Chapter 8

Multicarrier Modulation
andOFDM

8.1 Preview

In Chapter 7 we considered modulation of a single carrier for transmitting digital infor­

mation. In this chapter, we describe the transmission of digital information on multiple
carriers. In multicarrier modulation, the available channel bandwidth is subdivided into
a number of equal-bandwidth subchannels, as illustrated in Figure 8 .1. Thus, we create

IC(f)I

0 --1 f-­ w f
il.f

Figure 8 .1: Subdivision of the channel bandwidth in multicarrier modulation

K = W / f).j subchannels, where W is the overall bandwidth of the channel and f).j
is the frequency separation between adjacent subchannels, or, equivalently, f).j is the
frequency separation between adjacent subcarrier frequencies. Then, different informa­
tion symbols are transmitted simultaneously and synchronously in the K subchannels.
Thus, the data are transmitted by frequency division multiplexing (FDM).

377
Copyright 2011 Cengage Leaming. 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 materiaJly affect the overall learning eAperience. Cengage Leaming reserves the right to remove additional content at any time if subsequent rights restrictions require iL
 378 CHAPTER 8. MULTICARRIER MODULATION AND OFDM

For each subchannel, let us associate a carrier signal-that is,

Xk(t) = sin 2rrfkt, k = 0, 1,... ,K - 1 (8.1.1)

where fk is the midfrequency in the kth subchannel. We select the symbol rate l/T
on each of the subchannels to be equal to the frequency separation tlj of adjacent
subcarriers. Consequently, the subcarriers are orthogonal over the symbol duration T,
independent of the relative phase relationship between any two subcarriers-that is,

J: sin(2rrfkt + cf>k) sin(2rrfjt + </>j) dt = 0 (8.1.2)

where fk -fj = n/T, n = 1,2,..., for any arbitrary cf>k and</>;. With this restriction,
we have orthogonal frequency division multiplexing (OFDM).
An OFDM system can be designed to avoid the intersymbol interference effects
that are usually encountered in a single-carrier system. If Ts is the symbol interval in a
single-carrier system, the symbol interval in an OFDM system having K subchannels is
T = KTs. By selecting K to be sufficiently large, the symbol interval T in the OFDM
system can be made much larger than the time dispersion of the channel. Thus, the
intersymbol interference in an OFDM system can be made arbitrarily small by proper
selection of K. In such a case, the bandwidth of each subchannel is sufficiently small
and appears to have a fixed frequency response C(jk), k = 0, 1,2,... ,K - 1.

Illustrative Problem 8.1 [Orthogonality Property of Subcarriers]

Two subcarrier signals in an OFDM signal are

xk(t) = sin(2rrfkt + </>k), 0:::; t:::; T

Xj(t) = sin(2rrfjt + </>j), 0:::; t:::; T

Let fk
2 Hz and fj
= = fk + 'Tf, where n cf>j are = 1,2, or 3, and the phases cf>k and
arbitrary over [O, 2rr].
Demonstrate the orthogonality property given in Equation 8.1.2
by using the sampled signals Xk(mTs) and Xj(mTs), where Ts 1I5 s, T 10 s and = =

m 0, 1,2,... , M - 1, where M TITs 50.

= = =

The sampled signals are

Xk(mTs) = sin(4rrm/5 + cf>k), m = 0, 1,... ,49

Xj(mTs) = sin(4rrm/5 + 2rrmn/50 + </>j), m = 0, 1,... ,49

Using MATLAB, we compute

49

I xk(mTs)Xj(mTs)
m=O

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.
 8.2. GENERATION OF AN OFDM SIGNAL 379

for n = 1, 2 and 3. The results of the computation are 4.9 x 10-14, 2.28 x 10-14, and
5.07 x 10-14.
The MATLAB script for the problem is given below.

