ch01.

I 4/21/00 12:02 PM Page 1

CHAPTER 1
Chapter Objectives
How communication
systems work
Frequency allocation and
propagation characteristics
Computer solutions
(MATLAB and MATHCAD)
Information measure
Coding performance

INTRODUCTION

The subject of communication systems is immense. It is not possible to include all topics and
keep one book of reasonable length. In this book, the topics are carefully selected to accentuate
basic communication principles. Moreover, the reader is motivated to appreciate these princi-
ples by the use of many practical applications. Often, practical applications are covered before
the principles are fully developed. This provides instant gratification and motivates the read-
er to learn the basic principles well. The goal is to experience the joy of understanding how
communication systems work and to develop an ability to design new communication systems.
What is a communication system? Moreover, what is electrical and computer engi-
neering (ECE)? ECE is concerned with solving problems of two types: (1) production or
transmission of electrical energy and (2) transmission or processing of information. Com-
munication systems are designed to transmit information.
It is important to realize that communication systems and electric energy systems have
markedly different sets of constraints. In electric energy systems, the waveforms are usual-
ly known, and one is concerned with designing the system for minimum energy loss.
In communication systems, the waveform present at the receiver (user) is unknown until
after it is receivedotherwise, no information would be transmitted, and there would be no

1
ch01.I 4/21/00 12:02 PM Page 2

2 Introduction Chap. 1

need for the communication system. More information is communicated to the receiver when
the user is more surprised by the message that was transmitted. That is, the transmission of
information implies the communication of messages that are not known ahead of time (a priori).
Noise limits our ability to communicate. If there were no noise, we could communi-
cate messages electronically to the outer limits of the universe by using an infinitely small
amount of power. This has been intuitively obvious since the early days of radio. However,
the theory that describes noise and the effect of noise on the transmission of information was
not developed until the 1940s, by such persons as D. O. North [1943], S. O. Rice [1944], C. E.
Shannon [1948], and N. Wiener [1949].
Communication systems are designed to transmit information bearing waveforms to
the receiver. There are many possibilities for selecting waveforms to represent the informa-
tion. For example, how does one select a waveform to represent the letter A in a typed mes-
sage? Waveform selection depends on many factors. Some of these are bandwidth
(frequency span) and center frequency of the waveform, waveform power or energy, the ef-
fect of noise on corrupting the information carried by the waveform, and the cost of gener-
ating the waveform at the transmitter and detecting the information at the receiver.
The book is divided into eight chapters and four appendices. Chapter 1 introduces
some key concepts, such as the definition of information, and provides a method for evalu-
ating the information capacity of a communication system. Chapter 2 covers the basic tech-
niques for obtaining the spectrum bandwidth and power of waveforms. Baseband waveforms
Awhich have frequencies near f=0B are studied in Chapter 3, and bandpass waveforms Afre-
quencies in some band not near f=0B are examined in Chapters 4 and 5. The effect of noise
on waveform selection is covered in Chapters 6 and 7. Case studies of wire and wireless
communications, including personal communication systems (PCS) are emphasized in Chap-
ter 8. The appendices include mathematical tables, a short course on probability and ran-
dom variables, standards for computer communications, and an introduction to MATLAB.
Standards for communications systems are included, as appropriate, in each chapter. The per-
sonal computer is used as a tool to plot waveforms, compute spectra of waveforms, and an-
alyze and design communications systems.
In summary, communication systems are designed to transmit information. Commu-
nication system designers have four main concerns:

1. Selection of the information-bearing waveform

2. Bandwidth and power of the waveform
3. Effect of system noise on the received information
4. Cost of the system.

11 HISTORICAL PERSPECTIVE
A time chart showing the historical development of communications is given in Table 11.
The reader is encouraged to spend some time studying this table to obtain an appreciation
for the chronology of communications. Note that although the telephone was developed late
in the 19th century, the first transatlantic telephone cable was not completed until 1954.
Previous to that date, transatlantic calls were handled via shortwave radio. Similarly,
ch01.I 4/21/00 12:02 PM Page 3

TABLE 11 IMPORTANT DATES IN COMMUNICATIONS

Year Event

Before 3000 B.C. Egyptians develop a picture language called hieroglyphics.

A.D. 800 Arabs adopt our present number system from India.
1440 Johannes Gutenberg invents movable metal type.
1752 Benjamin Franklins kite shows that lightning is electricity.
1827 Georg Simon Ohm formulates his law (I=ER).
1834 Carl F. Gauss and Ernst H. Weber build the electromagnetic telegraph.
1838 William F. Cooke and Sir Charles Wheatstone build the telegraph.
1844 Samuel F. B. Morse demonstrates the Baltimore, MD, and Washington, DC, tele-
graph line.
1850 Gustav Robert Kirchhoff first publishes his circuit laws.
1858 The first transatlantic cable is laid and fails after 26 days.
1864 James C. Maxwell predicts electromagnetic radiation.
1871 The Society of Telegraph Engineers is organized in London.
1876 Alexander Graham Bell develops and patents the telephone.
1883 Thomas A. Edison discovers a flow of electrons in a vacuum, called the Edison ef-
fect, the foundation of the electron tube.
1884 The American Institute of Electrical Engineers (AIEE) is formed.
1887 Heinrich Hertz verifies Maxwells theory.
1889 The Institute of Electrical Engineers (IEE) forms from the Society of Telegraph En-
gineers in London.
1894 Oliver Lodge demonstrates wireless communication over a distance of 150 yards.
1900 Guglielmo Marconi transmits the first transatlantic wireless signal.
1905 Reginald Fessenden transmits speech and music by radio.
1906 Lee deForest invents the vacuum-tube triode amplifier.
1907 The Society of Wireless Telegraph Engineers is formed in the United States.
1909 The Wireless Institute is established in the United States.
1912 The Institute of Radio Engineers (IRE) is formed in the United States from the Soci-
ety of Wireless Telegraph Engineers and the Wireless Institute.
1915 Bell System completes a U.S. transcontinental telephone line.
1918 Edwin H. Armstrong invents the superheterodyne receiver circuit.
1920 KDKA, Pittsburgh, PA, begins the first scheduled radio broadcasts.
1920 J. R. Carson applies sampling to communications.
1923 Vladimir K. Zworkykin devises the iconoscope television pickup tube.
1926 J. L. Baird, (England) and C. F. Jenkins (United States) demonstrate television.
1927 The Federal Radio Commission is created in the United States.
1927 Harold Black develops the negative-feedback amplifier at Bell Laboratories.
1928 Philo T. Farnsworth demonstrates the first all-electronic television system.
1931 Teletypewriter service is initiated.
1933 Edwin H. Armstrong invents FM.
1934 The Federal Communication Commission (FCC) is created from the Federal Radio
Commission in the United States.
1935 Robert A. Watson-Watt develops the first practical radar.
1936 The British Broadcasting Corporation (BBC) begins the first television broadcasts.
1937 Alex Reeves conceives pulse code modulation (PCM).

3
ch01.I 4/21/00 12:02 PM Page 4

4 Introduction Chap. 1

TABLE 11 (cont.)

Year Event

1941 John V. Atanasoff invents the digital computer at Iowa State College.
1941 The FCC authorizes television broadcasting in the United States.
1945 The ENIAC electronic digital computer is developed at the University of Pennsylva-
nia by John W. Mauchly.
1947 Walter H. Brattain, John Bardeen, and William Shockley devise the transistor at Bell
Laboratories.
1947 Steve O. Rice develops a statistical representation for noise at Bell Laboratories.
1948 Claude E. Shannon publishes his work on information theory.
1950 Time-division multiplexing is applied to telephony.
1950s Microwave telephone and communication links are developed.
1953 NTSC color television is introduced in the United States.
1953 The first transatlantic telephone cable (36 voice channels) is laid.
1957 The first Earth satellite, Sputnik I, is launched by USSR.
1958 A. L. Schawlow and C. H. Townes publish the principles of the laser.
1958 Jack Kilby of Texas Instruments builds the first germanium integrated circuit (IC).
1958 Robert Noyce of Fairchild produces the first silicon IC.
1961 Stereo FM broadcasts begin in the United States.
1962 The first active satellite, Telstar I, relays television signals between the United
States and Europe.
1963 Bell System introduces the touch-tone phone.
1963 The Institute of Electrical and Electronic Engineers (IEEE) is formed by merger of
the IRE and AIEE.
196366 Error-correction codes and adaptive equalization for high-speed error-free digital
communications are developed.
1964 The electronic telephone switching system (No. 1 ESS) is placed into service.
1965 The first commercial communications satellite, Early Bird, is placed into service.
1968 Cable television systems are developed.
1971 Intel Corporation develops the first single-chip microprocessor, the 4004.
1972 Motorola demonstrates the cellular telephone to the FCC.
1976 Personal computers are developed.
1979 64-kb random access memory ushers in the era of very large-scale integrated (VLSI)
circuits.
1980 Bell System FT3 fiber-optic communication is developed.
1980 Compact disk is developed by Philips and Sony.
1981 IBM PC is introduced.
1982 AT&T agrees to divest its 22 Bell System telephone companies.
1984 Macintosh computer is introduced by Apple.
1985 FAX machines become popular.
1989 Global positioning system (GPS) using satellites is developed.
1995 The Internet and the World Wide Web become popular.
2000present Era of digital signal processing with microprocessors, digital oscilloscopes, digitally
tuned receivers, megaflop personal computers, spread spectrum systems, digital satellite
systems, digital television (DTV), and personal communications systems (PCS).
ch01.I 4/21/00 12:02 PM Page 5

Section 12 Digital and Analog Sources and Systems 5

although the British began television broadcasting in 1936, transatlantic television relay was
not possible until 1962, when the Telstar I satellite was placed into orbit. Digital transmis-
sion systemsembodied by telegraph systemswere developed in the 1850s before analog
systemsthe telephonein the 20th century. Now, digital transmission is again becoming
the preferred technique.

