7

Orthogonal Frequency-Division
Multiplexing

7.1 Introduction

Orthogonal Frequency-Division Multiplexing (OFDM) forms the basis for 4G, i.e., Fourth
Generation wireless communication systems. OFDM is used in 4G wireless cellular standards
such as Long-Term Evolution (LTE) and WiMAX (Worldwide Interoperability for Microwave
Access). OFDM is a key broadband wireless technology which supports data rates in excess of
100 Mbps. Similarly, the wireless local area (LAN) standards such as 802.11 a/g/n are based
on OFDM. Next we describe multicarrier transmission, which is the motivation and key idea
behind OFDM.

7.2 Motivation and Multicarrier Basics

Consider a bandwidth B = 2W available for communication, where W is the one-sided

bandwidth, or, in other words, the maximum frequency. For a single carrier communication
system, the symbol time T is given as

1
T =
B

1
basically implying that symbols can be transmitted at intervals of B seconds each. Therefore,
the symbol rate is given as

1
Rate = =B (7.1)
1/B
Such a system is termed a single-carrier communication system. In such a system, a single
carrier is employed for the entire baseband bandwidth of B . Therefore, roughly speaking, the
symbols are transmitted as symbol X(0) from 0 ≤ t < T , symbol X(1) from T ≤ t < 2T ,
and so on, i.e., roughly one symbol transmitted every T = B1 seconds.
Consider now dividing the total bandwidth B into N sub-bands of bandwidth B/N each as
shown in Figure 7.1. Each subcarrier can now be represented by a subcarrier. Therefore, the
B B
subcarriers are placed at . . . , − N , 0, N , . . ., as shown in the figure. For instance, consider the
bandwidth B = 256 kHz with N = 64 subcarriers. The bandwidth per sub-band is equal to
256
64 = 4 kHz, which is also the frequency spacing between the subcarriers. We now implement
a multi-carrier transmission system as follows. Consider the ith subcarrier at the frequency
B
fi = i N , with − N2 − 1 ≤ i ≤ N2 . Let Xi denote the data transmitted on the ith subcarrier.
Then, the signal si (t) corresponding to the ith subcarrier is given as

B
si (t) = Xiej2πfi t = Xiej2πi N t

where fi is the ith subcarrier centre frequency, as described above, and ej2πfi t is the ith
subcarrier. The above equation shows the data modulation process over the ith subcarrier.
The N different data symbols Xi are modulated over the N different subcarriers with centre
frequencies fi . Hence, there are a total of N data streams. Next we illustrate the scheme for
multicarrier transmission.
Subcarrier
Sub-band spacing N subcarriers

-(N/2-1)B/N -B/N B/N 2B/N (N/2)B/N

Total bandwidth B
Figure 7.1 Multi-carrier concept
7.2.1 Multicarrier Transmission
Consider now the different modulated signals si (t) corresponding to the N different
subcarriers. These signals are then superposed at the transmitted to form the composite signal
s (t) given as

s (t) = si (t)
i

= Xi ej2πfi t
i
B
= Xi ej2πi N t (7.2)
i

This composite signal s (t) is then transmitted over the wireless channels. Thus, N different
data streams are transmitted over N subcarriers in parallel in this multicarrier system. At the
receivers, the individual data streams have then to be isolated from the composite signal s (t)
above. This is accomplished as follows. Consider the signal y (t) received as

y (t) = s (t) = Xi ej2πfi t

i

For simplicity, to illustrate the demodulation procedure at the receiver, we have assumed noise
to be absent above. We will consider the general case of a noisy received signal later. From the
expression for the composite signal s (t) in Eq. (7.2), it can be readily seen that the expression
on the right-hand side is indeed the Fourier series respresentation s (t), corresponding to the
fundamental frequency f0 = (B) N and the various Xi representing the Fourier coefficients.
B
Indeed, all the frequencies i N are multiples of the fundamental frequency f0 = T10 = N B
.
Therefore, to extract Xl , which is the Fourier coefficient corresponding to the frequency
fl = lf0 , one needs to follow the procedure similar to compute the Fourier series as
N
T0
j2πfl t
∗ B B B B
f0 y (t) e dt = Xi ej2πi N t e−j2πl N t dt
0 N 0 i

N
B B
= Xi ej2π(i−l)f0 t dt
N 0
i
 N N
B B B B
= Xl dt + Xiej2π(i−l)f0 t dt
N 0 N 0
i=l
i=l

N
B B
= Xl + Xi ej2π(i−l)f0 t dt
N 0
i=l
=0

= Xl

T
where we have used the fact that 0 0 ej2π(i−l)f0 t dt = 0 for i = l , since this is basically
integrating a sinusoid of frequency (i − l) f0 , which is a multiple of the fundamental frequency
f0 over the period T0 . Therefore, since there are an integer number of cycles of the sinusoid
of frequency (i − l) f0 , this integral is 0. In fact, this basically implies that the different
sinusoids ej2πif0 t and ej2πlf0 t are orthogonal. It is this key property of orthogonality which
helps extract the different streams Xi modulated over the different subcarriers. This property
of orthogonality can be summarized as
⎧
N/B ⎨0 i=l
B
ej2π(i−l) N t =
0 ⎩N i= l
B

Therefore, all the subcarriers other than the l th subcarrier are orthogonal to the l th
∗
subcarrier. Further, observe that multiplying with ej2πfl t and integrating is basically
coherent demodulation, i.e., demodulation with the carrier matched to the subcarrier frequency
B
fl = l N . Thus, Xl , the data modulated on the different subcarriers, can be conveniently
recovered by coherently demoudulating with each of the subcarriers corresponding to
l = − N2 − 1 , . . . , N2 . The above scheme of transmission on multiple orthogonal subcarriers
and the associated data recovery at the receiver is termed MultiCarrier Modulation (MCM).
Also observe that the window of time associated with detection of this multicarrier signal
is N 1
B = f0 = T0 , which is basically the time period of integration. Hence, MCM basically
transmits N symbols using N subcarriers in a time period of N B . The symbol rate is, therefore,
N
N/B = B . Thus, the overall symbol rate in single carrier vs multicarrier systems is unchanged.
The transmitter and receiver block schematics for this MCM system are shown in figures 7.2
and 7.3 respectively.
 Composite
signal

Bank
S/P Summer
of
Symbols Demux + To channel
modulators

Serial to parallel i th data stream is Sum all the

conversion because modulated onto subcarriers
we are transmitting N the i th subcarrier
symbols in parallel

Figure 7.2 Multicarrier modulation transmitter

N Information
symbols

Bank
of
Repeater P/S
correlators
y(t) Mux Serail symbol
or
from channel demodulators stream

Coherent Parallel to serial

demodulation multiplexing
with exp (-j2pflt)

