Chapter 7

Noise E¤ects in Simple Modulation

Systems

7.1 Problems
Problem 7.1
Degrees kelvin (degrees absolute) are converted to to the Celsius scale by substracting 273
(actually 273.15). Degrees Celsius are converted to degrees Fahrenheit by the familiar
expression
9
degrees F = degrees C + 32
5
Thus
9
degrees F = (kelvin 273) + 32
5
Thus, 3 degrees on the kelvin scale is

9
(3 273) + 32 = 454 degrees F
5
and 290 degrees on the kelvin scale is

9
(290 273) + 32 = 62:6 degrees F
5

Problem 7.2
The signal power at the output of the lowpasss …lter is PT . The noise power is N0 BN ,
where BN is the noise-equivalent bandwidth of the …lter. From (6.116), we know that the

1
2 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

noise-equivalent bandwidth of an nth order Butterworth …lter with 3 dB bandwidth W is

W=2n
Bn (n) =
sin ( =2n)

Thus, the signal-to-noise ratio at the …lter output is

PT sin ( =2n) PT PT
SN R = = = f (n)
N0 BN =2n N0 W N0 W

so that f (n) is given by

sin ( =2n)
f (n) =
=2n
Thus we have the values given in the following table.

n f (n) SNR
1 0:6366 0:6366 NP0TW
3 0:9549 0:9549 NP0TW
5 0:9836 0:9836 NP0TW
10 0:9959 0:9959 NP0TW

Note that for large n, sin( =2n)=( =2n) 1, so that the SNR! PT =N0 W as n ! 1. The
plot is shown In Figure 7.1.

Problem 7.3
For DSB, the received signal and noise are given by

xr (t) = Ac m (t) cos (2 fc t + ) + n (t)

At the output of the predetection …lter, the signal and noise powers are given by
1
ST = A2c m2 NT = n2 = N0 BT
2
The predetection SNR is
A2c m2
(SNR)T =
2N0 BT
If the postdetection …lter passes all of the nc (t) component, yD (t) is

yD (t) = Ac m (t) + nc (t)

The output signal power is A2c m2 and the output noise PSD is shown in Figure 7.2.
7.1. PROBLEMS 3

0.9

0.8

0.7

0.6
f(n)

0.5

0.4

0.3

0.2

0.1

0
1 2 3 4 5 6 7 8 9 10
n

Figure 7.1: Normalized SNR at the …lter output.

Case I: BD > 12 BT
For this case, all of the noise, nc (t), is passed by the postdetection …lter, and the output
noise power is
Z 1 BT
2
ND = N0 df = 2N0 BT
1
B
2 D

This yields the detection gain

(SNR)D A2 m2 =N0 BT
= c =2
(SNR)T A2c m2 =2N0 BT

Case II: BD < 21 BT

For this case, the postdetection …lter limits the output noise and the output noise power
is Z BD
ND = N0 df = 2N0 BD
BD

This case gives the detection gain

(SNR)D A2 m2 =2N0 BD BT
= c =
(SNR)T A2c m2 =2N0 BT BD
4 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

()nS
c f
N
0

f
−
12TB 12TB
0

Figure 7.2: Plot of the spectrum of BT for Problem 7.3.

For an AM signal the predetection signal power is

1 h i
ST = A2c 1 + a2 m2n
2
and the postdetection signal power is

SD = A2c a2 m2n

The noise powers do not change.

Case I: BD > 12 BT

(SNR)D A2 a2 m2n =N0 BT 2a2 m2n

= h c i =
(SNR)T A2c 1 + a2 m2n =2N0 BT 1 + a2 m2n

Case II: BD < 12 BT

(SNR)D A2 a2 m2n =2N0 BD a2 m2n BT

= hc i = =
(SNR)T A2c 1 + a2 m2n =2N0 BT 2
1 + a mn 2 BD

Problem 7.4
We express n (t) as

1 1
n (t) = nc (t) cos 2 fc t (2 W ) t + + ns (t) sin 2 fc t (2 W ) t +
2 2
7.1. PROBLEMS 5

Sn(),c f S
ns f

N
0

f
−
12W 12W

Figure 7.3: Plots of Snc (f ) and Sns (f ) for Problem 7.4.

where we use the plus sign for the U SB and the minus sign for the LSB. The received
signal is
b (t) sin (2 fc t + )]
xr (t) = Ac [m (t) cos (2 fc t + ) m
Multiplying xr (t) + n (t) by 2 cos (2 fc t + ) and lowpass …ltering yields

yD (t) = Ac m (t) + nc (t) cos ( W t) ns (t) sin ( W t)

From this expression, we can show that the postdetection signal and noise powers are given
by

SD = A2c m2 ND = N0 W
ST = Ac m2 NT = N0 W

This gives a detection gain of one. The power spectral densities of nc (t) and ns (t) are
illustrated in Figure 7.3.Problem 7.5

Since the message signal is de…ned by

m(t) = A cos (2 f1 t + 1) + B cos (2 f2 t + 2)

the Hilbert transform is given by

b
m(t) = A sin (2 f1 t + 1) + B sin (2 f2 t + 2)

Squaring and averging m(t) gives

Efm2 (t)g = EfA2 cos2 (2 f1 t + 1 )g +

2EfA cos (2 f1 t + 1 ) B cos (2 f2 t + 2 )g + EfB 2 cos2 (2 f2 t + 2 )g
6 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

The …rst and third terms are A2 =2 and B 2 =2, respectively. If 1 and 2 are assumed
independent, the middle term is zero. Thus
A2 B 2
Efm2 (t)g = +
2 2
In exactly the same manner
A2 B 2
b 2 (t)g =
Efm +
2 2
b
and we see that the powers in m(t) and m(t) are equal.
We now compute

b
m(t)m(t) = [A cos (2 f1 t + 1) + B cos (2 f2 t + 2 )]
[A sin (2 f1 t + 1) + B sin (2 f2 t + 2 )]

which is

b
m(t)m(t) = A2 cos (2 f1 t + 1 ) sin (2 f1 t + 1) +
AB cos (2 f1 t + 1 ) sin (2 f2 t + 2) +
AB cos (2 f2 t + 2 ) sin (2 f1 t + 1) +
2
B cos (2 f2 t + 2 ) sin (2 f2 t + 2)

Applying the appropriate trig identities gives

A2 B2
b
m(t)m(t) = sin (4 f1 t + 2 1 ) + sin (4 f2 t + 2 2 ) +
2 2
AB [sin (2 (f2 f1 )t 1 + 2 ) + 2 sin (2 (f1 + f2 )t + 1 + 2 )] +
AB sin (2 (f1 f2 )t + 1 2)

All …ve terms in the preceding equation are sinusoids of di¤ering frequencies. Since the
average value of a sinusoid is zero

b
E fm(t)m(t)g =0

Problem 7.6
Expanding the noise about fc W=4 gives
W W
n (t) = nc (t) cos 2 fc t t+ + ns (t) sin 2 fc t t+
2 2
where we use the plus sign for the U SB and the minus sign for the LSB. The received
signal is
b (t) sin (2 fc t + )]
xr (t) = Ac [m (t) cos (2 fc t + ) m
7.1. PROBLEMS 7

Multiplying xr (t) + n (t) by 2 cos (2 fc t + ) and lowpass …ltering yields

W W
yD (t) = Ac m (t) + nc (t) cos t ns (t) sin t
2 2
From this expression, we see that the postdetection signal and noise powers unchanged from
those given in Problem 7.3 and are given by

SD = A2c m2 ND = N0 W
ST = Ac m2 NT = N0 W

This gives a detection gain of one. The power spectral densities of nc (t) and ns (t) are
easily plotted.