12 DIGITAL AND ANALOG SOURCES AND SYSTEMS

DEFINITION. A digital information source produces a finite set of possible messages.

A typewriter is a good example of a digital source. There is a finite number of char-
acters (messages) that can be emitted by this source.
DEFINITION. An analog information source produces messages that are defined on a
continuum.
A microphone is a good example of an analog source. The output voltage describes the
information in the sound, and it is distributed over a continuous range of values.
DEFINITION. A digital communication system transfers information from a digital
source to the intended receiver (also called the sink).
DEFINITION. An analog communication system transfers information from an analog
source to the sink.
Strictly speaking, a digital waveform is defined as a function of time that can have
only a discrete set of amplitude values. If the digital waveform is a binary waveform, only
two values are allowed. An analog waveform is a function of time that has a continuous
range of values.
An electronic digital communication system usually has voltage and current waveforms
that have digital values; however, it may have analog waveforms. For example, the information
from a binary source may be transmitted to the receiver by using a sine wave of 1,000 Hz to rep-
resent a binary 1 and a sine wave of 500 Hz to represent a binary 0. Here the digital source in-
formation is transmitted to the receiver by the use of analog waveforms, but the system is still
called a digital communication system. From this viewpoint, we see that a digital communica-
tion engineer needs to know how to analyze analog circuits as well as digital circuits.
Digital communication has a number of advantages:
Relatively inexpensive digital circuits may be used.
Privacy is preserved by using data encryption.
Greater dynamic range (the difference between the largest and smallest values) is possible.
Data from voice, video, and data sources may be merged and transmitted over a com-
mon digital transmission system.
In long-distance systems, noise does not accumulate from repeater to repeater.
Errors in detected data may be small, even when there is a large amount of noise on
the received signal.
Errors may often be corrected by the use of coding.
ch01.I 4/21/00 12:02 PM Page 6

6 Introduction Chap. 1

Digital communication also has disadvantages:

Generally, more bandwidth is required than that for analog systems.

Synchronization is required.

The advantages of digital communication systems usually outweigh their disadvantages.

Consequently, digital systems are becoming more and more popular.

13 DETERMINISTIC AND RANDOM WAVEFORMS

In communication systems, we are concerned with two broad classes of waveforms: deter-
ministic and random (or stochastic).
DEFINITION. A deterministic waveform can be modeled as a completely specified
function of time.
For example, if
w(t)=A cos (0 t+0 ) (11)
describes a waveform, where A, 0 , and 0 are known constants, this waveform is said to
be deterministic because, for any value of t, the value w(t) can be evaluated. If any of the
constants are unknown, then the value of w(t) cannot be calculated, and consequently, w(t)
is not deterministic.
DEFINITION. A random waveform (or stochastic waveform) cannot be completely
specified as a function of time and must be modeled probabilistically.
Here we are faced immediately with a dilemma when analyzing communication sys-
tems. We know that the waveforms that represent the source cannot be deterministic. For ex-
ample, in a digital communication system, we might send information corresponding to any
one of the letters of the English alphabet. Each letter might be represented by a determinis-
tic waveform, but when we examine the waveform that is emitted from the source, we find
that it is a random waveform because we do not know exactly which characters will be
transmitted. Consequently, we really need to design the communication system by using a
random signal waveform. Noise would also be described by a random waveform. This re-
quires the use of probability and statistical concepts (covered in Chapters 6 and 7) that make
the design and analysis procedure more complicated. However, if we represent the signal
waveform by a typical deterministic waveform, we can obtain most, but not all, of the re-
sults we are seeking. That is the approach taken in the first five chapters of this book.

14 ORGANIZATION OF THE BOOK

Chapters 1 to 5 use a deterministic approach in analyzing communication systems. This ap-
proach allows the reader to grasp some important concepts without the complications of sta-
tistical analysis. It also allows the reader who is not familiar with statistics to obtain a basic

A more complete definition of a random waveform, also called a random process, is given in Chapter 6.
ch01.I 4/21/00 12:02 PM Page 7

Section 15 Use of a Personal Computer and MATLAB 7

understanding of communication systems. However, the important topic of performance of

communication systems in the presence of noise cannot be analyzed without the use of sta-
tistics. These topics are covered in Chapters 6 and 7 and Appendix B. Chapter 8 gives prac-
tical case studies of wire and wireless communication systems.
This textbook is designed to be reader friendly. To aid the student in problem solv-
ing, several study-aid problems with abbreviated solutions are included at the end of each
chapter. In addition, the personal computer (PC) is used to solve problems as appropriate.
The book is also useful as a reference source for mathematics (Appendix A), statistics
(Appendix B and Chapter 6), computer communication systems (Appendix C), and
MATLAB (Appendix D) and as a reference listing communication systems standards that
have been adopted (Chapters 3, 4, 5, and 8 and Appendix C).
Communications is an exciting area in which to work. The reader is urged to browse
through Chapter 8, looking at case-study topics that are of special interest, such as personal
communication systems (PCSs).

15 USE OF A PERSONAL COMPUTER AND MATLAB

This textbook is designed so that a PC may be used as a tool to plot waveforms; compute
spectra (using the fast Fourier transform); evaluate integrals; and, in general, help the read-
er to understand, analyze, and design communication systems. MATLAB was chosen as the
program language since it is very efficient at these tasks and a student version is available
at a reasonable cost. For a brief summary of MATLAB programming concepts, see Appen-
dix D2 (Programming in MATLAB).
To solve a problem using a MATLAB, the MATLAB program is first run on the PC.
MATLAB is an interpretive program. That is, results are computed after each line of code is
entered. One has the option of keying in MATLAB statements one line at a time for immediate
execution; or, alternately, a script file containing MATLAB code statements may be called up
and run by MATLAB. The script or text file is also called an M file because the filename has
the form xxxx.M. For programs with more than a couple of lines of code, the M-file method is
usually used. The computed results may be shown in tabulated or graphical form. The M-files
may be created by the MATLAB text editor or by another text editor, such as Notepad (running
under Windows on the PC).
M-files are provided for solving selected equations and study-aid problems. The se-
lected equation are marked with a PC ( ) symbol. The M-files can be downloaded from
the World Wide Web at the Internet site
http:www.couch.ece.ufl.edu
or
http:www.prenhall.comcouch
or, by using anonymous file transfer protocol (FTP) from
ftp.ece.ufl.edu
Appendix B covers the topic of probability and random variables and is a complete chapter in itself. This

allows the reader who has not had a course on this topic to learn the material before Chapters 6 and 7 are studied.
ch01.I 4/21/00 12:02 PM Page 8

8 Introduction Chap. 1

under the subdirectory pubCOUCH6edMATLAB. Alternatively, the M-files may be ob-

tained from your instructor. MATHCAD template files can also be downloaded from the pre-
viouly listed sites. (Additional MATLAB and MATHCAD files for the homework problems
marked with the PC ( ) symbol are made available to the instructor on a floppy disk that
is included with the Solutions Manual.) For instructions on running the M-files, see Appen-
dix D1 (Quick Start for Running M-files). As an example, run the file e1_006.m, which
computes the results for Equation (16) of this chapter. Table 2_3.m, is the M-file shown in
Table 23 and produces the MATLAB plots shown in Fig. 221.

