2019 IEEE 5th International Conference on Computer and Communications

A Study on Parameter Optimization for PC-CFR Algorithm

Minglei Teng Lintao Li

School of Information and Electronics School of Electronic Engineering
Beijing Institute of Technology, Beijing University of Posts and Telecommunications
Beijing, China Beijing, China
e-mail: [email protected] e-mail: [email protected]

Bohao Sun Rui Wang

Department of Electronic and Electrical Engineering Beijing Galaxy Information&Communication co. LTD
University College London Beijing, China
London, UK e-mail: [email protected]
e-mail: [email protected]

Abstract—Due to the low complexity, the peak cancellation Researchers have proposed a number of PAPR reduction
algorithm (PC-CFR) can be applied to reduce the high peak- techniques to solve the high PAPR problem, which includes
to-average ratio (PAPR) of orthogonal frequency division clipping-and-filtering (CF), coding, selected mapping (SLM)
multiplexing (OFDM) signals. The core of the PC-CFR and so on [8-11]. Currently, CF is the most widely used
algorithm is that the spectrum of its correcting signal is nearly technique because of its superior trade-off between signal
identical to the spectrum of the OFDM signal, so the signal distortion and implementation complexity. The clipping-and-
waveform design of the correcting signal becomes the key filtering technique can be classified into two categories:
component of the algorithm. In this paper, the two key iterative clipping and filtering (ICF) [12] and crest factor
parameters of the correcting signal, filter type and filter order,
reduction (CFR) methods. The ICF method is proposed by
are studied. A number of filters and filter orders are compared
to demonstrate that the fir1 with 122(order) performs great
Armstrong in 2002 [13]. This method can reduce the PAPR
balance among PAPR reduction, in-band distortion and out-of- by repeatedly performing clipping and filtering operations on
band radiation based on the complementary cumulative the OFDM signal until the PAPR of the signal reaches a
distribution function(CCDF), error vector magnitude(EVM) predetermined threshold. Meanwhile, the out of band
and Power Spectral Density(PSD) that are main indicators to radiation can be suppressed effectively after PAPR reduction
measure PAPR reduction technology. The experiment also [14]. However, due to the iterative processing [15], the
proves that under the same conditions, the PC-CFR algorithm algorithm processing time and implementation complexity
using these two parameters is superior to other CFR are increased. The CFR method presented in [16,17] reduced
algorithms in reducing the peak-to-average ratio and the out- the PAPR by processing signals samples that exceed the
of-band rejection. predetermined threshold, including peak windowing (PW-
CFR) [18], noise shaping (NS-CFR) [19], and peak
Keywords-component; Orthogonal Frequency Division cancellation (PC-CFR) [20]. The basic strategy of CFR
Multiplexing; peak-to-average ratio; peak cancellation algorithm technique is to add a correcting signal over the OFDM signal
which has a high PAPR. For PW-CFR and NS-CFR method,
I. INTRODUCTION the correcting signals are window function and noise shaping
filters, respectively. These two methods have a common
OFDM is a high speed data transmission scheme based
shortcoming: there is a great difference between the
on multi-carrier modulation. Due to its high spectrum
correcting signal and the original OFDM signal. The
efficiency and robustness against multipath fading effects,
mismatch may lead to in-band distortion and out-of-band
OFDM has been widely used in many wireless
radiation. For PC-CFR method, the correcting signal is a
communication systems such as LTE, Wi-Fi (IEEE 802.11)
pulse shaping signal that has a nearly identical spectrum
and WiMAX (IEEE 802.16) [1-5]. The main drawback of
shaping as the original OFDM signal. Thus, signal waveform
OFDM is the high peak-to-average ratio (PAPR), which may
design of the correcting signal becomes the key component
lead to reduction of average transmit power if the peak
of the algorithm.
transmit power is limited. As a result, the maximum
communication distance of OFDM will be significantly In this work, the improvement of PC-CFR scheme based
constrained. Otherwise, signal distortion may arise as the on parameter optimization for correcting signal will be
linear dynamic range of the power amplifier is exceeded. investigated. As the correcting signal are always pulse
Therefore, the application of OFDM systems have been signals scaled by the amplitude and phase of the signal peaks
severely restrained because of the high PAPR problem, that exceed the predefined threshold, a prototype filter that is
especially in power-constrained applications [6,7]. matched to the spectrum of the OFDM signal should be

