International Journal of Computer Applications (0975 – 8887)

Volume 98– No.2, July 2014

PAPR Reduction in LTE Down Link using a Hybrid

Repeated Clipping and SQRT Companding

Zainab S. Hadi Buthaina M. Omran, Ph.D

University of Baghdad, College of Engineering University of Baghdad, College of Engineering
Electronic & Communications Engineering Electronic & Communications Engineering
Department Department

ABSTRACT amplitude is observed. The value of this peak may be N

The Orthogonal Frequency Division Multiplexing is one of times of the average amplitude [5].
the widely used modulation techniques in the broadband The drawback of a large dynamic range is that it places
wireless technology. One of the main problems of the pressure on the design of components such as the word
OFDM is the high peak-to-average power ratio of length of the IFFT/FFT pair, mixer stages, and most
transmitting signal due to the superposition of many importantly the HPA, which must be designed to handle
subcarriers. This paper presents a new hybrid peak-to- irregularly occurring large peaks, decreases the SQNR
average power ratio reduction technique, which repeated (Signal-to-Quantization Noise Ratio) of ADC (Analog-to-
clipping and frequency domain filtering with SQRT Digital Converter) and DAC (Digital-to-Analog Converter).
companding over LET channel. The paper highlights the Failure in design components with a sufficiently large linear
performance and advantages of the mixed technique. The range result in saturation of the HPA. Saturation creates both
simulations show that the proposed technique realizes an in band distortion, increasing the BER and out of band
improved PAPR With acceptable BER. distortion, or spectral splatter, which causes ACI. One
obvious solution is to design the components to operate
within large linear regions, however this is impractical as the
Keywords: components will be operating inefficiently and the cost
Complementary cumulative distribution function (CCDF), becomes prohibitively high. This is especially apparent in
OFDM, peak-to-average power ratio (PAPR), SQRT the HPA where much of the cost and ~50% of the size of a
companding, repeated clipping and filtering. transmitter lies [6, 7].
Many PAPR reduction schemes based on different
1. INTRODUCTION techniques, such as clipping and filtering, window shaping,
The basic principle of OFDM is to split a high data rate block coding, partial transmit sequence (PTS) technique [8],
streams into a number of lower data rate streams and then and selective mapping (SLM) technique [9], phase
transmitted these streams in parallel using several optimization, tone reservation [10] and injection [11], and
orthogonal sub-carriers (parallel transmission). When the nonlinear companding transform schemas. However, no
subcarriers have appropriate spacing to satisfy the specific PAPR reduction technique is the best solution for
orthogonality, their spectra will overlap [1]. OFDM has been the OFDM system. Various parameters like loss in data rate,
adopted in many wireless communication standards, transmit signal power increase, BER increase, computational
including European digital audio broadcasting, terrestrial complexity increase should be taken into consideration
digital video broadcasting and satellite–terrestrial interactive before choosing the appropriate PAPR technique [12, 13].
Multiservice infrastructure in China. In addition, OFDM has
been considered and approved by many IEEE standard 2. DEFINITION OF PAPR
working groups, such as IEEE 802.11a/g/n, IEEE 802.15.3a, The PAPR is defined as the ratio between the maximum
and IEEE 802.16d/e [2 - 3]. The applications of OFDM instantaneous power and its average power. If is a
include wireless personal area networks, wireless local area transmitted baseband OFDM signal, then PAPR is defined
networks, and wireless metropolitan networks. Currently, as:
Orthogonal Frequency Division Multiple Access (OFDMA)
is being investigated as one of the most promising radio
transmission techniques for LTE of the 3rd Generation (1)
Partnership Project (3GPP), International Mobile
Telecommunications—Advanced Systems [4]. Where, Is the average power of and can be
OFDM offers many advantages over single-carrier computed in frequency domain because IFFT is a unitary
modulations but high PAPR of transmitting signals is one of transformation Is useful duration of an OFDM symbol
the major issues of the OFDM system. A large dynamic [14].
range of input data symbols is the main cause of getting high The above power characteristics can also be described in
PAPR. An OFDM signal consists of independent data terms of their magnitudes (not power) by defining the crest
symbols modulated on N orthogonal subcarriers, and when factor (CF) as:
these N signals are added to the same phase, higher peak
CF = (2)

28
 International Journal of Computer Applications (0975 – 8887)
Volume 98– No.2, July 2014

Iterative clipping and filtering fft/iffft

N*
N* Nonlinear N*
Point
Point Processing Point inverse Add
inverse DFT over Zeroes 0 DFT CP
0 Clipping sampling
DFT over
over rate 0 samplin
0 Ratio =
sampling g rate
rate CR