16 BLOCK DIAGRAM OF A COMMUNICATION SYSTEM

Communication systems may be described by the block diagram shown in Fig. 11. Re-
gardless of the particular application, all communications systems involve three main sub-
systems: the transmitter, the channel, and the receiver. Throughout this book, we use the
symbols as indicated in this diagram so that the reader will not be confused about where the
signals are located in the overall system. The message from the source is represented by the
information input waveform m(t ). The message delivered by the receiver is denoted by

m (t). The [  ] indicates that the message received may not be the same as that transmit-

ted. That is, the message at the sink, m (t ), may be corrupted by noise in the channel, or
there may be other impairments in the system, such as undesired filtering or undesired non-
linearities. The message information may be in analog or digital form, depending on the par-
ticular system, and it may represent audio, video, or some other type of information. In
multiplexed systems, there may be multiple input and output message sources and sinks. The

spectra (or frequencies) of m(t) and m (t) are concentrated about f=0; consequently, they
are said to be baseband signals.
The signal-processing block at the transmitter conditions the source for more efficient
transmission. For example, in an analog system, the signal processor may be an analog low-
pass filter that is used to restrict the bandwidth of m(t ). In a hybrid system, the signal
processor may be an analog-to-digital converter (ADC), which produces a digital word that
represents samples of the analog input signal (as described in Chapter 3 in the section on
pulse code modulation). In this case, the ADC in the signal processor is providing source
coding of the input signal. In addition, the signal processor may add parity bits to the dig-
ital word to provide channel coding so that error detection and correction can be used by
the signal processor in the receiver to reduce or eliminate bit errors that are caused by noise

Noise
TRANSMITTER n(t) RECEIVER

Information
s( t) Transmission r( t ) ~ t)
m(
input Signal Carrier Carrier Signal
medium
m(t) processing circuits circuits processing
(channel) To information
sink (user)

Figure 11 Communication system.

ch01.I 4/21/00 12:02 PM Page 9

Section 16 Block Diagram of a Communication System 9

in the channel. The signal at the output of the transmitter signal processor is a baseband sig-
nal, because it has frequencies concentrated near f=0.
The transmitter carrier circuit converts the processed baseband signal into a frequen-
cy band that is appropriate for the transmission medium of the channel. For example, if the
channel consists of a fiber-optic cable, the carrier circuits convert the baseband input Ai.e.,
frequencies near f=0B to light frequencies, and the transmitted signal, s(t), is light. If the
channel propagates baseband signals, no carrier circuits are needed, and s(t) can be the out-
put of the processing circuit at the transmitter. Carrier circuits are needed when the trans-
mission channel is located in a band of frequencies around fc  0. (The subscript denotes
carrier frequency.) In this case, s(t ) is said to be a bandpass, because it is designed to
have frequencies located in a band about fc . For example, an amplitude-modulated (AM)
broadcasting station with an assigned frequency of 850 kHz has a carrier frequency of
fc =850 kHz. The mapping of the baseband input information waveform m(t ) into the
bandpass signal s(t) is called modulation. [m(t) is the audio signal in AM broadcasting.]
In Chapter 4, it will be shown that any bandpass signal has the form
s(t )=R(t) cos[c t  (t)] (12)
where c=2fc. If R(t)=1 and (t)=0, s(t) would be a pure sinusoid of frequency
f=fc with zero bandwidth. In the modulation process provided by the carrier circuits,
the baseband input waveform m(t) causes R(t) or (t) or both to change as a function of
m(t). These fluctuations in R(t) and (t) cause s(t) to have a nonzero bandwidth that de-
pends on the characteristics of m(t) and on the mapping functions used to generate R(t)
and (t). In Chapter 5, practical examples of both digital and analog bandpass signaling are
presented.
Channels may be classified into two categories: wire and wireless. Some examples of
wire channels are twisted-pair telephone lines, coaxial cables, waveguides, and fiber-optic
cables. Some typical wireless channels are air, vacuum, and seawater. Note that the gener-
al principles of digital and analog modulation apply to all types of channels, although chan-
nel characteristics may impose constraints that favor a particular type of signaling. In
general, the channel medium attenuates the signal so that the noise of the channel or the

noise introduced by an imperfect receiver causes the delivered information m to be deteri-
orated from that of the source. The channel noise may arise from natural electrical distur-
bances (e.g., lightning) or from artificial sources, such as high-voltage transmission lines,
ignition systems of cars, or switching circuits of a nearby digital computer. The channel may
contain active amplifying devices, such as repeaters in telephone systems or satellite
transponders in space communication systems. These devices are necessary to help keep the
signal above the noise level. In addition, the channel may provide undesirable multiple paths
between its input and output that have different time delays and attenuation characteristics.
Even worse, these characteristics may vary with time, which makes the signal fade at the
channel output. You have probably observed this type of fading when listening to distant
shortwave stations.
The receiver takes the corrupted signal at the channel output and converts it to a base-
band signal that can be handled by the receiver baseband processor. The baseband proces-

sor cleans up this signal and delivers an estimate of the source information m (t) to the
communication system output.
ch01.I 4/21/00 12:02 PM Page 10

10 Introduction Chap. 1

The goal is to design communication systems that transmit information to the receiv-
er with as little deterioration as possible while satisfying design constraints, of allowable
transmitted energy, allowable signal bandwidth, and cost. In digital systems, the measure of
deterioration is usually taken to be the probability of bit error (Pe)also called the bit er-

ror rate (BER)of the delivered data m . In analog systems, the performance measure is
usually taken to be the signal-to-noise ratio at the receiver output.

17 FREQUENCY ALLOCATIONS
Wireless communication systems often use the atmosphere for the transmission channel.
Here, interference and propagation conditions are strongly dependent on the transmission fre-
quency. Theoretically, any type of modulation (e.g., amplitude modulation, frequency mod-
ulation, single sideband, phase-shift keying, frequency-shift keying, etc.) could be used at
any transmission frequency. However, to provide some semblance of order and to minimize
interference, government regulations specify the modulation type, bandwidth, power, and
type of information that a user can transmit over designated frequency bands.
Frequency assignments and technical standards are set internationally by the Inter-
national Telecommunications Union (ITU). The ITU is a specialized agency of the Unit-
ed Nations, and the ITU administrative headquarters is located in Geneva, Switzerland,
with a staff of about 700 persons (see http:www.itu.ch). This staff is responsible for ad-
ministering the agreements that have been ratified by about 200 member nations of the
ITU. The ITU is structured into three sectors. The Radiocommunication Sector (ITU-R)
provides frequency assignments and is concerned with the efficient use of the radio fre-
quency spectrum. The Telecommunications Standardization Section (ITU-T) examines
technical, operating, and tariff questions. It recommends worldwide standards for the pub-
lic telecommunications network (PTN) and related radio systems. The Telecommunication
Development Sector (ITU-D) provides technical assistance, especially for developing
countries. This assistance encourages a full array of telecommunication services to be eco-
nomically provided and integrated into the worldwide telecommunication system. Before
1992, the ITU was organized into two main sectors: the International Telegraph and Tele-
phone Consultative Committee (CCITT) and the International Radio Consultative Com-
mittee (CCIR).
Each member nation of the ITU retains sovereignty over the spectral usage and stan-
dards adopted in its territory. However, each nation is expected to abide by the overall fre-
quency plan and standards that are adopted by the ITU. Usually, each nation establishes an
agency that is responsible for the administration of the radio frequency assignments within its
borders. In the United States, the Federal Communications Commission (FCC) regulates and
licenses radio systems for the general public and state and local government (see
http:www.fcc.gov). In addition, the National Telecommunication and Information
Administration (NTIA) is responsible for U.S. government and U.S. military frequency as-
signments. The international frequency assignments are divided into subbands by the FCC to
accommodate 70 categories of services and 9 million transmitters. Table 12 gives a general
listing of frequency bands, their common designations, typical propagation conditions, and
typical services assigned to these bands.
ch01.I 4/21/00 12:02 PM Page 11

Section 17 Frequency Allocations 11

TABLE 12 FREQUENCY BANDS

Frequency Propagation
Band a Designation Characteristics Typical Uses

330 kHz Very low Ground wave; low Long-range navigation;

frequency (VLF) attenuation day and submarine
night; high atmospheric communication
noise level
30300 kHz Low frequency Similar to VLF, slightly Long-range navigation and
(LF) less reliable; absorption marine communication
in daytime radio beacons
3003000 kHz Medium frequency Ground wave and night Maritime radio, direction
(MF) sky wave; attenuation finding, and AM
low at night and high in broadcasting
day; atmospheric noise
330 Mhz High frequency Ionospheric reflection Amateur radio;
(HF) varies with time of day, international broadcasting,
season, and frequency; military communication,
low atmospheric noise long- distance aircraft and
at 30 Mhz ship communication,
telephone, telegraph,
facsimile
30300 MHz Very high Nearly line-of-sight (LOS) VHF television, FM two-
frequency (VHF) propagation, with way radio, AM aircraft
scattering because of communication, aircraft
temperature inversions, navigational aids
cosmic noise
0.33 GHz Ultrahigh LOS propagation, cosmic UHF television, cellular
frequency (UHF) noise telephone, navigational
aids, radar, GPS,
microwave links, personal
communication systems
Letter designation
1.02.0 L
2.04.0 S
330 GHz Superhigh LOS propagation; rainfall Satellite communication,
frequency (SHF) attenuation above 10 radar microwave links
GHz, atmospheric
Letter designation attenuation because of
2.04.0 S oxygen and water
4.08.0 C vapor, high water vapor
8.012.0 X absorption at 22.2 GHz
12.018.0 Ku
18.027.0 K
27.040.0 Ka
26.540.0 R
30300 GHz Extremely high Same; high water-vapor Radar, satellite,
frequency (EHF) absorption at 183 GHz experimental
and oxygen absorption
at 60 and 119 GHz
a kHz = 10 3 Hz; MHz = 10 6 Hz; GHz = 10 9 Hz.
ch01.I 4/21/00 12:02 PM Page 12

12 Introduction Chap. 1

TABLE 12 FREQUENCY BANDS (cont.)

