PEAK-TO-AVERAGE

CHAPTER 5 POWER RATIO

O ne of the major drawback of OFDM system is high PAPR of transmitted signal, which
causes serious degradation in performance when a non-linear HPA is used. Therefore, it is
compulsory to use a proper PAPR reduction scheme at the transmitter. In this chapter, the detailed
analysis of PAPR, its definition, distribution and different methods of PAPR reduction are given. One
new method of PAPR reduction (combination of pre-coding and clipping) is proposed. It is also proposed
that MBH window can be used for the generation of pre-coding matrix. Also, a new algorithm of clipping,
based on iterative clipping concept, has been proposed.

5.1 INTRODUCTION

High PAPR in OFDM is observed due to large dynamic range of its symbol waveforms. An
OFDM signal consists of a number of independently modulated subcarriers, which can give a large
PAPR, when added up coherently. When N signals are added with the same phase, they produce a peak
power that is N times the average power. Most radio frequency communication systems have a high
power amplifier (HPA) in the transmitter to obtain sufficient transmit power. This high PAPR forces the
HPA to have a large back-off in order to ensure linear amplification of the signal, which significantly
reduces the efficiency of the amplifier. On the other hand, if an amplifier work with nonlinear
characteristics, it will cause undesired distortion of OFDM signal (like in-band distortion and the out-of-
band radiation) and degrades the BER performance of the system [122]. Thus, the PAPR reduction is one
of the most important research interests for the OFDM systems. In the absence of PAPR-reduction
techniques, the signal is clipped, making the receiver prone to errors.

5.2 DEFINITION OF PAPR

For a continuous time baseband OFDM signal, the PAPR is defined as the ratio between the
maximum instantaneous power and its average power. If ( ) is a transmitted baseband OFDM signal,
then PAPR is defined as [122]
max [| ( )| ]
[ ( )] = ,

(5.2.1)
[103]
where, is the average power of ( ) and is useful duration of an OFDM symbol. For a discrete
OFDM signal, the PAPR is computed from time oversampled OFDM signal as

max [| ( )| ]

[ ( )] = ,
[| ( )| ]
(5.2.2)
where, [. ] denotes the expectation operator and is total number of sub-carriers. The PAPR of pass
band OFDM signal is approximately twice that of baseband PAPR [122].

5.3 DISTRIBUTION OF PAPR

To design and develop an effective PAPR reduction technique, it is very important to accurately
identify the distribution of PAPR in OFDM systems. The distribution of PAPR plays an important role for
the design of whole OFDM system. The distribution of PAPR can be used in determining the proper
output back-off of HPA to minimize the total degradation. It can be used directly to calculate the BER
and to estimate the achievable information rates [101, 114, 121, and 122]. The distribution of PAPR is
often expressed in terms of Complementary Cumulative Distribution Function (CCDF).

In probability theory and statistics, the CCDF describes the probability that a real-valued random
variable X with a given probability distribution will be found at a value greater than or equal to x [9]. The
CCDF can be used to estimate the bounds of the PAPR and is generally used as performance evaluation
parameter for most of the PAPR reduction schemes. A lot of work has been reported in the literature on
the determination of bounds and the CCDF of the PAPR distribution in OFDM systems. According to
[55], the upper bound of PAPR in M-QAM OFDM systems can be derived as

3 3 1
[ ( )]
1 1
(5.3.1)
Using conventional analysis, the CCDF of PAPR in OFDM systems for small value of sub-
carriers N can be expressed as [101] -

{ > } = 1 (1 ) , (5.3.2)

where, is the given reference level.

[104]
However, the (5.3.2) does not fit well in OFDM systems with a very large N. An empirical approximation
of the CCDF of PAPR in OFDM systems has been given in [101] as -

{ } = 1 (1 ) .
> (5.3.3)

But this expression lacks theoretical justification and also yields discrepancies with the results for a large
N. Recently, a more accurate analytical expression of the PAPR distribution is given by Tao Jiang et.al
in [121] by assuming both equal and unequal power distribution. Their work was based on the extreme
value theory for Chi-squared-2 process. They have given the expressions of CCDF for both types of
power distribution i.e., equal and unequal. When all active sub-carries allocated equal power, the CCDF
can be approximated as [121] -

{ > } 1 2 , (5.3.4)

where, 0 /

And when unequal power is allocated to different sub-carriers, the CCDF of PAPR in OFDM systems can
be approximated as [121] -

{ > } 1 2 ,


(5.3.5)
where, is the power allocated to sub-carrier and = /2, if the sub-carrier at DC is
inactive, otherwise =( 1)2 and denotes the number of active sub-carriers. The
average transmitted power is denoted by .