----tllli" ------
% MATLAB script for Illustrative Problem 8.1

M= 50;
m=O:M-1;
phLk=2*pi*rand;
phLj=2*pi*rand;
% The sampled signals:
x_k=sin(4*pi*m/5+phLk);
n=1;
x_j_l=sin(4*pi*m/5+2*pi*m*n/M+phLj);
n=2 ;
x_j_2=sin(4*pi*m/5+2*pi*m*n/M+phLj);
n=3;
x_j_3=sin(4*pi*m/5+2*pi*m*n/M+phLj);
% Investigating the orthogonality of the sampled signals:
Suml=sum(x_k.*x_j_l);
% Displaying the results:
disp([' The result of the computation for n=l is: ',num2str(Suml)])
Sum2=sum(x_k.*x_j_2);
disp([' The result of the computation for n=2 is: ',num2str(Sum2)])
Sum3=sum(x_k.*x_j_3);
disp([' The result of the computation for n=3 is: ',num2str(Sum3)])

8.2 Generation of an OFDM signal

The modulator and demodulator in an OFDM system are usually implemented by use
of a parallel bank of tuned filters based on the discrete Fourier Transform (DFT). To
demonstrate this implementation, let us consider an OFDM modulator that generates
K independent subcarriers, where each subcarrier is modulated by symbols selected
from a QAM signal constellation. We denote the complex-valued signal points corre­
sponding to the information symbols on the K subchannels by Xk, k = 0, 1, ... , K - 1.
Hence, these information symbols {Xk} represent the values of the DFT of a multicar­
rier OFDM signal x ( t), where the modulation on each carrier is QAM. Because x ( t)

must be a real valued signal, its N-point DFT {Xk} must satisfy the symmetry property
XN-k = x:. Therefore, from the K information symbols {Xk}, we create N = 2K
symbols by defining

XN-k=x:, k=l, 2,...,K-1

x0 =Real(Xo) (8.2.1)
X� =lm(Xo)

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.
 380 CHAPTER 8. MULTICARRIER MODULATION AND OFDM

O bserve that the information symbol Xo is split into two parts, both of which are real.
It is convenient to denote the new sequence of symbols as {X�, k =0, 1, ... ,N - 1}.
Then the N-point inverse DFf (IDFf ) of {X�} yields the real valued sequence
N-1
Xn
1_ "' X' ej2rrn� , n=0, l,... ,N - 1
- L k (8.2.2)
__

.JN k=O

where }N is simply a scale factor. The sequence {Xn,0 � n �N - 1} corresponds to

the samples of the multicarrier OFDM signal x(t), consisting ofK=N / 2 subcarriers,
which may be expressed as

N-1
x(t) = - 1"' X�eJ·2rry
.JN k
L
kt
0 � t � T (8.2.3)
=O

where T is the signal duration (also the signal interval) and = Xn = x ( �), n
0, 1, ... ,N - 1. We note that the information signal Xo, which is represented by X0
and Xl< as given in (8.2.1), corresponds to the de component <Jo = 0) in (8.2.3). It is
convenient to set Xo=0, so that the multicarrier OFDM signal given by (8.2.3) has no
de component. Then, by making use of the symmetry conditions given by (8.2.1), the
multicarrier OFDM signal in (8.2.3) is expressed as

x(t) =
2 K-l 2rrkt
IXkl cos -----y-
( + ek
) , 0 � t � T (8.2.4)
.JN z;.1

where the information symbols xk= 1Xklei8k, k= 1, 2,... ,K - 1.

Illustrative Problem 8.2 [Generation of OFDM Signal] Using the 16-point QAM
signal constellation shown in Figure 7.21, select pseudorandomly each of the informa­
tion symbols X1,X2,... ,Xg. With T = 100 seconds, generate the transmitted signal
waveform x(t) given by (8.2.3) for t = 0, 1, ... , 100 and plot it. Then compute the
IDFT values Xn, for n = 0, l,... ,N - 1, by using (8.2.2). D emonstrate that x(t),
evaluated at Tn /N, n = 0, 1, ... ,N - 1, corresponds to the IDFT values. Finally,
using the IDFf values {Xn, 0 � n � N - 1}, compute the DFf, defined as

N-1
Xk= �Lxne-i2rrk N k=O,l,... ,N - l (8.2.5)
vN n=O

and, thus, demonstrate that the information symbols {Xk, 1 � k � 9} are recovered
from the samples of x(t), where t=nT /N, 0 � n �N - 1.

In this exampleK=lOandN= 20. The plot is shown inFigure 8.2 and theMATLAB
script for this problem is given next.

