NOISE IN COMMU·NICATION

SYSTEMS

A
11 communication systems are affected by noise. The extent of noise v~es from system to system
and channel to channel. For example, some channels such as fiberopt1c channels are less noisy
in comparison to radio channels. Noise is a random process and. characterized by probabilistic
description (probability density/distribution function) and spectral description (power spectral density); these
concepts have already been discussed in Chapter 6. Furthermore, in addition to channel noise, the receiver
itself contributes to the noise. What it means is that even if the channel is completely noise-free, the received
signal will be affected by the receiving system noise which in turn is generated by its constituent subsystems
and components.
In this chapter we shall first discuss the noise in the receiving system and relevant parameters to describe
it. Later, the effect of channel noise on analog modulated signals; both amplitude modulation and its variants
and angle modulation will be examined. The performance measure will be signal-to-noise ratio.
The receiving system noise is primarily caused by two mechanisms. The first is the noise generated due
·to system being at a physical temperature other than absolute zero. This is known as thermal noise and is
discussed below in more detail. The second m~chanism is the signal detection process itself where the signal
is detected in discrete chunks even for the continuously varying signal. This is referred to as shot noise. This
type of noise forms an important component in detection used in optical communication systems.

7. 1 THERMAL NOISE
Thermal noise _has its origin in random motion of electrons in passive resistors. This was first measured by
J. ~- Johnson m 1928 at Bell Labs. He con~uc~ed an experiment wherein an amplifier was excited by a
resistor and the output cun:ent was measured m different frequency bands. He observed that the mean square
?utput current was a func?on of_temperatu~e and was essentially independent of frequency.1 Thennal noise
is also c_alled Jo~n no1Se. _This observation was further supported in the same year by H. Nyquist of Bell
Labs usmg theoretical analysis based on thermodynamics and statisti"cal h · 2
Th rmal · • . mec arucs .
.e n01se is essentially fluctuation phenomenon caused by electron-atom collisions. This pro-
duces random
.h voltage across termmals
. of a resistance · The phenomenon is . govemed bya thennodyn"mics
Iaw wh1c states·that the mean kinetic energy per degree of fr d f th . . . b
(U . . .

w·parti·
. . . ( 1) ee om o e particle 1s given y eq
non of energy which is 2 kT, where k is Boltzmann's constant (1. 38 x 10-23 Jj°K) and Tis the absolute
temperature. Johnson observed that the , mean square noise voltage , 2 · the f·requency band B 1s
vn, m · gt·ven by

v~ = 4kTRB (7.0
448
 7. 1 Thermal Noise 449

noise voltage is proportional to resistance R and absolute temperature T. The above expres-
Jllean square_ artition of energy. We note that v~ is the noise power generated by resistance R in the
~e refleCts equipB Let Sn(f) be the two-sided noise power spectral density (PSD). Then in view of the
~on band.2 . , .. d '
0encY the observed nmse power IS m ependent of frequency in band B the following relation
'
~serv
\olds· Sn(f) 2B = 4 kTRB (7.2a)

or Sn(f) = 2 kTR (7.2b)

noise is modeled as white and Gaussian. Latter is the consequence of central limit theorem. We
,... 5 therJilal . · · • •·
1uu ' , represent a n01sy open-crrcmt resistor as a n01se-free resistor excited by the root mean square
can there1 ore _
source P, = kT RB.
voltarshould be noted that the flat power d~nsity spectrum inq.icated by Eq. (7.2) is not entirely correct.
fu the iiroit, the integral leading to the total thermal noise power diverges which is physically inadmissible.
Toe theoretical reasons for arriving on the whiteness of the power spectrum are based on classical statistical
mechanics. The quantum mechanical treatment of the phenomenon shows an upper cut-off frequency beyond
which spectrum tends to roll-off, thus leading to non-white power spectrum. This cut-off frequency is roughly
!h =(¥), where his the Planck's constant(= 6.636x 10- 34 Js), which is about 6 THz. This cut-off frequency
ismuch beyond the normal operating radio frequencies. Therefore, for all practical purposes, thermal noise
can accurately be modeled as white.
The spectral excitation model of the resistance is a simple modification of Fig. 7 .1, where noise power
spectrum is considered. This is shown in Fig. 7 .2. · ·
This model allows us to study the frequency dependence of systems following noise sources.
:
·,
_____... . . ., 1---.;- - ---<l
R 1

..... -.............. _..... ..

-- -~--.
..,

..
'
Noise-free

'' R
'
Noisy

Figure 7.1 Equivalence of noisy resistance with noise-free resistance and a noise source.

,- ..... -.............. -
:
'
'
R
.:
Noise-free resistance

~ Sn(f) =2 kTR

Figure 7.2 Equivalent noise source and noise-free resistance .

~II.,•
-~ls•-
ln ~-1 '-'lJ's in S
de _&. 7.3 . eries
t ns1tie8, ' resistor (f) d S (/)
a.n1-_ · Sine s R1 and R 2 are noise free and Sn1 an n2 are respective noise power spectral
ve au e the · · · 11 · ct
ded and noise sources due to resistors are s~UStica Y m ependent, tneir power spectral densities
th e resultant equivalent circuit for noise power spectrum is shown in Fig. 7.4.
M
450 NOISE IN COMMUNICATION SYSTEMS

,-.., S,df) =2 kT(R 1 + Ri)

Figure 7 .3 Noise sources for resistors in series .. Figure 7.4 Equivalent noise circuit for res· .
1sors
1 1n leri
es.