Figure 7.3 Multi-carrier modulation receiver

It is very important now to note the following fact. Observe from Eq. (7.1) and the above
rate for an MCM system. It is clear that the symbol rate in both these systems is exactly
identical, i.e., B . The single-carrier system transmits each symbol in time B1 , while the
MCM system transmits N symbols in parallel in time N B . What then is the advantage of
an MCM system over the single-carrier system? To understand this, consider an example
with a transmission bandwidth of B = 1.024 MHz, i.e., 1024 kHz. As seen in an earlier
chapter, notice that this bandwidth B is much greater than the coherence bandwidth Bc
which is typically around 250 kHz, i.e., Bc ≈ 250 kHz. Therefore, since the transmission
bandwidth B >> Bc , the single-carrier system experiences frequency-selective fading and
inter-symbol interference. However, consider an OFDM system with employs N = 256
subcarriers in the same bandwidth. The bandwidth per subcarrier is Bs = 1024
256 = 4 kHz. It can
be readily seen that the subcarrier bandwidth of 4 kHz is significantly lower than the coherence
 B
bandwidth of 250 kHz. Thus, since N << Bc , each subcarrier experiences flat fading.
Hence, there is no inter-symbol interference in the data transmitted on any of the subcarriers.
Thus, the most critical and key benefit of this MCM system is that through parallel transmission
using multiple narrowband subcarriers, it eliminates the Inter-Symbol interference (ISI), thus
avoiding distortion of the received symbols.
However, the above MCM system suffers from a significant bottleneck. Implementing the
bank of N modulators and N demodulators with closely spaced subcarrier frequencies is an
extremely challenging task. This was solved by the key idea proposed by Weinstein and Ebert
in 1972, in the paper titled "Data Transmission by Frequency Division Multiplexing using the
Discrete Fourier Transform". Both of them were engineers at Bell Telephone Laboratories.
Their idea can be described as follows. Consider the MCM transmt signal s (t). Observe that it
is band-limited to the bandwidth B (total bandwidth). Therefore, the Nyquist sampling rate is
B and the associated sampling time is Ts = B1 . Consider now the composite MCM sigal given
in Eq. (7.2). The uth sample at time instant uTs = Bu is given as
B u
s (uTs ) = x (u) = Xi ej2πi N B
i

iu
x (u) = Xi ej2π N
i
DFT

Observe from the expression above that the sample x (u) is basically the In-
verse Discrete Fourier Transform (IDFT) coefficient of the information symbols
X (0) , X (1) , . . . , X (N − 1) at the uth time point. Thus, the Inverse Fast Fourier Transform
(IFFT) can be conveniently employed to generate the sample MCM signal. This scheme of
generating the composite transmit signal through IDFT was proposed by Weinstein and Ebert
in 1971. Thus, it drastically reduces the complexity of implementing an OFDM system since
it eliminates the need for the bank of modulators corresponding to the different subcarrier
frequencies. This technique, where the MCM signal is generated by employing the IFFT
operation is termed Orthogonal Frequency Division Multiplexing, or OFDM. At the receiver,
to recover the information symbols, one can correspondingly employ an FFT operation.
Schematic figures of the OFDM transmitter and receiver with the IFFT and FFT blocks are
given in figures 7.4 and 7.5 respectively.
 Figure 7.4 OFDM transmitter schematic with IFFT

Figure 7.5 OFDM receiver schematic with FFT

7.2.2 Cyclic Prefix in OFDM

In this section, we explain the concept of cyclic prefix, which is an important component
of an OFDM system. Consider a frequency-selective channel modelled with channel taps
h (0) , h (1) , . . . , h (L − 1). Thus, the received symbol y at a given time instant n can be
expressed as

y (n) = h (0) x (n) + h (1) x (n − 1) + . . . + h (L − 1) x (n − L + 1),

ISI component

from which it can be seen that the received symbol y (n) at the time instant n experiences inter
symbol interference from the previous L − 1 transmitted symbols. Consider now two OFDM
symbols as follows. Let x (0) , x (1) , . . . , x (N − 1) denote the IFFT samples of the modu-
lated symbols X (0) , X (1) , . . . , X (N − 1), while x̃ (0) , x̃ (1) , . . . , x̃ (N − 1) denote the
IFFT samples of the previous modulated symbol block X̃ (0) , X̃ (1) , . . . , X̃ (N − 1). Thus,
the samples corresponding to these two blocks of OFDM symbols are transmitted sequentially
as

x̃ (0) , x̃ (1) , . . . , x̃ (N − 1) , x (0) , x (1) , . . . , x (N − 1)

Previous block Current block

Now, consider the received symbol y (0) corresponding to the transmission of x (0). This can
be expressed as

y (0) = h (0) x (0) + h (1) x̃ (N − 1) + . . . + h (L − 1) x̃ (N − L + 1) .

ISI from previous OFDM symbol

It can be seen from the above equation that the received symbol y (0) experiences inter-symbol
interference from x̃ (N − 1) , x̃ (N − 2) , . . . , x̃ (N − (L − 1)). Thus, there is inter-OFDM
symbol interference in this new OFDM system. The initial samples of the current OFDM
symbol block are being subject to interference from the N − 1 samples of the previous OFDM
block. This is shown in Figure 7.6. Similarly, the received symbol y (1) is given as

y (1) = h (0) x (1) + h (1) x (0) h (2) x̃ (N − 1) + . . . + h (L − 1) x̃ (N − L + 2),

ISI from previous OFDM symbol

which can again be seen to experience inter-OFDM symbol interference from the previous
OFDM block symbols x̃ (N − 1) , x̃ (N − 2) , . . . , x̃ (N − L + 2). Let us now consider a
modified transmission scheme as follows. To each transmitted OFDM sample stream, we pad
the last Lc symbols to make the transmitted stream as follows.
x̃ (0) , x̃ (1) , . . . , x̃ (N − 1) , x (N − Lc ) , x (N − Lc + 1) , . . . , x (N − 1) x (0) , x (1) , . . . , x (N − 1) .
| {z }| {z }| {z }
Previous block Cyclic prefix Current block

Initial samples, of subject to inter OFDM symbol interference

OFDM symbol size = N/B samples

Figure 7.6 Inter-OFDM symbol interference

Observe that we are prefixing the transmitted sample block x (0) , x (1) , . . . , x (N − 1) of
the current block with the Lc samples x (N − Lc ) , x (N − Lc + 1) , . . . , x (N − 1). Further,
this prefix is cyclic in nature, since the same samples from the end of the block are being cycled
towards the beginning. Therefore, this is known as the cyclic prefix and is an important aspect
of OFDM systems. Consider now the received symbol corresponding to x (0). This is given as

y (0) = h (0) x (0) + h (1) x (N − 1) + . . . + h (L − 1) x (N − L + 1)