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.
 8.2. GENERATION OF AN OFDM SIGNAL 381

20 40 60 80 100

Figure 8.2: The plot in Illustrative Problem 8.2

____,,... ·------
% MATLAB script for Illustrative Problem 8.2.
echo on
K=1O;N=2*K;T=100;
a=rand(1,36);
a=sign(a-0.5);
b=reshape(a,9,4);
% Generate the 16QAM points
XXX=2*b(:,1)+b(:,2)+j*(2*b(:,3)+b(:,4));
XX=XXX•;
X=[O XX 0 conj(XX(9:-1:1 ))];
xt=zeros(1,101);
for t=0:100
for k=O:N-1
xt(1,t+1)=xt(1,t+1)+1/sqrt(N)*X(k+1)*expU*2*pi*k*t/T);
echo off
end
end
echo on
xn=zeros(1,N);
for n=O:N-1
for k=O:N-1
xn(n+1)=xn(n+1)+1/sqrt(N)*X(k+1)*expU*2*pi*n*k/N);
echo off
end
end
echo on
pause % press any key to see a plot of x(t)
plot([0:100],abs(xt))
% Check the difference between xn and samples of x(t)
for n=O:N-1

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.
 382 CHAPTER 8. MULTICARRIER MODULATION AND OFDM

d(n+1)=xt(T/N*n+1)-xn(1+n);
echo off
end
echo on
e=norm(d);
Y=zeros(1,10);
for k=1 :9
for n=O:N-1
Y(1,k+1)=Y(1,k+1)+1/sqrt(N)*xn(n+1)*exp(-j*2*pi*k*n/N);
echo off
end
end
echo on
dd=Y(1:10)-X(1:1 O);
ee=norm(dd);

8.3 Demodulation of OFDM Signals

As we have observed above, the modulator in an OFDM system can be implemented by

computing the IDFT, given in (8.2.2). The demodulator that recovers the information
symbols {Xk} from the received signal samples is implemented by computing the DFT,
given in (8.2.5). The detector is described below. When the number of subcarriers is
large, say K > 30, the modulator and demodulator in the OFDM system are efficiently
implemented by using the fast Fourier transform algorithm (FFT algorithm) to compute
the DFT and IDFT.
If the number of subcarriers is sufficiently large so that the time dispersion effects
of the subchannels are negligible, then the signal transmitted through each subchannel
is simply attenuated and phase shifted by an amount corresponding to the frequency
response characteristics of each subchannel. Furthermore, the received signal is cor­
rupted by additive noise. Hence, the received symbols at the outputs of the OFDM
demodulator may be expressed as

(8.3.1)

where Ck = C(fk) is the frequency response (magnitude and phase) of the kth sub­
channel and 'lk is the additive noise corrupting the kth symbol. The detector that
follows the OFDM demodulator must eliminate the effect of the channel frequency re­
sponse characteristics by dividing each received symbol Xk by Ck. Typically, this is
accomplished in practice by measuring the frequency response characteristics of each
subchannel and using these measurements (estimates of Ck) to eliminate the magnitude
and phase effects of the subchannel from the received symbols. Thus, the compensated
received symbols at the detector may be expressed as

(8.3.2)

Upon observing each received symbol, the detector compares Xk with each of the pos­
sible transmitted signal points in the QAM signal constellation and selects the symbol
from the constellation that is closest in distance to xk.

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.
 8.3. DEMODULATION OF OFDM SIGNALS 383

Illustrative Problem 8.3 [Effect of Additive Noise on OFDM] Consider the OFDM
system described in Illustrative Problem 8.2. Suppose the IDFT samples given by
(8.2.2) of the transmitted signal are corrupted by additive Gaussian noise having zero
2
mean and variance cr . Hence, the received signal samples are represented as

rn=Xn+Bn n=O,l,... ,N-1

where Bn is the noise component. Compute the DFT of the sequence {rn} and then
obtain the estimates of the received symbols {Xk} when the noise variance is cr2 =l, 2,
2
and 4. For each value of cr , detect the sequence Xk and, thus, obtain the output
symbols {Xd from the detector. Comment on the accuracy of the detected symbols.