5.4 DIFFERENT PAPR REDUCTION TECHNIQUES

Several PAPR reduction techniques are included in Chapter-2. In this section, the popular and
frequently used techniques are discussed in detail.

5.4.1 Non-Linear Companding

Non-linear companding is a well known PAPR reduction method, which do not have any
restriction on system parameters like number of sub-carriers, frame format, and constellation type. Also, it

[105]
has good PAPR reduction capability with no bandwidth expansion and low computational complexity.
The other advantage of companding is that the signal can be recovered at the receiver through inverse
companding transform. A lot of companding techniques are available in the literature [56, 123, 134, and
136]. The exponential companding (EC) is most popular companding techniques. The companding
function ( ) of EC transform is given as [123] -

( ) = ( ) 1 exp

(5.4.1.1)
where, ( ) is the signum function, d is the degree of companding scheme and is the variance of
input signal applied for companding. The positive constant determines the average power of output
signal. In order to keep the input and output average power at same level, the is set as-

= [| ( )| ] [ 1 exp( | ( )| )] (5.4.1.2)

At the receiver side, the inverse function of ( ) is used in the de-companding operation which is
defined as

( )= ( ) 1

(5.4.1.3)

5.4.2 Selective Mapping (SLM)

In SLM, the input data sequences are multiplied by each of the phase sequences to generate
alternative input symbol sequences. Each of these alternative input data sequences are then applied to
IFFT operation, and then the one with the lowest PAPR is selected for transmission [95]. A block diagram
of SLM techniques is shown in Figure-5.4.2.1. The input data is partitioned into a data block Y of length
( ) ( )
N. Then these data block is multiplied element by element with phase sequences =
( ) ( )
[ , , , ,, , ] , = 1, 2, , , resulting into U modified data blocks =
[ , , , , , , ] , where , = . , , = 0, 1, , 1 . After that, the N-point IFFT
( )
of each data block is taken, the resulting OFDM signal is given as

[106]
 1 2
( )= , exp , = 0, 1, , 1 = 1,2, , 1

(5.4.2.1)
Among the OFDM data blocks ( ) ( = 0,1, , 1), only one with the lowest PAPR is selected for
transmission and the corresponding selected phase factor , also transmitted to receiver as side
information. For implementation of SLM OFDM systems, the SLM technique needs U- IFFT operation
and the number of required bits as side information is for each data block. Therefore, the ability
of PAPR reduction in SLM depends on the number of phase factors and the design of the phase factors.
The major drawback of SLM method is that it is more computationally complex and less bandwidth
efficient (side information is required). Therefore, a lot of work has been suggested as a modified SLM to
reduce the computational complexity [24] and to reduce or to remove the side information transmitted
[118].

Figure-5.4.2.1: Block diagram of selective mapping (SLM) technique for PAPR reduction

5.4.3 Partial Transmit Sequence (PTS)

In PTS method, the original frequency-domain data sequence is divided into multiple disjoint
sub-blocks, which are then weighted by a set of phase sequences to create a set of candidates. Finally, the
candidate with the lowest PAPR is chosen for transmission [116]. A block diagram of PTS techniques is
shown in Figure-5.4.3.1. The input data block in Y is divided in to disjoint sub-blocks, which are

[107]
 ( )
represented by the vectors { , = 0,1, , 1}. The input data block Y can be written in terms of
( )
as-

( )
= , = 0, 1, 2, , 1 ,

( )
where, =[ , , , ] = 0 (5.4.3.1)

( )
After that, the sub-blocks are transformed into time-domain partial transmit sequences by taking
the IFFT of length N. These partial transit sequences can be written as

( ) ( )
=[ ]= , = 0, 1, 2, , 1 (5.4.3.2)

( )
These partial sequences are then independently rotated by phase factors ={ = ,
= 0,1, , 1}. The rotated partial sequences are then optimally combined to obtain the OFDM
signals with lowest PAPR. The time domain signal after combining is given by

( )
= ,

(5.4.3.3)

Figure-5.4.3.1: Block diagram of partial transmit sequence (PTS) technique for PAPR reduction

