Multiple Antenna Technologies

Transcription

1 Multiple Antenna Technologie Manar Mohaien YuPeng Wang KyungHi Chang The Graduate School of Information Technology and Telecommunication INHA Univerity ABSTRACT Multiple antenna technologie have received high attention in the lat few decade for their capabilitie to improve the overall ytem performance. Multiple-input multiple-output ytem include a variety of technique capable of not only increae the reliability of the communication but alo impreively boot the channel capacity. In addition, mart antenna ytem can increae the link quality and lead to appreciable interference reduction. I. Introduction Multiple antenna technologie propoed for communication ytem have gained much attention in the lat few year becaue of the huge gain they can introduce in the communication reliability and the channel capacity level. Furthermore, multiple antenna ytem can have a big contribution to reduce the interference both in the uplink and the downlink by employing mart antenna technology. To increae the reliability of the communication ytem, multiple antenna can be intalled at the tranmitter or/and at the receiver. Alamouti code i conidered a the implet tranmit diverity cheme while the receive diverity include maximum ratio, equal gain and election combining method. Recently, cooperative communication wa deeply invetigated a a mean of increaing the communication reliability by not only conidering the mobile tation a uer but alo a a bae tation (or relay tation). The idea behind multiple antenna diverity i to upply the receiver by multiple verion of the ame ignal tranmitted via independent channel. On the other hand, multiple antenna ytem can tremendouly increae the channel capacity by ending independent ignal from different tranmit antenna. BLAST patial multiplexing cheme are a good example of uch category of multiple antenna technologie that boot the channel capacity. In addition, mart antenna technique can ignificantly increae the data rate and improve the quality of wirele tranmiion, which i limited by interference, local cattering and multipath propagation. Through haping the antenna radiation pattern and adaptively adjuting the antenna weight vector, mart antenna improve the communication link quality by increaing the received ignal power and uppreing the interference.

2 Fig. 1. Multiple antenna technologie. Beide, on-line calibration technique i alo adopted to correct the error due to the ditortion and nonlinearity of the radio frequency component in the antenna array ytem. Fig. 1 ummarize the different multiple antenna technologie and give ome example of thee technologie. Thi paper i organized a follow: in ection II we preent multiple antenna diverity cheme employed at the tranmitter or/and at the receiver. Spatial multiplexing preented by BLAST cheme i detailed in ection III. Section IV i dedicated to ome advanced multiple input multiple-output (MIMO) ytem including multi-uer MIMO and cooperative communication. While technique related to the mart antenna uch a phaed antenna array, witched beam antenna array, and adaptive antenna array are decribed in Section V. Finally, we conclude in Section VI. II. MIMO Diverity In communication ytem, we have to increae the reliability of the communication operation between tranmitter and receiver while maintaining a high pectral efficiency. The ultimate olution relie in the ue of diverity, which can be viewed a a form of redundancy [1]. There are many diverity technique that can be applied to communication ytem; we mention herein time diverity, frequency diverity, and patial diverity or any combination of thee three diveritie. In time diverity, the ame information-bearing ignal i tranmitted in different time lot where a good gain can be achieved when the duration between the two lot, in which the ame ymbol i tranmitted, i greater than the coherence time of the channel. In frequency diverity, the ame information-bearing ignal i tranmitted on different ubcarrier where a good diverity gain can be achieved when the eparation between ubcarrier i greater than the coherence bandwidth.

3 Finally, in patial diverity, the ame information-bearing ignal i tranmitted or received via different antenna where the maximum gain can be achieved when the fading occurring in the channel i independent (or low correlated). In the receiver, diverity gain can be achieved by combining the redundant ignal arriving via independent (or lowly correlated) channel. Fig. how ome poible combination of tranmit diverity which can be achieved when employing multiple tranmit antenna. In the following ection, we preent ome famou pace-time block code applied at the tranmitter ide. We preent alo the combining technique ued when different verion of the information-bearing ignal are received. Finally, we preent a cheme that include tranmit and receive diveritie..1 Space Diverity at the Tranmit Side The baic idea of the ue of tranmit diverity i to reduce the mobile tation (MS) receiver complexity while improving the detection performance. The pioneering work in the tranmit diverity wa done by Alamouti where he propoed hi famou 1 pace-time code. Alamouti cheme achieve diverity gain while requiring only a linear decoder. Later on, Tarokh et al. propoed a generalized theory of the complex orthogonal pace-time code. Baed on Tarokh work, more than two antenna can be ued and the code rate can be fractional. In the following we preent the two different type of pace-time code..1.1 Complex Orthogonal Space- Time Code For thi type of pace time code, the following condition mut be atified Square tranmiion matrix (number of tranmit antenna N t equal to number of ued time lot m) A unity code rate (number of ued time lot m equal to number of tranmitted ymbol l) Orthogonality of the tranmiion matrix in the time and pace H H domain ( SS =S S ) where S H i the conjugate tranpoe of S. Fig.. Tranmit diverity.

