0% found this document useful (0 votes)

30 views64 pages

Chap7 - Digital Transmission Via Carrier Modulation

Uploaded by

toptrum294

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

30 views64 pages

Chap7 - Digital Transmission Via Carrier Modulation

Uploaded by

toptrum294

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 64

Chapter 7

Digital Transmission via

Carrier Modulation

7.1 Preview

In the two preceding chapters, we considered the transmission of digital information

through baseband channels. In such a case, the information-bearing signal is trans
mitted directly through the channel without the use of a sinusoidal carrier. However,
most communication channels are bandpass channels; hence, the only way to transmit
signals through such channels is by shifting the frequency of the information-bearing
signal to the frequency band of the channel.
In this chapter, we consider four types of carrier-modulated signals that are suitable
for bandpass channels: amplitude-modulated signals, quadrature-amplitude-modulated
signals, phase-shift keying, and frequency-shift keying.

7.2 Carrier-Amplitude Modulation

In baseband digital PAM, the signal waveforms have the form

(7 .2.1)

where Am is the amplitude of the mth waveforms and gy(t) is a pulse whose shape
determines the spectral characteristics of the transmitted signal. The spectrum of the
baseband signals is assumed to be contained in the frequency band Iii � W, where
W is the bandwidth of I Gy(j) 12, as illustrated in Figure 7 .1. Recall that the signal
amplitude takes the discrete values

Am= (2m-l-M)d, m=l,2,... ,M (7 .2.2)

where 2d is the Euclidean distance between two adjacent signal points.

313
Copyright 2011 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).

Editorial review has deemed that any suppressed content does not material1y affect the overall learning experience. Cengage Leaming reserves the right to remove additional content at any time if subsequent rights restrictions require it.
31 4 CHAPT ER 7. DIGITAL T RANSMISSION VIA CARRIER MODULATION

--=------L---=L-�/
-W 0 W

Figure 7.1 : Energy density spectrum of the transmitted signal Br(t)

Baseband
signal --
'
---1•� x f----·
Bandpass signal
sm(t) cos 21tfc t
sm(t)

Carrier
cos (27tfc t)

Figure 7.2: Amplitude modulation of a sinusoidal carrier by the baseband PAM signal

To transmit the digital signal waveforms through a bandpass channel, the baseband
signal waveforms Sm(t), m = 1,2,... ,M, are multiplied by a sinusoidal carrier of the
form cos 2rrJct, as shown in Figure 7.2, where Jc is the carrier frequency (Jc > W)
and corresponds to the center frequency in the passband of the channel. Hence, the
transmitted signal waveforms are expressed as

Um(t) = AmBr(t) COS 2rrJct, m = 1,2,... ,M (7.2.3)

In the special case when the transmitted pulse shape gy(t) is rectangular-that is,

gy(t) = {�·0,
0 �

otherwise
t � T

the amplitude-modulated carrier signal is usually called amplitude-shift keying (ASK).

In this case the PAM signal is not bandlimited.
Amplitude modulation of the carrier cos 2rrJct by the baseband signal waveforms
Sm(t) shifts the spectrum of the baseband signal by an amount Jc and, thus, places the
signal into the passband of the channel. Recall that the Fourier transform of the carrier
is [ 8 (j - Jc) + 8 (j + Jc)] I2. Since multiplication of two signals in the time domain
corresponds to the convolution of their spectra in the frequency domain, the spectrum
of the amplitude-modulated signal is

(7.2.4)

Thus, the spectrum of the baseband signal Sm(t) = AmBr(t) is shifted in frequency by
the carrier frequency Jc. The bandpass signal is a double-sideband suppressed-carrier
(DSB-SC) AM signal, as illustrated in Figure 7.3.

Copyright 2011 Cengage Leaming. AJI Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to eJectronic rights, some third party content may be suppressed from the eBook andlor eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions l'«tuire it.
7.2. CARRIER-AMPLITUDE MODULATION 315

�---J
-t:==----��-1-��=-��
-W 0 W

(a)

A,,/2

��-- /
O
-Jc - W -fc -fc +
W fc- W fc fc + W
(b)

Figure 7 .3: Spectra of (a) baseband and (b) amplitude-modulated signals

-5d -3d -d 0 d 3d 5d

Figure 7 .4: Signal point constellation for PAM signal

We note that impressing the baseband signal Sm(t) onto the amplitude of the carrier
signal cos 2rrJc (t) does not change the basic geometric representation of the digital
PAM signal waveforms. The bandpass PAM signal waveforms may be represented in
general as
Um(t) = Sml./J(t) (7 .2.5)
where the signal waveform l./J(t) is defined as

t.jJ(t) = Br(t) cos 2rrJct (7 .2.6)

and
m=l,2, . . . ,M (7.2.7)
denotes the signal points that take the M values on the real line, as shown in Figure 7.4 .
The signal waveform t.jJ(t) is normalized to unit energy; that is,

(7.2.8)

Consequently,

oo 1 oo 1 oo
f _
00
Bf (t) cos2 2rrJct dt = 2
J _
00
Bf (t) dt + 2
J_
00
Bf (t) cos 4rrJct dt

=1 (7 .2.9)

But

(7.2.10)

because the bandwidth W of BT(t) is much smaller than the carrier frequency-that
is,Jc » W. In such a case, BT(t) is essentially constant within any one cycle of
cos4rrJct; hence, the integral in (7.2.10) is equal to zero for each cycle of the inte
grand. In view of (7.2.10), it follows that

� f:00 Bf (t) dt = 1 (7.2.11)

Therefore, BT(t) must be appropriately scaled so that (7.2 .8) and (7.2.11) are satisfied.

72 1. . Demodulation of PAM Signals

The demodulation of a bandpass digital PAM signal may be accomplished in one of
several ways by means of correlation or matched filtering. For illustrative purposes we
consider a correlation-type demodulator.
The received signal may be expressed as

r(t) =AmBT(t) cos 2rrJct + n(t) (7.2.12)

where n(t) is a bandpass noise process, which is represented as

n(t) = nc(t) cos 2rrJct - ns (t) sin 2rrJct (7.2.13)

and where nc(t) and n5(t) are the quadrature components of the noise. By cross
correlating the received signal r(t) with 1.fJ (t) given by (7.2.6), as shown in Figure 7.5,
we obtain the output

f:00 r(t)ljJ(t) dt =Am+ n =Sm+ n (7.2.14)

where n represents the additive noise component at the output of the correlator.
The noise component has a zero mean. Its variance can be expressed as

(7.2.15)

Received --�
signal r(t)
f�()dt Sampler To detector

Signal
pulse Clock
generator

Oscillator

Figure 7.5: Demodulation of bandpass digital PAM signal

where 'I' (j) is the Fourier transform of lfJ (t) and Sn(j) is the power spectral density
of the additive noise. The Fourier transform of lfJ (t) is

(7.2.16)

and the power spectral density of the bandpass additive noise process is

Sn(j)
{ No
2 '
If - fc I � W
(7.2.17)
0,
=

otherwise

By substituting (7.2.16) and (7.2.17) into (7.2.15) and evaluating the integral, we obtain
a� = No/2.
It is apparent from (7.2.14), which is the input to the amplitude detector, that the
probability of error of the optimum detector for the carrier-modulated PAM signal is
identical to that of baseband PAM. That is,

2(M -
1) Q ( 6(log2 M)Eavb ) (7.2.18)
(M2 - l)No
PM
M
=

where Eavb is the average energy per bit.

Illustrative Problem 7 .1 [PAM Signal Spectrum] In an amplitude-modulated digi

tal PAM system, the transmitter filter with impulse response gy(t) has a square-root
raised-cosine spectral characteristic as described in Illustrative Problem 6.8, with a
rolloff factor ex = 0.5. The carrier frequency is fc = 40/T. Evaluate and graph the
spectrum of the baseband signal and the spectrum of the amplitude-modulated signal.

Figure 7.6 illustrates these two spectral characteristics. The M ATLAB script for this
computation is given next.

120 -�-
, -
, �-
, -
, �, �-
, ,� �-
, ,�

100 �

150

80 �

100 60 �

40 �

50
20 �

0 o .__...__,__.___,__.___.___.___,__..._____.
-1 -0.5 0 0.5 -50 -40 -30 -20 -10 0 10 20 30 40 50

f f

Figure 7.6: Spectra of baseband signal and amplitude-modulated (bandpass) signal

-41111i" ------
% MATLAB script for Illustrated Problem 7.1.
echo on
T=1;
delta_T=T/200; % sampling interval
alpha=0.5; % rollojf factor
fc=40/T; % carrier frequency
A_m=1; % amplitude
t=-5*T +delta_T:delta_T:5*T; % time axis
N=length(t);
for i=1:N,
if (abs(t(i)r=T/(2*alpha)),
g_T(i) = sinc(t(i)/T)*(cos(pi*alpha*t(i)/f)/(1-4*alphaA2*t(W2/fA2));
else
g_T(i) = O; % The value of g_T is 0 at t=Tl(2*alpha)
end; % and at t=-Tl(2*alpha).
echo off ;
end;
echo on;
G_T=abs(fft(g_ T)); % spectrum of g_T
u_m=A_m*g_T.*cos(2*pi*fc*t); % the modulated signal
u_m=abs(fft(u_m)); % spectrum of the modulated signal
% actual frequency scale
f=-0.5/delta_ T:1/(delta_ T*(N-1)):0.5/delta_ T;
% Plotting commands follow.
figure(1);
plot(f,fftshift(G_ T));
axis([-1/T 1/f 0 max(G_T)]);
figure(2);
plot(f,fftshift(U_m));

---llMllJ§il;ful�Nii;t•1:Jl§@I
Illustrative Problem 7 .2 [Demodulation of Bandpass Digital PAM Signal]
Figure 7 .5 illustrates the demodulation of a bandpass digital PAM signal, which
involves cross-correlating the received signal

r(t) = AmBT(t) cos 2rrfct + n(t)

with the reference waveform

<jJ(t) = gy(t) cos 2rrfct

where Am is the transmitted signal amplitude,gy(t) is the rectangular signal pulse

gy(t) = { ti
0,
·
0:5t:5T

otherwise

and n(t) is a bandpass Gaussian noise process, which is represented as

n(t) = nc(t) cos 2rrfct - ns(t) sin2rrfct

and where nc(t) and ns(t) are the quadrature components of the noise.
Let us implement the demodulator in the discrete time, so that the correlator output
is
n

y(nTs) = L r(kTs)tfl(kTs), n= 1, 2, ...

k=O

where the sampling interval is Ts = T/100 and the carrier frequency Jc = 30/T. The
noise samples nc(kTs) and ns(kTs) are statistically independent, zero-mean Gaussian
with variance cr2• Perform the computation and plot y( nTs) for n= 1, 2, ..., 100 and
2 2 2
cr = 0, cr = 0.0 5, and cr = 0.5.

•H•1llill•1ijt
For convenience we set T = 1. Figure 7 .7 shows the demodulator output over the entire
signal interval. First we note that the double frequency terms average out to zero, as
best observed in the case where cr2 = 0. Secondly, we observe the effect of the additive
noise on the correlator output as cr2 increases.
The MATLAB script for the problem is given below.

----®Ii" ------
% MATLAB script for Illustrative Problem 7.2

Am= 1; % Signal Amplitude

T= 1;
Ts= 100ff;
fc= 30tr;
t= O:T/100:T;
Lt= length(t);
g_T= sqrt(2/f)*ones(1,Lt);
si= g_T .* cos(2*pi*fc*t);
var= [ 0 0.05 0.5]; % Noise variance vector
for k= 1 : length(var)
% Generation of the noise components:
n_c= sqrt(var(k))*randn(1,Lt);
n_s= sqrt(var(k))*randn(1,Lt);
noise= n_c.*cos(2*pi*fc*t) - n_s.*sin(2*pi*fc*t);
r = Am*g_T.*cos(2*pi*fc*t)+noise; % The received signal
y= zeros(1,Lt);
for i= 1:Lt
y(i)= sum(r(1:i).*si(1:i)); % The correlator output
end
% Plotting the results:
subplot(3,1,k)
plot([O 1:length(y)-1],y)
title([ ' \ s i grna" 2 = ' ,num2str(var(k))])
xlabel( ' n' )
ylabel('y(nT_s) ' )
end

2
(J = 0

100

20 40 60 80 100
n

2
cr = 0.05

100

20 40 60 80 100
n

2
(J = 0.5

100

20 40 60 80 100
n

Figure 7.7: Demodulator output in Illustrative Problem 7.2

7 .3 Carrier-Phase Modulation

In carrier-phase modulation the information that is transmitted over a communication

channel is impressed on the phase of the carrier. Because the range of the carrier phase
is 0 :::; e < 2rr' the carrier phases used to transmit digital information via digital-phase
modulation are em = 2rrm/M, form = 0,1,... ,M 1. Thus, for binary phase -

modulation (M = 2), the two carrier phases are eo = 0 and e1 = TT rad. For M-ary
phase modulation, M = zk, where k is the number of information bits per transmitted
symbol.
The general representation of a set of M carrier-phase-modulated signal waveforms
is

m = 0,1,... , M - 1 (7.3.1)

where BT(t) is the transmitting filter pulse shape, which determines the spectral char
acteristics of the transmitted signal, and A is the signal amplitude. This type of digital
phase modulation is called phase-shift keying (PSK). We note that PSK signals have
equal energy; that is,

(7.3.2)
00

2rrm
=
f _ A2Bf(t) cos2
( 2rr fc t + �
) dt
"" ""
00

1 1 4rrm
= 2
J _
00
A2Bf(t) dt + 2
J _
00
A2Bf(t) cos
( 4rr fc t + �
) dt
A2 "" Bf(t) dt
= -

2 J -
00
(7.3.3)

for allm (7.3.4)

where E5 denotes the energy per transmitted symbol. The term involving the double
frequency component in (7.3.2) averages out to zero when fc » W, where W is the
bandwidth of BT(t).
W hen BT(t) is a rectangular pulse, it is defined as

BT(t) = �' 0:::; t:::; T (7.3.5)

In this case, the transmitted signal waveforms in the symbol interval 0 :::: t :::: T may
be expressed as (with A = {Es)

2rrm
( )
{2E;
um(O = \;-r cos 2rr fc t + � , m = 0,1,... , M - 1 (7.3.6)

Note that the transmitted signals given by (7.3.6) have a constant envelope, and the
carrier phase changes abruptly at the beginning of each signal interval. Figure 7 .8
illustrates a four-phase (M = 4) PSK signal waveform.