978-1-7281-4743-7/19/$31.00 ©2019 IEEE 458

Authorized licensed use limited to: Nokia. Downloaded on October 23,2020 at 02:27:00 UTC from IEEE Xplore. Restrictions apply.
designed firstly. The two major parameter, filter type and the signal distortion is, otherwise, the closer the signal after
filter order have great effect on the performance of PC-CFR clipping is to the original signal.
algorithm. A number of filter types and filter orders are
considered for optimization in our work, we have III. PEAK CANCELLATION ALGORITHM
demonstrated that the fir1 with 122(order) performs great The peak cancellation algorithm reduces the PAPR by
balance among PAPR reduction, in-band distortion and out- subtracting a pulse signal that the spectrum is similar to
of-band radiation. original signal frequency spectrum so that the PAPR of the
The paper is organized as follows. In Section II, the corrected signal reaches a predetermined requirement. The
definition of PAPR and evaluation index of PAPR reduction peak cancellation algorithm includes the peak detection
technique are introduced. In Section III, the PC-CFR block, the peak scaling factor calculation block, the
algorithm process is presented. In Section IV, the parameter cancellation pulse allocation block, and the cancellation
optimization process for PC-CFR is introduced in detail. In pulse signal calculation block. In practice, the
Section V, performance comparison between the proposed implementation framework is shown in Fig.1. The
method and the traditional methods are given and the implementation can be described as:
superiority of the proposed scheme is proved. Finally, the 1) The peak detection block identifies the maximum
conclusion is drawn in Section VI. amplitude point sp on the region beyond the preset threshold.
2) The peak scale block generates a peak scaling factor,
II. OFDM AND PAPR
which can be expressed as K  s p  Ath  ˜ e jI , determined by
p

The basic idea of OFDM system is that the high speed

serial data stream is decomposed into a plurality of low the peak amplitude |sp | and the peak phase ϕp. Ath represents
speed data streams transmitted in parallel by using different the threshold level. The threshold should be reasonably
and mutually orthogonal subcarriers. In an OFDM system, selected, if the threshold is too low, resulting the higher evm,
each symbol modulates N multiple orthogonal sub-carriers. which increase the resource consumption caused by peak
Xk(K = 0, 1,…,N − 1) is the data symbol carried by the kth clipping and the risk of leakage. On the contrary, the peak-
subcarrier, the power-normalized complex baseband symbols to-average ratio is still high.
x(n) obtained by IFFT calculation are: 3) The cancellation pulse signal formed by the
cancellation pulse signal calculation block can be expressed
1 N 1 2S as K ˜ c pulse , cpulse is the prototype cancellation pulse signal
¦X
j nk
x ( n) k e N
,n 0,1,..., N  1 (1) generated from the spectral shape of the OFDM signal, so
N n 0
the spectral characteristics of the cancellation pulse signal is
nearly consistent with the original OFDM signal.
The PAPR of the transmitted signal can be defined as:
4) The peak cancellation is achieved by subtracting the
cancellation pulse signal from the peak point.
PAPR(dB) 10log10
max x(n) ^ 2
` , 0 d n d N 1 (2) Since the spectrum of the correcting signal is consistent
^
E x ( n)
2
` with the spectrum of the original OFDM signal, the PC-CFR
algorithm can reduce the peak-to-average ratio while
The complementary cumulative distribution function resisting the signal distortion and out-of-band radiation. In
(CCDF) of PAPR is one of the most commonly used the peak cancellation process, only the signal near the peak
indicators to measure PAPR reduction techniques, indicating point is processed, the implementation complexity and
the probability that PAPR exceed a given threshold z, then distortion are reduced, the EVM performance is better under
the CCDF of PAPR can be expressed as: the same PAPR.
2
FZmax
m
( z ) P( PAPR ! z ) 1  (1  e z ) N (3)

The complementary cumulative distribution function

(CCDF) of PAPR is one of the most commonly used
indicators to measure PAPR reduction techniques, indicating
the probability that PAPR exceed a given threshold z, then
the CCDF of PAPR can be expressed as:

¦ E  xˆ 
N 1
2
n  xn
EVM n 0
˜100% (4)
E xn   2
Figure 1. The PC-CFR algorithm.