The MATLAB script for this problem is given next. This script computes the number
2
of symbols received in error for cr = 1. By changing the variable "variance" in
this script, we can obtain the number of errors for different values of noise variance.
2
The results of sample runs of this script for cr = 1, 2, and 4 are summarized in the
following table. The first row of this table denotes the transmitted signals and the
2
following three rows show detected values for cr = 1, 2, and 4.

xk 1- 3j -l+j 3+j 1+3j 1- 3j 1- 3j 3+3j -1+3j 1+ 3 j

xb CJ2 = 1 1- 3j -1+j l+j 1+3j 1- 3j 2�t4i§1i2 0P410=i'J2 liJ.:N.4.t 97�HJ@j
xb CJ2 = 2 1- 3j 1+3j -1+j 1+3j 3- 3j 1- 3j 3+3j -1+3j -1+3j
xb CJ2 = 4 l+j -1+j -1- j 1+3j -1+j 1- 3j 3- 3j 3+3j 1+ 3 j

____,.... ·-----
% MATLAB script for Illustrative Problem 8.3.
echo on
K=1O;N=2*K;T=1OO;variance=1;
noise=sqrt(variance}*randn(1,N};
a=rand(1,36};
a=sign(a-0.5);
b=reshape(a,9 ,4);
% Generate the 16QAM points
XXX=2*b(:, 1 }+b(:,2)+j*(2*b(:,3}+b(:,4));
XX=XXX';
X=[O XX 0 conj(XX(9:-1:1}}];
x=zeros(1,N);
for n=O:N-1
for k=O:N-1
x(n+1)=x(n+1)+1/sqrt(N}*X(k+1)*expU*2*pi*n*k/N);
echo off
end
end
echo on
r=x+noise;
Y=zeros(1,10);

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.
 384 CHAPTER 8. MULTICARRIER MODULATION AND OFDM

for k=1 :9
for n=O:N-1
Y(1,k+1)=Y(1,k+1)+1/sqrt(N)*r(n+1)*exp(-j*2*pi*k*n/N);
echo off
end
end
echo on
% Detect the nearest neighbor in the 16QAM constellation
for k=1 :9
if real(Y(1,k+1))>0
if real(Y(1,k+1))>2
Z(1,k+1)=3;
else
Z(1,k+1)=1;
end
else
if real(Y(1,k+1))<-2
Z(1,k+1)=-3;
else
Z(1,k+1)=-1;
end
end
if imag(Y(1,k+1))>0
if imag(Y(1,k+1))>2
Z(1,k+1)=Z(1,k+1)+3*j;
else
Z(1,k+1)=Z(1,k+1)+j;
end
else
if imag(Y(1,k+1))<-2
Z(1,k+1)=Z(1,k+1)-3*j;
else
Z(1,k+1)=Z(1,k+1)-j;
end
end
echo off
end
echo on
error=max(size(find(Z(1,2:1O)-X(1,2:1O))));

8.4 Use of a Cyclic Prefix to Eliminate

Channel Dispersion
In our treatment of OFDM given above, we assumed that there is no time dispersion
introduced by the channel on each of the subcarrier signals. However, that may not be
the case in many communication channels in which OFDM may be employed. In this
section, we consider a method to mitigate channel dispersion.
When x(t) is the input to the channel, the channel output at the receiver may be
expressed as
r(t) = x(t) * c(t) + 17(t), (8.4.1)

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.
 8.4. USE OF A CYCLIC PREFIX TO ELIMINATE CHANNEL DISPERSION 385

where c ( t) is the impulse response of the channel and * denotes convolution. Because
the bandwidth t:!f of each subchannel is selected to be very small relative to the overall
channel bandwidth W = K t:!f, the symbol duration T = 1 I t:!f is larger than the
duration of the channel impulse response. To be specific, suppose that the channel
impulse response spans m + 1 signal samples, where m « N. A simple way to
completely avoid intersymbol interference (ISi) is to insert a time guard of duration
m T /N between the transmission of successive data blocks. This allows the response
of the channel to die out before the next block of K symbols is transmitted.
An alternative method of avoiding ISi is to append a so-called cyclic prefix to each
block of N signal samples {Xn, 0 .::; n .::; N - 1}. The cyclic prefix for the block of
samples contains the samples XN-m. XN-m+1 • . . . , XN-1 ·These samples are appended
to the beginning of the block, thus creating a signal sequence of length N + m samples,
which may be indexed from n = -m to n = N - 1, where the first m samples
constitute the cyclic prefix. Then, if the sample values of the channel response are
{en, 0 .::; n .::; m } , the convolution of {en} with {Xn, -m .::; n .::; N - 1} produces
the received signal {rnl. Because the ISi in any pair of successive signal transmission
blocks affects the first m signal samples, we discard the first m samples of {rn} and
demodulate the signal based on the received signal samples {rn. 0 .::; n .::; N - 1}.
If we view the channel characteristics in the frequency domain, the channel fre­
quency response at the subcarrier frequencies fk = k/T is
m
ck= c (2rrk/N) = L Cne-j2rrnk/N, k= 0, 1, ... , N - 1 (8.4.2)
n=O

