IMPROVED SPATIAL MODULATION FOR HIGH SPECTRAL EFFICIENCY

Transcription

1 IMPOVED SPIL MODULION FO HIGH SPECL EFFICIENCY ajab M. Legnain, ohdy H.M. Hafez 1 and bdelgader M. Legnain 2 1 Department of Sytem and Computer Eng., Carleton Univerity, Ottawa, Canada {rlegnain, hafez}@ce.carleton.ca 2 Department of Electrical and Electronic Eng., Garyouni Univerity, Benghazi, Libya. legnain@ieee.com BSC Spatial Modulation (SM) i a technique that can enhance the capacity of MIMO cheme by exploiting the index of tranmit antenna to convey information bit. In thi paper, we decribe thi technique, and preent a new MIMO tranmiion cheme that combine SM and patial multiplexing. In the baic form of SM, only one out of M available antenna i elected for tranmiion in any given ymbol interval. We propoe to ue more than one antenna to tranmit everal ymbol imultaneouly. hi would increae the pectral efficiency. t the receiver, an optimal detector i employed to jointly etimate the tranmitted ymbol a well a the index of the active tranmit antenna. In thi paper we evaluate the performance of thi cheme in an uncorrelated ayleigh fading channel. he imulation reult how that the propoed cheme outperform the optimal SM and V-BLS (Vertical Bell Laboratorie Layered pace-time at high ignal-to-noie ratio (SN). For example, if we eek a pectral efficiency of 8 bit//hz at bit error rate (BE) of 10-5, the propoed cheme provide 5dB and 7dB improvement over SM and V-BLS, repectively. KEYWODS Spatial Modulation (SM), MIMO ytem, Maximum Likelihood detection 1. INODUCION Wirele communication ytem uing MIMO (Multiple Input Multiple Output) have been hown to achieve ignificantly higher pectral efficiencie than conventional ingle-antenna ytem. In [1], thi performance improvement wa demontrated uing the Vertical-Bell Laboratorie Layered Space-ime (V-BLS) ytem. In [2][3], the patial modulation (SM) technique wa introduced. Intead of the normal twodimenional modulation (e.g. QM), the SM introduce a third dimenion which i the index of the antenna where the ymbol i emitted from. In the baic form of SM, the tranmitter ha acce to M antenna, but only one out of the M antenna i ued to tranmit in any given ymbol interval. he receiver mut determine which of the M antenna wa elected for tranmiion. he choice of one out of M antenna convey log 2 (M ) bit of information. t the receiver, iterative-maximum ratio combining (i-mc) i ued to etimate both the tranmitted ymbol and the index of the active antenna. hi technique achieve comparable performance with V-BLS, but with ignificantly lower complexity at the receiver [2]. In [4] Jeganathan et al. propoed an optimal detector for SM, which howed ignificant improvement over V-BLS and conventional SM (SM with i-mc detector) with reaonable increae in receiver complexity. In [5] Youni et al. propoed the Sphere decoder (SD) for SM to reduce the receiver complexity. It wa hown that SM with SD can achieve comparable performance to the optimal SM decoder with lower complexity. DOI : /ijdp