Problem 7.7
First we determine a general form by evaluating
Z fb
A
m(t) = A ej2 f t df = ej2 fb t
ej2 fa t
fa j2 t
There are three cases to consider.
Case 1 (fa = 0, fb = f2 ): This gives
Z f2
A
m(t) = A ej2 ft
df = ej2 f2 t
1
0 j2 t
which is
A
m(t) = ej2 f2 t=2
e j2 f2 t=2
ej2 f2 t=2
=
j2 t
or
A
sin( f2 t)ej f2 t
m(t) =
t
Since taking the Hilbert transform adds =2 to the phase for f > 0 and substracts =2 to
the phase for f < 0 we have
A
b
m(t) = sin( f2 t)ej[ f2 t+( =2)]
; f >0
t
and
A
b
m(t) = sin( f2 t)ej[ f2 t ( =2)] ; f <0
t
Case 2 (fa = f2 =2, fb = f2 ): This gives
Z f2
A
m(t) = A ej2 f t df = ej2 f2 t
e j f2 t
f2 =2 j2 t
8 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

which can be written

A
m(t) = ej3 f2 t
1 e j f2 t
j2 t
Thus
A A
m(t) = ej3 f2 t=2
e j3 f2 t=2
e j( 3 =2)f2 t
= sin(3 f2 t=2)ej f2 t=2
j2 t t
The Hilbert transform is
A
b
m(t) = sin(3 f2 t=2)ej[( f2 t=2)+( =2)]
; f >0
t
and
A
b
m(t) = sin(3 f2 t=2)ej[( f2 t=2) ( =2)]
; f <0
t
Case 3 (fa = f2 , fb = f2 ): This gives
Z f2
A A
m(t) = A ej2 f t df = ej2 f2 t
e j2 f2 t
= sin(2 f2 t)
f2 j2 t t

The Hilbert transform is

A
b
m(t) = sin(2 f2 t)ej =2
; f >0
t
and
A
b
m(t) = sin(2 f2 t)e j =2 ; f <0
t
b
Note that m(t) is real and m(t) is imaginary. The plots are easily constructed. All magni-
tude spectra have the familiar jsin(x)=xj form. The phase spectra are piecewise linear with
jumps for negative sin(x)=x.

Problem 7.8
The …rst step is to determine the value of M . Since

Pf X< M g + P fX > M g = 0:005, P fX > M g 0:0025

First we compute
Z 1
1 y 2 =2 2 M
P fX > M g = p e dy = Q = 0:0025
M 2
This gives
M
= 2:807
7.1. PROBLEMS 9

Thus,
M = 2:807
and
m (t)
mn (t) =
2:807
Thus, since m2 = 2

m2 2
m2n = = = 0:178
(2:807 )2 5:614 2

Since a = 0:7, we have

0:178 (0:7)2
Ef f = = 0:0802 = 8:02%
1 + 0:178 (0:7)2

and the detection gain is

(SNR)D
= 2E = 0:1604
(SNR)T
Problem 7.9

The message signal is

m (t) = 12 cos (8 t)
Thus,
mn (t) = cos (8 t)
so that
1
m2n =
2
The e¢ ciency is therefore given by

(0:5) (0:6)2
Ef f = = 0:1525 = 15:25%
1 + (0:5) (0:6)2

From (7.36), the detection gain is

(SNR)D
= 2Ef f = 0:305 = 5:157 dB
(SNR)T

Relative to baseband, the output SNR is

(SNR)D
= 5:157 dB
PT =N0 W
10 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

If the modulation index in increased to 0:9, the e¢ ciency becomes

(0:5) (0:9)2
Ef f = = 0:2883 = 28:83%
1 + (0:5) (0:9)2

This gives a detection gain of

(SNR)D
= 2Ef f = 0:5765 = 2:39 dB
(SNR)T

which, relative to baseband, is

(SNR)D
= 2:39 dB
PT =N0 W

This represents an improvement of 2:76 dB.

Problem 7.10
The output of the predetection …lter is

e (t) = Ac [1 + amn (t)] cos [! c t + ] + rn (t) cos [! c t + + n (t)]

The noise function rn (t) has a Rayleigh pdf. Thus

r r2 =2N
fRn (rn ) = e
N
where N is the predetection noise power. This gives
1 1
N = n2c + n2s = N0 BT
2 2
From the de…nition of threshold
Z Ac
r r 2 =2N
0:99 = e dr
0 N

which gives
A2c =2N
0:99 = 1 e
Thus,
A2c
= ln (0:01)
2N
which gives
A2c = 9:21 N
7.1. PROBLEMS 11

The predetection signal power is

1 h i 1 2
PT = Ac 1 + a2 m2n A [1 + 1] = A2c
2 2 c
which gives
PT A2
= c = 9:21 9:64 dB
N N

Problem 7.11
Equation (7.59) is, suppressing t,
2
n2c +n2s = Ef[n2c + n2s ]2 g E 2 fn2c + n2s g

The …rst term is

Ef[n2c + n2s ]2 g = Efn4c g + Efn4s g + 2Efn2c gEfn2s g

where the last term follows from the independence of nc and ns . Since (See Problem 5.22)

Efn4c g = Efn4s g = 3 4
n

and
Efn2c g = Efn2s g = 2
n

we have
Ef[n2c + n2s ]2 g = 3 4
n +3 4
n +2 2 2
n n =8 4
n

Also
2 2 2 2
E 2 fn2c + n2s g = Efn2c + n2s g = Efn2n + n2n g = 2 n =4 4
n

Thus
2 2
n2c +n2s = E 2 fn2c + n2s g Efn2c + n2s g =8 4
n 4 4
n =4 4
n

Problem 7.12
Let the predetection …lter bandwidth be BT and let the postdetection …lter bandwidth be
BD . The received signal (with noise) at the predetection …lter output is represented

xr (t) = Ac [1 + amn (t)] cos ! c t + nc (t) cos ! c t ns (t) sin ! c t

which can be represented

xr (t) = fAc [1 + amn (t)] + nc (t)g cos ! c t ns (t) sin ! c t

12 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

The output of the square-law device is

y (t) = fAc [1 + amn (t)] + nc (t)g2 cos2 ! c t

fAc [1 + amn (t)] + nc (t)g ns (t) cos ! c (t)
+n2s (t) sin2 cos ! c (t)

Assuming that the postdetection …lter removes the double frequency terms, yD (t) can be
written
1 1
yD (t) = fAc [1 + amn (t)] + nc (t)g2 + n2s (t)
2 2
Except for a factor of 2, this is (7.53). We now express the preceding expression as

1 1 1
yD (t) = A2c [1 + 2amn (t) + a2 m2n (t)] + Ac [1 + amn (t)] nc (t) + n2c (t) + n2s (t)
2 2 2
Let m(t) = cos(! m t) + cos(2! m t), which clearly has a maximum value of 2. Thus

