Selected mapping technique for reducing PAPR of single-carrier signals

Transcription

1 WIRELESS COMMUNICATIONS AND MOBILE COMPUTING Wirel. Commun. Mob. Comput. (2016) Publihed online in Wiley Online Library (wileyonlinelibrary.com) RESEARCH ARTICLE Selected mapping technique for reducing PAPR of ingle-carrier ignal Amnart Boonkajay * and Fumiyuki Adachi Reearch Organization of Electrical Communication, Tohoku Univerity, Katahira, Aoba-ku, Sendai-hi, Miyagi, Japan ABSTRACT Single-carrier tranmiion with frequency-domain equalization (SC-FDE) i widely known a a promiing tranmiion technique providing low error probability with low peak-to-average power ratio (PAPR) of tranmit ignal. However, the low-papr property of SC-FDE cannot be maintained if multi-level data modulation i introduced. The low-papr property of SC-FDE can be maintained by applying tranmit filtering with roll-off factor at the expene of pectrum efficiency. In thi paper, we propoe two type of elected mapping (SLM) to reduce the PAPR of SC-FDE tranmit ignal. The firt SLM technique i conducted in the frequency domain, where the phae rotation i applied to ubcarrier imilar to the SLM technique for orthogonal frequency diviion multiplexing tranmiion. The econd SLM technique i conducted in the time domain, where the phae rotation i applied directly to data-modulated ymbol equence. Computer imulation confirm that both SLM technique are able to reduce the PAPR of SC-FDE ignal without ignificant degradation of bit-error rate performance and pectrum efficiency. Copyright 2016 John Wiley & Son, Ltd. KEYWORDS ingle-carrier (SC) tranmiion; elected mapping (SLM); frequency-domain equalization (FDE); peak-to-average power ratio (PAPR) *Correpondence Amnart Boonkajay, Reearch Organization of Electrical Communication, Tohoku Univerity, Katahira, Aoba-ku, Sendai-hi, Miyagi, Japan. amnart@riec.tohoku.ac.jp 1. INTRODUCTION Ditributed antenna network [1], which achieve high pectrum efficiency (SE) and energy efficiency, i conidered a a promiing network architecture for the fifth-generation ytem. However, broadband wirele channel i characterized a a frequency-elective fading channel, in which inter-ymbol interference ignificantly degrade the biterror rate (BER) performance [2]. Orthogonal frequency diviion multiplexing (OFDM) [3] i a robut multicarrier tranmiion technique, but it high peak-to-average power ratio (PAPR) of tranmit ignal i the main drawback. On the other hand, ingle-carrier tranmiion with frequency-domain equalization (SC-FDE) [4] i attractive for uplink communication becaue of it lower PAPR, while the ue of FDE can effectively uppre the impact of inter-ymbol interference. Peak-to-average power ratio of SC ignal become higher a higher level modulation i ued. PAPR increae by 1 db in SC-FDE when the modulation level i changed from 4 quadrature amplitude modulation (QAM) to 16 QAM [5]. Thi fact indicate the neceity of PAPR reduction technique even for SC tranmiion. Even though ditributed antenna network can reduce the tranmit power becaue of horter range of tranmiion, PAPR reduction remain a ignificant iue in fifth-generation in order to further reduce the power conumption of linear power amplifier at the mobile terminal. We are alo aiming at achieving very low-papr tranmiion (for example, 3 db lower than the conventional SC-FDE) in uplink tranmiion in order to further improve the energy efficiency of uer equipment. In general, PAPR of SC ignal can be reduced by applying pule haping. Square-root raiedcoine (SRRC) pule with roll-off factor of 0.5 can achieve 2 db reduction of tranmit PAPR in 16 QAM modulation [6], but the pule with roll-off factor of 0.5 require 1.5 time of tranmiion bandwidth and conequently degrade SE. Many tranmit filtering technique providing lower PAPR than SRRC pule have been propoed a the technique [7,8], but the PAPR reduction i obviou only when high roll-off factor i ued. A tranmit filtering baed on minimum variance of intantaneou tranmit power criterion can reduce the PAPR without SE degradation [9], but the PAPR reduction i marginal. Copyright 2016 John Wiley & Son, Ltd.

2 Selected mapping technique for reducing PAPR of ingle-carrier ignal A. Boonkajay and F. Adachi The exiting PAPR reduction technique for SC-FDE [7 9] are pectrum inefficient a dicued previouly. Thi motivated u to develop a pectrum-efficient PAPR reduction technique for filtered SC ignal. A imple approach i to tudy the PAPR reduction technique for OFDM tranmiion. Many PAPR reduction technique were extenively tudied for OFDM tranmiion uch a clipping [10] and coding [11]. However, thee technique till have drawback: clipping intentionally generate waveform ditortion, and coding require redundant bit. Among variou PAPR reduction technique, elected mapping (SLM) [12,13] i very attractive becaue it i a mall-overhead, ditortionle PAPR reduction technique with imple implementation by applying phae rotation to the ubcarrier. It i hown in [12,13] that the SLM can effectively reduce the PAPR of OFDM ignal with moderate computational complexity, and it PAPR-complexity tradeoff i much better than partial tranmit equence technique [14]. However, the SLM technique ha been extenively tudied only for OFDM ignal. In thi paper, we introduce two approache for implementing SLM in SC-FDE tranmiion and propoe two SLM technique correponding to thoe implementation approache for reducing the PAPR of SC-FDE ignal. The firt SLM technique i implemented by utilizing the fact that SC-FDE ignal can be generated in frequency domain, that i, by inerting dicrete Fourier tranform (DFT) into conventional OFDM tranmitter a a linear precoder [15]. Thi technique i called frequency-domain SLM (FD-SLM). The frequency component obtained by DFT are multiplied with a elected phae rotation pattern before performing invere DFT (IDFT). The phae rotation pattern that provide the minimum PAPR i elected. The BER performance i kept intact; however, the tranmiion of o-called ide information i neceary. On the other hand, the econd SLM technique i implemented by utilizing the fact that the peak power of SC-FDE ignal depend on an arrangement of data ymbol in a tranmiion block [16]. In the econd SLM technique, the phae rotation pattern i directly multiplied to data-modulated equence block in time domain, in order to eliminate the ymbol arrangement that caue high peak power. Thi technique i called the time-domain SLM (TD-SLM). The reultant time-domain tranmit block candidate after multiplying each phae rotation pattern are pule haped by Nyquit pule (which i equivalent to a proce of applying a band-limited filtering in frequency domain), and then the bet phae rotation pattern that provide the minimum PAPR i elected. The novelty and contribution of thi paper can be ummarized a follow: A problem of high-papr ignal in SC-FDE due to higher level modulation and then, the requirement of pectrum-efficient PAPR reduction technique for SC-FDE are dicued. To reduce the PAPR without ignificant SE degradation, two PAPR reduction technique baed on SLM for SC-FDE are propoed. Thi i our new contribution in thi paper. The multiplication of elected phae rotation pattern can be carried out either in frequency domain or time domain. We alo would like to mention that the phae rotation ued in conventional SLM [12,13] for OFDM ignal i available only in frequency domain. Performance evaluation of SC-FDE uing the propoed SLM technique are performed by computer imulation in apect of PAPR, BER, throughput and computational complexity and are compared with that of the conventional SC-FDE (DFT-precoded SC-FDE) and SRRC-filtered SC-FDE. The evaluation reult how that the propoed SLM technique can reduce the PAPR of SC-FDE ignal without ignificant degradation on BER and throughput performance. The TD-SLM provide lower PAPR than FD-SLM if the ame number of phae rotation pattern i ued. The analyi on PAPR reduction effect of both FD-SLM and TD-SLM i alo provided in order to clarify why the TD-SLM give lower PAPR than FD-SLM. Note that the SLM require moderate computational complexity, and the tudy of complexity reduction i left a our future work becaue the problem of PAPR and pectrum efficiency are conidered a higher priority than computational complexity in thi paper. The remaining of thi paper i organized a follow. FD-SLM algorithm and it application in multi-level modulated SC-FDE are explained in Section 2. TD-SLM algorithm and it application in SC-FDE are decribed in Section 3. Section 4 how the performance evaluation. Finally, Section 5 conclude the paper. 2. SC-FDE USING FD-SLM An explanation of FD-SLM algorithm, together with it implementation in conventional SC-FDE block tranmiion, i provided in thi ection. Tranmitted and received ignal block are repreented a column vector, where the ignal proceing in each tage i repreented by matrix throughout thi paper FD-SLM algorithm Selected mapping [12,13] ha been introduced a a frequency-domain-baed ditortionle PAPR reduction technique with mall overhead. In thi paper, an FD-SLM i implemented to SC-FDE tranmiion, where the phae rotation i applied to the frequency-domain SC ignal after DFT and prior to IDFT. The FD-SLM algorithm and it ignal proceing are imply illutrated by Figure 1. Auming that an N c -length time-domain tranmit block i repreented by a vector D Œ.0/,.1/, :::,.N c 1/ T, PAPR of the time-domain tranmit ignal calculated over an overampled tranmiion block i expreed by Wirel. Commun. Mob. Comput. (2016) 2016 John Wiley & Son, Ltd.