4 1 S= * * - 1 Fig. 3. Alamouti cheme example with QPSK modulation. A aid before, the implet complex orthogonal pace-time code i the Alamouti code which ue two tranmit antenna and one receive antenna. Furthermore, Alamouti cheme require that the fading channel envelope remain contant over two time lot. Fig. 3 how an example of the encoding proce of Alamouti cheme with QPSK modulation [], [3]. Fig. 4 how the receiver tructure ued for decoding the combined received ymbol. At the receiver the following ignal are received (with applying the complex conjugate to the received ignal at t ) y1 h1 h 1 n1 * * * * y = h h + 1 n (1) The linear combiner multiplie the received ymbol by the Hermitian tranpoe of the channel matrix (for implicity, we conider that channel i perfectly etimated). The output of the linear combiner i then given by x 1 1 w 1 h1 h x = + + w () Maximum-Likelihood (ML) decoder i then applied to get the tranmitted ymbol. A one can ee, the implicity of the receiver i due to the patio-temporal orthogonality of the tranmiion matrix. ĥ 1 ĥ A complex orthogonal pace-time code uing 4 or 8 antenna wa propoed by Tarokh et al. in [4]. ŝ 1 ŝ Fig. 4. Alamouti code receiver..1. Generalized Complex Orthogonal Space-Time Code The earch for pace-time code with more than two antenna wa tarted by Tarokh, Jafarkhani, and Calderbank. Their work

5 ha built the bai for a theory of generalized complex orthogonal deign. Generalized complex orthogonal deign are ditinguihed from Alamouti code by the following A non-quare tranmiion matrix (number of ued time lot number of Tx antenna) A fractional code rate (number of tranmitted ymbol < number of ued time lot) Orthogonality of the tranmiion matrix i only guaranteed in the time ene. A a conequence of thee characteritic, the pectral efficiency i reduced and the number of time lot over which the channel hould be contant i increaed. The tranmiion matrix of a generalized complex pace-time code with 3 antenna, 4 tranmitted ymbol and 8 ued time lot i given by [5] G 3 = * * * 1 3 * * * 1 4 * * * * * * 4 3 (3) In the literature, more reearch wa done to increae the rate of the pace-time code. For more detail refer to [6]. Table 1 ummarize the difference between Alamouti and pace-time code characterized by the tranmiion matrix G 3. Thee hown coding cheme can be tranmitted in the pace-time domain, pace-frequency domain or in pacefrequency-time domain. Thee coding cheme are thu known a ST, SF, and STF coding, repectively [7]..1.3 Cyclic Delay Diverity (CDD) CDD can be conidered a a very imple tranmit diverity cheme. CDD can achieve tranmit artificial frequency diverity by electing appropriate tranmit delay. In thi method, multiuer diverity, obtained by cheduling baed on frequency domain channel repone, can be improved by adjuting the delay pread (at the tranmitter) which i done by controlling the delay value dependent on the channel condition [8], [9].. Space Diverity at the Receive Side In pace diverity at the receive ide, multiple antenna are ued in the receiver with ufficient pacing between antenna in uch a way mutual correlation between antenna i reduced and a conequence diverity gain i increaed [10]. To get diverity gain at the receiver, received ignal from different antenna are combined. There are four combining method, namely, elect combining (SC), maximal-ratio combining (MRC), equalgain combining (EGC), and quare-law combining. The firt three cheme are linear while the lat require a non-linear receiver. Fig. 5 how a implified block diagram of the linear combining cheme which differ in the weighting vector w. In SC, the ignal at the branch with maximum ignal to noie ratio (SNR) i elected and other received ignal are dicarded. The weighting vector w = (w 1, w,, w M ) i the N th column of the identity matrix of ize M where the N th branch ha the maximum SNR.

