PRINCIPLES

OF COMMUNICATION
SYSTEMS
Second Edition

Herbert Taub
Donald L. Schilling
Professors of Electrical Engineering
The City College of New York

McGraw-Hill Book Company

New York St. Louis San Francisco Auckland Bogota Hamburg
London Madrid Mexico Montreal New Delhi
Panama Paris Siio Paulo Singapore Sydney Tokyo Toronto

RPX-Farmwald Ex. 1042, p 1

This book was set in Times Roman.

The editor was Sanjeev Rao;
the cover was designed byt'ohn Hite;
the production supervisor was Marietta Breitwieser.
Project supervision was done by Santype International Limited.

PRINCIPLES OF COMMUNICATION SYSTEMS

Copyright 1986, 1971 by McGraw-Hill, Inc. All rights reserved.
Printed in the United States of America. Except as permitted under
the United States Copyright Act of 1976, no part of this publication
may be reproduced or distributed in any form or by any means, or
stored in a data base or retrieval system, without the prior written
permission of the publisher.
567890

BRBBRB

89

ISBN a-o?-062955-2

Ubrary of Coagress Cataloglna In Publlcadon Data

Taub, Herbert, 1918Principles of communication systems.
(McGraw-Hill series in electrical engineering.
Communications and signal processing)
Includes bibliographies.
1. Telecommunication systems. I. Schilling.
Donald L. II. Title. III. Series
TK5101.T28 1986
621.38
85-11638
ISBN 0-07-062955-2 (text)
ISBN 0-07-062956-0 (solutions manual)

RPX-Farmwald Ex. 1042, p 2

COMMUNICATION SYSTEM AND NOISE CALCULATIONS

615

in which the left-hand member is the average energy stored on the capacitor. This
result is an example of the famous equipartition theorem of classical statistical
mechanics. The equipartition theorem states that a system in equilibrium with its
surroundings, all at a temperature T, shares in the general molecular agitation
and has an average energy which is !k T for each degree of freedom of the system.
Thus, an atom of a gas, which is free to move in three directions, has three
degrees of freedom and correspondingly has an average kinetic energy which is
3 x !kT = !kT. At the other extreme, a macroscopic system such as a speck of
dust suspended in a gas similarly flits about erratically and has an average energy
associated with this random motion of ikT. Since the dust speck is much more
massive than an atom, the average velocity of the dust speck will be correspondingly much smaller. As another example, consider a wall galvanometer, which,
being free only to rotate, has a single degree of freedom. The kinetic energy
associated with such rotation is !llP where I is the moment of inertia and () is the
angular velocity. Such a galvanometer shares in the thermal agitation of the air
in which it is suspended, and f1() 2 = !kT. If the beam of light reflected from the
galvanometer mirror is brought to focus on a scale sufficiently far removed, the
slight random rotation of the galvanometer may be observed with the naked eye.
Altogether, it is interesting to note that the noise generated by a resistor is not a
phenomenon restricted to electrical systems alone, but is a manifestation of, and
obeys, the same physical laws that characterize the general thermal agitation of
the entire universe.
Returning now to the RC circuit of Fig. 14.4-1, we observe that it has one
degree of freedom, i.e., the circuit has one mesh, and a single current is adequate
to describe the behavior of the system. On this basis, then, Eq. (14.4-5) is seen to
be an example of the equipartition theorem.

14.5 AVAILABLE POWER

The available power of a source is defined as the maximum power which may be
drawn from the source. If, as in Fig. 14.5-1, the source consists of a generator v.
in series with a source impedance z. = R + jX, then maximum power is drawn
the complex conjugate of z..
when the load is ZL = R - jX, that is, ZL =
The available power is, therefore,

z:,

(14.5-1)

z.