Because the ISi is eliminated through either the cyclic prefix or the time guard band,
the demodulated sequence of symbols may be expressed as

k= 0, 1, ... ,N - 1 (8.4.3)

where {Xk} is the output of the N-point DFf computed by the demodulator and {1Jk} is
the additive noise corrupting the signal. By estimating the channel parameters {Cb 0 .::;

k .::; N - 1} and normalizing each Xk by Ck. we obtain the desired transmitted data,
corrupted by the additive noise; that is,

X'k= Xk + 1'1k· k= 0, 1, ... ,N - 1 (8.4.4)

Figure 8.3 illustrates the basic building blocks in the implementation of the trans­
mitter and the receiver in an OFDM system. As previously indicated, the FFT algo­
rithm is usually employed to efficiently implement the DFf and IDFT.

Illustrative Problem 8.4 [Use of Cyclic Prefix]

Using the OFDM signal generated in Illustrative Problem 8.2, add a cyclic prefix
of four samples to account for the channel dispersion and, thus, modify the MATLAB
script given in Illustrative Problem 8.2.

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.
 386 CHAPTER 8. MULTICARRIER MODULATION AND OFDM

Input Serial-to­ Multi carrier Transmitted

DIA
Parallel Modulator
Symbols Converter Signal x(t)
Buffer (IDFT)
{X<}

Transmitter

Parallel-
Output OFDM Received
to- AID
Demodulator r(t)
Symbols Serial Converter Signal
(DFT)
Converter

Receiver

Figure 8 .3: Block diagram of OFDM transmitter and receiver

In this problem K = 10 and N = 20. The signal values from the N-point DFT are
Xo,X1, ... ,X19. To this sequence, we append the values X15,X17,X18, and X19 at the
front end, prior to xo. Thus, the transmitted signal sequence at the input of the DIA
converter is

The modification to the MATLAB script is given below.

--®Ii" 11-----
% MATLAB script for Illustrative Problem 8.4

echo on
K=1O;N=2*K;T=1OO;m=4;
a=rand(1,36);
a=sign(a-0.5);
b=reshape(a,9,4);
% Generate the l 6QAM points
XXX=2*b(:,1)+b(:,2)+1 i*(2*b(:,3)+b(:,4));
XX=XXX';
X=[O XX 0 conj(XX(9:-1:1))];
xt=zeros(1,101);
for t=0:100
for k=O:N-1
xt(1,t+1)=xt(1,t+1)+1/sqrt(N)*X(k+1)*exp(1 i*2* pi*k*t!f);
echo off
end
end
echo on
xn=zeros(1,N+m);
for n=O:N-1

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.
 8.5. SPECTRAL CHARACTERISTICS OF OFDM SIGNALS 387

for k=O:N-1
xn(n+m+1)=xn(n+1)+1/sqrt(N)*X(k+1)*exp(1 i*2*pi* n*k/N);
echo off
end
end
xn(1 :m)=xn(N-m+1 :N);
echo on
pause % press any key to see a plot of x(t)
plot([0:100],abs(xt))
% Check the difference between xn and samples of x(t)
for n=O:N-1
d(n+1)=xt(T/N*n+1 )-xn(1 +n+m);
echo off
end
echo on
e=norm(d);
Y=zeros(1,1O);
for k=1 :9
for n=O:N-1
Y(1,k+1)=Y(1,k+1)+1/sqrt(N)*xn(n+m+1)*exp(-1 i*2*pi*k*n/N);
echo off
end
end
echo on
dd=Y(1:10)-X(1:10);
ee=norm(dd);

In the discussion above, the same QAM signal constellation is used to generate the
information symbols for all the subcarriers. However, OFDM allows us to transmit a
different number of bits/symbol on each subcarrier. Specifically, subcarriers that have a
higher signal-to-noise-ratio (SNR) due to lower attenuation can be modulated to carry
more bits/symbol than subchannels that have a higher attenuation. Hence, QAM with
different constellation sizes may be used on the different subcarriers.
OFDM using QAM on each of the subcarriers as described above has been im­
plemented for a variety of applications, including high-speed data transmission over
telephone lines, such as digital subscriber lines. This type of OFDM modulation has
also been called discrete multitone (DMT) modulation. OFDM is also used in digital
audio broadcasting in Europe and other parts of the world and in digital wireless local
area networks (LANs).