1 1
mn (t) = cos(! m t) + cos(2! m t)
2 2
and
1 1 1 1 1
m2n (t) = + cos(2! m t) + + cos(4! m t) + cos(! m t) cos(2! m t)
8 8 8 8 2
or
1 1 1 1 1
m2n (t) = + cos(! m t) + cos(2! m t) + cos(3! m t) + cos(4! m t)
4 4 8 4 8
Putting this together gives

1 2 a2
yD (t) = Ac 1 + +
2 4
1 a 1 a
A2c a + cos(! m t) + + cos(2! m t) +
2 8 2 16
1 1
A2c a2 cos(3! m t) + cos(4! m t) +
8 16
Ac nc (t) +
1 1
Ac a cos(! m t) + cos(2! m t) nc (t) +
2 2
1 2 1 2
n (t) + ns (t)
2 c 2
The PSD therefore consists of 6 basic terms as follows:
Term 1: A dc term (f = 0) which is an impulse.
7.1. PROBLEMS 13

Term 2: Impulse functions at frequencies f = fm and f = 2fm . The weights are given
on line 2 of the preceding equation when squared and divided by two. Note that the terms
do not have the same weight due to intermodulation distortion. Thus the demodulated
output is distorted.
Term 3: Impulse functions at frequencies f = 3fm and f = 4fm resulting from
harmonic distortion.
Term 4: The PSD of nc (t) (‡at) about f = 0 and weighted by Ac .
Term 5: The PSD of nc (t) (‡at) translated to frequencies f = fm and f = 2fm and
weighted by Ac a=2.
Term 6: The PSD of n2c (t) (triangular) about f = 0 The PSD of n2s (t) is identical to
the PSD of n2c (t)

Problem 7.13
Since m (t) is a sinusoid, m2n = 12 , and the e¢ ciency is
1 2
2a a2
Ef f = =
1 + 12 a2 2 + a2

and the output SNR is

a2 PT
(SNR)D = Ef f =
2 + a2 N0 W
In dB we have
a2 PT
(SNR)D;dB = 10 log10 + 10 log10
2 + a2 N0 W
For a = 0:4,
PT
(SNR)D;dB = 10 log10 11:3033
N0 W
For a = 0:5,
PT
(SNR)D;dB = 10 log10 9:5424
N0 W
For a = 0:7,
PT
(SNR)D;dB = 10 log10 7:0600
N0 W
For a = 0:9,
PT
(SNR)D;dB = 10 log10 5:4022
N0 W
The plot of (SNR)dB as a function of PT =N0 W in dB is linear having a slope of one. The
modulation index only biases (SNR)D;dB down by an amount given by the last term in the
preceding four expressions.
14 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

Problem 7.14
Letting the message signal signal be a sinusoid in (7.63) so that m2n = 0:5 gives

a2 =2 PT =N0 W
(SNR)D = 2
[1 + a2 =2] 1 + N0 W=PT

Plotting these for a = 0:1, 0.2, 0.5, and 1 veri…es Figure 7.6.
The distortion to noise ratio is

DD A4 a4 m4n
= c
ND ND
where
* +
2
1 1 1 1 1 3
m4n = sin4 (! m t) = + cos(2! m t) = + cos(2! m t) + cos2 (2! m t) =
2 2 4 2 4 8

and, from (7.60)

a2
ND = 4A2c 2
n 1+ + 4
n = 2A2c 2 + a2 N0 W + (N0 W )2
2

and, …nally, from (7.32)

1 1 a2 1
PT = A2c (1 + a2 m2n ) = A2c (1 + ) = A2c (2 + a2 )
2 2 2 4
We can therefore write ND as
4PT
ND = 2A2c N0 W + (N0 W )2 = N0 W (8PT + N0 W )
A2c

and
3
DD = a4
8
Putting all of this togethen allows us to write DD =ND in terms of a, PT , N0 , and W . This
gives
DD A4 a4 m4n 3a4 4PT 1
= c =
ND ND 8 2 + a 8PT (N0 W ) + (N0 W )2
2

or
DD 3 a4 PT =N0 W
= 2
ND 2 2 + a 8PT + N0 W
We see that, like for DD =SD , the ratio DD =ND goes to zero as a ! 0.
7.1. PROBLEMS 15

Problem 7.15
Assume sinusoidal modulation since sinusoidal modulation was assumed in the development
of the square-law detector. For a = 1 and mn (t) = cos (2 fm t) so that m2n = 0:5, we have,
for linear envelope detection (since PT =N0 W 1, we use the coherent result),
1
PT a2 m2n PT PT 1 PT
(SNR)D;C = Ef f = = 2 1 =
N0 W 1 + a2 m2n N0 W 1+ 2
N0 W 3 N0 W

For square-law detection we have, from (6.61) with a = 1

2
a PT 2 PT
(SNR)D;SL = 2 =
2 + a2 N0 W 9 N0 W

Taking the ratio

2 PT
(SNR)D;SL 9 N0 W 2
= 1 PT
= = 1:76 dB
(SNR)D;C 3 N0 W
3
This is approximately 1:8 dB.

Problem 7.16
By de…nition
m(t)
mn (t) =
k m
and
m2 2 1
m
m2n = 2 = 2 2
= 2
(k m ) k m k
so that
2
PT 2 ka PT
(SNR)D = 2Ef f = 2
N0 W 1 + ka N0 W
or
2 PT
(SNR)D = 2
(k=a) + 1 N0 W
The plot is easily generated. We conclude that for k a
a 2 PT
(SNR)D 2
k N0 W
and for k a
PT
(SNR)D 2
N0 W
Problem 7.17
16 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

The transfer function of the RC highpass …lter is

R j2 f RC
H (f ) = 1 =
R + j2 C 1 + j2 f RC

so that
(2 f RC)2
jH(f )j2 =
1 + (2 f RC)2
Thus, The PSD at the output of the ideal lowpass …lter is
( 2
N0 (2 f RC)
2 1+(2 f RC)2 , jf j < W
Sn (f ) =
0, jf j>W

The noise power at the output of the ideal lowpass …lter is

Z Z
W W
(2 f RC)2
N= Sn (f ) df = N0 df
W 0 1 + (2 f RC)2

with x = 2 f RC, the preceding expression becomes

Z 2 RCW
N0 x2
N= dx
2 RC 0 1 + x2

Since
x2 1
=1
1 + x2 1 + x2
we can write Z Z
2 RCW 2 RCW
N0 dx
N= df
2 RC 0 0 1 + x2
or
N0 1
N= 2 RCW tan (2 RCW )
2 RC
which is
N0 tan 1 (2 RCW ) 2 RCW tan 1 (2 RCW )
N = N0 W = N0
2 RC 2 RC

The output signal power is

1 A2 (2 fc RC)2
S = A2 jH (fc )j2 =
2 2 1 + (2 fc RC)2
7.1. PROBLEMS 17

Thus, the signal-to-noise ratio is

S A2 (2 fc RC)2 2 RC
=
N 2N0 1 + (2 fc RC)2 2 RCW tan 1 (2 RCW )
S
Note that as W ! 1; N ! 0.

Problem 7.18
From (7.78) and (7.84), we have
2
2 n
0:05 = + 2
; SSB
m
3 2
2 2 n
0:05 = + 2
; DSB
4 m

Thus we have the following signal-to-noise ratios

2 1
m
2
= 2 ; SSB
n 0:05
2 1
m
2
= 2; DSB
n 3 2
0:05 4

The SSB characteristic has an asymptote de…ned by

2
= 0:05

and the DSB result has an asymptote de…ned by

3 2 2
= 0:05
4
or
2
= 0:258
The curves are shown in Figure 7.4. The appropriate operating regions are above and to
the left of the curves. It is clear that DSB has the larger operating region.

