I

r I
 F. R. Connor 1982

1979
Second edition 1982

may be
retI"ieval . . "... L .......... _or transmitted in any
means, ..........'''' ......."" . . . . '..... 1I1ecJt1al1l1CalJ, p,n01:0CIDp~(lnJ~,
nth.;coT''UI,nc:l'.;co without the

l.PubHs:hers) Ltd.

British t.:altaU)2UID2 in Publication Data

F. R.
Noise.-2nd ed.
1. Electronic circuits-Noise
I. Title
621.3815'3 TK7867.5
ISBN 0-7131-3459-3

A"''''V& , .....,'"', Scotland

Preface

In this new edition, certain parts of the text have been extensively revised. A new
section on random variables is introduced in Chapter 2 and some basic ideas
concerning matched filtering, decision theory, and estimation theory are
presented in Chapter 3. A further treatment of circuit noise is made in Chapter 4
and a new section on low-noise amplifiers is included in Chapter 5. In Chapter 6,
a comparative study of the signal-to-noise performance of various systems has
been extended to cover digital systems and satellite systems. As an alternative
approach, the energy-to-noise density ratio and its effect on the bit error rate is
also included. A further feature of the book is the extended use of appendices to
cover such topics as narrowband noise, decision theory, estimation theory, and
the probability of error. It is intended for the reader seeking a deeper
understanding of the text and is supplemented by a large number of useful
references for further reading. The book also includes several worked examples
and a set of typical problems with answers.
The aim of the book is the same as in the first edition, with the difference that
Higher National Certificates and Higher National Diplomas are being
superseded by Higher Certificates and Higher Diplomas of the Technician
Education Council.
In conclusion, the author wishes to express his gratitude to those of his
readers who so kindly sent in various corrections for the earlier edition.

1982 FRC

Preface to the first edition

This is an introductory book on the important topic of Noise. Electrical noise is
of considerable importance in communication systems and the book presents
basic ideas in a coherent manner. Moreover, to assist in the understanding of
these basic ideas, worked examples from past examination papers are provided
to illustrate clearly the application of fundamental theory.
The book begins with a survey of the various types of electrical noise found in
communication systems and this is followed by a description of some
mathematical ideas concerning random variables; Circuit noise, noise factor,
and noise temperature are considered in the following chapters, and the book
ends with a comparative study of some important communication systems.
IV
1 1
1.1 1
1 2
1.3 4

2 and 6
6
7

9
10
12
17

4
4.1
VI ConlenlS

4.4 tranMstor noise

FET nOIse

5 measurement
5.1

63

6
6.1

B
C
D
E
F

H
I
J

L
m

F
Vlll Symbols

Pc average carrier power

PD detection probability
PF false alarm probability
PT transmitted power
R bit rate
resistance
Req equivalent thermal noise resistance
R(r) autocorrelation function
R.Jt) autocorrelation function of variable x
Rxy{-r;) cross-correlation function of variables x and y
S average signal power
SIN signal-to-noise ratio
StiNt input signal-to-noise ratio
SolN 0 output signal~to-noise ratio
S(w) power spectral density
T absolute temperature
periodic time
antenna noise temperature
effective noise temperature
receiver noise temperature
system temperature
energy
highest modulating frequency

y noise power of standard source

Y-factor = ------------
, noise power of antenna
rx transistor forward current gain
cX o transistor d.c. forward current gain
5 (t) Dirac delta function
A wavelength
v frequency of events
p correlation coefficient
(J standard deviation
L time interval
OJ angular frequency
 III

I I
2

1.1
 3

o~O'-1----------~~-----------1~O------------1~O~O
f(GHz)---

1.2
4
5
 il

statistics. 6

if rnB is occurrences Bwe

=
n- co n
 statistics

n events

p
B

mAB n
--------
n

p
 \..'IIJUJl.iIIJl.IJlk the two v\..I1I.IUI.I.\..II.&..J

P P IA)P

or

are 'nr1jQ ...... jQni'"1jQj"'lt events =1'

p =p
A B is

Three coins are tossed random. What is the Pr<)b~LbIjllty for all heads and for
all heads or all tails?

Drclbaloili1tv of all tails P i, the total uu,aul,u", for

sum of P i.e.!

A box contains six red beads and three blue beads. Two beads are drawn out In
succession. If the first bead is what is the that both beads are of dIfferent
colour?

dniwlln2 out a red bead is six times out of nine or

the conditional Drc)b~lbillitv
been i.e.
we obtain
 statistics 9

The Drc.oaltn1l1tv of dra.WlrUl out the two beads is and is

theorem as
=
or

10

030 08
025 X 06
8: 0-20 it
a.. 015
010 0-2
005

(a) (b)

2.1

=
 Noise

a x

2.2

or

etc. are p

+ ... +
or
 statistics 11

Two random variables x and y have standard deviations (J x and

Determine the variance of their sum.

Solution
If z = x + y, the '1118'1"'18,." ....450

y) =

Now

and o
(y y=o
Hence +
12 Noise

P(A) = 0
P(B) 75

2 3 4

2.3
 statistics 13

m
Noise

025

020'-
vT::~3

015~

010!-

0051-

0 2 5
I
6
I
7
I

8 m

2.4

=1

10

05

o x ...............
(a) (b)
2.5
 m=

u
u

m
 Noise

or =

or =
is

2.6
(J

= l-e

where the variance. 1'\11 {",\,-p.,("n,,,",- we have

du

Since

Hence
2ku

or
18 Noise

Xl (t)

t __

Fig. 2.7

wherea~, In time averaging, a partIcular sample function only is considered, e.g.

Xl (I),and averaging IS performed over a lime interval of this sample. Two of the
most u~eful quantltIes are the average or expected value E[X] and the
autocorrelation function R,(r) (whIch IS discussed In the next chapter).

Stationarity
A random process 111 which the average valuc of thc random variable X IS the
same usmg ensemble averaging at any time t = r lor l = r2 is called a stationary
process a~ its statistics are independent of the particular time position chosen.
For such a process, the average or expected value IS given by

E[X] = f~:XP(X)dX
where pIx) IS the probability density function of the random variable X in the
Interval 0 < x < x + dx. Similarly, the autocorrelation function R,(r) over the
- d

R '!)
r

()
o

o
 have zero mean IS

and 'lI<:lor1#:lnr"DoCO
and
if is a r"f"U"'Icoi"'r:1ni"

since

since

or

p= x

Hence + 0"2+

and Show there is

Solution

sin with 0"1

V2 =

Hence =0
or v t v,2 0
VI V ,2 0
Now p=
0'10',2 0'10',2

Hence p 0

=:

or

to

=:
 :E:
)(

Vo ~

0 TI2 T t-

-Vo

-, Vo
I
I
I
I
I
I
I
0 TI2 T t-
I
I I
I I
..J
Vo

3.2

+
24

(1 - with R

r)dt

sin i dt

sin { r) dt

dt

function over zero. Hence

and R

.3.3
Derive autocorrelation function of the random tcol,con-Y"fJI't"'l.h

shown in the time instants zero ,...,.,........... 9"11,.......

distribution.

+ Vf- ~ ~

0 t--

Vi--- '--- '"--

3.4

Assume that For m ,..h''.:Ir...... coC'

the time interval T, which more
cO]t1ve~nH~ntly written

To -r)

of

variable in Section as

Here
of chalngc~s

Since the any interval of

or
 3.5
and is shown in 3.5. SInce
-t)
more Flo",,,,,,,,,,c;a.U.J

26
Comnlent
We observe that the curve in 3.5 is narrow at low values the narrower is
is the number of in the interval T = t, nu"I .....~lt ..,.n the presence
treaUlen(:v clomponellts. This will be confirmed later when the spe:ctral
obtained for this

Two random are

where A and B are constants and
cross-correlation function := O.
Solution

-'t")dt
1" ....

1
+ sin dt
T

AB
:::::: sin 8)
T

x -cos SIn

or AB 8 sin (J) 1 t sin dt

+sin 8 cos cos (J) 1 t sin dt

-cos8 sin sin cos

sin 8 SIn cos cos

Each of the 1nt,{!!Ian1"I!.lIIC! is of standard form and is

is
28 Noise

2.

",""''::'''<'or if

With the of the Wiener-Khintchine 'the:'r.r,p'l"\""t determine the function

of the random 3.4. What the T l I r . ' I I . . " P . . . Slgn:lhcan(~e
of
the result?