ISI from same OFDM symbol

The inter-symbol interference can be seen to now be from

x (N − 1) , x (N − 2) , . . . , x (N − L + 1), if Lc ≥ L − 1. Thus, with the cyclic prefix
of appropriate length, i.e., Lc ≥ L − 1, inter-OFDM symbol interference can be avoided and
inter-symbol interference is restricted to samples from the same OFDM symbol. Therefore,
the samples y (0) , y (1) , . . . , y (N − 1) are given as

y (0) = h (0) x (0) + h (1) x (N − 1) + . . . + h (L − 1) x (N − L + 1)

y (1) = h (0) x (1) + h (1) x (0) + . . . + h (L − 1) x (N − L + 2)

..
.
y (N − 1) = h (0) x (N − 1) + h (1) x (N − 2) + . . . + h (L − 1) x (N − L)

It can now be clearly seen that the output y (n) is a circular convolution between the channel
filter h (n) and the input x (n).
This can, therefore, be expressed as

[y (0) , y (1) , . . . , y (N − 1)] = [h (0) , h (1) , . . . , h (L − 1) , 0, . . . , 0] ∗N [x (0) ,

x (1) , . . . , x (N − 1)]

where ∗N denotes circular convolution of modulo N . Therefore, the output y can be written as

y = h ∗N x

Therefore, taking the DFT of y (n) at the output, we have

Y (k) = H (k) X (k) , 0 ≤ k ≤ N − 1 (7.3)

where Y (k) , 0 ≤ k ≤ N − 1, denotes the N -point DFT of y (n). Similarly, X (k) denotes
the N -point DFT of x (n). Further, observe that the samples x (n) have been generated as the
IDFT of X (n). Therefore, the DFT of the samples x (n) yields back the original transmitted
symbols X (n). The coefficients H (k) denotes the DFT of the zero-padded channel filter,

h (0) , h (1) , . . . , h (L − 1) , 0, . . . , 0 . (7.4)

(N −L)

Thus, observe that Eq. (7.3) represents the flat-fading channel across the kth subcarrier in
the OFDM system. The quantity Y (k) represents the output symbol, while H (k) denotes
the equivalent flat-fading channel coefficient. This holds true for each subcarrier k, i.e., for
0 ≤ k ≤ N − 1. Thus, the frequency-selective fading channel is converted into a group of
narrowband flat-fading channels, one channel across each subcarrier. Observe that if a single
carrier system was used, and the symbols X (0) , X (1) , . . . , X (N − 1) were transmitted
directly then the received symbol y (n) would be given as

y (n) = h (0) X (n) + h (1) X (n − 1) + . . . + h (L − 1) X (n − L + 1)

Each symbol X (n) would experience inter-symbol interference of L − 1 past symbols.

Therefore, using this novel scheme of OFDM, we have been able to totally eliminate
the inter-symbol interference arising out of the frequency-selective nature of the channel.
The set of parallel flat-fading channels can be summarized by the expressions

Y (0) = H (0) X (0)

Y (1) = H (1) X (1)

..
.
Y (N − 1) = H (n − 1) X (N − 1)

This conversion of the frequency-selective wideband channel into N narrowband flat-fading

channels is shown schematically in Figure 7.7. Also, the modified transmitter and receiver
schematics with the blocks corresponding to the cyclic prefix are given in Figures 7.8 and
7.9 respectively. Now, considering the noise at the receiver, the received symbol Y (k) can be
expressed as

Y (k) = H (k) X (k) + N (k) (7.5)

 Figure 7.7 OFDM parallel subchannels

where N (k) denotes the noise across the kth subcarrier. A simple detection scheme for
X (k) is to use the zero-forcing detector for the subcarrier as

1 N (k)
X̂ (k) = Y (k) = X (k) +
H (k) H (k)
Ñ (k)

Further, for a simplistic BPSK or QPSK-modulated transmission, the coherent or matched

filter detector can be simply obtained by multiplying with H ∗ (k), i.e., the complex conjugate
of H (k) as

H ∗ (k) Y (k) = |H (k)|2 X (k) + H ∗ (k) N (k)

N (k)

X(0) x(0) Samples at

rate B

S/P IFFT P/S Add

Symbols Demux N Mux CP To channel

X(N-1) x(N-1)
Serial to parallel N = # of Block for addition of
conversion Subcarriers cyclic prefix (CP)

Figure 7.8 OFDM transmitter schematic with CP

Also, one can employ the MMSE detector as

H ∗ (k)
X̂MMSE (k) = Y (k)
|H (k)|2 + σn2

The above equation gives the MMSE receiver across the kth subcarrier in this OFDM system.
N information
symbols

x(0) X(0)

Samples CP S/P N-pt P/S

from channel removal demux FFT mux

x(N-1) X(N-1)
Removes CP symbols To demodulate Parallel to serial
subject to inter OFDM symbol on each multiplexing
symbol interference subcarrier

Figure 7.9 OFDM receiver schematic with CP

7.2.3 Impact of Cyclic Prefix on Data Rate

Consider the transmitted samples x (n) with the cyclic prefix, as given below.

x (N − Lc ) , x (N − Lc + 1) , . . . , x (N − 1) x (0) , x (1) , . . . , x (N − 1) .
Cyclic prefix Current block

The minimum required length of the cyclic prefix is L − 1 as was described in the previous
section. Also, observe that L − 1 is the delay spread of the wireless channel. Therefore, it fol-
lows that the length of the cyclic prefix should be greater than the delay spread of the channel.
However, since the samples in the tail, i.e., x (N − Lc ) , x (N − Lc + 1) , . . . , x (N − 1) are
simply repeated in the beginning, they do not constitute any additional information. Hence, the
effect of the addition of a long CP is lost in the throughput of the system. More specifically,
the loss in efficiency can be calculated as
Cyclic prefix
Loss in effciency =
Total OFDM symbol length
L−1
=
N +L−1
L−1
=
N +L−1
However, as the block length N becomes very large, we have

L−1
lim →0
N →∞ N + L − 1

Thus, the loss in throughput approaches 0 as the number of subcarriers N increases, for a fixed
length of the delay spread L. Also observe that as the number of subcarriers N increases, the
symbol time N B increases as shown in Figure 7.10. Increasing N results in increasing OFDM
symbol time, thus restricting the ISI to a small fraction of the OFDM symbol block, i.e., the
L
fraction N is progressively smaller. However, as the block length N increases, the decoding
delay at the receiver also increases as one has to wait for arrival of the entire block of N samples
before it can be demodulated. Hence, there is a trade-off for increasing N vs decoding delay.
Now, we present another intuitive framework to understand the effect of various parameters.
As we have said previously, the duration of the cyclic prefix has to be greater than the delay
spread.