[108]
There are two main issues of any PTS scheme: to reduce the computational complexity for searching the
optimal phase factors and to reduce the overhead by minimizing the side information. Suppose that there
are phase angles to be allowed, thus can has the possibility of different values. Therefore, there
are alternative representations for an OFDM symbol. The search complexity increases exponentially
with the number of sub-blocks . To reduce the search complexity and overhead (by reducing/avoiding
the usage of side information), various techniques have been suggested in the literature [4]. These
methods achieve significant reduction in search complexity with marginal PAPR performance
degradation. In 2007, R. J. Baxley et.al [96] gave a useful comparison between PTS and SLM techniques.
It has been shown that the PTS outperforms SLM in terms of PAPR reduction at the cost of increase side
information. Also the complexity of PTS is found less than SLM.

5.5 PROPOSED METHOD OF PAPR REDUCTION

Out of these available techniques, the clipping is the simplest method for the reduction of PAPR
by simultaneously increasing the average value and minimizing the peak value. On the other hand, the
pre-coding has also been considered as a best among all these techniques, because it improves the PAPR
without increasing much complexity and destroying the orthogonality between subcarriers. The pre-
coding also improves the BER in comparison to normal OFDM system because of diversity gain obtained
due to the spreading of data symbol on more than one subcarrier.

With the understanding on clipping and pre-coding techniques, an idea emerged to combine the
philosophy of pre-coding and clipping. Therefore, an attempt is made in this study to reduce the PAPR
without increasing roll-off factor by introducing clipping after pre-coding. This hybrid technique shows
good results, because of first pre-coding makes the envelope almost constant and then clipping reduces
the peak value to any desire level. This dual operation of pre-coding and clipping provides better PAPR
than both conventional pre-coding and clipping method. With the proposed method, the hardware
complexity is also similar as in the case of only pre-coding technique. The proposed technique gives good
result because the clipping noise of combined scheme would be less than that of only clipping.

For the clipping purpose, a new clipping algorithm is also proposed which is based on averaging
of high amplitude samples. This clipping algorithm can be called as soft clipping or iterative clipping.
This algorithm can be applied both with normal OFDM system (without pre-coding) and with pre-coded
OFDM system. The performance of proposed method has been compared with EC method in normal

[109]
OFDM system and it has been shown that this clipping method performs better than conventional
clipping.
In the pre-coding scheme, the design of pre-coding matrix plays an important role. The
performance of pre-coding matrix on PAPR reduction depends on the pulse shape used. In [75] and
[105], the performance of raised cosine (RC) and square root raised cosine (SQRC) pulse shapes has been
evaluated. In this chapter, a new pulse shape which is derived from pulse shape given by J. K. Gautam
et.al in 1996 [58] has been proposed for generating pre-coding matrix. The performance of proposed
pulse shape is better than RC and SQRC pulse shapes. The contribution of work for PAPR reduction can
be summarized as:

a) Proposed a new method which is a combination of Pre-coding and Clipping.

b) Proposed a new pulse shape for the generation of Pre-coding matrix.
c) Proposed a new algorithm for clipping operation.

5.5.1 OFDM System Model with Pre-coding and Clipping

In this study, a discrete time baseband OFDM system with N orthogonal sub-carriers has been
considered. It consists of transmitter, channel and receiver blocks which are described below.

Figure-5.5.1.1: Block diagram of OFDM system with pre-coding and clipping

[110]
5.5.1.1 Transmitter model

The discrete time pre-coding-based OFDM system with N sub-carriers, as given in [105], is
shown in Figure-5.5.1.1. This OFDM system is similar to the system shown in Chapter-2 (Figure-2.2.1),
except the pre-coding and clipping block at the transmitter and inverse per-coding block at the receiver.
At the transmitter, the N complex data symbols ( , = 0, 1, 2, , 1) (either M-PSK or M-QAM
modulated) with data rate (1/ ) are converted from serial-to-parallel and fed to the pre-coder block. In
the pre-coder block, N input symbols are multiplied with a pre-coding matrix of size Z x N resulting
into an output symbol of length Z.
Consider an input vector , obtained after encoding (M-PSK or M-QAM), applied to pre-coding
process
=[ , , ,, ] (5.5.1.1)

where [. ] denotes the matrix transpose. The pre-coding process (matrix) transforms this vector into a
new vector of length , obtained as