The autocorrelation function of the random t-.:o.","or')1"'\h was ........c..,1r~'."'I .. , obtained as

we

dr

e dr

or e -2vl_1 e - Jwr
+

]
 e -2\,11 + dr

the form

=2 -2v +

or

and the maximum is It

(0-

3.6

IS amount of

> Wo
30 Noise

Solution
From the Wiener-Khintchine theorem we have

R(T.) = f~: S(f)eJWTdJ

or R(O) = f~: S(f)dJ

which is the average output power of the network. Here
. Sowo 0 Sowo
R(O) = - - e = - - watts
. 4 4
If V rrns is the output voltage in a 1 Q load then

or volts

3.6 White noise

Random noise signals generated by various sources are known to have a
uniform power distribution over a very wide range of frequencies up to about
10 13 Hz, which is in the ultra-violet region, after which it falls ofT as predicted by
quantum theory. Such noise is defined as white noise by analogy with white
light which has a uniform power distribution over the band of optical
frequencies. The nearest examples are thermal noise in a resistor and shot noise
in valves or transistors which also have a Gaussian amplitude distribution and
are known as Gaussian white noise.
If No is the noise power spectral density per Hz for positive frequencies only
then, assuming both positive and negative frequencies (for mathematical
purposes), the noise power spectral density is N o/2, as shown in Fig. 3.7(a).
Hence, we obtain for the autocorrelation function
f+ 1'+
R(r) =~
1
""II.;
I
('+ oc'

-<:I)
S(w)e lWt dw = ~
1
""II
oc

-0:'
N e lWr
0
2
dw =
N
~-
2
1
J oc

2II -"'0
eJ"H dm

N
or R(T) = ~o(r)
2
where 6(T) is the Dirac delta function shown in Fig. 3.7(b). Since R(T.) has a value
at T = 0 only, there is no correlation between any two samples of white noise
separated by an interval, > 0 and they are therefore statistically independent.
From Fig. 3.7(a) it will be observed that the average power, which is given by

Pay = +- f+ ""
.... 1f - oc
S(w)dw
 Correlation techniques 31

R(.)

I
No/2
S (Ul) .....

____ L _ _ _ __
............... -Lv 0 +Ul-
(a) (b)

~vl R(.)

~B
1 ("I o B f- T-

(c) (d)

Fig. 3.7

becomes infinite and cannot be physically realised in any practical circuits. As

most communication circuits are band-limited, it is more practical to consider
the results of passing white noise through a filter with some defined bandwidth.
The output noise is then called band-limited white noise or coloured noise.

3.7 Band-limited white noise

Ifwhite noise is passed through a low-pass ideal filter with a bandwidth B Hz,
the output noise can be obtained by means of the transfer function H(w) of the
filter. Hence, we have
SI)(W) = IH(w)12
Sj(w)
where S, (w) and So(w) are the input and output power spectral densities and
IH(w)1 = 1, with
1 "+cc. 1 r+ cc No No f+B
Pav = 2
n -00
J
So(W) dw =? I
_n,; _ -oo
-2 dw = -
2_B
df = NoB watts

and is illustrated in Fig. 3.7(c).

The autocorrelation function R (r) of the filtered white noise is

R(r)
1
=;;-- f+ 00
So(w)eiW!dw = -1 f+ 00
Sj(w) IH(w)1 2 e JWt dw
.t..1t -co 2n:",_oo

1 J"+ ~eJW!dw=~!
N. N i+B N [eJ21<ft] +B
eJ21fJtdf=~ - . -
ex)
=-
2n _ (f) 2 2 .! -B 2 J2m -8
 Noise

x
or R
x

iovv-uass filter shown in 3.8 is white noise with a power 1IlI .... ""...,"I.Ir.l.1.

Determine the power and average

noise power.

3.8

Solution
If is the network transfer function we have

or

and are the and power 1IlI ........'...,"I.'u,& densities respec:tlvely then
1

or

The average noise power is

 the substitution u = wRC with du = RC dw we obtain
ex; du
-1
4nRC 4nRC

=--
4nRC

or

A sine..wave carrier ....... with Gaussian white noise is

",.0. ............

IF and detected a linear detector. that

follows or a distribution oel>enOll1l2

If the Vcsin then the carrier band..limited noise from the IF

= {Vc+ + cos
where the and

or
where r
The that

and =
where is the variance.
The of the
Since x and y r
y)dx

Also y)
34 Noise

or

since dxdy = rdrd(p. Hence

re - [,1 - 2rf', cos+ v~J;2al
p (I', ) = - - - - - - - : : - - - - -

To obtain the amplitude distribution p (r) we must integrate this expression over all
values of from 0 to 2n. Hence

p{r) = { " p(r, )d

= J "2
n re - [" 0"
r -~"" COS
"'.LV
'Y'
1 J,,2
, ~a d
o 271.a 2
I"e - (,' + ~. ~ )/2a' f21r
= . , e d ', cos 1>/,,2 d4)
27[a~ 0

To evaluate thIs II1tegral we use the standard integral 10 (z) where

which is a modified Bessel function of zero order. Hence

The quantity Vc /a 2 can be related to_tb~input carrier and noise powers C, and N,
respectively since V~/2 = C 1 and a 2 = n 2 (t) = N j Hence

rV, r.j2C,
,-
(J .IN,
r
or z=_ )2((',11'1,)
(J

p(/')=----
(J2

re - r2 '20'2
or p(r).~ - - , .... (since V~/2(J2 ~ 0)
(J.

which is a Rayleigh distribution with the peak value at r = a.

 ]) e=

at t'

3.9

3.8
Noise

3.10

can

R + +
 37

are

n ]

or f n
38 Noise

digital signals, it is pointless having an output anywhere between 0 and 1. Hence,

on the basis of the received waveform, it is necessary to decide which of the two
states the signal is in.
In this decision approach, the filter at the receiver must be matched to the
received waveform to achieve the maximum output signal-to-noise ratio. Such a
filter is designed specifically to maximise the output signal-to-noise ratio and is
called a matched jilter. The matched filter is the optimum filter for detecting
signals received with additive white noise, i.e. noise with a uniform power
density spectrum over a wide frequency band.
It is shown in Appendix F that the transfer function of such a filter is given by
H (01) = kS* (w)e -JW!d
where k is an arbitrary gain constant, S*(o1) is the complex conjugate of 5(w),
the spectrum of the received signal s (t), and td is the time-delay of the filter.
Furthermore, the impulse response of such a filter is given by
h(t) = kS(td -t)
where s(t) is the input signal to which the filter is matched. The time-delay td is
required to make the filter physically realisable and the filter has its maximum
output at some time t = to. The optimum decision is then made at time t = to to
determine the nature of the output corresponding to a 0 or 1, and the output
depends only on the original signal energy and is independent of its waveform.
A practical implementation of the matched filter for a rectangular pulse is
shown in Fig. 3.11.

Input _ Output
Pre-filter Comb filter
Signal i pulse
I
Fig. 3.11

In many practical cases, it is not feasible to provide a suitable matched filter as

it may be physically unrealisable and so somewhat simpler filters are used with
some loss in signal-to-noise ratios. Table 3.11ists some typical examples of these
and it is observed that the maximum loss is only about 1 to 2 dB.
In Appendix F it is shown that the matched filter is mathematically
equivalent to a correlation detector. In practical applications it may therefore
be more convenient to implement the matched filter by means of a correlation
detector. In the correlation detector shown in Fig. 3.12(a), the transmitted signal
waveform s(t) is stored at the receiver and correlated (multiplied) with the
received signal plus noise. After integration over the time interval to, the
detector then decides whether a 0 or 1 has been received.
 Table 3.1

s
' - - - - - - - v(t)

s(t)

t --+--

3.12
40 Noise

which converts the non~white noise to true white noise and modifies the input
signal slightly. The modified signal is now mixed with white noise and can be
optimally detected as was shown earlier. The modification to the input signal
leads to some intersymbol interference in the case of digital signals which can be
minimised if the bandwidth of the pre-whitening filter is large compared to 1/1'
where Tis the duration of the digital signal.

3.10 Decision theory 4

In the design of optimum receivers, an important problem is the detection of the
received signal in a background of noise. This may involve simply the detection
of a given signal or its absence or it may involve the detection of a zero or one, as
in digital data communications. In either case, the receiver must make the best
possible decision on the basis of various criteria. The statistical nature of the
decision making process involves hypothesis testing and is known as decision
theory or detection theory.
Receivers designed to minimise the average cost of making a decision use, as a
basis, Bayes' decision rule. When a sufficient knowledge of costs and a priori
probabilities is available, the decision process can be optimised according to an
expected cost criterion.
It is shown in Appendix G that such decision procedures which are based on
expected cost minimisation involve testing a likelihood ratio which is the ratio of
the a posteriori probabilities of the observations against a threshold which
depends on the a priori probabilities and costs. It is given by the expression

L[ (t)J~PdC21-Cll)
Y ~~p2(e12-e22)
where the Lh.s. is denoted by the threshold value L t , P 1 and P 2 are the a priori
probabilities associated with the hypotheses Hl and H2 respectively, and ell,
e 12, e 21, and C 22 are the conditional costs of making decisions.
The Bayes' criterion is therefore characterised by the average cost or risk
involved in making a decision. In many cases, it is more useful to express the risk
in terms of the probabilities of detection PD and false alarm PF , provided the
a priori probabilities and costs are available.
If the costs and a priori probabilities are not available, a useful decision
strategy is the Neyman-Pearson criterion which maximises P D while holding P F
at some acceptable value. This type of criterion can also be reduced to a
likelihood ratio, where the threshold is determined by the allowed false alarm
probability. It is shown in Appendix G that its value is given by
dPD
L[y(t)] = -
dPF
where (dPD/dPr) is the slope of the receiver operating characteristic at any
given point.
 Correlation techniques 41

3.11 Estimation tbeory 16

It was shown earlier that decision theory can be used to detect the presence of a
signal. Similarly, estimation theory can be used to make an estimate of some
unknown parameter of a signal such as its frequency or phase. Since the signal is
received with additive noise, only an estimate of the signal parameter is possible.
Two useful estimators are Bayes' estimate, which endeavours to minimise a cost
function, and the maximum-likelihood estimate, which tends to maximise a
likelihood function. As the latter does not require any a priori information, it is
often employed and will be considered here.

