IEEE TRANSACTIONS ON BROADCASTING, VOL. 57, NO.

2, JUNE 2011

291

Efcient PAPR Reduction in OFDM Systems Based on a Companding Technique With Trapezium Distribution
Shiann-Shiun Jeng, Member, IEEE, and Jia-Ming Chen, Student Member, IEEE

AbstractOrthogonal Frequency Division Multiplexing (OFDM) has become the most widely adopted technology in wireless communication systems. OFDM is limited mainly by its high Peak-to-Average Power Ratio (PAPR). A uniformly distributed nonlinear companding scheme efciently reduces PAPR with a low Bit Error Rate (BER). However, the uniformly distributed companding scheme cannot perform variably to satisfy the different performance requirements for the systems. Therefore, this work proposes a novel scheme that transforms the OFDM signals into a trapezium distribution. The uniformly distributed companding scheme is a special case of the proposed scheme. The general formulas of the proposed scheme are derived and the trade-off between PAPR reduction and BER performance is achieved by setting the value of a parameter. Then, the simulation results show the PAPR reduction and the BER over the AWGN and multipath channels, indicating that the proposed scheme provides a favorable trade-off between the PAPR reduction and the BER. Index TermsCompanding, OFDM, PAPR.

I. INTRODUCTION RTHOGONAL frequency division multiplexing (OFDM) systems have been extensively applied in wireless communication systems, e.g. Worldwide Interoperability for Microwave Access (WiMAX). OFDM systems have one major disadvantage, i.e. a very high Peak-to-Average Power Ratio (PAPR) at the transmitter [1]. When the OFDM signals with high PAPR are transmitted through a nonlinear device, such as a high-power amplier (HPA) or a digital-to-analog converter (DAC), a high peak signal generates out-of-band energy and in-band distortion. These degradations would seriously affect the performance of OFDM systems. In order to reduce the PAPR of OFDM systems, methods [2][11] have been studied for a few years. They fall into two categories [2]. The rst category of the PAPR reduction schemes changes the formation of the OFDM signals with high PAPR before multicarrier modulation, e.g. coding, selective mapping (SLM), and partial transmit sequence (PTS). The researches [4][6] reduce the computational complexity for the SLM scheme. The authors in [7] reduce the computational complexity for the PTS scheme. The researches [4][7] maintain the original BER of OFDM systems but require a large number of computational complexity and side information.
Manuscript received October 15, 2010; revised January 13, 2011; accepted January 19, 2011. Date of publication February 17, 2011; date of current version May 25, 2011. The research is supported in part by the National Science Council of the Republic of China, Taiwan, under Grant 99-2221-E-259-004. The authors are with the Department of Electrical Engineering, National Dong Hwa University, Hualien 97401, Taiwan (e-mail: ssjeng@mail.ndhu. edu.tw; d9523005@ems.ndhu.edu.tw). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TBC.2011.2112237

The second category of the PAPR reduction schemes transforms the OFDM signals after multicarrier modulation, such as companding techniques [8][13]. The -law companding scheme [8] is a common companding scheme. Then, the other companding scheme [9] utilizes a nonlinear operation to transform the original OFDM signals into the uniform distribution and efciently reduce PAPR with a low Bit Error Rate (BER). Similarly, the exponential companding scheme [10] transforms the amplitude of the original OFDM signals into the uniform distribution. The referred uniformly distributed companding schemes [9], [10] can keep the same average power as that of the original signal and signicantly outperform the -law companding scheme [8]. Furthermore, the piecewise nonlinear companding scheme [11] has been also proposed to compare with the exponential companding scheme [10], i.e. one uniformly distributeduniformly-distributed scheme [9]. The research [12] offers a trade-off between BER performance and PAPR reduction by adjusting two parameters, and its companding function is a piecewise function with three pieces. Thus, the companding function of the research [12] is more complicated than those of the others [8][11]. Two nonlinear companding schemes with iterative receiver are proposed in [13] and also outperform the -law companding scheme [8]. However, the reference [13] requires additional FFT and IFFT devices. The investigations [8][13] offer efcient PAPR reduction but get a little worse BER without side information. Although the uniformly-distributed scheme [9] can offer the efcient PAPR reduction with low BER, it can not fulll the performance requirements for various systems. Therefore, this work proposes a nonlinear companding scheme to transform the original OFDM signals into the trapezium distribution by regulating a parameter that governs the proposed trapezium distribution, and the companding function is a continuous function. The case of the uniform distribution is a special case of the proposed scheme. Then, the general formulas of the attenuation factor caused by the nonlinear operation and the theoretical complementary cumulative distribution function (CCDF) of PAPR for the proposed scheme are derived. The performance of the proposed scheme is simulated and compared with those of the uniformly-distributed scheme [9] and the piecewise nonlinear companding scheme [11], while the different modulation schemes and the AWGN or multipath channel are considered. The remainder of this paper is organized as follows. Section II introduces the PAPR problem of the OFDM system and the uniformly-distributed companding scheme [9]. Section III shows the derivation of the general formulas for the proposed scheme, and then the attenuation factor caused by the nonlinear operation is derived for the proposed scheme. Section IV shows the theoretical analysis of the CCDF of the PAPR for the proposed scheme and the simulation results of the CCDF of the PAPR, the