8.5 Spectral Characteristics of OFDM Signals

As indicated above, the signals transmitted on the subcarriers of an OFDM system are
mutually orthogonal in the time domain; that is,

f: Xk(t)Xj(t) dt = 0, k-/= j (8.5.1)

where Xk(t) is defined in Equation 8.1.1. However, these signals have significant
overlap in the frequency domain. This can be observed by computing the Fourier

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.
 388 CHAPTER 8. MULTICARRIER MODULATION AND OFDM

transform of the signal

x,(t) �Re [Jt x,eJ2rrht ]

= Jt Ak cos(2rrfkt + ek), 0 � t � T (8.5.2)

for several values ofk.

---lil!IJ§ll;§il�tj§lij;t•):Jl§j@I
Illustrative Problem 8.5 [Spectrum of OFDM Signal] Determine the magnitude
IUk(j) I of the Fourier transform of the signal given by Equation 8.5.2 for fk = kIT,
k= 0, 1, 2, 3, 4, 5. For simplicity, let Ak = 1 and ek = 0 for allk. Plot IUk(j) I on the
same graph fork= 0, 1, 2, 3, 4, 5 for 0 � f � 4 /T.

The Fourier transform of Xk(t) = ,J2 / T cos 2rrfkt for 0 � t � T may be expressed
as the convolution of G(j) with V(j) where

1
V(j) = 2 [8(j - fk) + 8 (f + fk)] - cos 2rrfkt
G(j) = me-j(rrfT-rrf2>sinc(jT)

Hence,

IUk(j) I= ,ff- I sinc(j - fk)T + sinc(j + fk)TI

The plot of IUk(j) I is shown in Figure 8.4. Note the large spectral overlap of the main
lobes of each IUk(j) I. Also note that the first sidelobe in the spectrum is only 13 dB

20 I I I I I I I

0
-. - ,.-'.'.:: �
-20 � � � --<! � -.;;;:: � �

-40
r
ID -60
:s.
-80
s ""

2 -100

-120

-140

-160 -

I I I I I I I
-180
0 0.5 1.5 2 2.5 3 3.5 4

Figure 8.4: IUk(j) I in Illustrative Problem 8.5

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.
 8.6. PEAK-TO-AVERAGE POWER RATIO IN OFDM SYSTEMS 389

down from the main lobe. Therefore, there is a significant amount of spectral overlap
among the signals transmitted on different subcarriers. Nevertheless, these signals are
orthogonal when transmitted synchronously in time.
The MATLAB script for the problem is given below.

% MATLAB script for Illustrative Problem 8.5

T = 1;
k = 0 : 5;
Lk = kif ;
f = -4tr : O.D1*4ff : 4ff;
U_k_abs = zeros(length(k),length(t));
for i = 1 : length(k)
U_Labs(i,:) = abs(sqrt(T/2)*(sinc((f-Lk(i))*T) + sinc((f+Lk(i))*T)));
end
plot(f,U_k_abs(1,:),'. -',f,U_k_abs(2,:),'--',f,U_k_abs(3,:), 'c-' ,f,U_k_abs(4,:),'. ',f,U_k_abs(5,:),
f,U_k_abs(6,:))
xlabel(' f ' )
ylabel(' I U_k ( f ) I ')

8.6 Peak-to-Average Power Ratio in OFDM Systems

A major problem in multicarrier modulation is the high peak-to-average power ratio
(PAR) that occurs in the modulated signal. Generally, large signal peaks in the trans­
mitted signal occur when the signals in the K subchannels add constructively in phase.
Such large signal peaks may saturate the power amplifiers at the transmitter, thus caus­
ing intermodulation distortion in the transmitted signal. Intermodulation distortion can
be reduced and often avoided by reducing the power in the transmitted signal and, thus,
operating the power amplifier at the transmitter in its linear range. Such a power re­
duction or "power back-off" results in inefficient operation of the OFDM system. For
example, if the PAR is 10 dB, the power back-off may be as much as 10 dB to avoid
intermodulation distortion.
One of the simplest methods to reduce the PAR is to limit the peak amplitude of the
transmitted signal to some value above the average power. However, such amplitude
clipping will result in signal distortion. The following illustrative problem investigates
this approach.