Problem 7.19
The mean-square error
"2 (A; ) = E [y (t) Ax (t )]2
can be written

"2 (A; ) = E y 2 (t) 2Ax(t )y (t) + A2 x2 (t )

18 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

1000

900

800
SSB
700
DSB
600

500

400
SNR
300

200

100

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35

Phase error variance

Figure 7.4: Plots for Problem 7.18.

In terms of autocorrelation and cross correlation functions, the mean-square error is

"2 (A; ) = Ry (0) 2ARxy ( ) + A2 Rx (0)

Letting Py = Ry (0) and Px = Rx (0) gives

"2 (A; ) = Py 2ARxy ( ) + A2 Px

In order to minimize "2 we choose = m such that Rxy ( ) is maximized. This follows
since the crosscorrelation term is negative in the expression for "2 . Therefore,

"2 (A; m) = Py 2ARxy ( m) + A2 P x

The gain Am is determined from

d"2
= 2ARxy ( m) + A2 Px = 0
dA
which yields
Rxy ( m)
Am =
Px
This gives the mean-square error
2 (
Rxy 2 (
m) Rxy m)
"2 (Am ; m ) = Py 2 +
Px Px
7.1. PROBLEMS 19

R(t )e jψ (t ) rn (t )sin[ψ (t ) − φ (t )] = rn (t )sin[φe (t )]

rn (t )e jφn (t ) rn (t ) cos[ψ (t ) − φ (t )] = rn (t ) cos[φe (t )]

Ac e jφ ( t )

Figure 7.5: Phasor diagram for SNR 1.

which is
2 (
Pxy m)
"2 (A m; m) = Py
Px
The output signal power is
2
SD = E [Am x (t m )] = A2m Px

which is
2 (
Rxy 2 (
m) Rxy m)
SD = =
Px Rx (0)
Since ND is the error "2 (Am ; m) we have
2 (
Rxy
SD m)
= 2 (
ND Rx (0) Ry (0) Rxy m)

Note: The gain and delay of a linear system is often de…ned as the magnitude of the
transfer and the slope of the phase characteristic, respectively. The de…nition of gain and
delay suggested in this problem is useful when the magnitude response is not linear over
the frequency range of interest.

Problem 7.20
The phasor diagram for SNR 1 is illustrated in Figure 7.5. The phasor diagram for
SNR 1 is shown in Figure 7.6. We see that the angle (t), which determines the demodu-
lated output, is approximately n (t), whereas for SNR 1 the angle (t) is approximately
(t). For (t) (t) the error in the demodulated output is small.

Problem 7.21
20 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

rn (t )sin[ψ (t ) − φ (t )] = rn (t )sin[φe (t )]

rn (t ) cos[ψ (t ) − φ (t )] = rn (t ) cos[φe (t )]

rn (t )e jφn ( t )

R(t )e jψ (t )

Ac e jφ (t )

Figure 7.6: Phasor diagram for SNR 1:

The single-sided spectrum of a stereophonic FM signal and the noise spectrum is shown in
Figure 7.7. The two-sided noise spectrum is given by

KD2
SnF (f ) = N0 f , 1<f <1
A2C

The predetection noise powers are easily computed. For the L + R channel
Z 15;000 2 2
KD 2 12 KD
Pn;L+R = 2 N 0 f df = 2:25 10 N0
0 A2C A2C

For the L R channel

Z 53;000 2 2
KD 2 12 KD
Pn;L R =2 N 0 f df = 91:14 10 N0
23;000 A2C A2C

Thus, the noise power on the L R channel is over 40 times the noise power in the L + R
channel. After demodulation, the di¤erence will be a factor of 20 because of 3 dB detection
gain of coherent demodulation. Thus, the main source of noise in a stereophonic system
is the L R channel. Therefore, in high noise environments, monophonic broadcasting is
preferred over stereophonic broadcasting.

Problem 7.22
7.1. PROBLEMS 21

S()nNoise
Pilot
()L f
+ R f ()L f
+ R f
F fSpectrum

()f kHz

0 15 23 53

Figure 7.7: Sterophonic FM baseband signal and noise.

With BT = 2(D + 1)W (7.118) becomes

2 2
BT PT 3 BT PT
(SNR)DF = 3 1 m2n = 2 m2n
2W N0 W 4 W N0 W

This clearly gives (7.119) for the case in which BT W.

The required plot follows by setting this equal to 1000 (30 dB). In order to make the plot
independent of unknown parameters normalize (SNR)DF by plotting (SNR)DF =m2n NP0TW .
The asymptotic value of (SNR)DF for BT W di¤ers from the preceding equation only
by the approximation
2
BT BT 2
2 =
W W
These di¤er by 0.5 dB when
" # " #
3 BT 2 2 PT 3 BT 2
PT
10 log mn 10 log 2 m2n = 0:5
4 W N0 W 4 W N0 W

or
BT BT
20 log 20 log 2 = 0:5
W W
This gives !
BT
W 2
log BT
= 0:025
W
22 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

or
2W 0:025
1 = 10 = 0:9441
BT
Finally
2W
=1 0:9441 = 0:0559
BT
which gives a value of
BT 2
= = 35:75
W 0:0559
Problem 7.23

From (7.132) and (7.119), the required ratio is

K2 W 1 W
2 A2D N0 f33 f3 tan f3
C
R= 2
2 KD 3
3 AC N0 W
2

or
3
f3 W 1 W
R=3 tan
W f3 f3
This is shown in Figure ??.
For f3 = 2:1kHz and W = 15kHz, the value of R is
2
2:1 15 1 15
R=3 tan = 0:047
15 2:1 2:1

Expressed in dB this is
R = 10 log10 (0:047) = 13:3 dB
The improvement resulting from the use of preemphasis and deemphasis is therefore 13:3 dB.
Neglecting the tan 1 (W=f3 ) gives an improvement of 21 8:75 = 12:25 dB. The di¤erence
is approximately 1 dB.

Problem 7.24
The received FDM spectrum is shown in Figure 7.8. The k th channel signal is given by

xk (t) = Ak mk (t) cos 2 kf1 t

Since the guardbands have spectral width 3W , 1f1 = W + 3W + W = 5W , 2f1 = f1 + W +

3W + W = 10W and so on. Clearly kf1 = 5kW and the k th channel occupies the frequency
band
(5k 1) W f (5k + 1) W
7.1. PROBLEMS 23

Since the noise spectrum is given by

KD2
SnF = N0 f 2 , jf j < BT
A2C

The noise power in the k th channel is

Z (5k+1)W
BW 3 h i
Nk = B f 2 df = (5k + 1)3 (5k 1)3
(5k 1)W 3

where B is a constant. This gives

BW 3
Nk = 150k 2 + 2 = 50BW 3 k 2
3
The signal power is proportional to A2k . Thus, the signal-to-noise ratio can be expressed as
2
A2k Ak
(SNR)D = =
k2 k

where is a constant of proportionality and is a function of KD ,W ,AC ,N0 . If all channels

are to have equal (SNR)D , Ak =k must be a constant, which means that Ak is a linear
function of k. Thus, if A1 is known, Ak = k A1 ; k = 1; 2; :::; 7. A1 is …xed by setting the
SNR of Channel 1 equal to the SNR of the unmodulated channel. The noise power of the
unmodulated channel is Z W
B
N0 = B f 2 df = W 3
0 3
yielding the signal-to-noise ratio
P0
(SNR)D = B 3
3W

where P0 is the signal power.