180°-phase 0°-phase -90°-phase

shift shift shift

t t t
!\ (\ f\ (\ (\ (:(\ (\ )lo t

0 T 2T 3T 4T

Figure 7 .8: Example of a four-phase PSK signal

By viewing the angle of the cosine function in (7.3 .6) as the sum of two angles, we
may express the waveforms in (7.3.1) as

Um(t) = ffs BT(t) cos ( 2:m ) cos 2rrfct - ffs BT(t) sin ( 2:m ) sin2rrfct
= Smcl.fJ1(t) + Smsl.fJ2(t) (7.3.7)

where

m 2rr
Smc = ffs COS �

Sms
re: . 2rrm
= v.c:.s Sill � (7 .3 .8)

and l.fJ1(t) and l.fJ2(t) are orthogonal basis functions defined as

l/JI (t) = BT(t) cos 2rrJct

1.fJ2(t) = -BT(t) sin2rrfct (7.3.9)

By appropriately normalizing the pulse shape BT(t), we can normalize the energy of
these two basis functions to unity. Thus, a phase-modulated signal may be viewed as
two quadrature carriers with amplitudes that depend on the transmitted phase in each
signal interval. Hence, digital phase-modulated signals are represented geometrically
as two-dimensional vectors with components Smc and Sms -that is,

Sm = ( .JEs cos Z�m .JEs sin Z�

m ) (7.3.10)

Signal point constellations for M = 2, 4, and 8 are illustrated in Figure 7.9. We observe
that binary phase modulation is identical to binary PAM (binary antipodal signals).
The mapping, or assignment, of k information bits into the M = 2k possible phases
may be done in a number of ways. The preferred assignment is to use Gray encoding,
in which adjacent phases differ by one binary digit, as illustrated in Figure 7.9. Conse
quently, only a single bit error occurs in the k-bit sequence with Gray encoding when
noise causes the erroneous selection of an adjacent phase to the transmitted phase.

01 00
·-----

010. 000

I
I
I
I
I uo
I
I
•
I
I
-- ---- ----·
11 10 111• •101

M=2 M=4 M=8

Figure 7 .9: PSK signal constellations

----tlMQ1".il;f4iiltJ#IW;t•1=JM§I·•------
Illustrative Problem 7.3 [PSK Waveform] Generate the constant-envelope PSK sig
nal waveforms given by (7.3.6) for M = 8. For convenience, the signal amplitude is
normalized to unity.

Figure 7 .10 illustrates the eight waveforms for the case in which Jc 6/T. The
MATLAB script for this computation is given next.

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

Figure 7 .10: M = 8 constant-amplitude PSK waveforms

-41111i" ------
% MATLAB script for Illustrative Problem 7.3.
echo on
T=1;
M=B;
Es=T/2;
fc=6/T; % carrier frequency
N=100; % number of samples
delta_ T=T/(N-1);
t=O:delta_ T:T;
u0=sqrt(2*Es/T)*cos(2*pi*fc*t);
ul=sqrt(2*Es/T)*cos(2*pi*fc*t+2*pi/M);
u2=sqrt(2*Es/T)*cos(2*pi*fc*t+4*pi/M);
u3=sqrt(2*Es/T)*cos(2*pi*fc*t+6*pi/M);
u4=sqrt(2*Es/T)*cos(2*pi*fc*t+8*pi/M);
u5=sqrt(2*Es/T)*cos(2*pi*fc*t+1 O*pi/M);
u6=sqrt(2*Es/T)*cos(2*pi*fc*t+12*pi/M);
u7=sqrt(2*Es/T)*cos(2*pi*fc* t+14*pi/M);
% plotting commands follow
subplot(8,1,1);
plot(t,uO);
subplot(8,1,2);
plot(t,ul);
subplot(8,1,3);
plot(t,u2);
subplot(8,1,4);
plot(t,u3);
subplot(8,1,5);
plot(t,u4);
subplot(8,1,6);
plot(t,u5);
subplot(8, 1,7);
plot(t,u6);
subplot(8,1,8);
plot(t,u7);

7.3.1 Phase Demodulation and Detection

The received bandpass signal in a signaling interval from an AWGN channel may be
expressed as

r(t) = Um(t) + n(t)

= Um(t) + nc (t) cos 2rr Jct - n5 (t) sin 2TTfct (7.3.11)

where nc (t) and n5 (t) are the two quadrature components of the additive noise.
The received signal may be correlated with l/Ji (t) and 1./12 (t) given by (7.3.9). The
outputs of the two correlators yield the noise-corrupted signal components, which may

be expressed as

r =Sm+ n

( v EsCOS
rr 2rrm
=
� +nc (7.3.12)

where nc and n5 are defined as

nc = 2
1
Joo gy(t)nc(t) dt
_00

1 oo
ns = 2 J gy(t)ns(t) dt
_00 (7.3.13)

nc(t) and n5(t) are zero-mean Gaussian ran

The quadrature noise components
dom processes that are uncorrelated. As a consequence, E(nc) = E(n5) = 0 and
E(ncn5) = 0. The variance of nc and n5 is
E(n�) = E(n;) = �o (7.3.14)

----tlll4-il;fu1Wj§;t•1:JM§I·•------
Illustrative Problem 7 .4 [Demodulation of PSK Signals]
In this problem, we consider the demodulation of a M = 4 PSK signal waveform
r(t) as given in (7.3.11), where the transmitted signal is given in (7.3.7) and n(t) is
the additive Gaussian noise process. The pulse shape gy(t) is rectangular; that is,

gy(t) =
{.ft 0 :5, t :5, T
0 otherwise

The demodulator employs two quadrature correlators to compute the two inputs to the
detector at the sampling time T.

Yc(t) = J: r(T)l./J1 (T) dT

Ys(t) = f: r(T)l./Jz(T) dT
where l./J1 (t) and l./Jz(t) are two orthogonal basis waveforms given by (7.3.9).
Let us implement the two correlators in discrete time. Thus, the two outputs of the
correlators are
n

Yc(nTs) = L r(kTs)l./J1 (kTs), n = 1, 2, ...

k=O
n

Ys(nTs) = L r(kTs)l./Jz(kTs), n = 1, 2, ...

k=O

T5 = T /100 and the carrier frequency Jc = 3 0/T . The

where the sampling interval is
noise samples nc(kT5) and n5(kT5) are statistically independent, zero-mean Gaus
sian with variance u2• Perform the computation and plot Ye (nT5) and Ys(nT5) for

n = 1, 2, ... , 10 0, cr2 = 0, cr2 = 0.0 5, cr2 = 0.5 and each of the phases in a four
phase PSK signal.

For convenience we set T = 1. Figure 7 .11 illustrates the correlator outputs over the
entire signal interval for the four possible transmitted phases. Note that the double fre
quency terms average out to zero, as best observed in the case where cr2 = 0 . Secondly,
2
we observe the effect of the additive noise on the correlator outputs as cr increases.
The MATLAB script for the problem is given below.

----411®1@11 ------
% MATLAB script for Illustrative Problem 7.4

M= 4;
Es= 1; % Energy per symbol
T= 1;
Ts= 100ff;
fc= 30ff;
t= O:T/1
OO:T;
Lt= length(t);
g_T= sqrt(2ff)*ones(1,Lt);
sLl = g_T.*cos(2*pi*fc*t);
sL2= -g_T.*sin(2*pi*fc*t};
for m= 0: 3
% Generation of the transmitted signal:
s_mc= sqrt(Es} * cos(2*pi*m/M);
s_ms= sqrt(Es} * sin(2*pi*m/M);
u_m= s_mc.*sLl + s_ms.*sL2;
var= [ 0 0.05 0.5]; % Noise variance vector
if (m== 2)
figure
end
for k= 1: length(var)
% Generation of the noise components:
n_c= sqrt(var(k}}*randn(1,Lt);
n_s= sqrt(var(k}}*randn(1,Lt);
% The received signal:
r= u_m + n_c.*cos(2*pi*fc*t) - n_s.*sin(2*pi*fc*t);
% The correlator outputs:
y_c= zeros(1,Lt);
y_s= zeros(1,Lt);
for i= 1
:Lt
y_c(i)= sum(r(1 :i));
:i).*sL1 (1
y_s(i)= sum(r(1 :i));
:i).*sL2(1
end
% Plotting the results:
subplot(3,2,2*k-1+mod(m,2))
plot([O :length(y_c)-1],y_c,'
1 • - ')
hold
plot([O :length(y_s)-1],y_s)
1
title([ '\ sigma" 2 = ',num2str(var(k}}]}

xlabel(['n (rn=',num2str(m),') '])

axis auto
end
end

.!Ao00' ,'.0,0°90'
150 100

�.. 4C

-50 -20
0 10 20 80 40 50 60 70 80 90 100 0 10 20 90 40 50 60 70 80 90 100

.'00,0°180 ,'.0,0°210'
50 50

'
? ?
�,
-50 �· -5C

-100 -100

-150 -150
0 10 20 80 40 50 60 70 80 90 100 0 10 20 90 40 50 80 70 80 90 100

o'.o.os,o.o ,'.o.os,0.90
150 150

100 100

�� 50 ! "SC

-50 -50
0 10 20 80 40 50 60 70 80 90 100 0 10 10 so 40 50 60 70 80 90 100

o'0Q05,0°180' o'00.05,0°270'
50 20

-20

�" -4C

-60

-80

-150 -100
0 10 20 80 40 50 60 70 80 90 100 0 10 10 so 40 50 80 70 80 90 100

,'.05,••0· ,'.o.5,•·90
150 150

100 100

�.. SC

-50 -50
0 10 20 so 40 50 60 70 80 90 100 0 10 10 311 40 50 60 70 80 90 100

o':0.5,0.180 o':OS,0•270
10 20

�.. -4C

-50

-80

-100
10 20 so 40 50 60 70 80 90 100 0 10 10 311 40 50 60 70 80 90 100

Figure 7 11 : Correlator outputs in Illustrative Problem 7.4

The optimum detector projects the received signal vector r onto each of the M
possible transmitted signal vectors {Sm} and selects the vector corresponding to the
largest projection. Thus, we obtain the correlation metrics

C(r, Sm) = r ·Sm, m = 0, 1, . . . ,M - 1 (7.3.15)

Because all signals have equal energy, an equivalent detector metric for digital phase
modulation is to compute the phase of the received signal vector r = (r1, rz) as
_1 r2
er= tan - (7.3.16)
r1
and select the signal from the set {Sm} whose phase is closest to er.
The probability of error at the detector for phase modulation in an AWGN channel
may be found in any textbook that treats digital communications. Because binary phase
modulation is identical to binary PAM, the probability of error is

(7.3.17)

where Eb is the energy per bit. Four-phase modulation may be viewed as two bi
nary phase modulation systems on quadrature (orthogonal) carriers. Consequently, the
probability of a bit error is identical to that for binary phase modulation. For M > 4,
there is no simple closed-form expression for the probability of a symbol error. A good
approximation for PM is

PM � 2Q (\j�
No sm
. TT

M
)
� 2Q ()2kEb --

No
.
sm-
TT

M
) (7.3.18)

where k = log2M bits per symbol. Figure 7.12 illustrates the symbol-error probability
as a function of the SNR Eb I No.
The equivalent bit-error probability for M-ary phase modulation is also difficult to
derive due to the dependence of the mapping of k-bit symbols into the corresponding
signal phases. When a Gray code is used in the mapping, two k-bit symbols corre
sponding to adjacent signal phases differ in only a single bit. Because the most proba
ble errors due to noise result in the erroneous selection of an adjacent phase to the true
phase, most k-bit symbol errors contain only a single bit error. Hence, the equivalent
bit-error probability for M-ary phase modulation is well approximated as

(7.3.19)

---ti!!IJ§ji;MiW§lij;t•]:'9¥@·1-----
Illustrative Problem 7 .5 [PSK Simulation] Perform a Monte Carlo simulation of an
M = 4 PSK communication system that models the detector as the one that computes

2
1
1 0-
5
....
0
2
� Z
'+-< 10 -
0
5
-�
� 2

t 10 -
3

Q...::s 5

2
4
10-
5

2
S
10-
-4 0 4 8 12 16 20 24
SNR/bit, dB

Figure 7.12: Probability of a symbol error for M-ary PSK

the correlation metrics given in (7.3 .15). The model for the system to be simulated is
shown in Figure 7.13.

Uniform random
number generator Gaussian RNG

4-PSK
Detector 1---r-...- 2-bit symbol
mapper

Gaussian RNG

Compare

Bit-error Symbol-error
counter counter

Figure 7.13: Block diagram of an M = 4 PSK system for a Monte Carlo simulation

As shown, we simulate the generation of the random vector r given by (7.3.12), which
is the output of the signal correlator and the input to the detector. We begin by

generating a sequence of quaternary (2-bit) symbols that are mapped into the corre
sponding four-phase signal points, as shown in Figure 7 .9 for M 4. To accomplish
this task, we use a random number generator that generates a uniform random num
ber in the range (0, 1). This range is subdivided into four equal intervals, (0,0.25),
(0.25,0.5), (0.5,0.75), and (0.75, 1.0), where the subintervals correspond to the pairs
of information bits 00, 01, 11, and 10, respectively. These pairs of bits are used to
select the signal phase vector Sm.