=[ ] [ ] =[ , , ,, ] (5.5.1.2)

where, is the pre-coding matrix, defined as [105]-

p 0 ,0 p 0 ,1 . . . p 0 , N 1
p p1,1 . . . p1, N 1
1,0
. . . .
P (5.5.1.3)
. . . .
. . . .

p Z 1,0 p Z 1,1 . . . p Z 1, N 1

where, , is the element of pre-coder matrix , = + is the total number of sub-carriers required
after pre-coding, and is the extra sub-carriers (overhead) used with a condition 0 . When no
pre-coding is used the matrix reduced to identity matrix of size and without overhead. After
multiplication with pre-coding matrix, the resultant pre-coded symbols are fed to the IFFT block. After
IFFT, the guard band in the form of cyclic prefix is added between consecutive OFDM symbols. Thus the
baseband pre-coded OFDM signal with cyclic prefix can be written as [105]

[111]
 (, )= ,

where, (, )= (, + ) 0 (5.5.1.4)

where, (= ) is the useful duration of one OFDM symbol, (= ) is the length of cyclic prefix,
G is the number of samples in guard band (cyclic prefix duration), and , = 0, 1, 2, , 1 is
the output of pre-coder block and it is defined as

= , , = 0,1, , 1

(5.5.1.5)
where, is the complex data symbol of data rate 1/ and , is the element of pre-coding matrix .
After IFFT, the pre-coded OFDM signal is given to clipping block, where the signal is clipped either
using conventional clipping method or using clipping algorithm defined in next sub-section. The
conventional clipping operation is defined as [101 and 135]

| ( )|, | ( )| <
| ( )| = (5.5.1.6)
, | ( )|

where, A is a preset clipping level. The normalized clipping level is called clipping ratio and is defined as

= (max [ ( )]) (5.5.1.7)

When = 1, it means no clipping and as decreases from 1 to 0.9, 0.8. clipping increases.

5.5.1.2 Receiver model

The tap-delay line model with -path is considered for multipath fading channel. After
considering the effect of multipath fading channel, the samples of received signal can be expressed as

(, )= ( ) (, )+ ( , ), = 0,1, , 1

(5.5.1.8)

[112]
where, ( ) is the impulse response of multipath fading channel with path gains { ( ): = 0, 1, ,
1}, is the path delay of path, and ( , ) is a zero-mean, unit variance complex Gaussian noise.
After discarding first G sample of the received signal and taking Z-point FFT, the output of FFT block is
given as
= + , = 0, 1, , 1 (5.5.1.9)

where, the term is the channel response to the subcarrier frequency and is the AWGN term in
frequency domain. To compensate the fading affect of channel, one-tap equalizer is used and each
element of vector is multiplied by an equalized gain factor , the output of equalizer may be written
as
= + (5.5.1.10)
where, is defined as

| | +
= (5.5.1.11)

The output of equalizer is now multiplied with matrix , where, is the Hermitian transpose of .
After multiplication, a vector of length N is obtained, given as -

= , = 0,1, , 1 (5.5.1.12)

This vector is then given to decoder to recover transmitted symbol (M-PSK / M-QAM).

5.5.2 Proposed Pulse Shape for Pre-coding Matrix Generation

The performance of pre-coding method depends on the time domain function used for generating
pre-coding matrix. In this section, an analysis of PAPR of pre-coded OFDM system is given and then a
new pulse shape is proposed. The PAPR of the transmitted signal ( , ) can be defined as follows

max[| ( , )| ]
=
[| ( , )| ]
(5.5.2.1)
where, | ( , )| is the instantaneous power and [| ( , )| ] is average power of the pre-coded OFDM
signal ( , ).

[113]
 At the receiver, the guard band samples are removed before demodulating the OFDM using FFT,
therefore, the guard band is ignored in PAPR analysis included in this section. The transmitted OFDM
signal ( , ) (without CP) can be written as follows

(, )= , 0

(5.5.2.2)
After substituting the value of

(, )= , , 0

(5.5.2.3)

After interchanging the summation

(, )= , , 0

( )

(5.5.2.4)

(, )= ( ), 0

(5.5.2.5)

where, ( ) is a set of time limited function, used to generate pre-coding matrix and is defined as

, , 0

( )= ,


0,

= 0, 1, , 1 (5.5.2.6)

[114]
The upper bound of PAPR of pre-coded OFDM signal ( ) at any time t can be written as

