University of Bristol - Explore Bristol Research. Peer reviewed version. Link to published version (if available): /WCNC.2009.

Transcription

1 Noune, MB., & Nix, AR. 2009). Frequency-domain tranmit proceing for MIMO SC-FDMA in wideband propagation channel. In IEEE Wirele Communication and Networing Conference WCNC2009), Budapet, Hungary pp. 1-6). Intitute of Electrical and Electronic Engineer IEEE). DOI: /WCNC Peer reviewed verion Lin to publihed verion if available): /WCNC Lin to publication record in Explore Britol Reearch PDF-document Univerity of Britol - Explore Britol Reearch General right hi document i made available in accordance with publiher policie. Pleae cite only the publihed verion uing the reference above. Full term of ue are available:

2 hi full text paper wa peer reviewed at the direction of IEEE Communication Society ubject matter expert for publication in the WCNC 2009 proceeding. Frequency-Domain ranmit Proceing for MIMO SC-FDMA in Wideband Propagation Channel Mohamed Noune and Andrew Nix Centre for Communication Reearch, Univerity of Britol Merchant Venture Building, Woodland Road, BS8 1J, Britol UK Abtract Recently there ha been coniderable interet in the ue of Single Carrier Frequency Diviion Multiple Acce SC-FDMA) a the uplin tranmiion cheme for the 3GPP Long erm Evolution LE) tandard. hi paper invetigate different Multiple Input Multiple-Output MIMO) technique for uplin SC-FDMA. Here we demontrate how the tranmitter can exploit the available channel information, through precoding and pre-equalization, in order to combat the frequency elective nature of the propagation channel and improve the ytem performance under fading. In thi wor we invetigate the achievable ytem capacity, and the Pea-to-Average Power Ratio PAPR) characteritic of the precoded and pre-equalized SC- FDMA waveform. We conider unitary precoding for MIMO, a well a pre-equalization for a Multiple-Input Single-Output MISO) ytem with Maximum Ratio Combining MRC), Equal Gain Combining EGC), Zero Forcing ZF), and Minimum Mean Square Error MMSE) algorithm. Reult reported in thi paper how that ZF pre-equalization for SISO SC-FDMA i capable of achieving the Shannon Capacity limit for faded channel, with a BER performance that i identical to an AWGN channel, and that pre-equalized increae the capacity of MIMO SC-FDMA ytem. Keyword: 3GPP LE, MIMO, SC-FDMA, Frequency-Domain Pre-equalization, PAPR, Equalization, Beamforming. I. INRODUCION HE ignificant expanion een in mobile and cellular technology over the lat two decade i a direct reult of the increaing demand for high data rate tranmiion over bandwidth and power limited wirele channel. hi requirement for high data rate reult in ignificant interymbol interference ISI) for ingle carrier ytem, and thereby require the ue of robut coding and powerful ignal proceing technique in order to overcome the time and frequency elective nature of the propagation channel. Recent example include different tranmitter/receiver optimization technique for high data rate tranmiion a well a the ue of multiple antenna at both the tranmitter and receiver, nown a Multiple-Input Multiple-Output MIMO). MIMO tae advantage of the patial eparation between antenna element to create uncorrelated patial channel, and to exploit higher level of patial diverity. hi tranlate to improved pectral and power efficiency. hee technique are particularly attractive at the bae tation, where large antenna pacing are more eaily accommodated. he hird Generation Partnerhip Project 3GPP) Long erm Evolution LE) radio acce tandard i baed on hared channel acce providing pea data rate of 75 Mbp in the uplin and 300 Mbp in the downlin and upport different carrier bandwidth MHz) in both Frequency Diviion Duplex FDD) and ime Diviion Duplex DD) mode 1]. A woring aumption in the LE tandard i the ue of