6 Table 1. Comparion between Alamouti and generalized complex pace-time code. Space-time code Number of Tx antenna Number of tranmitted ymbol, l Number of ued time lot, m Orthogonality of Tx matrix S Spatiotemporal ene G Only temporal ene Rate = l/m 1 1/ Table. Comparion between diverity combining cheme. Scheme Requiring CSI Outage Probability F(x) Application SC No / M c 1 e x y No contraint MRC Ye k M 1 No contraint x / y 1 x c 1 e k = 0 k! yc EGC Ye No cloed form for M > No contraint SLC No - FSK or DS-CDMA A one can ee, the SC cheme doe not require any channel information except that of SNR. On the other hand, MRC and EGC cheme require the channel tate information (CSI) or a part of it (channel envelope, phae, delay). MRC cheme weight the received ignal according to their reliability; a more reliable ignal ha a high weight while a le reliable ignal ha a mall weight. Alo, the channel phae ditortion i compenated. Finally, ignal are aligned then combined. On the other hand, EGC cheme can be viewed a a implified verion of MRC where ignal are weighted equally (i.e. the weighting vector w = [1, 1,,1] M ) then aligned before being combined coherently. In practice, the phae at different branche can t be often etimated. So, EGC and MRC can t be employed. In uch ituation, quare law combining (SLC) can be applied to obtain patial diverity without requiring phae etimation. Unlike linear combining cheme, SLC cheme can only be applied to modulation cheme which preerve ome ort of orthogonality including frequency-hift keying (FSK) or directequence CDMA [5], [11]. Table ummarize a comparion between combining cheme. It i known that, from a performance point of view, MRC i optimum and give the bet performance among the pre-decribed combining cheme. y 1 y y M. w 1 x x w x w M. Linear combiner Fig. 5. Simple block diagram of linear combining cheme..3 Combined Tranmit/Receive Diverity Spatial diverity cheme explained in the previou two ection can be combined together to achieve diverity at both receive and tranmit ide. An example of uch a hybrid patial diverity cheme i y c

7 the Alamouti/MRC MIMO ytem [1]. III. Spatial Multiplexing A hown in the previou ection, MIMO diverity can be ued in the tranmitter or the receiver ide or in both to increae the reliability of the communication. In thi ection we talk about patial multiplexing cheme which are for goal to increae the channel capacity. The mot known patial multiplexing cheme are the BLAST family which include Vertical-BLAST, Diagonal- BLAST, and Turbo-BLAST. The acronym BLAST tand for Bell Laboratorie Layered Space-Time. 1,1 1, 1,3 K 0,1,3 S 0 0 3, Diagonal-BLAST D-BLAST wa originally propoed by Fochini [13]. In D-BLAST, the ymbol to be tranmitted are arranged on the diagonal of the pace-time tranmiion matrix where element under the diagonal are padded with zero. Fig. 6-a depict the tructure of the D-BLAST tranmitter for four tranmit antenna. At firt, the bit tream i de-multiplexed into four parallel tream which are encoded and modulated independently. Encoded-modulated tream are cycled over time. Equation (4) i an example of the tranmiion matrix when uing four tranmit antenna. 1,4 1, 1 1, K 0 0 0,4, K, K 1, 0 0 K = 3, 3, K 3 3, K 3, K 1 3, K 0 4,1 4, K 4 4, K 3 4, K 4, K 1 4, K (4) Fig. 6. Tranmitter block diagram for BLAST family uing four tranmit antenna.

8 Table 3. MIMO ytem diverity order. MIMO Configuration Diverity order STBC N t N r BLAST N r - N t + 1 The firt diagonal of S i tranmitted via the firt antenna; the econd diagonal i tranmitted via antenna, and o on. 3. Vertical-BLAST A implified verion of D-BLAST wa propoed by Wolnianky known a Vertical-BLAST or V-BLAST [14]. In V- BLAST, incoming data tream i demultiplexed into N t tream each of which i encoded and modulated independently and ent on an antenna of it own. V- BLAST high-level diagram i depicted in Fig. 6-b where four antenna are ued at the tranmit ide. Compared to D-BLAST, V-BLAST doe not include cycling over time, the complexity i ignificantly reduced. In addition, unlike D-BLAST, V- BLAST doe not include any pace-time watage. At the receiver, tranmitted ymbol can be decoded uing ordered erial interference-cancellation (OSIC) detector. For the OSIC to work properly, the number of receive antenna N r mut be at leat a large a the number of tranmit antenna. 3.3 Turbo-BLAST Turbo-BLAST wa firt decribed by Sellathurai and Haykin [15]. The Turbo- BLAST tranmitter tructure i depicted in Fig. 6-c. The data tream bit are firtly demultiplexed into N t parallel tream which are encoded independently uing the block encoder (outer encoder) (i.e. channel coding). The output tream of the outer encoder are interleaved independently and paed to the inner encoder. The miion of the outer encoder i to achieve random-layered pace-time (RLST) coding. The tructure of the RLST encoder, with periodical cyclic pace-time interleaving i depicted in Fig. 6-d. For optimal performance of the RLST code, the receiver hould employ the maximum a poteriori probability (MAP) decoding algorithm. Neverthele, the complexity of the MAP decoding algorithm i very high (increae exponentially with N t ). To decreae the complexity of the receiver, the near-optimal turbo-like receiver can be ued. Thi near-optimal turbo-like receiver i known a iterative detection and decoding (IDD) receiver. Before going further, we lit in Table 3 a comparion between diverity order of the different pace-time coding and the BLAST family cheme. IV. Advanced Topic 4.1 Single and Multi CodeWord MIMO In ingle codeword (SCW) MIMO, an encoded packet i ditributed acro many tream to form the MIMO tranmiion. Feedback i ued to control the rank of the MIMO tranmiion (number of tream ued) a well a the overall rate of tranmiion. In multiple codeword MIMO, everal eparately encoded packet are tranmitted independently over the multiple tream. Here the rate of each tream can be controlled with feedback [16] and [17]. 4. Single-Uer MIMO and Multi-Uer MIMO In ingle-uer MIMO, already explained technique in previou ection are ued where the channel capacity grow linearly with min(nt, Nr) [18].