----tiMIU-il;tfilW§l§;t•1:JM§®I
Illustrative Problem 8.6 [Computation of PAR]
Generate samples of the OFDM signal

x(t) =
K-l
2:: cos
( �kt
2
+ ek
) ' 0 � t � T
k=l

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.
 390 CHAPTER 8. MULTICARRIER MODULATION AND OFDM

where K = 3 2, T = 1 s, the sample rate F5 = 200 samples per second, and the
modulation on each subcarrier is 4-phase PSK; that is, ek takes the possible values 0,
TT/ 2, TT, 3 rr / 2, selected pseudorandomly. For each realization of x ( t), determine the
PAR. Repeat the computation of the PAR for 20 different realizations of x ( t) and plot
the values of the PAR for the 20 different realizations.

The average power of the sample {Xn} is

1 199 2
Pav= I Xn
200
n=O
and the peak power is
Ppeak: = max{x�}
n
Hence, the PAR= Ppeak/Pav· The plot of the PAR is shown in Figure 8.5.

12 I I I I I I I I I

'

10

'

'

�
I
6

0
2 4 6 8 10 12 14 16 18 20

Realization

Figure 8.5: PAR values in Illustrative Problem 8.6

The MATLAB script for the problem is given below.

----tl&li" ------
% MATLAB script for Illustrative Problem 8.6

T =1;
Fs = 200;
t = 0:1/(Fs*T):T-1/(Fs*T);
K =32;
k=1:K-1;
rlz =20; % No. of realizations
PAR =zeros(1,rlz); % Initialization for speed

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.
 8.6. PEAK-TO-AVERAGE POWER RATIO IN OFDM SYSTEMS 391

forj=1:rlz
theta =pi*floor(rand(1,length(k))/0.25)/2;
x=zeros(1,Fs); % Initialization for speed
echo off;
for i=1:Fs
for I=1:K-1
x(i)=x(i) + cos(2*pi*l*t(i)!r+theta(l));
end
end
echo on;
% Calculation of the PAR:
P_peak=max(x."2);
P_av=sum(x."2)/Fs;
PARU)=P_peak/P_av;
end
% Plotting the results:
stem(PAR)
axis([1 20 min(PAR) max(PAR)])
xlabel(' Realization')
ylabel(' PAR')

----tll!4-il;tfilrijij;t•1:Jl§@I
Illustrative Problem 8.7 [Limiting the PAR by Clipping the Peaks]
Repeat Illustrative Problem 8.6, but in this case clip the peak amplitude of the
samples so that the PAR � 3 dB. Define the clipped signal as {.Xn} and compute the
signal distortion defined as

1 199
D = 200 I (Xn - Xn)2
n=O
0
Plot D for the 2 signal realizations.

The plot of the signal distortion Dis shown in Figure 8.6.

The MATLAB script for the problem is given below.

--®iii! ----
% MATLAB script for Illustrative Problem 8.7

T=1;
Fs=200;
t= 0: 1/(Fs*T):T-1/(Fs*T);
K=32;
k=1:K-1;
rlz=20; % No. of realizations
% Initialization for speed:
PAR=zeros(1,rlz);

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.
 392 CHAPTER 8. MULTICARRIER MODULATION AND OFDM

PAR_dB =zeros(1,rlz);
D =zeros(1,rlz);
echo off;
forj=1:rlz
theta = pi*floor(rand(1,length(k))/0.25)/2;
x =zeros(1,Fs); % Initialization for speed
for i =1:Fs
for 1 =1:K-1
x(i) =x(i) + cos(2*pi*l*t(i)/T +theta(l));
end
end
x_h =x;
% Calculation of the PAR.:
[P_peak idx] =max(x.A2);
P_av =sum(x.A2)/Fs;
PAR(j) =P _peak/P_av;
PAR_dB(j) =10*log10(PARU));
% Clipping the peak:
if P_peak/P_av > 1.9953
while P_peak/P_av > 1.9953
x_h(idx) =sqrt(10A0.3*P_av);
[P_peak idx] =max(x_h.A2);
P_av =sum(x_h.A2)/Fs;
PAR_dB(j) =1 O*loglO(P_peak/P_av);
end
end
D(j) =sum((x-x_h).A2)/Fs; % Distortion
end
echo on;
% Plotting the results:
stem(D)
axis ([1 20 min(D) max(D)])
xlabel('Realization')
ylabel(' Distortion ( D) ')