Orthogonal Frequency Diviion Multiple Acce OFDMA). OFDMA i an OFDM-baed multiple acce cheme 2] that provide each uer with a unique fraction of the ytem bandwidth. OFDMA i highly uitable for broadband wirele acce networ particularly the downlin) ince it combine calability, multipath robutne and MIMO compatibility 2]. OFDMA i very enitive to frequency offet and phae noie, and thu require accurate frequency and phae ynchronization. In addition, OFDMA i characterized by a high PAPR, and for a given pea-power limited amplifier thi reult in a lower mean tranmit level at the handet. For thee reaon, OFDMA i not well-uited to the uplin tranmiion. SC- FDMA, alo nown a DF precoded OFDMA 1], ha been propoed for uplin cellular application. Although the 3GPP tandard only adopt MIMO tranmiion for downlin, MIMO can be ued in conjunction with SC- FDMA for uplin tranmiion. hi will improve ytem capacity and overcome the power limitation of thee ytem. MIMO can be divided into three main type or categorie. Firtly, in precoding, alo nown a beamforming, the ame ignal i emitted from each of the tranmit antenna with appropriate phae and gain weighting uch that the ignal power i maximized at the receiver input. hi increae the ignal gain from contructive combining 5]- 7]. In 11], a tudy of the PAPR characteritic of beamforming with unitary precoding for SC-FDMA ha hown that the increae in PAPR of the precoded waveform i not ignificant when compared to the ingle antenna ytem or the non-precoded MIMO SC- FDMA cae. In addition, amplitude clipping can be ued in order to maintain the PAPR below a certain level at the expene of a performance degradation. Secondly, Spatial Multiplexing SM), 6]- 8], aim to increae the capacity of the ytem for the ame bandwidth without any increae in the tranmit power. Finally, ranmit Diverity D) aim to increae the diverity order of the ytem. In the context of SC-FDMA, 12] how the poibility of employing D for SC-FDMA in the form of Space-ime/Frequency Bloc Coding SBC/SFBC), or the more novel Single Carrier Space-Frequency Bloc Coding SC-SFBC). In 4], a tudy of the PAPR characteritic together with the error performance of MIMO SC-FDMA wa preented for both SM and D. A wa hown, both the PAPR and error performance are highly influenced by the choice of the ub-carrier mapping and vary from one MIMO cheme to the other. It wa alo hown that pule haping can be ued to reduce the high PAPR of MIMO SC-FDMA, reulting in a performance degradation for both SM and D. he goal of thi article i to demontrate the lin performance of MIMO SC-FDMA when patial equalization i ued at the tranmitter. Here, we preent a tudy of the Pea-to-Average Power Ratio PAPR) characteritic and the achievable ytem /09/$ IEEE Authorized licened ue limited to: UNIVERSIY OF BRISOL. Downloaded on October 13, 2009 at 10:09 from IEEE Xplore. Retriction apply.

3 hi full text paper wa peer reviewed at the direction of IEEE Communication Society ubject matter expert for publication in the WCNC 2009 proceeding. i-th uer input data DF DF Spatial Multiplexing/ SFBC Mapping Mapping IDF IDF P/S P/S Inert Inert Remove Remove P/S P/S DF DF Spatial Demultiplexing & Equalization Demapping Demapping IDF IDF i-th uer utput data a) ranmitter Structure Fig. 1. ranceiver tructure for MIMO SC-FDMA. b) Receiver Structure capacity of the precoded MIMO and pre-equalized MISO SC- FDMA waveform. We conider unitary precoding, a well a pre-equalization uing a range of diverity combining cheme, namely Maximum Ratio Combining MRC), Equal Gain Combining EGC), Zero Forcing ZF), and the Minimum Mean Square Error MMSE) algorithm. hi paper i organized a follow. In ection II the MIMO SC-FDMA tranmiion model i given. Section III and IV provide an overview of the MIMO precoding and ranmit Diverity pre-equalization technique for MIMO SC-FDMA, repectively. In ection V we