The additive noise components nc and ns are statistically independent, zero-mean

2
Gaussian random variables with variance a- • For convenience, we may normalize the
2
variance to a- 1 and control the SNR in the received signal by scaling the signal
energy parameter Es, or vice versa.
The detector observes the received signal vector r Sm+ n as given in (7.3.12) and
computes the projection (dot product) of r onto the four possible signal vectors Sm. Its
decision is based on selecting the signal point corresponding in the largest projection.
The output decisions from the detector are compared with the transmitted symbols, and
symbol errors and bit errors are counted.
Figure 7.14 illustrates the results of the Monte Carlo simulation for the transmission
of N=lOOOO symbols at different values of the SNR parameter Eb/No, where Eb
Es /2 is the bit energy. Also shown in Figure 7.14 is the bit-error rate, which is defined
as Pb �PM /2, and the corresponding theoretical error probability, given by (7.3.18).
The MATLAB scripts for this Monte Carlo simulation are given next.

llE Simulated bit-error rate

O Simulated symbol-error rate

- Theoretical bit-error rate

10-4

10-s

10-6 '--�-'--�-'-�-'-�--'-�_,_�__,_�---'�---''---
0 2 3 4 5 6 7 8 9 10

Figure 7.14: Performance of a four-phase PSK system from the Monte Carlo simulation

--411ili" -------
% MATLAB script for Illustrative Problem 7.5.
echo on
SNRindB 1=0:2:1O;
SNRindB2=0:0.1:1O;
for i=1:length(SNRindBI),
[pb,ps]=cm_sm32(SNRindBl(i)); % simulated bit and symbol error rates
smld_biLerr_prb(i)=pb;
smld_symboLerr_prb(i)=ps;
echo off ;
end;
echo on;
for i=1:length(SNRindB2),
SNR=exp(SNRindB2(i)*log(10)/1O); % signal-to-noise ratio
theo_err_prb(i)=Qfunct(sqrt(2*SNR)); % theoretical bit-error rate
echo off
end;
echo on ;
% Plotting commands follow
semilogy(SNRindBI,smld_biLerr_prb,' * ' );

hold
semilogy(SNRindBI,smld_symboLerr_prb, 'o' );
semilogy(SNRindB2,theo_err_prb);

--411®1i" -------
function [pb,ps]=cm_sm32(snr_in_dB)
% [pb,ps]=cm...sm32(snr_in_dB)
% CM..SM32 finds the probability of bit error and symbol error for the
% given value of snr _in_dB, signal-to-noise ratio in dB.
N=10000;
E=1; % energy per symbol
snr=10"(snr_in_dB/10); % signal-to-noise ratio
sgma=sqrt(E/snr)/2; % noise variance
% the signal mapping
s00=[1 O];
sOI=[O 1];
sll=[-1 O];
slO=[O -1];
% generation of the data source
for i=1:N,
temp=rand; % a uniform random variable between 0 and 1
if (temp<0.25), % With probability 114, source output is "00."
dsourcel(i)=O;
dsource2(i)=0;
elseif (temp<0.5), % With probability 114, source output is "01."
dsourcel(i)=O;
dsource2(i)=1;
elseif (temp<0.75), % With probability 114, source output is "10."
dsourcel(i)=1;
dsource2(i)=O;
else % With probability 114, source output is "11."

dsourcel(i)=1;
dsource2(i)=1;
end;
end;
% detection and the probability of error calculation
numofsymbolerror=O;
numofbiterror=O;
for i=1 :N,
% The received signal at the detector, for the ith symbol, is:
n(1)=gngauss(sgma);
n(2)=gngauss(sgma);
if ((dsourcel(i)==O) & (dsource2(i)==0)),
r=sOO+n;
elseif ((dsourcel(i)==O) & (dsource2(i)==1)),
r=sOl+n;
elseif ((dsourcel(i)==1) & (dsource2(i)==0)),
r=slO+n;
else
r=sll+n;
end;
% The correlation metrics are computed below.
cOO=dot(r,sOO);
cOl=dot(r,sOl);
clO=dot(r,slO);
cll=dot(r,sl1);
% The decision on the ith symbol is made next.
c_max:=max:([cOO cOl clO ell]);
if (cOO==c_max:),
decis1=0; decis2=0;
elseif (cOl==c_max:),
decis1=0; decis2=1;
elseif (clO==c_max:),
decis1=1; decis2=0;
else
decis1=1; decis2=1;
end;
% Increment the error counter, if the decision is not correct.
symbolerror=O;
if (decisi-=dsourcel(i)),
numofbiterror=numofbiterror+1;
symbolerror=1;
end;
if (decisZ-=dsource2(i)),
numofbiterror=numofbiterror+1;
symbolerror=1;
end;
if (symbolerror==1),
numofsymbolerror = numofsymbolerror+1;
end;
end;
ps=numofsymbolerror/N; % since there are totally N symbols
pb=numofbiterror/(2*N); % since 2N bits are transmitted

732. . Differential Phase Modulation and Demodulation

The demodulation of the phase-modulated carrier signal, as described above, requires

that the carrier-phase components l./J1(t) and l./Jz(t) be phase locked to the received
carrier-modulated signal. In general, this means that the receiver must estimate
the carrier-phase offset of the received signal caused by a transmission delay through
the channel and compensate for this carrier-phase offset in the cross-correlation of the
received signal with the two reference components 1./11(t) and l./Jz(t). The estimation
of the carrier-phase offset is usually performed by use of a phase-locked loop (PLL).
Thus, we achieve coherent phase demodulation.
Another type of carrier-phase modulation is differential phase modulation, in which
the transmitted data are differentially encoded prior to the modulator. In differential
encoding, the information is conveyed by phase shifts relative to the previous signal
interval. For example, in binary phase modulation, the information bit 1 may be trans
mitted by shifting the phase of the carrier by 180° relative to the previous carrier phase,
whereas the information bit 0 is transmitted by a zero-phase shift relative to the phase
in the preceding signaling interval. In four-phase modulation, the relative phase shifts
between successive intervals are 0°, 90°, 180°, and 2 7 0° , corresponding to the infor
mation bits 00, 01, 11, and 10, respectively. The generalization of differential encoding
for M > 4 is straightforward. The phase-modulated signals resulting from this encod
ing process are called differentially encoded. The encoding is performed by a relatively
simple logic circuit preceding the modulator.
Demodulation and detection of the differentially encoded phase-modulated signal
1
may be performed as follows. The received signal phase er tan- rz I r1 at the detec
tor is mapped into one of the M possible transmitted signal phases {em} that is closest
to er. Following the detector is a relatively simple phase comparator that compares the
phases of the detected signal over two consecutive intervals to extract the transmitted
information.
We observe that the demodulation of a differentially encoded phase-modulated
signal does not require the estimation of the carrier phase. To elaborate, suppose
that we demodulate the differentially encoded signal by cross-correlating r(t) with
Br(t) cos 2rrfct and -Br(t) sin2rrfct. At the kth signaling interval, the two com
ponents of the demodulator output may be represented in complex-valued form as

(7 .3.20)

where ek is the phase angle of the transmitted signal at the kth signaling interval,
c/> is the carrier phase, and nk nkc + }nks is the noise. Similarly, the received
signal vector at the output of the demodulator in the preceding signaling interval is the
complex-valued quantity

(7.3.21)

The decision variable for the phase detector is the phase difference between these two
complex numbers. Equivalently, we can project rk onto rk-1 and use the phase of the

Received Delay Phase Output

signal byT comparator decision

Figure 7.15: Block diagram of DPSK demodulator

resulting complex number; that is,

rkr:_1 EseJCBk-ek-1) + ffs eJCBk-<f>lnk-1 + ffs e-JCBk-i-<f>lnk + nknz_1 (7.3.22)

which, in the absence of noise, yields the phase difference ek - ek-1· Thus, the mean
value of rkr:_1 is independent of the carrier phase. Differentially encoded PSK sig
naling that is demodulated and detected as described above is called differential PSK
(DPSK). The demodulation and detection of DPSK are illustrated in Figure 7.15.
The probability of error for DPSK in an AWGN channel is relatively simple to
derive for binary (M 2) phase modulation. The result is

P2
.!e-Eb!No (7.3.23)
2
The graph of (7.3.23) is shown in Figure 7.16. Also shown in this figure is the probability

10-1

2
10-2
5
.....

5 2

] 10-3
'H
0 5
.£
r.:::= 2
�
..c 10-4
0
tl.. 5
r::...""
2
10-5
5

2
10--{i
0 2 4 6 8 10 12 14
SNR/bit, dB

Figure 7.16: Probability of error for binary PSK and DPSK

of error for binary PSK. We observe that at error probabilities below io-4, the differ
ence in SNR between binary PSK and binary DPSK is less than 1 dB.
For M > 2, the error probability performance of a DPSK demodulator and detector
is extremely difficult to evaluate exactly. The major difficulty is encountered in the
determination of the probability density function for the phase of the random variable
rkrt_1, given by (7.3.22). However, an approximation to the performance of DPSK is
easily obtained, as we now demonstrate.
W ithout loss of generality, suppose the phase difference ek - ek-1 0. Further
-i
more, the exponential factors e-j(ek -</>l and ejCek-</>l in (7.3.22) can be absorbed into
the Gaussian noise components nk- l and nk without changing their statistical proper
ties. Therefore, rkrL1 in (7.3.22) can be expressed as

(7.3.24)

The complication in determining the probability density function of the phase is the
term nknt_1. However, at SNRs of practical interest, the term nknL 1 is small relative
to the dominant noise term $s (nk + nt_1). If we neglect the term nknt_1 and we
also normalize rkrt_1 by dividing through by .,[E;, the new set of decision metrics
becomes

x ffs + Re(nk + nt_1)

y Im(nk + nt_1) (7.3.25)

The variables x and y are uncorrelated Gaussian random variables with identical vari
ances er� No. The phase is
1y
er tan- (7.3.26)
x

At this stage we have a problem that is identical to the one for phase-coherent demod
ulation. The only difference is that the noise variance is now twice as large as in the
case of PSK. Thus we conclude that the performance ofDPSK is 3 dB poorer than that
for PSK. This result is relatively good for M � 4, but it is pessimistic for M 2, in
the sense that the loss in binary DPSK relative to binary PSK is less than 3 dB at large
SNR.

----tlll4-il;tfilrijij;J•):)!§i·•-----
Illustrative Problem 7 .6 [DPSK Encoder] Implement a differential encoder for the
case of M 8 DPSK.

The signal points are the same as those for PSK shown in Figure 7.9. However, for
DPSK these signal points represent the phase change relative to the phase of the previ
ous transmitted signal points.

The MATLAB script for implementing the differential encoder is given next.

----tl&li" ------
% MATLAB script for Illustrative Problem 7.6.
mapping=[O 1 3 2 7 6 4 5]; % for Gray mapping
M=B;
E=1;
sequence=[O 1 0 0 1 1 0 0 1 1 1 1 1 0 0 0 O];
[e]=cm_dpske(E,M,mapping,sequence); % e is the differential encoder output.

Illustrative Problem 7.7 [Demodulation and Detection of DPSK Signals]

The demodulator and detector for DPSK signals is shown in Figure 7.15. The
quadrature correlators are identical to those employed in PSK (see Illustrative Prob
lem 7.4). Using the two correlator outputs, implement a differential detector for the
case of M = 8 DPSK.

The MATLAB script for implementing differential detector is given below.

----tl&li" ------
% MATLAB script for Illustrative Problem 7.7

M = 8;
mapping=[O 1 3 2 7 6 4 5]; % Gray mapping
Es = 1; % Energy per symbol
T = 1;
Ts = 100ff;
fc = 30ff;
t = T/100:T/100:2*T;
Lt = length(t};
g_T = sqrt(2tr)*ones(1,Lt};
si_l = g_T.*cos(2*pi*fc*t);
sL2 = -g_T.*sin(2*pi*fc*t);
var = 0.05; % Noise variance
% Determine the differential phase:
m = 2; % 0 <= m <= 7
theta_d = 2*pi*m/M;
% Assuming the phase of the first txed symbol, i.e., the reference phase is 0:
s_mcl = sqrt(Es) • 1;
s_msl = sqrt(Es} • O;
u_ml = s_mcl*sL1(1:Lt/2) + s_msl*sL2(1:Lt/2);
s_mc2 = sqrt(Es} * cos(theta_d);
s_ms2 = sqrt(Es) * sin(theta_d);
u_m2 = s_mc2*sL1(1:Lt/2) + s_ms2*sL2(1:Lt/2);
% Generation of the noise components:
n_c = sqrt(var}*randn(1,Lt};

n_s = sqrt(var)*randn(1,Lt);
% The received signals:
rl = u_ml +n_c(1:Lt/2).*cos(2*pi*fc*t(1:Lt/2)) - n_s(1:Lt/2).*sin(2*pi*fc*t(1:Lt/2));
r2 = u_m2+n_c(Lt/2+ 1:Lt).*cos(2*pi*fc*t(Lt/2+1 :Lt)) - n_s(Lt/2+1:Lt).*sin(2*pi*fc*t(Lt/2+1:Lt));
r = [rl r2];
% Detec tion of the mapped symbol and differential phase:
[detect phase_d] = dd_dpsk(r,M,mapping);

�111u.11;m•�1#1W;t•1:'93®·•------
Illustrative Problem 7 .8 Perform a Monte Carlo simulation of an M 4 DPSK com
munication system. The model for the system to be simulated is shown in Figure 7.17.

UnifonnRNG Gaussian RNG

M=4
M=4DPSK 2-bit
Delay DPSK 1----.--.
mapper output
detector

Gaussian RNG

Compare

Symbol-error
counter

Figure 7.17: Block diagram of M 4 DPSK system for the Monte Carlo simulation

The uniform random number generator (RNG) is used to generate the pairs of bits
{00, 01, 11, 10}, as described in Illustrative Problem 7.5. Each 2-bit symbol is mapped
into one of the four signal points Sm [cos rrm/2 sin rrm/2], m 0, 1, 2, 3, by
the differential encoder. Two Gaussian RNG are used to generate the noise components
[nc n5 ]. Then the received signal-plus-noise vector is

r [cos rr;n + nc sin rr;n + ns ]

The differential detector basically computes the phase difference between rk and Yk- I .

Mathematically, this computation can be performed as in (7.3.22)-that is,

rkr:_1 (rck +jrsd(rc k-1-jrsk-1)

rckrc k-1 +rskrsk-1 +j(rskrc k-1-rckrsk-1)
Xk + j Y k

and ek tan-1(Ykl Xk) is the phase difference. The value of ek is compared with
the possible phase differences {0°, 90°, 180°, 2 70°}, and a decision is made in favor
of the phase that is closest to ek. The detected phase is then mapped into the pair of
information bits. The error counter counts the symbol errors in the detected sequence.
Figure 7.18 illustrates the results of the Monte Carlo simulation for the transmis
sion of N=l0,000 symbols at different values of the SNR parameter Eb/No, where
Eb E5/2 is the bit energy. Also shown in Figure 7.18 is the theoretical value of the
symbol error rate based on the approximation that the term nk n -l is negligible. We k
observe from Figure 7.18 that the approximation results in an upper bound to the error
probability.

Figure 7.18: Performance of four-phase DPSK system from Monte Carlo simulation
(the solid curve is an upper bound based on approximation that neglects the noise term
n n
l k-1)
The MATLAB scripts for this Monte Carlo simulation are given next.

----41113'
111 ' �---��
% MATLAB script for Illustrative Problem 7.8.
echo on
SNRindB1=0:2:12;

SNRindB2=0:0.1:12;
for i=1 :length(SNRindBl),
srnld_err_prb(i)=cm_sm34(SNRindB 1 (i)); % simulated error rate
echo off
end;
echo on ;
for i=1 :length(SNRindB2),
SNR=exp(SNRindB2(i)*log(10)/10); % signal-to-noise ratio
theo_err_prb(i)=2*Qfunct(sqrt(SNR)); % theoretical symbol error rate
echo off
end;
echo on ;
% Plotting commands follow
semilogy(SNRindB 1, srnld_ err_prb, ' * ' );
hold
semilogy(SNRindB2,theo_ err_prb);

----®Ii" ------
function [p]=cm_sm34(snr_in_dB)
% [p]=cm....sm34(snr _in....dB)
% CM__sM34 finds the probability of error for the given
% value of snr ...in....dB, signal-to-noise ratio in dB.
N=10000;
E=1; % energy per symbol
snr=10�(snr_in_dB/10); % signal-to-noise ratio
sgma=sqrt(E/(4*snr)); % noise variance
% Generation of the data source follows.
for i=1 :2*N,
temp=rand; % a uniform random variable between 0 and 1
if (temp<0.5),
dsource(i)=O; % With probability 112, source output is " O ."