Problem 7.25
From the plot of the signal spectrum it is clear that
A
k=
W2
Thus the signal power is
Z W
A 2 2
S=2 2
f df = AW
0 W 3
24 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

Noise
PSD

f
1 f
2 f
3 f
4 f
5 f
6 f
7
0

Figure 7.8: FDM signal and noise for Problem 7.24.

The noise power is N0 B. This yields the signal-to-noise ratio

2 AW
(SNR)1 =
3 N0 B
If B is reduced to W , the SNR becomes
2 A
(SNR)2 =
3 N0
This increases the signal-to-noise ratio by a factor of B=W .

Problem 7.26
From the de…nition of the signal we have

x (t) = A cos 2 fc t
dx
= 2 fc A sin 2 fc t
dt
d2 x
= (2 fc )2 A cos 2 fc t
dt2
The signal component of y (t) therefore has power
1
SD = A2 (2 fc )4 = 8A2 4 4
fc
2
The noise power spectral density at y (t) is
N0
Sn (f ) = (2 f )4
2
7.1. PROBLEMS 25

so that the noise power is

Z W
N0 16
ND = (2 )4 f 4 df = N0 4
W5
2 W 5

Thus, the signal-to-noise ratio at y (t) is

4
5 A2 fc
(SNR)D =
2 N0 W W

Problem 7.27
Since the signal is integrated twice and then di¤erentiated twice, the output signal is equal
to the input signal. The output signal is

ys (t) = A cos 2 fc t

and the output signal power is

1
S D = A2
2
The noise power is the same as in the previous problem, thus
16 4
ND = N0 W5
5
This gives the signal-to-noise ratio

5 A2
(SNR)D =
32 4 N0 W 5

Problem 7.28
The signal power is
Z W Z W=f3
df dx 1 W
SD = 2A = 2Af3 = 2Af3 tan
0 1 + (f =f3 )2 0 1 + x2 f3
Since
ND = N0 W
the signal-to-noise ratio at y (t) is

2A f3 1 W
(SNR)D = tan
N0 W f3
26 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

The normalized (SNR)D , de…ned by (SNR)D =(2A=N0 ); is illustrated in Figure ??.

Problem 7.29
Note that the form of m(t) a¤ects both the numerator and the denominator of (7.139).
The numerator is a¤ected through m2n and the denominator is a¤ected through f . By
de…nition,
N
1 X
mn (t) = Cn cos(2 nf t + n )
M
n=1

where M is the maximum value of m(t). Since the phases n are unknown, the valueP of M
cannot be determined and this is as far as we can go. If n = 0 for all n, then M = Cn .
2
This, however, is a very special case. In general, given the phases n , mn , is computed as
*" N
#2 +
1 X
m2n = 2 Cn cos(2 nf t + n)
M
n=1

where, as always, h i denotes a time average value. However,

* N N N
+
1 X C2 X X
n
m2n = + Cn Ck cos(2 nf t + n ) cos(2 kf t + k)
M2 2
n=1 n=1 k=1(k6=n)
N
X
1
= Cn2
2M 2
n=1

Using m(t) and M allows us to write (7.140). Using the preceding result in (7.146), with
replaced by D since the signal is no longer sinusoidal,allows determination of (7.148) as was
done in Example 7.4. This completes the problem. Note that the di¢ culty lies in …nding
the normalization constant M . Finding M for a speci…c signal will, in general, require using
MATLAB or similar tool to plot m(t).

Problem 7.30
The instantaneous frequency deviation in Hz is

f = fd m (t) = x (t)

where x (t) is a zero-mean Gaussian process with variance 2 = fd2 2 . Thus,

x m
Z 1
jxj x2 =2 2
j f j = jxj = p e x dx
1 2 x
7.1. PROBLEMS 27

Recognizing that the integrand is even gives

r Z 1 r
2 1 x2 =2 2 2
j fj = xe x dx = x
x 0

Therefore,
r
2
j fj = fd m

Substitution into (7.148) gives

2
fd
3 W m2n NP0TW
(SNR)D = q q h i
p A2C A2
1 + 2 3 BWT Q N0 B T + 6 2 fdWm exp 2N0 BCT

The preceding expression can be placed in terms of the deviation ratio, D, by letting
fd
D=
W
and
BT = 2 (D + 1) W
Problem 7.31

The input to the PLL is

xc (t) = Ac cos(2 fc t + ) + nc (t) cos(2 fc t + ) ns (t) sin((2 fc t + )
which can be written
R(t) cos[2 fc t + + (t)] = fAc + nc (t)g cos(2 fc t + ) ns (t) sin(2 fc t + )
The phase deviation is
1 ns (t)
(t) = tan
Ac + nc (t)
Since the signal-to-noise ratio is su¢ ciency large to justify the linear model, the phase
deviation is written
ns (t)
(t) =
Ac
We know that the PSD of the quadrature noise is N0 for jf j < B=2, where B is the
predetection bandwidth. Thus, the PSD of the input signal to the baseband (lowpass)
model of the PLL is
N0
S (f ) = 2 ; jf j < B=2
Ac
28 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

Using the PLL transfer function (The linear model is assumed valid since the input SNR is
assumed large, which yields small phase errors.) gives

N0
S (f ) = S (f ) jG(f )j2 = jG(f )j2
A2c

where G(f ) is the transfer function relating the phase error (f ) to the input phase (f )
(See (3.221)). The variance of the phase error is
Z Z
N0 1 2N0 1
2
= 2 jG(f )j2 df = 2 jG(f )j2 df
Ac 1 Ac 0

The integral represents the equivalent noise bandwidth BL of the PLL based on the closed-
loop transfer function G(f ). We have obviously assumed BL B, which is typically the
case. Thus
2 2N0 BL N0 BL
= =
A2c S
where S = A2c =2 is the input signal power. The input noise is assumed Gaussian. The
phase error is therefore Gaussian since it is a linear transformation, through the transfer
function G(f ), of the input noise. The preceding expression shows that the phase error
variance is the reciprocal of the signal-to-noise ratio, where the noise power is measured in
the bandwidth of the transfer function G(f ).