preent the channel capacity for MIMO SC-FDMA with precoding and pre-equalization. Simulation reult and concluion are preented in ection VI and VII repectively. II. MIMO SC-FDMA RANSMISSION 3GPP LE 1]- 3] i a driving force in the improvement of mobile communication tandard 3]. A woring aumption for 3GPP LE i the ue of SC-FDMA on the uplin. he principle of SC-FDMA ignaling i preented in 9]- 10]. For high data rate wirele communication, multiuer tranmiion can be achieved through OFDMA and/or SC- FDMA. Since OFDMA waveform are characterized by their high dynamic range a a reult of the IDF pread, which tranlate to a high PAPR, SC-FDMA i favored for the uplin tranmiion. Fig. 1 how the tructure of the N -by-n R MIMO SC-FDMA ytem conidered in thi paper. For the i-th uer, for the j-th tranmit antenna, and for each bloc of M data ample x i) j = x i) j,0,xi) j,1 j,m 1],,xi), the tranmitter map the correponding M frequency component of the bloc, X i) j = X i) j,0, Xi) j,1,, Xi) j,m 1], reulting from an M point DF of the data ample, onto a et of M active ub-carrier elected from a total of N = QM ubcarrier Q >1). In thi paper we conider ditributed and localized SC-FDMA D-FDMA and L-FDMA repectively) 9]. We denote the ubcarrier mapping tranform matrix for the i-th uer by D i.he entrie of thi matrix for both D-FDMA and L-FDMA are given in Equation 1) and 2) repectively: ) 0i 1)M M ; u M D i = i ; 0Q M ; ) u M i+q ; ; u i+mq ; 0 ]1) Q Q )M M D i = ] 0 i 1)M M ; I M ; 0 Q i)m M 2) he ub-carrier mapping produce X i) j = D i X i) j uch that ] X i) j = Xi) j,0, Xi) j,1,, Xi) j,n 1. X i) j i proceed by the Invere DF IDF) to produce the time-domain tranmitted ignal x i) j = x i) j,0, xi) j,1,, xi) j,n 1]. he SC-FDMA tranmitted ignal can be repreented by: x i) j = PF 1 N D if M x i) j 3) where F 1 N and F M are the IDF and DF matrix repectively. he generic K-point DF matrix ha pq j2π entrie F K ] p,q = e K, and it invere i F 1 K = FH K, where ) H denote the Hermitian tranpoe. P repreent the cyclic prefix ) inertion matrix: P =C, I N ], C = 0 P N P ), I P ] he received SC-FDMA ignal on the m-th receive antenna, r i) m,n at time n, for a multipath fading MIMO channel corrupted by Additive White Gauian Noie, w m,n, i) with variance N o i given by: N r m,n i) = L j=1 l=0 h j,m x i) j,n l + wi) m,n 4) We denote R i), X i) and W i) be the -th ub-carrier of the received ignal, the tranmitted ignal and the noie ignal at the receiver front-end, repectively, where: R i) = ] R i),1, Ri),2,, Ri),N, X i) ] = X i),1, X i),2,, X i),n ] and W i) = W i),1, Wi),2,, Wi),N. By taing the DF of both ide in equation 4), the -th ub-carrier of i-th uer received data i given by: R i) = H i) X + W i) 5) where H denote the -th ub-carrier of the MIMO channel frequency repone given by H 1,1 H 1,2 H 1,N H H = 2,1 H 2,2 H 2,N H N R,1 H N R,2 H N R,N III. UNIARY PRECODING FOR MIMO SC-FDMA When the channel information i available at the tranmitter, via feedbac from the receiver or through the reciprocity principle in a duplex ytem, it i poible to increae the channel capacity by exploiting the patial ub-channel 6]. hi i achieved by ditributing the tranmit energy acro the ub-patial channel for the frequency flat channel, and acro pace and frequency for the frequency elective MIMO channel, in order to increae the pectral efficiency. Optimal power allocation i achieved through the waterpouring algorithm, a decribed in 6]. ranmit eigen-beamforming with unitary precoding exploit the eigen-tructure of the channel in order to achieve patial multiplexing 16]. From the Singular Value Decompoition SVD) of the MIMO channel matrix H, H = U Λ V H, it i poible to acce the patial mode of the MIMO channel matrix through the Authorized licened ue limited to: UNIVERSIY OF BRISOL. Downloaded on October 13, 2009 at 10:09 from IEEE Xplore. Retriction apply.