3 A. Boonkajay and F. Adachi Selected mapping technique for reducing PAPR of ingle-carrier ignal Figure 1. Signal proceing in frequency-domain elected mapping. PAPR D n o max j.n/j 2, n D 0, V 1, V 2, :::, N c 1 1 P Nc, (1) 1 N c nd0 j.n/j2 where V i overampling factor. According to Figure 1, a et of U different N c N c diagonal matrice repreenting phae rotation pattern P u D diagœp u.0/, :::, P u.n c 1/, u D 0 U 1 i defined. Candidate for the frequency-domain tranmit ignal block are generated by multiplying the phae-rotation matrix to the frequency-domain component after tranmit filtering and before IDFT, repreented by S D H T F Nc d,whereh T and F Nc repreent tranmit filtering and N c -point DFT operation, repectively. The vector d D Œd.0/, :::, d.n c 1/ T repreent the time-domain tranmit ymbol block. The definition of H T and F Nc will be decribed in detail in Section 2.2. In addition, the phae-rotation matrix for the firt candidate P 0 i et to be an N c N c identity matrix I Nc a a repreentative of original tranmit block, where the other candidate are generated either in determinitic or random approach. We aume the phae rotation pattern to be real valued and unit magnitude, that i, P u.k/ D 1, k D 0 N c 1. The reaon of etting P u.k/ a real valued i to reduce the number of complex-valued multiplication of the tranmitter, epecially when the low-papr SC-FDE are expected to be exploited in uplink communication. The unit-magnitude property i required in order to meet tranmit power contraint. In thi paper, we ue a et of phae rotation pattern generated randomly, which i confirmed in [17,18], where it provide the bet PAPR reduction performance. The intantaneou PAPR of tranmit block candidate u D F H N c P u S for all u are calculated by referencing (1), and the elected tranmit ignal Ou D Œ Ou.0/, :::, Ou.N c 1/ T with the correponding elected phae rotation pattern index Ou i elected by the following criterion: Ou D arg min PAPR u D F H N ud0,1,:::,u 1 c P u H T F Nc d. (2) Note that the matrix and vector repreentation for tranmit ignal proceing will be decribed in more detail in Section Tranceiver ytem model Single-uer N c -length SC-FDE block tranmiion with N g -length cyclic prefix (CP) inertion i conidered in thi paper. DFT and IDFT are employed in the tranceiver for reaching frequency-domain proceing, at which the tranmit filtering and receive FDE can be applied a imple one-tap multiplication. The SC-FDE uing FD-SLM tranceiver i illutrated in Figure Tranmitter. Tranmitter of SC-FDE equipped with FD-SLM i hown in Figure 2(a). We begin with a block coniting of N c data-modulated ymbol d D Œd.0/, :::, d.n c 1/ T. The block d i tranformed into frequency domain by N c - point DFT, yielding the frequency-domain ignal vector D D ŒD.0/, :::, D.N c 1/ T a D D F Nc d, (3) where the N c -point DFT matrix F Nc i given by F Nc D p 1 1 e j2.1/.1/ Nc e j2.1/.nc 1/ Nc Nc , (4) 5 1e j2.nc 1/.1/ Nc e j2.nc 1/.Nc 1/ Nc and it Hermitian tranpoe repreent an invere operation. Next, D i multiplied by tranmit filtering matrix H T D diagœh T.0/, :::, H T.N c 1/. We aume the tranmit filtering in thi paper to be SRRC filtering with roll-off factor D 0, that i, ideal rectangular filtering, reulting in H T.k/ D 1forallk D 0 N c 1. The frequency-domain filtered ignal i repreented by S D H T F Nc d. (5) The filtered ignal S i then ued a input ignal in FD-SLM algorithm a decribed in Section 2.1 and Figure 1. The candidate are generated by multiplying S with P u, u D 0 U 1, following by IDFT Wirel. Commun. Mob. Comput. (2016) 2016 John Wiley & Son, Ltd.