IV. PARAMETER OPTIMIZATION

where xˆn represents a signal with distortion and xn represents
Table 1 shows the influence of various parameters on the
an ideal signal. Thus, the larger the EVM is, the more severe
performance of clipping in the PC-CFR algorithm. It can be

459

Authorized licensed use limited to: Nokia. Downloaded on October 23,2020 at 02:27:00 UTC from IEEE Xplore. Restrictions apply.
seen that the order and type of the filter are the key four filter functions in Matlab are compared to each other,
parameters that determine the pulse signal. namely firceqrip, firls, fir1, fir2. The firceqrip function
generates an equal-wave FIR filter. The firls function
TABLE I. THE INFLUENCE OF VARIOUS PARAMETERS IN THE PC- produces a least squares linear phase FIR filter, the fir1
CFR ALGORITHM ON THE PERFORMANCE OF PC-CFR function produces a FIR filter that is smoothed by a
Parameter Significance Influence Hamming window, and the fir2 function generates a FIR
Ath Threshold filter by frequency sampling and then IFFT.
Determined by the desired PAPR
level Fig. 5 shows a comparison of the effects of four filters on
CPG The number of
Ensure that the number of CPG is the signal power spectrum and can be seen that the transition
as small as possible to reduce band of the filter generated by the firls function is wider than
Number CPG
complexity without missing peaks
The peak Determined by the peak amplitude
the other three, and the out-of-band rejection is also the
N worst. The performance of the transition band of the filter
scaling factor and phase of the OFDM signal
The spectrum is consistent with the generated by the fir2 function is also not ideal. Compared to
cancellation
cpulse
pulse signal
spectrum of the original OFDM firceqrip, the fir1 function produces a better out-of-band
signal rejection of the filter. In addition, different filter functions
Determine the degree of convergence have little effect on the CCDF value of the suppressed signal,
Ntaps Filter order
of the cancellation pulse signal
Determine the shape of the pulse
as shown in Fig. 6. Table 2 shows that the EVM values are
Type Filter type also basically the same, with a difference of less than 0.1%
signal
Comprehensive consideration, the fir1 function is selected as
The criterion for selecting the filter type and order is that the filter generation function of the cancellation pulse.
the filter has a finite response in both the time and frequency
domains, allowing for a balance between out-of-band and in-
band distortion.
1) The experimental conditions are as follows: As
mentioned in the section III, the threshold needs to strike a
balance between EVM and PAPR. Considering this, the 7 dB
is set as the initial target. The OFDM signal is used for
simulation, the number of subcarriers is 1024 and the
subcarrier spacing is 15KHz, the IFFT/FFT points is 4096,
the sampling rate is 61.44MHz, the bandwidth is 15.36 MHz,
and the peak-to-average ratio is 9.63dB. In order to ensure
that the experimental results are not affected by the number
of CPG and iterations, the number of CPG and the number of
iterations is set to 6 and 2.
2) Consider how to choose the appropriate filter order.
According to the theory of digital signal processing, the
order of the digital filter affects the convergence speed and
side lobe size of the transition band. The higher the order is, Figure 2. The comparison of the signal PSD before and after suppression
the faster the transition band converges but the side lobes with the different filter ntaps.
increase slightly. In this paper, the firceqrip function in
Matlab is used to generate a low-pass equalripple filter. Fig.
2 shows the power spectral density comparison before and
after suppression when the filter order is fs/2, fs, 2fs, and 4fs.
fs is 61.44 which is 1/106 of the sampling rate, ntaps
represents the value of order. The red curve in the figure is
the power spectral density curve of the original signal before
the peak-to-average ratio suppression, and the blue curve is
the signal power spectral density curve after the suppression.
It can be observed that the lower filter order causes spectral
distortion of the transition band and affects out-of-band
rejection. but the higher order could increase the side lobes,
which also affects the performance of the out-of-band
rejection. However, the CCDF value of the suppressed signal
is not affected by filter order, which is proved in Fig. 3. The
EVM tends to be stable as it increases to a certain value as
the order increases, as shown in Fig. 4. Considering
comprehensively, the filter order is determined to be 2fs=122.
3) Consider the type of filter. After determining the filter Figure 3. The comparison of the signal CCDF after suppression with the
order, compare the effects of different filter to the signal. The different filter ntaps.