0018-9316/$26.00 2011 IEEE

292

IEEE TRANSACTIONS ON BROADCASTING, VOL. 57, NO. 2, JUNE 2011

Fig. 1. OFDM system using companding techniques.

TABLE I LIST OF THE USED SYMBOLS

. The PAPR for the OFDM signals within a symbol frame can be dened as (3) The CCDF of the PAPR for the original OFDM signals with subcarriers [1] can be represented as

(4) is a given threshold value. The OFDM signals are where , and the comtransformed by the companding function, . The inpanded signal can be represented as verse of the companding function is the decompanding function, . The uniformly-distributed companding scheme [9] transforms the OFDM signal into the uniform distribution in the . A random variable is considered interval, to be the amplitude of the companded signal, . Since is designed to be the uniform distribution, the CDF for the referred scheme can be represented as (5) Because and are strictly monotonically increasing functions, there are homologous inverse functions for both funccan be repretions. The homologous inverse function of sented as (6) Then, the relationship between sented as and can be repre-

power spectrum density (PSD) and the BER over the AWGN or multipath channels, while the different modulation schemes are used. Finally, Section V draws the conclusion for this work. II. PRELIMINARY Fig. 1 shows the block diagram of the OFDM system using companding techniques. Table I lists the symbol used in this is the -th complex modulated symbol. Then, is work. the IFFT output signal and can be represented as (1) Suppose that the input information symbols are statistically independent and identically distributed (i.i.d.). By the central limit theorem, as increases (e.g., ), the real and imaginary parts of become i.i.d Gaussian random variables with zero . Therefore, has the mean and variance distribution with the cumulative distribution function (CDF) [9] as following (2) where is the error function, is the variance of the original OFDM signals, and is the amplitude of the OFDM signal,

(7) Solving (7) yields the companding function of the referred scheme, (8). (8)

IEEE TRANSACTIONS ON BROADCASTING, VOL. 57, NO. 2, JUNE 2011

293

, the distribution of is uniform, as in When is equal to the uniformly-distributed scheme [9]. The CDF of the proposed scheme can be then represented as

(12) When is equal to , is equal to , i.e. is a strictly monotonically (12) is equal to (5). Because increasing function, it has a homologous inverse function. Then, and can be represented the relationship between as

Fig. 2. Trapezium distribution of the proposed scheme.

Substituting (2) and (6) into (8), the companding function of the referred scheme can be written as

(9) where is , and arg(.) is an angle function. In order the same as the power of , to make the power of is equal to . The OFDM signals are transmitted over the AWGN and multipath channels. According to the uniformly-distributed scheme [9], the received signals can be , where is the shown by is the AWGN noise. channel response with paths, and , , can be estimated by the The Fourier transform of , , can be pilot signals, and the inverse function of derived. In order to compensate for the effect caused by the channel distortion, the received signals are equalized when is assumed. Therefore, the received signals, which are through the analog-to-digital (A/D) converter, . The companding function, are rewritten as , is a strictly monotonically increasing function and has its inverse function, . The decompanding function of the referred scheme can be represented as (10) where is the inverse error function, and . plitude of the received signals, III. PROPOSED COMPANDING TECHNIQUE This section rstly derives the general formulas for the OFDM signal with the trapezium distribution, and then analyzes the attenuation effect of the proposed companding scheme. This work transforms the distribution of the original OFDM signal into the trapezium distribution in the interval . It generates the transformed signals, , and the trapezium distribution of the proposed scheme is shown in Fig. 2. The parameter, , is used to specify the forms of . Then, the the trapezium distribution, where probability density function (p.d.f.) of the proposed scheme can be represented as (11) is the am-