Resistors in Parallel
The noise sources and noise-free resistances are shown in Fig. 7.5.
Since the two noise sources are independent, superposition can be used. The contribution of R toSn0(J]
while keeping Sn2 (f) = 0 (noise source due to R2 is dead) is given by
1

2
2
Sno1 (f) = Sn1 (f) ( Ri R+ R )
2
(7.Ja)

Similarly, the contribution of R2 to Sno(f) is given by

(7.3b)

Therefore, the output noise PSD, Sno(f), is given by

(7.4a)
Sno(f) = Sno1 (f) + Sno2(f)
or
2 2 (7.4b)
Snolf) = Sn1 (f) ( R R2 ) + Sn2(f) ( Ri )
1 + R2 R1 +R2
UsingSn1(f) = 2kTR1 andSn2(f) = 2kTR2

Sno(f) = 2kT [ R1 ( R2 )2 + R2 ( R1 )2]

R1 +R2 R1 +R2
or
Sno(f) = 2kT RiR 2
(R1 + R2) = 2kTReq
where Req = .!JJ..!.!L
R i + R2 • ·ve tl'IO'
. . . . a pas~1 1ed b1
The r~sul~ nt n01_s~ equivalent c1rcu1t is shown in Fig. 7 .6. In the general case ~be: contfibll
terminal. crrcmt _conta.J.mng reactance is represented by its Thevenin impedance, the noise t at ~e
its resistive (which can be frequency dependent) part ed tb 1el1
1101
· 7. .7 , Z T represe nts th e Th,evenin impedance
· .,ninalS I·s s
looking into terminals. It sboUId be
0
In Fig.
circuit con~ms _no external voltage or current source, and the voltage appearing across te
due to the c1rcult components.
 7. 1 Thermal Noise 451

Req

Noise sources for resistances in parallel. Figure 7.6 Equivalent noise circuit for resistances in parallel.
risure 7,5

Z,(f)

Passive
R,L,C
circuit
1-------0~ t NWso solrage

(a)

Z,(f) = Rr(f) + jXT(f) Noise voltage

Noise voltage

(b) (c)

rigure 7,7 (a) Two-terminal passive R, L, C circuit; (b) Thevenin equivalent of (a); (c} Equivalent noise circuit for (a),(b).

Example 7.1
Dete .
fl!l!ne the output noise PSD, Snif), and the output noise power in the following circuit.

R C

Soluli-0
II
Order to .
compute Sn(f) , we first obtain Thevenin impedance, Zr, looking into terminals. This is given by:

(R Jzhc)
Zr=
(R + f2;1c)
~fter ·
Silttp\"1
ficatio n,
R . 2rr/C
Zr = 1 + 4rr2J2C2R2 - 11 + 4rr2f2C2R2
ill
'I
452 NOISE IN COMMUNICATION SYSTEMS

Thus,
R
R(J) = 1 + 41r2.(2C2-R2

2kTR
and
Snif)=2kT R(J)= l + 41r2f2C2-R2
The noise power at the output is given by

l
oo 2kTR
N = _ 00
Sn(J)df = l + 4 rr 2{2C2-R2 df
or,

N==c
KT

It is observed that the noise power does not depend on resistance R, even though it generates noise. This i
the same result as shown in Example 6.39.

Example 7.2

. In the following circuit, obtain the noise PSD acros·s terminals a-b when the resistances R1, R2, and Rim
at temperatures T 1, T 2, and T 3 respectively. Verify your result when T 1 = T 2 = T 3 = T.
R3
r------,,.__ __.1LII 1 1 - - - - 0 a

_ _ ___.__ _ _ _ _ _-Ob
Solution
The noise PSD due to individual resistances are

Sn1 (J)= 2kT1R1

-
Snz(J) = 2kT2R2

Sn3(J) = 2kT3R3
. PSD due to R 1, R2 , and R3 bes
Let the output noise (J) S (J) dS (J) Then using superPo
no I , no2 , an no3 • •

Sno(J) = Snol if)+ Snoz(J) + Sn J(J) 0

where

Sno2(J) =(
R1 +R2
R1 )2 2kT2R2
S,ro3(J) = 2kT3R3
 7.1 Thermal Noise 453

s,,(f) = 2k [r,R, (R, R,) 2+ T2R2 (R, ~R,)' + T,R,]

- T2 ::::: T3 == T
~lbe0 TI - R1R2 ]
Snoif) = 2kT [ - - - + R3 = 2kT R
R1+R _ eq
2

'" ;/;rarion: . . •
~nr . resistance, looking mto a-b 1s
nieve(llll
Rth = R. 3 + R1R2
R1 +R3

and Sno (f) = 2kT Rth = 2kT [R3 .+ R1RiR

+R3
2
]

1bis is the same expression as obtained earlier.

7.1.1 Noise Temperature and Noise Figure

We now relook at the basic noise generation by a resistor [Eq. (7:1), Fig. 7:1]. •We note that~ is the mean
!iJU3fe noise power produced at the open-circuited terminals across the 'resistarice. We now compute the
~ai/able noise power generated by the resistor. The maximum power transfer theorem in the basic circuit
anal~sis states that maximum power from a source ·with iritemar resistance R will be transferred when it is
lerminated in a load resistance RL (Fig. 7 .8). This is true in case of noise source too. ,
The maximum noise power available from the resistance and dissipated in the load RL = R is

_ v~ _ 4 kTRB = kTB (7.5)

Or . Na - 4R - 4R
eq11Ivalent1y, available noise power spectral density (both sided) from the resistor is

kT
Snaif) == (7.6)
2
It is observed that both Na and Snaif) are indepenqent of the resistance
ln!Pedan . value. In case the source is an
aJldF.qs.0.;:::: R+jX, then the required load impedance for r~imum P?wer,transfe~ is ZL = Z_* = R-!X,
ll1wer Which) and _(7.6) still .hold for Na and Sna(f). Further, 1t 1s emphasized that Na 1s the maximum nmse
under co . can be extracted from the source. Equation (7.5) is now used to characterize the noise source
ns1derat·
ion. It should be noted that
Na
U
nique! T =
kB (7. 7)
iPrlld Ycharacte
. Ueect th · a noise source. In other words it
nzes · c_an be sai"d th at a noise
• source at temperature T
w~0te it by eTava1·1_able noise power Na. We shall call this temperature
<en · · tempera~ure of the source and
a n~1se
ChctePe n- It is remarked that T, obtained at the output of a two-termmal network 1s a fictitious quantity
nd s on the output noise power
n
across the termm · al Th s
s. u
Na (7.8)
Tn == kB
454 NOISE IN COMMUNICATION SYSTEMS