Maximum-likelihood estimation (MLE)

Since the received signal is a function of time and contains additive noise, it can
be represented by
z(t) = s(t, e) + n (t)
where s(t, 0) is the transmitted signal, e is the unknown parameter to be
estimated, and n (t) is assumed to be Gaussian white noise. By observing the
signal continuously or by sampling it, it is possible to make an estimate of e
which is denoted by ti.
lt is shown in Appendix H that, for a sampled signal, the likelihood function
ofz(t) is

p(zIO)=Aexp { -NoIf T
0 [z(t)-S(t,e)]2 d tJ1
where A is an unknown constant. To obtain the maximum-likelihood function,
it is convenient to differentiate In p (zIO) with respect to e and then equate it to
zero. This yields
a
-[lnp(zIO)]=-
2A IT [z(t)-s(t,O)]
as(t,O)
~'e dt
ao No 0 0

and so the maximum-likelihood estimate of 0 is a solution of the equation

I T
o [z(t)-s(t,O)]
~ as(t,
ao
ti)
dt=O

where 0 has been replaced by the estimate ti.

The estimate ti depends on the signal received and on the number of
observations made. Hence, it can vary as a random variable with a mean and
variance. The variance of an estimator, under certain conditions, cannot be less
than the lower bound known as the Cramer-Rao bound.
It is shown in Appendix H that, if the expectation ofti, Le. E [tiIB], is equal to
the true value e, the variance of ti is given by
2 1
u
o
~ -=~------~==
E[ {aln~~zIO) fJ
42 Noise

Example 3.10
A signal of known amplitude and frequency has the form s(t, 0) = A sin (wot + t/ where
o ::;; t ::;; T.
Make an estimate of the phase angle cPo

Solution
We have
s(t, 0) = A sin (wot + 4
T=
os(t 8)
A cos (Wot + cP)

and the phase estimate ;p IS a solutIOn of the equation

f: [z(t) - A sm (wot + $)] A cos (wot + $)dt = 0

or ITo
~
z(t)cos(wot+cP)dt =
fT -sin2(w
0
A
2
A

o t+t/dt

z(t)

(a)

zIt)
tan- 1 ylxl---__ ,p

(b)

Fig. 3.13
The ...... t,:>rT ....... 1 the if where is an or 1. Hence
T
o

= tan

n"'Il111".r.I1,o ... " and matched In

4
Circuit noise

The two most important types of noise associated with electronic components,
such as valves, transistors, and resistors, are thermal noise and shot noise. The
physical basis of each will be considered and, as both of them give rise to noise
power in the system, they can be regarded as producing one combined noise
effect.

4.1 Thermal noise

A metallic conductor or resistor contains a number of free electrons. Due to
thermal agitation, these free electrons are moving about continuously in the
conductor causing collisions with the atoms and a continuous exchange of
energy takes place. This accounts for the resistance property of the conductor
and, though there is no current in the conductor on open-circuit, the random
motion of electrons in the conductor produces voltage fluctuations across the
conductor which accounts for a mean-square noise voltage v~ at its terminals.
The thermal noise effect was investigated experimentally by Johnson 17 and
theoretically by Nyquist. 18 Experimental results showed that the thermal noise
voltage depends upon temperature and its mean-square value v~ is given by

v~ = 4kTBR
where k is Boltzmann's constant, T is the absolute temperature, B is the
bandwidth of the system, and R is the resistance of the conductor. For example,
if R = 1 kQ, B = 5 MHz, and T= 290K then with k = 138 X 10- 23 11K we
obtain vf = 80 X 10- 12 or Vrms = [vtJ 1 / 2 ::::: 9/lV.
Nyquist's investigation of the effect was based on thermodynamical reason-
ing and similar results were obtained. He showed that the thermal noise power
Pn associated with any resistor is given by
Pn = kTB watts
where k, T, and B have their previous meaning.
The derivation is given in Appendix I and is based on the assumption of
available noise power. This implies matched conditions, as is usually the case in
most communication channels, since it is necessary to transfer the maximum
 Circuit noise 45

signal power through the system. However, in practice, the concepts of noise
voltage, noise power, or noise power spectral density can be equally well
employed in the study of noise problems. If the noise voltage spectral density is
S. (f), it can be shown that*
Sv(f) = 2kTR
a result which depends on T and R but is independent of frequency up to about
10 13 Hz. This implies that thermal noise covers a broad band of frequencies and
has a uniform response. Hence, it is often called 'Johnson noise' or white noise
due to an analogy with white light which has a uniform power distribution over
the band of optical frequencies.

Equivalent circuit
It is convenient in practice to represent thermal noise in a resistor as due to a
thermal noise source v~ in series with a noiseless resistor R, which is based on
Tbevenin's theorem. Alternatively, a current source it in shunt with a
conductance G may be used and this is based on Norton's theorem. This is
illustrated in Fig. 4.1.

v~-)
TL-_ _o
Fig. 4.1

Under matched conditions, the load is also R (assumed noiseless) and the
maximum noise power available from the source is obtained as follows.
In Fig. 4.2, iUis the current in the circuit then i = ~/2R and the maximum

R
R

Fig. 4.2
* See Appendix C
 we

or
a

Two in series paIau~~1. If the

and 'l'"A."~A"tnlAI" ~~.~n.~~Atherms

noise tenmIIlals in each case.

Solution
Series 'l'"A." . ."t.,. . 'I'"."

4.3
If vn is the rms noise .. '. . . 11'11" ............ at the ........,"" .........."... then

where

Hence +
 noise

or vn =
In .......................'. . . . . _ if = T, as is gerlentUY the case in pnlctllce. then

and the two resistors behave as a resistor of value

Parallel resistors

4.4

The noise at the terminals due to and

the noise terminals due to 1.'.2
these two are rms the total rms at the terminals is Vn

where vn = we have

with

and

or

In nQ11"f'11"'nl~I" when practlc:e. then

or
 Noise

and the two resistors behave as a resIstor of va]ue

o t-

4.5
 noise

4.6

IS
50 Noise

Fig. 4.7

with an equivalent resistor Req. For a triode, the hypothetical resistor is inserted
in series with the grid and it is given by
RCq :::::: 25/g m
where gm is the mutual conductance of the triode. The equivalent circuit for
combined thermal and shot noise is shown in Fig. 4.8.

+HT +HT +HT

Rg

Fig. 4.8

The equivalent rms 'thermal' noise voltage VI of Fig. 4.8 is given by

VI = Jv~ + v~
where vf = 4kTBRS!
v~ = 4kTBRcq

Hence

Example 4.2
State two sources of noise encountered in high-gain amplifier circuits and briefly explain
their origins.
 Circuit 51

r"" ..... "'f.'... ........... "" of 4 kQ over

with

the

4.9

10

or

Also with = 4kTB

or

or

kQ

noise factor in Section 5.1

F
52 Noise

where (S,IN,)'is defined at the source and (So/No) may be referred to the input side at
points P and Q, with the amplifier noise represented by Req. Hence

F =
(2vY I (Vsl2
4DOOkii3, (iooo+:i3jO)I~fB
4
= 4000 x 3330

or F = 333 or 52dB

4.3 Partition noise

In multigrid valves, such as tetrodes and pentodes, the division of current to one
or other electrode is sut-ject to random fluctuations also. This gives rise to a
further noise effect which is basically similar to shot noise. It is known as
partition noise and can be evaluated using statistical ideas similar to that for
calculating shot noise.
This effect can be accounted for by increasing the value of the- equivalent
noise resistance Req obtained previously for the case of a triode. The value of
Req for a pentode is given by

Req =
Ia ) [2'-+-2-
5 20Is]
(Ia + Is grn gm

where 1a is the anode current, I, is the screen current, and gm is the mutual
conductance of the valve. Typical values for RCq are between 1000 0 to 10 kO.
Due to partition noise, multigrid valves are more noisy than triodes and
should be avoided in the early stages of an amplifier if noise is of primary
concern, as in low-noise amplifiers for space communications. The noise
generated in the early stages is amplified in subsequent stages and will give a
large noise output. In the case of multigrid mixers, the conversion conductance
gc is used instead of gm for obtaining the value of Req. Since gc is much smaller
than gm' values of ReQ around 100 kO are possible and so multigrid mixers are
quite noisy.

4.4 Bipolar transistor noise 20 21

Noise in junction transistors shows some similarity to that in valves and the
three types of valve noise, namely thermal noise, shot noise, and partition noise,
are found to exist in transistors. This is basically because random fluctuations in
the movement of the charge carriers (electrons and holes) cause variations in the
various transistor currents. However, their exact nature and evaluation is more
complicated and still not clearly understood.
For the bipolar transistor, it is found that thermal noise is associated with the
b

4.10
54 Noise

trans-
III

5.1
 F

power
power

the the '-J~"'-'''''''' IS

""' .. A''''''....,'..... power

IS
 measurement

an

1.

3.
4. a .......,,-........,.LJ

Discuss the sources of noise in an ampll:ner and the manner in which limit

nU!n-l!aln am,pU1tler deSilgIled to from a 75 n source contains

SU:U"UlU ratio 1 the first valve
enl11V~=llelnt noise resistance of the valve to be draw the LJ ..............:LILJ.:I ..

arr'an,geIne][}t and hence calculate the source emf that would be reCIUllred
ratio at the in bandwidth of 200 kHz. The value of
414 W.