Problem 7.32
The peak value of the signal is 12:5 so that the signal power is
1
S= (12:5)2 = 78:125
2
If the A/D converter has a wordlength n, there are 2n quantizing levels. Since these span
the peak-to-peak signal range, the width of each quantization level is
25 n
S= = 25 2
2n
This gives
1 2 1
"2 = S = (25)2 2 2n
= (52:08)2 2n
12 12
This results in the signal-to-noise ratio
S 78:125 2n 3 2n
SN R = = 2 = 2
"2 52:08 2
7.1. PROBLEMS 29

Problem 7.33
PPM was de…ned in Chapter 3. Assume that the pulse position is represented by

(t) = o + 1 mn (t)

where o is the nominal pulse position for m(t) = 0 and 1 establishes the peak o¤set from
the nominal pulse position and is adjusted so that the pulse does not deviate outside of its
assigned slot. We assume a zero-mean signal so that

m(t) = 0

for which = o . The pulse is detected when the received pulse crosses a threshold. When
noise is added to the pulse we can approximate the error in detecting the threshold crossing
as
(t1 ) n(t1 )
=
TR A
where TR is the pulse risetime, t1 is the point at which the pulse would cross the threshold
in the absence of noise, and A represents the pulse amplitude. Thus, in general
2 2
2
TR TR
= n2 = No BT
A A
As we saw in Chapter 2, the relationship between risetime and bandwidth gives TR = 1=2BT ,
where BT is the transmission bandwidth. We let
1 1
0 = 2TR = 2 =
2BT BT
This gives
2
2
1 No
= No BT =
2ABT 4A2 BT
The average transmitted signal power is
o
PT = A2
Ts
where Ts is the sampling frequency or 1=2W assuming Nyquist rate sampling. Using this
in the preceding equation by substituting for A2 gives

2
No o
=
4PT Ts BT
Since 1 mn (t) represents the signal, the received signal power is
2 2
SD = 1 mn
30 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

we have
2 m2 4PT Ts BT 1 4PT BT
1 n 2 2 2 2
(SNR)D = = 1 mn = 1 mn
2 No o 2W No o
or
2 2 2BT PT PT
(SNR)D = 1 mn = 2 21 m2n BT2
o No W No W
where we have used o = 1=BT . The peak-to-peak pulse displacement is 2 1 = Ts = 1=2W
so that 1 = 1=4W . Substituting this into the preceding equation yields
2
1 BT PT
(SNR)D = m2n
8 W No W

Therefore with all the assumptions made K = 1=8.

Note: We have made many assumptions in this problem including Nyquist rate sampling.
It is useful, however, for illustrating the thought process for applying the concepts of this
chapter to modulation processes other than those considered in this chapter.

7.2 Computer Exercises

Computer Exercise 7.1
From the expression for the normalized error
2
2 2 n
NQ = + 2
m

the following MATLAB program results:

% File: ce7_1a.m
snrdb = [0 1 2];
snr = 10.^(snrdb./10);
varphas = 0:0.1:1;
hold on
for k=1:3
nerrqpsk = varphas + (1/snr(k));
plot(varphas,nerrqpsk’)
end
hold off
grid, xlabel(’Phase Error Variance’), ylabel(’Normalized MSE (QDSB)’)
% End of script file.
7.2. COMPUTER EXERCISES 31

1.8

1.6
Normalized MSE (QDSB)

1.4

1.2

0.8

0.6
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Phase Error Variance

Figure 7.9: Normalized mse for QDSB and SSB with the SNR as a parameter.

Executing the program gives the results illustrated in Figure 7.9. We see from the de…ning
equation that the mse is linear in the phase error variance for a given SNR. The SNR simply
provides an o¤set (SNR = 1 (0 dB) curve is on top).

For the DSB system we have the MATLAB program

% File: ce7_1b.m
snrdb = [0 1 2];
snr = 10.^(snrdb./10);
varphas = 0:0.1:1;
hold on
for k=1:3
nerrqpsk = 0.75*varphas.*varphas + (1/snr(k));
plot(varphas,nerrqpsk’)
end
hold off
grid, xlabel(’Phase Error Variance’), ylabel(’Normalized MSE (DSB)’)
32 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

1.8

1.6
Normalized MSE (DSB)

1.4

1.2

0.8

0.6
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Phase Error Variance

Figure 7.10: Normalized mse for DSB with the SNR as a parameter.

% End of script file.

This yields the result shown in Figure 7.10. The mse is now quadradic in the phase
error variance for a given SNR. The SNR again simply provides an o¤set (SNR = 1 (0 dB)
curve is on top).

Computer Exercise 7.2

The MATLAB program for solving this computer exercise follows.

% File: ce7_2.m
zdB = 0:50; % predetection SNR in dB
z = 10.^(zdB/10); % predetection SNR
beta = [1 5 10 20]; % modulation index vector
hold on % hold for plots
for j=1:length(beta)
bta = beta(j); % current index
a1 = exp(-(0.5/(bta+1)*z)); % temporary constant
7.2. COMPUTER EXERCISES 33

80

70
20
60
Postdetection SNR in dB

50
5
40
1
30

20

10

-10
0 5 10 15 20 25 30 35 40 45 50
Predetection SNR in dB

Figure 7.11: Plot for Computer Exercise 7.2.

a2 = q(sqrt(z/(bta+1))); % temporary constant

num = (1.5*bta*bta)*z;
den = 1+(4*sqrt(3)*(bta+1))*(z.*a2)+(12/pi)*bta*(z.*a1);
result = num./den;
resultdB = 10*log10(result);
plot(zdB,resultdB,’k’)
end
hold off
xlabel(’Predetection SNR in dB’)
ylabel(’Postdetection SNR in dB’)
% End of script file.

Executing the program gives the results shown in Figure 7.11. We see that the threshold
e¤ect is less pronounced. A smaller value of yields a smaller bandwidth and decreased
detection gain.

Computer Exercise 7.3

34 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

The MATLAB program which solves this computer exercise is a modi…cation of the program
used in Computer Example 7.2 and follows:

% File: ce7_3.m
zdB = 0:0.1:40; % predetection SNR in dB
z = 10.^(zdB/10); % predetection SNR
beta = [2 4 6 8 10 15 20]; % modulation index vector
hold on % hold for plots
for j=1:length(beta)
a2 = exp(-(0.5/(beta(j)+1)*z));
a1 = q(sqrt((1/(beta(j)+1))*z));
r1 = 1+(4*sqrt(3)*(beta(j)+1)*a2);
r2 = r1+((12/pi)*beta(j)*a2);
semilogy(zdB,r2,’k’)
axis([10 30 0 4])
grid
end
hold off % release
xlabel(’Predetection SNR in dB’)
ylabel(’1+D_2+D_3’)
% End of script file.

This program gives the results shown in Figure 7.12 for = 2, 4, 6, 8 10, 15, and 20. We
see that denominator has a value of two for = 2 at approximately 13 dB. This increases
to approximately 23.5 dB at an index of 20. These values can be taken o¤ Figure 7.12
and plotted if desired. It is clear that the predetection SNR for which the denominator of
(7.148) is equal to 2 increases as the modulation index increases.

Computer Exercise 7.4

The MATLAB program for solving this computer exercise follows. The strategy used is
to plot two curves for each modulation index; one curve with the modulation not included
and one curve with the modulation included. In order to keep the results from being too
cluttered only two modulation indexes are used; = 5 and = 20.

% File: ce7_4.m
zdB = 0:50; % predetection SNR in dB
z = 10.^(zdB/10); % predetection SNR
beta = [5 20]; % modulation index vector
hold on % hold for plots
7.2. COMPUTER EXERCISES 35

3.5

2.5
1+D2+D3

4 6
2
2 8 10 15
1.5
20

0.5

0
10 12 14 16 18 20 22 24 26 28 30
Predetection SNR in dB

Figure 7.12: Values of PT =N0 W for which 1 + D2 + D3 = 2.