Frequency Propagation
Band a Designation Characteristics Typical Uses

Letter designation
27.040.0 Ka
26.540.0 R
33.050.0 Q
40.075.0 V
75.0110.0 W
110300 mm (millimeter)
10 310 7 GHz Infrared, visible LOS propagation Optical communications
light, and
ultraviolet

18 PROPAGATION OF ELECTROMAGNETIC WAVES

The propagation characteristics of electromagnetic waves used in wireless channels are high-
ly dependent on the frequency. This situation is shown in Table 12, where users are as-
signed frequencies that have the appropriate propagation characteristics for the coverage
needed. The propagation characteristics are the result of changes in the radio-wave velocity
as a function of altitude and boundary conditions. The wave velocity is dependent on air tem-
perature, air density, and levels of air ionization.
Ionization (i.e., free electrons) of the rarified air at high altitudes has a dominant ef-
fect on wave propagation in the medium-frequency (MF) and high-frequency (HF) bands.
The ionization is caused by ultraviolet radiation from the sun, as well as cosmic rays. Con-
sequently, the amount of ionization is a function of the time of day, season of the year, and
activity of the sun (sunspots). This results in several layers of varying ionization density lo-
cated at various heights surrounding the Earth.
The dominant ionized regions are D, E, F1, and F2 layers. The D layer is located clos-
est to the Earths surface at an altitude of about 45 or 55 miles. For f>300 kHz, the D lay-
er acts as a radio-frequency (RF) sponge to absorb (or attenuate) these radio waves. The
attenuation is inversely proportional to frequency and becomes small for frequencies above
4 MHz. For f<300 kHz, the D layer provides refraction (bending) of RF waves. The D
layer is most pronounced during the daylight hours, with maximum ionization when the sun
is overhead, and almost disappears at night. The E layer has a height of 65 to 75 miles, has
maximum ionization around noon (local time), and practically disappears after sunset. It pro-
vides reflection of HF frequencies during the daylight hours. The F layer ranges in altitude
between 90 and 250 miles. It ionizes rapidly at sunrise, reaches its peak ionization in early
afternoon, and decays slowly after sunset. The F region splits into two layers, F1 and F2, dur-
ing the day and combines into one layer at night. The F region is the most predominant
medium in providing reflection of HF waves. As shown in Fig. 12, the electromagnetic
spectrum may be divided into three broad bands that have one of three dominant propaga-
tion characteristics: ground wave, sky wave, and line of sight (LOS).
ch01.I 4/21/00 12:02 PM Page 13

Section 1-8 Propagation of Electromagnetic Waves 13

Signal propagation

Transmit Receive
antenna antenna

Earth

(a) GroundWave Propagation (Below 2 MHz)

Signal propagation

Ionosphere

Transmit Earth
antenna Receive
antenna

(b) SkyWave Propagation (2 to 30 MHz)

Signal propagation
Transmit Receive
antenna antenna

Tower
Earth

(c) Line-of-Sight (LOS) Propagation (Above 30 MHz)

Figure 12 Propagation of radio frequencies.

Ground-wave propagation is illustrated in Fig. 12a. It is the dominant mode of prop-

agation for frequencies below 2 MHz. Here, the electromagnetic wave tends to follow the
contour of the Earth. That is, diffraction of the wave causes it to propagate along the sur-
face of the Earth. This is the propagation mode used in AM broadcasting, where the local
coverage follows the Earths contour and the signal propagates over the visual horizon. The
ch01.I 4/21/00 12:02 PM Page 14

14 Introduction Chap. 1

following question is often asked: What is the lowest radio frequency that can be used? The
answer is that the value of the lowest useful frequency depends on how long you want to
make the antenna. For efficient radiation, the antenna needs to be longer than one-tenth of
a wavelength. For example, for signaling with a carrier frequency of fc=10 kHz, the wave-
length is
c
=
fc

(3
108 m  s)
= =3*10 4 m (13)
104
where c is the speed of light. (The formula =cfc is distance=velocity*time, where
the time needed to traverse one wavelength is t=lfc.) Thus, an antenna needs to be at
least 3,000 m in length for efficient electromagnetic radiation at 10 kHz.
Sky-wave propagation is illustrated in Fig. 12b. It is the dominant mode of propaga-
tion in the 2- to 30-MHz frequency range. Here, long-distance coverage is obtained by re-
flecting the wave at the ionosphere, and at the Earths boundaries. Actually, in the
ionosphere the waves are refracted (i.e., bent) gradually in an inverted U shape, because the
index of refraction varies with altitude as the ionization density changes. The refraction in-
dex of the ionosphere is given by [Griffiths, 1987; Jordan and Balmain, 1968]

81N
n= 1 (14)
f2

where n is the refractive index, N is the free-electron density (number of electrons per cubic
meter), and f is the frequency of the wave (in hertz). Typical N values range between 10 10
and 10 12, depending on the time of day, the season, and the number of sunspots. In an ion-
ized region n<1 because N>0, and outside the ionized region n L 1 because N L 0.
In the ionized region, because n<1, the waves will be bent according to Snells law; viz,

n sin r=sin i (15)

where i is the angle of incidence (between the wave direction and vertical), measured just
below the ionosphere, and r is the angle of refraction for the wave (from vertical), mea-
sured in the ionosphere. Furthermore, the refraction index will vary with altitude within the
ionosphere because N varies. For frequencies selected from the 2- to 30-MHz band, the re-
fraction index will vary with altitude over the appropriate range so that the wave will be bent
back to Earth. Consequently, the ionosphere acts as a reflector. The transmitting station will
have coverage areas as indicated in Fig. 12b by heavy black lines along the Earths surface.
The coverage near the transmit antenna is due to the ground-wave mode, and the other cov-
erage areas are due to sky wave. Notice that there are areas of no coverage along the Earths
surface between the transmit and receive antennas. The angle of reflection and the loss of
signal at an ionospheric reflection point depend on the frequency, the time of day, the sea-
son of the year, and the sunspot activity [Jordan, 1985, Chap. 33].
During the daytime (at the ionospheric reflection points), the electron density will be
high, so that n<1. Consequently, sky waves from distant stations on the other side of the
ch01.I 4/21/00 12:02 PM Page 15

Section 1-8 Propagation of Electromagnetic Waves 15

world will be heard on the shortwave bands. However, the D layer is also present during
the day. This absorbs frequencies below 4 MHz.
This is the case for AM broadcast stations, where distant stations cannot be heard dur-
ing the day, but at night the layer disappears, and distant AM stations can be heard via sky-
wave propagation. In the United States, the FCC has designated some frequencies within
the AM band as clear channels (as shown in Table 51). On these channels, only one or
two high-power 50-kw stations are assigned to operate at night, along with a few low-pow-
er stations. Since these channels are relatively free of interfering stations, night sky-wave sig-
nals of the dominant 50-kw station can often be heard at distances up to 1,500 miles from
the station. For example, some clear-channel 50-kw stations are WSM, Nashville, on 650
kHz; WCCO, Minneapolis, on 830 kHz; and WHO, Des Moines, on 1040 kHz.
Sky-wave propagation is caused primarily by reflection from the F layer (90 to 250
miles in altitude). Because of this layer, international broadcast stations in the HF band can
be heard from the other side of the world almost anytime during the day or night.
LOS propagation (illustrated in Fig. 12c) is the dominant mode for frequencies above
30 MHz. Here, the electromagnetic wave propagates in a straight line. In this case, f 2  81N,
so that n L 1, and there is very little refraction by the ionosphere. In fact, the signal will prop-
agate through the ionosphere. This property is used for satellite communications.
The LOS mode has the disadvantage that, for communication between two terrestrial
(Earth) stations, the signal path has to be above the horizon. Otherwise, the Earth will block
the LOS path. Thus, antennas need to be placed on tall towers so that the receiver antenna
can see the transmitting antenna. A formula for the distance to the horizon, d, as a func-
tion of antenna height can be easily obtained by the use of Fig. 13. From this figure,

d 2+r 2=(r+h) 2
or
d 2=2rh+h 2

where r is the radius of the Earth and h is the height of the antenna above the Earths sur-
face. In this application, h 2 is negligible with respect to 2rh. The radius of the Earth is 3,960

Tangent to surface of Earth

d

r
r

Earth
Figure 13 Calculation of distance to horizon.
ch01.I 4/21/00 12:02 PM Page 16

16 Introduction Chap. 1

4
statute miles. However, at LOS radio frequencies the effective Earth radius is 3 (3,960)
miles. Thus, the distance to the radio horizon is
d= 2h miles (16)
where conversion factors have been used so that h is the antenna height measured in feet and
d is in statute miles. For example, television stations have assigned frequencies above
30 MHz in the VHF or UHF range (see Table 12), and the fringe-area coverage of high-
power stations is limited by the LOS radio horizon. For a television station with a 1,000-ft
tower, d is 44.7 miles. For a fringe-area viewer who has an antenna height of 30 ft, d is
7.75 miles. Thus, for these transmitting and receiving heights, the television station would
have fringe-area coverage out to a radius of 44.7+7.75=52.5 miles around the transmit-
ting tower.
In addition to the LOS propagation mode, it is possible to have ionospheric scatter propa-
gation. This mode occurs over the frequency range of 30 to 60 MHz, when the radio frequency
signal is scattered because of irregularities in the refractive index of the lower ionosphere (about
50 miles above the Earths surface). Because of the scattering, communications can be carried out
over path lengths of 1,000 miles, even though that is beyond the LOS distance. Similarly,
tropospheric scattering (within 10 miles above the Earths surface) can propagate radio frequen-
cy signals that are in the 40-MHz to 4-GHz range over paths of several hundred miles.
For more technical details about radio-wave propagation, the reader is referred to text-
books that include chapters on ground-wave and sky-wave propagation [Griffiths, 1987; Jor-
dan and Balmain, 1968] and to a radio engineering handbook [Jordan, 1985]. A very
readable description of this topic is also found in the ARRL handbook [ARRL, 1997], and
personal computer programs, (e.g., MINIMUF, and PropMan) that predict sky-wave propa-
gation conditions [Rose, 1982, 1984; Rockwell, 1995] are available.