9 For multi-uer MIMO, which i of high interet reearch topic, it wa hown that for N t tranmitting antenna (at the bae tation) and N r uer, the ame overall capacity can be achieved. Thi later work wa encouraged by applying dirty paper coding [19] where reult howed that if the tranmitter know the interfering ignal, then the channel capacity will not be affected by the preence of the interference [0]. On the other hand, multi-uer MIMO can integrate beamforming to apply patial diviion multiple acce (SDMA). 4.3 Cooperative Communication and Virtual MIMO In cooperative communication, a mobile can act a both a uer and relay. A conequence, mobile end to the bae tation it own data bit and ome of other mobile (ometime called partner) information bit. Fig. 7 how a cooperative cellular ytem where for implicity we conider three cooperative uer and one bae tation [1]. A depicted in Fig. 7, uer 1 cooperate with uer and 3 to end it own information. A a reult, the overall cooperative ytem can be een a virtual-mimo (V-MIMO) and in the above example it i 3 N BS MIMO ytem (for the uplink) where N BS i the bae tation number of receive antenna. Uer and 3 can imply amplify and forward uer 1 received information or detect and forward []. Another method of cooperation i the coded cooperation where different coded portion are ent via different fading channel [3]. Fig. 7. Cooperative communication and virtual-mimo. 4.4 Pre-Coded MIMO with Rank Adaptation Per Antenna Rate Control (PARC) PARC can be conidered a a cloed-loop MIMO ytem where tranmitter ue channel quality indication (CQI) fed by the receiver to elect the bet modulation and coding cheme per antenna. Fig. (8) how a general PARC tranmitter tructure with 4 tranmit antenna [4] Per Group Rate Control (PGRC) In PARC, a CQI feedback i neceary for each tranmit antenna. Thi increae the uplink overhead.

10 Fig. 8. Per Antenna Rate Control (PARC). To olve thi problem, PGRC i ued where a feedback i required per group of antenna. Thi reduce the feedback information while maintaining almot the ame performance of PARC [4] Per Uer Unitary Rate Control (PU RC) PU RC i a multi-uer cloed-loop MIMO ytem. Each uer feed back the CQI to the bae tation. The bae tation ue the CQI to determine the modulation and coding cheme per uer. In addition, bae tation can apply unitary pre-coding and adaptively elect the number of tranmit antenna (rank adaption) [5]. V. Smart Antenna 5.1 Introduction Smart antenna wa born in the early 1990 when well developed adaptive antenna array originate from Radar ytem. Later, Smart antenna technique i applied in wirele communication ytem. Recently, Smart antenna technique ha been propoed a a promiing olution to the future generation of wirele communication ytem, uch a the Fourth-Generation mobile communication ytem, broadband wirele acce network, where a wide variety of ervice through reliable high-data rate wirele channel are expected. Smart antenna technique can ignificantly increae the data rate and improve the quality of wirele tranmiion, which i limited by interference, local cattering and multipath propagation [6], [7]. Smart antenna offer the following main application in high data-rate wirele communication ytem [8], [9]: Spatial Diverity Co-channel interference reduction Angle reue or pace diviion multiple acce (SDMA) Spatial multiplexing Smart antenna ytem can be categorized into three main group: Phaed antenna array ytem, witched beam ytem, and adaptive antenna array ytem. To match the characteritic in each radio frequency chain of the tranmitter and receiver, on-line calibration i required in mart antenna ytem. On-line calibration technique can compenate the error uch a the ditortion of radio frequency component due to mall environment change, the nonlinear characteritic of mixer, amplifier and attenuator, I/Q imbalance error, etc. 5. Phaed Antenna Array Sytem Phaed antenna array i a group of antenna in which the relative phae of the repective ignal feeding the antenna are varied in uch a way that the effective radiation pattern of the array i reinforced in a deired direction and uppreed in undeired direction.