36 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

for j=1:length(beta)
bta = beta(j); % current index
a1 = exp(-(0.5/(bta+1)*z)); % temporary constant
a2 = q(sqrt(z/(bta+1))); % temporary constant
num = (1.5*bta*bta)*z;
den1 = 1+(4*sqrt(3)*(bta+1))*(z.*a2);
den2 = 1+(4*sqrt(3)*(bta+1))*(z.*a2)+(12/pi)*bta*(z.*a1);
result1dB = 10*log10(num./den1); % w/o modulation
result2dB = 10*log10(num./den2); % with modulation
plot(zdB,result1dB,’k’,zdB,result2dB,’k-.’)
end
hold off
legend(’w/o modulation’,’with modulation’,4)
xlabel(’Predetection SNR in dB’)
ylabel(’Postdetection SNR in dB’)
% End of script file.

Executing the program gives the results shown in Figure 7.13. We see, as expected, that
above threshold the results are identical. This results since the above-threshold performance
is only a function of the predetection SNR and the modulation index. Below threshold,
a region in which we typically have little interest, the inclusion of modulation reduces the
postdetection SNR, since the denominator of the postdetection SNR expression is increased.

Computer Exercise 7.5

The MATLAB program used in Computer Example 3.4 is modi…ed in order to allow the
addition of noise to the input phase deviation. The only parameter to be entered in this
program is the variance of the phase due to noise.

% File: ce7_5.m
% beginning of preprocessor
clear all % be safe
fdel = 40; fn = 10; zeta = 0.707;
varn = input(’Enter phase noise variance > ’);
npts = 2000; % number of simulation points
fs = 2000; % default sampling frequency
T = 1/fs;
t = (0:(npts-1))/fs; % time vector
nsettle = fix(npts/10); % set nsettle time as 0.1*npts
Kt = 4*pi*zeta*fn; % loop gain
7.2. COMPUTER EXERCISES 37

80

70

60

Postdetection SNR in dB 50

40

30

20

10
w/o modulation
with modulation
0
0 5 10 15 20 25 30 35 40 45 50
Predetection SNR in dB

Figure 7.13: FM performance results with and without modulation.

a = pi*fn/zeta; % loop filter parameter

filt_in_last = 0; filt_out_last=0;
vco_in_last = 0; vco_out = 0; vco_out_last=0;
% end of preprocessor - beginning of simulation loop
for i=1:npts
if i<nsettle
fin(i) = 0;
phin = 0;
else
fin(i) = fdel;
phin = 2*pi*fdel*T*(i-nsettle)+sqrt(varn)*randn(1);
end
s1 = phin - vco_out;
s2 = sin(s1); % sinusoidal phase detector
s3 = Kt*s2;
filt_in = a*s3;
filt_out = filt_out_last + (T/2)*(filt_in + filt_in_last);
filt_in_last = filt_in;
filt_out_last = filt_out;
vco_in = s3 + filt_out;
vco_out = vco_out_last + (T/2)*(vco_in + vco_in_last);
38 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

vco_in_last = vco_in;
vco_out_last = vco_out;
phierror(i)=s1;
fvco(i)=vco_in/(2*pi);
freqerror(i) = fin(i)-fvco(i);
end
% end of simulation loop - beginning of postprocessor
subplot(2,1,1), plot(t,fin,t,fvco), grid
xlabel(’Time - Seconds’)
ylabel(’Frequency - Hertz’)
subplot(2,1,2), plot(phierror/2/pi,freqerror), grid
xlabel(’Phase Error / 2*pi’)
ylabel(’Frequency Error - Hz’)
% end of postprocessor
% End of script file.

The program is executed twice, …rst with a phase noise variance of zero as a reference and
then with a phase noise variance of 0.4. The results are illustrated in Figures 7.14 and
7.15. We see that the acquisition time is increased. The student is warned that with noise
present, the number of cycles slipped, and consequently the acquisition time, is a random
variable. This should be kept in mind when attempting to reproduce these results.

Computer Exercise 7.6

(Note: This computer exercise is closely related to Problem 7.31.) The input to the PLL is
assumed to be a sinusoid plus bandlimited white noise. Thus the PLL input is modeled as

xc (t) = Ac cos(2 fc t + ) + nc (t) cos(2 fc t + ) ns (t) sin(2 fc t + )

or
xc (t) = [Ac + nc (t)] cos(2 fc t + ) ns (t) sin(2 fc t + )

This can be represented as

xc (t) = R(t) cos[2 fc t + + (t)]

where
1 ns (t) 1 ns (t) ns (t)
(t) = tan tan =
Ac + nc (t) Ac Ac
7.2. COMPUTER EXERCISES 39

60

Frequency - Hertz
40

20

-20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time - Seconds
60
Frequency Error - Hz

40

20

-20
0 0.5 1 1.5 2 2.5 3 3.5
Phase Error / 2*pi

Figure 7.14: Results for phase noise variance 2 = 0.

pn

80
Frequency - Hertz

60

40

20

-20
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time - Seconds
60
Frequency Error - Hz

40

20

-20

-40
-1 0 1 2 3 4 5 6
Phase Error / 2*pi

Figure 7.15: Results for phase noise variance 2 = 0:4.

pn
40 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

The approximation results since the input SNR is assumed large. We therefore use (t) as
the input to the baseband PLL model. The SNR of the input signal is

A2c =2 A2c
= 2
=
n 2 2n

where 2 = n2s . The variance of (t) is

n

2 1
2 n
= =
A2c 2

In the following program we set the phase error variance. Note that this does not determine
the SNR. One can set 2n at some value (such as 1) and then adjust Ac to give the required
SNR.

% File: ce7_6.m
% beginning of preprocessor
clear all % be safe
fn = 10; zeta = 0.707;
varn = input(’Enter noise variance > ’);
npts = 50000; % number of simulation points
fs = 2000; % sampling frequency
T = 1/fs;
t = (0:(npts-1))/fs; % time vector
Kt = 4*pi*zeta*fn; % loop gain
a = pi*fn/zeta; % loop filter parameter
filt_in_last = 0; filt_out_last=0;
vco_in_last = 0; vco_out = 0; vco_out_last=0;
% end of preprocessor - beginning of simulation loop
for i=1:npts
phin = sqrt(varn)*randn(1);
s1 = phin - vco_out;
aa(1,i) = s1;
s2 = sin(s1); % sinusoidal phase detector
s3 = Kt*s2;
filt_in = a*s3;
filt_out = filt_out_last + (T/2)*(filt_in + filt_in_last);
filt_in_last = filt_in;
filt_out_last = filt_out;
vco_in = s3 + filt_out;
vco_out = vco_out_last + (T/2)*(vco_in + vco_in_last);
7.2. COMPUTER EXERCISES 41

4500

4000

3500

3000
Number of Samples

2500

2000

1500

1000

500

0
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
Histogram Bin

Figure 7.16: Histogram of phase error.

vco_in_last = vco_in;
vco_out_last = vco_out;
phierror(i)=s1;
end
% end of simulation loop - beginning of postprocessor
[N_samp,x] = hist(aa,40); % get histogram parameters
bar(x,N_samp,1)
xlabel(’Histogram Bin’), ylabel(’Number of Samples’)
% end of postprocessor
% End of script file.

Executing the program gives the histogram illustrated in Figure 7.16. The histogram sug-
gests that the pdf of the phase error is closely gaussian.