Input Noise-free system

Sai(f) Available gain= Ga

Figure 7.8 Load matching for available noise power. Figure 7.9 Noise-free two-port.

and
kTn
Sna(f) = 2 (7.9)

Noise in Two Ports

We now consider two-port systems and characterize their intrinsic noise. For this purpose we define th:
available gain of a noise-free two-port (Fig. 7.9) as

G (f) = Sa;(f) (7.10)

a Sao(f)

where S 0 ;(f) and S00 (f) are available input and output PSD respectively. It must be noted that Ga(f) is defined
for the noise-free two-port and it-is power-sense gain (or loss).
In ·g eneral two-port systems (e.g., amplifiers) will generate their own internal noise as they operatei
some elevated (non-absolute zero) temperature. Such two-ports shown in Fig. 7 .10 are called noisy two-Po~
Consider a noisy two-port with available gain G0 (f) . It is excited by a noise source with available nmse~
Sna;(f). The output noise PSD, Snao(f), then will be sum of the noise PSD due to two-port alone, Snaso '
and the noise PSD produced by Sna;(f).
(7.lll
Snao(f) = Sna;(f) Ga(f) + Snasoif)

In Eq. (7.11), Snaso(f) is the noise PSD produced by the two-port appearing at the output tenni;:~1
first term in the output is due the passage of input noise through the two-port. Further, it is note ·se pSD
!
. PSD 1s
output n01se . always greater th an the n01se
. PSD produced at the output solely by th e 1·nput no1
(Sna;(f)), i.e. Snao(f) > Sna;(f) Ga(f), as Snaso is greater than zero for noisy two-ports.

-
Inp ut Noisy two-port
-
Output
s=-
;if) G.(f) snao(f)
- - -

(a)

sai s QQ

Noisy two-port
N ai Ga
N QQ

{b)

. . teri:z:otion•
Figure 7.10 Noisy two-port: la)_noise characterization and lb) signal and noise charac
 7. 1 Thermal Noise 455

, jse Fi~re measure of noisiness of two-ports when it is defined for a particular frequency. It is known
~.0 efigure is a Tue spot noise figure is defined as
~01S 1,,. figure-
q,ot JIO,.,.,
P Total noise PSD at the output
F(f) = Output nmse
. PSD solely due to input noise PSD (7.12)

Using F,q. (? .11),

F(f) = Snao<J)
Sna;(f)Ga(f)
_ Sna;(f)Ga(f) + Snaso(f)
- Sna;(f)Ga(f)
=l+ Snaso(f) (7.13)
Sna;(f)Ga<J)

When available noise power at input and output are considered, then the average noise figure is defined as

Total available noise power at output

F--------,-----=-------=-----
- Output noise power solely due to input noise power
(7.14a)

Nao (7.14b)
-: NaiGa

And NQJJ is given by

· Nao = Naso + GaNai (7.15)
Therefore

p = Naso + GaNai
GaNai
= 1+ Naso (7.16)
GaNai

~~-th(?.! 4}-(7.16), Nai is the available noise at the input, Nao is the total available noise at the output and

1
e avaiJabJ . ,
Both e noise due to two-port appearing at the output.
Doise.free <7:13) and (7.16) show that noise figure 1. The equality sign holds when two-ports are
lllssible bt c~ is not possible in any practical system. The noise figure can be brought to as near to I as
IIowev~ng the two-port as low noise as possi~le. .
1
Cose to I. B~t ::th~r Variable in Eq. (7.16) is Na; which can_ be in~re~se~ to a very large value to bring F
lbere i kind of solution is not proper and also Nai 1s not mtrms1c to the two-port.
Considered : ::~rnative way to define noise figure of a two-port. Here, signal-to-noise ratio (SNR) is
1
let the av . input and output ports
ai1able signal
· . . and output ports be Sa1· an d Sao respectively
power at mput · · 7 .1Ob). Then
(Fig.

Sao= GaSai (7.17)

llie avera
ge Doise fi .
gure " defined ., F
- (Mi (7.18)
- (No
S)
 456 NOISE IN COMMUNICATION SYSTEMS

where (SIN); and (S/N) 0 are SNR at input and output ports respectively. These are given by

(i} = Sa;
Na;
0 I9a)

(it = - =
Saa Ga Sa;
Nao Naso+ Ga Na; (7,19b)

Therefore, using Eq. (7.18),

F=
Sa; Nao Sa; Naso + GaNa;
Na; Sao Na; · Ga Sa;
or
F = I+ Naso
(7,19c)
GaNai
which is the same expression as given by Eq. (7.16). Thus Eqs. (7.18) and (7.14a) are equivalent definitionsof
noise-figure of a two-port. In Eq. (7 .18), the numerator will always be greater than the denominator because
SNR at the output port will include noise due to two-port in addition to the noise appearing due to the input
noise alone. Therefore the ratio of input and output SNR must be greater than l.
It is remarked that noise figure is also known as noise factor, but the former is more commonly used. In
system noise calculations F is often used in dB, i.e.,

Fin dB = 10 log 10 F

Equivalent Noise Bandwidth

Two-ports are usually characterized as filters (Fig. 7.11), say with transfer function H(f). .
When the input to the two-port is available noise PSD, Saniif), the available output noise PSD ts