% else.
else
dsource(i)=1; % With probability 112, source output is "1 "

end;
end;
% Differential encoding of the data source follows
mapping=[O 1 3 2];
M=4;
[difLenc_output] = cm_dpske(E,M,mapping,dsource);
% Received signal is then
for i=1 :N,
[n(1) n(2)]=gngauss(sgma);
r(i,:)=diff_enc_output(i,:)+n;
end;
% detection and the probability of error calculation
numoferr=O;
prev_ theta=O;
for i=1 :N,
theta=angle(r(i,1 )+j*r(i,2));
delta_ theta=mod(theta-prev_ theta,2*pi);
if ((delta_theta<pi/4) I (delta_theta> 7*pi/4)),

decis=[O OJ;
elseif (delta_ theta<3*pi/4),
decis=[O 1 ];
elseif (delta_theta<5*pi/4)
decis=[1 1 ];
else
decis=[1 O];
end;
prev_theta=theta;
% Increase the error counter, if the decision is not correct.
((decis(1r=dsource(2*i-1)) I (decis(2f=dsource(2*i))),
if
numoferr=numoferr+1;
end;
end;
p=numoferr/N;

----tl®li" -------
function [enc_comp] = cm_dpske(E,M,mapping,sequence);
% [enc_comp] cm-<ipske(E,M,mapping,sequence)
% CMJ>PSKE differentially encodes a sequence.
% E is the average energy, M is the number of constellation points,
% and mapping is the vector defining how the constellation points are
% allocated. Finally, "sequence" is the uncoded binary data sequence.
k=log2(M};
N=length(sequence);
% If N is not divisible by k, append zeros, so that it is ...
remainder=rem(N,k};
if (remainder-=0),
for i=N+1:N+k-remainder,
sequence(i)=O;
end;
N=N+k-remainder;
end;
theta=O; % Initially, assume that theta=O.
for i=1:k:N,
index=O;
for j=i:i+k-1,
index=2*index+sequenceU);
end;
index=index+ 1 ;
theta=mod(2*pi*mapping(index)/M+theta,2*pi);
enc_comp((i+k-1)/k,1)=sqrt(E}*cos(theta);
enc_comp((i+k-1)/k,2)=sqrt(E}*sin(theta);
end;

7.4 Quadrature Amplitude Modulation

A quadrature-amplitude-modulated (QAM) signal employs two quadrature carriers,
cos 2 rr fc t and sin 2 TT fc t, each of which is modulated by an independent sequence

of information bits. The transmitted signal waveforms have the form

Um(t) =AmcBT(t) cos 2rrfct + AmsBT(t) sin2rrfct, m= 1,2,... ,M

(7.4.1)
where {Ame} and {Ams} are the sets of amplitude levels that are obtained by mapping
k-bit sequences into signal amplitudes. For example, Figure 7.19 illustrates a 16-QAM
signal constellation that is obtained by amplitude modulating each quadrature carrier by
M = 4 PAM. In general, rectangular signal constellations result when two quadrature
carriers are each modulated by PAM.

• • • •

Figure 7.19: M = 16-QAM signal constellation

More generally, QAM may be viewed as a form of combined digital amplitude

and digital phase modulation. Thus the transmitted QAM signal waveforms may be
expressed as

Umn(t) = AmBT(t) cos(2rrfct + 8n), m=l,2,... ,M1, n=l,2,... ,M2

(7.4.2)
If M1 = 2k1 and Mz = 2k2, the combined amplitude- and phase-modulation method
results in the simultaneous transmission of k1 + kz = log2M1M2 binary digits occurring
at a symbol rate Rb I ( k i + kz). Figure 7.20 illustrates the functional block diagram of
a QAM modulator.

Transmitting
Balanced
filter
modulator
gT(t)

Oscillator
Serial-to
Binary Transmitted
parallel
data 90° phase QAM signal
converter
shift

Transmitting
Balanced
filter
modulator
gT(t)

Figure 7.20: Functional block diagram of modulator for QAM

t"- _ ... - -- - ----

M=64
-e--•--•--
T
I I

+ •
/
M=32
... - --..- ---- - --' •
+
I / ' I
/ M= 16 '
.. • ...
t t
·--·
I
- --
T I
I
I I I M=81 I I
• • T--t- • •
I -r---t
1M=4
I I
I
• '
I
+--� _._ _ __.
I
' •
I I I I I
I I
I I I I
' ' ..__ _ ___ .. __ .. JI '
I ' / I
'
T
I
•
'... - --+-- -- -- -
/
/
•
I
T
I I
M=8 M= 16
L --e---e--•--·--... ---+----l

(a) (b) (c)

Figure 7.21: (a) Rectangular and (b), (c) circular QAM signal constellations

It is clear that the geometric signal representation of the signals given by (7 . 4.1)
and (7.4.2) is in terms of two-dimensional signal vectors of the form

Sm = ( .JEs Ame .JEs Ams) , m = 1,2, . . . ,M (7.4.3)

Examples of signal space constellations for QAM are shown in Figure 7.21. Note that
M = 4 QAM is identical to M = 4 PSK.

74 1 . . Demodulation and Detection of QAM

Let us assume that a carrier-phase offset is introduced in the transmission of the signal
through the channel. In addition, the received signal is corrupted by additive Gaussian
noise. Hence, r(t) may be expressed as

r(t) = AmcBT(t)cos(2rrfct + cp) + AmsBT(t) sin(2rrfct + cp) + n(t) (7.4.4)

where cp is the carrier-phase offset and

n(t) = nc(t)cos 2rrfct - ns(t) sin 2rrfct

The received signal is correlated with the two phase-shifted basis functions

l./J1(t) = gy(t)cos(2rrfct + cp)

l./Jz(t) = gy(t) sin(2rrfct + cp) (7.4.5)

as illustrated in Figure 7.22, and the outputs of the correlators are sampled and passed
to the detector. The PLL shown in Figure 7.22 estimates the carrier-phase offset cp
of the received signal and compensates for this phase offset by phase shifting I/Ji (t)
and l./J 2 (t) as indicated in (7.4.5). The clock shown in Figure 7.22 is assumed to be

Sampler

Clock

Compute
Output
distance
decision
metrics D(sm)
90 phase
shift

f�()dt Sampler

Figure 7.22: Demodulation and detection of QAM signals

synchronized to the received signal so that the correlator outputs are sampled at the
proper instant in time. Under these conditions, the outputs from the two correlators are

r1 = Ame + nc cos cf> - ns sin cf>

r2 = Ams + nc sin cf> + ns cos cf> (7.4.6)

where

nc = 2
1 f T nc(t)gy(t) dt
0

ns = � J: ns(t)gy(t) dt (7.4.7)

The noise components are zero-mean, uncorrelated Gaussian random variables with
variance No/2.
The optimum detector computes the distance metrics

D(r,sm) =Ir - sml2, m=l,2, . . . ,M (7.4.8)

where rt = (r1, r2) and Sm is given by (7.4.3).

----t11••�i•;MINjj;t•'=''a®·•------
Illustrative Problem 7 .9 [Demodulation of QAM Signals]
The demodulator for a QAM signal is shown in Figure 7.22. It employs two quadra
ture correlators that cross-correlate the received signal given by (7.4.4) with the phase
quadrature basis functions in (7.4.5). The pulse shape BT(t) is rectangular; that is,

BT(t) � {f Q:5,t:5,T
otherwise

Let us implement the two correlators in discrete time. Thus, the two outputs of the
correlators are

Yc (nTs) = L r (kTs)l}J 1 (kTs), n = 1, 2, ...

k=O
n

Ys (nTs) = L r (kTs)l}J2 (kTs), n = 1, 2, ...

k=O

where the sampling interval is Ts = T/ 100 and the carrier frequency Jc = 30 / T. The
carrier phase</> may be selected to be uniformly distributed in the interval (0, 2rr), and
the additive noise samples nc (kTs) and ns (kTs) are statistically independent, zero
mean Gaussian with variance a-2• Perform the computation and plot yc (nTs) and
Ys (nTs) for n = 1, 2, ... , 1 0 0, a-2 0, a-2 0.0 5, a-2
= 0.5 for the M
= 8 QAM = =

signal constellation shown in Figure 7.21(b). We may select any one of the eight signal
points for transmission.

The position of the eight signal points are (1, 1), (1, -1), (-1, 1), (-1, -1), (1 +
J3, 0), (-1, -J3, 0), (0, 1 + J3), and (0, -1 - J3). For convenience, we set T 1. =

Figure 7.23 illustrates the correlator outputs over the signal interval when the transmit
ted symbol is (1, -1) . Note that the double frequency terms average out to zero, as
best observed in the case where 0. Furthermore, a-2 = we observe the effect of the
additive noise on the correlator output as a-2 increases.
The MATLAB script for the problem is given below.

% MATLAB script for Illustrative Problem 7.9

M = 8;
Es = 1; % Energy oer symbol
T = 1;
Ts = 100ff;
fc = 30ff;
t = O:T/1 OO:T;
Lt = length(t);
A_mc = 1/sqrt(Es); % Signal Amplitude
A_ms = -1/sqrt(Es); % Signal Amplitude
g_T = sqrt(2/f)*ones(1,Lt);
phi = 2*pi*rand;
sLl = g_T.*cos(2* pi*fc*t + phi);
sL2 = g_ T.*sin(2*pi*fc*t + phi};
var = [ 0 0.05 0.5]; % Noise variance vector
for k = 1 : length(var)
% Generation of the noise components:
n_c = sqrt(var(k}}*randn(1,Lt};
n_s = sqrt(var(k}}*randn(1,Lt};
noise = n_c.*cos(2*pi*fc+t) - n_s.*sin(2*pi*fc+t);
% The received signal
r = A_mc*g_T.*cos(2*pi*fc*t+phi) + A_ms*g_T.*sin(2*pi*fc*t+phi} + noise;
% The correlator outputs:

y_c = zeros(1,Lt);
y_s = zeros(1,Lt);
for i = 1 :Lt
y_c(i) = sum(r(1 :i).*sL1(1 :i));
y_s(i) = sum(r(1 :i).*sL2(1 :i));
end
% Plotting the results:
subplot(3,1,k)
plot([O 1 :length(y _c)-1],y _c,' • - ')
hold
plot([O 1 :length(y_s)-1],y_s)
title([ ' \ s i grna" 2 = ',num2str(var(k))])
xlabel( ' n ')

axis auto
end

cr2 = 0

100

-100 .__��-'-��

0 20 40 60 80 100
n
cr2 = 0.05

100

-100 .__��-'--��-'-��-'-��--'

0 20 40 60 80 100
n
(l = 0.5

100

-100 ��

0 20 40 60 80 100
n

Figure 7.23: Correlator outputs in Illustrative Problem 7 .9

7.4.2 Probability of Error for QAM in an AWGN Channel

In this section, we consider the performance of QAM systems that employ rectangular
signal constellations. Rectangular QAM signal constellations have the distinct advan
tage of being easily generated as two PAM signals impressed on phase quadrature
carriers. In addition, they are easily demodulated.
For rectangular signal constellations in which M = 2k, where k is even, the QAM
signal constellation is equivalent to two PAM signals on quadrature carriers, each hav
ing JM= 2k/Z signal points. Because the signals in the phase quadrature components
are perfectly separated by coherent detection, the probability of error for QAM is eas
ily determined from the probability of error for PAM. Specifically, the probability of a
correct decision for the M-ary QAM system is

(7.4.9)

where P -JM is the probability of error of a JM-ary PAM with one-half the average
power in each quadrature signal of the equivalent QAM system. By appropriately
modifying the probability of error for M-ary PAM, we obtain

(7.4.10)

where Eav /No is the average SNR per symbol. Therefore, the probability of a symbol
error for the M-ary QAM is

(7.4.11)

We note that this result is exact for M = 2k when k is even. On the other hand, when
k is odd, there is no equivalent JM-ary PAM system. This is no problem, however,
because it is rather easy to determine the error rate for a rectangular signal set. If we
employ the optimum detector that bases its decisions on the optimum distance metrics
given by (7.4 .8), it is relative!y straightforward to show that the symbol-error probabil
ity is tightly upper bounded as

[ ( ) 2
�
J
3Eav
PM 1 - 1 - 2Q
(M l)No

)
-

3kEavb
(7.4.12)
(M - l)No

for any k � 1, where Eavb I No is the average SNR per bit. The probability of a symbol
error is plotted in Figure 7.24 as a function of the average SNR per bit.

Illustrative Problem 7 .10 [QAM Simulation] Perform a Monte Carlo simulation of

an M= 16-QAM communication system using a rectangular signal constellation. The
model of the system to be simulated is shown in Figure 7.25.

1
10-

2
2
10-

� 5

0 2
�
I'< 10-
3

�
O'd 5

S
10-

2
6
10- ��
-6 -4 -2 0 2 4 6 8 10 12 14 16 18 20
SNR/bit, dB

Figure 7.24: Probability of a symbol error for QAM

UniformRNG GaussianRNG

4-bit symbol
r
c
M= 16-QAM
Detector -�--- 4 -bit symbol
signal selector

GaussianRNG

Compare

Symbol-error
counter

Figure 7.25: Block diagram of an M = 16-QAM system for the Monte Carlo simula
tion

The uniform random number generator (RNG) is used to generate the sequence of infor
mation symbols corresponding to the 16 possible 4-bit combinations of bi, b2, b3, b4.
The information symbols are mapped into the corresponding signal points, as illustrated

• • 3 • •

• • 1 • •

-3 -2 -1 2 3
• • • •
-1
-2

•
.-3 • •

Figure 7.26: M = 16-QAM signal constellation for the Monte Carlo simulation

in Figure 7.26 , which have the coordinates [Ame, Ams]. Two Gaussian RNG are used
to generate the noise components [ ne, ns]. The channel-phase shift cf> is set to 0 for
convenience. Consequently, the received signal-plus-noise vector is

r = [Ame + ne Ams + ns ]
The detector computes the distance metrics given by (7.4.8) and decides in favor of
the signal point that is closest to the received vector r. The error counter counts the
symbol errors in the detected sequence. Figure 7.27 illustrates the results of the Monte
Carlo simulation for the transmission of N = 10 ,000 symbols at different values of the
SNR parameter Eb I No, where Eb = Es I 4 is the bit energy. Also shown in Figure 7.27
is the theoretical value of the symbol-error probability given by (7.4.10) and (7.4.11).

10-2

10·3

10·4

10-5

10·6

10·7 '--��--'-��-'--'
0 5 10 15

Figure 7.27: Performance of M = 16-QAM system from the Monte Carlo simulation.

The MATLAB scripts for this problem are given next.