4 hi full text paper wa peer reviewed at the direction of IEEE Communication Society ubject matter expert for publication in the WCNC 2009 proceeding. multiplication of the tranmitted ignal X i) by V prior to tranmiion and the multiplication of the received ignal R i), by U H at the receiver. he input-output relationhip of thi operation i given by: R i) = U H H V X i) + W i) ) = Λ X i) + W i) 6) hi how that with channel nowledge at the tranmitter, H can be decompoed into r parallel SISO channel, where r i the ran of H, atifying: R i) j, = λ j, X i) j, + W i) j,, j =1, 2,,r 7) Given a MIMO ytem with an average power contraint, the concept of water filling aim to exploit the patial decompoition of the MIMO channel given channel nowledge at the tranmitter, by allocating optimal energy onto each patial of the mode, which will increae the capacity of each patial mode. Uing the optimal power allocation cheme decribed in 14], the optimum power, γ j,, allocated to the patial mode j, i given by: { μ 1 λ γ j, = j, : μ 1 λ j, 8) 0 : otherwie where μ i the cutoff SNR level below which no eigen-mode tranmiion i performed. μ i given by: ] μ = N 1+ N r o 1 9) E λ j, IV. MISO SC-FDMA WIH PRE-EQUALIZAION he main focu of our invetigation in thi ection i contrained pre-equalization for Multiple-Input Single-Output MISO) SC-FDMA, where the mean tranmit power i contrained to be the ame a the tandard tranmiion cae. We tart by deriving the pre-equalizer weight without any power contraint and then introduce the A. Uncontrained Pre-equalization We define G i) j, the pre-equalization coefficient without power contraint. he different deign criteria conidered in thi paper are: 1) Maximum Ratio Combining: MRC pre-equalization aim to maximize the intantaneou SNR at the receiver front end. Under the MRC criterion, the pre-equalization weight are: G i) j, = ) H j, 10) 2) Equal Gain Combining: In order for all the ub-carrier to arrive at the receiver in phae, phae equalization can be performed at the tranmitter in the form of EGC. Under the EGC criterion, the pre-equalization weight are: G i) j, = i=1 Hj, ) Hj, 11) 3) Zero-Forcing: he pre-equalizer contructed under the ZF criterion repreent the invere of the channel frequency repone. For the i-th uer and the j-th tranmit antenna, the ZF pre-equalization weight are therefore: G i) j, = ) Hj, Hj, 2 12) 4) Minimum Mean Square Error: he pre-equalization weight under the MMSE frequency-domain pre-equalizer for the j-th tranmit antenna atify: e 2 = G i) j,)h j, ) Ψ i he pre-equalizer weight are therefore given by: ) G i) Hj, j, = H j, 2, 14) + λ λ mut be found by numerical computation a it depend on the channel delay profile. If N o, the variance of the additive noie at the receiver front-end, i nown at the tranmitter, then λ in equation 15) equal N o. B. Contrained Pre-equalization Hj, ) G i) j, = H j, 2 15) + No For the j-th tranmit antenna, the -th ub-carrier weight of the power contrained frequency-domain pre-equalization filter, G i) j,,forthei-th uer atify: G i) i) 2 X = X i) 2 16) Ψ i j, j, Ψ i Let G i) j, be the -th ub-carrier pre-equalization coefficient without power contraint. We define P j the normalization factor for the j-th tranmit antenna, uch that G i) j, = Gi) j,p j. From equation 16): 1 G i) j, 2 = 1 G i) 2 M M j,p j =1 17) Ψ i Ψ i j, It follow from equation 17) P j i given by: ) 1 P j = M G i) 2 Ψ i V. CAPACIY OF MIMO SC-FDMA he channel capacity of the i-th uer, C i),atthe-th ubcarrier i: C i) = log 2 det I NR + E ) H H 18) N N o j, he intantaneou capacity of the i-th uer, C i), can therefore be given by: C i) = log 2 det I NR + E ) H H H 19) N N o Ψ i A. Capacity Under Unitary Precoding Given the Eigenvalue Decompoition EVD) of H H H, H H H = Q ΛQ H, we can re-write equation 19) a: C i) = r log 2 1+ E ) λ j, 20) N N o Ψ i j=1 A mentioned previouly, optimal power allocation cheme decribed in 14], the optimum power C i) = r log 2 1+ E ) γ j, λ j, 21) N N o Ψ i j=1 Authorized licened ue limited to: UNIVERSIY OF BRISOL. Downloaded on October 13, 2009 at 10:09 from IEEE Xplore. Retriction apply.