Figure 7.10 Inter OFDM symbol interference with increasing OFDM symbol time
Therefore, we need

Lc × T s ≥ T d

where Ts denotes the sample time and Td denotes the delay spread. Also, the sample time
Ts = B1 , where B is the total bandwidth of the system and Td = B1c , where Bc is the coherence
bandwidth of the system. The above condition implies
Td
Lc ≥ ,
Ts
B
=
Bs
Combining this with the earlier condition that N >> Lc for efficiency in terms of the effective
data rate, we have

B
N >> Lc ≥ .
Bc

B
This can also be recast as Bc >> N . Interestingly, this is the same condition for frequency flat
B
fading across each subcarrier since this implies that the subcarrier bandwidth N is required to
be much less than the coherence bandwidth Bc . Thus, an appropriately designed OFDM system
converts a frequency-selective fading channel into a set of parallel narrowband flat-fading
channels across the subcarriers. The example next illustrates an OFDM system design example.

7.3 OFDM Example

In this section, we consider a practical WiMAX example to illustrate the impact of the various
parameters in the design of a complete OFDM system. As already stated in the beginning,
WiMAX, which stands for Worldwide Interoperability for Microwave Access, is a prominent
4G wireless standard. The total number of subcarriers N = 256, with a bandwidth of 15.625
kHz per subcarrier. Therefore,
B
= 15.625 kHz
N

⇒ B = N × 15.625 = 256 × 15.625

= 4 MHz
Also, observe that the subcarrier bandwidth is less than the coherence bandwidth, i.e.,
Bs = 15.625 kHz << Bc = 250 kHz. Therefore, each subcarrier experiences frequency flat
fading. The OFDM symbol time without CP is
N 256
= = 64 μs.
B 4 × 106
The raw OFDM symbol time, corresponding to the N = 256 IFFT samples, is 64 μs. WiMAX
employs a cyclic prefix which is 12.5% of the symbol time. Therefore, the duration of the cylic
prefix is

Duration of cyclic prefix = 12.5% of symbol time

12.5
= × 64μs,
100

= 8 μs

Thus, the total transmitted OFDM symbol duration with cyclic prefix is 64 μs + 8 μs = 72 μs.
Also, the number of samples in the CP is
CP duration
# Samples in CP =
Sample time
8 μs
=
1/B

= 8 μs × 4 × 106

= 32

Thus, the length of the cyclic prefix Lc = 32 samples and the total number of samples is
256 + 32 = 288. This break-up of the OFDM symbol in terms of the regular samples and the
cyclic prefix is shown in Figure 7.11. Finally, the loss in spectral efficiency is
32
Loss in spectral efficiency =
288
8 μs
=
72 μs

= 11.1%

This is the loss in spectral efficiency arising because of the addition of the cyclic prefix.
 Figure 7.11 WiMAX OFDM symbol with cyclic prefix

7.4 Bit-Error Rate (BER) for OFDM

Consider the OFDM subcarrier system model given in Eq. (7.5), i.e.,

Y (k) = H (k) X (k) + N (k) (7.6)

where N (k) is the subcarrier noise obtained from the FFT of the noise samples at the output
of the receiver as

N −1
km
N (k) = n (m) e−j2π N

m=0

where N is the number of subcarriers, and n (0) , n (1) , . . . , n (N − 1) are additive noise
samples for each of the output samples y (0) , y (1) , . . . , y (N − 1). We now deduce the
statistical properties of these noise samples N (k), which are required to characterize the
BER performance of the OFDM system. Firstly, observe that the noise N (k) is the linear
combination of Gaussian noise samples n (0) , n (1) , . . . , n (N − 1). Hence, it is Gaussian in
nature. Further, the mean or expected value of N (k) is given as
N −1
km
E {N (k)} = E n (m) e−j2π N

m=0

N −1
km
= E {n (m)} e−j2π N

m=0
 2
Further, the variance σN of the noise sample N (k) is given as
2
σN = E |N (k)|2

N −1 N −1 ∗
km kl
=E n (m) e−j2π N n (l) e−j2π N

m=0 l=0

N −1 N −1
k
=E n (m) n∗ (l) e−j2π(m−l) N (7.7)
m=0 l=0

Observe that since the noise samples n (m) are independent identically distributed Gaussian
of variance σn2 , it follows that E {n (m) n∗ (l)} = 0 if m = l and σn2 if m = l . Therefore, the
above expression for the noise variance can be simplified as
N −1 N −1
k
2
σN = E {n (m) n∗ (l)} e−j2π(m−l) N
m=0 l=0

N −1
= σn2
m=0

= N σn2

Further, let us assume that each of the channel taps h (0) , h (1) , . . . , h (L − 1) is Rayleigh
fading in nature, i.e., has a complex symmetric Gaussian distribution of mean 0 and variance
1. Therefore, the channel coefficient across the kth subcarrier is given as
N −1
km
H (k) = h (m) e−j2π N

m=0

As can be seen, H (k) is a linear combination of Gaussian random variables h (k) , 0 ≤ k ≤

L − 1. Therefore, H (k) is indeed complex Gaussian i.e. has a Rayleigh fading envelope.
Further, since each h (k) is zero mean, H (k) also has mean zero. Further, identical to the
development of the noise variance above, it follows on similar lines that the channel power
gain E |H (k)|2 is
2
E |H (k)|2 = E
km
m = 0L−1 h (m) e−j2π N

L−1
2
E |h (m)|2
km
= e−j2π N

m=0

=L

Therefore, the system model in Eq. (7.6) represents a standard Rayleigh fading channel of
power gain L with receiver noise N (k) of variance N σn2 . Therefore, the average SNR is NLP2 .
σn
Hence, the BER is given by the standard expression for that of a Rayleigh fading wireless
channel as
⎛ ⎞
LP
1⎝ N σn
2
BEROFDM = 1− LP
⎠
2 2 + N σ2
n

7.5 MIMO-OFDM

MIMO-OFDM is a combination of the Multiple-Input Multiple-Output (MIMO) wireless

technology with that of OFDM, to further increase the rate in broadband multi-antenna wireless
systems. Similar to OFDM, MIMO-OFDM converts a frequency-selective MIMO channel into
multiple parallel flat fading MIMO channels. Hence, MIMO-OFDM significantly simplifies
baseband receive processing by eliminating the need for a complex MIMO equalizer. We have
already seen that the frequency-selective SISO channel is modelled as an FIR channel filter,
with the output y (n) at time instant n given as
L−1
y (n) = h (l) x (n − l) + w (n) ,
l=0

= h (0) x (n) + h (1) x (n − 1) + . . . + h (L − 1) x (n − L + 1) +w (n)