(13) Solving (13) yields the companding function of the proposed scheme, (14). (14) Substituting (2) and (12) into (14), the companding function of the proposed scheme can be expressed as

(15) The companding function is restricted to being a strictly monotonically increasing function and has an inverse transform function, which is the decompanding function. Therefore, the decompanding function of the proposed scheme can be represented as

(16) To make the power of the companded signal the same as that of the original OFDM signal, the power of the companded signal can be represented as

(17)

By substituting (11) into (17) and assuming it yields , given by (18).

, solving

(18)

294

IEEE TRANSACTIONS ON BROADCASTING, VOL. 57, NO. 2, JUNE 2011

Substituting (18) and into (15) and (16) enables the companding and decompanding functions of the proposed scheme to be rewritten as (19) and (20), respectively.

The recovered signal by the decompanding function can be represented as [11] (23) Eq. (23) reveals that the decompanding function increases the channel noise to . As shown in the analysis of the reference [11], if the decompanding function is not used, the equivand the alent noise is composed of the companding noise channel noise . Then, the equivalent noise with and without and the decompanding function can be represented as , respectively. denote the attenuation factor of the proposed scheme. Let It can be calculated and represented as (24) (shown at the bottom can be obtained to generate the of this page), where different equivalent noise by setting , and the different equivalent noise may result in different BER performance. IV. PERFORMANCE EVALUATION This section rstly derives the general formulas of the theoretical CCDF of the PAPR, evaluates the performance of the proposed scheme, and then compares with those of the uniformly-distributed and piece schemes [9], [11]. In order to compare the PAPR of the companding scheme with that of the original OFDM signals, the transform gain , which is the ratio of the PAPR of the original signals to that of the companded signals by the companding scheme, is dened as (25)

(19)

(20) The nonlinear operation results in the attenuation effect [11], [14], [15]. Therefore, the attenuation effect of the nonlinear companding technique has to be analyzed. According to [14], the companded signal, which is composed of an attenuated signal component and companding noise , can be represented as (21) where is an attenuation factor. The reference [14] shows that the OFDM signal is not stationary, and the reference [15] shows that an OFDM signal is time invariant. The reference [15] also depicts that is given by (22)

This work considers the OFDM system using companding schemes through AWGN and multipath channels. Cyclic prex (CP) is used to overcome the effect of intersymbol interference (ISI). After compensating for the effect caused by the channel . distortion, the received signal can be expressed as

where is the PAPR of the companded signal. and are the largest and average power of the original OFDM and are the largest and average signals, respectively. power of the companded signals by the companding schemes, respectively. For the theoretical analysis of the CCDF of the PAPR, the transform gain of the proposed scheme must be determined by

(24)

IEEE TRANSACTIONS ON BROADCASTING, VOL. 57, NO. 2, JUNE 2011

295

computing the largest and average power of the companded siginto (19) yields the largest power of the nals. Substituting companded signals by the proposed scheme, given by (26).