20

18�
I
16�

14�

e: 12'""
"
0
"€JO,_
.� 8,..
0

6�

4�

z,_

0
2 4 6 8 10 12 14 16 18 20

Realization

Figure 8.6: Distortion plots in Illustrative Problem 8.7

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.
 8.6. PEAK-TO-AVERAGE POWER RATIO IN OFDM SYSTEMS 393

Another method to reduce the PAR in OFDM is to insert different phase shifts
in each of the subcarriers. These phase shifts can be selected pseudorandomly or by
means of some algorithm. For example, we may have a small set of stored pseudoran­
domly selected phase shifts that can be used when the PAR in the modulated subcarrier
is large. The information on which set of pseudorandom phase shifts is used in any sig­
nal interval can be transmitted to the receiver on one of the K subbands. This approach
is explored in the following illustrative problem.

----tll!IX-ii;Miri§l#IJ;r•]:J!§®·•---
Illustrative Problem 8.8 [Reduction of PAR by Random Phase Shifting]
Generate samples of the OFDM signal

=
K-l
( 27;.kt
x(t) I
k=l
cos + ek + ct>k
) •
0 � t � T

where the phase c/>k is selected from a uniform distribution on interval (0, 2rr) and the
remaining signal parameters are identical to those in Illustrative Problem 8 .6. Generate
four sets of uniformly distributed phases {cf> k} , and for each realization of x(t) , select
20 different
the set that yields the smallest PAR. Repeat this process for realizations of
x(t) and plot the resultant PAR for the 20 realizations of x(t).

The plot of the PAR is shown in Figure 8.7.

5 10 15 20
Realization

Figure 8.7: PAR values in Illustrative Problem 8.8

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.
 394 CHAPTER 8. MULTICARRIER MODULATION AND OFDM

The MATLAB script for the problem is given below.

----tllli" ------
% MATLAB script for Illustrative Problem 8.8

T=1;
Fs=200;
t=0:1/(Fs*T) T-1/(Fs*T);
K =32;
k=1:K-1;
rlz =20; % No. of realizations
PAR=zeros(1,rlz); % Initialization for speed
echo off;
forj=1:rlz
theta =pi*floor(rand(1,length(k))/0.25)/2;
phi =2*pi*rand(4,length(k));
PAR_phi= zeros(1,size(phi, 1)); % Initialization for speed
for m=1: size(phi,1)
x=zeros(1,Fs); % Initialization for speed
for i=1:Fs
for l=1:K-1
x(i)=x(i) + cos(2*pi*l*t(i)ff+theta(l)+phi(m,l));
end
end
% Calculation of the PAR:
P _peak=max(x. A2);
P _av=sum(x. A2)/Fs;
PAR_phi(m)=P _peak/P _av;
end
[PARU) idx_theta]=min(PAR_phi);
end
echo on;
% Plotting the results:
stem(PAR)
axis ([1 20 min(PAR) max(PAR)] )
xlabel(' Realization')
ylabel(' PAR')

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.
 PROBLEMS 395

Problems

8.1 Repeat Illustrative Problem 8 .2 by generating an OFDM signal that transmits the
four 16-point QAM information symbols (Xi, X2, X3, X4).

8.2 Repeat Illustrative Problem 8 .2 by generating an OFDM signal that transmits the
K = 16 information symbols (Xo, Xi, ..., Xis) selected from the four-phase PSK con­
stellation . Use the FFf algorithm to compute the DFf and IDFT .

8.3 Repeat Illustrative Problem 8 .3 for the K = 16 OFDM signal that consists of the in­
formation symbols (Xo, Xi, ..., Xis) selected from the four-phase PSK constellation .
Use the FFT algorithm to compute the DFf and IDFT .

8.4 Compute the Fourier transform of the OFDM signal given in (8 .2 .3), where K = 16
and the symbols (Xo, Xi, ..., Xis) selected from the four-phase PSK constellation .
Plot the magnitude IX(j) I for the OFDM signal.

8.5 Repeat Illustrative Problem 8 .6 for the same signal parameters .

8.6 Repeat Illustrative Problem 8 .7 for the PAR:::; 4 dB and for the PAR:::; 6 dB .

8.7 Repeat Illustrative Problem 8 .8 for K = 8, four-phase PSK transmitted symbols .

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.