11 Phaed antenna array ytem i uually utilized in radio frequency (RF) or intermediate frequency (IF) with the ytem central frequency larger than 10 GHz, uch a atellite communication ytem [30]. There are two main different type of phaed array, alo called beamformer. There are time domain beamformer and frequency domain beamformer. 5.3 Switched Beam Sytem The witched beam method i conidered a an extenion of the current ectorization cheme. In the witched beam approach, the ector coverage i achieved by multiple predetermined fixed beam pattern with the greater gain placed in the centre of a beam [30]. When a mobile uer i in the vicinity of a beam, then the ignal at the output port will be given a in (5). Thi enable the witched beam ytem to elect the ignal from the output port correponding to that beam. A the mobile move to the coverage of another beam during the call, the ytem monitor the ignal trength and witche to other output port a required. A baic witched beam antenna architecture i hown in Fig. 9. And Fig. 10 illutrate the produced antenna pattern with 4 antenna. L y () t = () t G () t I () t G ( θ ) (5) i i i l li l l= 1 where y i (t) i the total ignal appearing at port i, i (t) i the ignal ource, I l (t) i the interfering ignal ource located at arbitrary angle θ l, G i i the tranfer function between ignal ource along the main beam and their correponding output port, G li i the tranfer function between interference ignal l and port i. Fig.9. Functional block diagram of witched beam ytem. Fig.10. Produced antenna pattern of witched beam ytem with 4 antenna. Switched beam ytem can offer everal advantage, including Low complexity and cot. Since witched beam ytem only require a beamforming network, RF witche, and imple control logic, they are relatively eay and cheap to implement. Moderate interaction with bae tation receiver. In practice, witched beam ytem can imply replace conventional ector antenna without requiring ignificant modification to the radio bae tation antenna interface or the baeband algorithm implemented at the receiver. Coverage extenion. The antenna array aperture gain will boot the link

12 budget, which could be tranlated to a coverage extenion. 5.4 Adaptive Antenna Array Sytem (AAA) Adaptive antenna date back to The original work wa attributed to L. C. Van Atta work, Electromagnetic Reflection. Since then, adaptive beamforming technique have been employed to remove unwanted noie and jamming from the output, mainly in military application. With the thriving commercial wirele communication indutry and the advancing microproceor technologie, the adaptive beamforming technique have found their application in commercial wirele communication. With powerful digital ignal proceing (DSP) hardware at the bae-band, algorithm could control antenna beam pattern adaptively to the real ignal environment, forming beam toward the deired ignal while forming null to co-channel interferer. Thu, the ytem performance i optimized in term of link quality and ytem capacity [31]. Adaptive antenna array can be utilized in the tranmitter ide, which i known a tranmit beamforming (TxBF) or in the receiver ide, which i called receive beamforming (RxBF) Tranmit Beamforming (TxBF) The implementation of adaptive antenna array technique in a handet i difficult with today` hardware due to it limitation in ize, cot, and energy torage capability, while it i feaible to adopt antenna array at bae tation. Tranmit beamforming provide a powerful method for increaing downlink capacity [3]-[35]. The idea of TxBF i imilar to the pre-coded MIMO technique but with different trategie to calculate the tranmit weight vector. TxBF adjut the antenna main lobe toward to the deired uer and reduce the interference to other uer. A imple illutration of TxBF i hown in Fig. 11. Fig.11. An illutration of TxBF. Eigenvector TxBF Algorithm Eignenvector TxBF algorithm i widely ued for TxBF. The eigenvector of the patial covariance channel matrix i calculated a R = λh (6) where R i the autocovariance matrix of the deired uer` ignal, and H i the patial covariance channel matrix. The eigenvector λ max which correpond to the larget eigenvalue will be elected a the weight vector [36]. One example of beam pattern for 4 uniform linear array element i hown in Fig 1.

13 Fig. 1. Example beam pattern of 4 antenna element in a ectorized ytem for a ingle ector (main beam direction i 40 ). Tranmit Adaptive Array (TxAA) Algorithm Tranmit adaptive array (TXAA) i a technique in which the uer periodically end quantized etimate of the optimal tranmit weight to the BS via a feedback channel. The tranmitter weight are optimized to deliver maximum power to the uer. The optimal tranmit weight are given by H H w = H / HH (7) where w i the tranmit weight vector and H i the channel matrix. The weight are normalized o that the total tranmitted power i not altered. In the cae of multipath channel emanating from each antenna, the optimal weight will be given by the principal eigenvector H of the channel correlation matrix H H Receive Beamforming (RxBF) Beamforming alo can be applied in the uplink to improve the link quality and uppre the co-channel interference, which i known a receive beamforming (RxBF). Through RxBF, mart antenna ytem can receive predominantly from a deired direction (direction of the deired ource) compared to ome undeired direction (direction of interfering ource). Thi implie that the digital proceing ha the ability to hape the radiation pattern to adaptively teer beam in the direction of the deired ignal and put null in the direction of the interfering ignal. Thi enable low co-channel interference and large antenna gain to the deired ignal. Baed on the reference ignal adopted in the beamforming algorithm, RxBF can be claified into patial reference beamforming (SRB), temporal reference beamforming (TRB), and ignal tructure reference beamforming (SSRB). Spatial Reference Beamforming (SRB) Spatial reference beamforming method i ometime referred a direction of arrival (DoA) method. SRB etimate the direction of arrival of the ignal baed on the patial reference ignal, uing any of the technique like multiple ignal claification or etimation of ignal

