Analytical calculations of CCDF for some common

PAPR reduction techniques in OFDM systems

Srinivas Ramavath and Rakhesh Singh Kshetrimayum
Department of Electronics and Electrical Engineering,
Indian Institute of Technology (IIT) Guwahati, Assam 781039, India.
E-mail: [email protected], [email protected]

AbstractHigh PAPR is one of the major drawbacks of intervals [1]. Another problem which is experienced by OFDM
OFDM systems. It drives the transmitters power amplier into signals that have high PAPR involves the development of
its nonlinear region, thus causing nonlinear distortions. Square non-linear distortions [2]. Numbers of techniques have been
rooting technique (SQRT) and -law companding technique are
the simple techniques to reduce PAPR in OFDM systems. In proposed for reducing the PAPR in OFDM systems, which
this paper, we derive analytical CCDF for both the techniques. basically can be divided in three categories [3]. First, there are
The comparisons of SQRT and -law companding techniques signal distortion techniques, which reduce the peak amplitudes
are carried out using theoretical and simulation results. It is simply by nonlinearly distorting the OFDM signal at or around
veried that the -law companding technique is better than the peaks [4]. Second, there are coding techniques that use
SQRT technique for PAPR reduction.
a special FEC code set that excludes OFDM symbols with
Keywords- CCDF, -law compander, OFDM, PAPR, SQRT. a large PAPR. The third technique scrambles each OFDM
symbol with different scrambling sequences and selecting the
sequence that gives the smallest PAPR [5], [6].
I. I NTRODUCTION OFDM has been adopted in some wireless standards such
RTHOGONAL frequency division multiplexing
O (OFDM) is an efficient parallel data transmission
scheme, in which high data rates can be achieved by
as Digital Audio broadcasting (DAB), Terrestrial digital video
broadcasting (DVB-T), HIPER LAN12, IEE802.11 standard
for WLAN, and IEE802.16 standard for WiMAX. Moreover,
transmitting a number of orthogonal subcarriers. The main OFDM has been considered for fourth generation (4G) trans-
reason for increasing interest in the OFDM system is the way mission technique [7].
it handles multipath fading phenomenon that is inherent with The rest of the manuscript is organized as follows. In
the wireless channels. Multipath is the result of reflection of Section II describes the system model and defines the PAPR.
wireless signals by objects in the environment between the The theoretical CCDF are calculated in Section III. The paper
transmitter and receiver. Multipaths arrive at the receiver with is concluded in Section IV.
random phase offsets, because each reflected wave follows
a different path from transmitter to reach the receiver. The II. OFDM SYSTEM MODEL AND PAPR
OFDM signals can be efficiently modulated and demodulated
using inverse fast Fourier transform (IFFT) and fast Fourier In an OFDM system, a high-rate data stream is split into
transform (FFT) respectively. N low-rate streams that are transmitted simultaneously by the
It also has many advantages like spectral efficiency, simple subcarriers. Each of them may be modulated using quadrature
implementation by fast Fourier transform, low receiver com- amplitude modulation (QAM) or quadrature phase-shift keying
plexity, etc. However, OFDM suffer from problems like high (QPSK) [8]. The input data symbol X(k) for 0 k N 1
peak-to-average power ratio (PAPR), sensitivity to time and is modulated by a set of N orthogonal frequencies of IFFT,
frequency synchronization errors. OFDM has various prop- then the OFDM output sequence x(n) for 0 n N 1 is
erties that make it desirable over existing single carrier sys- represented by
tems, the main advantage is OFDMs immunity to frequency { }
1 N1 2nk
selective fading. Single carrier systems can increase their xn = k
N k=0
X exp j
N
, 0 n N 1 (1)
data rate by shortening the symbol time, thereby increasing
the occupied bandwidth. Wideband channels are sensitive to The OFDM symbol x[n] is a complex number and its real and
frequency selective fading which require complex equalizers in imaginary parts are denoted by xre [n] and xim [n]. According
the receiver to recover the original signal. OFDM overcomes to the central limit theorem as N gets large, the distributions
this problem by dividing the wideband channel into a series of xre [n] and xim [n] become normal (Gaussian). This means
of narrowband channels in which each narrowband channel that x[n] has Rayleigh distribution as shown in Fig. 1 and its
experience flat fading. Therefore only 1 tap equalizers are probability density function is given as
required in the receiver, reducing complexity greatly. The
x2
inter symbol interference (ISI) and inter channel interference x
fX n (x) = 2e
22 (2)
in OFDM systems are mitigated by the insertion of guard