460

Authorized licensed use limited to: Nokia. Downloaded on October 23,2020 at 02:27:00 UTC from IEEE Xplore. Restrictions apply.
 which is why PW-CFR and NS-CFR use window functions.
The out-of-band rejection of PC-CFR mentioned in this
paper is only worse than NS-CFR a little, as shown in Fig.9.
In summary, the performance of the PC-CFR algorithm is
superior to other limiting algorithms.

TABLE II. THE COMPARISON OF THE SIGNAL EVM AFTER

SUPPRESSION WITH THE DIFFERENT FILTE

Filter type the signal EVM after suppression

firceqrip 3.9568 %

firls 3.9694 %

firls 3.9322 %

fir2 3.9103 %

Figure 4. The comparison of the signal EVM after suppression with the
different filter ntaps.

Figure 6. The comparison of the signal CCDF after suppression with the
different filter.

Figure 5. The comparison of the signal PSD after suppression with the
different filter.

V. SIMULATION RESULTE
The experimental validation was carried out, considering
a OFDM signal, the subcarrier frequency is set 15KHz, the
number of subcarrier is 1024, the IFFT/FFT points are 4096,
the 7dB is the initial target. Compare the optimized peak-to-
average ratio suppression algorithm PC-CFR with the
previously mentioned additional peak-to-average ratio
algorithm, namely peak windowing (PW-CFR) and noise
shaping (NS-CFR) and direct limiting (CFR). It can be seen
from Fig.7 that due to the window function, both PW-CFR
and NS-CFR appear to regenerate the peak, but the direct
clipping or the PC-CFR mentioned in this paper can suppress
the peak well. from Fig. 8, the CCDF results also show that
both achieve the desired goals. However, the out-of-band
rejection of direct limiting method is worse than PC-CFR, Figure 7. The PAPR suppression of four clipping method.