14 Fig. 13. A general tructure of SRB. parameter via rotational invariance technique algorithm or their derivative. They involve finding a patial pectrum of the antenna/enor array, and calculating the DoA from the peak of thi pectrum [37]. A general architecture of SRB algorithm i hown in Fig. 13. The general tep of SRB method are hown a follow: DoA Etimation Arbitrary Array: MUSIC, etc. Linear Array: ESPRIIT, etc. Beam Synthei Gram-Schmidt, etc. Combining EGC, MRC, Wiener Filter, etc. Multiple ignal claification (MUSIC) algorithm etimate the DoA of the deired ignal by uing an eigen-pace method baed on a patial reference ignal. MUSIC require intenive calculation of eigenvalue and eigenvector of an autocorrelation matrix of the input vector from the receiving antenna array. A general tep of MUSIC algorithm i hown below: Collect received ample and etimate the covariance matrix of the received ample. Perform eigen-decompoition of the covariance matrix. Calculate patial pectrum. Etimate DoA by locating peak in the pectrum. Etimation of ignal parameter via rotational invariance technique (ESPRIT) i alo well known for the SRB method. In addition, ESPRIT ha many important advantage over MUSIC algorithm [38]: No knowledge of the array geometry and element characteritic are required. Much le complex on computation. No calibration of the array i required. The algorithm imultaneouly etimate the number of ource and DoA`

15 Fig. 14. A general tructure of TRB. Temporal Reference Beamforming (TRB) Temporal Reference Beamforming hown in Fig. 14, i a method ued to create the radiation patter of the antenna array by adding contructively the phae of the ignal in the DoA of the deired uer, and nulling the pattern of the interfering uer baed on the temporal reference ignal [39]. Baed on the temporal reference ignal and ome predefined adaptive weight calculation criterion, ome adaptive algorithm uch a LMS (Leat Mean Square), RLS (Recurive Leat Square), and DSMI (Direct Sample Matrix Invere) algorithm, are ued to adjut the weight vector of the antenna array to improve the link quality. The general characteritic of TRB are a follow: Good performance in multipath channel environment Computationally inexpenive Requiring Training equence Difficult to apply TxBF becaue of the abence of DoA information Criterion of Adaptive Weight Calculation In the minimum mean-quare error (MMSE) criterion, the weight are choen to minimize the mean-quare error (MSE) between the beamformer output and the temporal reference ignal. While in the maximum ignal-to-interference ratio (MSIR) criterion, the weight are choen to directly maximize the ignal-tointerference ratio (SIR). And The minimum variance (MV) criterion chooe the weight that minimize the variance of the output power. All the above three criterion ha the ame form of w = βr v (8) opt 1 i 1 where R i the invere of the covariance i matrix of the interference ignal received in the antenna array and V i the antenna array propagation vector [40]. Let u aume that d(t) i the tranmitted temporal reference ignal and R u i the covariance matrix of interference ignal at the output of the beamformer. The calculation of β for MMSE, MSIR

16 Table 4. β Calculation. Criterion MMSE MSIR MV β Ed { ( t)} 1 + v R v 1 Ed { ( t)} H i Ed SIR { ( t)} H opt g v w H 1 v R v u and MV criterion are ummarized in Table 4. Adaptive Beamforming Algorithm The leat mean quare (LMS) algorithm ue the temporal reference ignal to update the weight at each iteration. In the LMS algorithm, we are earching for the optimal weight that would make the array output either equal or a cloe a poible to the reference ignal, which i the weight that minimize the MSE. Since the MSE ha a quadratic form, moving the weight in the negative direction of the gradient of the MSE hould lead u to the minimum of the error urface. The weight update equation i hown in (9) [30]. problem by normalizing with the power of the input. The weight updating function of NLMS algorithm i hown a ˆ μ * w( t 1) w( t) x ( t) ε (10) + = + a + x() t Recurive leat quare (RLS) algorithm i derived to overcome the drawback of low convergence peed in the LMS algorithm, when the eigenvalue pread of the correlation matrix R of received ignal vector x i large. RLS algorithm replace the tep ize μ with the invere of R. The weight are then updated uing (11). 1 * w( t+ 1) = w( t) R x ( t+ 1) ε (11) ( 1) ( ) ( 1) * w t+ = w t μx t+ ε (9) where μ i a contant, called the tep ize, which determine how cloe the weight approach the optimum value after each iteration and it control the convergence peed of the algorithm. Andε i the error ignal between the temporal reference ignal and the received ignal at the beamformer output. x(t+1) i the received ignal vector at the antenna array at time t+1. The main drawback of the LMS algorithm i that it i enitive to the caling of it input. Thi make it very hard (if not impoible) to chooe a tep ize μ that guarantee tability of the algorithm. The normalized leat mean quare (NLMS) algorithm i a variant of the LMS algorithm that olve thi Fig. 15. Performance comparion among LMS, NLMS, and RLS algorithm. Fig. 15 how a imple performance comparion of the above three algorithm in OFDMA ytem under the Rayleigh fading channel with 8 antenna element [41]. From thi figure, we ee that RLS algorithm perform bet due to it fater