Input data Clipped

Interpolated
zero padded Interpolated Frequency
baseband signal
baseband signal domain filtering
Fig 1: Clarifies the clipping and filtering process

3. REPEATED CLIPPING AND amplification. Although frequency domain filtering is a

common signal processing technique the form shown in
FREQUENCY DOMAIN FILTERING figure (1) is unusual. In most filtering applications the filter
In the clipping technique hard limiting is applied to the is designed to meet particular specifications in the
amplitude of the complex values of the IFFT output. The continuous frequency domain. In this application, the
filtering technique is designed to alleviate or cancel out-of- wanted signal is an OFDM signal, which is the sum of
band distortion dependent on oversampling value, but discrete frequency components in each symbol period.
cannot correct in-band distortion [16] The filter must therefore have as little effect as possible on
the in-band discrete frequency domain while attenuating as
At first the input data signal as shown in figure
much as possible any out-of-band components. Because the
(1) is oversampled by a factor I by placing zeroes
filter operates on a symbol by symbol basis, there is no
in the middle of the vector. This oversampled signal is
filtering across symbol boundaries and so no resultant
transformed into the time domain using an IFFT. This
intersymbol interference. The filtering does cause some peak
results in trigonometric interpolation of the time domain
regrowth. However, this is much less than for clipping
signal.
before interpolation. [15, 17]
After an IFFT, the original signal is clipped in the time
The clipping noise is added at the transmitter rather than the
domain. The clipping can be described as shown below:
receiver. In fading channels this means that in general the
clipping noise will cause less degradation in bit error rate
(3) than noise added in the channel since the clipping noise
fades along with the signal.
Where represents the output of the time domain signal
, Is the threshold clipping level
; Is the mean power The 4. SQUARE-ROOT COMPANDING
clipping ratio, is defined as the ratio of the clipping level to The block diagram of a typical OFDM system using the
the mean power of the unclipped baseband signal The original SQRT technique for PAPR reduction is shown in
clipping is followed by frequency domain filtering to reduce Fig.2. By using the SQRT technique, the original OFDM
out-of-band power. output signals Is processed by (4) before they are
The clipped time domain signal c is then converted back into converted into analog waveforms and amplified by the
the discrete frequency domain using an FFT, The in band power amplifier
discrete frequency components of the clipped signal (4)
Are passed unchanged to the Is the new OFDM signal, and Is the phase of During
inputs of the second IFFT while the out-of-band the entire signal processing, the phases of the OFDM output
signals Are kept unchanged while only the amplitudes
components, Are nulled [16]. This
are treated and changed [18].
technique is repeated iteration number, usually choose For the complex Gaussian distributed signals, such as
between one and four. OFDM output signals, SQRT process changes the Rayleigh
The second IFFT is followed by serial to parallel distribution of these signals into a Gaussian-like, close to
conversion, digital to analogue conversion, modulation and

29
 International Journal of Computer Applications (0975 – 8887)
Volume 98– No.2, July 2014

Fig 2: Block diagram of an OFDM system using SQRT technique

Gaussian, distribution [18,19]; while the Chi-square reduced in all sizes of OFDM blocks, N. In [18, 19] the
distribution is, converted, according to the analysis of these SQRT process applies to the signals of all OFDM output
signals given in the previous section, to Rayleigh symbols; therefore, the PAPR reduced without the need to
distribution. The latter is because the Rayleigh distribution send side information. The SQRT process changes the
in such signals represents voltage while the Chi-square distribution of the power signals to Rayleigh distribution and
distribution represents the power of the same signals. reduces the value of average power from N to N1/2. The
However, not only the statistical distribution is changed by variance of the Rayleigh distribution equals
the SQRT process, but the values of the mean and [20] which is approximately equal to half the value of
variance of the processed OFDM output signals are also variance of the Gaussian distributed signals. The SQRT
changing, and subsequently the values of the average power process in the SQRT OFDM system performs this statistical
and peak power of these signals are altered also. To transformation, and therefore results in a constant
understand the effect of the SQRT process on the power degradation in the BER rate equal to 3 dB because of
values of OFDM output signals, we assume normalized decreasing of variance to the half of that of the conventional
average power ( When the average power is OFDM system .
normalized, the value of the peak power is diminished by N
because for the same PAPR. This assumption is
applicable for all OFDM symbols as the average power is 5. THE PROPOSED PAPR
constant and equal to Hence, the PAPR can REDUCTION METHOD
be analyzed according to (1) through the peak power only. We propose hybrid approach, which consist of repeated
The new value of normalized peak power is always greater clipping and frequency domain filtering and SQRT
than one because is constantly greater than in all compander as shown in figure (3). The main idea for
OFDM symbols. Therefore, the SQRT process always combining the two methods is relaying on the observation
causes a reduction in the value of the peak power of the that two different types of signal processing for PAPR
normalized OFDM symbols, and as a result the PAPR is reduction will increase the overall performance.