ISI from previous symbols
 Figure 7.12 MIMO OFDM transmitter schematic

where w (n) denotes the noise. Hence, a MIMO frequency-selective channel can be modelled
as a MIMO FIR filter, which can be described as
L−1
y (n) = H (l) x (n − l) + w (n)
l=0

= H (0) x (n) + H (1) x (n − 1) + . . . + H (L − 1) x (n − L + 1) +w (n)

ISI from previous symbol vectors

Therefore, the symbol vector y (n) at the time instant n is affected by inter-symbol vector inter-
ference from x (n − 1) , x (n − 2) , . . . , x (n − L + 1). This is an L-tap frequency-selective
MIMO channel. As can be seen, in a MIMO frequency-selective channel, the interference
occurs between current and previous transmit symbol vectors. In a MIMO-OFDM system,
one needs to perform the IFFT operation at each transmit antenna. The schematic figures
showing clearly the processing at the transmitter and receiver of the MIMO-OFDM system
are shown in figures 7.12 and 7.13 respectively. Hence, employing MIMO-OFDM, the MIMO
frequency-selective channel can be converted into a set of parallel flat-fading MIMO channels.
These can be described as,

ỹ (0) = H̃ (0) x̃ (0)

ỹ (1) = H̃ (1) x̃ (1)

..
.
ỹ (N − 1) = H̃ (N − 1) x̃ (N − 1)

The model across the kth subcarrier is ỹ (k) = H̃ (k) x̃ (k), where ỹ (k) and x̃ (k) are the
received and transmitted symbol vectors corresponding to the kth subcarrier, and H̃ (k) is
the flat-fading channel matrix corresponding to the subcarrier k. Each of the received vectors
ỹ (0) , ỹ (1) , . . . , ỹ (N − 1) can be processed by a simple MIMO zero-forcing receiver or a
MIMO-MMSE receiver for detection of the vectors x̃ (0) , x̃ (1) , . . . , x̃ (N − 1). The zero-
forcing MIMO receiver is given as
†
ˆ ZF (k) = H̃ (k)
x̃ ỹ (k)

−1
= H̃H (k) H̃ (k) H̃H (k) ỹ (k)

Also, the MMSE receiver for the subcarrier k of the MIMO-OFDM system is given as
†
ˆ MMSE (k) = H̃ (k)
x̃ ỹ (k)

−1
= Pd Pd H̃H (k) H̃ (k) + σw
2
I H̃H (k) ỹ (k)

where Pd denotes the data power. The channel matrices H̃ (0) , H̃ (1) , . . . , H̃ (N − 1)
corresponding to the OFDM subcarriers are given as follows. Let hu,v (k) , h̃u,v (k) denote
the (u, v)th entries of the matrices H (k) , H̃ (k) respectively. Then, h̃u,v (k) is given as the
N -point DFT of the zero-padded coefficients

hu,v (0) , hu,v (1) , . . . , hu,v (L − 1) , 0, . . . 0

(N −L)

In effect, the channel matrix H̃ (k) is the kth frequency point corresponding to the FFT of the
zero padded channel matrices [H (0) , H (1) . . . , H (L − 1) , 0r×t , . . . , 0r×t ].
 MIMO detection
for 1st subcarrier

1st receive
antenna CP S/P N pt MIMO P/S
removal demux FFT detection mux

2st receive CP S/P N pt MIMO P/S

antenna removal demux FFT detection mux

MIMO detection P/S

for 2st subcarrier demux

rth receive
antenna CP S/P N pt MIMO P/S
removal demux FFT detection mux

Figure 7.13 MIMO-OFDM receiver schematic

7.6 Effect of Frequency Offset in OFDM

OFDM divides the available wideband amongst a set of orthogonal overlapping subcarriers.
Hence, the presence of a carrier-frequency offset can introduce severe distortion in an OFDM
system, as it results in a loss of orthogonality amongst the subcarriers. Hence, the presence
of a carrier-frequency offset introduces Inter-Carrier Interference (ICI) in OFDM systems. In
this section, we characterize the effect of frequency offset on the performance of the OFDM
system. Consider a frequency offset Δf such that

Δf
=
B/N

where denotes the normalized frequency offset, normalized with respect to the subcarrier
bandwidth B/N . Corresponding to the frequency offset , the baseband received samples y (n)
are given as

N
2
1 k+
y (n) = X (k) H (k) ej2πn N + w (n) (7.8)
N
k=−N
2
To verify the above equation, set = 0. The above equation then reduces to

N
2
1 k
y (n) = X (k) H (k) ej2πn N + w (n)
N
k=− N
2

Performing now the FFT of y (0) , y (1) , . . . , y (N − 1) at the receiver, Y (l), which
corresponds to the symbol received on the l th subcarrier, is given as
1 nl
Y (l) = y (n) e−j2π N
N n

N
2
1 k nl 1 nl
= X (k) H (k) ej2πn N e−j2π N + w (n) e−j2π N
N n k=− N
N n
2

W (l)
N
2
1 k−l
= X (k) H (k) ej2πn N + W (l)
N n k= −N
2

N
2
1 k−l
= X (l) H (l) + X (k) H (k) ej2πn N +W (l)
N n
k= −N
2
, k=l
0

= X (l) H (l) + W (l)

k−l
where we have used the fact that n ej2πn N = 0 if k = l . Thus, in the absence of a carrier-
frequency offset, i.e., = 0, the system in Eq. (7.8) reduces to the earlier flat-fading OFDM
system across each subcarrier, i.e.,

Y (l) = X (l) H (l) + W (l)

Now consider the received symbols y (n) in the presence of a carrier-frequency offset

N
2
1 k+
y (n) = X (k) H (k) ej2πn N + w (n)
N
k=− N2
Therefore, the demodulated symbol Y (l) in the presence of a carrier-frequency offset is given
as
1 nl
Y (l) = y (n) e−j2π N
N n

N
2
1 k−l+ 1 nl
= X (k) H (k) ej2πn N + w (n) e−j2π N
N n k=− N
N n
2

W (l)
N
2
1 1 k−l+
= X (l) H (l) ej2πn N + X (k) H (k) ej2πn N + W (l)
N n
N
k= −N
2
, k=l n

= X (l) H (l) + W (l)

To simplify the above expression, we will use the result below

N −1 Nθ
sin
ejθn = 2
θ
ej φ̃
n=0
sin 2

where ej φ̃ is a phase factor, which does not affect the power at the output. Using the relation
above, the expression for Y (l) can be simplified as
N
2
sin (π ) 1 j φ̃l sin (π )
Y (l) = H (l) X (l) e + H (k) X (k) ej φ̃kl +W (l)
sin πN N N sin π l−N
k+
k=− N
2
, k=l
Desired signal
Inter-carrier interferenceIl