1 2
( ) , ,

( )
(5.5.2.7)
The maximum value of PAPR is given as [105]

= max ( ),

(5.5.2.8)
After substituting the value of ( ) from (5.5.2.7) in (5.5.2.8)

1 2
= max , ,

( )
(5.5.2.9)
After replacing the inner summation term by ( ), it becomes

1
= max | ( )| ,

(5.5.2.10)

where, it has assumed that | | = 1, = 0, 1, , 1, which represents the average energy per
transmitted symbol. From above expression (5.5.2.9 and 5.5.2.10), it is clear that the PAPR of the pre-
coded OFDM signal depends on the number of sub-carriers and the entries of the pre-coding matrix.
Since, the number of sub-carriers (N) is fixed for a particular application; one can reduce the PAPR by a
proper selection of the pre-coding matrix . It is also visible that PAPR depends now on the sum of N
different functions ( ) in time interval 0 to . The maximum value of (5.5.2.10) is reduced, if the
peak amplitudes of the N functions ( ), = 0, 1, , 1 do not occur at the same time instant
within the time interval 0 to Tu. For selecting different functions ( ) which satisfied this condition,
choose one basic function ( ) and then shift this in cyclic order to generate other function. This method
is suggested by S. B. Slimane [105]. Later on, in the year 2010, M. J. Hao [75] gave the condition for

[115]
choosing optimal pre-coding matrix which provides minimum error probability. Hence, if ( ) is the
basic function used, then the element of pre-coding matrix can be determined from the basic
function ( ) in the following way as

1
, = ( ) ,

(5.5.2.11)
After solving (5.5.2.11) -
1
, = (1) ,

where, ( ) is the Fourier transform of function ( ). (5.5.2.12)

Two window functions namely Square root Raised Cosine (SQRC) [105] and Modified Bartlett-Hanning
(MBH) [58] have been compared in this study for reduction of PAPR. The expression, in the context of
PAPR reduction, of SQRC function as given in [105] is

, 0<
2

1
( )= , <

( 1) 1 1+
+ , <
2 2
(5.5.2.13)

The expression for MBH window function after due changes for PAPR improvement is given as

4 2
+ ( ) + (1 ) ( ) ,
2

0<


1
( )= , <


4 2
( ) + (1 ) ( ) ,
2

1 1+
<

[116]
 (5.5.2.14)
where, is a pulse shape parameter and is a roll-off factor. The range of is from 0.5 to 1.88 as
given in [58] and can take any value between 0 and 1. The is given as-

= = ( )/
(5.5.2.15)
where, is extra sub-carrier (overhead) required with pre-coded OFDM systems. Figure- 5.5.2.1 shows
the CCDF of the PAPR of OFDM signal with N=64 subcarriers for SQRC and MBH window function. In
this simulation, oversampled OFDM sequences with the oversampling rate of 4 are used to analyze PAPR
reduction.

Figure- 5.5.2.1: Comparison of CCDF of the PAPR of the pre-coded OFDM signal with
N=64 subcarriers for SQRC and MBH window function

It is clearly visible from the graph that the pre-coder based on MBH window function with =
1.5 performs better than the SQRC based pre-coder. Also it is visible that as roll-off factor increases
(more overhead), the PAPR reduces more.

[117]
Necessary Condition for the Optimal Pre-coding Matrix

In order to reduce the PAPR of OFDM signal by using pre-coding method without destroying the
orthogonality between sub-carriers, the pre-coding matrix should satisfy some condition. These condition
are given as-

(a) The pre-coding matrix should be an orthogonal matrix [105], i.e., it should satisfy the following
relation:

= (5.5.2.16)

where I is the identity matrix, and represents the Hermitian transpose of the matrix .
The above condition can be written in this way also-

1, =

, , =
0,
(5.5.2.17)

(b) For obtaining minimum BER, all the singular values of pre-coding matrix must be equal to one
[75].
The pre-coding matrix generated using proposed pulse shape satisfied both these condition.