461

Authorized licensed use limited to: Nokia. Downloaded on October 23,2020 at 02:27:00 UTC from IEEE Xplore. Restrictions apply.
 [2] Vishnumohan, Nivish,Vellingiri, Krishnaveni,Thangavelu,
Kesavamurthy. Design of OFDM Transceiver for LTE System-A
Survey[J]. International Journal of Communications, Network and
System Sciences,2015,8(6).
[3] M. N. GEETHA, M. S. SAHANA and R. A. Lal, "The Relative Study
of PAPR Reduction Techniques for LTE OFDM Scheme," 2017
International Conference on Current Trends in Computer, Electrical,
Electronics and Communication (CTCEEC), Mysore, 2017, pp. 558-
561.
[4] Lee, Suchul,Kim, Chong-Kwon. D-Fi: A diversity-aware Wi-Fi using
an OFDM-based Bloom filter[P]. Network Protocols (ICNP), 2012
20th IEEE International Conference on,2012.
[5] Seyran Khademi,Thomas Svantesson,Mats Viberg,Thomas Eriksson.
Peak-to-Average-Power-Ratio (PAPR) reduction in WiMAX and
OFDM/A systems[J]. EURASIP Journal on Advances in Signal
Processing,2011,2011(1).
[6] Xin Guan,Yang Huang,Qingjiang Shi.Joint Subcarrier and Power
Allocation for Multi-UAV Systems[J].China
Communications,2019,16(01):47-56.
Fiure 8. The comparison of the signal CCDF after suppression with the [7] S.H. Han and J. H. Lee.:An overview of peak-to-average power ratio
different clipping method. reduction techniques for multicarrier transmission, IEEE Wireless
Commun 12(2):56-65,2005.
[8] Byung Moo Lee,Youngok Kim. An Adaptive Clipping and Filtering
Technique for PAPR Reduction of OFDM Signals[J]. Circuits,
Systems, and Signal Processing,2013,32(3).
[9] Electronics and Communications; Study Data from Myongji
University Provide New Insights into Electronics and
Communications (Neural Network Based Simplified Clipping and
Filtering Technique for PAPR Reduction of OFDM Signals)[J].
Computers, Networks & Communications,2015.
[10] Xuan G, Yang X, Zhao Y, et al.:Implementation of PAPR reduction
in OFDM system based on clipping technique with EVM, 2012 IEEE
14th International Conference on Communication Technology,
Chengdu, 93-98,2012.
[11] Mishra P, Amin M U.:PAPR reduction in OFDM using various
coding techniques, International Journal of Wireless and Mobile
Computing 15(1): 16-20,2018..
[12] K. Anoh, C. Tanriover, and B. Adebisi.:On the optimization of
iterative clipping and filtering for PAPR reduction in OFDM
systems,IEEE Access 5:12004C12013,2017.
[13] J. Armstrong.:Peak-to-average power reduction for OFDM by
repeated clipping and frequency domain filtering, Electronics Letters,
Figure 9. The comparison of the signal CCDF after suppression with the 28TH ,38(5):246-247,2002..
different clipping method. [14] K. Anoh, C. Tanriover, B. Adebisi and M. Hammoudeh.:A New
Approach to Iterative Clipping and Filtering PAPR Reduction
Scheme for OFDM Systems,IEEE Access 6:17533-17544,2018.
VI. CONCLUSION [15] Y.C. Wang and Z.Q. Luo.:Optimized iterative clipping and filtering
for PAPR reduction of OFDM signals, IEEE Trans.Commun
In this paper, PC-CFR is considered to be the most 59(1):33-37, 2011.
suitable algorithm for reducing PAPR in power-constrained [16] zhao C, R.J. Baxley,et al.:Constrained Clipping for Crest Factor
applications. After analyzing the basic principles, the Reduction in Multiple-user OFDM, 2007 IEEE Radio and Wireless
parameters that determine PC-CFR performance are Symposium :341-344, 2007.
identified. Taking CCDF, EVM and signal power spectrum [17] da Silva L D, de Lima E G.:A Novel Limiter with Application in
density as indicators to determine the clipping performance, Crest Factor Reduction Techniques for Wireless ommunications,2018
31st Symposium on Integrated Circuits and Systems Design (SBCCI):
a large number of simulations are performed on MATLAB to 1-6, 2018.
deeply analyze the influence of this parameter on PAPR
[18] Liu P, B.G. Stewart and S. McMeekin.:Peak-to-average power ratio
suppression and optimize these parameters to obtain the most reduction of FDRM, 2005 International Symposium on Intelligent
suitable value under consideration. It is clear from the Signal Processing and Communication Systems:429-432,2005.
performance analysis and comparison that the PC-CFR is [19] Hemphill E, Tarn H, Pecot M, et al.: High density WCDMA digital
better than other limiting PAPR reduction techniques. front end reference design[J]. Xilinx Application Notes, Document
Available on Request.
REFERENCES [20] E.Hemphill, S.Summerfield, G.Wang, and D. Hawke.: Cancellation
[1] J. Bingham.:Multicarrier modulation for data transmission: an idea Crest Factor Reduction Reference Design, Xilinx Application Note
whose time has come, IEEE Communm 28(5):5-14, 1990 XAPP1033, 2007.

462

Authorized licensed use limited to: Nokia. Downloaded on October 23,2020 at 02:27:00 UTC from IEEE Xplore. Restrictions apply.