978-1-4673-4700-6/12/$31.00 2012
c IEEE 393
 0.4
QAM modulation (N=16) of both the input samples and IFFT operation.The statistical
pdf of x [n]
distribution as well as the mean and variance values of the
re
0.2 signal is varied by applying square rooting.
0
0.02 0.01 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07

SQUARE ROOTING
MODULATOR
x [n]

CONVERTER

CONVERTER
re

CONVERTER

D/A &UP
(NPOINT)
Data In

....

....
0.4

S/P

IFFT

P/S
pdf of x [n]
im

0.2

0 AWGN
0.02 0.01 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
xim[n] channel

DEMODULATOR
0.4

A/D &DOWN
pdf of |x[n]|

CONVERTER

CONVERTER

CONVERTER
SQUARING
(NPOINT)

....
FFT
....
P/S

S/P
0.2
Data out

0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
|x[n]|
Fig. 2: Block diagram of an OFDM system using SQRT technique
Fig. 1: Probability distribution function of OFDM signal (a)
real part of x[n], (b) imaginary part of x[n], (c) amplitude of
As well as the amplitudes of the OFDM output signals are
x[n]
changed while the phases are kept unchanged [10].
. The peak power and average power of OFDM symbol are
defined by the following equations y= x
1
Ppeak =0nN1
max
xn 2 y =
2 x
1 N1 We can denote the Probability density function after SQRT
Pavg = xn 2
N n=0 technique as
The PAPR of the OFDM signal is expressed as fX n (x)
fY n (y) =
y
max x[n]2
PAPR = [ ] (3) x
x2

E x[n]2 2
e 22
fY n (y) = 1

Note that the maximum of Xn is equivalent to crest factor. Let 2 x
Xmax denote the crest factor i.e., Xmax =n=0,..,N1
max Xn . Now, the where x = y2 , substitute in above equation then PDF is
cumulative distribution function (CDF) of Xmax is given as [9]
2y3 y24
e 2 fY n (y) = (6)
FX max (x) = P(Xmax < x) 2
x
= fY n (y) dy The peak power and average power of square rooting technique
0 of OFDM symbol are computed as [11]
x y2
y
= e 22 dy Psqrt peak = 0nN1
max 0.5
Ppeak (7)
0 2
x2 N1
1
CDF = 1 e 22 Psqrt av = [Pn (t)]0.5 (8)
N n=0
The probability that the PAPR is below some threshold level
can be written as where Ppeak = x2 , the PAPR formula for the square rooting
( )N technique in OFDM system is given by
x2
P(PAPR x) = 1 e 22 (4) max 0.5
0nN1 Ppeak
PAPRsqrt = N N1 (9)
We consider the following complementary cumulative distri- n=0 [Pn (t)]0.5
bution function Note that the maximum of Yn is equivalent to crest factor. Let
( )N
x2 Ymax denote the crest factor i.e., Ymax =n=0,..,N1
max Y . Now, the
n
CCDF = 1 1 e 22 (5) cumulative distribution function (CDF) of Ymax is given as
FY max (y) = P(Ymax < y)
III. CCDF FOR PAPR REDUCTION TECHNIQUES y
A. Square rooting technique = fX n (x) dx
0
y 3 x4
As every sample of the IFFT output xn is coming from the 2x
= e 22 dx
summation of all the input complex valued samples. There 0 2
are two factors affecting amplitude value at the output (1) y4

amplitude of the input sample (2) combination of phases CDF = 1 e 22

394 2012 International Conference on Communications, Devices and Intelligent Systems (CODIS)
The probability that the PAPR of Square rooting technique is where is the -law parameter of the compander, x is the
below some threshold level can be written as input signal and v is the maximum value of the signal x. And
( )N at the receiver site, -law expander restores original signal by
y4
P(PAPR y) = 1 e 22 (10) v Z ln(1+)
x = (e v 1)sgn(Y ) (13)