2 2. ELED WOK In [6] and [7] Jeganathan et al. preented a new modulation cheme baed on SM, called generalized pace hift keying (GSSK) and pace hift keying (SSK), repectively. In SSK, the information bit are conveyed uing the antenna index, and in GSSK, the information bit are conveyed uing combination of active antenna indexe. Generalied SM wa propoed in [8] and [9] which extended the concept of SM. In thee cheme, at each time interval one ymbol i tranmitted by a combination of active antenna. In [10] Başar et al. ued Space-ime Block coding (SBC) for SM. he SBC-SM applie the tranmit diverity of the SBC to the SM cheme. heir imulation reult howed that the SBC-SM provide better performance compared to SM and can offer 3-5 db improvement in bit error rate performance over SM and V-BLS (depending on pectral efficiency). In thi paper, we propoe a new MIMO tranmiion cheme baed on SM. Unlike the SM which ue only one out of M antenna to tranmit a ymbol in any given ymbol interval [2][3], the propoed cheme ue M antenna to tranmit M ymbol imultaneouly, where M < M. he propoed cheme group a equence of independent random bit into block, in which each block contain log 2 ( M M M ) bit, (M i modulation order). he firt log 2 (M ) bit are ued to elect M tranmit antenna, with indexe i, i+1, (i+m -1), where i i the antenna index. hen, the lat log 2 ( M M ) bit are tranmitted on the M antenna after being modulated uing a conventional modulation cheme (e.g., M-PSK, M-QM, etc.). For example, conider M = 4, M = 2, and M = 2, then four bit can be tranmitted imultaneouly. Suppoe a block of four bit to be tranmitted a, ay [ ], then, the active antenna will be antenna indexe 2 and 3, and the tranmit ymbol will be 1 and -1 on antenna 2 and 3, repectively. able 1 illutrate the mapping of the propoed cheme. Note that, when M = 1 the cheme become conventional SM. able 1. Propoed cheme mapping: M = 2, M = 4, M = 2. Block Input ctive antenna ranmit ymbol vector, x ,2 [ ] ,2 [ ] ,2 [ ] ,2 [ ] ,3 [ ] ,3 [ ] ,3 [ ] ,3 [ ] ,4 [ ] ,4 [ ] ,4 [ ] ,4 [ ] ,1 [ ] ,1 [ ] ,1 [ ] ,1 [ ] 14

3 he ret of the paper i organized a follow: In Section 3 we decribe the mapper and the detector of the new MIMO tranmiion cheme. Simulation reult and concluion are provided in Section 4 and 5, repectively. hroughout the paper, the following notation are ued. Bold lowercae and bold uppercae letter denote vector and matrice, repectively. We ue [.], r[.], [.] *, and [.] H to denote tranpoe, trace, conjugate and Hermitian of a matrix or a vector, repectively. Furthermore, we ue. F to denote Frobeniu norm of a matrix or a vector, and E[.] to denote the expectation. 3. SYSEM MODEL We conider a MIMO ytem where M i the total number of tranmit antenna and M i the total number of receive antenna. During a tranmiion period, the number of active antenna i M, (M < M ). ll channel element are aumed to be mutually uncorrelated flat fading channel. he ytem model of the propoed cheme i hown in Fig. 1. In the Figure, b i a equence of independent random bit to be tranmitted. he new cheme group the incoming bit into block of log 2 ( M M M ) bit. Each block i mapped into a vector, which i then tranmitted over MIMO channel, (e.g. ee table 1) b M M M iˆ, ˆ iˆ bˆ Figure 1. he propoed cheme ytem model he M 1 tranmitted ignal vector i given by = Ξ i x, (1) where Ξ denote the Circular hift operation, for example, if a = [ a1, a2,..., am ] then Ξ 2 a = [ am, a1,..., a( M 1+ 2) ]. = [ 1, 2,..., M,0,...,0( )], M M M 1, 2,..., are the tranmitted ymbol which are elected from an M -ary ignal contellation. he covariance matrix of, E[ H = ], mut atify the power contraint, r[ ] = E. In other word, he average tranmit energy per ymbol i E, where E i the average tranmit energy. hen, the propoed M M M cheme can tranmit log 2 ( M ) bit imultaneouly. t the receiver, the received ample vector on the receive antenna can be expreed a y = H x + n (2) where y = i the M 1 received ample vector, and n = n, n,..., i the [ y1, y2,..., y M ] [ 1 2 n M ] M 1 additive noie vector, in which each element i aumed to be an independent and 15