Solution
The main sources of noise described in Sections 4.1 to 4.3. The
overall effect of the noise SH!nal-IO-IHHseratio of the If
this ratio below a certain minimum of value and the
the is limited.

Problem
The circuit is shown in
circuit is

5.2
 Noise

and IS the SHlnal-H)-n.OI~;e ratio the ..., . . . 'LU ...., ... =1. we have

s =---
e:
I 4 x 1200 300
k7 B = 414 X 10- 21 200 x
or 828 x 10 16

noise of the ampUJher and IS the

1200+ 700
F= =158
1200

Also F= 1)

Hence e; 158 x 828 16

or e; = 0-392 10 12

and es = 0-63

Define the terms nOl~e ratio and a

~1~nal-U) .. ntOI!\e
between them In network.
5 ~ho\\'~ a netVtrork with ItS nOise properties
. n l l .......... ,... ..... two
of nlcan-~quare and the other of mean-
dn\en from a source an interna1 resIstance
to hc uncorrelated.
n&>'I"U''''r'lll'"'ltl'tc" an for the nOIse factor
sourcc and that It IS a mlnlmum when
expressIons for 2 and Ifthc nctwork consists shunt-connected r&>:"I:''I'''''''

or a ~cnc~-connectcd rC\I~tor J~.

5.3
Solution
answer to first part IS gIven In Section 5.1.
 circuit is the at

---+

+ 4kTB

5.5

Shunt R
 Series resistor
=0
Hence 4kTBR

.5.6

1+

or F 1+

where J1
or 1+
if J1
 Noise measurement

Common-base

5.7

=1 +
J

F=
(1-
+----------------------~--------
J

=1+

IS source
62 Noise

,-----0 + VDD

JFET IGFET

Fig. 5.8

where Rn is an equivalent noise resistance, g,n is the input conductance, and gm is

the forward transconductance.

5.3 Cascaded networks

When two or more active networks such as amplifiers are connected in cascade
to give greater amplification, the overall noise factor F of the arrangement is
important.
Consider two networks in cascade with noise factors of FI and F2 and power
gains of Gland G 2 respectively. Let F be the overall noise factor of the
combIned networks with a relevant bandwidth B. If a noise signal with power
kTB is fed into the first network we have
output noise of first network = FI G} kTB
output noise of ideal network = GjkT B
nOlse generated 1I1 first network = F 1G 1H B-G 1k7 B
= (F I -l)H BG I
The noise generated in the Ilrst network therefore appears to be due to
a hypothetical input noise component which is equal to [(F1 - l)k TBGdG l ]
= (FI - l)k TB and is shown dotted in Fig. 5.9. Similarly, the equivalent noise
generator associated with the second network is (F2 -l)kTB. Hence, the first
two networks can be regarded as 'ideal' for the purposes of this analysis if the
noise contributions from the two input signals are taken into account and the
total noise output from the second network = kTBG 1 G 2 + (F I - l)kTBG J G 2
+ (F z -1)kTBG J = FlkTBG} Gz + (F2 -1)k7BG 2
 measurement

1)

5.9

or F= +

2. a Lean to
64 Noise

~) o+HT

RF ~
choke ')

?s
R

Fig. 5.10

Active
NOIse Attenuator f-------i Power meter
network
diode
F1 .G 1

Fig. 5.11

equivalent circuit in Fig. 5.10 is a shot noise current generator delivering a shot
noise mean-square current i; .
Let the input resistance to the active network which is used be Rgo The
measurement consists in noting the output power Po with the diode current zero
initially and wilh zero attenuation in circuit. The diode current is then adjusted
to a value I a which gives the same output reading with 3 dB attenuation in
circuit, i.e. the output power has been doubled.
To obtain a suitable meter reading for Po, the network gain is set at the same
convenient value for both readings. The use of 3 dB attenuation maintains the
reading at Po in both cases and avoids any scale non-linearity. Hence, we have
e2
noise power available from resistor Rg := _ 1 _ = kTB
4R g

. . e; eIaRgB
nOIse power output of dIOde = _0- = -----=~-
4R g 2
Hence, when Ia = 0 we have
Po = kTBGF (since F = No/kTBG)
 Noise measurement 65

When the diode current is set at the value 18 we have

Po+Po = kTBGF + elaR g BG/2
or 2kTBGF = kTBGF + el a Rg BG/2
Hence kTBGF = el a R g BG/2
with
Typically, e/2kT ~ 20, hence
F::::: 20I a R g
and it can easily be evaluated since the values of 18 and Rg are known directly in
the measurement.

Comment
At higher frequencies in the microwave region, the diode noise power is
insufficient for measuring large noise factors and so a gas discharge tube is used.
It is placed at a small angle across the waveguide to produce an impedance
match. At present, solid state noise sources are also available. (See Section 5.9.)

5.5 Noise temperature

The thermal noise power P" available from a resistor in a bandwidth B is kTB
where T is the absolute temperature of the resistor. Hence, an alternative
concept associated with noise power is the effective noise temperature Te which
is given by
Te = Pn /kB
for a resistor at temperature T K.
The idea of effective noise temperature can be extended to other noisy
sources which are not necessarily associated with a physical temperature as is
the case with a resistor. For example, a non-thermal device, e.g. an antenna
which picks up noise power due to the random radiation it receives from
various directions, may also be associated with an effective noise temperature Ta
If the noise power received by the antenna is Po in a bandwidth B then
Ta = P,jkB

The value ofTa depends on the direction in which the antenna points and its
radiation pattern. Different parts of the sky are associated with sources of
random radiation usually called galactic noise, solar noise, etc. Hence, the
sky effectively has a 'noise temperature' and it varies with frequency as shovm in
Fig. 5.12.
The concept of effective noise temperature may also be applied to an active
network, such as an amplifier. It is found to be more useful and meaningful for
low-noise amplifiers, such as masers or parametric amplifiers. In the case of
 5.12

.5.13

F 1+
 measurement
1-0...,r'\ ....... ,~ ... ..,,1- TO here the
value
K
F

F
where IS

a HHilJ-It':vel the

A 6dB
B dB
C 20dB
2-0
..... r' ...... the
1"'11 ..... lt:>,C'

arnlpllheJrS must
the . . . . . ,,.+."'. 1,,....
 B,C

F +---

025 +0-047
F=
For B

+---

1)
+--+---
4 4 x 100
+ +0025
F = 2475
then B,

aJ.J.A ........ .., .......... are, the

the no1s iness'.
4

are inter-related the .".V ...H.".""'.r-, .....

3. Both
4. effect.

In
""-.. '", ..... '""' systems.
 Noise measurement 69

case of an antenna connected directly to a receiver, Ts is defined by

1~ = 1~ +1'r
where Ta is the antenna noise temperature associated with its radiation
resistance in thermal equilibrium with the environment and Tr is the effective
input noise temperature of the receiver due to its internal noise sources.
However, for convenience, a transmission line often connects the antenna to
a receiver and gives rise to thermal losses, while a preamplifier may be used
ahead of the receiver to improve system performance. In this case, it is
convenient to refer noise temperatures of the 'receiver chain' to a point just
behind the antenna, as shown in Fig. 5.14. The system noise temperature T" is
then given by
Ts = 1'a + (L -1)To + LTp + LTr/G
where L is the loss factor of the transmission line, To is the ambient temperature,
Tp is the noise temperature of the preamplifier with power gain G, and Tr is the
effective noise temperature of the receiver.

>r l"s
L,To
I

I
Tp,G

Fig. 5.14
I

I
Tr

Example 5.4
An antenna with a noise temperature of 57 K is connected by means of a cable to a
preamplifier and receiver. The cable loss is 1 dB and the preamplifier has a 20 dB power
gain and a noise temperature of90 K. If the effective temperature of the receiver is 290 K,
what is the system noise temperature? Assume an ambient temperature of 290 K.

Solution
We have
Ts = Ta + (L -l)To + LTp + LTr/G
with L = 1 dB = 126
and G = 20dB = 100
Hence
Ts = 57 + (1'26 -1'0)290 + (1'26 x 290) + (1'26 x 290)/100
= 57 + 754+ 113-4+ 365
or Ts = 2495K
Noise

factor L. is

the nOIse t43rnn,o.ro:Jtl1'43 of the receiver

28

>

1
>
C)

> d>
C) c
w
d>
c
w

- - - ' ' - - - - - - ' - - - - W,

Th ree-Ievel
maser maser

5.15
 71

.5.16
72

t _______ Output

5.17
 measurement

Input

Input

Idler
circuit
5.18
74 Noise

performance and is fabricated by the planar process. The thin conducting GaAs
epitaxial layer is deposited on a semi-insulating GaAs substrate and the
Schottky barrier gate is formed by the deposition of a metal on the epitaxial
layer. The construction of a typical device is shown in Fig. 5.19.

Schottky
barrier
\
s \ G D

s D
Semi-insulating
GaAs

-1~
Gate
length
Fig. 5.19

At present, the most important application of GaAs FETs is in low-noise

amplifiers. GaAs FETs designed for low-noise applications have Schottky
barrier gates with typical gate length of 05 ,urn. Noise figures range from 1 dB at
4GHz to about 2dB at 12GHz with a power gain of about 10-12 dB.
A microwave low-noise amplifier may employ one or two GaAs FET stages
followed by a bipolar amplifier. Alternatively, the first stage can be an ultra low-
noise parametric amplifier followed by a low-noise FET stage. An overall noise
figure of 3 dB can' be achieved over a 1 IvlHz band in the freqi,lency range
725-775 GIlz.
FET amplifiers have other advantages, such as long shelf and working lives,
small size, and low power dissipation. They are finding increasing use in the
front end of various kinds of microwave receivers for both radar and satellite
communications. In earth station applications, they are a very reliable and cost-
effective way of implementing a low-cost satellite ground station with an
excellent figure of merit GIT.