19 INFORMATION MEASURE
As we have seen, the purpose of communication systems is to transmit information from a
source to a receiver. However, what exactly is information, and how do we measure it? We
know qualitatively that it is related to the surprise that is experienced when we receive the
message. For example, the message The ocean has been destroyed by a nuclear explosion
contains more information than the message It is raining today.

DEFINITION. The information sent from a digital source when the jth message is
transmitted is given by

Ij=log2  P1  bits
j
(17a)

where Pj is the probability of transmitting the jth message.

The refractive index of the atmosphere decreases slightly with height, which causes some bending of ra-

dio rays. This effect may be included in LOS calculations by using an effective Earth radius that is four-thirds of
the actual radius.
The definition of probability is given in Appendix B.
ch01.I 4/21/00 12:02 PM Page 17

Section 19 Information Measure 17

From this definition, we see that messages that are less likely to occur (smaller value
for Pj) provide more information (larger value of Ij). We also observe that the information
measure depends on only the likelihood of sending the message and does not depend on pos-
sible interpretation of the content as to whether or not it makes sense.
The base of the logarithm determines the units used for the information measure. Thus,
for units of bits, the base 2 logarithm is used. If the natural logarithm is used, the units
are nats and for base 10 logarithms, the unit is the hartley, named after R. V. Hartley,
who first suggested using the logarithm measure in 1928 [Hartley, 1948].
In this section, the term bit denotes a unit of information as defined by Eq. (17a). In lat-
er sections, particularly in Chapter 3, bit is also used to denote a unit of binary data.These two
different meanings for the word bit should not be confused. Some authors use binit to denote
units of data and use bit exclusively to denote units of information. However, most engineers use
the same word (bit) to denote both kinds of units, with the particular meaning understood from
the context in which the word is used. This book follows that industry custom.
For ease of evaluating Ij on a calculator, Eq. (17a) can be written in terms of the
base 10 logarithm or the natural logarithm:
1 1
Ij= log10 Pj= ln Pj (17b)
log10 2 ln 2
In general, the information content will vary from message to message because the Pjs
will not be equal. Consequently, we need an average information measure for the source,
considering all the possible messages we can send.

DEFINITION. The average information measure of a digital source is

m m
H=  PjIj=  Pj log2
j1 j1
 P1  bits
j
(18)

where m is the number of possible different source messages and Pj is the probability
of sending the jth message (m is finite because a digital source is assumed). The av-
erage information is called entropy.

Example 11 EVALUATION OF INFORMATION AND ENTROPY

Find the information content of a message that consists of a digital word 12 digits long in which
each digit may take on one of four possible levels. The probability of sending any of the four
levels is assumed to be equal, and the level in any digit does not depend on the values taken on
by previous digits.
In a string of 12 symbols (digits), where each symbol consists of one of four levels, there
are 4  4 p 4=4 12 different combinations (words) that can be obtained. Because each level
is equally likely, all the different words are equally likely. Thus,


1 1 12
Pj= 12

4 4
or

Ij=log2  A 1B =12 log (4)=24 bits

1 12
4
2
ch01.I 4/21/00 12:02 PM Page 18

18 Introduction Chap. 1

In this example, we see that the information content in every one of the possible messages equals
24 bits. Thus, the average information H is 24 bits.
Suppose that only two levels (binary) had been allowed for each digit and that all the
words were equally likely. Then the information would be Ij=12 bits for the binary words, and
the average information would be H=12 bits. Here, all the 12-bit words gave 12 bits of in-
formation, because the words were equally likely. If they had not been equally likely, some of
the 12-bit words would contain more than 12 bits of information and some would contain less,
and the average information would have been less than 12 bits. For example, if half of the 12-
bit words (2,048 of the possible 4,096) have probability of occurence of Pj=105 for each of
these words (with a corresponding Ij=16.61 bits) and the other half have Pj=4.78*104
(for a corresponding Ij=11.03 bits), then the average information is H=11.14 bits.

The rate of information is also important.

DEFINITION. The source rate is given by
H
R= bitss (19)
T
where H is evaluated by using Eq. (18) and T is the time required to send a message.
The definitions previously given apply to digital sources. Results for analog sources
can be approximated by digital sources with as much accuracy as we desire.

110 CHANNEL CAPACITY

AND IDEAL COMMUNICATION SYSTEMS
Many criteria can be used to measure the effectiveness of a communication system to see if
it is ideal or perfect. For digital systems, the optimum system might be defined as the sys-
tem that minimizes the probability of bit error at the system output subject to constraints on
transmitted energy and channel bandwidth. Thus, bit error and signal bandwidth are of prime
importance and are covered in subsequent chapters. This raises the following question: Is it
possible to invent a system with no bit error at the output even when we have noise intro-
duced into the channel? This question was answered by Claude Shannon in 19481949
[Wyner and Shamai, 1998; Shannon, 1948, 1949]. The answer is yes, under certain as-
sumptions. Shannon showed that (for the case of signal plus white Gaussian noise) a chan-
nel capacity C (bitss) could be calculated such that if the rate of information R (bitss)
was less than C, the probability of error would approach zero. The equation for C is

 
S
C=B log2 1  (110)
N
where B is the channel bandwidth in hertz (Hz) and SN is the signal-to-noise power ratio
(wattswatts, not dB) at the input to the digital receiver. Shannon does not tell us how to
build this system, but he proves that it is theoretically possible to have such a system. Thus,
Shannon gives us a theoretical performance bound that we can strive to achieve with prac-
tical communication systems. Systems that approach this bound usually incorporate error-
correction coding.
ch01.I 4/21/00 12:02 PM Page 19

Section 111 Coding 19

In analog systems, the optimum system might be defined as the one that achieves the
largest signal-to-noise ratio at the receiver output, subject to design constraints such as chan-
nel bandwidth and transmitted power. Here, the evaluation of the output signal-to-noise ra-
tio is of prime importance. We might ask the question, Is it possible to design a system with
infinite signal-to-noise ratio at the output when noise is introduced by the channel? The an-
swer is no. The performance of practical analog systems with respect to that of Shannons
ideal system is illustrated in Chapter 7. (See Fig. 727.)
Other fundamental limits for digital signaling were discovered by Nyquist in 1924 and
Hartley in 1928. Nyquist showed that if a pulse represents one bit of data, noninterfering
pulses could be sent over a channel no faster than 2B pulsess, where B is the channel band-
width in hertz. This is now known as the dimensionality theorem and is discussed in Chap-
ter 2. Hartley generalized Nyquists result for the case of multilevel pulse signaling, as
discussed in Chapters 3 and 5.
The following section describes the improvement that can be obtained in digital systems
when coding is used and how these coded systems compare with Shannons ideal system.

111 CODING

If the data at the output of a digital communication system have errors that are too frequent
for the desired use, the errors can often be reduced by the use of either of two main
techniques:

Automatic repeat request (ARQ)

Forward error correction (FEC)

In an ARQ system, when a receiver circuit detects parity errors in a block of data, it requests
that the data block be retransmitted. In an FEC system, the transmitted data are encoded so
that the receiver can correct, as well as detect, errors. These procedures are also classified
as channel coding because they are used to correct errors caused by channel noise. This is
different from source coding, described in Chapter 3, where the purpose of the coding is to
extract the essential information from the source and encode it into digital form so that it can
be efficiently stored or transmitted using digital techniques.
The choice between using the ARQ or the FEC technique depends on the particular
application. ARQ is often used in computer communication systems because it is relatively
inexpensive to implement and there is usually a duplex (two-way) channel so that the re-
ceiving end can transmit back an acknowledgment (ACK) for correctly received data or a
request for retransmission (NAC) when the data are received in error. (See Appendix C, Sec-
tion C4, for examples of ARQ signaling.) FEC techniques are used to correct errors on
simplex (one-way) channels, where returning of an ACKNAC indicator (required for the
ARQ technique) is not feasible. FEC is preferred on systems with large transmission delays,
because if the ARQ technique were used, the effective data rate would be small; the trans-
mitter would have long idle periods while waiting for the ACKNAC indicator, which is re-
tarded by the long transmission delay. Since ARQ systems are covered in Appendix C, we
concentrate on FEC techniques in the remainder of this section.
ch01.I 4/21/00 12:02 PM Page 20

20 Introduction Chap. 1

Noise
Transmitter

Encoder g(t) s(t) r(t)

Digital m Carrier Transmission medium
and other
source signal processing circuits (channel)

~ Decoder
r(t) g(t) ~
Carrier m Digital
and other
circuits signal processing sink

Receiver
Figure 14 General digital communication system.