4 Selected mapping technique for reducing PAPR of ingle-carrier ignal A. Boonkajay and F. Adachi Figure 2. Tranceiver ytem model of SC-FDE uing FD-SLM. SC-FDE, ingle-carrier tranmiion with frequency-domain equalization; FD-SLM, frequency-domain elected mapping; DFT, dicrete Fourier tranform; IDFT, invere DFT; PAPR, peak-to-average power ratio. operation for obtaining the time-domain tranmit block candidate u. Selection i employed by referencing to (2) for obtaining the time-domain tranmit block providing the lowet PAPR among U candidate, that i, Ou, together with the elected phae rotation pattern P Ou.In ummary, the time-domain tranmit ignal after paing through all procee Ou D Œ Ou.0/, :::, Ou.N c 1/ T can be expreed a Ou D F H N c P Ou H T D D F H N c P Ou H T F Nc d. (6) In cae that H T i SRRC filtering with roll-off factor D 0, (6) can be implified a Ou D F H N c P Ou F Nc d. (7) Finally, the lat N g ample of tranmit block are copied a a CP and inerted into the guard interval, then a CPinerted ignal block of N g C N c ample i tranmitted Receiver. The wirele propagation channel in thi paper i aumed to be a ymbol-paced L-path frequency-elective block fading channel [2], where it impule repone i given by L 1 X h./ D h l ı. l /, (8) ld0 where h l and l are complex-valued path gain and time delay of the l-th path, repectively. ı./ i the delta function. Time-domain received ignal vector after CP removal r Ou D Œr Ou.0/, :::, r Ou.N c 1/ T i expreed by r Ou D 2E T h Ou C n, (9) where Ou D F H N c P Ou H T F Nc d i obtained from (6). E i ymbol energy, and n i noie vector in which each element i zero-mean additive white Gauian noie having the variance 2N 0 =T with T i ymbol duration and N 0 being the one-ided noie power pectrum denity. Channel repone matrix h i a circular matrix repreenting time-domain channel repone, which i 2 3 h 0 h L 1 h 1. h h h D 0 0 h L 1. (10). h L 1 h h L 1 h 0 The received ignal i tranformed into frequency domain by N c -point DFT, obtaining the frequency-domain received ignal vector R Ou D ŒR Ou.0/, :::, R Ou.N c 1/ T a Wirel. Commun. Mob. Comput. (2016) 2016 John Wiley & Son, Ltd.

5 A. Boonkajay and F. Adachi Selected mapping technique for reducing PAPR of ingle-carrier ignal R D D D 2E T F Nc h Ou C F Nc n 2E F Nc hf H N T c P Ou H T D C F Nc n 2E H c P T Ou H T D C N, where the frequency-domain channel repone H c i (11) F Nc hf H N c D diagœh c.0/, :::, H c.n c 1/ H c. (12) Frequency-domain equalization baed on minimum mean-quare error criterion (MMSE-FDE) i multiplied to R Ou in order to mitigate the effect of frequency-elective fading. The FDE i repreented by N c N c diagonal matrix W R D diagœw R.0/, :::, W R.N c 1/, which i equivalent to imple one-tap multiplication. The FDE weight with repect to each frequency index W R.k/ are derived o a to minimize the mean-quare error between frequencydomain tranmit ignal D (defined by (3)) and received ignal OD D W R R Ou, yielding Hc W R.k/ D.k/P Ou.k/H T.k/ jh c.k/p Ou.k/H T.k/j 2. (13) C.E =N 0 / 1 It i oberved from (13) that the information of channel repone, tranmit filtering coefficient, and the elected phae rotation pattern are indipenable. In general, the receiver know H c.k/ through channel etimation [19], where the tranmit filtering coefficient i predetermined. To inform the receiver about the phae rotation pattern, at leat log 2 U bit are required a a ide information containing the pattern number. We aume the perfect channel etimation and ideal ide information detection (i.e., error-free ide information detection) in thi paper for implicity. Finally, the frequency-domain equalized ignal OD D W R R Ou i tranformed back into time domain by N c -point IDFT. The time-domain received ignal before demodulation Od D ŒOd.0/, :::, Od.N c 1/ T i expreed by Od D F H N c W R R Ou 2E D F H N T c W R H c P Ou D C F H N c W R N. 3. SC-FDE USING TD-SLM (14) In thi ection, an explanation of TD-SLM algorithm and it implementation in SC-FDE are provided. The TD-SLM algorithm i different from the FD-SLM becaue the generation of tranmit block candidate (i.e., phae rotation) i employed in time domain. Thi alo reult in the difference between the tranceiver of SC-FDE uing TD-SLM and SC-FDE uing FD-SLM TD-SLM algorithm The TD-SLM algorithm and it ignal proceing are depicted in Figure 3. A et of U different N c N c diagonal phae rotation pattern matrice P u D diagœp u.0/, :::, P u.n c 1/, u D 0 U 1 i defined, which i imilar to FD-SLM. However, P u i multiplied to time-domain tranmit ymbol block d intead of in frequency domain, and then obtaining the time-domain candidate d u D P u d. The pattern generation i exactly the ame a in FD-SLM, that i, P u.n/ D 1, n D 0 N c 1 i generated in random approach. (Note that we ue the time index n intead of frequency index k for emphaizing that phae rotation i applied in time domain.) All of the time-domain block candidate in U branche are then paed through tranmit ignal proceing (the detail are decribed in Section 3.2). The intantaneou PAPR of time-domain tranmit ignal candidate after paing through tranmit ignal proceing u D Œ u.0/, :::, u.n c 1/ T are calculated baed on (1), and the elected tranmit ignal Ou D Œ Ou.0/, :::, Ou.N c 1/ T with the correponding elected phae rotation pattern index Ou i elected by the following criterion: Ou D arg min PAPR u D F H N ud0,1,:::,u 1 c H T F Nc P u d. (15) It i oberved that (15) i different from (2) becaue P u i multiplied before DFT operation. Note that the matrix Figure 3. Signal proceing in TD-SLM. TD-SLM, time-domain elected mapping; DFT, dicrete Fourier tranform; IDFT, invere DFT; PAPR, peak-to-average power ratio. Wirel. Commun. Mob. Comput. (2016) 2016 John Wiley & Son, Ltd.