Example 5.5
Sketch and discuss the variation of the sky noise temperature as a function of frequency.
A receiver has a system noise factor of 10 dB and it is proposed to improve its
sensitivity by adding a preamplifier of 3 dB noise factor and 10 dB power gain.
 Noise measurement 75

Solution

minimum when
horizon.
behaves

Problem
of aerial
without

290 2900K

----=841K

2900
1rn.r"\rr'l.''I7o.r'\'''!ol'>'nt factor 3-45

2668 K

with pre:amlplrDer 7., +---- K

lrn'l"'Iorr'l.'l.l01r'\"'!ot:>lnt factor 64dB

the latter dB when the

76 Noise

what the main

Assume that the bandwidth of linear IS same and noise

defined with reference 290 K.

K
1

5.20

with Lis Hence

When referred the mput we have

290
=1
280
or

Also

or 40dB

the

F= and or
or
 Noise measurement 77

5.8 Noise temperature measurement 31

The effective noise temperature Ta of an antenna is usually measured using the
'Y-factor method'. The principle of the method is to compare the noise power
received by the antenna to the noise power generated by a standard noise
source and, from the ratio cif these noise powers, Ta can be determined. The
circuit arrangement used is shown in Fig. 5.21.

Low-noise
preamplifier

rn~~wer I
! meter ,
! i

Matched
load

Fig. 5.21

With the standard liquid-cooled load connected to the input, the output
noise power Ns is noted on the meter. The antenna is now connected to the
input and the output noise power Na is noted on the meter. If the ratio NjNa is
denoted by Yand the receiver noise temperature is TR we have
k(T, + T R)B
Y - -Ns - - -'----
- Na - k(Ta + TR)B
where B is the relevant system bandwidth. Hence
Y= Ts+TR
Ta+TR
or YTa + YTR = 1~ + T R
T = ~+TR(l- Y)
and a Y
To measure TR , the output noise power N h of a 'hot load', i.e. a load at room
temperature, is compared with the noise power output N s of the standard
cooled load. If the ratio of these noise powers is Y' then we have

Y' N = k(Th+ T)B

=_h R
Ns k(Ts + TR)B
with Y'Ts + Y'T R = Th + TR
or Y'T R - TR = Th - 'f'T,
78 Noise

Th - Y'Ts
Hence T -------
R - (Y'-1)

which can be determined since Y', T h , and Ts are known by direct measurement.
The value ofTa can then be determined from the previous expression using the
known values of 7'., TR , and Y.

5.9 Excess noise ratio (ENR)32

Microwave noise generators are usually of the solid-state or gas discharge types
which use room temperature as the standard reference by simply de-energising
the noise source and assuming room temperature as the standard value To. In
this case, the effective receiver temperature T R is related to the noise factor F by
T - Y'T
T =
R
h
(Y' _ 1)
= (F -1)1'
F _ _ (Th/1'o) - Y'
with 1- (Y'-l)

or F = _(T-"h-,---IT--,o,-,-)-,---_1
(Y' -1)
where Th is the hot-load temperature of the energised source.
The quantity in the nUlneratQr is a measure of the power output of the :Q.Qise
source and is called th~_~xc,=-~s noise rqUo.(ENR) which is giveI1l:>i" - --
ENR = 10 log (ThiTo -1)dB
and F = ENR - 10 10g(Y' - 1) dB
Solid-state noise sources are semiconductor p--n diodes operating in the
avalanche region. The randomness of the avalanche multiplication process
produces fluctuations in the avalanche current which generate random noise
over a wide frequency range. Typically, the diodes operate at voltages of around
20 to 30 volts and are driven from a constant current source from between 5 to
20 rnA.
The choice between solid-state and gas discharge noise sources is based on
frequency coverage. For laboratory work below 18 GHz requiring operation at
many frequencies, solid-state units otTer economic advantages, in addition to
small size, low mass, and low power consumption. Usually, solid-state noise
sources are calibrated at several frequencies and, typically, may have an ENR of
around 15 dB in the frequency range 1-124 GHz or, in certain cases, an ENR as
high as 40 dB. They can be used to determine the noise figures of amplifiers,
mixers, or receivers and also to check the performance of radar and
communication systems.
34
80

is

components
~
'0 1------
>

o f--

6.1
 Systems 81

If S (f) is the noise power spectral density, then the power in a bandwidth (~f
associated with the single 'impulse' is S (f) (jj = No (5j where No is the power
spectral density in watts/Hz. If the corresponding single noise component is
vn(r) we have
V,,(t) = Vn sin 2nUIF+ jn)t
where };F is the IF frequency and fn is the frequency of the particular noise
frequency component with a peak value Vn and No (5} = V~/2.
When of --+ 0, all the noise components cover the IF bandwidth 2B
continuously and the total noise power N is

J::
I

N, = No 6} = 2NoB

S m2 V 2 J m 2 V2 m2 P
and 1/\';, = ~j 2N oB = 8No~ = 4No~
where Pc is the average carrier power.
To obtain the output noise power oflhe detector, each noise component v,,(t)
within a bandwidth ~rwill beat with the carrier and the resultant voltage is given
approximately by
Vn(t) = Vc [1 +(Vn/Vc)coswntJsinwct
if Vc }> Vn . When this is applied to the detector, the output noise voltage
no(t) = Vncoswnt. Hence, the average noise power (jNo in a bandwidth of is
6N o = V;/2 = No~f (as obtained earlier)
and for the whole IF bandwidth this becomes

No = f +B

-B
No df = 2N () B

S m2 V 2 I m l V2 m2 P
Hence -;'- = _ _c / 2N 0 B = - - " = - - "
No 2 I 4N oB 2NoB
which is equal to twice S,/N, or
(So/NolAM = 2(SjNJAM
and amounts to a 3 dB improvement at the detector. The 3 dB improvement is
due to arithmetic addition of power in the sidebands and quadratic addition of
the independent noise in each sideband.

DSBSC system
The only difference between an AM system and a DSBSC system is the carrier
power which is present in the former. Hence, for equal average power in the
X2 N01.\e

the ratlO5J the two must same,

and

amounts to a 6

on
. . . . . . . . . . .,. . . . . '. . . """ ...., it can

opf~ratlon of the en~/eU)ne detectors for use in the

C'uY"U''''hrn1nnllC'

an amlplltu(le-lmodu.latc~d wave under conditions of very low

 83

or

v(t)

carner and the

The and with the blocked

84 Noise

[
and ...... "' ..... .,. ....~ ,...,.. """I .. ' from to due to

If

since is small. Hence

cos
The does not contain the modulation n'V''O',ty,.'fh, as last
,...,,,,"'I"1""'A'11C" from 0 to
'\1<:lr''\l1r1,n 1,\1 the modulation
called the AM

Comment
The results obtained above show in the of the C" .. ,9',,.. ...................... ,,'" detector. there no
threshold effect and the modulation is """ ..."'.... ""'' ' ' . . embedded
noise. In the case the ~n"~I'~ del:ect:or. the threshold
effect noise is be useful
when SI

an IS
 Systems 85

where K is a constant of proportionality. The average signal power is So where

Kl !J.f2
So = 2 watts

To evaluate the effect of random noise, we observe that each noise frequency
componentJ;, will beat with the carrier wave to produce amplitude modulation
and angle modulation as illustrated in Fig. 6.3.

Fig. 6.3

If the noise component has a peak voltage Vn where Vc ~ Vn we obtain

vn(t) = (Vc + Vn cos wnt) -+ j Vn sin wnt
= Vc [{l + (Vnl Vc)cos wnt} + j (V,jVc) sin wnt]
or Vn (t) :::::: Vc sin (wet + 8)
(J = -1 (VnIVc)sinWnl (T' I ).
where tan :::::: Vn Vc smwnt
1 + (Vn/ Vc )cos wnt '
since Vc ~ Vn
The output noise voltage from the discriminator Vd(t) will be proportional to
the frequency modulation produced by the noise signal Vn (t) which is related to
the phase modulation it produces by
1 de Vn
Vd(t)
.
= K2n:- -dt = K-.
~/ c
fn cos Wilt

The average output noise power bIVo in a bandwidth of is then given by

K2 V 2j'2
~ 11.1 D n
OlV 0 = 2
2Vc
Previously it was shown that N oof = V;/2 and Pc = V~ /2 is the average
carrier power. Hence
= K2(N0 (1)f2n
.sN
o 2Pc
 In BHz

or

Hence

or
B

B ~f

.6.4

ratio an ""' . . . . 1=............ ""

B
Hence

In particular, if ~f = 75 kHz and B = 15 kHz we have m f = 5 and so the

signal-to-noise improvement due to FM is 3 x 25 = 75 or 19 dB. The factor of
75 can be increased further by the use of pre-emphasis at the transmitter and de-
emphasis at the receiver. It can be shown'~ that this amounts to about 4 dB.
giving an overall SIN improvement compared to AIv1 of 23 dB.