----tlili" -------
% MATLAB script for Illustrative Problem 7.10.
echo on
SNRindB1=0:2:15;
SNRindB2=0:0.1:15;
M=16;
k=log2(M);
for i=1:length(SNRindBl),
srnld_err_prb(i)=cm_sm41(SNRindB1(i)); % simulated error rate
echo off;
end;
echo on ;
for i=1:length(SNRindB2),
SNR=exp(SNRindB2(i)*log(10)/1O); % signal-to-noise ratio
% theoretical symbol error rate
theo_err_prb(i)=4*Qfunct(sqrt(3*k*SNR/(M-1)));
echo off
end;
echo on ;
% Plotting commands follow.
semilogy(SNRindBl,srnld_err_prb,' * ');
hold
semilogy(SNRindB2,theo_err_prb);

----tlili" ------
function [p]=cm_sm41 (snr_in_dB)
% [p]=cm...sm41(snr_in....d B)
% CM...SM41 finds the probability of error for the given
% value of snr_in....dB, SNR in dB.
N=10000;
d=1; % min. distance between symbols
Eav=10*d"2; % energy per symbol
snr=10"(snr_in_dB/1O); % SNR per bit (given)
sgma=sqrt(Eav/(8* snr)); % noise variance
M=16;
% Generation of the data source follows.
for i=1 :N,
temp=rand; % a uniform R. V. between 0 and 1
dsource(i)=1 +fioor(M*temp); % a number between 1 and 16, uniform
end;
% Mapping to the signal constellation follows.
mapping=[-3*d 3*d;
-d 3*d;
d 3*d;
3*d 3*d;
-3*d d''
-d d;
d ct·'
3*d ct·'

-3*d -d;
-d -d;
d -d;
3*d -d;
-3*d -3*d;
-d -3*d;
d -3*d;
3*d -3*d];
for i=1 :N,
qam_sig(i, :)=mapping(dsource(i), :);
end;
% received signal
for i=1 :N,
[n(1) n(2)]=gngauss(sgma);
r(i, :)=qam_sig(i, :)+n;
end;
% detection and error probability calculation
numoferr=O;
for i=1 :N,
% Metric computation follows.
for j=1 :M,
metricsU)=(r(i,1)-mappingU,1W2+(r(i,2)-mappingU,2W2;
end;
[min_metric decis] = min(metrics);
if (decis-=dsource(i)),
numoferr=numoferr+1;
end;
end;
p=numoferr/(N);

7 .5 Carrier-Frequency Modulation

We have described methods for transmitting digital information by modulating either

the amplitude of the carrier, the phase of the carrier, or the combined amplitude and
phase. Digital information can also be transmitted by modulating the frequency of the
carrier.
As we will observe from our treatment below, digital transmission by frequency
modulation is a modulation method that is appropriate for channels that lack the phase
stability that is necessary to perform carrier-phase estimation. In contrast, the lin
ear modulation methods that we have introduced-namely, PAM, coherent PSK, and
QAM-require the estimation of the carrier phase to perform phase-coherent detection.

7.5.1 Frequency-Shift Keying

The simplest form of frequency modulation is binary frequency-shift keying (FSK).

In binary FSK we employ two different frequencies-say, f1 and f2 = f1 + !).j-to

transmit a binary information sequence. The choice of frequency separation (j.j =

fz -Ji is considered later. Thus the two signal waveforms may be expressed as

uz (t)= \jrm
T; cos2rrfzt, (7.5.1)

where Eb is the signal energy per bit and Tb is the duration of the bit interval.
More generally, M-ary FSK may be used to transmit a block of k =l o g 2M bits per
signal waveform. In this case, the M signal waveforms may be expressed as

Um(t)= \J{2E;
r cos(2rrfct+2rrm/j. f t), m=O,l,... ,M-1, 0 � t� T
(7.5.2)
where Es = kEb is the energy per symbol, T = kTb is the symbol interval, and fj.f is
the frequency separation between successive frequencies-that is, fj.f = fm- fm-1
for all m= 1, 2,... ,M -1, where fm= fc+ mfj.f.
Note that the M FSK waveforms have equal energy, Es. The frequency separation
(j.j determines the degree to which we can discriminate among the M possible trans
mitted signals. As a measure of the similarity (or dissimilarity) between a pair of signal
waveforms, we use the correlation coefficient :Ymn:

1 I
T
Ymn= Um(t)Un(t)dt (7.5.3)
Es Jo

By substituting for Um(t) and Um(t) in (7.5.3), we obtain

1 I
T 2E
Y mn
=Es Jo
/ cos(2rrfct+ 2rrmfj.jt)cos(2rrfct+ 2rrn(j.jt)dt
1 I
T 1 (
= cos2rr(m- n)fj.jtdt+ cos[4rrfct+2rr(m+ n)fj.jt]dt
T J T J
o o
sin 2rr( m- n)fj.f T
(7.5.4)
2rr( m- n)fj.f T

where the second integral vanishes when fc » 1IT. A plot of Ymnas a function of the
frequency separation fj.f is given in Figure 7.28. We observe that the signal waveforms
are orthogonal when fj.f is a multiple of l/2T. Hence, the minimum frequency sepa
ration between successive frequencies for orthogonality is 1I2T. We also note that the
minimum value of the correlation coefficient is Ymn=-0.217, which occurs at the
frequency separation fj.f=0.715 IT.

�-1-��--��fl--�----'._��· �t
0

Figure 7.28: Cross-correlation coefficient as a function of frequency separation for

FSK signals

M-ary orthogonal FSK waveforms have a geometric representation as M M-dimen

sional orthogonal vectors, given as

so= (.JE:;",o ,... ,o) (7.5.5)

s1 = (O,.Jf'.;,O,... ,O) (7.5.6)

(7.5.7)
SM-I= (0,0,... ,0,.JE:;") (7.5.8)

where the basis functions are t/Jm (t) = ../2/Tcos2rr(jc + mtl.f)t. The distance
between pairs of signal vectors is d = ,f[E; for all m, n, which is also the minimum
distance among the M signals. Note that these signals are equivalent to the M-ary
baseband orthogonal signals described in Section 5.4.
The demodulation and detection of the M-ary FSK signals are considered next.

7.5.2 Demodulation and Detection of FSK Signals

Let us assume that the FSK signals are transmitted through an additive white Gaussian
noise channel. Furthermore, we assume that each signal is delayed in the transmis
sion through the channel. Consequently, the filtered received signal at the input to the
demodulator may be expressed as

r(t) = \j{2E;
T
cos(2rrfct + 2rrmtl.f t + c/>m) + n(t) (7.5.9)

where cf>m denotes the phase shift of the mth signal (due to the transmission delay)
and n(t) represents the additive bandpass noise, which may be expressed as

n(t) = nc (t) cos2rrfct - ns (t) sin 2rrfct (7.5.10)

The demodulation and detection of the M FSK signals may be accomplished by

one of two methods. One approach is to estimate the M carrier-phase shifts { cf>m} and
perform phase-coherent demodulation and detection. As an alternative method, the
carrier phases may be ignored in the demodulation and detection.

f�()dt

f�()dt
Received Output
signal Detector
decision

f�()dt
PLLM cos (21tfct + 2rt(M - 1),1':,.ft �M)
+

Figure 7.29: Phase-coherent demodulation of M-ary FSK signals

In phase-coherent demodulation, the received signal r (t) is correlated with each of

the M possi�le received signals, cos(2rrfct+2rrm !j.f t+cPm), form 0, 1, ... , M =

l, where {c/:>m} are the carrier-phase estimates. A block diagram illustrating this type
of demodulation is shown in Figure 7 .29. It is interesting to note that when cPm !- cf:>m
form 0, l, ... , M -1 (imperfect phase estimates), the frequency separation required
=

for signal orthogonality at the demodulator is !j.f 1/T, which is twice the minimum =

separation for orthogonality when cf:> Cb. =

The requirement of estimating M carrier phases makes coherent demodulation of

FSK signals extremely complex and impractical, especially when the number of signals
is large. Therefore, we shall not consider coherent detection of FSK signals.
Instead, we consider a method for demodulation and detection that does not require
knowledge of the carrier phases. The demodulation may be accomplished as shown
in Figure 7 .30. In this case, there are two correlators per signal waveform, or a total
of 2M correlators, in general. The received signal is correlated with the basis func
tions (quadrature carriers) ,J2/T cos(2rrfct + 2rrmfj.f t) and ,J2/Tsin(2rrfct +
2rrm !j.f t), for m 0, 1, ... , M - 1. The 2M outputs of the correlators are sampled
=

at the end of the signal interval and are passed to the detector. Thus, if the mth signal
is transmitted, the 2M samples at the detector may be expressed as
[ sin2rr(k-m)!j.fT coscf:>m- cos2rr(k-m)!j.fT- smcf:>m
.
J
rr l
rkc= vr.s +nkc
2rr(k-m)!j.fT 2rr(k-m)!j.fT
[ cos2rr(k-m)fj.fT-1 coscf:>m sin2rr(k-m)!j.fT smcf:>m
.
rks = ./Es 2rr(k m)!j.fT _ 2rr(k m)!j.fT J (7 +
_
+nks

. 5 11)
.

where nkc and nks denote the Gaussian noise components in the sampled outputs.
We observe that when k m, the sampled values to the detector are
=

rmc = ffs COS cf:>m + nmc

Yms = ffs sincf:>m + nms (7 .5.12)

V2i'f COS 21tfct

x f�()dt rlc
Sample at t = T
V2ff sin 2rtfJ

f�()dt rls
Sample at t = T
V2i'f COS 2rt<fc l'.f)t
+

x f�()dt r2c
Sample at t = T
V2ff sin 2rt<fc + l'.f)t

Received f�()dt r2s Output

signal Sample at t = T Detector
decision

V2ff cos 2rt[fc + (M - 1)1'.flt

f�()dt
Sample at t = T
V2ff sin 2n[fc + (M -1)1'.flt

�--�X>----� f�()dt
Sample at t = T

Figure 7.30: Demodulation of M-ary signals for noncoherent detection

Furthermore, we observe that when k !- m, the signal components in the samples rkc
and rks will vanish, independent of the values of the phase shift cf>k, provided that the
frequency separation between successive frequencies is tlf l/T. In such a case, the =

other 2 (M - 1) correlator outputs consist of noise only-that is,

k !- m (7.5.13)

In the following development we assume that tlj 1/ T, so that the signals are or
thogonal.
It can be shown that the 2M noise samples { nkc} and { nks} are zero-mean, mu
tually uncorrelated Gaussian random variables with equal variance cr2 = No/2. Con
sequently, the joint probability density function for rmc and rms conditioned on cf>m
is

__l_e-[(rmc-JE,coscf>ml2+(rm,-JE,smcf>ml2]1 a2
frm (rme. rms I ,.1..
'1-' m ) 2 (7 5 14)
Zrrcr2
= · ·

and form !- k, we have

l
frk (rkc , rks ) ___e-(rfc+rfs)f2a2 (7.5.15)
Zrrcr2
=

Given the 2M observed random variables { rkc, rks} f=-01 , the optimum detector se
lects the signal that corresponds to the maximum of the posterior probabilities- that is,

P [Sm was transmitted Ir] = P(sm Ir), m = 0,1,... ,M 1 - (7 .5.16)

1
where r is the 2M-dimensional vector with elements {rkc,rks}f=-0 . W hen the signals
are equally probable, the optimum detector specified by (7.5.16) computes the signal
envelopes,defined as

m = 0,1,... ,M 1 - (7.5.17)

and selects the signal corresponding to the largest envelope of the set {rm}. In this
case the optimum detector is called an envelope detector.
An equivalent detector is one that computes the squared envelopes

m = 0,1,... ,M 1 - (7.5.18)

and selects the signal corresponding to the largest, { rk} . In this case, the optimum
detector is called a square-law detector.

----tllli-il;@IW§l§;t•1:Jl#@I
Illustrative Problem 7 .11 [FSK Signaling] Consider a binary communication system
that employs the two FSK signal waveforms given as

ui(t) =cos2rrfit, 0 � t �Tb

Uz(t) =cos2rrfzt, 0 � t �Tb
where f1 = 1000/Tb and fz = Ji + l/Tb. The channel imparts a phase shift of
cf> = 45° on each of the transmitted signals, so that the received signal in the absence
of noise is

r(t) =cos ( 2rrfit + �) , i= 1,2,

Numerically implement the correlation-type demodulator for the FSK signals.

We sample the received signal r(t) at a rate F5 = 5000/Tb in the bit interval Tb. Thus
r(t) is represented by the 5000 samples {r(n/F5)}. The correla
the received signal
tion demodulator multiplies {r(n IF5)} by the sampled version of u1 (t) =cos2rrf1t,
vi(t) = sin2rrf1t,u2(t) =cos2rrf2t,andv2(t) = sin2rrfzt,as illustrated inFig
ure 7 .30. Thus the correlator outputs are

k=l,2,... ,5000

k= 1,2,... ' 5000

k = 1, 2, ... ' 5000

k = 1, 2' ... ' 5000

The detector is a square-law detector that computes the two decision variables

r1 = Yfc(5000) + Yfs(5000)
r2 = r}c(5000) + r}s(5000)

and selects the information bit corresponding to the larger decision variable.
A MATLAB program that implements the correlations numerically is given next.
The graphs of the correlator outputs are shown in Figure 7.31, based on the signal
U1 (t) being transmitted.

1500

1000

500

-500

-1000

-1500

-2000 �-�-�-�-
-�-�-�-�-��-
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
m

Figure 7.31: Output of correlators for binary FSK demodulation

----41&1
1 1" ------
% MATLAB script for Illustrative Problem 7.11.
echo on
Tb=1;
f1=1OOOffb;
f2=fl +1ffb;
phi=pi/4;
N=5000; % number of samples
t=O:Tb/(N-1):Tb;
ul=cos(2*pi*fl *t);
u2=cos(2*pi*f2*t);

% Assuming that ul is transmitted, the received signal r is

sgma=1; % noise variance
for i=1:N,
r(i)=cos(2*pi*fl*t(i)+phi}+gngauss(sgma);
echo off;
end;
echo on;
% The correlator outputs are computed next.
vl=sin(2*pi*fl*t);
v2=sin(2*pi*f2*t);
rlc(1}=r(1)*u1(1);
rls(1}=r(1}*vl(1);
r2c(1}=r(1)*u2(1);
r2s(1}=r(1)*v2(1);
for k=2:N,
rlc(k}=rlc(k-1}+r(k}*ul(k};
rls(k}=rls(k-1}+r(k}*vl(k};
r2c(k}=r2c(k-1}+r(k}*u2(k};
r2s(k}=r2s(k-1)+r(k}*v2(k};
echo off;
end;
echo on;
% decision variables
rl=rlc(5000} "2+rls(5000}"2;
r2=r2c(5000)"2+r2s(5000)"2;
% Plotting commands follow.