5.5.3 Proposed Peak Clipping Algorithm

The PAPR reduction with pre-coding depends upon the roll-off factor (). As increase, the
PAPR reduces but BER degrades. Also it reduces the bandwidth efficiency because of extra sub-
carrier . Therefore, to reduce PAPR at the same roll off factor, the peak clipping is applied on OFDM
signal. The proposed peak clipping is different than normal peak clipping method because in normal peak
clipping only one peak of OFDM symbol is clipped whereas in proposed method, the Q high amplitude
samples of OFDM symbol are taken. These samples are reduced in amplitude by averaging. The proposed
algorithm is given in Table 5.1. It is similar to iterative clipping method. After applying peak clipping
algorithm, the amplitude of clipped signal ( ) is raised so that the average power is same as before
clipping.

[118]
 Table-5.1: Peak clipping algorithm

Input parameters required:

1. "NOI" - Number of Iteration

2. " " - Total number of high amplitude samples taken for averaging

First arrange the OFDM signal ( ), ( = 0,1,2, , 1) in ascending order of their amplitude.

1: ( ) (| ( )| // Arrange x(n) in ascending order of its amplitude //

2: 1 // Initialize main loop counter //
3: While <= do // Outer main loop starts//
4: // Initialize inner loop counter //
5: While > + 1 do // Inner main loop starts//

| ( )| | ( )|
6: | ( )| // average of and ( 1) samples //
7: = 1 // Decreases counter of Inner loop //

8: End while // Inner loop ends //

9: +1 // Increases counter of Outer loop //
10: End while // Outer main loop ends //
11: ( ) (| ( )| // re-arrange all the samples of y(n) at their original
location //

5.6 PERFORMANCE EVALUATION OF PROPOSED METHOD

In this section, the performance of proposed method is evaluated via simulation. The CCDF of
PAPR, BER, and power spectral density (PSD) has been taken as performance evaluation parameter. The
BER performance is evaluated in frequency selective fading channel and PSD plots are given with four
times oversample data. For this simulation, a 64 sub-carrier OFDM system with QPSK modulation is
taken. The tap-delay line model with L-path is considered for frequency selective fading channel. The
pre-coder is based on MBH window function with roll-off factor = 11% and pulse shape parameter =
1.5.

[119]
5.6.1 Performance of Pre-coding with Conventional Clipping

The comparison between the CCDF of the PAPR of OFDM signal with normal OFDM system
(without PAPR reduction), conventional clipping, conventional pre-coding and proposed hybrid method
is shown in Figure-5.6.1.1 and Figure- 5.6.1.2. Following are the observations from the graph:

(a) Pre-coding is better than clipping method.

(b) Clipping after pre-coding (proposed hybrid method) reduces the PAPR at same roll-off factor .
(c) As the clipping level increases, the PAPR reduces.

The BER performance of OFDM system with proposed PAPR reduction technique is shown in
Figure-5.6.1.3. The BER performance is evaluated over Rayleigh frequency selective fading channel
(with L=10). It is clearly visible from graphs that the pre-coded based OFDM system performs better than
the original OFDM system. This is because of diversity gain obtained via pre-coding matrix. Also it is
shown in the plots that the BER degrades as the clipping level increases. However, this degradation is not
significant.

Figure- 5.6.1.1: Comparison of CCDF of PAPR between conventional clipping, pre-coding, and
proposed technique

[120]
Figure- 5.6.1.2: Complementary cumulative distribution function of the PAPR of the OFDM
signal with N=64 subcarriers for roll-off factor =11% at different clipping level

Figure- 5.6.1.3: BER for 64- subcarrier QPSK-OFDM with pre-coding at =11% and clipping
at 0.8 and 0.7 over multipath fading channel with L=10

[121]
5.6.2 Performance of Proposed Peak Clipping Algorithm

In this section, the performance of proposed peak clipping algorithm is evaluated and compared
while taking normal and pre-coded OFDM system. Figure- 5.6.2.1 shows the CCDF plots of normal
OFDM system with and without peak clipping algorithm. It is clearly visible from the graphs that as the
number of iteration (NOI) increases, the PAPR is reduced more. This reduction is not linear because after
some iterations (for example, NOI = 40 in present case), the PAPR is not reducing much, this can be
considered the saturation level of iteration. For further reduction in PAPR, Q has to be changed.

The proposed algorithm is also compared with EC method [123]. The comparison is shown in
Figure- 5.6.2.2. It is clearly visible from the graphs that proposed method gives better result than EC
method ( = 2). The BER performance of proposed algorithm and EC method is shown in Figure-
5.6.2.3. In conclusion, the proposed algorithm gives better PAPR than the EC method with almost same
BER. The PSD plots support this statement as shown in Figure- 5.6.2.4.