6 Selected mapping technique for reducing PAPR of ingle-carrier ignal A. Boonkajay and F. Adachi and vector repreentation for tranmit ignal proceing in (15) will be decribed in more detail in Section Tranceiver ytem model Similar to Section 2.2, ingle-uer N c -length block tranmiion with N g -length of CP inertion i aumed. In a ytem uing TD-SLM, time-domain ymbol are mapped by multiplying with phae rotation pattern prior to applying DFT. Tranceiver ytem model of SC-FDE uing TD-SLM i hown in Figure Tranmitter. Tranmitter of SC-FDE uing TD-SLM i hown in Figure 4(a). A tranmit block coniting of N c datamodulated ymbol d D Œd.0/, :::, d.n c 1/ T i ued for generating U candidate for SLM d u by multiplying with different phae rotation pattern. The u-th tranmit block candidate i expreed by d u D P u d, (16) where P u D diag ŒP u.0/, :::, P u.n c 1/ repreent phae rotation pattern matrix. The detail of phae rotation pattern generation i already dicued in Section 3.1. Then, each tranmit block candidate i tranformed into frequency domain by N c -point DFT, yielding frequencydomain tranmit ignal of the u-th candidate D u D ŒD u.0/, :::, D u.n c 1/ T a D u D F Nc P u d, (17) where N c -point DFT matrix F Nc i already defined in (4). Next, D u i multiplied by the tranmit filtering matrix H T D diag ŒH T.0/, :::, H T.N c 1/, obtaining the frequency-domain ignal after filtering for the u-th candidate S u D H T D u. Note that the SRRC filtering with roll-off factor D 0 i aumed for the tranmit filtering, imilar to SC-FDE uing FD-SLM. After that, S u i tranformed back into time domain by N c -point IDFT F H N c. PAPR calculation i applied in order to earch and elect the time-domain tranmit ignal with the lowet PAPR baed on (15), yielding Ou D Œ Ou.0/, :::, Ou.N c 1/ T. The elected time-domain tranmit ignal baed on TD-SLM i expreed by Ou D F H N c H T F Nc d Ou D F H N c H T F Nc P Ou d. (18) In addition, if H T i SRRC filtering with roll-off factor D 0, (18) can be implified a Ou D P Ou d. (19) Finally, the lat N g ample of tranmit block are copied a a CP and inerted into the guard interval, then a CPinerted ignal block of N g C N c ample i tranmitted Receiver. The propagation channel i aumed to be the ame a in Section 2.2, that i, a ymbol-paced L-path frequencyelective block fading channel [2], where it impule repone and channel repone matrix are repreented by Figure 4. Tranceiver ytem model of SC-FDE uing TD-SLM. SC-FDE, ingle-carrier tranmiion with frequency-domain equalization; TD-SLM, time-domain elected mapping; DFT, dicrete Fourier tranform; IDFT, invere DFT; PAPR, peak-to-average power ratio. Wirel. Commun. Mob. Comput. (2016) 2016 John Wiley & Son, Ltd.

7 A. Boonkajay and F. Adachi Selected mapping technique for reducing PAPR of ingle-carrier ignal (8) and (10), repectively. The time-domain received ignal vector after CP removal r Ou D Œr Ou.0/, :::, r Ou.N c 1/ T i expreed by r Ou D 2E T h Ou C n, (20) where Ou D F H N c H T F Nc P Ou d i obtained from (18). The received ignal vector r Ou i tranformed into frequency domain by N c -point DFT, obtaining the frequency-domain received ignal R Ou a R Ou D D 2E F Nc hf H N T c H T D Ou C F Nc n 2E H c H T D T Ou C N, (21) where H c i already defined in (12). MMSE weight matrix W R D diagœw R.0/, :::, W R.N c 1/ i applied at the receiver in order to reduce the effect from frequency electivity. The equalized ignal OD Ou D W R R Ou i obtained, where W R.k/ i determined by Hc W R.k/ D.k/H T.k/ jh c.k/h T.k/j 2. (22) C.E =N 0 / 1 It i oberved that the MMSE-FDE weight in (22) i different from (13) becaue the elected phae rotation pattern P u i not conidered; meanwhile, the de-mapping i employed in time domain intead. The frequency-domain equalized ignal OD Ou i then tranformed back into time domain by N c -point IDFT, yielding time-domain equalized ignal vector before applying demapping Qd Ou D ŒQd Ou.0/, :::, Qd Ou.N c 1/ T a Qd Ou D F H N c OD Ou 2E D F H N T c W R R Ou C F H N c W R N. (23) Finally, de-mapping i applied in order to obtain the original time-domain tranmit block. De-mapping i imply performed by multiplying the time-domain equalized ignal with the Hermitian tranpoe of elected phae rotation pattern matrix, obtaining time-domain received vector Od D ŒOd.0/, :::, Od.N c 1/ T a Od D P H Ou Q d Ou. (24) Similar to FD-SLM, it i oberved from (24) that the receiver need to know which phae rotation pattern i elected at the tranmitter; otherwie, de-mapping cannot be organized accurately and conequently lead to BER degradation. Thi indicate that explicit ide information tranmiion i alo indipenable in TD-SLM. In ummary, the imilarity and difference among conventional SLM technique in OFDM tranmiion [12], the propoed FD-SLM, and TD-SLM for SC-FDE tranmiion can be decribed a follow. The main concept of the conventional SLM technique in OFDM tranmiion and the propoed SLM technique for SC-FDE are imilar; in other word, thoe SLM technique aim to increae the degree-offreedom of PAPR of tranmit ignal by generating the tranmit ignal candidate by multiplying the variou phae rotation pattern to the original ignal. However, candidate generation and the place where the multiplication of phae rotation pattern are carried out for each SLM technique are different. For example, in OFDM tranmiion, phae rotation pattern i multiplied to frequency component where each component contain data-modulated ymbol. In SC-FDE tranmiion uing FD-SLM, phae rotation pattern multiplication i applied to frequency component, but each component contain the DFT-precoded data-modulated ymbol. In contrat, phae rotation pattern multiplication i applied directly to time-domain data-modulated ymbol (not ubcarrier) in SC-FDE uing TD- SLM. We alo would like to mention that the original SLM method for OFDM ignal doe not employ phae rotation pattern multiplication in time domain. In addition, we would like to leave the computational complexity problem a our future work becaue the problem of PAPR reduction i conidered a higher priority than computational complexity. 4. PERFORMANCE EVALUATION Numerical and imulation parameter are ummarized in Table I. We aume 4 QAM, 16 QAM, and 64 QAM SC block tranmiion with the number of available ubcarrier N c D 64. Overampling factor for PAPR evaluation i aumed to be V D 8. Sytem performance of conventional SRRC filtered SC-FDE, SC-FDE uing FD-SLM, and SC-FDE uing TD-SLM are evaluated in term of PAPR of tranmit ignal, average BER (uncoded), throughput, and computational complexity Peak-to-average power ratio The complementary cumulative ditribution function (CCDF) of PAPR wa firt obtained. Then, the PAPR value at CCDF = 0.001, called PAPR 0.1%, i found and dicued. Figure 5 how the PAPR 0.1% performance of SC-FDE uing FD-SLM and SC-FDE uing TD-SLM, a a function of number of candidate (U) and with variou data modulation level. The original SC-FDE mentioned in thi Wirel. Commun. Mob. Comput. (2016) 2016 John Wiley & Son, Ltd.