(7.20)
Sunoif) = Snaiif) IH(f)l 2

and Nuo = 1_: Sna;(f) IH(f)l 2 df

(7.21)

functio»sort
Often in communication systems these filters are modeled as lowpass or bandpass (other_filt~r ?.12,
not precluded). The magnitude of typical filter functions H(f) may be depicted as shown in fig.
When Snaiif) is due to a resistance at temperature T, i.e. S,.a;if) is !sf , the S 11 a0 if) is given as
(7.221
Snaoif) = -kT
2
IH(f)l 2

1
and (7,2Jl
00

Nao= -kT IH(f)l 2 df bOsen

2 - oo ' 'is often' now
Thus, output noise PSD is weighted by IH(f)l 2 and shaped accordingly. It is remarked th at_~~rn 100, we
as the center frequency of the passband and the value of IH(f)I can be other than maxi

s.~.___H_(f_)

Figure 7.11 Two-port as a filter.

 7 . 1 Thermal Noise 457

IH(f)I
IH(0)I -IH(-fo)I H(f) IH(fci)I

figure 7.12 Typical filter functions (a) lowpass, (b) bandpass.

filter H(f) with another filter with constant magnitude IH(f)I = IH(fo)I for bandpass filters and
reJJiace~~H(O)I for lowpass filters with the bandwidth BN (see dashed spectral functions) such that output
1H(f)I - N. is same for both actual two-port (H(f)) and the hypothetical filter (constant magnitude).
0oise Power ao
Therefore, for lowpass filters ,

kT f'X)
Nao = 2 }_ IH(f)f df
00

= -kT2 iBN IH(O)l 2

df (7.24a)
2 o
= kT IH(O)l 2 BN
And for bandpass filters

Nao = -kT2
2
f~=¥ IH(fo)I 2 df (7.24b)

= kT IH(fo)l 2 "/JN
lbebandwidth BN is called eq~ivalent noise bandwidth. Using the above equations, it is given by

J IH(J)l 2 df for lowpass systems

BN = -oo
00

(7.25a)
2 IH(O)l 2

J
BN = -oo
00 2
IH(f)l df
for bandpass systems (7.25b)
2
lbeva1u 2 IH(fo)I
0indpas esofBct .
= 0 for lowpass systems and J = Jo for
8 syste::s. eterrruned using Eqs. (7 .25) are with respect to f

I ~
I Q"'Ple 7.3
(a) Ob · ·
lain eq .
uivalent noise
· bandwidth for the Gaussian filter spect·fi e d b y

H(J) = e-a212
(b) Oblai
'" n the ratic 'ii;;
B
where B, is half-power bandwidth.
458 NOISE IN COMMUNICATION SYSTEMS

Solution
(a) Using Eq. (7.25a)
1 f~00
IH(f)l df
2

BN =2 JH(0)l 2
Here,
IH(0)I = 1
Thus,

,_2
Using the integral for area under the Gaussian curve viz. f~oo e -2;;2" dx =

(b) AtBh

or
e
-2a2B2
h = -21
which leads to
Bh = _I_, ../1n2 = 0.588
a,./2 a
Therefore, -
BN
Bh
= 0.626
0.588
= 1.065.

Effective Noise Temperature the noise

Recall that a source generating the available noise power Na is uniquely characterized by .
temperature [Eqs. (7.7-7.9)). This concept is also useful in characterizing noise in two-ports- of noise
Consider a noisy two-port (Fig. 7.13) with available gain, Ga, excited by a noise source
temperature Ts. The input noise N; and the output noise N 0 are given by
(7.26)
N; = kTsB (7,27)
J
·s oisY
1 11
· ban d wi"dth of th e system (two-port) is assumed to be B. The two-port itself
Here the operatrng
produces additional noise·NX() at the output.
Thus, the total output noise is ( ,zs)
7
N, = No + Nxo = GakTsB + Nxo
 7. 1 Thermal Noi.se 459

Noisy two-port
1----+- No<Tso) -
N,;== kT,;B Noise-free two-port
-
Gain: Ga .__.,._ Nw<Txo) Tei ==Ts+ Txi N 1 == Nxo
Gain: Ga
a
N;
(f,) -
Noise temperature in a two-port.
_7,13 Figure 7.14 Effective noise temperature for two-port.
Fi9"'y

noise can be referred at the input port as if the two-port were noise-free. This el+ective noise
t,o
ve total :JJ'
111' 8 written as
:,nut can be Nr N
·~ N er· -
- - - k-T Bs + - XO (7.29)
Ga Ga

!fl us denote Nxo

Nxi=- (7.30)
Ga

Toen
Nei = kTsB + Nxi (7.31)
We can equivalently write
Nxi = kTxiB (7.32)
F,juation (7.32) can be inter:preted as additional noise source at t~e input with noise temperature Txi. This
further implies that the system is now excited by a noise source with temperature (Ts+ Txi) and produces the
total output noise power, Nr, in the noise-free two-port (Fig. 7.14). This power is the same which is observed
for the noisy two-port. Thus, the effective noise temperature at the input port is given by

Tei= Ts+ Txi (7.33)

Fllllher,
Nr = No +Nxo
or

Nr = kTsBGa + kTxiBGa
= k (Ts + Tx;) BGa = kTeaBGa
lbisi,,, Ii
=P es that the effective noise temperature at the output port is
· given
· bY

(7.34)
'Ive shau
01input~:~s._(7.33) and (7.34) in system noise calculations. It is remarked that raising the temperature
'we re-ex rn~lies a~rnitting additional noise, which may be extraneous to the system also.
atnine noise figure [Eq. (?.14a)], which in the present context means

Nr Total available noise power at the output

lns F = N0 = Available noise power solely due to the input
.,,n .
Yste"'
Oise calculations, the usual source temperature is taken to be To= 290°K. With this condition

GakT0 B + GaTxi
F-
- GakToB
460 NOISE IN COMMUNICATION SYSTEMS

or
Txi
F=l+-
To
or the equivalent relation is
Txi = (F - 1) To = Te
It is emphasized that Tx; is the effective noise temperature of the system alone when referred at th . (7,Ji1
Therefore the two-port system can be characterized by (F, Ga, B) or (Te, Ga, B) where Te=== T. ~~PUtfwlrt
X! 1g, 7,]jj_

-- Noisy two-port
--
(F, Ga, B) or (T,, Ga, B)
- -
-

Figure 7.15 Noisy two-port characterization.