4 2 identically ditributed (iid) zero mean complex Gauian random variable with variance σ. H N i the channel matrix between tranmit antenna and receive antenna, and it i given by h1,1 h1,2 L h1, M = h2,1 h2,2 M M H, M M O M hm h,1 M h,2 L M, M (3) where h j,i i the complex fading coefficient between the i th tranmit antenna and the j th receive antenna. h j,i i aumed to be iid complex zero mean Gauian with variance one. he receiver ue maximum likelihood detector to etimate the index, î, and the tranmitted ymbol vector, ŝ. he ML detector etimate the index i, and the tranmitted ymbol vector,, a [11] where Pr [ˆ, i ˆ] = arg max Pr Ξi ( y H, ) i, Ξi ( y H, ) = arg min y H, = π i, M Ξ 2 i F 1 y H, exp 2M 2 σ N σ N Ξ 2 i Ξi i the conditional probability denity function (PDF) of y given H and. Equation (4) etimate both the index and the tranmitted ymbol jointly by earching over all combination of the index i and the ymbol vector. 4. SIMULION ESULS In thi ection, we provide imulation reult for the propoed MIMO tranmiion cheme and compare it with the reult of optimal SM and V-BLS. he V-BLS ytem ue minimum mean quare error ordered ucceive interference cancellation (MMSE-OSIC) detection [12]. Monte Carlo imulation are ued to evaluate bit error rate (BE) performance of the propoed cheme, SM and V-BLS for different pectral efficiencie (η ), number of tranmit antenna (M ) and number of active antenna (M ). We aume uncorrelated flat ayleigh fading channel. M-QM modulation with Gray mapping i ued in the imulation. Figure 2 how the BE performance for 6 bit//hz of QM new cheme with M =2, QM, SM and QM V-BLS. From Figure 2, the new cheme outperform the optimal SM and V-BLS. t BE of 10-5, the new cheme provide SN gain of about 2.2 db over SM and V-BLS. Figure. 3 how the BE performance for 8 bit//hz of QM new cheme with M =3, QM, SM and QM V-BLS. From the figure, the new cheme provide SN gain of 5 db and 7 db over optimal SM and V-BLS at BE of 10-5, repectively. F (4) (5) 16

5 Figure. 4 how the BE performance for 12 bit//hz of QM with M =3, and QM with M =2 for the new cheme, QM for SM and QM V- BLS. It can be een that, the new cheme provide huge performance improvement over optimal SM and V-BLS. t 10-3, the 8 4 new cheme provide about 7 db and 10 db SN gain over optimal SM V-BLS, repectively. From Figure 2, 3 and 4, we conclude that the propoed cheme ha better BE performance than V-BLS, becaue it provide a full diverity order and it ue the index of antenna to convey information which lead to lower the modulation order. lo the propoed cheme outperform the optimal SM, due to the fact that the modulation order ued in the propoed cheme i lower than that ued in optimal SM. 5. CONCLUSIONS In thi paper, we propoed a new MIMO tranmiion cheme to improve the pectral efficiency. In the cheme, we combine SM with patial multiplexing. hi new propoed SM cheme ue everal antenna to tranmit different ymbol at the ame time lot, where the active antenna are ubet of a larger et of antenna. By computer imulation, BE performance for the propoed cheme wa evaluated for uncorrelated ayleigh fading channel and wa compared to optimal SM and V-BLS. he imulation reult how that the new MIMO tranmiion cheme outperform optimal SM and V-BLS at high SN. Furthermore, the performance improvement of the new cheme over optimal SM and V-BLS increae a the tranmiion rate increae, which make it a potential candidate for high data rate tranmiion ytem e.g., WiMX and LE-dvanced SM 4x4 16 QM VBLS 3x4 4 QM New cheme 4x4 4 QM M = BE SN(dB) Figure 2. BE performance for 6 bit//hz. 17

6 10 0 SM 4x4 64 QM VBLS 4x4 4 QM New cheme 4x4 4 QM, M = BE SN(dB) Figure 3. BE performance for 8 bit//hz BE SM 8x4 512 QM VBLS 4x4 8 QM New cheme, 4x4 32 QM, M =2 New cheme 8x4 8 QM, M = SN(dB) Figure 4. BE performance for 12bit//Hz. 18