Communication systems with FEC are illustrated in Fig. 14, where encoding and de-
coding blocks have been designated. Coding involves adding extra (redundant) bits to the
data stream so that the decoder can reduce or correct errors at the output of the receiver.
However, these extra bits have the disadvantage of increasing the data rate (bitss) and, con-
sequently, increasing the bandwidth of the encoded signal.
Codes may be classified into two broad categories:
Block codes. A block code is a mapping of k input binary symbols into n output bina-
ry symbols. Consequently, the block coder is a memoryless device. Because n>k,
the code can be selected to provide redundancy, such as parity bits, which are used by
the decoder to provide some error detection and error correction. The codes are de-
noted by (n, k), where the code rate R is defined by R=kn. Practical values of R
range from 41 to 78, and k ranges from 3 to several hundred [Clark and Cain, 1981].
Convolutional codes. A convolutional code is produced by a coder that has memory.
The convolutional coder accepts k binary symbols at its input and produces n binary
symbols at its output, where the n output symbols are affected by v+k input sym-
bols. Memory is incorporated because v>0. The code rate is defined by R=kn.
Typical values for k and n range from 1 to 8, and the values for v range from 2 to 60.
The range of R is between 14 and 78 [Clark and Cain, 1981]. A small value for the code
rate R indicates a high degree of redundancy, which should provide more effective er-
ror control at the expense of increasing the bandwidth of the encoded signal.

Block Codes
Before discussing block codes, several definitions are needed. The Hamming weight of a
code word is the number of binary 1 bits. For example, the code word 110101 has a Ham-
ming weight of 4. The Hamming distance between two code words, denoted by d, is the
number of positions by which they differ. For example, the code words 110101 and 111001

Do not confuse the code rate (with units of bitsbits) with the data rate or information rate (which has units

of bitss).
ch01.I 4/21/00 12:02 PM Page 21

Section 111 Coding 21

have a distance of d=2. A received code word can be checked for errors. Some of the er-
rors can be detected and corrected if d s+t+1, where s is the number of errors that
can be detected and t is the number of errors that can be corrected (s t). Thus, a pattern
of t or fewer errors can be both detected and corrected if d 2t+1.
A general code word can be expressed in the form
i1i2i3  ikp1p2p3  pr
where k is the number of information bits, r is the number of parity check bits, and n is the to-
tal word length in the (n, k) block code, where n=k+r. This arrangement of the informa-
tion bits at the beginning of the code word followed by the parity bits is most common. Such a
block code is said to be systematic. Other arrangements with the parity bits interleaved between
the information bits are possible and are usually considered to be equivalent codes.
Hamming has given a procedure for designing block codes that have single error-
correction capability [Hamming, 1950]. A Hamming code is a block code having a Hamming
distance of 3. Because d 2t+1, t=1, and a single error can be detected and corrected.
However, only certain (n, k) codes are allowable. These allowable Hamming codes are

(n, k)=(2 m-1, 2 m-1-m) (111)

where m is an integer and m 3. Thus, some of the allowable codes are (7, 4), (15, 11),
(31, 26), (63, 57), and (127, 120). The code rate R approaches 1 as m becomes large.
In addition to Hamming codes, there are many other types of block codes. One popu-
lar class consists of the cyclic codes. Cyclic codes are block codes, such that another code
word can be obtained by taking any one code word, shifting the bits to the right, and plac-
ing the dropped-off bits on the left. These types of codes have the advantage of being easi-
ly encoded from the message source by the use of inexpensive linear shift registers with
feedback. This structure also allows these codes to be easily decoded. Examples of cyclic
and related codes are BoseChaudhuriHocquenhem (BCH), Reed-Solomon, Hamming,
maximallength, ReedMller, and Golay codes. Some properties of block codes are given
in Table 13 [Bhargava, 1983].

TABLE 13 PROPERTIES OF BLOCK CODES

Code a

Property BCH Reed-Solomon Hamming Maximal Length

Block length n=2 m-1 n=m(2 m-1) bits n=2 m-1 n=2 m-1
m=3, 4, 5, p
Number of r=m2t bits r=m
parity bits
Minimum d 2t+1 d=m(2t+1) bits d=3 d=2 m-1
distance
Number of k n-mt k=m
information bits

am is any positive integer unless otherwise indicated; n is the block length; k is the number of information bits.
ch01.I 4/21/00 12:02 PM Page 22

22 Introduction Chap. 1

Convolutional Codes
A convolutional encoder is illustrated in Fig. 15. Here k bits (one input frame) are shifted
in each time, and, concurrently, n bits (one output frame) are shifted out, where n>k.
Thus, every k-bit input frame produces an n-bit output frame. Redundancy is provided in
the output, because n>k. Also, there is memory in the coder, because the output frame
depends on the previous K input frames, where K>1. The code rate is R=kn, which
is 43 in this illustration. The constraint length, K, is the number of input frames that are held
in the kK-bit shift register. Depending on the particular convolutional code that is to be gen-
erated, data from the kK stages of the shift register are added (modulo 2) and used to set the
bits in the n-stage output register.
For example, consider the convolutional coder shown in Fig. 16. Here, k=1, n=2,
K=3, and a commutator with two inputs performs the function of a two-stage output shift
register. The convolutional code is generated by inputting a bit of data and then giving the
commutator a complete revolution. The process is repeated for successive input bits to pro-
duce the convolutionally encoded output. In this example, each k=1 input bit produces
n=2 output bits, so the code rate is R=kn=12. The code tree of Fig. 17 gives the en-
coded sequences for the convolutional encoder example of Fig. 16. To use the code tree,
one moves up if the input is a binary 0 and down if the input is a binary 1. The corre-

Convolutional encoder

k bit frame Constraint length=K input frames

Shift register
(kK bits)

Uncoded input data

Logic

n bit frame

Shift register
(n bits)
Coded output data
n bits

Figure 15 Convolutional encoding Ak=3, n=4, K=5, and R=34B.

Several different definitions of constraint length are used in the literature [Blahut, 1983; Clark and Cain,
1981; Proakis, 1995].
ch01.I 4/21/00 12:02 PM Page 23

Section 111 Coding 23

Constraint length = K = 3

Three stage shift register

Data = x
S1 S2 S3
Input

Modulo 2 adders

Commutator

Convolutional code
Output

Figure 16 Convolutional encoder for a rate 12, constraint length 3 code.

sponding encoded bits are shown in parentheses. For example, if the input sequence
x11=1010 is fed into the input (with the most recent input bit on the right), the corre-
sponding encoded output sequence is y11=11010001, as shown by path A in Fig. 17.
A convolutionally encoded signal is decoded by matching the encoded received data
to the corresponding bit pattern in the code tree. In sequential decoding (a suboptimal tech-
nique), the path is found like that of a driver who occasionally makes a wrong turn at a fork
in a road but discovers the mistake, goes back, and tries another path. For example, if
y11=11010001 was received, path A would be the closest match, and the decoded data
would be x11=1010. If noise was present in the channel, some of the received encoded
bits might be in error, and then the paths would not match exactly. In this case, the match
is found by choosing a path that will minimize the Hamming distance between the selected
path sequence and the received encoded sequence.
An optimum decoding algorithm, called Viterbi decoding, uses a similar procedure. It
examines the possible paths and selects the best ones, based on some conditional probabili-
ties [Forney, 1973]. The Viterbi procedure can use either soft or hard decisions. A soft-
decision algorithm first decides the result on the basis of the test statistic being above or
below a decision threshold and then gives a confidence number that specifies how close
the test statistic was to the threshold value. In hard decisions, only the decision output is

The test statistic is a value that is computed at the receiver, based on the receiver input during some spec-

ified time interval. [See Eq. (74).]

ch01.I 4/21/00 12:02 PM Page 24

24 Introduction Chap. 1

x 1 = 0000
0 y 1 = 00000000
0 (00)
x 2 = 0001
(00) 1
y 2 = 00000011
0 (11)
x 3 = 0010
(00) 0 y 3 = 00001101
1 (01)
x 4 = 0011
(11) 1
y 4 = 00001110
0 (10)
(00) x 5 = 0100
0 y 5 = 00110111
0 (11)
x 6 = 0101
(01) 1
y 6 = 00110100
1 (00)
x 7 = 0110
(11) 0 y 7 = 00111010
0 1 (10)
x 8 = 0111
(10) 1
y 8 = 00111001
(01)
x 9 = 1000
0 y 9 = 11011100
0 (00)
x 10 = 1001
(11) 1
1 y 10 = 11011111
(11)
0
x 11 = 1010
(01) 0 y 11 = 11010001
1 (01)
x 12 = 1011
Path A (00) 1
y 12 = 11010010
(10)
1
x 13 = 1100
(11) 0 y 13 = 11101011
0 (11)
x 14 = 1101
(10) 1
y 14 = 11101000
(00)
1
x 15 = 1110
(10) 0 y 15 = 11100110
1 (10)
x 16 = 1111
(01) 1
y 16 = 11100101
(01)