8 Selected mapping technique for reducing PAPR of ingle-carrier ignal A. Boonkajay and F. Adachi Table I. Simulation parameter. Data modulation 4 QAM, 16 QAM, 64 QAM No. of ubcarrier N c =64 Cyclic prefix length N g =16 Tranmit filtering Tranmitter SRRC ( =0, 0.22) Phae rotation Random equence type No. of candidate U = 1512 PAPR 0.1% threhold 6 db Channel Receiver Fading type Power delay profile Channel etimation Side information haring FDE Frequency-elective block Rayleigh Symbol-paced 16-path uniform Ideal Ideal detection MMSE-FDE QAM, quadrature amplitude modulation; PAPR, peak-to-average power ratio; FDE, frequency-domain equalization; MMSE-FDE, frequency-domain equalization baed on minimum mean-quare error criterion. Figure 5. PAPR veru the number of phae rotation pattern. SC-FDE, ingle-carrier tranmiion with frequency-domain equalization; TD-SLM, time-domain elected mapping; FD-SLM, frequency-domain SLM; PAPR, peak-to-average power ratio; QAM, quadrature amplitude modulation; OFDM, orthogonal frequency diviion multiplexing. paper refer to the DFT-precoded SC-FDE; in other word, the tranmit ignal proceing i carried out by inerting the DFT into the conventional CP-inerted OFDM tranmitter. The SC-FDE ignal produced by inerting DFT a a precoder into the OFDM tranmitter i ued in Long- Term Evolution wirele network [20]. It PAPR value i equal to the cae when U D 1 (equivalent to tranmiion without SLM) in Figure 5. The PAPR 0.1% of conventional SC-FDE i 7.5 db for 4 QAM, 8.4 db for 16 QAM, and 8.6 db for 64 QAM. For comparion and analyi on PAPR of SC-FDE ignal, the PAPR 0.1% of OFDM tranmiion uing FD-SLM auming 16 QAM modulation i plotted. TD-SLM i not available for OFDM ignal becaue time-domain ignal after IDFT i already pule haped and there i no change on PAPR when phae rotation pattern i multiplied. The CCDF of PAPR of OFDM ignal can be approximated by uing the central limit theorem to aume that the real and imaginary part of time-domain ignal follow a Gauian ditribution [12,21], where the CCDF can be expreed by prob.papr > z/ 1.1 e z /ˇN c. Here, N c repreent the number of ubcarrier and ˇ i an approximation parameter for overampling, which i et to be ˇ D 2.3 [22]. The CCDF of PAPR of OFDM uing FD-SLM can be approximated by auming that all the phae-rotated tranmit block are independent to each other and i expreed by prob.papr FD-SLM > z/ Œ1.1 e z /ˇN c U,whereU i the number of phae rotation pattern [22]. In addition, it i hown in [12,13] that the SLM can effectively reduce the PAPR of OFDM ignal with moderate computational complexity and it PAPR-complexity trade-off i much better than partial tranmit equence technique [14]. Therefore, SLM can be conidered a a trong candidate among other PAPR reduction technique for OFDM tranmiion and conequently ha potential to be a promiing PAPR reduction technique for SC-FDE ignal. To the bet knowledge of the author, the PAPR reduction for SC-FDE ha not been extenively examined except the ue of tranmit filtering [6 9]. Therefore, in thi paper, the SRRC filter [6] i ued a a reference in performance comparion. It i oberved from Figure 5 that PAPR 0.1% decreae when U increae in OFDM uing FD-SLM, SC-FDE uing FD-SLM, and TD-SLM. It can be een that more than 3 db reduction of PAPR compared with the original SC-FDE can be achieved when U D 512. Thi i becaue increaing U lead to an increaing of degree-of-freedom, and hence reult in higher probability to obtain a low- PAPR ignal from tranmit candidate. In addition, there are two more obervation obtained from the reult in Figure 5: the firt one i the PAPR 0.1% of OFDM uing FD- SLM and SC-FDE uing FD-SLM become converging to the ame value when U i large, and the econd one i TD- SLM outperform FD-SLM by achieving lower PAPR in every modulation level and number of phae rotation pattern. The reaon for upporting the aforementioned two obervation can be analyzed by referring the following computer imulation reult. Figure 6 how the CDF of PAPR reduction effect defined by the difference between the PAPR of the tranmit ignal, which i phae rotated by the elected phae rotation pattern (i.e., SLM output ignal) and that of the original tranmit ignal, that i, PAPR. Ou / PAPR./, where Ou i defined in (6) and (18). The number of phae rotation patterniaumedtobeu D 64 and auming 16 QAM modulation. It i obviouly een that the OFDM uing FD-SLM achieve the bet PAPR reduction performance, Wirel. Commun. Mob. Comput. (2016) 2016 John Wiley & Son, Ltd.

9 A. Boonkajay and F. Adachi Selected mapping technique for reducing PAPR of ingle-carrier ignal Figure 6. CDF of PAPR reduction. SC-FDE, ingle-carrier tranmiion with frequency-domain equalization; TD-SLM, timedomain elected mapping; FD-SLM, frequency-domain SLM; PAPR, peak-to-average power ratio; CDF, cumulative ditribution function; OFDM, orthogonal frequency diviion multiplexing; SRRC, quare-root raied-coine. following by the SC-FDE uing TD-SLM and SC-FDE uing FD-SLM, repectively. The reult in Figure 6 i conitent with Figure 5 a the OFDM uing FD-SLM can ignificantly reduce the PAPR a up to 4.3 db of PAPR reduction i achievable, while up to 2.0 and 3.1 db of PAPR reduction are achievable in SC-FDE uing FD-SLM and TD-SLM, repectively. The reaon i becaue the original OFDM ignal ha high PAPR, reulting in higher probability to obtain a phae rotation pattern among U pattern, which can reduce the PAPR. On the other hand, there i probability that no phae rotation pattern can reduce the PAPR among U pattern in SC-FDE becaue the original SC-FDE ignal ha low PAPR. The phenomenon that there i no phae rotation pattern among U pattern, can reduce the PAPR i obviouly een in SC-FDE uing FD-SLM a the probability that the reduction efficiency i 0 db i approximately 30%. Figure 7 how the complex envelope of tranmit ignal Ou.n/ D Ou,I.n/ C j Ou,Q.n/, where Ou.n/ i the n-th element in Ou defined in (6) and (18), in I-Q plane auming U D 64, N c D 16, and 4 QAM modulation. The tranmit complex envelope of conventional SC-FDE, SC-FDE uing FD-SLM, and SC-FDE uing TD-SLM are plotted from the ame et of data-modulated ymbol. It i een from both Figure 7(a) and (b) that the complex envelope of SC-FDE uing FD-SLM and TD-SLM travel in narrower range compared with that of the conventional SC-FDE, reulting in lower peak amplitude and lower PAPR. In addition, it i een that the complex envelope of SC-FDE uing TD-SLM alway pa through the original tranmit ignal contellation (here i 4 QAM contellation), where ome part of SC-FDE uing FD-SLM doe not pa through the original contellation. Thi i becaue applying phae rotation to SC-FDE ignal in time domain alway reult in the phae-rotated ignal which maintain original modulation cheme, for example, applying the phae rotation to 4 QAM data-modulated tranmit block alway Figure 7. Complex envelope of the SC-FDE uing SLM. SC- FDE, ingle-carrier tranmiion with frequency-domain equalization; TD-SLM, time-domain elected mapping; FD-SLM, frequency-domain SLM; PAPR, peak-to-average power ratio; SRRC, quare-root raied-coine. obtain 4 QAM data-modulated ymbol. By maintaining the original modulation level, the amplitude of reulting phae-rotated ignal i bounded, and then, candidate generation via multiplication of phae rotation pattern can reach the nearly optimal olution, while the amplitude of Wirel. Commun. Mob. Comput. (2016) 2016 John Wiley & Son, Ltd.