Example 7.4

An amplifier having available power gain of 46 dB and operating bandwidth 25 kHz is excited by a noise
source at temperature To(=290°K). It produces the output noise power of 32 pW.

(a) What will be the output noise power when the noise source temperature is raised to 2To?
(b) What is the noise figure of the amplifier?

Solution
(a) Total available noise power= Nao = 32 x 10- 12 W.
Let Te be the amplifier equivalent noise temperature (referred at input) and Ts be the temperanrreoftlie
noise source. Then,
k(Ts + Te)BGa = Nao
where

B = operating bandwidth = 25 x 103 Hz

Ga = Available gain of the amplifier = 46 dB ~ 4 x 104

Also given, Ts = To = 290°K.

Therefore

Nao
Ts + Te = To + Te =
kBGa
= _____________
32 X 10- 12 __,.
1.38 X 10-23 X 25 X 103 X 4 X }04
= 2318
Thus Te= 2318 - Ts= 2318 - To= 2318 - 290 = 2028°K
Now the raised value of Ts= 2To = 580°K. ·
~;:;P

l
s-.J'
7 . l Thermal Noise 461

. e power with the raised value of Ts is

(heno1s
~eoce,
N~ 0 = k(2To + Te)B Ga
= 1.38 X 10- 3(580 + 2028)
2
X 25 X 103 X 4 X 104
= 35.99 X 10-l2W

36pW

. figure of the amplifier is given by

) 11Je noise
'

F = l + Te = l 2028
To + 290

= 7.993 8 = 9.027dB

----------------~
-..
Noise Figure of an Attenuator
Here we are interested in obtaining noise figure of a purely resistive attenuator. Let us consider exciting the
anenuatorintroducing loss L by a noise source of noise temperature To (standard temperature) (Fig. 7 .16).
Let us first consider the combined noise source-attenuat'or which is kept at the standard temperature
1
To=290°K. The total output noise power Nao is given by

Nao= kToB (7 .37)

Here Nao includes noise generated by the attenuator. Next we consider the above system in which the attenua-
tor is treated as noise-free and its noise is represent~d by effective noise temper~ture, Te, at its input terminals.
Hence, in this representation, the total effective temperature at the input of attenuator is (To+ Te) (Fig. 7 .17).
Th The temperature To is due to the noise source and Te is due to the noise generated by the attenuator.
us,
Nao = k(To +Te)GaB (7 .38)
W
hereG-1 · " tehoss
a - I 1s the gain of the attenuator (Ga < 1), and L (> 1) 1s
l .mtroducedby the attenuator.

• --------------------------- I
I l----r-C>---1 Attenuator
: Noise To (Resistive)
1
source 1-_.1-c>---1 Gain= Ga= 1/L

-----------------------
I '------~ I

Figure 7 .16 Attenuator excited by a noise source.

-
..., Noise-free
attenuator
-
Na 0
-

t ...,- Gain= Ga= 1/L -

(T, +T0)

Figure 7.17 Equivalent noise-free attenuator with noise referred at input.

~ I

462 NOISE IN COMMUNICATION SYSTEMS

I
11 Equating the right-hand side in Eqs. (7.37) and (7.38)

kToB = k (To + Te) GaB

or
TeGa = To (1 - Ga)

or
Te= To (~a - 1) = To(L-1)

Using Eq. (7.36),

Te T0 (L- 1)
F=1+-=1+ Ti"
To . o
or
F= L
(l.~1
Therefore, noise figure of an atte~uator is_its loss value_ (in ratio), _and its effectiv~ ~oise temJ)erature i
To(L -1 ). These results are useful m modeling feeders which are passive lossy transmission lines collllectini
an antenna to the front-end of the receiver. ·

Noise Characterization of Cascaded Systems

In general, communication systems consist of many stages. For example, a heterodyne receiver has afrau.
end r-f amplifier, followed by a mixer, an IF amplifier, and a baseband demodulator. Each stage contri'ooll!
to the total system noise. Our intent is to characterize the composite cascaded system and determine ovetlil
noise figure/effective noise temperature in terms of noise figure/noise temperature of individual stages. W1
first consider a two-stage system (Fig. 7.18) where the first stage is characterized by (Ga1, Fi, Tt1 ) and th:
second stage by (Ga2, F2, Te2). The composite two-stage cascade is characterized by (Ga, F, Te)-
Let the cascade be excited by a noise source with available power Na;, and the noise generated soltfy~
the first stage be Nxol which appears at the output of the first stage along with the amplifiernoiseduetotll!
input noise source. Thus, the total noise appearing at the'output of the first stage is
(7.41)
Naol = Ga1Nai + Nxol
This noise is amplified by the second stage and appears at the output in addition to the noise solely generatio
by the second stage (Nx02) . The total output becomes

Nao = Ga2Naol + Nx02

or (7.42)
Nao = Gal Ga2Nai + G a2Nxol + N xo2
:---------------,---------------,
I I I
I I I
I
I
CG.1, F1, T, 1) I - (G02 , F2, T,i) Nao
I T I
I
I Nwl I Nxo2 I
!_~•:_:.:.:!___
------------------------
N aal

Figure 7.18 Two-stage noisy cascade.