17 convergence peed than LMS and NLMS algorithm. Signal Structure Reference Beamforming (SSRB) SSRB method i baed on inherent tructure of the tranmit ignal of the implicit kind reference ignal. Algorithm uch a blind beamforming, leat quare, and contant modulu algorithm, are baed on the SSRB method. SSRB method i robut againt different propagation condition and doe not require the array manifold knowledge. But the convergence problem become the main drawback of the SSRB method. VI. Concluion In thi paper we introduced the multi antenna technologie which can be conidered a one of the mot vivid area of reearch. Multiple antenna technologie were categorized into two main group where in the firt group we introduced ome technique related to patial diverity and patial multiplexing by outlining the gain achieved by thee cheme. Furthermore, we introduced the mart antenna technique and the up-todate reearch progre in thi field. The advantage of multiple antenna ytem make of them a very trong candidate to increae link reliability, increae channel capacity and reduce interference in both uplink and downlink. Reference [1] A. Goldmith, Wirele Communication, Cambridge Univ. Pre, 005. [] M. Alamouti, A imple tranmit diverity technique for wirele communication, IEEE J. Sel. Area Commun., 16, pp , [3] G. Toulo, MIMO Sytem Technology for Wirele Communication, Taylor and Franci, 006. [4] V. Tarokh et al., Space-time block code from orthogonal deign, IEEE Tran. Inf. Theory, vol. 45, no. 5, pp , [5] S. Haykin and M. Mohr, Modern Wirele Communication, Pearon Prentice Hall, 005. [6] W. Su and X-G. Xi, Two generalized complex orthogonal pace-time block code of rate 7/11 and 3/5 for 5 and 6 tranmit antenna, IEEE Tran. Inf. Theory, vol. 49, no. 1, pp , Jan 003. [7] K. Suto, and T. Ohtuki, Space-timefrequency block code over frequency elective fading channel, IEICE Tran. Commun., vol. E87-B, no. 7, pp , July 004. [8] NTT DoCoMo, Multi-degree cyclic delay diverity with frequency-domain channel dependent cheduling, 3GPP TSG-RAN WG1 meeting #44bi, R , Mar [9] Samung, Adaptive cyclic delay diverity, 3GPP TSG-RAN1 43#, R , 7 th 11t, Nov. 005, Seoul, Korea. [10] G. Stüber, Principle of Mobile Comm-unication, Kluwer, 001. [11] M. K. Simon and M-S. Alouini, Digital Communication over Fading Channel, Wiley, 005. [1] E. Biglieri et al., Diverity, interference cancellation and patial

18 multiplexing in MIMO mobile WiMAX ytem, to appear in IEEE WiMAX 07, 007. [13] G. Fochini, Layered pace-time architecture for wirele communication in fading environment when uing multi-element antenna, Bell Lab Technical Journal, Autumn [14] P. Wolnianky et al., V-BLAST: an architecture for realizing very high data rate over the rich-cattering wirele channel, URSI International Sympoium on Signal, Sytem and Electronic, [15] M. Sellathurai and S. Haykin, TURBO-BLAST for wirele communication: theory and experiment, IEEE Tran. Signal Proceing, vol. 50, no., 10 pp , Oct. 00. [16] A. Hottinen et al., Indutrial embrace of mart antenna MIMO, IEEE Wirele Commun. Mag., Aug [17] A. Jette et al. UMBFDD candidate propoal for IEEE 80.0, IEEE C /09, Mar [18] A. Goldmith et al., Capacity limit of MIMO channel, IEEE J. on Sel. Area in Commun., vol. 1, no. 5, pp , Jun [19] M. Cota, Writing on dirty paper, IEEE Tran. Info. Theory, vol. 9, no. 3, pp , May [0] Q. Spencer et al., An introduction to the multi-uer MIMO downlink, IEEE Commun. Mag., pp , Oct [1] A. Noratinia and A. Hedayat, Cooperative communication in wirele network, IEEE Communication Magazine, pp , Oct [] J. Laneman et al., An efficient protocol for realizing cooperative diverity in wirele network, Proc. IEEE ISIT, June 001, pp.94. [3] T. Hunter and A. Noratinia, Diverity through coded cooperation, IEEE Tran. on Wirele Commun., vol. 5, no., pp. 1-7, Feb [4] Texa Intrument, MIMO OFDMA E-UTRA propoal for different antenna configuration, 3GPP TSG RAN1 WG1 #43, R , Nov [5] Samung, Downlink MIMO for EUTRA, 3GPP TSG RAN1 WG1 #43, R , Nov [6] S. Ohmori, Y. Yamao, and N. Nakajima, The future generationof mobile communication baed on broadband acce technologie, IEEE Commun. Mag., vol. 38, pp , Dec [7] K. Sheikh, D. Gebert, D. Gore, and A. Paulraj, Smart antenna for broadband wirele acce network, IEEE Commun. Mag., vol. 37, pp , Nov [8] A. Lozano, F.R. Farrokhi, and R.A. Valenzuela, Lifting the limit on high-peed wirele data acce uing antenna array, IEEE Commun. Mag., pp , Sept [9] Jeffrey H. Reed, Smart antenna: A ytem level overview for oftware defined radio for creating an API, SDRF-04-I-0057-V0.00, Software Defined Radio Forum, Jan [30] Ahmed EI Zooghby, Smart Antenna Engineering, Artech Houe, 005. [31] Michael Chryomalli, Smart antenna, IEEE Antenna and Propagation Mag., vol. 4, pp ,, June 000. [3] H. Boche and M. Schubert, Theoretical and experimental comparion of optimization criteria for downlink beamforming, European