753 . . Probability of Error for Noncoherent Detection of FSK

The derivation of the performance of the optimum envelope detector for the M -ary
FSK signals can be found in most texts on digital communication. The probability of a
symbol error may be expressed as

PM=
M-1
I (-l)n+l
( M -1 ) __l e-nkEb/No(n+l) (7.5.19)
n n + 1
n=l

W hen M = 2, this expression reduces to the probability of error for binary FSK, which
is

(7.5.20)

For M > 2, the bit-error probability may be obtained from the symbol-error probability
by means of the relation
k
2 -1
Pb= PM (7.5.21)
2k - l

The bit-error probability is plotted in Figure 7.32 as a function of the SNR per bit

......
for M = 2, 4, 8, 16, 32. We observe that for any given bit-error probability, the SNR per
bit decreases as M increases. In the limit as M oo, the error probability can be made
arbitrarily small provided that the SNR per bit exceeds -1.6 dB. This is the channel

10-1
5

2
10-2
.... 5

�
�
2
iiS 10-3
o:I
""' 5
0
c
� 2
� 10-
4
..0
0
� 5
Q..,:-.,
2
S
10- M==
5 Channel Capacity
: Limit (-1.6 dB)
2 I
10-6
-2 0 2 4 6 8 10 12 14
SNR/Bit, dB

Figure 7.32: Probability of a bit error for noncoherent detection of orthogonal FSK
signals

capacity limit, or Shannon limit, for any digital communication system transmitting
information through an AWGN channel.
The cost of increasing M is the bandwidth required to transmit the signals. Be
cause the frequency separation between adjacent frequencies is !:l.f = l/T for signal
orthogonality, the bandwidth required for the M signals is W MIT. The bit rate is =

R = k/T , where k = log2M. Therefore, the bit rate-to-bandwidth ratio is

R log2M
(7.5.22)
w M
We observe that RI W --+ 0 as M --+ oo.

-tiMQi.ii;tfjiW§liflj;t•]:JM§j®I
Illustrative Problem 7 .12 [Binary FSK Simulation] Perform a Monte Carlo simula
tion of a binary FSK communication system in which the signal waveforms are given
by (7.5.1), where fz = Ji + 1 ITb and the detector is a square-law detector. The block
diagram of the binary FSK system to be simulated is shown in Figure 7 .33.

Since the signals are orthogonal, when U1 (t) is transmitted, the first demodulator out
put is

Yic = ./Eb cos</>+ n1c

r1s = ./Eb sin</>+ nIs

Uniform Gaussian Gaussian

RNG RNG RNG

rlc r lc 2

'is r ls
FSK +
Signal Detector Output Bit
r2 c 2
Selector

r2s r 2s

Gaussian Gaussian
RNG RNG

Compare

Bit-Error
Counter

Figure 7 .33: Block diagram of a binary FSK system for the Monte Carlo simulation

and the second demodulator output is

r2c = n2c

r2s = n2s

where n1c, n15, n2c, and n2s are mutually statistically independent, zero-mean Gaus
2
sian random variables with variance <J" and cf> represents the channel-phase shift.
In the above expression, the channel-phase shift cf> may be set to zero for conve
nience. The square-law detector computes

and selects the information bit corresponding to the larger of these two decision vari
ables. An error counter measures the error rate by comparing the transmitted sequence
to the output of the detector.

------
The MATLAB programs that implement the Monte Carlo simulation are given next.
Figure 7 .34 illustrates the measured error rate and compares it with the theoretical error
probability given by (7 .5 .20).

ii!
----@I
% MATLAB script for Illustrative Problem 7.12.
echo on
SNRindB 1=0:2:15;
SNRindB2=0:0.1 :15;

for i=1 :length(SNRindBl),

smld_err_prb(i)=cm_sm52(SNRindB 1 (i)); % simulated error rate
echo off ;
end;
echo on ;
for i=1 :length(SNRindB2),
SNR=exp(SNRindB2(i)*log(10)/1O); % signal-to-noise ratio
theo_err_prb(i)=(1 /2)*exp(-SNR/2); % theoretical symbol error rate
echo off;
end;
echo on;
% Plotting commands follow.
semilogy(SNRindBl,smld_err_prb,' * ');
hold
semilogy(SNRindB2,theo_err_prb);

100

10-1

10-2

10-3

10-4

10-5

10-7
0 2

Figure 7 .34: Performance of a binary FSK system from the Monte Carlo simulation

----tl®li" ------
function [p]=cm_sm52(snr_in_dB)
% [p]=cm-Ym52(snr_in..dB)
% CM..SM52 Returns the probability of error for the given
% value of snr_in..dB, signal-to-noise ratio in dB.
N=10000;
Eb=1;
d=1;
snr=1OA(snr_in_dB/1O); % signal-to-noise ratio per bit
sgma=sqrt(Eb/(2*snr)); % noise variance
phi=O;
% Generation of the data source follow s.

for i=1 :N,

temp=rand; % a uniform random variable between 0 and 1
if (temp<0.5),
dsource(i)=O;
else
dsource(i)=1;
end;
end;
% detection and the probability of error calculation
numoferr=O;
for i=1 :N,
% demodulator output
if (dsource(i)==O),
rOc=sqrt(Eb)*cos(phi)+gngauss(sgma);
rOs=sqrt(Eb)*sin(phi)+gngauss(sgma);
rlc=gngauss(sgma);
rls=gngauss(sgma);
else
rOc=gngauss(sgma);
rOs=gngauss(sgma);
rlc=sqrt(Eb)*cos(phi)+gngauss(sgma);
rls=sqrt(Eb)*sin(phi)+gngauss(sgma);
end;
% square-law detector outputs
rO=r0c"2+rOs"2;
rl=rlc"2+rls"2;
% Decision is made next.
if (rO>rl),
decis=O;
else
decis=1;
end;
% If the decision is not correct the error counter is increased.
if (decis-=dsource(i)),
numoferr=numoferr+1;
end;
end;
p=numoferr/(N);

7 .6 Synchronization in Communication Systems

In Section 3 .3 we described the demodulation process for AM signals. In particular, we

showed that we can classify the demodulation schemes into coherent and noncoherent
schemes. In coherent demodulation, the AM signal is multiplied by a sinusoid with
the same frequency and phase of the carrier and then demodulated. In noncoherent
demodulation, which can be applied only to the conventional AM scheme, envelope
demodulation is employed and there is no need for precise tracking of the phase and
the frequency of the carrier at the receiver. We furthermore showed in Illustrative
Problem 3 .6 that correct phase synchronization in coherent demodulation is essential
and phase errors can result in considerable performance degradation.

In this chapter we discussed the demodulation schemes for digital carrier-modulated

systems. In the demodulation of PAM, PSK, and QAM, we assumed that we have com
plete knowledge of the carrier frequency and phase.
In this section we discuss methods to generate sinusoids with the same frequency
and phase of the carrier at the demodulator. These methods are studied under the title
of carrier synchronization and apply to both analog and digital carrier modulation sys
tems discussed in this chapter and Chapter 3. Another type of synchronization, called
timing synchronization, clock synchronization , or timing recovery, is encountered only
in digital communication systems. We briefly discuss this type of synchronization
problem in this section as well.

761 . . Carrier Synchronization

A carrier synchronization system consists of a local oscillator whose phase is controlled

to be in synch with the carrier signal. This is achieved by employing a PLL. A phase
locked loop is a nonlinear feedback-control system for controlling the phase of the
local oscillator. In the following discussion, for simplicity we consider only binary
PSK modulation systems.
The PLL is driven by a sinusoidal signal at carrier frequency (or a multiple of it).
In order to obtain the sinusoidal signal to drive the PLL, the DSB-modulated signal

u(t) = Acm(t) cos(2rrfct - </>(t)) (7 .6.1)

where m(t) = ±1, is squared to obtain

u2(t) = A�m2(t) cos2(2rrfct - </>(t))

A2 A2
= -f m2(t) + -f m2(t) cos(4rrfct - 2</>(t))
A2 A2
=

T T cos(4TTfct - 2</>(t))
+ (7 .6.2)

Obviously, this signal has a component at 2fc· The reason that we do not deal directly
with u(t) is that usually the process m(t) is zero-mean, so the power content of u(t)
at fc is zero. Now, if the signal u2(t) is passed through a bandpass filter tuned to
2fc, the output will be a sinusoidal signal with the central frequency 2fc, a phase of
-2</>(t), and an amplitude of A� H (2fc) I 2. Without loss of generality we can assume
that the amplitude is unity; that is, the input to the PLL is

r(t) = cos(4rrfct - 2</>(t)) (7 .6.3)

The PLL consists of a multiplier, a loop filter, and a voltage-controlled oscilla

tor (VCO), as shown in Figure 17 .35l If we assume that the output of the VCO is
sin(4TTfct - 2(/>(t)) , then at the input of the loop filter we have
e(t) = cos(4rrfct - 2</>(t)) sin(4rrfct - 2(/>(t))
1 1
2 sin(2</>(t) - 2</>(t)) 2 sin(8rrfct - 2</>(t) - 2</>(t))
� �

= + (7 .6.4)

Input Signal e(t)

x Loop Filter
r(t)

sin(4rrfct - 2</>)

v (t)
vco

Figure 7.35: The phase-locked loop

Note that e(t) has a high- and a low-frequency component. The role of the loop filter
is to remove the high-frequency component and to make sure that (b(t) follows closely
the variations in </>(t). A simple loop filter is a first-order lowpass filter with a transfer
function of
1 + T1s
G(s) = (7.6.5)
1 + T25

where T2 » T1. If we denote the input of the VCO as v (t), then the output of the VCO
will be a sinusoid whose instantaneous frequency deviation from 2fc is proportional to
v (t). But the instantaneous frequency of the VCO output is
1 d �

2f.c + - -</>(t)
rrdt
Therefore,

(7.6.6)

or, equivalently,

2(b(t) = K f 00 V ( T) d T (7.6.7)

where K is some proportionality constant. After removal of the second and fourth
harmonics, the PLL reduces to the one shown in Figure 7.36.
Assuming that (b(t) closely follows changes in </>(t), the difference 2</>(t) - 2(b(t)
is very small, and we can use the approximation

� sin (2</>(t) - 2</>(t)) � </>(t) - </>(t) (7.6.8)

With this approximation the only nonlinear component in Figure 7.36 is replaced by a
linear component, resulting in the linearized PLL model shown in Figure 7.37. Note
that this model is represented in the transform domain, and the integrator is replaced
by its transform domain equivalent, 1 / s.

+
2</>(t) - + -
� sin(2<f>(t) - 2<f>(t)) � G(s) -

2</> (t)

vco v(t)
2cP(t) =Kf�00V(T)dT

Figure 7.36: The phase-locked loop after removal of high-frequency components

+
</>(t) + - G(s)
- �

4>(t)

K v(t)
s

Figure 7.37: The linearized model for a phase-locked loop

The model shown in Figure 7.37 is a linear control system with a forward gain of
G(s) and a feedback gain of K/s; therefore, the transfer function of the system is given
by
ci>(s) KG(s)/s
H(s) = = (7.6.9)
(s) 1 + KG(s)/s

and with the first-order model for G(s) assumed before,

1 + T1S
G(s) = (7.6.10)
1 + T2S
H (s) is given as

(7.6.11)

With H(s) as the transfer function, if the input to the PLL is (s), the error will be

�(s) = (s) - ci>(s)

= (s) - (s)H(s)
= [l - H(s)] (s)
(1 + T2S)S
<l>(s) (7.6.12)
K + (1 + KT1)s + T2s2

Now let us assume that up to a certain time cp(t) ;::::: 4><t), so Licp(t) ;::::: 0. At this
time some abrupt change causes a jump in cp(t) that can be modeled as a step-that is,
(s)=K1 Is. With this change we have
(1+Tzs)s K1
ii(s)=
2
K+(1 +KT1 )s+Tz5 s
K1 (1 +Tzs)
(7 .6.13)

Now, by using the final value theorem of the Laplace transform, which states that

limj(t)=IimsF(s) (7.6.14)
t-oo s-0

as long as all poles of sF(s) have negative real parts, we conclude that

lim iicp(t)=lims(s)
t-oo s-0

K1 s(l+Tzs)
=lim
s-o K+(1 +KTi)s+Tzs2
=0 (7.6.15)

In other words, a simple first-order loop filter results in a PLL that can track jumps in
the input phase.
The transfer function (7.6.11) can be written in the standard form

(2(wn - w�/K)s+w�
H(s)= (7.6.16)
2
S 2 +2(WnS +Wn
Here,

\j{K
Wn=
Tz
Wn(T1 +l/K)
S=
2
where Wn is the natural frequency and ( is the damping factor.

Illustrative Problem 7 .13 [First-Order PLL] Assuming that

1 + O.Ols
G(s)=
l+s
and K=1, determine and plot the response of the PLL to an abrupt change of height 1
to the input phase.

Here T1 =0.01 and Tz=1; therefore,

Wn=1
( =0.505

which results in

H(5) =
052 .1.001515 1 1
+
+

Thus the response to cp(t)= u(t)-that is, <1>(5) 1/ = 5-is given by

(5) =
53 + 1.00.011525 1 1 +

+ 5 +

In order to determine and plot the time response ¢ (t) to the input u(t), we note that
we have to determine the output of a system with transfer function H (5) to the input
u(t). This can be done most easily by using state-space techniques. We will employ the
MATLAB function tf2ss.m, which returns the state-space model of a system described
by its transfer function. After determining the state-space representation of the system,
we obtain the step response numerically.
The function tf2ss.m takes the numerator and the denominator of the transfer func
tion H (5) and returns A, B, C, and D, its state-space representation, in the form

{: t
x(t) = Ax(t) + Bu(t)

y(t)= Cx(t) + Du(t)

This representation can be approximated by

{ x(t + Llt)= x(t)

y(t)= Cx(t)
+

+
Ax(t) Llt
Du(t)
+ Bu(t) 6.t

or, equivalently,

{ 1)
x(i + = x(i)
y(i)= Cx(i)
+ Ax(i) 6.t
+ Du(i)
+ Bu(i) Llt

For this problem it is sufficient to choose u(t) to be a step function and the nu
merator and the denominator vectors of H(5) [0.01 1 to be ] and [l 1.01 l],
respectively. With this choice of numerator and denominator vectors, the state-space
parameters of the system will be

A= [-1.01 1 -01]
B �
[�]
c= [ 0.01 ] l

D=O

The plot of the output of the PLL is given in Figure 7 .38.

Figure 7.38: The response of the PLL to an abrupt change in the input phase in Illus
trative Problem 7.13

As we can see from Figure 7.38, the output of the PLL eventually follows the input;
however, the speed by which it follows the input depends on the loop filter parameters
and K, the VCO proportionality constant.

The MATLAB script for this problem is given next.

__.,,... ·-----
% MATLAB script for Illustrative Problem 7.13.
echo on
num=[0.01 1];
den=[1 1.01 1];
[a,b,c,d)=tf2ss(num,den);
dt=0.01;
u=ones(1,2000);
x=zeros(2,2001);
for i=1:2000
x(:,i+1)=x(:,i)+dt.*a*x(:,i)+dt.*b*u(i);
y(i)=c*x(:,i);
echo off;
end
echo on;
t=[O:dt:20];
plot(t(1:2000),y)

762 . . Clock Synchronization

In Chapter 5 and in this chapter we have seen that a popular implementation of the
optimal receiver makes use of matched filters and samplers at the matched filter out
put. In all these cases we assumed that the receiver has complete knowledge of the
sampling instant and can sample perfectly at this time. Systems that achieve this type
of synchronization between the transmitter and the receiver are called timing recovery,
clock-synchronization, or symbol-synchronization systems.