5 hi full text paper wa peer reviewed at the direction of IEEE Communication Society ubject matter expert for publication in the WCNC 2009 proceeding. B. Capacity of Pre-equalized MISO Sytem For MISO tranmiion, a a reult of pre-equalization, the -th ub-carrier of the received ignal for the i-th uer can be expreed a: R i) = H H i) X + W i) 22) where H i the equivalent channel frequency repone a a reult of pre-equalization at the -th ub-carrier: where ] H = H 1, G i) 1,, H 2,G i) 2,,, H N, Gi) N, he intantaneou capacity of the i-th uer, C i), can therefore be given by: C i) = log 2 1+ E ) H H N N H 23) o Ψ i = log 2 1+ E N P j H j, G i) j, 2 N N o Ψ i j=1 he upper limit for the i-th uer capacity under preequalization, correpond to the ZF pre-equalizer. hi i due to the channel flattening a a reult of pre-equalization on each antenna, which tranform the frequency elective channel, on each antenna, into a narrowband frequency non-elective channel. he i-th uer capacity under ZF pre-equalization, denoted by C i) ZF, i given by: C i) ZF = log 2 1+ E N N P j 2 N N o Ψ i j=1 = log 2 1+ E N P j 2 24) N o Ψ i j=1 VI. RESULS AND DISCUSSIONS able I how the main imulation parameter ued in thi paper. We aume that the total number of ub-carrier N i 512 and that each uer ha acce to 128 ub-carrier with a preading factor Q of 4. For the 3GPP LE tandard the Spatial Channel Model SCM) ha been tandardized to provide a ocalled geometric or ray-baed model 1]. he SCM define three environment, namely i) Suburban Macro, ii) Urban Macro and iii) Urban Micro. Baed on the 3GPP-SCM channel model, 15] preent a number of different channel model for indoor, rural urban and uburban microcell. Here, we aume the ue of Urban LoS cenario Range1 LoS1 and LoS2) and Range2 B5b LoS3) a well a Urban NLoS cenario C3 15]. Carrier Frequency 2 GHz ranmiion Bandwidth 5MHz otal Number of Sub-carrier N) 512 Number of Sub-carrier per Uer M) 128 Guard Interval P ) 64 SC-FDMA Symbol Duration µ) 150µ Channel Knowledge Perfect Fading per tap i.i.d. Rayleigh Channel Coding Not included Signal Contellation QPSK Delay Profile Model Dependent ABLE I SC-FDMA SIMULAION PARAMEERS In thi ection we preent the PAPR characteritic and the achievable capacity of the cheme dicued in ection III and IV. he reult preented in thi ection were obtained uing the channel model decribed in 15]. For each cenario, different channel realization were imulated, where the phae of each tap wa uniformly ditributed in the interval 0, 2π]. All the reult are then averaged over SC- FDMA ymbol per uer. We aume that perfect nowledge of the channel tate information CSI) i available at the tranmitter. Channel etimation i performed at the bae-tation, where the mean of the wideband channel i normalized to unity. he channel information i then ent bac to the mobile unit. We ignore any overhead and latency iue. Fig. 2. PrPAPR>χ) EGC Pre equalizer MRC Pre equalizer ZF Pre equalizer SC FDMA Beamforming MMSE Pre equalizer PAPR db) PAPR per Antenna of Precoded and Pre-equalized MIMO SC-FDMA Fig. 2 how the Complementary Cumulative Ditribution Function CCDF) of the average PAPR per antenna for different precoded and pre-equalized two tranmit antenna SC- FDMA ytem for the NLoS channel cenario. A can be een, the average PAPR varie with the utilized tranmit proceing technique. hi i due to the different phae and gain on each of the ub-carrier a a reult of precoding and pre-equalization, which increae the dynamic range of the precoded and pre-equalized waveform. We alo oberve that the PAPR of the ZF pre-equalized waveform i lower than the PAPR of the non pre-equalized SC-FDMA waveform. In addition, although EGC pre-equalization increae the PAPR compared to the tandard SC-FDMA waveform, thi cheme offer a lower PAPR compared to MRC, MMSE and beamforming. hi i explained by the phae rotation introduced by EGC, which doe not increae the dynamic range of the SC-FDMA waveform, but rather ditort the phae on all the ub-carrier. Furthermore, the average PAPR per antenna a a reult of beamforming i i higher than the other pre-equalized SC-FDMA waveform. Fig. 3 how the ergodic capacity per ub-carrier of the preequalizer preented in thi wor, for a SISO SC-FDMA ytem under a fading environment. We have only hown the ergodic capacity for cenario LoS1 and NLoS. It wa aumed that the multipath for each channel realization for the channel fading cae are complex independent and identically ditributed i.i.d) Rayleigh variable. hee reult how that the uncontrained ZF pre-equalizer achieve the Shannon AWGN capacity, which repreent the upper bound of the achieved capacity for a ingle carrier ytem. In addition, Authorized licened ue limited to: UNIVERSIY OF BRISOL. Downloaded on October 13, 2009 at 10:09 from IEEE Xplore. Retriction apply.