(26) The average power of the companded signals by the proposed scheme is designed to be the same as that of the original OFDM . Thus, the CCDF of the PAPR achieved signal, i.e. using the proposed scheme can be expressed as

. The AWGN and multipath channels the input streams, are used as the channel models in this simulation. When a multipath channel model is used, the Stanford University Interim 5 (SUI-5) channel [17], corresponding to a multipath and power prole prole with three paths of delay [0 4 10] [0 5 10] dB, is applied in the simulations. Then, the CP with a length of 1/8 OFDM symbol, corresponding to the CP duration of 11.4 , is utilized to overcome the effect of ISI. The SUI channel model is suitable for evaluating the performance of IEEE 802.16 systems. For most wireless systems, high power-ampliers (HPAs) are normally used to provide adequate transmit power. Moreover, a solid state power amplier (SSPA) [18], [19] is considered in this work and can be represented as (29) and are the input and output signals, respecwhere tively. The parameter is set to be the typical value 2 in this work. Furthermore, this simulation assumes perfect frame synchronization and symbol timing and absence of frequency offset effect. Then, this study considers that the word lengths of A/D and D/A converters are long enough to ignore the quantization error. In addition to comparing this work with the uniformly-distributed scheme [9], the piecewise companding scheme [11] is also compared in this work. The piecewise scheme is composed of the linear operation and the nonlinear operation. The companding function of the piecewise scheme is represented as

(27) Substituting and into (27) yields (28).

(28) According to (28), the different PAPR reduction can be obtained by adjusting the value of , i.e. . By using (24) and (28), therefore, the proposed scheme can achieve a good trade-off between the BER performance and PAPR reduction . by adjusting the value of , i.e. Fig. 1 shows the OFDM system that employs companding schemes. In the simulation of performance evaluation, the number of the subcarriers is 128. The oversampling factor is set to 4. Therefore, the length of the FFT and IFFT is 512. The [16]. Quadrature Phase-Shift useful symbol time is 91.4 Keying (QPSK) and 16-Quadrature Amplitude Modulation (16-QAM) schemes are used as the modulation schemes in this work. QPSK is used in Figs. 48, and 16-QAM is used in Figs. 9 and 10. The random symbol generator generates

(30) The proposed scheme can provide performance that meets the requirements for various systems by adjusting the value of . However, the uniformly-distributed scheme and the piecewise scheme cannot deliver the performance that satises the various requirements for the systems. For example, in this work, to meet , the paramthe performance of . To meet the pereter of the proposed scheme, , is set to at , is set to , formance of while the QPSK scheme is used. Fig. 3 plots the transfer curves of the uniformly-distributed scheme [9], the piecewise scheme and . The [11] and the proposed scheme with transfer curve of the proposed scheme is varied by adjusting . This work compares the performance of the proposed scheme with those of the uniformly-distributed and piecewise schemes. Figs. 48 are obtained by using the QPSK scheme. Fig. 4 plots the waveforms of the original OFDM signal, the uniformly-distributed scheme, the piecewise scheme and the proposed scheme and . The waveforms of the companded with signals by the proposed scheme are limited to smaller range than that of the original OFDM signal. Fig. 5 shows the theoretical and simulation results of the CCDF of the PAPR for the proand , indicating that the posed scheme with theoretical and simulation results for the proposed scheme with

296

IEEE TRANSACTIONS ON BROADCASTING, VOL. 57, NO. 2, JUNE 2011

Fig. 3. Transfer curves of the referred schemes and the proposed scheme with a = h and = h .

=2 5

23

Fig. 5. CCDF of the PAPR for the referred schemes and the proposed scheme with a = h and = h .

=2 5

23

Fig. 4. Waveforms for the original OFDM signal, the referred schemes, and the proposed scheme with a = h and = h .

=2 5

23

2=5h

Fig. 6. PSD for the referred schemes and the proposed scheme with a and = h .

23

are very close under and the difference between the theoretical and simulation results for the prois about 0.5 dB. When CCDF posed scheme with , the greater CCDF, the more difference beis greater than tween the theoretical and simulation results becomes because the theoretical CCDF of the PAPR in (2) is an approximation. The proposed scheme can offer specic PAPR reduction by adjusting the value of , although the simulation results are a little different from the theoretical analysis. Fig. 5 also shows the CCDF of the PAPR achieved using the uniformly-distributed scheme, the piecewise scheme and the proposed scheme with and , indicating that the CCDF of the PAPR of is lower than those of the the proposed scheme with uniformly-distributed and piecewise schemes, i.e. decreasing by about 0.5 and 0.7 dB, respectively. Fig. 6 presents the PSD of the uniformly-distributed scheme, the piecewise scheme and and , indicating the proposed scheme with that the spectrum side-lobe of the PSD of the proposed scheme is similar to those of the uniformly-distributed with and piecewise schemes. Fig. 6 also reveals that the spectrum side-lobe of those of the uniformly-distributed and piecewise schemes. Fig. 7 demonstrates the BER over the AWGN channel