The SINR is given as

Signal power
SINR =
Interference + Noise power
Signal power
=
E |Il |2 + σn2

The desired signal power can be calculated as

! "2
2 2 sin (π )
Signal power = E |H (l)| E |X (l)|
N sin πN
For a large value of N , i.e., number of subcarriers, we have
π π
lim sin ≈
N →∞ N N
π π
⇒ N sin
→N =π
N N
Hence, the signal power for a large number of subcarriers N is given as
! "
2 sin (π ) 2
Signal Power = E |H (l)| P
π
! "2
2 sin (π )
= P |H |
π

where P = E |X (l)|2 is the power of the transmitted data symbols and

|H |2 = E |H (l)|2 denotes the average channel-power gain across each subcarrier.
The interference power is given as,

k= N2 2
2 2 2 sin (π )
E |Il | = E |X (l)| E |H (l)|
N sin π l−N
k+
k=− N2 , k=l

Setting k − l = u and letting N → ∞, we have

∞ ! "2
2 2 sin π
E |Il | = P |H | u
N sin π N
u=−∞,u=0
∞ ! "2
2 2 1
= P |H | (sin π ) u
N sin π N
u=−∞,u=0

2θ
Employing the inequality sin θ ≥ π, we have
πu 2πu/N 2u
sin ≥ =
N π N
πu
⇒ N sin ≥ 2u (7.9)
N
The interference power E |Il |2 can, therefore, be approximated as
∞ ! "2
2 2 2 1
E |Il | ≤ P |H | (sin π )
2u
u=−∞, u=0

"2∞ !
2 1
2
= P |H | 2 (sin π )
2u
u=1
∞ ! "2
1 2 2 1
= P |H | (sin π )
2 u=1
u
π2
6

π2
= P |H |2 sin2 π
12

= 0.822P |H |2 sin2 π

Hence, the SINR in the presence of carrier-frequency offset of is given as

2
P |H |2 sin π
π
SINR = 2 2
(7.10)
0.822P |H | sin π + σn2

Example 7.1 illustrates the effect of ICI in reducing the SINR at the output of the OFDM
receiver.

EXAMPLE 7.1

Consider |H |2 = 1, and data power P = 10 dB with noise power σn2 = 0 dB. Derive the
SNR/SINR with and without a carrier-frquency offset of = 5% = 0.05 in a WiMAX
system.

Solution: If the carrier-frequency offset = 0, the SNR at the receiver is given as

P |H |2
SNR = = 10 dB
σn2
Considering now a carrier-frequency offset of = 0.05 and employing the expression in
Eq. (7.10), the SINR is given as
2
10 × sinπ0.05
π0.05
SINR = = 8.25
0.82210 × sin2 (π0.05) + 1
Therefore, the reduction in SINR is a factor of 1.75 or, basically, 17.5%. Further, the WiMAX
subcarrier bandwidth is 15.625 kHz. The absolution value of the frequency offset Δf , is
therefore,

Δf = Bs = 0.05 × 15.625 ≈ 0.78kHz

Further, considering a typical 4G carrier frequency of 2.4 GHz, the carrier-frequency offset as
a fraction of the centre frequency is
0.78 × 103 1
9
= × 10−6
2.4 × 10 3
This is also termed as 0.33 ppm or parts per million with respect to the carrier frequency since
1
1 × 10−6 = 1000000 is equivalent to 1 part per million.

7.7 OFDM–Peak-to-Average Power Ratio (PAPR)

The Peak-to-Average Power Ratio (PAPR) is a critical problem in OFDM systems, which needs
to be handled effectively in order to limit the distortion at the receiver. Consider a non-OFDM
or single-carrier system with BPSK modulated symbols. For example, let the symbol stream
x (0) , x (1) , x (2) , . . . be given as +a, −a, +a, . . . and so on. The power in each symbol
equals a2 . Further, also observe that this is the peak power at any given instant of time.
Therefore, we have

Peak power = Average power = E |x (k)|2 = a2

Thus, since the peak and average power are equal, the peak-to-average power ratio, or PAPR,
is given as
Peak power
PAPR =
Average power

a2
=
a2
= 1 = 0 dB

X(0) x(0)

IFFT
N

X(N-1) x(N-1)
N = # of
subcarriers

Figure 7.14 OFDM subcarrier loading

The above relation clearly shows that there is no significant deviation of the instantaneous
power level from the mean power level. Now, consider an OFDM system in which the different
information symbols X (0) , X (1) , X (2) , . . . given by +a, −a, +a, . . . for instance, are
loaded onto the subcarriers. This is shown schematically in Figure 7.14. The actual samples
transmitted over the wireless channel are x (0) , x (1) , x (2) , . . . , x (N − 1), which are the
IFFT samples of the information symbols X (0) , X (1) , X (2) , . . . , X (N − 1). Consider
the kth IFFT sample x (k) given as

N −1
1 ki
x (k) = X (i) ej2π N
N
i=0

where X (i) denotes the information symbols. The average power in the symbols is given as

Average power = E |x (k)|2

 N −1
1 2
E |X (i)|2 E
ki
= ej2π N
N
i=0
1

N −1
1
= 2 E |X (i)|2
N
i=0

N −1
1
= 2 a2
N
i=0

1 2 a2
= 2a N =
N N
As can be seen from the above equation, the average power of transmission is aN . The peak
2

power can be found as follows. Observe that the peak of the OFDM sample arises for all
symbols X (i) = +a or X (i) = −a. This can be verified as follows.
N −1
1 ki
|x (k)| = X (i) ej2π N
N
i=0

N −1
1 ki
≤ X (i) ej2π N
N
i=0

N −1
1 ki
= |X (i)| ej2π N
N
i=0 a
1

N −1
1
= a
N
i=0

=a

Therefore, the peak power is given as a2 . Hence, the peak-to-average power ratio in an OFDM
system is given as

a2
OFDM PAPR = =N
a2 /N
 OFDM CCDF of PAPR for Various Subcarriers N
1

0.9

0.8

0.7
Probability (PAPR > x)
0.6

0.5

0.4

0.3

0.2

0.1

4 5 6 7 8 9 10 11 12
PAPR in dB
Figure 7.15 OFDM PAPR for various number of subcarriers N

From the above expression, it can be seen that the peak-to-average power ratio in an OFDM
system is N , which is significantly higher compared to that of the single-carrier system, which
is 1. Further, interestingly, this PAPR rises with N , i.e., the number of subcarriers. Larger the
number of subcarriers, larger is the PAPR. This high PAPR of the OFDM arises because of the
IFFT operation. The data symbols across the subcarriers can add up to produce a high peak
valued signal as seen above. For instance, in an OFDM system with 512 subcarriers and BPSK
modulation, the PAPR at the output can be as high as 10 dB. The PAPR of an OFDM system
is characterized using the CCDF, i.e., the complementary cumulative distribution function.
The CCDF F X (x) of a random variable X is given as the probability that X > x, expressed as