A simple implementation of clock synchronization employs an early-late gate.

The operation of an early-late gate is based on the fact that in a PAM communication
system the output of the matched filter is the autocorrelation function of the basic pulse
signal used in the PAM system (possibly with some shift). The autocorrelation function
is maximized at the optimum sampling time and is symmetric. This means that, in the
absence of noise, at sampling times T+ = T + 6 and T- = T - 6, the outputs of the
sampler will be equal- that is,

(7.6.17)

In this case, the optimum sampling time is obviously the midpoint between the early
and late sampling times:

r+ + r
T = ---- (7.6.18)
2

Now let us assume that we are not sampling at the optimal sampling time T, but instead
we are sampling at T1. If we take two extra samples at r+ = T1 + 6 and r- = T1 - 6,
these samples are not symmetric with respect to the optimum sampling time T and,
therefore, will not be equal. A typical autocorrelation function for positive and negative
incoming pulses and the three samples is shown in Figure 7 .39.

Matched Filter Output

Optimum Sample
2214752 2014/07/21 173.244.197.106
Early Sample
Late Sample

t r- T r+ t

T- T r+

Figure 7 .39: The matched filter output and early and late samples

Here

T- = T- 81
y+ = T + 82

where
(7.6.19)

and, as the figure shows, this results in

(7.6.20)

Also, in this case

y- + y+
T < T1 (7.6.21)
2
=

Therefore, when ly(T-) I > ly(T+) I, the correct sampling time is before the as
sumed sampling time, and the sampling should be done earlier. Conversely, when
Iy(T-) I Iy(T+) I, the sampling time should be delayed. Obviously, when
<

Iy ( y-) I Iy ( y+) I, the sampling time is correct and no correction is necessary.

The early-late gate synchronization system therefore takes three samples at T1,
y- T1- 8, and y+
= T1 + 8 and then compares Iy(T-) I and Iy(T+) I and,
=

depending on their relative values, generates a signal to correct the sampling time.

Illustrative Problem 7 .14 [Clock Synchronization] A binary PAM communication

systems uses a raised-cosine waveform with a rolloff factor of 0.4. The system trans
mission rate is 4800 bits/s. Write a MATLAB file that simulates the operation of an
early-late gate for this system.

Because the rate is 4800 bits/s, we have

1
T = (7.6.22)
4800

and with oc = 0.4, the expression for a raised-cosine waveform becomes

. cos(4800 x 0.4rrt)
x (t) = smc (4BOOt)
1 - 4 x 0.16 x 48002tZ

. cos l 920rrt
smc(4800t) (7.6.23)
1 - 1. 4746 l07tZ
=

This signal obviously extends from -oo to + oo. The plot of this signal is given in
Figure 7.40.
From Figure 7.40 it is clear that, for all practical purposes, it is sufficient to consider
only the interval It I .::; 0.6 x 10-3, which is roughly [-3T, 3T]. Truncating the raised
cosine pulse to this interval and computing the autocorrelation function result in the
waveform shown in Figure 7.41.

0.6

0.4

0.2

-3
-1 --0.8 0.8 1x10

-0.2

Figure 7.40: The raised-cosine signal in Illustrative Problem 7.14

--3
-1 x 10

Figure 7.41: The autocorrelation function of the raised-cosine signal

In the MATLAB script given next, the raised-cosine signal and the autocorrelation
function are first computed and plotted. In this particular example the length of the
autocorrelation function is 1201 and the maximum (i.e., the optimum sampling time)
occurs at the 600th component. Two cases are examined: one when the incorrect

sampling time is 700 and one when it is 500. In both cases the early-late gate corrects
the sampling time to the optimum 600.

__.,.... ·-----
% MATLAB script for Illustrative Problem 7.14.
echo on
alpha=0.4;
T=1/4800;
t=[-3*T:1.001*T/100:3*T];
x=sinc(t.tr). *(cos(pi*alpha*t.ff)./(1-4*alpha A2*t. A 2/f A2));
pause % Press any key to see a plot of x(t).
plot(t,x)
y=xcorr(x);
ty=[t-3*T,t(2:length(t))+3*T];
pause % Press any key to see a plot of the autocorrelation of x(t).
plot(ty,y);
d=60; % Early and late advance and delay
ee=0.01; % Precision
e=1; % Step size
n=700; % The incorrect sampling time
while abs(abs(y(n+d))-abs(y(n-d)))>=ee
if abs(y(n+d))-abs(y(n-d))>O
n=n+e;
elseif abs(y(n+d))-abs(y(n-d))<O
n=n-e;
end
echo off
end
echo on ;
pause % Press any key to see the corrected sampling time
n
n=500; % Another incorrect sampling time
while abs(abs(y(n+d))-abs(y(n-d)))>=ee
if abs(y(n+d))-abs(y(n-d))>O
n=n+e;
elseif abs(y(n+d))-abs(y(n-d))<O
n=n-e;
end
echo off
end
echo on ;
pause % Press any key to see the corrected sampling time
n

Problems

7 .1 In a carrier-amplitude-modulated PAM system, the transmitter filter has a square

root raised-cosine spectral characteristic with rolloff factor ex = 0.5. The carrier fre
quency is fc = 40/ T. Evaluate and graph the spectrum of the baseband signal and the
amplitude-modulated PAM signal.

7 .2 Repeat Problem 7.1 when the carrier frequency is fc = 80 IT.

7 .3 Repeat Problem 7.1 when the transmitter has a square-root duobinary spectral char
acteristic.

7.4 The purpose of this problem is to demonstrate that (7.2.9) and (7.2.10) hold by
evaluating (7.2.9) numerically using MATLAB. The pulse Br(t) may be assumed to
be rectangular-that is,

{1, 0 �t�2
Br(t)
0,
=

otherwise

Let the carrier frequency fc

2000 Hz. Use a sampling rate of F5 20,000 samples
= =

per second on the signal waveform ljJ(t) given by (7.2.6) and compute the energy of
1fJ (t) by approximating the integral in (7.2 .8) by the summation

1 N-1
, 2
L. l/J (nTs)
1 N-1
,
L. l/J
2 (n)
F5
=

N N
n=O n=O

where N = 40,000 samples. Write a MATLAB program to generate the samples

1fJ (n/ F5) and perform the computation of the signal energy as described above.

7.5 The cross-correlation of the received signal r(t) with lfJ (t) as given by (7.2.14)
may be performed numerically using MATLAB. Write a MATLAB program that com
putes the correlator output

n = 0, 1, ... , N - 1
where F5 is the sampling frequency. Evaluate and graph y(n) when r(t) ljJ(t),
where ljJ(t) is the waveform described in Problem 7.4 and F5 20,000 Hz. =

7.6 Evaluate and graph the correlation {y (n)} in Problem 7.5 when the signal Br(t)
is

Br(t)
{� (1 - cos rrt), O�t�2
0,
=

otherwise

for the same parameters given in Problem 7.4.

7.7 In Illustrative Problem 7.3, the eight PSK waveforms had a constant amplitude.
Instead of the rectangular pulse BT (t), suppose that the signal pulse shape is

gy(t)=
{� (1 - cos2rrt/T), 05't5'T
0, otherwise

Write a MATLAB program to compute and graph the M = 8-PSK signal waveforms
for the case in which Jc=6 IT.

7.8 Write a MATLAB program that numerically computes the cross-correlation of the
received signal r(t) for a PSK signal with the two basis functions given by (7.3.9).
That is, compute

Yc(n) = � (�) (�),

k O
r qi1 n=0,1,... ,N - 1

Ys( n)= � (�) (�),

k O
r qiz n=0,1,... ,N - 1

where N is the number of samples of r(t), qi1 (t), and qiz (t). Evaluate and plot these
correlation sequences when

{
r(t)=SmcqlI (t) + Smsqlz (t)
2, 0 5, t 5, 2
gy(t)=
0, otherwise

Jc= 1000 Hz, F5 = 10000 samples per second, and the transmitted signal point is as
given.

a. Sm=( Smc . Sms) =(1,0)

b.Sm=(-1,0)

C. Sm=(0,1)

7.9 Write a MATLAB program that performs a Monte Carlo simulation of an M= 4-

PSK communication system, as described in Illustrative Problem 7 .5, but modify the
detector so that is computes the received signal phase er as given by (7.3. 1 6) and
selects the signal point whose phase is closes to er.

7.10 Write a MATLAB program that implements a differential encoder and a differ
ential decoder for a M = 4-DPSK system. Check the operation of the encoder and
decoder by passing a sequence of 2-bit symbols through the cascade of the encoder
and decoder and verify that the output sequence is identical to the input sequence.

7.11 Write a MATLAB program that performs a Monte Carlo simulation of a binary
DPSK communication system. In this case, the transmitted signal phases are e 0 =

and e 180°. A e 0 phase change corresponds to the transmission of a 0. A

= =

e 180° phase change corresponds to the transmission of a 1. Perform the simulation

for N 10000 bits at different values of the SNR parameter Eb! No. It is convenient to
=

normalize Eb to unity. Then, with a-2 No/2, the SNR is Eb/No l/2 a-2 , where a-2
= =

is the variance of the additive noise component. Hence, the SNR can be controlled by
scaling the variance of the additive noise component. Plot and compare the measured
error rate of the binary DPSK with the theoretical error probability given by (7.3 23) . .

7.12Write a MATLAB program that generates and graphs the M 8-QAM signal =

waveforms given by (7.4.2) for the signal constellation shown in Figure P7.12.

(-3, 1) (3, 1)

(-3, -1) (3, -1)

Figure P7.12

Assume that the pulse waveform BT (t) is rectangular-that is,

{l, 0 � t � T
BT (t) =

0, otherwise
and the carrier frequency is Jc = 8 IT.

7.13 Repeat Problem 7.12 when the pulse waveform BT (t) is given as

{�(1- cos2rrt/T), 0 � t � T
BT (t) =

0, otherwise

7.14 Repeat Illustrative Problem 7.1 when the carrier frequency Jc = 45/T.

7.15 Repeat Illustrative Problem 7.4 when the carrier frequency Jc = 45 IT.

7.16 Repeat Illustrative Problem 7 .9 when the carrier frequency Jc = 50 /T.

7.17 Write a MATLAB program that performs a Monte Carlo simulation of an M =

8-QAM communication system for the signal constellation shown in Figure P7.12.
Perform the simulation for N = 10, 000 (3-bit) symbols at different values of the SNR
parameter Eavb I No. It is convenient to normalize Eavb to unity. Then, with a-2 = No I 2,
the SNR is Eavb I No = 1I2 a-2, where a-2 is the variance of each of the two additive
noise components. Plot and compare the measured symbol-error rate of the QAM
system with the upper bound on the theoretical error probability given by ( 7 .4 .12).

7.18 Repeat the simulation in Problem 7.17 for the M = 8-signal constellation shown
in Figure P7.18. Compare the error probabilities for the two M = 8-QAM signal
constellations and indicate which constellation gives the better performance.

(l + '13, 0)

Figure P7.18

7.19 Consider the binary FSK signals of the form

ui(t) = .j¥f cos2rrf1t,

u2(t) = .j¥f cos2rrf2t,

1
f2 =fl+
2Tb

Let !1 = 1000 I Tb. By sampling the two waveforms at the rate F5 = 5000 I Tb , we
obtain 5000 samples in the bit interval 0 t :::;; Tb. Write a MATLAB program
:::;;

that generates the 5000 samples for each of u 1(t) and u2(t) and compute the cross
correlation

and, thus, verify numerically the orthogonality condition for U1(t) and u2(t).

7 .20 Use the MATLAB program given in Illustrative Problem 7.11 to compute and
graph the correlator outputs when the received signal is

r(t) =cos 2rrf1t + ( ;) ,

7 .21 Use the MATLAB program given in Illustrative Problem 7.11 to compute and
graph the correlator outputs when the transmitted signal is U2 (t) and the received
signal is
r(t) =cos 2rrfzt + ( :) , 0 � t � T

7 .22 Write a MATLAB program that performs a Monte Carlo simulation of a quater
nary (M = 4) FSK communication system and employs the frequencies

k
fk =Ji+ ' k=0,1,2,3
T

The detector is a square-law detector. Perform the simulation for N = 10,000 (2-bit)
symbols at different values of the SNR parameter Eb I No, and record the number of
symbol errors and bit errors. Plot and compare the measured symbol- and bit-error rates
with the theoretical symbol- and bit-error probabilities given by (7.5.19) and (7.5.21).

7 .23 In Illustrative Problem 7.13 it was assumed that the input phase has an abrupt
jump, and the simulation showed that a first-order loop filter can track such a change.
Now assume that the input changes according to a ramp-that is, starts to increase
linearly. Simulate the performance of a first-order PLL in this case and determine
whether the loop is capable of tracking such a change.

7 .24 Repeat Illustrative Problem 7.13 when the loop filter is given as

1
G(s) = 5 J2
+

Also, determine the closed-loop transfer function H(s) and the corresponding natural
frequency and damping factor of the loop. Is the loop stable?

7 .25 Repeat Illustrative Problem 7.14 with a rectangular pulse shape in the presence of
AWGN for SNR values of 20 , 10, 5, and 2 dB.