10 Selected mapping technique for reducing PAPR of ingle-carrier ignal A. Boonkajay and F. Adachi Table II. Optimal U for SLM algorithm. 4QAM 16QAM 64QAM PAPR 0.1% 6dB FD-SLM U D128 (6.0 db) U D 256 (5.86 db) U D 256 (5.87 db) TD-SLM U D 4 (5.97 db) U D 16 (5.88 db) U D 16 (5.99 db) PAPR 0.1% 4QAM(7.55dB) FD-SLM U D 4(7.54dB) U D 8(7.19dB) TD-SLM U D 2(7.53dB) U D 4 (6.97 db) PAPR 0.1% SRRC filtered SC-FDE w/ =0.22 FD-SLM U 512 U D 64 (6.25 db) U D 16 (6.83 db) (4.3 db for 4 QAM, 6.44 db for 16 QAM, TD-SLM U D 256 (4.29 db) U D 8 (6.27 db) U D 4 (6.95 db) 6.96 db for 64 QAM) SLM, elected mapping; TD-SLM, time-domain SLM; FD-SLM, frequency-domain SLM; QAM, quadrature amplitude modulation; PAPR, peak-toaverage power ratio; SC-FDE, ingle-carrier tranmiion with frequency-domain equalization; SRRC, quare-root raied-coine. phae-rotated ignal in SC-FDE uing FD-SLM become cloe to Gauian random variable (imilar to OFDM uing FD-SLM). Regarding to the aforementioned dicuion, SC-FDE uing TD-SLM ha potential to achieve lower PAPR than FD-SLM. In general, an optimal U for achieving a particular PAPR 0.1% value need to be dicued to avoid inufficient computational complexity and ide information bit. In thi paper, the optimal U for achieving a particular PAPR 0.1% value i dicued in three approache: the optimal U for achieving PAPR 0.1% 6 db, the optimal U for keeping the PAPR 0.1% of multi-level modulated SC-FDE to be le than or equal to that of 4 QAM-modulated SC-FDE, and the optimal U for keeping the PAPR 0.1% to be le than or equal to that of SRRC filtered SC-FDE with D The optimal U for thoe three apect are hown in Table II, where the TD-SLM give lower U than FD-SLM in every comparion. Figure 8 how the PAPR 0.1% performance comparion between SLM uing real-valued phae rotation pattern (i.e., e j0 and e j ) and SLM uing complex-valued phae rotation. We aume 16 QAM modulation, while complexvalued phae rotation with phae rotation interval of 90 ı (i.e., e j where D 0, =2, ) and 45 ı (i.e., e j where D 0, =4, =2, ) are ued for comparion. It i obviouly een in Figure 8 that there i no difference of PAPR 0.1% performance even though the phae rotation pattern i real valued or complex valued in both FD- SLM and TD-SLM. Thi reult i alo conitent with [18] and auming OFDM tranmiion that the SLM provide the bet performance when the phae rotation i dicrete and uniformly ditributed over Œ0, 2/. Regarding to thi imulation reult, the ue of real-valued phae rotation i attractive becaue it can reduce the unneceary complexvalued multiplication operation becaue of candidate generation without degrading PAPR performance BER performance To confirm that the propoed SLM algorithm for SC- FDE provide BER preervation property, BER perform a function of average received bit energy-to-noie power pectrum denity ratio E b =N 0 D.1=N mod /.E =N 0 /.1 C N g =N c /,wheren mod repreent modulation level (two for 4 QAM, four for 16 QAM, and ix for 64 QAM), of SC- FDE uing FD-SLM and TD-SLM at U D 512 are hown in Figure 9 and compared with conventional SC-FDE. Figure 8. PAPR performance of complex-valued phae rotation pattern. TD-SLM, time-domain elected mapping; FD-SLM, frequency-domain SLM; PAPR, peak-to-average power ratio; SRRC, quare-root raied-coine; QAM, quadrature amplitude modulation. Figure 9. BER performance. SC-FDE, ingle-carrier tranmiion with frequency-domain equalization; TD-SLM, time-domain elected mapping; FD-SLM, frequency-domain SLM; BER, biterror rate; QAM, quadrature amplitude modulation. Wirel. Commun. Mob. Comput. (2016) 2016 John Wiley & Son, Ltd.

11 A. Boonkajay and F. Adachi Selected mapping technique for reducing PAPR of ingle-carrier ignal Ideal ide information detection i aumed in thi paper. It i obviouly een from Figure 9 that the BER performance of conventional SC-FDE, SC-FDE uing FD-SLM, and SC-FDE uing TD-SLM are the ame, indicating that the SLM algorithm can reduce the PAPR without degrading the BER. To achieve the BER preervation property, however, up to log 2 U-bit of ide information need to be tranmitted in order to achieve an accurate de-mapping at the receiver. The ide information hould be coded by uing forward error-correction coding. However, thi further degrade the throughput [23]. The impact of ide information on throughput performance will be dicued in the next ubection Throughput performance In thi ubection, the throughput performance of SC- FDE uing SLM algorithm are initially evaluated by average throughput performance a a function of peak tranmit E =N 0, where the throughput performance in bp/hz i defined a follow [24]; D N mod.1 PER/ 1 1 C 1 1 C N, (25) gcn SI N c where PER i packet-error rate and N SI i the number of required ide information ymbol, which i aumed to be N SI D d.log 2 U/=N mod e. Uncoded ide information tranmiion and ideal ide information detection are aumed for implicity. The packet length i aumed to be 3072 bit in thi paper. The peak tranmit E =N 0 i conidered becaue it refer to the required peak tranmit power of a power amplifier, while the peak tranmit E =N 0 i defined a a ummation of average received E =N 0 and PAPR 0.1% [5]. Figure 10 how the throughput performance a a function of peak tranmit E =N 0 of the conventional SC-FDE, SRRC filtered SC-FDE with D 0.22, and SC-FDE uing the propoed SLM. The number of candidate for both TD- SLM and FD-SLM are et a the minimum U for keeping the PAPR 0.1% to be le than or equal to that of SRRC filtered SC-FDE with D 0.22 (Table II), except in cae of 4 QAM-modulated SC-FDE uing FD-SLM where U i et to be 512 (the maximum value conidered in thi paper). In Figure 10(a), it can be oberved from the figure that SRRC filtered SC-FDE with D 0.22 achieve better throughput performance becaue of it low-papr property. However, it peak throughput degrade by a factor of 1/(1+ ). SC-FDE uing FD-SLM can provide imilar throughput performance at low peak E =N 0 region compared with SRRC filtered SC-FDE with D 0.22 (except in 4 QAM cae), where there i a light peak throughput degradation a a reult from ide information. In Figure 10(b), it i een that SC-FDE uing TD-SLM provide imilar throughput performance to SRRC filtered SC-FDE with D 0.22 at low peak E =N 0 region in every modulation cheme. TD-SLM require le number of Figure 10. Throughput performance. SC-FDE, ingle-carrier tranmiion with frequency-domain equalization; TD-SLM, time-domain elected mapping; FD-SLM, frequency-domain SLM; SRRC, quare-root raied-coine; QAM, quadrature amplitude modulation. candidate (U) than FD-SLM for achieving the ame PAPR performance, which i equivalent to le ide information bit. Thi contribute to improving the peak throughput, even though the reulting peak throughput i till lightly le than conventional SC-FDE. Wirel. Commun. Mob. Comput. (2016) 2016 John Wiley & Son, Ltd.