L'

= R 0 +jX 0

Figure 14.5-1 A source of impedance z. is loaded by

a complex conjugate impedance ZL =
in order to
draw maximum power.

z:

RPX-Farmwald Ex. 1042, p 3

616

PRINCIPLES OF COMMUNICATION SYSTEMS

Note that the available power depends only on the resistive component of the
source impedance.
Using Eq. (14.5-1), we have that the available thermal-noise power (actual
power, not normalized power) of a resistor R in the frequency range df is
Pa=

4k~; df =

kT df

(14.5-2)

The two-sided available thermal-noise power spectral density is

(14.5-3)
Observe that Ga does not depend on the resistance of the resistor but only on the
physical constant k and on the temperature. If the source consists of a combination of resistors (all at temperature T) together with inductors and capacitors, then in Eq. (14.5-2) the R in the numerator and the R in the denominator
are both replaced by R(f), where R(f) is the (usually frequency-dependent)
resistive component of the impedance seen looking back into the network. These
R(f)'s will cancel, as do the R's. Hence, whether the network is a single resistor
or a complicated RLC network, the available noise-power spectral density is
Ga= kT/2 quite independently of its component values and circuit configuration.
Equation (14.5-3) expresses the available noise-power spectral density as predicted by the principles of classical physics, which also predict that this value of
Ga applies at all frequencies; i.e., the noise is white. This result is manifestly
untenable, since it predicts that the total available power
Pa=

J_:

(14.5-4)

Ga(/) df

is infinite. This prediction was one of a series of similar inconsistencies which

were, in part, responsible for the development of the branch of physics called
quantum mechanics. The quantum mechanical expression for Ga(f) is
hf/2

Ga(/)=

(14.5-5)

ehf{kT - 1

Ga({)

ltT
2

0.45/tT

4.3xl0 7'
2.6Xl010 T

Figure 14.5-2 Available power spectral density of thermal noise as given by Eq. (14.5-5).

RPX-Farmwald Ex. 1042, p 4

COMMUNICATION SYSTEM AND NOISE CALCULATIONS

617

in which h = 6.62 x 10- 34 J/s is Planck's constant. Equation (14.5-5) yields a

finite value for Pa and reduces to Eq. (14.5-3) when hf~ kT.
The power spectral density of Eq. (14.5-5) is plotted in Fig. 14.5-2. Note that
the density is lower than kT/2 by 1 dB or more only whenf~ 4.3 x 109 T , which
at room temperature, T0 ~ 290K, corresponds to f ~ 1.3 x 10 12 =
1.3 x 103 GHz. Hence we may certainly use Ga= kT/ 2 at radio and even microwave frequencies ( ~ 10 GHz). Note that a microwave receiver may employ a
maser amplifier operating at a temperature as low as 4K in order to minimize
the noise due to the amplifier. Even at these low temperatures, it is still appropriate to assume that the noise is white. At optical frequencies this assumption is no
longer valid, and Eq. (14.5-5) must be employed.

14.6 NOISE TEMPERATURE

Solving Eq. (14.5-2) for T, we have
(14.6-1)
When we apply Eq. (14.6-1) to a passive RLC circuit in which the noise is due
entirely to the resistors, then T is the actual common temperature of the resistors.
Consider, however, the noise which may appear across a set of terminals connected to a more general type of circuit, including possibly active devices.
Suppose that we measure the available power at the terminals and find that the
noise is white, i.e., the available power Pa increases in proportion to the bandwidth, so that PJdf is the same at all frequencies. We may then take Eq. (14.6-1)
to be the definition of the noise temperature of the network. The noise temperature of the network need not be the temperature of any part of the network.
Consider, for example, the simple idealized situation represented in Fig.
14.6-1. Here a resistor R, which is a thermal-noise source at a temperature T, is
connected to the input terminals of an amplifier of gain A. We assume that the
input impedance of the amplifier is infinite and assume further, for simplicity, that
the amplifier output resistance is a noiseless resistor R0 Then the noise power, in
a frequency range df, available at the amplifier output terminals is

v2 kTRA 2 df
p =--2..=---a Ro
Ro

(14.6-2)

Figure 14.6-1 Illustrating that the noise temperature seen looking

back into a set of terminals a - b may assume any value.