RuinMasters Classic Fantasy RolePlay
86% (7)
RuinMasters Classic Fantasy RolePlay
281 pages
Littlebearcatbunny
100% (4)
Littlebearcatbunny
7 pages
Principles of Radio Transmitters and Receivers 2
No ratings yet
Principles of Radio Transmitters and Receivers 2
92 pages
Woven Garments Quality Control Procedure
100% (2)
Woven Garments Quality Control Procedure
47 pages
Lecture4 Characterization of Communication Signals and Systems
No ratings yet
Lecture4 Characterization of Communication Signals and Systems
25 pages
Bandpass Modulation & Demodulation: Engr. Ghulam Shabbir
No ratings yet
Bandpass Modulation & Demodulation: Engr. Ghulam Shabbir
114 pages
Ch2 4 DSB QAM
No ratings yet
Ch2 4 DSB QAM
123 pages
Chapter 2 Modulation
No ratings yet
Chapter 2 Modulation
81 pages
Comm Chapter - 4 and 5 Amplitude Modulation and Demodulation
No ratings yet
Comm Chapter - 4 and 5 Amplitude Modulation and Demodulation
69 pages
Lec 04 PCS
No ratings yet
Lec 04 PCS
86 pages
MTG Keywords
No ratings yet
MTG Keywords
4 pages
Ch3 AM Mod Sept2024
No ratings yet
Ch3 AM Mod Sept2024
48 pages
EN1055 Modulation Updated 2
No ratings yet
EN1055 Modulation Updated 2
40 pages
نظرية اتصالات (42 96)
No ratings yet
نظرية اتصالات (42 96)
55 pages
Lecture 03
No ratings yet
Lecture 03
43 pages
Hafta) Presentation 2
No ratings yet
Hafta) Presentation 2
47 pages
Analog Communication Systems
No ratings yet
Analog Communication Systems
42 pages
04 Amplitute - Modulation XX
No ratings yet
04 Amplitute - Modulation XX
61 pages
Week-6 EE-351
No ratings yet
Week-6 EE-351
31 pages
Lecture 2 AM 2024
No ratings yet
Lecture 2 AM 2024
34 pages
Poinsettiacrochetpattern
100% (1)
Poinsettiacrochetpattern
11 pages
3318 Exam Prep
No ratings yet
3318 Exam Prep
42 pages
Communication Engineering-Ece
No ratings yet
Communication Engineering-Ece
59 pages
Lecture 2 - AM
No ratings yet
Lecture 2 - AM
32 pages
Lecture 4
No ratings yet
Lecture 4
64 pages
Normal Amplitude Modulation Time and Frequency Domain Characteristics Lecture Outline
No ratings yet
Normal Amplitude Modulation Time and Frequency Domain Characteristics Lecture Outline
58 pages
Digital Phase Modulation
No ratings yet
Digital Phase Modulation
29 pages
EELE44514 L25-27prelim
No ratings yet
EELE44514 L25-27prelim
24 pages
Communication System CH#3
No ratings yet
Communication System CH#3
27 pages
3nd Week (AM)
No ratings yet
3nd Week (AM)
37 pages
IN 1501 Analog Data Analog Signal
No ratings yet
IN 1501 Analog Data Analog Signal
23 pages
Analog Communication Systems (ECE3001) : Dr. Jeetashee Aparajeeta Assistant Professor (SR) Sense
No ratings yet
Analog Communication Systems (ECE3001) : Dr. Jeetashee Aparajeeta Assistant Professor (SR) Sense
90 pages
Lab 1
No ratings yet
Lab 1
13 pages
Introduction To Communication Systems: Chapter Two
No ratings yet
Introduction To Communication Systems: Chapter Two
48 pages
Lecture 6
No ratings yet
Lecture 6
16 pages
Modulation Part1 PDF
No ratings yet
Modulation Part1 PDF
30 pages
Chapter Five
No ratings yet
Chapter Five
14 pages
Digital Carrier Modulation
No ratings yet
Digital Carrier Modulation
15 pages
Lab 7: Amplitude Modulation and Complex Lowpass Signals: ECEN 4652/5002 Communications Lab Spring 2020
No ratings yet
Lab 7: Amplitude Modulation and Complex Lowpass Signals: ECEN 4652/5002 Communications Lab Spring 2020
28 pages
Module 2 - Modulation and Demodulation - Analogue
No ratings yet
Module 2 - Modulation and Demodulation - Analogue
17 pages
Chapter Two
No ratings yet
Chapter Two
41 pages
What Is A Signal ?
No ratings yet
What Is A Signal ?
62 pages
Digital Modulation Types and Applications: College of Engineering Electrical Engineering Department
No ratings yet
Digital Modulation Types and Applications: College of Engineering Electrical Engineering Department
19 pages
History of Games
No ratings yet
History of Games
23 pages
Digital Communication: Gaurav S. Kasbekar Dept. of Electrical Engineering IIT Bombay
No ratings yet
Digital Communication: Gaurav S. Kasbekar Dept. of Electrical Engineering IIT Bombay
19 pages
Linear Continuous Wave (CW) Modulation: August 2013 Lectured by Assoc Prof. Thuong Le-Tien
No ratings yet
Linear Continuous Wave (CW) Modulation: August 2013 Lectured by Assoc Prof. Thuong Le-Tien
40 pages
Modulations Part1 PDF
No ratings yet
Modulations Part1 PDF
21 pages
TE311 Lecture07 AM
No ratings yet
TE311 Lecture07 AM
29 pages
Digital Communication Fundamentals Part II
No ratings yet
Digital Communication Fundamentals Part II
7 pages
Channel Encoding & Decoding
No ratings yet
Channel Encoding & Decoding
31 pages
Analog Modulation System (1) : 4-1, 4-2,4-3, 5-1 Lecture 7, 2008-9-26
No ratings yet
Analog Modulation System (1) : 4-1, 4-2,4-3, 5-1 Lecture 7, 2008-9-26
23 pages
Digital Communication: Gaurav S. Kasbekar Dept. of Electrical Engineering IIT Bombay
No ratings yet
Digital Communication: Gaurav S. Kasbekar Dept. of Electrical Engineering IIT Bombay
19 pages
Pulse Amplitude Modulated (PAM) Signals
No ratings yet
Pulse Amplitude Modulated (PAM) Signals
10 pages
DSB and AM PDF
No ratings yet
DSB and AM PDF
16 pages
Digital Communication: Modulation and Coding Course
No ratings yet
Digital Communication: Modulation and Coding Course
22 pages
Digital Communication I: Modulation and Coding Course: Term 3 - 2008 Catharina Logothetis
No ratings yet
Digital Communication I: Modulation and Coding Course: Term 3 - 2008 Catharina Logothetis
22 pages
Bandpass Modulation
No ratings yet
Bandpass Modulation
15 pages
Woodworking For Everyone 13 Step by Step Projects For Your Home and Family by Peter Alkema
No ratings yet
Woodworking For Everyone 13 Step by Step Projects For Your Home and Family by Peter Alkema
164 pages
Modulation AM e FM - Bem Explicado Sobre Processo AM
No ratings yet
Modulation AM e FM - Bem Explicado Sobre Processo AM
8 pages
Poc Formula
No ratings yet
Poc Formula
5 pages
Communication Systems - 4th Edition - Carlson (616-631) MI APRTE PDF
No ratings yet
Communication Systems - 4th Edition - Carlson (616-631) MI APRTE PDF
16 pages
Frequency Modulation: Analog and Digital Communications Autumn 2005-2006
No ratings yet
Frequency Modulation: Analog and Digital Communications Autumn 2005-2006
16 pages
AnalogCommunication PDF
No ratings yet
AnalogCommunication PDF
12 pages
Career - Preparation - Key
No ratings yet
Career - Preparation - Key
5 pages
Theatre of The Mind Ship Combat Rules - 17-End
100% (1)
Theatre of The Mind Ship Combat Rules - 17-End
4 pages
UK Ham Bands Chart - V0003 - 2 Pager
No ratings yet
UK Ham Bands Chart - V0003 - 2 Pager
2 pages
Connections in CLT Assemblies
No ratings yet
Connections in CLT Assemblies
59 pages
Skjoldehamn Hood Tutorial
100% (1)
Skjoldehamn Hood Tutorial
4 pages
Matt Hayek Photo Wedding Packages 3
No ratings yet
Matt Hayek Photo Wedding Packages 3
10 pages
Ofcom Ir 2009 Analogue and Digita1
No ratings yet
Ofcom Ir 2009 Analogue and Digita1
6 pages
(Ebook PDF) Focus On Grammar 4 With Essential Online Resources 5Th Edition Download
No ratings yet
(Ebook PDF) Focus On Grammar 4 With Essential Online Resources 5Th Edition Download
56 pages
Putrajaya Ebike Guided Tour
No ratings yet
Putrajaya Ebike Guided Tour
10 pages
6.recreation and Leisure
No ratings yet
6.recreation and Leisure
3 pages
Finite and Non-Finite Verbs
No ratings yet
Finite and Non-Finite Verbs
3 pages
Kiểm Tra Tiếng Anh 7 Giữa Kì 1 - Năm Học 2023 (Ms. Tham)
No ratings yet
Kiểm Tra Tiếng Anh 7 Giữa Kì 1 - Năm Học 2023 (Ms. Tham)
2 pages
Survey Questionnaire:: Survey Questions (Please Circle The Best Answer)
No ratings yet
Survey Questionnaire:: Survey Questions (Please Circle The Best Answer)
4 pages
Topics Hobbies Pre Intermediate
No ratings yet
Topics Hobbies Pre Intermediate
6 pages
BAO LONG - GRADE 3 - UNIT 5 (Lesson 3) - WS - 17122022
No ratings yet
BAO LONG - GRADE 3 - UNIT 5 (Lesson 3) - WS - 17122022
8 pages
5E CharacterSheet Fillable 2
No ratings yet
5E CharacterSheet Fillable 2
3 pages
Go Play Do Questionnaire 2
No ratings yet
Go Play Do Questionnaire 2
1 page
ACF MAINS 2021 401 (Compulsory English) 20032021104513
No ratings yet
ACF MAINS 2021 401 (Compulsory English) 20032021104513
6 pages
Solucion de Guia 0 de Ingles 11-6
No ratings yet
Solucion de Guia 0 de Ingles 11-6
6 pages
III. Analyze The Following Embroidered Articles Below. Identify What Is Being Asked
No ratings yet
III. Analyze The Following Embroidered Articles Below. Identify What Is Being Asked
3 pages
Sea Vines Dress in Hand Maiden Downloadable PDF - 2
No ratings yet
Sea Vines Dress in Hand Maiden Downloadable PDF - 2
1 page
Hashoo Foundation: Living A Better Future Proposal I
No ratings yet
Hashoo Foundation: Living A Better Future Proposal I
4 pages
Jacob The Great Final
No ratings yet
Jacob The Great Final
2 pages
Exercises in Electronics: Operational Amplifier Circuits
From Everand
Exercises in Electronics: Operational Amplifier Circuits
Roland Büchi
3/5 (1)
Analog Dialogue, Volume 47, Number 2
From Everand
Analog Dialogue, Volume 47, Number 2
Analog Dialogue
No ratings yet
Analog Dialogue, Volume 48, Number 1: Analog Dialogue, #13
From Everand
Analog Dialogue, Volume 48, Number 1: Analog Dialogue, #13
Analog Dialogue
4/5 (1)
Feynman Lectures Simplified 2C: Electromagnetism: in Relativity & in Dense Matter
From Everand
Feynman Lectures Simplified 2C: Electromagnetism: in Relativity & in Dense Matter
Robert Piccioni
No ratings yet

Chap7 - Digital Transmission Via Carrier Modulation

Uploaded by

Chap7 - Digital Transmission Via Carrier Modulation

Uploaded by

Chapter 7

Digital Transmission via

In the two preceding chapters, we considered the transmission of digital information

7.2 Carrier-Amplitude Modulation

In baseband digital PAM, the signal waveforms have the form

Am= (2m-l-M)d, m=l,2,... ,M (7 .2.2)

where 2d is the Euclidean distance between two adjacent signal points.

Figure 7.1 : Energy density spectrum of the transmitted signal Br(t)

Um(t) = AmBr(t) COS 2rrJct, m = 1,2,... ,M (7.2.3)

the amplitude-modulated carrier signal is usually called amplitude-shift keying (ASK).

Figure 7 .3: Spectra of (a) baseband and (b) amplitude-modulated signals

Figure 7 .4: Signal point constellation for PAM signal

t.jJ(t) = Br(t) cos 2rrJct (7 .2.6)

� f:00 Bf (t) dt = 1 (7.2.11)

72 1. . Demodulation of PAM Signals

r(t) =AmBT(t) cos 2rrJct + n(t) (7.2.12)

where n(t) is a bandpass noise process, which is represented as

n(t) = nc(t) cos 2rrJct - ns (t) sin 2rrJct (7.2.13)

f:00 r(t)ljJ(t) dt =Am+ n =Sm+ n (7.2.14)

Figure 7.5: Demodulation of bandpass digital PAM signal

where Eavb is the average energy per bit.

Illustrative Problem 7 .1 [PAM Signal Spectrum] In an amplitude-modulated digi­

Figure 7.6: Spectra of baseband signal and amplitude-modulated (bandpass) signal

r(t) = AmBT(t) cos 2rrfct + n(t)

with the reference waveform

<jJ(t) = gy(t) cos 2rrfct

where Am is the transmitted signal amplitude,gy(t) is the rectangular signal pulse

and n(t) is a bandpass Gaussian noise process, which is represented as

n(t) = nc(t) cos 2rrfct - ns(t) sin2rrfct

y(nTs) = L r(kTs)tfl(kTs), n= 1, 2, ...

Am= 1; % Signal Amplitude

Figure 7.7: Demodulator output in Illustrative Problem 7.2

In carrier-phase modulation the information that is transmitted over a communication

for allm (7.3.4)

BT(t) = �' 0:::; t:::; T (7.3.5)

180°-phase 0°-phase -90°-phase

Figure 7 .8: Example of a four-phase PSK signal

and l.fJ1(t) and l.fJ2(t) are orthogonal basis functions defined as

l/JI (t) = BT(t) cos 2rrJct

Sm = ( .JEs cos Z�m .JEs sin Z�

M=2 M=4 M=8

Figure 7 .9: PSK signal constellations

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

Figure 7 .10: M = 8 constant-amplitude PSK waveforms

7.3.1 Phase Demodulation and Detection

r(t) = Um(t) + n(t)

where nc and n5 are defined as

nc(t) and n5(t) are zero-mean Gaussian ran­

Yc(t) = J: r(T)l./J1 (T) dT

Yc(nTs) = L r(kTs)l./J1 (kTs), n = 1, 2, ...

Ys(nTs) = L r(kTs)l./Jz(kTs), n = 1, 2, ...

T5 = T /100 and the carrier frequency Jc = 3 0/T . The

xlabel(['n (rn=',num2str(m),') '])

Figure 7 11 : Correlator outputs in Illustrative Problem 7.4

C(r, Sm) = r ·Sm, m = 0, 1, . . . ,M - 1 (7.3.15)

Figure 7.12: Probability of a symbol error for M-ary PSK

The additive noise components nc and ns are statistically independent, zero-mean

llE Simulated bit-error rate

O Simulated symbol-error rate

- Theoretical bit-error rate

732. . Differential Phase Modulation and Demodulation

The demodulation of the phase-modulated carrier signal, as described above, requires

Received Delay Phase Output

Figure 7.15: Block diagram of DPSK demodulator

resulting complex number; that is,

rkr:_1 EseJCBk-ek-1) + ffs eJCBk-<f>lnk-1 + ffs e-JCBk-i-<f>lnk + nknz_1 (7.3.22)

Figure 7.16: Probability of error for binary PSK and DPSK

x ffs + Re(nk + nt_1)

Illustrative Problem 7.7 [Demodulation and Detection of DPSK Signals]

The MATLAB script for implementing differential detector is given below.

UnifonnRNG Gaussian RNG

r [cos rr;n + nc sin rr;n + ns ]

Mathematically, this computation can be performed as in (7.3.22)-that is,

rkr:_1 (rck +jrsd(rc k-1-jrsk-1)

7.4 Quadrature Amplitude Modulation

of information bits. The transmitted signal waveforms have the form

Illustrative Problem 7 .1 [PAM Signal Spectrum] In an amplitude-modulated digi

nc(t) and n5(t) are zero-mean Gaussian ran

M-ary orthogonal FSK waveforms have a geometric representation as M M-dimen