for the uniformly-distributed scheme, the piecewise scheme and and , while Fig. 8 the proposed scheme with displays the BER over the SUI-5 channel for the uniformly-distributed scheme, the piecewise scheme and the proposed scheme and . As shown in Figs. 7 and 8, the prowith posed scheme with yields a lower BER than do the uniformly-distributed and piecewise schemes. Then, Figs. 9 and 10 are generated by using the 16-QAM scheme. Figs. 9 and 10 show the BER over the AWGN and SUI-5 channels for the uniformly-distributed scheme, the piecewise scheme and the proand , respectively. The posed scheme with is BER achieved using the proposed scheme with lower than those achieved using the uniformly-distributed and piecewise schemes. When the various modulation schemes and the different channels are considered, the proposed scheme with yields a little lower BER than do the uniformly-distributed and piecewise schemes. As revealed by the simulation results above, the performance of the proposed scheme varies with the value of . When is set , the PAPR reduction of the proposed scheme exceeds to those of the uniformly-distributed and piecewise schemes, but the BER of the former is a little worse than those of the both

IEEE TRANSACTIONS ON BROADCASTING, VOL. 57, NO. 2, JUNE 2011

297

Fig. 7. BER over AWGN channel for the referred schemes and the proposed = h and = h when the QPSK modulation is used. scheme with a

=2 5

23

Fig. 10. BER over SUI-5 channel for the referred schemes and the proposed scheme with a = h and = h when the 16-QAM modulation is used.

=2 5

23

lower BER as increases and offers more efcient PAPR reduction as decreases. Accordingly, based on the requirements for the system, the proposed scheme can outperform both of the referred schemes. The proposed scheme provides a favorable trade-off between the PAPR reduction and the BER, while the QPSK or 16-QAM schemes and the AWGN or SUI-5 channels are considered. V. CONCLUSION Companding techniques can solve the high PAPR problem for OFDM systems. Two of companding scheme, i.e. the uniformlydistributed companding scheme and the piecewise companding scheme, are studied herein to provide efcient PAPR reduction with a low BER. However, both of the referred schemes cannot deliver the performance that satises the various requirements of the systems. In this work, the distribution of the OFDM signal is transformed into the trapezium distribution, and the general formulas for the proposed scheme are derived that enable the desired performance to be achieved by controlling the parameter. The uniformly-distributed companding scheme is the special case of the proposed scheme. Then, the simulation results reveal that the proposed scheme may offer the more efcient PAPR reduction or the lower BER than the uniformly-distributed and piecewise schemes under the condition of efcient PAPR reduction or efcient BER performance. Whether the QPSK or 16-QAM schemes and the AWGN or SUI-5 channels are considered, the simulation results indicate that the specic performance of the proposed scheme can be achieved by varying the parameter , to fulll the performance requirements of various systems. Therefore, the proposed scheme can deliver a good trade-off between the PAPR reduction and the BER. ACKNOWLEDGMENT The authors would like to thank the reviewers for their kind suggestions and review of this manuscript. REFERENCES
[1] J. Tellado, Multicarrier transmission with low PAR, Ph.D. dissertation, Stanford Univ., Stanford, CA, 1998.

Fig. 8. BER over SUI-5 channel for the referred schemes and the proposed scheme with a = h and = h when the QPSK modulation is used.