6 hi full text paper wa peer reviewed at the direction of IEEE Communication Society ubject matter expert for publication in the WCNC 2009 proceeding. Capacity bit//hz) Shannon Capacity for Rayleigh fading LOS Model NLOS Model Shannon Capacity for AWGN channel Contrained MMSE Pre Equalizer Contrained ZF Pre Equalizer BER NLoS ZF Equalizer LoS1 ZF Equalizer LoS2 ZF Equalizer LoS3 ZF Equalizer NLoS ZF Pre equalizer LoS1 ZF Pre equalizer LoS2 ZF Pre equalizer LoS3 ZF Pre equalizer AWGN SNR db) Fig. 3. Ergodic Channel Capacity of Pre-equalized SISO SC-FDMA for Faded Multipath Channel Scenario the uncontrained MMSE pre-equalizer converge toward the Shannon AWGN capacity for high SNR value. Furthermore, the achieved channel capacitie of the contrained preequalizer i below Shannon AWGN capacity curve, and equal the narrowband Rayleigh capacity. hi i due to the power contraint applied to the pre-equalizer, which mean that the pre-equalizer exploit the frequency electivity of the channel but doe not compenate for the fluctuation in the channel mean power. In other word, pre-equalization i only concerned with flattening the channel and doe not necearily perform fat power control. For intance, in a narrowband channel contrained pre-equalization would play no role at the tranmitter. BER SNR db) a) BER Performance under ZF NLoS MMSE Equalizer LoS1 MMSE Equalizer LoS2 MMSE Equalizer LoS3 MMSE Equalizer NLoS MMSE Pre equalizer LoS1 MMSE Pre equalizer LoS2 MMSE Pre equalizer LoS3 MMSE Pre equalizer SNR db) b) BER Performance under MMSE Ergodic Capacity Per Sub carrier bp/hz) MISO without Pre equalization 1 2 SIMO N 1 Pre equalized MISO 2 2 MIMO without pre equalization SISO without Pre equalization SNR db) Fig. 4. Ergodic Channel Capacity of Pre-equalized MIMO SC-FDMA for Faded Multipath Channel Scenario Fig. 4 how the ergodic capacity per ub-carrier for different MIMO SC-FDMA etting with and without contrained preequalization. A can be een, the capacity of the tandard preequalized MIMO SC-FDMA ytem increae by increaing both the number of tranmit and receive antenna, which demontrate how the patial MIMO channel tranlate into an increae in the ytem throughput. A a reult of preequalization, the capacity of the 2-by-1 increae compared to the the tandard MIMO SC-FDMA. hi capacity increae ha a fixed capacity level that i irrepective of N, a demontrated in equation 24). Fig. 5a) and Fig. 5b) how the BER performance of the Fig. 5. BER Performance of the ZF and MMSE Equalizer and Pre-equalizer for SC-FDMA ZF and MMSE equalizer and pre-equalizer, repectively for SISO SC-FDMA. he reult are hown for a ingle reource unit occupying the firt M ub-carrier and employing the L- FDMA ub-carrier mapping. We alo aume perfect channel information at the tranmitter. he reult were obtained uing the ytem parameter decribed in able I. It wa alo aumed that the value of λ for the MMSE pre-equalizer wa equal to N o. Overall, the ZF baed linear equalizer offer the wort BER performance compared to the other cheme for all channel cenario. hi i due to the noie enhancement reulting from the channel inverion applied by the ZF. he MMSE LE offer a performance that i uperior to that of the ZF LE a it tae into account the preence of the noie at the receiver front-end. In contrat, the performance of the MMSE pre-equalizer compared to the MMSE equalizer depend on the channel delay profile. he ZF precoder offer a BER performance that i identical to the AWGN channel a a reult of the channel inverion. VII. CONCLUSION In thi paper we preented the ue of frequency-domain precoding and pre-equalization for uplin MIMO SC-FDMA tranmiion. Overall, the ZF pre-equalizer offer the lowet PAPR compared to the other cheme. A a reult of ZF pre-equalization, the tranmiion channel for each uer i tranformed into a flat frequency non-elective channel for Authorized licened ue limited to: UNIVERSIY OF BRISOL. Downloaded on October 13, 2009 at 10:09 from IEEE Xplore. Retriction apply.