7 EFEENCES [1] P. W. Wolnianky, G. J. Fochini, G. D. Golden, and.. Valenzuela, V-BLS: an architecture for realizing very high data rate over the rich-cattering wirele channel,. USI- 98 International Sympoium on Signal, Sytem, and Electronic, Pia, Italy, pp , 29 Sep-2 Oct [2]. Meleh, H. Haa, C. hn, and S. Yun, Spatial modulation - a new low complexity pectral efficiency enhancing technique, Firt international Conference on Communication and Networking in China (ChinaCom 06), Beijing, China, pp.1-5, Oct [3]. Meleh, H. Haa, S. Sinanovic, C. W. hn, and S. Yun, Spatial modulation, IEEE ranaction on Vehicular echnology, vol. 57,no.4, pp , July 2008 [4] J. Jeganathan,. Ghrayeb, and L. Szczecinki, Spatial modulation: optimal detection and performance analyi, IEEE Commun. Lett., vol. 12, no. 8, pp , ug [5]. Youni,. Meleh, H. Haa, and P. M.Grant, "educed Complexity Sphere Decoder for Spatial Modulation Detection eceiver," Global elecommunication Conference (GLOBECOM 2010), Miami, Florida, US., 6-10 Dec [6] J. Jeganathan,. Ghrayeb, and L. Szczecinki, Generalized pace hift keying modulation for MIMO channel, IEEE 19th International Sympoium on Peronal, Indoor and Mobile adio Communication, ( PIMC 2008), Canne, France, Sep [7] J. Jeganathan,. Ghrayeb, L. Szczecinki, and. Ceron, Space hift keying modulation for MIMO channel, IEEE ran. Wirele Commun., vol. 8, no. 7, pp , Jul [8] Fu, J. and Hou, C. and Xiang, W. and Yan, L. and Hou, Y., " Generalied patial modulation with multiple active tranmit antenna,", GLOBECOM Workhop (GC Wkhp), 2010 IEEE, pp , 6-10 Dec [9] Youni,. and Serafimovki, N. and Meleh,. and Haa, H., "Generalied patial modulation," Conference ecord of the Forty Fourth ilomar Conference on Signal, Sytem and Computer (SILOM 2010), pp , 7-10 Nov [10] E. aşar, U. ygölü, E. Panayirci, H. V. Poor, "Space-time block coding for patial modulation," 21t International Sympoium on Peronal Indoor and Mobile adio Communication (PIMC), vol., no., pp , Sept [11] J. G. Proaki, M. Salehi, Digital Communication, 5th ed. New York: McGraw Hill, Inc., [12]. Bohnke, D. Wubben, V. Kuhn, and K.-D. Kammeyer, educed complexity MMSE detection for BLS architecture, GLOBECOM 03., pp uthor ajab M. Legnain i a Ph.D tudent at Sytem and Computer Engineering, Carleton Univerity, Ottawa, Canada. He received a B.Sc. and M..Sc in Electrical and Electronic Engineering both from Univerity of Garyouni, Benghazi, Libya, in 2000 and 2004 repectively. He had been a faculty member of Univerity of Garyouni at Electrical and Electronic Department from 2006 to Hi current reearch focue on mart antenna and relay network. ohdy H.M. Hafez obtained the Ph.D. in Electrical Engineering, from Carleton Univerity, Ottawa, Canada. He joined the Department of Sytem and Computer Engineering, Carleton Univerity a an aitant profeor, and he i now a full profeor. Dr. Hafez ha many year experience in the area of Wirele communication, F and pectrum engineering. He ha lectured extenively in wirele and related area. Hi current reearch focue on broadband wirele networking, 3G/4G/LE, wirele over fiber and enor network. 19