6.2 S/N ratnos

Typical graphs of input and output SIN ratios are shown in Fig. 6.5, assuming
the same input noise bandwidth for the vanous systems. A t Jaw values of S,,' N"
the AM systems appear to be the best, but for S,/ N, above the 10 dB threshold,
FM is superior to the AM systems. To achieve large values of Sol No> however,
PCM appears to offer the best advantage. As a comparison, the ideal system
shown is still about 8 to 10 dB better than the FM or PCM system.

LpCM
-----;~-i----:;7~1 O.dlglt code)
AM

30dB

..Y. per"l
(5dlglt code)

o (S;iN,) dB

Fig. 6.5

The most familiar pulse systems used are pulse amplitude modulatJOfl (PAM).
pulse position modulation (PPM), and pulse code modulation (peM).

" Sec F R Connor. ModliiarlOli. Edward Arnold (19R2)

88 Noise

PAM system
Pulse amplitude modulation is normally employed in the early stages of PPM
or PCM systems since it is easy to multiplex PAM pulses. It is generally not
used as the final system, however, since the SIN ratio obtainable is not as good
as those of the other pulse systems.
It can be shown that PAM gives results very similar to those obtained
previously for AM. Since noise directly affects the amplitude of the pulses, it
appears as direct amplitude modulation in the system, as with the modulating
signal. Hence, we have

PPM system
In pulse position modulation, the modulating voltage causes a time displace-
ment of the pulse. To evaluate the SIN ratio, assume that the maximum time
displacement due to the modulation is to and so the peak signal volts is equal to
K Co where K is a constant of proportionality. Hence, the average signal power
at the output of the detector is
So = K2t5/2
The effect of noise in the signal is to alter the time displacement which leads to
an error e, as shown in Fig. 6.6. The rms noise volts causing the error e produces
an rms time displacement lit such that
e I1t
=
Vc tr

where Vc is the peak pulse volts and tr is the rise-time of the leading pulse edge
which carries the modulation.

/With noise
v ..../ "

t-
Fig. 6.6

The output rms noise volts is thus K I1t and the average output noise power at
the detector is
Hence

.....o r ' .. 't'"orl to

IS mean

or

Quantlsed

Analogue signal

E~============~

Sampling Quantised signal

6.7
90 Noise

The error c: can take on all possible values between -/1 V/2 and + /1 V/2 and
may be considered as due to added noise in the signal. Hence, the mean-square
value of the error gives the mean-square value of 'quantisation noise'. To
calculate it, assume that over a long period of time all levels have an equal
probability of occurrence and so the occurrence of any level is the same. Hence,
we obtain

L1 V 2
or N
o
= -12- watts (for a 1 Q load)

To calculate the signal power for q levels spaced L1 V volts apart we have
v= (q -1).6. V volts

Assuming further that bJpolar pulses are used (since less power is consumed),
the pulse heIghts are 6 r /2. 311 Ii /2 ..... (q - 1)11 V/2. For equal
probabIllty of occurrence of all levels 111 a long message. we obtain the average
Signal power as

61 2 0 _, _0
or 51 = - 2q
-- [! - + j - + y +. .. + (I} - 1) 2 J
-"

This may be written as

NOlA

from whlc]) we obtaIn

,q(q-l)(2(/-1)
+ 2- + ... + (If -
0 ' .
1- 1)- = - ...._ -
6
_,
1~+2~+.,.+
(l -q _2')"
..... .
q(ij - l)(q - 2)

\ 2 / 4x6
L1 j 2
Hence So = -'1") [1/((/ - 1)(2ij - 1) - (j(q - l)(Cf - 2)J
Lit
,61 '2
= ._- (lj-l)I'{2(j- IJ-Uf-2)]
12 '.. ,
 Systems 91

AV 2
=U(q-l)(q+l)

AV 2
=_(q2_1)
12
AV2
or So ~ 12 q2 (for q ~ 1)

Hence, the signal-to-noise ratio due to quantisation noise becomes

AV2 /AV 2
(S /N )PCM ~ _ _ (q2 -1) - -
o 0 - 12 ! 12
~(q2_1)

or (for q ~ 1)
The result depends on the square of the number oflevels used and so a large
number of levels is required for a large SIN ratio. As an example, if q = 128,
(So/No) ~ (128)2 ~ 42dB. This requires theuseofa 7-digit code since 27 = 128
levels.

According to the Hartley-Shannon law of information, the communication

capacity C of a system with a bandwidth B and a signal-to-noise ratio SIN is
given by
C = B log] (1 + SIN) bits/s
This rate of information transmission may be regarded as the ideal if it is
assumed that the error rate is less than 1 in 10 5 bits/so Comparing this with a
binary PCM system, we observe that, for a sampling frequency of 2 HI where HI
is the highest modulating frequency and q quantised levels are used, the amount
of information H transmitted is given by
H = log2 q bits
For a sampling frequency of 2 W, at least one pulse is sent in each sampling
period and so the total number of pulses sent per second is n = 2 W. Hence
C = H' = nH = 2Wlog 2 qbits/s
or C = B 10g2 q2 bits/s
since B = 2 HI is the system bandwidth.
Previously it was shown that
AV2
So = 12 (q2 -1)
 C B

C=B

ratio

ratio of was obtained """..."'........."...... 1.. ,

where q is the number of . . . ",....... f-'C"oA

For

nB
or n
group

or
 Ideal

5
PCM

40 PCM

-
:::c
23
:3 8

peM

Error 10

6.8
 A rot
=0

~ one

two ';:).lJ:.".l.lU.l';:) p 1, we

and well the bit ratio is the same.

-E/No
Noise

Table 6.1

1
'2

SER 10- 3

6.9
 Systems 97

budget, the critical parameter to be considered is the effective isotropic radiated

power (EIRP) at the satellite which is important in achieving maximum power
output from its transmitter. It is given by the expression

EIRP
PTG T
=----------------
LFLS
or EIRP = PT+GT-LF-L s dB
where P T = transmitter power
LF = feeder and diplexer loss
Ls = free space loss
G T = transmitting antenna gain

Given below are typical values of the relevant parameters at an operating

frequency of 6 GHz for a geostationary satellite orbiting at a distance of about
36000 km above the earth. The losses include the large space loss (201 dB) and
the miscellaneous losses due to transmitter ageing (1 dB), antenna pointing
error (2 dB), and rain attenuation margin (2 dB).

Transmitter power 23dBW

Transmitting antenna gain 60dB
EIRP 83dBW
Free space loss (4nd 21A2) -201 dB
Satellite antenna gain 28dB
Miscellaneous losses -5dB
Received carrier power -95dBW
Satellite noise power -126dBW
Carrier-to-noise power density ratio 97dBHz
Carrier-to-noise power ratio 31 dB

The critical downlink parameter is the figure of merit G IT since it directly

determines the carrier-to-noise power density ratio (C / No), which is the
ultimate criterion at the receiver's demodulator for analogue signals, or the bit
energy-to-noise power density (E I No) for digitaJ signals. Hence, it is easily
shown that
C EIRP X GR
No kTsLsLm M
C
or -N = EIRP+GR-kTs-Ls-Lm-M dBHz
- 0

and
98 Noise

where C = carrier power at the demodulator

No = noise power spectral density
N = noise power
EIRP = effective isotropic radiated power
G R = receiving antenna gain
k = Boltzmann's constant
T, = system noise temperature
Ls = free space loss
Lm = miscellaneous losses
114 = margin for multiple carriers
J-V = relevant bandwidth
Typical figures for a ground station receiving a signal at 4 GHz from a
geostationary satellite are given below. It is convenient here to consider the
parameter (C I No) for analogue signals since it is independent of the system
bandwidth used. For digital signals, the ratio (EI No) is easily determined from
C Cr 1 ER
-=-x-=-
No No r No
E C
or -=--dB
No NoR
where T is the bit duration, E is the energy per bit, and R is the bit rate.

Satellite EIRP 18 dBW

Free space loss 197 dB
Receiving antenna gain 60dB
System noise temperature 70 K
GIT ratio 41-5dB/K
Miscellaneous losses - 5 dB
Received carrier power -124dBW
Noise power density -210dBW/Hz
Carrier-to-noise power density ratio 86 dB Hz
Carrier-to-noise power ratio 20dB
Bit rate (digital systems)* 85 x 10 6 bits/s
Energy-to-noise power density ratio 17 dB
BER 1 x 10"4

Here, the system noise temperature includes the antenna noise temperature
and that of the subsequent receiver chain, the losses include the large space loss
(197 dB) and the miscellaneous losses due to atmospheric attenuation (2 dB),
antenna aperture efficiency (1 dB), and mispointing and polarisation loss (2 dB).