Modulator Iterative clipping

S/P IFFT Add P/S
& filtering
QPSK CP

CH

Demodulator P/S FFT Remove S/P

QPSK CP

Fig 3: Proposed PAPR reduction

30
 International Journal of Computer Applications (0975 – 8887)
Volume 98– No.2, July 2014

Table 1. The value of PAPR and SNR at BER =

PAPR SNR at BER PAPR SNR at PAPR SNR at PAPR SNR at

= Clipping BER = SQRT BER = Hybrid BER =
Clipping SQRT Hybrid

25.3766 11.872 14.5624 8.85 13.8413 14.466 8.4279 11.571

6. SIMULATION AND RESULT because of filtering, While SQRT has worse BER but PAPR
To verify the proposed PAPR reduction technique, we tested Slightly larger repeated clipping and frequency domain
it in the LTE downlink system with parameters shown in filtering.
Table 2 0
10
Table 2. Simulation parameter
Orignal
FFT size 128
One clip and filter
Spacing frequency 15 KHz
Two clip and filter
BW 1.25MHz Three clip and filter
CP 32 Four clip and filter

CCDF (Pr[PAPR>PAPR0])
-1
No symbol 1000 10 hybrid
Sampling frequency 192MHz
No. of path delay in LTE 6
channel

0
Bit error probability curve for qpsk using OFDM
10
-2
orginal 10
clipping and filtering
sqrt
-1 hybrid
10

-3
10
0 2 4 6 8 10 12
BER

-2 PAPR0 [dB]
10
Fig 5: Comparison in PAPR between original signal,
repeated clipping and frequency domain filtering and
hybrid with CR = 4, oversampling = 2.
-3 As shown in figure (5) there are about 6.887 dB (10.755 -
10
3.868 =6. 887) Improvement in the PAPR value. Figure (5)
is for PAPR after companding, but the total PAPR for
original signal=25. 3766 as shown in table (1) for hybrid =8.
4279 (25.3766 - 8.4279 =16. 9487) 16.9487db is the
Improvement in the total PAPR. And with little
-4
10 improvement BER comparison with the original signal as
0 5 10 15 shown in figure (4)
snr
When we change the Papr and BER will change because of
Fig 4: Comparison in BER between original signal, frequency domain filtering that we use. We take I = 1.125,
repeated clipping and frequency domain filtering, SQRT 1.25, 1.5,2,3,4 for explanation I=1.125 that means
companding and proposed hybrid. 0.125*FFT size of zero that add in the center input vector to
make trigonometric interpolation.
As shown in figure (4) the hybrid has better PAPR and BER
compare, Hybrid result with oversampling equal to 2 and
CR = 4, and the same thing for Just repeated clipping and
frequency domain filtering, and we can see that repeated
clipping and frequency domain filtering has the best BER

31
 International Journal of Computer Applications (0975 – 8887)
Volume 98– No.2, July 2014

0
Bit error probability curve for qpsk using OFDM 0
Bit error probability curve for qpsk using OFDM
10 10
I4 CR 4
I3 CR 3
CR 2
I2
-1 CR 1.75
-1 I 1.5 10
10 CR1.5
I 1.25
I 1.125

BER
-2
10
BER

-2
10

-3
10
-3
10

-4
10
-4
0 5 10 15 20 25 30
10 snr
0 5 10 15
snr
Fig 9: Comparison in BER for different CR
Fig 7: Comparison in BER for different oversampling When CR decreases the PAPR improve on account of BER
value. but as shown in figures (8, 9) CR=3 has batter PAPR with
slightly degraded in BER compare with CR=4 And CR =2
As shown in figure (6) and (7) an inverse relationship has better than 3, 4 PAPR and acceptable BER degradation.
between BER and PAPR for oversampling =4 has a better
BER but worse PAPR compare with other oversampling 7. CONCLUSIONS
unlike oversampling =1. 125, we can say that for large
The hybrid PAPR has been proposed in this paper is better
oversampling the BER improve, but PAPR Getting worse
than just repeated clipping and frequency domain filtering or
compare with oversampling and vice versa.
just SQRT companding. It can significantly reduce high
peaks in the OFDM systems better than the both. At least 5
9
dB better than both types. But at the expense of increased
cr4
8.5 complexity a little bit compares with just SQRT or just
cr3
repeated clipping, but this technique still low complex
cr2
8
cr1.75
compare with other PAPR techniques such as PTS, SLM,
coding.
7.5
And PAPR can reduce depending on two parameters the
7 oversampling and clipping ratio When CR is low PAPR
reduce and the same thing for oversampling but on account
papr