=2 5

23

Fig. 9. BER over AWGN channel for the referred schemes and the proposed = h and = h when the 16-QAM modulation is used. scheme with a

=2 5

23

referred schemes. When is set to , the BER of the proposed scheme is lower than those of the uniformly-distributed and piecewise schemes, and the PAPR reduction of the proposed scheme is a little worse than those of the uniformly-distributed and piecewise schemes. Therefore, the proposed scheme obtains

298

IEEE TRANSACTIONS ON BROADCASTING, VOL. 57, NO. 2, JUNE 2011

[2] S. H. Han and J. H. Lee, An overview of peak-to-average power ratio reduction techniques for multicarrier transmission, IEEE Wireless Commun., vol. 12, pp. 5665, Apr. 2005. [3] T. Jiang and Y. Wu, An overview: Peak-to-average power ratio reduction techniques for OFDM signals, IEEE Trans. Broadcast., vol. 54, no. 2, pp. 257268, Jun. 2008. [4] C. P. Li, S. H. Wang, and C. L. Wang, Novel low-complexity SLM schemes for PAPR reduction in OFDM systems, IEEE Trans. Signal Process., vol. 58, no. 5, pp. 29162921, May 2010. [5] J. H. Wen, S. H. Lee, and C. C. Kung, SLM-based data position permutation method for PAPR reduction in OFDM systems, in Wireless Communication and Mobile Computing. Hoboken, NJ: Wiley InterScience, 2008. [6] S. H. Wang and C. P. Li, A low-complexity PAPR reduction scheme for SFBC MIMO-OFDM systems, IEEE Signal Process. Lett., vol. 16, no. 11, pp. 941944, Nov. 2009. [7] N. T. Hieu, S. W. Kim, and H. G. Ryu, PAPR reduction of the low complexity phase weighting method in OFDM communication system, IEEE Trans. Consum. Electron., vol. 51, no. 3, pp. 776782, Aug. 2005. [8] X. B. Wang, T. T. Tjhung, and C. S. Ng, Reduction of peak-to-average power ratio of OFDM system using a companding technique, IEEE Trans. Broadcast., vol. 45, no. 3, pp. 303307, Sep. 1999. [9] T. Jaing, W. Xiang, P. C. Richardson, D. Qu, and G. Zhu, On the nonlinear companding transform for reduction in PAPR of MCM, IEEE Trans. Wireless Commun., vol. 6, no. 6, pp. 20172021, Jun. 2007. [10] T. Jiang, Y. Yang, and Y. Song, Exponential companding transform for PAPR reduction in OFDM systems, IEEE Trans. Broadcast., vol. 51, no. 2, pp. 244248, June 2005.

[11] J. Hou, J. Ge, D. Zhai, and J. Li, Peak-to-average power ratio reduction of OFDM signals with nonlinear companding scheme, IEEE Trans. Broadcast., vol. 56, no. 2, pp. 258262, Jun. 2010. [12] J. Hou, J. H. Ge, and J. Li, Trapezoidal companding scheme for peak-to-average power ratio reduction of OFDM signals, Electron. Lett., vol. 45, no. 25, pp. 13491351, Dec. 2009. [13] T. Jiang, W. Yao, P. Guo, Y. Song, and D. Qu, Two novel nonlinear companding schemes with iterative receiver to reduce PAPR in multicarrier modulation systems, IEEE Trans. Broadcast., vol. 51, no. 2, pp. 268273, Jun. 2006. [14] D. Dardari, V. Tralli, and A. Vaccari, A theoretical characterization of nonlinear distortion effects in OFDM systems, IEEE Trans. Commun., vol. 48, no. 10, pp. 17551764, Oct. 2000. [15] P. Banelli and S. Cacopardi, Theoretical analysis and performance of OFDM signals in nonlinear AWGN channels, IEEE Trans. Commun., vol. 48, pp. 430441, Mar. 2000. [16] J. G. Andrews, A. Ghosh, and R. Muhamed, Fundamentals of WiMAX: Understanding Broadband Wireless Networking. Englewood Cliffs, NJ: Prentice Hall, 2007. [17] C. Oestges and B. Clerckx, MIIMO Wireless Communications: From Real-World Propagation to Space-Time Code Design. San Diego, CA: Academic Press, 2007, pp. 409410. [18] E. Costa, M. Midrio, and S. Pupolin, Impact of amplier nonlinearities on ofdm transmission system performance, IEEE Trans. Commun., vol. 3, pp. 3739, Feb. 1999. [19] A. Conti, D. Dardari, and V. Tralli, On the performance of cdma systems with nonlinear amplier and awgn, in Proc. 6th IEEE Int. Symp. Spread Spectrum Techniques and Applications (ISSSTA), Sep. 2000, vol. 1, pp. 197202.