 7.1 Thermal Noise 463

Nao!

Figure 7.19 Noisy first stage.

fthe composite cascade is given by [Eq. (7.14a)]

. figure o
nieno1se
Total output noise
F=~--~-~-------
Output noise solely due to input noise

or Nao
F=----
Ga1Ga2Nai

· Gal Ga2Nai + Ga2Nxo1 + Nxo2

= Ga1Ga2Nai

=l+ Nxol + Nxo2 (7.43)

Ga1Nai Gal Ga2Nai

Now let us consideronly stage 1 (Fig. 7 .19). Its noise figure is given by
I l J I \,

or

or

(7.44a)
!' .
llllilarly for stage 2

Nxo2 =
F2 - l (7.44b)
Usin .
I Ga2Nai
gEqs. (7.44a, b) in Eq. (7.43), the expression for F becomes

I or
F = 1 + (F1 - 1) + (F2 - l)

I lh
eforrn .
Pi r al1srn le di
F=F1 + -G-
· (F2 - l)
al
.
(7.45)

' ,1; k::::: 1 a ng to Eq. (7 .45) can be extended to an n-stage cascade (Fig. 7 .20) with parameters ( Ga1c,
1n . '2 ... n).
e noise fi
gure for the composite system is then given by

F = F1 + -
(F2 - 1)
- - + (F3 - _;_c.__
l) +.•.+ <Fn - 1)
Gal Ga1Ga2 Ga1Ga2 · · · Ga,n-1
(7 .46)
464 NOISE IN COMMUNICATION SYSTEMS

1-------------------
I
---, I
I I
1 (Gal• F I • Tel ) (G0 2, F2, Te2) (Gan, Fn, T-)
I .,,
I
I L----- I

L~~~!~-------------------------------:
Figure 7.20 n-stage noisy cascade.

The overall effective noise temperature, Te, is obtained by using the relationship between noise fi
noise temperature [Eq. (7.36)] which is re-written below

Te= (F- 1) To

or
Te
F=l+-
. To
Using the above for both the composite casca9e and individual stages, Eq. (7.46) becomes

or
T2 T3 T.
T=T
e el +-e-+
G e + ... + en (7.4n
al Gal Ga2 Gal Ga2 :.. Ga,n-l
Equations (7.46) and (7.47) are known as Friis formula for noise figure and noise temperature for cascaded
systems, respectively. 3 .

Example 7.5

A 3-stage receiving system operates at 30 MHz bandwidth. The gain and noise figure of these stages are
given as:

G1 = 40 dB G2 = 10 dB G3 = 30 dB
F1 = 2dB F2 = 9dB F3 = 6dB

(a) Calculate the noise figure of the system.

(b) What is the output noise power generated by the system?

Solution
(a) The gains and noise figures in ratio forms are

F1 = 1.584 F2 = 7.493 p3 =4
 7. 1 Thermal Noise 465
, ...nula for noise figure for cascaded systems
. (he 10, ...
vs1ng
F -- F l +-
F2 - 1 . F3 - 1
-+--
G1 G1G2
= 1.584 + 6.493 3
104 + 1()5
= 1.584679
'Jbeeffective noise te~perarui:e of the system= T, = (F - l)To = 0.584679 x 290 = 169.55oK.
M the equivalent input noise power of the noise-free system is
ttence .
Ni= kT,B = 1.38 x 10- 23 x 169.55 x 30 x 106
= 7.019 X 10-l 4 W

Hence, the output noise power

No = Ni x System gain,
(System gain = 80 dB = 108)
Thus, No = 7.019 x 10- 14 x 108 =;: 7.019 x 10-6w
~7µW
/'

Example 7.6
Consider the following system

Amplifier
Amplifier l---0--1
mixer
A' B
F2 =7dB
G2=20dB
Operating bandwidth= 30 MHz

(a) Obtai 0 the output noise power produced·by the system A-B-C. ·
(b) ~e ~bove system is connected to a feeder on the front with gain == 0.9 (loss = 1.111) shown as A' -A.
lain the output noise power for the system A'-A-B-C.
(c) Calculate the percentage change between output noise power in (a) and (b) and interpret the results.

Solution
(a) F
or the system A-B-C

Fi= 100.15 = 1.412 G1 = 40dB = 104

Fi= 10°-7 =-S.012 G2 = 20dB = 102
Hence F _ I 5.012 - 1
Fa = Fi + - 2- - = 1.412 + l04 = 1.4124
G1 ,
106
Ga = System gain= 60 dB =
p:
466 NOISE IN COMMUNICATION SYSTEMS

Hence, the effective noise temperature is

Tea = (Fa - l)To = (1.4124 - 1) ?< 290 = 119.59°K

The output noise power is

Noa= kTeaBG

= 1.38 X 10- 23 X 119.59 X 30 X 106 X 106 = 4.951 X 1Q-8 W

(b) For the systemA'-A-B -C

Noise figure of the feeder
1
Ft= Feeder loss= - = 1.111
0.9
Feeder gain = Gt = 0.9
System gain= Gb = 0.9 x 106
Hence, noise figure for the system is

Substituting values of Ft, F1, F2, Gt, and G1

1.412 - 1 5.012 - 1
Fb = 1.111 + -0.9
- - + 0.9 X 1
()4 = 1.5681

Hence the effective noise temperature is

Teb = (Fb - 1) x To= (1.5681 - 1) x 290 = 164.76°K

The output noise power is

Nob = k Teb B Gb
= l.38 X 10- 23 X 164.76 X 30 X 106 X 0.9 X 106
= 6.]389 X w-S W
Hence Nob > Noa •
(c) Percentage increase in output noise power

Nb
0 - N 0
= N ° X 100
oa
= 23.99% 24%

The effect of feeder is to increase the noise substantially.

