZW Lempel – Ziv –Welch
M Mega = 106 = Million
M.Eng. Master of Engineering
MAC Media Access Control
MAN Metropolitan Area Network
Mb Megabit = 1,000,000 bits
MB Megabyte = 220 bytes = 1,048,576 bytes
MF Multifrequency
MF Mainframe
MHz Megahertz
micro (m) 10-6
milli (m) 10-3
MIME Multipart Internet Mail Extensions
MIMO Multiple-Input, Multiple-Output
MPEG Moving Picture Experts Group
MPLS Multiprotocol Label-Switching
MSA Multiple Sourcing Agreement
MSB Most Significant Bit
MTP Message Transfer Part
MTSO Mobile Telephone Switching Office
MUX Multiplexer
MX Mail Exchanger
nano (n) 10-9
NAT Network Address Translator
NCTE Network Circuit Terminating Equipment
NHLFE Next Hop Label Forwarding Entry
NIC Network Interface Card
NMT Nordic Mobile Telephone System
NOC Network Operations Center
NSF National Science Foundation
NT1 Network Termination type 1

283
NTSC National Television Standards Committee
OC Optical Carrier (SONET)
OC3 OC level 3 = 3 DS3s
OC48 OC level 48 = 48 DS3s
OCU Office Channel Unit
OE Optical Ethernet
OFDM Orthogonal Frequency-Division Multiplexing
OLT Optical Line Terminal
ONT Optical Network Terminal
ONU Optical Network Unit
OPI Outside Plant Interface
OPX Off-Premise Extension
OSI Open Systems Interconnect
OSPF Open Shortest Path First
PABX Private Automated Branch Exchange
PBX Private Branch Exchange
PC Personal Computer
PCM Pulse Code Modulation
PCS Personal Communication Services
PDU Protocol Data Unit
PHB Per-Hop Behavior
PIC Preferred Inter-exchange Carrier
pico (p) 10-12
PON Passive Optical Network
POP Point of Presence
POP Post Office Protocol
POTS Plain Ordinary Telephone Service
PRI ISDN Primary Rate Interface
PSK Phase Shift Keying
PSTN Packet-Switched Telecommunications Network
PSTN Public Switched Telephone Network
PTT Post Telephone and Telegraph
PVC Permanent Virtual Circuit
QAM Quadrature Amplitude Modulation
QoS Quality of Service
QPSK Quadrature Phase Shift Keying
RBHC Regional Bell Holding Company
RBOC Regional Bell Operating Company
RFC Request for Comments
RIR Regional Internet Registry
RPR Resilient Packet Ring
RSA Rivest Shamir Adelman
RTP Real-Time Control Protocol
RTP Rusty Twisted Pair
SAC Subscriber Area Concept
SCP Service Control Point
SD Standard Definition
SDH Synchronous Digital Hierarchy
SDLC Synchronous Data Link Control
SDSL Symmetric Digital Subscriber Line
SFP Small Formfactor Pluggable
SIM Subscriber Information Module
SIP Session Initiation Protocol
SLA Service Level Agreement
SMS Short Message Service
SMTP Simple Mail Transfer Protocol
SNMP Simple Network Management Protocol
SOA Start of Authority
SONET Synchronous Optical Network
SPI Stateful Packet Inspection
SRC Source
SSH Secure Shell
SSID Service Set ID
SSL Secure Socket Layer
SSP Service Switching Point
STM Synchronous Transport Module
STP Signal Transfer Point
STS Synchronous Transport Signal
SVC Switched Virtual Circuit
SX Short Wavelength
T Tera = 1012 = Trillion (US), Billion (UK)