7 hi full text paper wa peer reviewed at the direction of IEEE Communication Society ubject matter expert for publication in the WCNC 2009 proceeding. SISO SC-FDMA. he ytem therefore achieve a performance that i identical to the AWGN channel. For wideband Rayleigh fading channel, a ingle antenna ingle carrier ytem operating under the contrained ZF pre-equalizer i capable of achieving the capacity of the narrowband Rayleigh fading channel. However, if the tranmit power of the ytem i not limited, in other word the pre-equalizer i not contrained, the ytem can achieve the capacity of an AWGN channel. hi reult can be ued to exploit multi-uer diverity when ZF preequalization i combined with a power and reource allocation cheme. In addition, for multiple tranmit antenna ytem, the ue of pre-equalization alo increae the ytem capacity of MIMO SC-FDMA compared to the tandard cae. On the other hand, a the number of the tranmit antenna increae, pre-equalization doe not exploit the available MIMO patial channel, a the overall capacity i equal to the cae of two tranmit antenna. Although ZF pre-equalization achieve a uperior BER performance compared to frequency domain equalization, there are two iue to be conidered in the pre-equalizer deign. Firtly, becaue the implementation of pre-equalization require perfect nowledge of the uplin channel, with no latency, further wor i required to addre how channel etimation and tracing can ait the pre-equalizer in the cae of mobility, channel etimation error and channel mimatch. Secondly, ince the PAPR of the pre-equalizer output i dependent on the channel fading, it i eential to employ PAPR reduction. Although PAPR reduction come at the expene of performance degradation, the overall PAPR reduction mut be greater than the lo in SNR for a given bit error rate. 14] A. Goldmith and P. Varaiya, Capacity of fading channel with channel ide information, Information heory, IEEE ranaction on, vol. 43, no. 6, pp , ] Final report on lin level and ytem level channel Model, IS WINNER D5.4 ver ] J. S. Hammerchmidt, C. Brunner, and C. Drewe, Eigenbeamforming - A Novel Concept in Array Signal Proceing, Proc. European Wirele Conference 2000, Dreden, Germany, Sep REFERENCES 1] 3GPP, echnical Specification Group Radio Acce Networ Phyical layer apect for evolved Univeral erretrial Radio Acce URA), 3GPP, echnical Specification R V7.1.0 Sep. 2006, Releae 7 2] R. van Nee, and R. Praad, OFDM for Wirele Multimedia Communication, Norwood, MA: Artech Houe, ] Hyung G. Myung, echnical Overview of 3GPP Long erm Evolution LE), Feb.8, ] N. avangaran, A. Wilzec,. Kaier, MIMO SC-FDMA ytem performance for pace time / frequency coding and patial multiplexing, Smart Antenna, WSA International IG Worhop on 2008), pp ] S. Jafar, S. Vihwanath and A. Goldmith, Channel capacity and beamforming for multiple tranmit and receive antenna with covariance feedbac, ICC 01, ] A. Paulraj, R. Nabar and D. Gore, Introduction to Space-ime Wirele Communication, Cambridge Univerity Pre, Cambridge, UK, ] G. J. Fochini and M. J. Gan, On limit of wirele communication in a fading environment when uing multiple antenna, Wirele Peronal Communication, vol. 6, pp , ] G.J. Fochini, Layered pace-time architecture for wirele communication in a fading environment when uing multielement antenna, Bell Lab ech. J., pp , ] Hyung G. Myung, Single Carrier Orthogonal Multiple Acce echnique for Broadband Wirele Communication, Polytechnic Univerity, January ] Hyung G. Myung, Junung Lim, and David J.Goodman, Single Carrier FDMA for Uplin Wirele ranmiion, 2006 IEEE Vehicular echnology Magazine 1 September ] H. G. Myung, J.-L. Pan, R. Oleen, and D. Grieco, Pea Power Characteritic of Single Carrier FDMA MIMO Precoding Sytem, IEEE Vehicular echnology Conference VC) 2007 Fall, Baltimore, USA, Oct ] C. Ciochina, D. Catelain, D. Mottier and H. Sari, A Novel Space- Frequency Coding Scheme for Single-Carrier Modulation, Proc. PIMRC 2007, September 2007, Athen, Greece. 13] David Falconer, S. Le Ariyaviitaul, Anader Benyamin-Seeyar, Brian Eidon, Frequency Domain Equalization for Single-Carrier Broadband Wirele Sytem, Communication Magazine, IEEE, Apr Authorized licened ue limited to: UNIVERSIY OF BRISOL. Downloaded on October 13, 2009 at 10:09 from IEEE Xplore. Retriction apply.