Ex.a mple 7.7

un1or
fthOutteil
Jn the following 11y11tem, attenuator introduces u loss of 2 dB. What should be the temperuture 0
no that the overall noiHe figure of' the system doeH not exceed 2.5 dB?
A1ten1111wr Ampllfler Mixer
IOHH • 2 dB 0 1• 1000 02 • 100
fi" • 1.3 di) P • 9dB
 7 .2 Shot Noise 467
•.1,JiDtl ork in terms of effective noise temperat
Ure.
)(I"" •• AtO w
we!P"' jred system noise gure
. fl
= F
m 2.5 dB
~re4u . . Fm l.778

~re
~- ,..,.,1 system effecttve n01se temperature < T

Tem = (Fm -
-
l)To:::: 0.778 2
w•· . x 90 :::: 225.62oK
uired attenuator temperature be TA, then th
"tthere<! e overall system n .
"" o1se temperature

Te= (FA - l)TA + Tel X LA+~

G1
FA= Attenuator noise figure
.,i,,,fP. =
Attenuator 1 _
1
SinCC LA = 2 dB = .
Wll"'' 584 an F 1 = 1.3 dB = 1.348oss - LA :::: 2 dB -- 1.584
d
Hence, Tei =
(l.384 - 1) x 2900 :::: l01.2oK
G~en F2 = 9 dB= 7.943

Hence, Te2 =(7.943 - 1) x 290:::: 2013.55oK

Using values of FA, Te 1, Te 2, LA, and G1 in the exp · c
ress10n 1or Te, we get

Te= 0.584 TA+ 163.488

Since condition is Te -< Tem
or
0.584 TA + 163.488 225.62
or
T 225.62 - 163.488
A~ 0.584 .
H 106.39° K
ence the ~eq . ed
' urr attenuator temperature is less than 106° K.

72
· SHOT NOISE
In electroni
1 ste
0rinulatect ch sy ms, shot noise arises due to discrete nature of charge carriers. It was first observed and
ilev·1 YSchottky · · th · · al 3 4
ces, elec m 1918 in context of current flow m erm1omc v ves. ' In modem semiconductor
~ulllbers (in:ons and/or holes flow past a potential barrier. Firstly, it should be noted that these are discrete
linieinterva1. gers),
th
e.g. a system,cannot have a half or a quarter of hole/electron. ·Secondly, in any given
~lllly 5 b' e charge Carriers arrive at random time instants and also the number of charge carriers vary
•us . ' Ut th · . .
&laij variation . eir number tends to converge towards a mean value. Therefore, m successive small intervals
ar Sticauy • ~s around the mean value. Thirdly, the number of carriers flowing in successive intervals is
~nd the: ependent. The above fluctuations around the mean val\,le result in variation of resultant current
00 to Counuean current, These fluctuations are termed as shot noise. The word "shot" in shot noise appears
8
de a lirface ng nature of events of occurrences. As an analogy, assume that rain drops (shots) are falling
60~ndent on~ aggregation of these drops results in the flow. The amount of flow in any time interval is
hecol11es e number of rain drops (shots) falling on the surface. Now replace shot with carrier and the
the current flow.
468 NOISE IN COMMUNICATION SYSTEMS

Figure 7.21 shows an example of nu'!1ber_of charge carriers (elec~on/hole) n:1oving Past a barn
long tenn average is n = 100 carriers/tJme mterval (t:..t). However, m successive non-overla . er, ]1'
intervals (shown for five such intervals) value of n is 102, 107, 95, 101 , and 98 resulting in fluctuPP~ng ti~
' th · ations
+7, -5, +1, and -2 around the mean value 100. ~owever, e runnmg mean value _for five interv +1,
n(5t:..t) = 100.6 and in the long run Jim n(kt:..t) n. Also, the number of charge earners arrivi . zjs1.c
. . . . . . ngma1¾
interval t:.t is statistically independent of that m other time mterval, 1.e. n; and nj are stat1stically ind
of each other for all values of i and} pertaining to time intervals t:..t; and t:..tj , Therefore, the resultan
due to motion of charged carriers fluctuates around a mean value, thus producing shot noise in th tcu11en1
This phenomenon is quite significant in the optical detection process where incident light intensity (e cuhrreni.
. . hih ' PotonsJ
generates photo-electrons and the phenomenon 1s called quantum n01se w c 1s essentially the shot .
The passage of each charge carrier (hereafter electron) through the electronic system produces no15e_
. d
pulse of the observed current; _each of these pulses occumng at a ran om time.
. In the long run a CUiTenta~
will measure the average current but fluctuations will be observed in short intervals. If p(t) is the CUiTent:
of duration T, then
qn = Cd 1T p(t)dt (7.4!)

where Cd is the detector constant and q is the electron charge (q = 1.6 x 10- 19 Coulombs). Supposearmli
of electron gen~rate~ a saw-tooth curren,t puls~, then a typical snapshot of the current generation mayre
depicted as shown rn Fig. 7.22. The mean current will be the long-tenn average of the pulse train.

Shot Noise Statistics

Shot noise involves counting of electrons arriving at random time instants. This is a discrete random variabk
In case of shot noise, it is modeled as the Poisson distribution which is written as
Prob. [Arrival of n electrons in time interval t:..t]

(7.49)

n I
1
n 1 = (102)
:
n - (107)
2- I n3 =(95) I n4 =(IOI) I n5 =(98) I

i - - - - - - - -7--
1
100 ------i--------
I
I I I
- - - - - - , - - -- ---,----
1 I I I I
I I I I I
I I I I I
I I I
0 Ill 2/l/ 3/l/ 4/l/ 5/l/

Figure 7.21 Depiction of lluct t· f

ua ion o number of carriers with time.