,. 132 channels
 Qlagra.ms the arr'an,gerneIlt of a microwave satellite
communication an active
.........'...., .. I.... ...., ......... ,"'... earth in geo-

par'amete:rs and

6.1

station eqlUl]:lm<~nt corlslsts chain and . . a,....,.,.l',.",,,,,

t-""'lInC'lI"""lt-t-a ....

which are isolated modulators drive

IF of 70 MHz The ",.n,n~rol'Ufj"'r arnpl1lher .. """" ......... &r,
C''lIf-,o.lh1~a the

trallsrrUSSlon back earth.

received with a bandwidth up to
MHz band at 4 GHz and so low-noise wideband are ""' ...... ,o.n1l ....
100 Noise

15 K followed
demodulators are 1"'1'"'\1n''I1t:~1nt11'"'\'I''\n
the and RX use conventional oa:set)an.a
 I

7
102 Noise

8 Derive an expression for the overall noise factor of a combination of three

two-port networks, connected in cascade, expressed in terms of the
individual noise factors and gains of the individual networks
A receiving system consists of a preamplifier connected through a length
of cable to a main receiver. The noise factor of the preamplifier is 6 dB,
while the corresponding values for the cable section and receiver are 8 and
13 dB. Given the attenuation of the cable is 8 dB, calculate the minimum
gain required in the preamplifier if the overall noise factor of the system is
not to exceed 9 dB. (U.L.)
9 (a) Describe briefly what is meant by the noise factor of a radio receiver.
(b) Two amplifiers, connected in series, are matched in impedance and
bandwidth. The first has a gain of 16 dB and a noise factor of 6 dB, the
second has a gain of 10 dB and a noise factor of 3 dB. Calculate (i) the
noise factor of the combination and (ii) the noise factor if the order of
the amplifiers is reversed. (CG,L.I.)
10 A parabolic antenna has a noise temperature of 60 K and it is connected by
a length of waveguide to a parametric preamplifier. The waveguide has a
loss of 1 dB while the preamplifier has an effective noise temperature of
77 K and a gain of 20 dB. If a receiver is used with the preamplifier and its
noise factor is 10 dB, what is the effective noise temperature of the system?
11 A communication receiver system has an input stage with a noise
temperature of 100 K and a loss of 5 dB. This is followed by three IF
amplifier modules, each with a gain of 10 dB, a bandwidth of 6 MHz, and a
noise figure of 13 dB. With the system matched throughout, calculate
(a) the system noise figure,
(b) the equivalent noise temperature of the system when connected to an
aerial with a noise temperature of 50 K,
(c) the smallest usable input signal power if the communication recog-
nition system operates correctly for an output signal-to-noise ratio
> 1,
(d) the effect on (a), (b), and (c) of adding a preamplifier with a noise
temperature of 100 K and gain of 20 dB. The receiver system is at a
temperature of 290 K. (eE.l.)
12 A signal s(t) is a triangular pulse of the form
s(t) = Kt O~t~T

s(t) = 0 at all other values of time

where K is a constant. Determine the output of a filter matched to this signal.
If Gaussian white noise of zero mean value and noise spectral density No
(positive frequencies only) is added to the signal, what is the maximum
signal-to-noise ratio at the output of the matched filter?
13 In a pulse radar system, the observed signal is received in the presence of
Gaussian noise of zero mean and unit variance. Assuming the received
signal is of 2 volt amplitude, determine for a Neyman-Pearson receiver
 Problems 103

the probability of detection when the false alarm probability is set at 02.
14 A 30 channel PCM system with uniform quantisation and a 7-bit binary
code has an output bit rate of 15 Mbits!s. Determine
(a) the maximum information band\vidth over which satisfactory oper-
ation is possible,
(b) the output signal-to-quantising noise ratio for an input sinusoidal
signal at a frequency of 3 kHz and maximum design amplitude.
15 Calculate the signal-to-noise ratio for a sinusoidal signal quantised into
M levels given that the total mean-square quantising noise voltage
(T2 = 0083(I\V)2 where 11v is the step size. What assumption has been made

about the quantisation? Hence, estimate the number of digits per character
required in a PCM system carrying the above signal if the quality has to be
satisfactory for the telephone system.
Discuss whether speech processed in a similar manner has the same
quality. Explain how the signal-to-noise ratio for speech can be improved
by analogue or digital signal processing. (C.E.I.)
16 A coherent binary data system uses on-otT pulses varying in amplitude
from 0 to V volts. The probability of a 0 or 1 in the presence of Gaussian
noise is the same. For a peak signal power to average noise power ratio of
13 dB, calculate the probability of error. Also, for a probability of error of
10 - 5, determine the signal-to-noise amplitude threshold required.
17 An FSK communication channel transmits binary information at a bit rate
of 100 kbits!s in the presence of Gaussian noise with a spectral density of
10- 19 W 1Hz. If the signal is transmitted with a peak voltage level of 1 volt,
determine for a probability of error of 10 - 4 the path loss of the channel for
bit by bit detection with
(a) incoherent FSI(,
(b) coherent FSK.
18 Explain why communications satellites fulfil an important role in world-
wide communications. Mention, particularly, aspects of the role that
cannot be readily fulfilled by alternative systems.
A satellite in a geostationary orbit at 35800 km has a 4 GHz downlink
transmitter which feeds 25 watts into an antenna with 20 dB gain. The
ground station receiving system has a total noise figure of 15 dB.
Calculate the antenna gain necessary at the ground station to maintain a
30 dB input signal-to-noise ratio over a 12 MHz signal band. (eE.L)
19 The parameters of a satellite-to-ship link are as follows

Satellite RF transmitter power per channel +PsdBW

Satellite aerial gain relative to isotropic + 17 dB
Free space path loss at 15 GHz -189 dB
Propagation margin at 5 elevation for 99 % of the time +5dB
Ship aerial gain relative to isotropic +GrdB
RF input power to ship receiver -152dBW
20
 4
5

8
9

11
 STEINBERG, J McGraw-Hill
2 Communication

and Communication Inter-

4
5

8
9 Modulation

1
12 JOHNSON,
cation r"h't:ln1n.t:lIl

13 WIENER. N Generalised harmonic

14 KHINTCHINE, A Korrelationstheorie
1934.
15

16 and RAJASEKARAN,
flP1JllOQlUJns. John
,,1" B
... \J.J" ....." ... .J Thermal n"".1'" .... 1'" ............. t:lIlt:lIt"t1"'1r"1i"u in C01001uct:or:s, ...,"',,,,,.ro,. . ,

18 ...nf"t:lf" ...,'n of electric

and

20

21

Electronic

Advances
24 t'r4ryCE~ealn~7S Institute

25 Electrical
26 Ultra low-noise 1I"'\f"lI1.. <'lI'I"1""""t ........ arnlpllhelrs in communication Sa(~eUl1[e
........ ,u...... " ......... ..,., Advances in 1971.
Low-Noise The

28 lortrnnlr Communication
29 and ROWE, H E t'rllcedeOljrzas Institute

30 POSNER, R of GaAs FET low-noise

arrlpllne:rs~ Microwave
31 KREUTEL, wand PACHOLDER, A 0 Measurement tprnf"loJo.rgtl11-p of a
satellite communications 1969.
R and aplpU(~at],on
sources. M ullard
33 CARLSON, A Communication
34 and SCHILLING, D L McGraw-Hill

35 F Introduction
36 PEEBLES, P z. Communication
OLIVER, B M. and PIERCE, J R The Ph110:S0t)hV
November 1948.
SHANNON, the presence of t'rc')Ct~ealnG'S Institute
Radio bnazn!eel's
39 LAWTON,J.G '"'-'v' ....... iJ ......... t..7'U.....

Convention on
40 MARTIN, J Communication Satellite Prentice-Hall
41 TURNER, L W Electronic Book. Newnes-Butterworth
Venn rI.:;;>,nr""'rrl

A.1

s
(a) s

(b)

A.2

s not
110 Noise

was theorem.

Error . . ,," ............ '1 ..........

To

Power

x
we
1
T
IS x
 Appendices 111

Table A.1

x erf x x erf x

0 0 1 50 09661
005 00563 1 55 09716
010 01124 1 60 09763
0'15 01679 1 65 09803
020 02227 1 70 09837
025 02763 1 75 09866
030 () 3286 1 80 09890
035 03793 1 85 09911
040 04283 1 90 09927
045 04754 1 95 09941
050 05205 200 09953
055 05663 205 09962
060 06038 210 09970
065 o 642() 2 15 09976
070 06778 220 09981
075 07111 22E, () 9985
080 074;;.1 2 30 09988
085 On06 2. 35 o 9991
090 07969 240 09993
095 08208 245 09994
1 00 08427 250 09995
1 05 08624 255 09996
110 08802 260 09997
1 15 08961 265 09998
1 20 09103 270 099986
1 25 09229 275 099989
1 30 09340 280 099992
1 35 09437 285 099994
1 40 () 9522 290 099995
1 45 09596 295 099996
300 099997

xiI) X{;) ,!