F X (x) = Pr (X > x)
Naturally, the CCDF is related to the CDF, i.e., cumulative distribution function FX (x) of X as

FX (x) = Pr (X ≤ x)

= 1 − Pr (X > x)

= 1 − FX (x)

The CCDF of the PAPR then shows the probability that the PAPR, which is a random quantity,
exceeds a particular threshold. A plot of the CCDF of the PAPR for various values of N , the
total number of subcarriers, is shown in Figure 7.15.
The impact of PAPR on the OFDM system hardware can be understood as follows. Every
communication system has a receiver amplifier, which serves to amplify the amplitude of the
receive signal, in order to boost its strength. However, the characteristic of the amplifier is
linear only for a limited amplitude range of the signal. Typically, the amplifier operates around
a bias point, as shown in Figure 7.16, which is roughly around the average power of the signal.
As long as the signal amplitude is restricted to the dynamic range of the amplifier around this
bias point, for which the amplifier characteristic is linear, there is no noninear distortion at the
output. However, in the case of OFDM, since the peak power deviates significantly from the
average power, there is a high chance that the signal crosses into the voltage region outside the

Figure 7.16 Nonlinear amplifier characteristic

dynamic range of the amplifier, thus resulting in a noninear distortion of the received signal.
This nonlinear effect, arising out of amplifier saturation, leads to loss of orthogonality of the
subcarriers and inter-carrier interference. The net result is a poor decoding performance and a
rise in the bit-error rate. A slightly modified OFDM technique, which can significantly reduce
the PAPR, is SC-FDMA, which is described next.

7.8 SC-FDMA

SC-FDMA, which stands for Single-Carrier Frequency Division for Multiple Access, can
be employed to reduce the peak-to-average power ratio in an OFDM system. Consider the
following hypothetical modification of the OFDM transmitter, shown in Figure 7.17, by the
insertion of an N -point FFT block before the N -point IFFT block. It can then be seen that the
FFT and the IFFT cancel the effect of each other and the net output is the exact input symbol
stream, i.e., corresponding to a single-carrier system. This drastically reduces the PAPR, since,
as seen previously, the PAPR of a single-carrier system is 0 dB. However, instead of using an
N -point FFT, one can use an M -point FFT, where M < N , to reduce the PAPR, while still
retaining the properties of the OFDM system. This proposed SC-FDMA schematic is shown in
Figure 7.18. Hence, introduction of the M -point FFT in SC-FDMA significantly reduces the
PAPR of the system. This is the central principle of SC-FDMA.
Transmitted
Symbols

Symbol S/P N pt N pt P/S Add

Stream Demux FFT IFFT Mux CP

Figure 7.17 Hypothetical modification of OFDM transmitter

Transmitted
symbols

Symbol S/P M pt Subcarrier N pt P/S Add

stream demux FFT mapping IFFT mux CP

2 new blocks in SC-FDMA

transmitter schematic
Figure 7.18 SC-FDMA transmitter schematic
7.8.1 SC-FDMA Receiver
The SC-FDMA receiver schematic is shown in Figure 7.19. The SC-FDMA receiver
incorporates two new blocks compared to the OFDM receiver. The purpose of these additional
blocks can be described as follows. After the N -point FFT operation at the receiver, the signals
are equalized across all the subcarrires, to remove the effect of the fading-channel coefficient
across the subcarriers. Following the above operation, they are demapped from the subcarriers,
which are N in number, to the original FFT block size of M . Finally, the M -point FFT is
performed on these samples to generate the symbol stream.

Equalizer
CP S/P N pt M pt P/S Symbol
and
removal demux FFT IFFT mux stream
demapping

2 new blocks in SC-FDMA

receiver schematic
Figure 7.19 SC-FDMA receiver schematic

7.8.2 Subcarrier Mapping in SC-FDMA

Subcarrier mapping, in which the M samples at the output of the M -point FFT are mapped
to the N subcarriers, is a key operation in SC-FDMA, a block representation of which
can be seen in the SC-FDMA transmitter schematic in Figure 7.18. The various possible
SC-FDMA subcarrier mappings are illustrated through the following example. Consider
M = 4 SC-FDMA symbols and N = 12 subcarriers. Let x (0) , x (1) , x (2) , x (3) denote
the symbols and X (0) , X (1) , X (2) , X (3) denote the corresponding M = 4-point FFT
samples which are to be loaded onto the subcarriers. Let the number of subcarriers be N = 12.
In Interleaved FDMA (IFDMA) shown in Figure 7.20, the samples X (i) are interleaved with
zeros. In LFDMA, which is also employed in the uplink of the 4G mobile standard LTE,
the samples are loaded as a block onto the subcarriers, with appropriate zero padding. This
is shown in Figure 7.20. Post this subcarrier mapping, the rest of the procedure prior to
transmission proceeds as shown in the SC-FDMA transmitter block.
 Interleaving Zeroes

I-FDMA

X(0) 0 0 X(1) 0 0 X(2) 0 0 X(3) 0 0

X(0) X(1) X(2) X(3) 0 0 0 0 0 0 0 0

L-FDMA
Employed in LTE Uplink Zero-Padding

Figure 7.20 SC-FDMA various subcarrier mapping schemes

1. Consider a three-tap wireless channel [h (0) , 0, h (2)] (i.e., h (1) = 0) with each tap a unif-
norm Rayleigh fading channel coefficient. Employ the WSSUS channel assumption. Let an
OFDM system with 256 subcarriers in the 2.1 GHz band be implemented over this channel
with delay spread Td = 8 μ s. The DFT operation at the receiver is given as

N −1
1 2πlk
F (k) = √ f (l) e−j N

N l=0

(a) With a bare minimum cyclic-prefix duration, the system has a useful bit rate of 8.6486
Mbps for QPSK transmission. What is the bandwidth of the system?
(b) For a cyclic prefix of 12 μs duration, what is the loss in spectral efficiency for this
system?
(c) Consider the channel coefficients H (0) and H (64) at the 0th and 64th subcarriers
respectively. What is their joint distribution?
(d) It is not difficult to see that every OFDM system must satisfy a very important constraint
that the wireless channel is constant during the transmission of the total OFDM symbol
(i.e., useful plus cyclic). Assuming a system with 20% loss of spectral efficiency and
an extreme vehicle moving at 200 km/h, what is the maximum possible number of
subcarriers?
2. OFDM System Design Consider an OFDM system with N = 256 subcarriers over a
bandwidth of B = 5 MHz. Let the corresponding frequency selective fading channel have
an impulse response with 3 multipath components at delays of [0, 0.40, 1.0] μs, with each
component of −3 dB average power. Noise power at the receiver is σv2 = 3 dB. Assume
that the IFFT and FFT operations are given respectively as