Figure 7,22 D
epiction of current pulses.
 7.2 Shot Noise 469

__ rnean number of_electron arrivals per unit time. Recall that mean and variance of Poisson
J.:::: 11 - Chapter 6) are given by
pere ,..;able (
,aoJ11 Vil' '
11
E [n] = ii = AD..t
Var [rt] = a; = AD.. 0

. taneous and average currents generated are

the instan
111Crefore,
i(t) = qn (7 .50)
/),.t

I = E[i(t)] = qn (7.51)
D..t

!) detector constant is assumed to be unity. The average power of the shot noise is given by the
5
,,Eq(?. ' 2 · · · b
"' · f '(t) <J'· which 1s given y
variance o I ' ' '. -

Var[i(t)] = E[i(t) - 1]
2

=( :J 2
var(n) [using Eq. (7.50)]

or

a'?-
I
= (_!!_)2
D,_f
a2
n

= (1J2 ii [ii= a; for _Poisson random variable]

- q
( -D..t
)2 -I D..t- [using Eq. (7.51)]
q

or
2 qi (7.52)
(J',
I
=D,_f
-
Assurnin g KiI = Bsh as the operational
. .
bandwidth

qi (7.53)
2
a-
I
= -Bsh
Q
= qlBsh
letus denote shot n01·se power as Nsh and associated
· · PSD as S.sh (f) , t h en
n01se
Bsh/2

lneabov
N sh = qlBsh = 1-Bsh/2
Sshif)df
(7.54)

e equation implies that the shot noise PSD (both sided) is given by

qi
t Ss1iif) =2 (7.55)

,,11/o 0bserv .
· llJakingations
" 1Us
. . are m · order. First the
shot noise power 1s · 1 to the average detected current,
· proport1ona
0
apPerar10n. l'hu stgn
It . . a l d ependent. Second,
' · P SD 1s
the shot n01se . constant over th e w h o le frequency range of
othars as s, It can be said that the shot noise is white. It is remarked that shot noise is multiplicative. It

te
· 80on as th e signal
freq1 e 81gna1. · is detected and, unlike the therma1 n01se,
· · cannot be present in the absence
1t
Uency is on1 shot noise is spectrally white until the averaging argument breaks down. This roll-off
he order of GHz. For relatively large average current, the white noise is modeled as Gaussian
 470 NOISE IN COMMUNICATION SYSTEMS

di s tn·bu te d w1·th mean an d vanance

. equ al to )../),.t · This approximation is fairly accurate and is
. the a .
of the central limit theorem. . . PPhcati~
A s h. as been sru•d eari.·1er, th e sign
• al -dependent shot nmse . ,, appears
t . m modem semicond Uctor
tectors. But, in such detection systems, an ~dd1tion · · al " no·1se curren 1s generated
. d ue to ranct Pho~
of charge earners · even m· the absence of th e mc1 · ·den t 11·ght· These charge earners are produced0dtn en..: ·•~ss1.on
mal generation and the current generated due to them is called dark current. The presence of d;~ to !her.
modifies Eq. (7.54) as curten
1
Nsh = q(I +ld)Bsh
(7.56)
where Id is the average dark current.
An alternative elaborate frequency domain treatment of shot noise is described by Rice.6 lb
shot noise on the. system can be studied by raising the effective . temperature of ·the system ate effect
. noise its . 0f
by Tesh which is caused by the shot noise. In other words T; = Terh + Tesh, where Terh is the effective input
temperature due to the thermal noise of the system. noise

Example 7.8

The depiction of shot noise as a series of pulses (Fig. 7.22) can be visualized as excitation of a filter byan
impulse train where impulses occur at random time instants. Determine the PSD of n(t) in the figure ~ven
below. ·

Solution
In the input process, time instance 1j is randomly distributed and is closely related with the number of occ_ur·
rences of the relevant events, say electron generation, in a time interval. It is modeled as Poisson distribunon
which is a counting process, z(t). It is illustrated below:
Count
z(t) =N(t)

f1
lz

lt
tj
x(t) = dz(t) t ...
dt

11
t t
t2
t t
Hence, t ...

P [Number of counts in the time interval(t _ ti) == m]

2

== P[(N(t2) - N(ti)) = m]
(A(t2 - t1)re->..Ctz-t1) 1,2,····
m === 0'
m!
 7.2 Shot Noise 471

umber of counts per unit time. Further, in the Poisson process modeling, the number of
;:; averag~:Vertapping intervals are statistically independent. It implies that
wbere . two non
ots ,n
cOll E [(N(tz) - N(t1 )) (N(t4) - N(t3))] = E [N(t2) - N(ti)] E [N(t4) - N(t3)]

_ t ) and (t4 - t3) are non-overlapping.

rvals 1
wbereinte (tz ·
that the process x(t) is o b ~aine
. d bY d"" . . the process z(t). We first obtain the autocor-
tuerenttatmg
weobsefV~ f z(t) and then determme the same for the process x(t) which in tum will be used to obtain
. nfunction o
ieJauo x(t) and n( t).
.
Assuming t2 > ti

Rz (t1,t2) = E [N(t1) N(t2)]

= E[N(t1) N(t1 + t2 - t1)]

= E [N(t1) (N(t1) + N(t2 - t1))]

= E [(N(t1)) 2] E [N(t1) N(t2 - t1)] ·

Since intervals (t 1 - 0) and (t2 - t1) are non-overlapping, Rz(t1, t2) becomes

Rz(t1,t2) = E[(N(t1)) 2]E[N(t1)]E[N(t2 - ti)]

Now E[(N(t 1))2] is the second moment of Poisson random variable. It is given by

Also,
E [N(t1)] = M1
and
E[N(t2 - t1)] = )..(t2 - t1)
Using the bo
a ve expression,

Or

SilllilarJy Wh
en t1 > t2,

Colllb'1ning th R2 (t1, t2) = M2 +)..2 t1t2

· e above two equations,