.V'\
-f~O
3T T t~
-"2 2"

(a) fb)
Fig, A,3

, Hence, rearranging the order of integration then yields

+ T I L2.
7J2
1
X (t)dt = -7-
r
+W

~rrT - '"
F(w)dw
[ f
-TIL
-I TIL-
X (t)ej<"'dt
]
112 Noise

or
1 ['+Tl2
- X2(t)dt = ----------
1 f+ 00
F(w)dw
f+ 00
X(t)eJw1dt
T ~ ---Tl2 2nT - a : ; - - - 00

since X (t) is zero over the intervals - 00 < t < - Tj2 and + Tj2 < t < + 00.
Also, we have
,,+ 00

J_ 00 X(t)eJw'dt = F* (w)

where F* (w) is the conjugate of F(w) such that F(w) F* (w) = 1F(w) 12. Hence
1__
-
f+1i 2 X2{l)dl

= -1 f+08IF(W)12 dw
T -7/2 2n - 00 T

As more 'samples' of x(t) are removed over time intervals of T and added to
X(t), the signal X(t) will eventually resemble the periodic signal x(t) and so in
the limit when T ---} 00 we obtain

1 J~ + J /2
lim ::-::
T ---+ ~ 7 _ Ti2
1
X2(t) dt = lim ::-::
7 -> oc 7
f + l/2
_ T2
x 2(t) dt = lim -
T -> oc;
1 : + ex) F (w) 12
i
2n J - 00
1

"dw
7

The quantity in the centre is the average power of the periodic signal x(t) and
so we obtain

P dV
.
= Jlln -
1 J' + 08 1 F(w) 12-dw = --
1 :
I
+ 00
S(w)dw
7-+0C;2n -00 T 2nj_00

from the defmition of S(w). Hence, by inspection, we obtain

1F(w) 12
S(w) = lim
l-+OC; 7

V oltage spectral density

The voltage spectral density Svtfl is defined by

f::SJf)dj=V~
where v ~ is the mean-square noise voltage. In the case of thermal noise, the
noise spectrum is constant over a finite bandwidth B. Hence

f +B
-B Sy(j)df=4k7BR

or SvCf) f +B

-B
dI = 4kTBR
 1

Hence
or

we

or

e
14

~"I'I\'U/h~nil1 noise

f --IJIoo-

f--.....
(b)

A.4

+
 115

or

Hence

A.5
116 Noise

lDDlendlx F:
a
+

or

and

IH

N
 Appendices 117

Furthermore, if E is the signal energy we have

+OO 1 f+oo
E = f S2(t) dt = - IS(wW dw
- 00 2n - 00

and substituting this expression into (S/N)max yields

a result which depends on the signal energy but is independent of its waveform.
If the impulse response of the matched filter is h(t), the Fourier transform of an
impulse <5{t) is unity and we have

H( ) = F[h(t)] = F[h(t)]
w F[b(t)]
where H(w) = S*(w)e- Jwto. Taking the inverse transform then yields
1
h(t)=-
f+ 00
S*(w)e-JWIOxeJwtdw
2n - 00

or h(t) = -1 f+ 00
S*(w)eJw(t-ro)dw
2n - 00

and the r.h.s. is simply the inverse transform expression for the signal s* (t - to).
Hence, it follows that
h(t) = s* (t - to) = s(to - t)
and the impulse response of the matched filter is the input signal delayed and
time-reversed. For h (t) to be real, or for the filter to be physically realisable, we
must have t ~ to, i.e. all the signal energy in s(t) must be received before the
decision is made at t = to.

Correlation detector
If the input signal to the matched filter is s(t), the output signal so(t) is given by
the convolution integral as

so(t) = f+OO.. 00 s(T)h(t - T) dT

where r is an arbitrary variable, h(t) = s(to - t) is the impulse response of the

matched filter, and h (t - T) = S(to - t + T). Hence
Noise

G:

. A.6
 costs

Hence C IP + P

or

L
Noise

+00
p

may

A.7

a
 Appendices 121

when it is present at the receiver, and so choosing H 2 correctly_ Similarly, we

have
P F =P(H 2 Ix 1 )= r+ cc
pdylxddy
~ },

which is the probability that x 2 (t) was not transmitted, yet deciding it is present
at the receiver, and so choosing H 2 incorrectly_
To apply this criterion, it is usual to decide on a maximum acceptable false
alarm probability PF and then to maximise the detection probability PD - The
resulting values of P D and P F when plotted yield the receiver operating
characteristics shown in Fig. A.8. Parameter d is the signal-to-noise voltage
ratio at the output of the receiver.

10~
I

075

t
05

o 025 05 075 10

PF -

Fig. A.S

For any given false alarm probability P F , a corresponding threshold value YI

is determined to yield the maximum value for the detection probability P D - If
y(t) > YI' hypothesis H2 is chosen and, if y(t) < Y1' hypothesis HI is chosen.
To show the relationship with Bayes' criterion, we obtain the derivatives of
 d
or

(z I (k or 1
 we

or

In

where

=0

=0
 1

we 1

{ =1

as

is as a minimum
 o

........----- I -------t

(a) (b)
A.9

or m

or
a t

dW

or N= watts

dW=---
e
126 Noise

where f is the frequency of oscillation and h is Planck's constant. For

frequencies up to about 1013 Hz, L1 W :::::: kT which is the classical value.

Appendix J: Shot noise

The shot noise rms current Is in a diode is due to the random emission of
electrons from the cathode. Each electron arriving at the anode carries a discrete
electronic charge e which gives rise to a current pulse i(t) in the anode during the
transit time T, as shown in Fig. A.l0(a). The actual shape of the current pulse is
immaterial if the time-average interval chosen is such that T ~ T. Each pulse can
be regarded as a Dirac delta function (5(t) and approximated by a short
rectangular pulse, as illustrated in Fig. A.I0(b). Hence, we have

l+: b{t) dt = e

l.e. the area of the rectangular pulse is such that e/T x T = e.

~
0
r- T
-1 t~

la)

Fig. A.10

If F (w) IS the Fourier transform of b(t) then

sin OJ, /2
F (OJ) = f+
-
'L

00
.
b (t) e - Jrot d t = e
OJT/2
.

and IF (wW = e2 [~-i-~;;~T

where IF (OJW is the energy spectral density and is shown in Fig. A.10(c).
From Fig. A.lO(c) we observe that if the transit time T is very small (about
10 - 9 s) then liT:::::: 10 9 Hz and the spectral density over a bandwidth L1f = B is
fairly constant, especially at lower frequencies. Hence, the total energy W in a
bandwidth B is given by

W= J1'+_""IF(w)
00
12 d f = 2
f+oo
IF(wWdf
0

or
 Appendices 127

If n electrons arrive at the anode in time T where T is sufficiently large, the

average shot noise power in a 1 Q load becomes
I; = n ~VIT = n2e 2 BIT
and substituting for the average anode current 1a = ne IT yields
I; = 2el"B
~
or Is =y'2el a B

Appendix K: Noise factors

Grounded-cathode circuit

Fig. A.11

Grounded-cathode amplifier
The equivalent circuit is shown in Fig. A.lI where Rs is the source resistance, Rg
is the grid-leak resistance, and Req is the equivalent shot noise resistance of the
valve. If their rms noise voltages are 1.:" vg, and Veq respectively, the total noise
voltage of the amplifier is Vo and for an ideal amplifier it is V1 where VI is due to
the source resistance only. Hence
mean-square noise voltage of amplifier v;
F= ................................ _ ...................................................... = -
mean-square noise voltage due to source vf
4kTB[Req + RsRg/{Rs + Rg)]
[vsR~/(Rs + Rg)] 2
Since v; = 4kTBR s, we obtain
[Req (Rs + Rg) + RsRgJ (R, + Rg)
F = --'---=---"----"-'----::---"-=----'-~
RsR;

or F= 1+RjRg +i q [1+RjR g]2

s
128 Noise

Grounded-grid amplifier
The eq uivalent circuit is shown in Fig. A.Il. Since the effective amplification
factor of the grounded-grid amplifier is (/1 + 1), the rms noise voltages Vs and Vg
appear as (/1 + 1)u, and (/1 + 1)v~ in the equivalent circuit. The grid noise voltage
veq , however, is common to both grid and anode circuits and so its value is
(p + l)vg - ug = pUg in the equivalent circuit. Hence, we obtain
mean-square noise voltage of amplifier
F=--------------------------------
mean-square noise voltage due to source

/124kTBReQ + (p + If 4kTBR sR g/(R, + Rg)

(/1 + 1)24k7BR, [Rg/(R, + Rg) J2

[/1/{fl + 1) J2 RCq + R,Rg/(R, + Rg)

i(tRJO~:+R~")ji---

or F = 1 + Rj Rg + (_/1_)2 Req [1 + R,,/R gJ2

\/1 + 1 R,

Common-base transistor

Common base circuit Commonemltter circuit

Fig. A.12

The equivalent T-circuit is shown in Fig. A.12 where R, is the source resistance
and rc , rb, and rc are the emitter, base, and collector resistances respectively. The
rms t101Se voltages v, and Vb are due to thermal noise in the source and base
resistances respectively, whIle ve and 1\ are due to shot noise and partition noise
Il1 the emitter and collector regions respectively. Hence, we have for a
bandwidth B and absolute temperature T
v: = 4kTBR,
v~ = 4kTBrb
vc2 = [2,.2
,e
= 2kTBr c
since f: = 2efoB and re = kT/elt if Ie is the d.c. emitter current Also, the
 is

(Ial = +

F = --------------

source

or + +

source

or

F
Noise

varIOUS

F = 1+ +-+
(1 re ]

Shot

(kHz)--
A.13

if

Vg

A.14
 current

F=

----+

1
t t

A.15

For
ance, gi
 to zero, we

error
IS .......... ,L ........................ " " .....

2E ]

two
 was the

If/=

or

-J
it

]
/2
I

VUJ.JlUVA"-' noise power, 55

Band-limited

coloured
communication 91
conditional nrl"\t'VJIt'\111T'U
correlation

113, 126
 91

14
UU ....,LIV.U. 9, 17

108

Low-noise . . . . . . 'I"\h1ho,... 74

lUCl.l\.lU.lUlJU-llA'-'lUI\...'VU '-'.:>lJIIIClllVJLI. 41
Random
7

Ul.:>IL.llU'ULlVll. 15

10

112

White
Wiener-Khintchine 1-1->,..,.,..._,... ......... 13

110