RPX-Farmwald Ex. 1042, p 5

622

PRINCIPLES OF COMMUNICATION SYSTEMS

14.10 NOISE FIGURE

Let us assume that the noise present at the input to a two-port may be represented as being due to a resistor at the two-port input, the resistor being at room
temperature T0 (usually taken to be T0 = 290K). If the two-port itself were
entirely noiseless, the output available noise-power spectral density would be
G~ 0 = ga(f)(kT0 /2). However, the actual output noise-power spectral density is
Gao, which is greater than G~0 The ratio GaJG~ 0 F is the noise figure of the
two-port, that is,

F(f)

=Gao = ga(f)(kTo/2)
Gao
G~ 0

(14.10-1)

If the two-port were noiseless, we would have F = 1 (0 dB). Otherwise F > 1.

Using Eq. (14.9-2) with T = T0 , and Eq. (14.10-1), we find that the noise figure F
and the effective temperature T,, are related by

T,, = T0 (F - 1)

(14.10-2)

F=l+T,,=T,,+To

or

To

To

(14.10-3)

The noise figure as defined by Eq. (14.10-1) is referred to as the spot noise.figure,
since it refers to the noise figure at a particular" spot" in the frequency spectrum.
If we should be interested in the average noise figure over a frequency range from
/ 1 to f 2 , then, as may be verified (Prob. 14.10-3), this average noise figure F is
related to F(f) by

_f:

F=

ga(f)F(f) df

f12ga(f) df

(14.10-4)

l,

Two-ports are most commonly characterized in terms of noise figure when the
driving noise source is at or near T0 , while the concept of effective noise temperature T,, is generally more convenient when the noise temperature is not
near T0
When following a signal through a two-port, we are not so much interested
in the noise level as in the signal-to-noise ratio. Consider, then, the situation indicated in Fig. 14.10-1. Here the noise at the two-port input is represented as being
R

Ba(fJ

Figure 14.10-J A signal v, and a noise source

are superimposed and applied at the input of
a two-port of available gain g0 (f).

RPX-Farmwald Ex. 1042, p 6

COMMUNICATION SYSTEM AND NOISE CALCULATIONS

623

due to a resistor R so that the available input-noise-power spectral density is

G~~> = kT/2. A signal is also present at the input with available power spectral
density G~~>. The available output-signal-power spectral density is
(14.10-5}
However, because of the noise added by the two-port itself, the available outputnoise spectral density is
(14.10-6}
Combining Eqs. (14.10-5) and (14.10-6), we have an alternative interpretation of
the spot noise figure, that is,
(14.10-7}
Thus F is a ratio of ratios. The numerator in Eq. (14.10-7} is the input-signal-tonoise power spectral density ratio, while the denominator is the output-signal-tonoise power spectral density ratio.
Let us assume that in a frequency range fromf1 to f 2 the power spectral densities of signal and noise are uniform. In this case it may be verified (Prob.
14.10-5) that the average noise figure F defined by Eq. (14.10-4) has the
significance
(14.10-8}
where S1 and N; are, respectively, the total input available signal and noise
powers in the frequency range / 1 to f 2 , and similarly Sa and Na are the total
output available signal and noise powers.
The noise figure F (or F) may be expressed in a number of alternative forms
which are of interest. If the available gain g0 is constant over the frequency range
of interest, so that F = F, then S0 = g0 S1 In this case Eq. (14.10-8} may be
written
(14.10-9)
Further, the output noise Na is
Na= gaNi

+ N,p

(14.10-10}

where g0 N 1 is the output noise due to the noise present at the input, and N,P is
the additional noise due to the two-port itself. Combining Eqs. (14.10-9} and
(14.10-10), we have

F=l+~
gaNi

(14.10-11}

or, the noise due to the two-port itself may be written, from Eq. (14.10-11), as
(14.10-12)

RPX-Farmwald Ex. 1042, p 7