284
TACS Total Access Communication System
TCO Telecommunications Certification Organization
TCP Transmission Control Protocol
TDM Time-Division Multiplexing
TDMA Time-Division Multiple Access
TIA Telecommunications Industries Association
TIU Terminal Interface Unit
TSB Technical Service Bulletin
μ micro
μm micron = 10-6 meters
U/D Up / Down
UBR Unspecified Bit Rate
UDP User Datagram Protocol
UMTS Universal Mobile Telecommunications Service
UNIX Harem Guards
URL Uniform Resource Locator
USB Universal Serial Bus
UTF Unicode Transformation Format
UTP Unshielded Twisted Pair
UTRAN Universal Terrestrial Radio Access Network
UUCP Unix-Unix Copy Protocol
VBR Variable Bit Rate
VCSEL Vertical-Cavity Surface-Emitting Laser
VDSL Very High Bit Rate Digital Subscriber Line
VLAN Virtual Local Area Network
VoD Video on Demand
VoIP Voice Over IP
VPI Virtual Path Identifier
VPLS Virtual Private LAN Service
VPN Virtual Private Network
VSAT Very Small Aperture Terminal
VSB-AM Vestigial Side Band Amplitude Modulation
VSP VoIP Service Provider
WAN Wide Area Network
WATS Wide Area Telephone Service
W-CDMA Wideband CDMA (= IMT-DS, UMTS)
WDM Wave Division Multiplexing
WEP Wired Equivalent Privacy
Wi-Fi Wireless Fidelity
WiMAX Worldwide Interoperability for Microwave Access
WLAN Wireless Local Area Network
WPA Wi-Fi Protected Access
xDSL Any DSL Technology

285
About Teracom
About the Author
Eric Coll is an international expert in telecommunications, data communications and network-
ing and has been actively involved in the industry since 1983. He holds Bachelor of Engineer-
ing and Master of Engineering (Electrical) degrees.
Mr. Coll has taught telecommunications technology training seminars to wide acclaim across
North America since 1992, and has broad experience working as an engineer in the telecom-
munications industry.
He has worked for Nortel’s R&D labs as a design engineer on projects including digital voice
and data communications research and digital telecom network equipment design; and on
satellite radar systems; consulting on Wide Area Network design for HMO applications; and
many other projects in capacities ranging from detailed design and implementation to systems
engineering, project leader and consultant.
In addition to being founder and Director of Teracom Training Institute, Mr. Coll provides con-
sulting to the telecommunications industry and acts as a telecommunications technology sub-
ject matter expert for tax and legal matters.

Public Seminars
Instructor-led training is the best you can get, allowing you to ask questions and interact with
classmates. Teracom’s public seminars are instructor-led courses geared for the non-engineer-
ing professional needing a comprehensive overview and update, and those new to the busi-
ness needing to get up to speed.
Teracom instructor-led seminars are career-enhancing training. They’re the ideal way to put in
place a solid base of knowledge.
We start at the beginning of the story, progress through it in a logical order and finish at the
end.
We will explain the jargon and buzzwords, and even more importantly, provide you with a
structured understanding on which you can build in the future. Visit teracomtraining.com to see
the latest schedule.

Private Onsite Seminars

Since 1992, we have provided high-quality on-site instructor-led training in telecommunica-
tions, data communications, IP, networking, VoIP and wireless at Bell Labs, Qualcomm, 3Com,
Cisco, Intel, Verizon, AT&T, Alcatel, Nortel, Teleglobe, MindSpring, APEX Telecom, Equifax,
Transamerica Insurance, The Hartford, Bell Canada, Bell Mobility, TELUS, Frontier Communi-
cations, SDN Communications, City of San Francisco, Shaw, Shentel, Western Iowa Tele-
phone, American Broadband, Cap Gemini, MicroCell Telecom, TDS Telecom, Kyocera, Wells
Fargo, Winstar, Western Wireless, US Cellular, Ericsson/Hewlett-Packard, Bloomberg, Depart-
ment of Homeland Security, the NSA, Office of Naval Intelligence, Defense Systems Informa-
tion Agency, Spawar, ComSec Establishment, US Coast Guard, US Air Force, General Ac-
counting Office, GSA, Intelsat, RangeTel, Alltel, Vertek, DSCI, Cox Cable, Rogers, Florida
Power and Light, Idaho Power, Entergy, New York Power Authority, Genuity, LG Electronics,
Panasonic, SouthEast Telephone, State of Nebraska, State of Montana, Ketchikan Public Utili-
ties, L-3 Communications, NOAA, Tektronix, Canadian Army and Air Force, Bermuda Telecom,
UTS, the Universal Service Administrative Company and US Army... to name a few.
We have built a solid reputation for delivering high-quality training programs that are a re-
sounding success. We would like to do the same for you! Please contact us for more informa-
tion.