19 Tran. on Telecomm., vol. 1, no. 5, pp , 001. [33] R. M. Buehrer, A. G. Kogianti, S.-C. Liu, J. Tai, and D. Uptegrove, Intelligent antenna for wirele communication -uplink, Bell Lab Technical Journal, vol. 4, no. 3, pp , [34] H. Holma and A. Tokala, WCDMA for UMTS, John Wiley & Son, 000. [35] A. Yener, R. D. Yate, and S. Uluku, Interference management for CDMA ytem through power control, multiuer detection, and beamforming, IEEE Tran. on Comm., vol. 49, no. 7, pp , 001. [36] S.J. Ko, J. Heo, and K.H. Chang, "An effective downlink reource allocation for upporting heterogeneou traffic data in an OFDM/SDMA-baed cellular ytem," in Proc. IEEE GLOBECOM, Nov. 006, WLC7-5. [37] Lal C. Godara, Application of antenna array to mobile communication, part II: beamforming and direction-of-arrival conideration, IEEE Proc., vol.85, pp , Aug [38] Haardt, M., Noek, J.A., Unitary ESPRIT: how to obtain increaed etimation accuracy with a reduced computational burden, IEEE Tran. on Signal Proceing, vol.43, pp , [39] B. L. P. Cheung, Simulation of Adaptive Array Algorithm for OFDM and Adaptive Vector OFDM Sytem, Mater of Science in Electrical Engineering of Virginia Polytech., Sept. 00. [40] John Litva and Titu Kwok-Yeung Lo, Beamforming in Wirele Communication, Artech Houe, [41] J. Heo and K.H. Chang, "Tranmit and receive beamforming for OFDMA/TDD ytem," in Proc. ISAP, Aug. 005, pp Manar Mohaien July, 001: BS, Communication and Control, Univerity of Gaza, Gaza, Paletine Sep., 005: MS, The School Polytechnic of Nice Univerity, Sophia- Antipoli, France Feb., 006 ~ Preent: Ph.D. tudent, The Graduate School of IT & T, INHA Univerity 001 ~ 003: The Paletinian Telecommunication Company (JAWWAL) <Rreearch Interet> MIMO Detection and Co- Channel Interference Cancellation Yupeng Wang July, 004: BS, Communication Engineering, Northeatern Univerity, Shenyang, China July, 006: MS, The Graduate School of IT & T, INHA Univerity Sep., 006 ~ Preent: Ph.D. tudent, The Graduate School of IT & T, INHA Univerity <Reearch Interet> 3GPP LTE Sytem, Radio Reource Management, MIMO Technique, and UWB KyungHi Chang Feb., 1985: BS, Electronic Engineering, Yonei Univerity Feb., 1987: MS, Electronic Engineering, Yonei Univerity Aug., 199: Ph.D., EE Dept., Texa A&M Univ ~ 1990: A Member of Reearch Staff, Samung Advanced Intitute of Technology (SAIT) 199 ~ 003: A Principal Member of Technical Staff (Team Leader), Electronic and Telecommunication Reearch Intitute (ETRI) 003 ~ Preent: Aociate Profeor, The Graduate School of IT & T, INHA Univerity

20 <Reearch Interet> RTT deign for IMT- Advanced & 3GPP LTE Sytem, WMAN Sytem Deign, Cognitive Radio, Cro-layer Deign, and Cooperative Relaying Sytem