Computer Exercise 7.7

The MATLAB program for this computer exercise is a simple modi…cation of the program
used for Computer Example 7.3. The only change is for the bit error probabilities for FSK
(frequency-shift keying) and BPSK (binary phase-shift keying). In order to keep the results
from being cluttered results are computed for n = 8. The MATLAB program follows.
42 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

% File: ce7_7.m
n = 8; % wordlength
snrtdB = 0:0.1:30; % predetection snr in dB
snrt = 10.^(snrtdB/10); % predetection snr
Pb1 = q(sqrt(snrt)); % bit error probability for FSK
Pb2 = q(sqrt(2*snrt)); % bit error for BPSK
Pw1 = 1-(1-Pb1).^n; % Pw for FSK
Pw2 = 1-(1-Pb2).^n; % Pw for BPSK
a = 2^(-2*n); % temporary constant
snrd1 = 1./(a+Pw1*(1-a));
snrddB1 = 10*log10(snrd1); % postdetection snr (FSK)
snrd2 = 1./(a+Pw2*(1-a));
snrddB2 = 10*log10(snrd2); % postdetection snr (BPSK)
plot(snrtdB,snrddB1,’k’,snrtdB,snrddB2,’k-.’)
legend(’FSK’,’BPSK’)
xlabel(’Predetection SNR in dB’)
ylabel(’Postdetection SNR in dB’)
% End of script file.

The results are shown in Figure 7.17. We see that above threshold, the postdetection SNR is
independent of the modulation scheme. This, of course, was expected since above threshold
the postdetection SNR is a function only of wordlength. We also note that the use of
BPSK extends the threshold over that of FSK due to the 3-dB improvement in bit error
probability.

Computer Exercise 7.8

The main calculations are accomplished in the function …le as follows.

function [gain,delay,px,py,rxy,rho,snrdb] = snrmse(x,y)

ln = length(x); % length of reference (x) vector
fx = fft(x,ln); % FFT the reference (x) vector
fy = fft(y,ln); % FFT the measurement (y) vector
fxconj = conj(fx); % Conjugate the FFT of reference vector
sxy = fy .* fxconj; % Cross PSD
rxy = ifft(sxy,ln); % Cross correlation function
rxy = real(rxy)/ln; % Take the real part and scale
px = x*x’/ln; % Power in reference vector
py = y*y’/ln; % Power in measurement vector
[rxymax,j] = max(rxy); % Find max of the crosscorrelation
7.2. COMPUTER EXERCISES 43

50
FSK
45 BPSK

40

35
Postdetection SNR in dB
30

25

20

15

10

0
0 5 10 15 20 25 30
Predetection SNR in dB

Figure 7.17: PCM performance for FSK and BPSK with n = 8:

gain = rxymax/px; % System gain

delay = j-1; % System delay
rxy2 = rxymax*rxymax; % Square rxymax for later use
rho = rxymax/sqrt(px*py); % Here’s the correlation coefficient
snr = rxy2/(px*py-rxy2); % SNR
snrdb = 10*log10(snr); % SNR in db
% End of function file.

The preceding function …le is called by a program that sets up the reference signal and
the measuremenjt signal. The system under study for this example is a channel having a
sinusoidal input. An interfering signal and noise are added in the channel. The paremeters
are established in the following program.

% File: snrmsepost.m
npts = 1024; % data points
n = 1:npts;
A = 20; % signal amplitude
B = 2; % interference amplitude
phii = rand(1); % interference phase
fd = 2; % signal frequency
fi = 3; % interference frequency
theta = 2*pi*n/npts; % phase increment
44 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

nvar = 0.1; % noise variance

noise = sqrt(nvar)*randn(1,npts); % noise samples
xx = A*cos(fd*theta); % reference signal
y1 = (A/2)*cos(fd*theta+pi/4);
y2 = B*cos(fi*theta+phii);
y3 = noise;
yy = y1+y2+noise; % measurement
[gain,delay,px,py,rxy,rho,snrdb] = snrmse(xx,yy);
% end of script

Prior to executing the program we …rst calculate the SNR at the point of measurement.
Some of the obvious results are as follows:

(20)2
Px = = 200
2
1 A 2 (2)2
Py = + + 0:1 = 50 + 2 + 0:1 = 52:1
2 2 2
1
Gain =
2
512
Delay = = 64
8

These values are easily understood. The reference signal is a sinusoid with a peak
amplitude A = 20, which has a power of 200 Watts. The power at the measurement
point is the sum of a signal of peak amplitude of A=2, an independent interfering sinusoid
with peak amplitude B = 2, and additive noise having variance 0.1. Since the signal
at the measurement point is a sinusoid of amplitude A=2 and the reference signal has a
peak amplitude of amplitude A, the theoretical gain is 0.5. The delay requires a bit more
thought. Since 1024 samples are taken in an FFT block, and the reference signal goes
through 2 periods in 1024 samples, the period of the signal is 512 samples. Since the phase
shift of the signal at the measurement point is =2, which is 1/8 of a period, the theoretical
delay is 512/8 or 64 samples. In addition, since the signal is periodic, a valid delay is
k(512) 64 samples where k is an arbitrary integer.
The theoretical SNR requires a bit of calculation. Since the interference is orthogonal
to the signal (di¤erent frequencies) we consider the interference to be noise. Also, the noise
is independent of the signal. Therefore the noise power is the sum of the interference power
plus the noise power. This gives
(2)2
N= + 0:1 = 2:1
2
7.2. COMPUTER EXERCISES 45

The signal power is

2
1 A
S= = 50
2 2

Thus the SNR in dB is

50
(SNR)dB = 10 log = 13:7675
2:1

Executing the program gives

» snrmsepost
» px
px = 200.0000
» py
py = 52.4432
» gain
gain = 0.5016
» delay
delay = 448
» snrdb
snrdb = 13.7322

All of these valuse are very close to the theoretical values previously determined. We should
recall that any calculation involving noise will be a random variable. One source of error
is that a su¢ cient number of samples must be processed to ensure that the variance of
the result must be small. Another source of error, and often the most important source of
error, results in the peak of the crosscorrelation function being missed. This problem can
be reduced by increasing the sampling frequency.

Computer Exercise 7.9

Since the signal is a sinusoid, we can use an inverse sin function as a compressor charac-
teristic. Note that this actually expands the signal but we will use the term ‘compressor’
since this is the …rst operation performed. The compressor output is a triangular wave so
that the compressor output has a uniform pdf ranging from =2 to =2. Thus, with 3 bit
46 CHAPTER 7. NOISE EFFECTS IN SIMPLE MODULATION SYSTEMS

2 1

0.8
1.5
0.6
1

Compressor Characteristic
0.4

Expander Characteristic
0.5
0.2

0 0

-0.2
-0.5
-0.4
-1
-0.6
-1.5
-0.8

-2 -1
-1 -0.5 0 0.5 1 -2 -1 0 1 2
Time Time

Figure 7.18: Compressor and expander characteristics.

words we have 8 quantization levels given by

L1 = 0 X =8
L2 = =8 X =4
L3 = =4 X 3 =8
L4 = 3 =8 X =2
L5 = =8 X 0
L6 = =4 X =8
L7 = 3 =8 X =4
L8 = =2 X 3 =8

The expander characteristic has a sinusoidal characteristic to restore the original sinusiodal
signal. The compressor and expander characteristics are given in Figure 7.18 and the original
signal, the compressed signal, and the expanded signal are shown in Figure 7.19.
7.2. COMPUTER EXERCISES 47

1
Orig. sig.

-1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2
Compressed sig.

-2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
1
Expanded sig.

-1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time

Figure 7.19: Original signal, compressed signal, and expanded signal.