Figure-5.6.2.5. shows the CCDF plots of pre-coded OFDM system with and without peak
clipping algorithm. The CCDF of PAPR without peak clipping with = 22% is similar to CCDF of
PAPR with peak clipping with = 11% only. Therefore, a 50% overhead is reduced. The BER
performance of pre-coded OFDM system with our proposed algorithm is shown in Figure- 5.6.2.6. It is
concluded that with proposed algorithm the PAPR is reducing more than conventional pre-coding with a
small BER degradation.

The overall performance comparison is summarized in Table-5.2. For example, when =

10 , the value of PAPR is 2.1, 2.6, 7.2, and 10.2 dB with proposed algorithm ( = 70, = 28), EC
method ( = 2), conventional clipping ( = 0.7) and without PAPR reduction respectively. The gain of
proposed method with respect to Original OFDM signal is 8.1 dB and with respect to EC method is 0.5
dB. In terms of BER, the performance of proposed clipping algorithm, conventional clipping and EC
method is almost similar and degrades from original OFDM system. For example, when =
15 , the BER is 0.008536, 0.01208, 0.0135 and 0.01453 with original OFDM (without PAPR
reduction), conventional clipping, and EC method and proposed clipping algorithm respectively.

[122]
 Figure-5.6.2.1: CCDF of the PAPR of OFDM signal with proposed algorithm

Figure-5.6.2.2: Comparison of CCDF of the PAPR of OFDM system with EC method and proposed
algorithm

[123]
 Figure-5.6.2.3: BER for 64- subcarrier QPSK-OFDM over frequency selective Rayleigh
fading channel with =

Figure-5.6.2.4: Comparison of power spectral density of 64- subcarrier OFDM system with EC
method and proposed algorithm

[124]
 Figure-5.6.2.5: CCDF of the PAPR of the pre-coded OFDM signal with N=64 subcarriers with
proposed algorithm

Figure-5.6.2.6: BER for 64- subcarrier pre-coded OFDM system over frequency selective Rayleigh
fading channel with L=10 with proposed algorithm

[125]
 Table-5.2: Performance comparison of proposed PAPR reduction method

Parameter PAPR in dB

Normal OFDM System Pre-coded OFDM System (MBH at = 1.5, = 11%)

CCDF Without Conventional Proposed Exponential Only Pre-coding Pre-coding with Pre-coding with
PAPR Clipping Clipping Companding Conventional Proposed
Reduction Method Algorithm clipping Clipping
Method Algorithm
(CR= 0.8), (Q=16, NOI= 40) (d = 2) (MBH at = 1.5) (CR= 0.8), (Q=16, NOI= 40)
10 10.2 dB
8.5 dB 3.7 dB 2.6 dB 5.4 dB 3.4 dB 3.3 dB
(CR= 0.7) (Q=28, NOI= 70) (CR= 0.7) (Q=28, NOI= 70)
7.2 dB 2.1 dB 2.4 dB 2.7 dB

BER in Rayleigh Fading Channel

Normal OFDM System Pre-coded OFDM System (MBH at = 1.5, = 11%)

Without PAPR Conventional Proposed Exponential Only Pre-coding Pre-coding with Pre-coding with
Reduction Clipping Method Clipping Companding Conventional Proposed
Method Algorithm clipping Clipping
Algorithm
(Q=16, NOI= 40) (CR= 0.8), (Q=16, NOI= 40)
15 dB 0.008536 (CR= 0.7) 0.0135 0.0004961
0.008836 0.0006823 0.0009766
(Q=28, NOI= 70) (CR= 0.7) (Q=28, NOI= 70)
0.01208
0.01453 0.00127 0.001307

[126]
5.7 SUMMARY

In this chapter, a novel algorithm of peak clipping is proposed in a pre-coded OFDM system to
reduce the PAPR. The proposed technique is better than conventional because it does not require any
increasing in roll off factor to reduce PAPR. Increasing the roll off factor degrades the BER as given in
[105]. The peak clipping after pre-coding reduces PAPR and reduced the overhead. The BER
performance is not affected. The proposed algorithm also applied in normal OFDM system and compared
with exponential companding. With the simulation studies, it has been shown that the proposed algorithm
reduces PAPR better than exponential companding with less side lobe level. Also a new pulse shape is
proposed to generate optimal pre-coding matrix.

***************************************************

[127]