We consider the following complementary cumulative distri-
bution function Companding is widely used in speech processing where high
( )N peaks occur infrequently. OFDM signal also exhibit similar
y4
characteristics where high peaks occur infrequently. Compand-
CCDF = 1 1 e 22 (11)
ing technique improves the quantization resolution of small
signals at the price of the reduction of the resolution of
The curves show theory and simulation results. The maximum
large signals, since small signals occur more frequently than
reduction in PAPR value of about 4.5 dB is achieved by the
large ones. PAPR is decreasing exponentially with increasing
Square rooting technique.
value of the companding parameters . In -law compander,
the signals with lower amplitudes are amplified with greater
Square rooting papr
0
10 gain. In OFDM systems, the occurrence of subcarriers having
very large peak amplitudes is less frequent, since most of the
subcarriers have low peak amplitudes. Because of less frequent
high amplitude subcarriers, the average power is low resulting
in high PAPR. The peak power of compressor is given by
( )
ln 1 + Ppeak
0.5
Pcomp peak = 0nN1
max
ccdf

1
10 (14)
ln(1 + )
The average power of compressor as follows

1 N1 ln (1 + [Pn (t)]0.5 )
Conve. Simul.
SQRT. Simul. Pcomp av =
N n=0 ln (1 + )
(15)
SQRT. Theory
Conve. Theory
2
10
0 2 4 6 8 10 12 The PAPR formula [11] for the companded OFDM is given
papr(dB) by
Fig. 3: CCDF curve for PAPR reduction performance of the square rooting ( [ ]0.5 )
techniques. N ln 1 + Ppeak
PAPRcomp = 0nN1
max ( ) (16)
B. -law companding 0.5
N1
n=0 ln 1 + [Pn (t)]
The companding method is a simple and practical to reduce  
the PAPR for OFDM signal. The compressor is simple loga-  X 
where Pn (t) is the normalized instantaneous power i.e  Xpeak 
rithm computation. The reverse computation of a compressor
of the nth subcarrier and the Ppeak is the normalized peak power
is called an expander.In this paper, the compression is done at
from the compander.
the transmitter after the IFFT block. The expansion is done at
the receiver prior to FFT block. The complementary cumulative distribution function
(CCDF) of the PAPR is one of the most frequently used
parameter for analyzing PAPR reduction. Before calculating
COMPRESSOR
MODULATOR

CCDF, we have to find probability density function after

CONVERTER

CONVERTER
CONVERTER

D/A &UP
(NPOINT)

Data In
....

....

companding. Here we can denote the probability density

S/P

IFFT

P/S

function (PDF) as
fX n (x)
AWGN
fZ n (z) = (17)
channel
Z
Here Z is the derivative of compressor output. Taking the
DEMODULATOR

A/D &DOWN
CONVERTER

CONVERTER

CONVERTER
EXPANDER
(NPOINT)

....
FFT
....

derivative of Equation (4.1), we have,

P/S

S/P

Data out

dY 1
= . (18)
Fig. 4: Block diagram of an OFDM system using compander technique dx ln (1 + ) 1 + .x
Substituting equations (2) and (18) in equation (17), then we
In general, -law compression characteristic is given by will get the PDF after simplification,

v ln(1 + x
v ) ln (1 + ) ( ) x 2

Z= sgn(x) (12) fZ n (z) = x + x 2

e 22 , n = 0, ....., N 1 (19)
ln (1 + ) 2

2012 International Conference on Communications, Devices and Intelligent Systems (CODIS) 395
Note that the maximum of Zn is equivalent to crest factor. Let 0
10
PAPR comparison

Zmax denote the crest factor(i.e., Zmax =n=0,..,N1

max Zn ). Now, the
cumulative distribution function (CDF) of Zmax is given as
z
FZ max (z) = P(Zmax < z) = fZ n (x) dx (20)
0
z x2 (1+) z x2
ln (1 + ) loge
FZ max (z) = xe 22 dx + x2 e 22 dx
2

ccdf
2
1
0 0 10
(21)
Here first integral part is CDF of Rayleigh distribution and
second integral part is CDF of Maxwell distribution, P(Zmax < Conven. Simul.
SQRT. Simul.
z) = P1 (Zmax < z) + P2 (Zmax < z) respectively. Compan. Simul.
Conven. Theory
P(Zmax < z) = P(Z0 < z).P(Z1 < z)........P(ZN1 < z) (22) SQRT. Theory
Compan. Theory
2
10
The Rayleigh distribution CDF is given as 0 2 4 6 8 10 12
( 2
) papr(dB)
z
Fig. 6: CCDF curve for different PAPR reduction techniques.
P1 (Zmax < z) = 1 e 22 (23)