12 Selected mapping technique for reducing PAPR of ingle-carrier ignal A. Boonkajay and F. Adachi Table III. Computational complexity performance. Conv. SC-FDE SC-FDE with FD-SLM SC-FDE with TD-SLM Phae rotation DFT.N c / 2.N c / 2 U.N c / 2 IFFT N c log 2.N c / UVN c log 2.VN c / UVN c log 2.VN c / PAPR calculation - UVN c UVN c SC-FDE, ingle-carrier tranmiion with frequency-domain equalization; TD-SLM, time-domain elected mapping; FD-SLM, frequency-domain SLM; PAPR, peak-to-average power ratio; DFT, dicrete Fourier tranform; IFFT, invere fat Fourier tranform. TD-SLM, auming 16-QAM modulation. TD-SLM outperform FD-SLM in term of computational complexity when the required PAPR i et to be the ame becaue TD-SLM require le number of candidate. For example, at the point that PAPR 0.1% =6 db, TD-SLM require only 10% of total computational complexity of FD-SLM. 5. CONCLUSION Figure 11. Computational complexity veru PAPR performance. SC-FDE, ingle-carrier tranmiion with frequencydomain equalization; TD-SLM, time-domain elected mapping; FD-SLM, frequency-domain SLM; PAPR, peak-to-average power ratio; QAM, quadrature amplitude modulation Computational complexity The computational complexity performance i evaluated by counting the number of complex-valued multiplication [25]. Computational complexity a a function of the number of ubcarrier (N c ) and candidate (U) for conventional SC-FDE, SC-FDE uing FD-SLM, and SC-FDE uing TD- SLM i ummarized in Table III. Note that IDFT can be replaced by invere fat Fourier tranform (IFFT) if the number of ubcarrier i a power of two. It i een from Table III, and indicated in [12,13], that SLM require high computational complexity compared with the conventional SC-FDE becaue of the requirement of additional IFFT operation. SC-FDE uing TD-SLM ha more complexity at the ame number of candidate compared with FD-SLM becaue it alo require additional DFT operation (typically U time of DFT operation compared with FD-SLM). In addition, the ue of real-valued phae rotation pattern can reduce unneceary computational complexity for generating the tranmit waveform candidate. However, the PAPR in Figure 5 how that TD-SLM achieve imilar PAPR to FD-SLM with le number of candidate. Figure 11 how the PAPR 0.1% againt complexity a a function of U for SC-FDE uing FD-SLM and In thi paper, two pectrum-efficient PAPR reduction technique for SC ignal called FD-SLM and TD-SLM were propoed. The propoed technique are baed on generating tranmit block candidate through phae rotation in different domain. The phae rotation i applied to ubcarrier either between DFT and IFFT for FD-SLM, or directly to the time-domain tranmit ignal vector prior to DFT for TD-SLM. Simulation reult confirmed that both FD-SLM and TD-SLM can efficiently reduce the PAPR of SC-FDE tranmiion without ignificant degradation on BER and throughput compared with SRRC filtered SC-FDE tranmiion. It wa alo clarified in thi paper that the TD-SLM achieve better PAPR reduction performance compared with FD-SLM, at which the TD-SLM require le number of candidate and le computational complexity than FD-SLM for achieving the ame target PAPR. ACKNOWLEDGEMENTS Thi paper include a part of reult of The reearch and development project for realization of the fifthgeneration mobile communication ytem (# , April 2016) commiioned to Tohoku Univerity and The Minitry of Internal Affair and Communication (MIC), Japan. REFERENCES 1. Adachi F, Takeda K, Yamamoto T, Matukawa R, Kumagai S. Recent advance in ingle-carrier ditributed antenna network. Wiley Wirele Communication and Moblie Comput 2011; 11: Goldmith A. Wirele Communication. Cambridge Univerity Pre: New York, Wirel. Commun. Mob. Comput. (2016) 2016 John Wiley & Son, Ltd.