Figure 17 Code tree for convolutional encoder of Figure 16.

known, and it is not known if the decision was almost too close to call (because the test
value was almost equal to the threshold value). The soft-decision technique can translate in-
to a 2-dB improvement (decrease) in the required receiver input EbN0 [Clark and Cain,
1981]. Eb is the received signal energy over a 1-bit time interval, and N02 is the power
spectral density (PSD) of the channel noise at the receiver input. Both Eb and N0 will be
defined in detail in later chapters. [For example, see Eq. (724b) or Eq. (844).]
ch01.I 4/21/00 12:02 PM Page 25

Section 111 Coding 25

Code Interleaving
In the previous discussion, it was assumed that if no coding was used, the channel noise
would cause random bit errors at the receiver output that are more or less isolated (that is,
not adjacent). When coding was added, redundancy in the code allowed the receiver decoder
to correct the errors so that the decoded output was almost error free. However, in some ap-
plications, large, wide pulses of channel noise occur. If the usual coding techniques are used
in these situations, bursts of errors will occur at the decoder output because the noise bursts
are wider than the redundancy time of the code. This situation can be ameliorated by the
use of code interleaving.
At the transmitting end, the coded data are interleaved by shuffling (i.e., like shuf-
fling a deck of cards) the coded bits over a time span of several block lengths (for block
codes) or several constraint lengths (for convolutional codes). The required span length
is several times the duration of the noise burst. At the receiver, before decoding, the da-
ta with error bursts are deinterleaved to produce coded data with isolated errors. The iso-
lated errors are then corrected by passing the coded data through the decoder. This
produces almost error-free output, even when noise bursts occur at the receiver input.
There are two classes of interleaversblock interleavers and convolutional interleavers
[Sklar, 1988].

Code Performance
The improvement in the performance of a digital communication system that can be achieved
by the use of coding is illustrated in Fig. 18. It is assumed that a digital signal plus chan-
nel noise is present at the receiver input. The performance of a system that uses binary-
phase-shift-keyed (BPSK) signaling is shown both for the case when coding is used and for
the case when there is no coding. For the no-code case, the optimum (matched filter) detec-
tor circuit is used in the receiver, as derived in Chapter 7 and described by Eq. (738). For
the coded case, a (23, 12) Golay code is used. Pe is the probability of bit erroralso called
the bit error rate (BER)that is measured at the receiver output. EbN0 is the energy-
per-bitnoise-density ratio at the receiver input (as described in the preceding section). For
EbN0=7 dB, Fig. 18 shows that the BER is 10 3 for the uncoded case and that the BER
can be reduced to 10 5 if coding is used.
The coding gain is defined as the reduction in EbN0 (in decibels) that is achieved
when coding is used, when compared with the EbN0 required for the uncoded case at some
specific level of Pe . For example, as can be seen in the figure, a coding gain of 1.33 dB is
realized for a BER of 103. The coding gain increases if the BER is smaller, so that a cod-
ing gain of 2.15 dB is achieved when Pe=105. This improvement is significant in space
communication applications, where every decibel of improvement is valuable. The figure al-
so shows noted that there is a coding threshold in the sense that the coded system actually
provides poorer performance than the uncoded system when EbN0 is less than the thresh-
old value. In this example, the coding threshold is about 3.5 dB. A coding threshold is found
in all coded systems.
For optimum coding, Shannons channel capacity theorem, Eq. (110), gives the
EbN0 required. That is, if the source rate is below the channel capacity, the optimum code
ch01.I 4/21/00 12:02 PM Page 26

26 Introduction Chap. 1

101

102
Polar baseband or
Shannon's ideal system
BPSK without
with coding (114)
Turbo code coding (738)
Pe = Probability of bit error

[Sklar, 1997]
103
Coding gain = 1.33 dB

BPSK with
(23, 12) Golay coding
104

Coding gain = 2.15 dB

105
Coding gain = 8.8 dB
Ideal coding gain = 11.2 dB

106
2 1 0 1 2 3 4 5 6 7 8 9 10
Eb/N0 (dB)
1.59
Figure 18 Performance of digital systemswith and without coding.

will allow the source information to be decoded at the receiver with Pe S 0 (i.e., 10 q),
even though there is some noise in the channel. We will now find the EbN0 required so
that Pe S 0 with the optimum (unknown) code. Assume that the optimum encoded signal
is not restricted in bandwidth. Then, from Eq. (110),

C= lim
B
B log 1  NS  = lim B log 1  ENBT 
2
B
2
b

0
b

= lim
x0
 log [1  (Ex N T )x]
2 b 0 b

where Tb is the time that it takes to send one bit and N is the noise power that occurs with-
in the bandwidth of the signal. The power spectral density (PSD) is n(f)=N02, and, as
shown in Chapter 2, the noise power is

   
B B
N0
N= n(f)df= df=N0B (112)
B B 2
ch01.I 4/21/00 12:02 PM Page 27

Section 111 Coding 27

where B is the signal bandwidth. LHospitals rule is used to evaluate this limit:

C= lim
x0
 1  (E 1N T )x NET  log e  N TE ln 2
b 0 b
b

0 b
2
0 b
b
(113)

If we signal at a rate approaching the channel capacity, then Pe S 0, and we have the max-
imum information rate allowed for Pe S 0 (i.e., the optimum system). Thus, 1Tb=C,
or, using Eq. (113),

1 Eb

Tb N0 Tb ln 2

or

EbN0=ln 2=1.59 dB (114)

This minimum value for EbN0 is-1.59 dB and is called Shannons limit. That
is, if optimum codingdecoding is used at the transmitter and receiver, error-free data will
be recovered at the receiver output, provided that the EbN0 at the receiver input is larg-
er than 1.59 dB. This brick wall limit is shown by the dashed line in Fig. 18, where
Pe jumps from 0 (10 q) to 12 (0.5*100) as EbN0 becomes smaller than 1.59 dB,
assuming that the ideal (unknown) code is used. Any practical system will perform worse
than this ideal system described by Shannons limit. Thus, the goal of digital system de-
signers is to find practical codes that approach the performance of Shannons ideal (un-
known) code.
When the performance of the optimum encoded signal is compared with that of BPSK
without coding (10 5 BER), it is seen that the optimum (unknown) coded signal has a cod-
ing gain of 9.61-(1.59)=11.2 dB. Using Fig. 18, compare this value with the coding
gain of 8.8 dB that is achieved when a turbo code is used. Table 14 shows the gains that
can be obtained for some other codes.
Since their introduction in 1993, turbo codes have become very popular because they
can perform near Shannons limit, yet they also can have reasonable decoding complexity
[Sklar, 1997]. Turbo codes are generated by using the parallel concatenation of two simple
convolutional codes, with one coder preceded by an interleaver [Benedetto and Montorsi,
1996]. The interleaver ensures that error-prone words received for one of the codes corre-
sponds to error-resistant words received for the other code.
All of the codes described earlier achieve their coding gains at the expense of band-
width expansion. That is, when redundant bits are added to provide coding gain, the overall
data rate and, consequently, the bandwidth of the signal are increased by a multiplicative fac-
tor that is the reciprocal of the code rate; the bandwidth expansion of the coded system rel-
ative to the uncoded system is 1R=nk. Thus, if the uncoded signal takes up all of the
available bandwidth, coding cannot be added to reduce receiver errors, because the coded
signal would take up too much bandwidth. However, this problem can be ameliorated by
using trellis-coded modulation (TCM).
ch01.I 4/21/00 12:02 PM Page 28

28 Introduction Chap. 1

TABLE 14 CODING GAINS WITH POLAR BASEBAND, BPSK OR QPSK

Coding Coding
Gain (dB) Gain (dB)
at 10 5 at 10 8 Data Rate
Coding Technique Used BER BER Capability

Ideal coding 11.2 13.6

Turbo code [Sklar, 1997] 8.8
Concatenated a 6.57.5 8.59.5 Moderate
ReedSolomon and
convolution (Viterbi
decoding)
Convolution with 6.07.0 8.09.0 Moderate
sequential decoding
(soft decisions)
Block codes 5.06.0 6.57.5 Moderate
(soft decision)
Concatenated a 4.55.5 6.57.5 Very high
ReedSolomon and
short block
Convolutional with 4.05.5 5.06.5 High
Viterbi decoding
Convolutional with 4.05.0 6.07.0 High
sequential decoding
(hard decisions)
Block codes 3.04.0 4.55.5 High
(hard decisions)
Block codes with 2.04.0 3.55.5 High
threshold decoding
Convolutional with 1.53.0 2.54.0 Very high
threshold decoding

a Two different encoders are used in series at the transmitter (see Fig. 14), and the corresponding decoders are used at

the receiver.
Source: Bhargava [1983] and [Sklar, 1997].