DVD-Video Courses
Teracom’s DVD video training courses are high-quality multimedia productions featuring the in-
structor, extensive animated graphics and bullets... like private lessons from the Director of the

286
Institute!
Our engaging and humorous instructor Eric Coll talks to you directly via the camera, using his
vast experience and acclaimed ability to explain key concepts, mainstream technologies and
how it all fits together – in plain English – while keeping your attention.
Our goal is to bust the buzzwords, explain the jargon and instill structured understanding -
knowledge that lasts a lifetime.
Every course comes with a comprehensive workbook with copies of all graphics and detailed
reference notes, sure to be a valuable reference for years to come. Online courses and certifi-
cations are included at no extra charge with the DVD courses. Visit teracomtraining.com for
details.

GSA Schedule
Teracom instructor-led courses, DVDs, online courses and certification packages are on our
US Government GSA schedule supply contract... which took two years and a 200-page appli-
cation to obtain... so you know you are getting quality.

Online Courses
Teracom Online Courses are based on our famous instructor-led courses, tuned and refined
over more than 20 years of teaching these topics to thousands of students and hundreds of
companies and organizations.
These courses are the same courses used by the biggest telecom carriers in the business to
train their employees - constantly updated to deliver the core technical knowledge required in
the telecom business today.
This is the best quality training of its kind available.

TCO Telecommunications Certification

Upgrade your knowledge - and your résumé - with high-quality telecom training courses by
Teracom coupled with certification from the Telecommunications Certification Organization.
Individuals: differentiate yourself from the rest of the crowd when applying for a job or angling
for a promotion, with solid knowledge of telecom and network technology – and a certification
to prove it.
For employers, TCO Certification is an extremely cost-effective way of implementing consis-
tent, comprehensive telecommunications and network technology fundamentals training
across your team, ensuring that both existing resources and new hires are up to the same
speed, with a common vocabulary, framework and knowledge base.
The exclusive Unlimited Plan gives you unlimited repeats of courses and exams... which
means guaranteed to pass... and refresh your knowledge anytime in the future!
You get a Certificate suitable for framing plus - another Teracom exclusive - a personalized
Letter of Reference / Letter of Introduction for your résumé detailing the knowledge your TCO
Certification represents and inviting the recipient to contact Teracom for verification.
Please visit teracomtraining.com/certification.htm for more information.
Teracom Online Courses and Certifications

287
Teracom is a Gold Training Partner of the Telecommunications Certification Organization, au-
thorized to administer exams for TCO certifications on the myTeracom Learning Management
System and award TCO Certifications.
✴ Certified Telecommunications Network Specialist (CTNS)
✴ Certified Telecommunications Analyst (CTA)
✴ Certified Wireless Analyst (CWA)
TCO Certification is proof of your knowledge of telecom, datacom and networking fundamen-
tals, jargon, buzzwords, technologies and solutions.
Guaranteed to Pass – and repeat courses anytime with the Unlimited Plan!
Join our thousands of satisfied customers including:

the FBI Training Academy, US Marine Corps Communications School, US Army, Navy, Air
Force, Coast Guard, CIA, IRS, CRA, CRTC, RCMP, banks, power companies, police forces,
manufacturers, government, local and regional telcos, broadband carriers, individuals and
many more.
Visit teracomtraining.com to register today. Scan this code with your smartphone to visit the
certifications page for full details!

288