13 A. Boonkajay and F. Adachi Selected mapping technique for reducing PAPR of ingle-carrier ignal 3. Han SH, Lee JH. An overview of peak-to-average power ratio reduction technique for multicarrier tranmiion. IEEE Tranaction on Wirele Communication 2005; 12(2): Falconer D, Ariyaviitakul SL, Benyamin-Seeyar A, Edion B. Frequency domain equalization for inglecarrier broadband wirele ytem. IEEE Communication Magazine 2002; 40(4): Okuyama S, Takeda K, Adachi F. MMSE frequencydomain equalization uing pectrum combining for nyquit filtered broadband ingle-carrier tranmiion. In Proceeding IEEE Vehicular Technology Conference (VTC 2010-Spring), Taipei, Taiwan, May Akaiwa Y. Introduction to Digital Mobile Communication (1t edn). Wiley: New Jerey, Rha P, Hu S. Peak-to-average ratio (PAR) reduction by pule haping uing a new family of generalized raied coine filter. In Proceeding IEEE Vehicular Technology Conference (VTC 2003-Fall), Florida, USA, Oct. 2003; Meza C, Lee K, Lee K. PAPR Reduction in ingle carrier FDMA uplink ytem uing parametric linear pule. In Proceeding International Conference on ICT Convergence (ICTC 2011), Seoul, Korea, Sept. 2011; Boonkajay A, Obara T, Yamamoto T, Adachi F. Excebandwidth tranmit filtering baed on minimization of variance of intantaneou tranmit power for low- PAPR SC-FDE. IEICE Tranaction on Communication 2015; E98-B(04): Li X, Cimini LJ. Effect of clipping and filtering on the performance of OFDM. IEEE Communication Letter 1998; 2(5): Caron N, Gulliver TA. PAPR reduction of OFDM uing elected mapping, modified RA code and clipping. In Proceeding IEEE Vehicular Technology Conference (VTC 2002-Fall), Vancouver, Canada, Sept. 2002; Bauml RW, Ficher RFH, Huber JB. Reducing the peak-to-average power ratio of multicarrier modulation by elected mapping. IEEE Electronic Letter 1996; 32(22): Gacanin H, Adachi F. Selective mapping with ymbol re-mapping for OFDM/TDM uing MMSE-FDE. In Proceeding IEEE Vehicular Technology Conference (VTC 2008-Fall), Calgary, Canada, Sept. 2008; Baxley RJ, Zhou GT. Comparion of elected mapping and partial tranmit equence for cret factor reduction in OFDM. In Proceeding 2006 IEEE Military Communication Conference (MILCOM 2006), Wahington D.C., USA, October 2006; Slimane S. Reducing the peak-to-average power ratio of OFDM ignal through precoding. IEEE Tranaction on Vehicular Technology 2007; 56(2): Wulich D, Goldfield L. Bound of the ditribution of intantaneou power in ingle carrier modulation. IEEE Tranaction on Wirele Communication 2005; 4(4): Ohkubo N, Ohtuki T. Deign criteria for phae equence in elected mapping. IEICE Tranaction Communication 2003; E86-B(9): Zhou GT, Peng L. Optimality condition for elected mapping in OFDM. IEEE Tranaction Signal Proceing 2006; 54(8): Takei Y, Ohtuki T. Uplink pre-equalization uing mme prediction for TDD/MC-CDMA ytem. In Proceeding IEEE International Sympoium on Peronal Indoor and Mobile Radio Communication (PIMRC 2005), Berlin, Gremany, Sept. 2013; Kawamura T, Kihiyama Y, Sawahahi M. Performance of tar 16QAM cheme conidering cubic metric for uplink DFT-Precoded OFDMA. IEICE Tranaction on Communication 2014; E97-A: Praad R. OFDM for Wirele Communication Sytem. Artech Houe: Maachuett, Haan ES, El-Khamy SE, Deouky MI -, El-Dolil SA, Abd El-Samie FE. Peak-to-average power ratio reduction in pace-time block coded multi-input multioutput orthogonal frequency diviion multiplexing ytem uing a mall overhead elective mapping cheme. IET Communication 2008; 3(10): Eom SS, Nam HW, Ko YC. Low-complexity PAPR reduction cheme without ide information for OFDM ytem. IEEE Tranaction Signal Proceing 2012; 60(7): Fukuda K, Nakajima A, Adachi F. LDPC-coded HARQ throughput performance of MC-CDMA uing ICI cancellation. In Proceeding IEEE Vehicular Technology Conference (VTC 2007-Fall), Baltimore, USA, Sept. 2007; Tenma K, Yamamoto T, Lee K, Adachi F. 2-tep QRM-MLBD for broadband ingle-carrier tranmiion. IEICE Tranaction Communication 2012; E95-B (4): Wirel. Commun. Mob. Comput. (2016) 2016 John Wiley & Son, Ltd.

14 Selected mapping technique for reducing PAPR of ingle-carrier ignal A. Boonkajay and F. Adachi AUTHORS BIOGRAPHIES Amnart Boonkajay received hi BE degree (1t-cla honor) in Telecommunication Engineering from Sirindhorn International Intitute of Technology (SIIT), Thammaat Univerity, Thailand in 2010, and ME and PhD degree in Communication Engineering from Tohoku Univerity, Japan in 2013 and 2016, repectively. He i currently working a a government-indutry-academia collaboration reearcher at Reearch Organization of Electrical Communication, Tohoku Univerity, where he i doing a reearch and development for implementing the fifth-generation mobile communication ytem. Hi reearch interet are baed on wirele and mobile communication phyical layer including tranmit and receive ignal proceing, filtering and equalization, coding and modulation. He wa the recipient of the Japanee Government (MEXT) Scholarhip from 2010 to 2016 and the Tohoku Univerity, Department of Electrical and Information Engineering Excellent Student Award in He i alo an active reviewer of technical journal and conference paper and received the IEEE Wirele Communication Letter Exemplary Reviewer in Fumiyuki Adachi received hi BS and Dr Engineering degree in Electrical Engineering from Tohoku Univerity, Sendai, Japan, in 1973 and 1984, repectively. In April 1973, he joined the Electrical Communication Laboratorie of Nippon Telegraph & Telephone Corporation (now NTT) and conducted variou type of reearch related to digitalcellular mobile communication. From July 1992 to December 1999, he wa with NTT Mobile Communication Network, Inc. (now NTT DoCoMo, Inc.), where he led a reearch group on wideband/broadband CDMA wirele acce for IMT-2000 and beyond. He contributed to the development of 3G air interface tandard, known a W-CDMA. Since January 2000, he ha been with Tohoku Univerity, Sendai, Japan. He wa a full Profeor until March 2016 and i now a pecially appointed Profeor for reearch. Hi reearch interet i in the area of wirele ignal proceing and networking including broadband wirele acce, equalization, tranmit/receive antenna diverity, MIMO, adaptive tranmiion, and channel coding. He wa a program leader of the 5-year Global COE Program Center of Education and Reearch for Information Electronic Sytem ( ), awarded by the Minitry of Education, Culture, Sport, Science and Technology of Japan. From October 1984 to September 1985, he wa a United Kingdom SERC Viiting Reearch Fellow in the Department of Electrical Engineering and Electronic at Liverpool Univerity. He i an IEICE Fellow and i a Co-recipient of the IEICE Tranaction bet paper of the year award 1996, 1998, and 2009 and alo a recipient of Achievement award He i an IEEE Fellow and i a VTS Ditinguihed Lecturer ince He i a Corecipient of the IEEE Vehicular Technology Tranaction bet paper of the year award 1980 and again 1990 and alo a Recipient of IEEE VTS Avant Garde award He i a Recipient of Thomon Scientific Reearch Front Award 2004, Ericon Telecommunication Award 2008, Telecom Sytem Technology Award 2009, Prime Miniter Invention Prize 2010, Britih Royal Academy of Engineering Ditinguihed Viiting Fellowhip 2011, KDDI Foundation Reearch Award 2012, IEEE VTS Conference Chair Award 2014, C&C Prize 2014, and Rinzaburo Shida Award He i lited in Highly Cited Reearcher 2001 ( highlycited.com/archive/). Wirel. Commun. Mob. Comput. (2016) 2016 John Wiley & Son, Ltd.