Trellis-Coded Modulation

Gottfried Ungerboeck has invented a technique called trellis-coded modulation (TCM) that
combines multilevel modulation with coding to achieve coding gain without bandwidth
expansion [Benedetto, Mondin, and Montorsi, 1994; Biglieri, Divsalar, McLane, and Simon,
1991; Ungerboeck, 1982, 1987]. The trick is to add the redundant coding bits by increasing the
number of levels (amplitude values) allowed in the digital signal without changing the pulse
width. (The bandwidth will remain the same if the pulse width is not changed, since the band-
width is proportional to the reciprocal of the pulse width.) This technique is called multilevel
signaling and is first introduced in Section 34. For example, the pulses shown in Fig. 314a
represent L=4 multilevel signaling, where each level carries two bits of information as
ch01.I 4/21/00 12:02 PM Page 29

Section 112 Preview 29

shown in Table 33. Now add one redundant coding bit to the two information bits to provide
eight amplitude levels for the pulses, but maintaining the same pulse width so that the wave-
form would have the same bandwidth. Then the redundant bit, due to the coding, could be ac-
commodated without any increase in bandwidth. This concept can be generalized to
complex-valued multilevel signaling, as shown at the end of Sec. 510. In summary, TCM in-
tegrates waveform modulation design, with coding design, while maintaining the bandwidth
of the uncoded waveform.
When a convolutional code of constraint length K=3 is implemented, this TCM tech-
nique produces a coding gain of 3 dB relative to an uncoded signal that has the same band-
width and information rate. Almost 6 dB of coding gain can be realized if coders of
constraint length 9 are used. The larger constraint length codes are not too difficult to gen-
erate, but the corresponding decoder for a code of large constraint length is very complicat-
ed. However, very-high-speed integrated circuits (VHSIC) make this such a decoder feasible.
The 9,600-bits CCITT V.32 (Table C7), 14,400-bits CCITT V.33bis (Table C8),
and 28,800-bits CCITT V.34 (Table C5) computer modems use TCM. The CCITT V.32
modem has a coding gain of 4 dB and is described by Example 4 of Weis paper [Wei, 1984;
CCITT Study Group XVII, 1984].
For further study about coding, the reader is referred to several excellent books on the
topic [Blahut, 1983; Clark and Cain, 1981; Gallagher, 1968; Lin and Costello, 1983;
McEliece, 1977; Peterson and Weldon, 1972; Sweeney, 1991; Viterbi and Omura, 1979].

112 PREVIEW
From the previous discussions, we see the need for some basic tools to understand and de-
sign communication systems. Some prime tools that are required are mathematical models
to represent signals, noise, and linear systems. Chapter 2 provides these tools. It is divided
into the broad categories of properties of signal and noise, Fourier transforms and spectra,
orthogonal representations, bandlimited representations, and descriptions of linear systems.
Measures of bandwidth are also defined.

113 STUDY-AID EXAMPLES

SA11 Evalution of Line of Site (LOS) The antenna for a television (TV) station is located
at the top of a 1,500-foot transmission tower. Compute the LOS coverage for the TV station if
the receiving antenna (in the fringe area) is 20 feet above ground.

Solution: Using Eq. (16), we find that the distance from the TV transmission tower to the
radio horizon is
d1= 2h  2(1,500)=54.8 miles
The distance from the receiving antenna to the radio horizon is
d2= 2(20)=6.3 miles
Then, the total radius for the LOS coverage contour (which is a circle around the transmission
tower) is
d=d1+d2=61.1 miles
ch01.I 4/21/00 12:02 PM Page 30

30 Introduction Chap. 1

SA12 Information Data Rate A telephone touch-tone keypad has the digits 0 to 9, plus
the  and # keys. Assume that the probability of sending  or # is 0.005 and the probability of
sending 0 to 9 is 0.099 each. If the keys are pressed at a rate of 2 keyss, compute the data rate
for this source.

Solution: Using Eq. (18), we obtain

H =Pj log2  P1 
j

10(0.099) log  0.099   2(0.005) log  0.005 
1 1 1
= 10 10
log10 (2)

or
H=3.38 bitskey

Using Eq. (19), where T=1(2 keyss)=0.5 skey, yields

H 3.38
R=  =6.76 bitss
T 0.5

SA13 Maximum Telephone Line Data Rate A computer user plans to buy a higher-speed
modem for sending data over his or her analog telephone line. The telephone line has a signal-
to-noise ratio (SNR) of 25 dB and passes audio frequencies over the range from 300 to 3,200
Hz. Calculate the maximum data rate that could be sent over the telephone line when threre are
no errors at the receiving end.
Solution: In terms of a power ratio, the SNR is SN=10(2510)=316.2 (see dB in Chap-
ter 2), and the bandwidth is B=3,200-300=2,900 Hz. Using Eq. (110), we get

 
S
R=B log2 1  =2,900 [log10 (1+316.2)]log10(2),
N

or
R=24,097 bitss

Consequently, a 28.8-kbits modem signal would not work on this telephone line; however, a
14.4-kbits modem signal should transmit data without error.

PROBLEMS

11 A high-power FM station of frequency 96.9 MHz has an antenna height of 1200 ft. If the sig-
nal is to be received 60 miles from the station, how high does a prospective listener need to
mount his or her antenna in this fringe area?
12 Using geometry, prove that Eq. (16) is correct.
13 A terrestrial microwave system is being designed. The transmitting and receiving antennas are
to be placed at the top of equal-height towers, with one tower at the transmitting site and one at
ch01.I 4/21/00 12:02 PM Page 31

Problems 31

the receiving site. The distance between the transmitting and receiving sites is 25 miles. Calcu-
late the minimum tower height required for an LOS transmission path.
14 A cellular telephone cell site has an antenna located at the top of a 60-ft tower. A typical cel-
lular telephone user has his or her antenna located 4 ft above the ground. What is the LOS ra-
dius of coverage for this cell site to a distant user?
15 A digital source emits 1.0- and 0.0-V levels with a probability of 0.2 each and 3.0- and
+4.0-V levels with a probability of 0.3 each. Evaluate the average information of the source.
16 Prove that base 10 logarithms may be converted to base 2 logarithms by using the identity
log2(x)=[1log10(2)] log10(x).
17 If all the messages emitted by a source are equally likely (i.e., Pj=P), show that Eq. (18)
reduces to H=log2(1P).
18 For a binary source:
(a) Show that the entropy H is a maximum when the probability of sending a binary 1 is equal
to the probability of sending a binary 0.
(b) Find the value of maximum entropy.
19 A single-digit, seven-segment liquid crystal display (LCD) emits a 0 with a probability of 0.25;
a 1 and a 2 with a probability of 0.15 each; 3, 4, 5, 6, 7, and 8 with a probability of 0.07 each;
and a 9 with a probability of 0.03. Find the average information for this source.
110 (a) A binary source sends a binary 1 with a probability of 0.3. Evaluate the average informa-
tion for the source.
(b) For a binary source, find the probability for sending a binary 1 and a binary 0, such that the
average source information will be maximized.
111 A numerical keypad has the digits 0, 1, 2, 3, 4, 5, 6, 7, 8, and 9. Assume that the probability of
sending any one digit is the same as that for sending any of the other digits. Calculate how of-
ten the buttons must be pressed in order to send out information at the rate of 2 bitss.
112 Refer to Example 11 and assume that words, each 12 digits in length, are sent over a system
and that each digit can take on one of two possible values. Half of the possible words have a
probability of being transmitted that is A 12 B 13 for each word. The other half have probabilities
equal to 3 A 12 B 13. Find the entropy for this source.
113 Evaluate the channel capacity for a teleprinter channel that has a 300-Hz bandwidth and an SNR
of 30 dB.
114 Assume that a computer terminal has 110 characters (on its keyboard) and that each character
is sent by using binary words.
(a) What are the number of bits needed to represent each character?
(b) How fast can the characters be sent (characterss) over a telephone line channel having a
bandwidth of 3.2 kHz and an SNR of 20 dB?
(c) What is the information content of each character if each is equally likely to be sent?
115 An analog telephone line has an SNR of 45 dB and passes audio frequencies over the range of
300 to 3,200 Hz. A modem is to be designed to transmit and receive data simultaneously (i.e.,
full duplex) over this line without errors.
(a) If the frequency range 300 to 1,200 Hz is used for the transmitted signal, what is the max-
imum transmitted data rate?
(b) If the frequency range 1,500 to 3,200 Hz is used for the signal being simultaneously re-
ceived, what is the maximum received data rate?
(c) If the whole frequency range of 300 to 3,200 Hz is used simultaneously for transmitting
and receiving (by the use of a hybrid circuit as described in Chapter 8, Fig. 84), what are
the maximum transmitting and receiving data rates?
ch01.I 4/21/00 12:02 PM Page 32

32 Introduction Chap. 1

116 Using the definitions for terms associated with convolutional coding, draw a block diagram for
a convolutional coder that has rate R=23 and constraint length K=3.
1-17 For the convolutional encoder shown in Fig. P117, compute the output coded data when the
input data is x=[10111]. (The first input bit is the leftmost element of the x row vector.)

Three-stage shift register

Data = x
Input S1 S2 S3

Modulo 2 adders

Commutator
Convolutional code
Output

Figure P117