Similarly, the Maxwell distribution CDF is given as IV. CONCLUSIONS

( 2
)
z 2 z2 In this paper we have considered the CCDF performance
P2 (Zmax < z) = er f ( ) ze 2 (24)
2 analysis of the PAPR reduction techniques in OFDM systems.
The analytical calculation of CCDF is carried out for combin-
The CDF for single channel from the above two expression is ing and square rooting techniques. Both theoretical analysis
[( ) ( )]
z2 z2 and simulation show that the PAPR performance is improved
ln (1 + ) z 2
= 1 e 22 + er f ( ) ze 22 in companding technique. In future work, we will focus on
2 2
the analytical Bit Error Rate performance analysis using these
(25) techniques.
The overall CDF for N subcarriers as follows
[ (( ) ( ))]N
z2 z2
ln (1 + ) z 2 R EFERENCES
= 1 e 22 + er f ( ) ze 22
2 2 [1] R. V. Nee and R. Prasad, OFDM for Wireless Multimedia Communica-
(26) tions, Boston, Artech House Publishers, March 2000.
[2] E. Costa, M. Midrio, and S. Pupolin, Impact of amplifier nonlinearities
As stated above CCDF = 1 CDF. on OFDM transmission system performance, IEEE Comm. Lett., Vol. 3,
no. 3, Feb. 1999, pp. 37-39.
COMPANDING PAPR [3] S. H. Han and J. H. Lee, An overview of peak-to-average power ratio
0
10 reduction techniques for multicarrier transmission,IEEE Wireless Comm.,
Vol. 12, no. 2, April 2005, pp. 56-65.
[4] X. Li and L. J. Cimini, Jr., Effects of clipping and filtering on the
performance of OFDM, IEEE Comm. Lett., Vol. 2, no. 5, May 1998,
pp. 131-133.
[5] H. Sakran, M. Shokair, A. Abou Elazm, PAPR Reduction Technique of
OFDM signals Using -Law Compander, IEEE Symposium on Wireless
Communication System, ISWCS08, 21-24 Oct., pp. 386-390, 2008.
[6] X. Wang, T. T. Tjhung, and C. S. Ng, Reduction of peak to average power
ccdf

1
10 ratio of OFDM system using a companding technique, IEEE Trans. on
Broadbasting, 1999.
[7] S.Hara, and R. Prasad, Multicarrier Techniques for 4G Mobile Commu-
nications, Nor- wood, MA: Artech House, 2003.
[8] J. G. Proakis, Digital Communications. New York: Mc Graw-Hill, 2004.
Conven. Simul. [9] Y. S. Cho, J. Kim, W. Y. Yang and C. G. Kang, MIMO OFDM Wireless
Comp. Simul. Communications With MATLAB, John Wiley and Sons (Asia) Pte Ltd.,
Comp. Theory 2010.
Conven. Theory [10] W. F. Al-Azzo, B. M. Ali, S. Khatun and S. Jamalullail, Time Domain
2
10
0 2 4 6 8 10 12 Statistical Control for PAPR Reduction in OFDM System, Asia-Pacific
papr(dB) Conference on Communications, 2007.
Fig. 5: CCDF curve for PAPR reduction performance of the companding [11] S. P. Vimal and K. R. Shankar Kumar, A New SLM Technique for
technique. PAPR Reduction in OFDM Systems, EuroJournals Publishing, Vol. 65
no. 2, pp. 221-230, 2011.
[12] H. Harada and R. Prasad, Simulation and software radio for mobile
The curves (theory and simulation) shows that the maximum communications, Boston, Artech House Publishers, 2002.
[13] H. Liu and G. Li, OFDM-Based Broadband Wireless Networks: design
reduction in PAPR value of about 5 dB is achieved by the and optimization, John Wiley and Sons, 2005.
companding method described above. PAPR reduction perfor- [14] L. L. Yang, Multicarrier Communications, Wiley, 2009.
mances of square rooting and -law companding techniques
are compared in Fig. 6.

396 2012 International Conference on Communications, Devices and Intelligent Systems (CODIS)