6.5
BER and vice versa.
6
8. REFERENCES
5.5 [1] R. Van. Nee and R. Prasad, 2000. OFDM for Wireless
Multimedia Communication. London: Artech house.
5
[2] IEEE Standard 802.11a, 1999. “Wireless LAN medium
4.5
access control (MAC) and physical layer (PHY)
specifications”.
4
8 10 12 14 16 18 20 22
[3] IEEE Standard 802.16a, 2003. “Air interface for fixed
snr at (ber =10-4)
broadband wireless access systems –Amendment 2:
Fig 8: Relationship between PAPR and BER for medium access control modifications and additional
different value of CR physical layer specifications for 2-11 GHz,”
[4] ITU-R, Report M.2134, 2008. Requirements related to
technical performance for IMT-Advance Radio
Interface.
[5] Hem Dutt Joshi, May 2012 “performance augmentation
of OFDM system,” Ph.D. dissertation, Jaypee Univ. of
engineering and Technology, India.

32
 International Journal of Computer Applications (0975 – 8887)
Volume 98– No.2, July 2014

[6] Young Soo Cho, Jaekwon Kim, Won Young Yang and [13] S. H. Han and J. H. Lee April 2005. “An overview of
Chung G. Kahn, 2010. MIMO OFDM wireless peak-to-average power ratio reduction techniques for
communications with Matlab, Asia: John Wiley & multicarrier transmission”, Wireless Communications,
Sons. IEEE ,vol.12 , no: 2 ,pp 56-65.
[7] Gavin Hill, March 2011 “Peak Power Reduction in [14] T. Jiang and Y. Wu, June 2008. “An overview: peak-to-
Orthogonal Frequency Division Multiplexing average power ratio reduction techniques for OFDM
Transmitters,” Ph.D. Thesis, Victoria University of signals,” IEEE Transactions on Broadcasting, vol. 54,
Technology. no. 2, pp. 257- 268.
[8] Leonard J. Cimini, Jr., and Nelson R. Sollenberger Jun [15] Carole A. Devlin, Anding Zhu, and Thomas J. Brazil
1999. “Peak-to-average power ratio reduction of an Jan. 2008. “Peak to Average Power Ratio Reduction
OFDM signal using partial transmit sequences,” Technique for OFDM Using Pilot Tones and Unused
Communications, 1999. ICC '99. 1999 IEEE Carriers” Radio and Wireless Symposium, 2008 IEEE ,
International Conference on, vol.1, pp 511-515, 06-10. pp 33 – 36, 22-24.
[9] Sarita Singh Bhadhoria and S. Gupta, March 2013. [16] Armstrong, J., February 2002. “Peak-to-average power
“PAPR Reduction of OFDM Signals Using Selected reduction for OFDM by repeated clipping and
Mapping Technique,” Current Research in Engineering, frequency domain filtering” ELECTRONICS
Science and Technology (CREST) Journals, vol.1,no. LErrERS, vol. 38, no. 5, pp. 246- 247, 28th.
1, pp 34-37.
[17] Armstrong, J., 2001 "New OFDM peak-to-average
[10] M. Chauhan and A. Chobey, August 2012. “PAPR power reduction scheme," IEEE 53rd Vehicular
Reduction in OFDM system Using Tone Reservation Technology Conference, Vol. 1, no. Pp. 756-760.
Technique,” International Journal of Computer
Technology and Electronics Engineering (IJCTEE), [18] Wisam F. Al-Azzo, Borhanuddin M. Ali, Sabira
vol. 2, no. 4, pp 57-60. Khatun, and S. M. Bilfagih, 2007. “Time domain
statistical control for PAPR reduction in OFDM
[11] J. Hou, C. Tellambura, and J. Ge June 2013, “Clipping system”, Proc. IEEE APCC 2007, Bangkok, Thailand,
Noise-based Tone Injection for PAPR Reduction in pp.141-144.
OFDM Systems,” Communications (ICC), 2013 IEEE
International Conference on, pp 5759 -5763, 9-13. [19] Wisam F. Al-Azzo, Borhanuddin M. Ali, march 2011.
“Adaptive Square-Rooting Companding Technique for
[12] M. K. Singh, S. Sharma, P. Singh, A. Gupta, S. K. PAPR Reduction in OFDM Systems”, World Academy
Singh Aug 2012. “Different PAPR Reduction of Science, Engineering and Technology, Vol. 5, 20th.
Techniques for Orthogonal Frequency Division
Multiplexed Signals”, International Journal of [20] S. Haykin and M. Moher, , 2005. Modern Wireless
Electronics and Computer Science Engineering. Communications. Prentice Hall.

IJCATM : www.ijcaonline.org
33