N −1 N −1
1 j2πkn/N
x (n) = X (k) e , X (k) = x (n) e−j2πkn/N
N
k=0 n=0

In that case, the system model after FFT at the receiver becomes Y (k) = H (k) X (k)
+ V (k), where V (k) is the FFT of the AWGN. Answer the following questions.
(a) Describe the time-domain model of the above frequency-selective channel.
(b) What is the minimum number of samples required in the cyclic prefix?
(c) What is the duration of this minimum cyclic prefix?
(d) If the actual cyclic prefix employed is twice the minimum length required with QPSK
modulated subcarriers, what is the effective bit rate of the OFDM system?
(e) What is the bit-error rate across each subcarrier if the total transmit power of 70 dB is
distributed equally across the subcarriers?
(f) What is the reduction in SNR across each subcarrier in the presence of a 5% carrier-
frequency offset relative to the subcarrier bandwidth?
3. OFDM System Design Consider a mobile OFDM profile with N = 512 subcarriers over a
bandwidth of B = 10MHz. Let the corresponding frequency-selective fading channel have
an impulse response with 4 multipath components at delays of 0 μs, 0.60 μs, 1.2 μs, 2.4 μs,
with each component of average power- −3 dB. Noise power at the receiver is σn2 =
3 dB per sample (i.e., before FFT). Assume that the IFFT and FFT operations are given
respectively as,

N −1 N −1
1 j2πkn/N
x (n) = X (k) e X (k) = x (n) e−j2πkn/N
N
k=0 n=0

In that case, the system model after FFT at the receiver becomes Y (k) = H (k) X (k) +
W (k), where W (k) is the FFT of the AWGN. Answer the following questions.
(a) Describe the time-domain model of the above frequency-selective channel, i.e., what
 is the length of the channel impulse response filter and what is the number of nonzero
channel taps?
(b) What is the minimum number of samples required in the cyclic prefix in this system?
(c) What is the duration of this minimum cyclic prefix?
(d) If the actual cyclic prefix employed is three times the minimum length required with
QPSK-modulated subcarriers, what is the effective bit rate of the OFDM system?
(e) What is the QPSK bit-error rate across each subcarrier if the total transmit power of
80 dB is distributed equally across the subcarriers?
(f) What is the reduction in SNR across each subcarrier in the presence of an 8% carrier-
frequency offset relative to the subcarrier bandwidth?
4. Consider an OFDM system with total passband bandwidth B = 5 MHz with N = 512
subcarriers. The channel has a maximum delay spread of Td = 4 μs. Answer the questions
that follow.
(a) What is the symbol time of a corresponding single-carrier system?
(b) What is the sample time of the OFDM system and the raw symbol time without the
cyclic prefix?
(c) What is the minimum number of samples required in the cyclic prefix?
(d) If the length of the cylic prefix is twice the required minimum calculated above, what
is the total OFDM symbol time?
(e) What is the loss in efficiency due to the overhead of the cyclic prefix?
(f) If the modulation employed is 16-QAM, what is the effective bit rate of the above
OFDM system?
(g) At a carrier frequency of fc = 2.4 GHz, what is the maximum possible velocity of a
mobile for the system to be able to function?
5. Alamouti Coded OFDM Consider an Alamouti-coded OFDM system with two transmit
and one receive antennas. Consider symbols X1 (k) , X2 (k) loaded onto transmit antennas
1,2 respectively on the kth subcarrier of the OFDM system in the first OFDM symbol.
Consider a frequency-selective channel with channel taps h (0) , h (1) , . . . , h (L − 1).
Answer the questions that follow.
(a) Describe the operation of the Alamouti-coded OFDM system above, i.e., the
transmission on each subcarrier and the decoding operation at the receiver.
(b) If the transmit power is P per subcarrier, noise power σn2 , find the instantaneous receive
SNR expression for each symbol.
 (c) Considering the channel taps to be IID Rayleigh random variables of average power
unity, derive the average Symbol Error Rate (SER) expressions for BPSK and 16-PSK
modulation.
(d) Compute the average SER above for both BPSK and 16-PSK for P = 35 dB, number
of subcarriers N = 128, L = 5, and noise power σn2 = −3 dB.
(e) Considering the channel taps to be independent Rayleigh random variables of average
power σl2, 0 ≤ l ≤ L − 1, derive the average SER expressions for BPSK and 16-PSK
modulation.
(f) Compute the average SER above for both BPSK and 16-PSK for P = 35 dB, number
of subcarriers N = 128, σl2 = l dB for 0 ≤ l ≤ 4 and noise power σn2 = −3 dB.
6. Consider a multipath channel with L i.i.d Rayleigh faded taps h(i), 0 ≤ i ≤ L − 1, each
distributed as the symmetric complex Gaussian CN (0, 1). Let H(u), H(v) denote the
complex channel coefficients corresponding to subcarriers u, v respectively in an OFDM
system with N subcarriers and bandwidth B . Consider uncorrelated scattering and answer
the questions that follow.
(a) Express H(u), H(v) in terms of the channel taps.
(b) Compute the auto-correlation R (0) corresponding to the the complex channel
coefficient across each subcarrier.
(c) Derive the exact expression for the correlation between the subcarrier coefficients
H(u), H(v).
(d) Derive a suitable bound for the correlation between the subcarrier coefficients
H(u), H(v) in terms of |u − v | and N .
(e) Derive an expression for the number of subcarriers per coherence bandwidth Bc in the
above system. Denote this by Nc.
(f) Let |u − v | = αNc . Derive a bound for the correlation between the subcarrier
coefficients H(u), H(v) as a function of R (0) and α. From this, deduce the rate at
which the correlation is decreasing as a function of α.
7. Consider an OFDM system with N = 4 subcarriers and L = 2 tap wireless channel with
channel coefficients h(0), h(1). Let noise power per subcarrier (i.e., after the FFT at the
receiver) be σ 2 = 3 dB. Let the total transmit power be PT . Answer the questions that
follow.
1
(a) Consider h(0) = h(1) = 2 and derive the complex channel coefficients across all the
subcarriers.
(b) Consider h(0) = h(1) = 12 , PT = 18 dB with equal power allocation and derive the
instantaneous BER across each subcarrier for BPSK modulation.
(c) Consider h(0) = h(1) = 12 , PT = 18 dB and derive the optimal power allocation for
each subcarrier.
(d) Consider coefficients h(0), h(1) to be Rayleigh fading i.i.d. with average power unity
and derive the average BER as a function of PT , σ 2 for BPSK modulation with total
power PT and equal power allocation across subcarriers.