2012 IEEE 16th International Symposium 2012 ISCE 1569579313 THE OPTIMIZATION CAPACITY OF THE MU-MIMO WITH CHANNEL QUALITY INFORMATION Jean-Baptiste YAMINDI Hong Ji Wu Mu-qing jbyamindi@yahoo.fr jihong@bupt.edu.cn wumuqing@bupt.edu.cn Institute, School of Information & Telecom. Engineering Beijing University of Posts and Telecommunications, Beijing, CHINA In this paper our main work focus is on feedback v Channel Quality Information (CQI) in order to optimize the MU-MIMO transmissions operation as important to further enhance system capacity. The rest of paper is organized as follows: the Signal Model and Transmission Schemes in Section II; System Proceeding of feedback information in section III; the PMI/CQI feedback in section IV; section V on Zero Forced Precoder Technology for feedback; section VI analyzes and provides simulation results and section VII is the conclusion. II. SIGNAL MODEL AND TRANSMISSION SCHEMES By the consideration of the signal model for the DL MIMO transmission, the receiver known as eNodeB is designed with a precoder which each terminal can use to get the information for its own data stream reception. The mobile stations are known as User Equipment (UE). eNodeB UE Abstract-- This paper represents recent research on Long Term Evolution-Advanced (LTE-A) technology which is a mobile communication system supported by Multiple-Input Multiple-Output (MIMO) antenna transmission based on Downlink transmission, whereby User Equipments (UEs) are scheduled for packet transmission via information such as channel quality indicator reports, rank indication and precoding matrices. The main focus of this research is on the feedback technique via channel quality information (CQI) in order to optimize the performance of Multi-User-MIMO (MU-MIMO) transmissions in both the ideal and the measured transmission channels, using a simulation application to compare the capacity of MU-MIMO performance and Single User (SU)-MISO. The simulation shows the advantage of MU-MIMO terms of optimization of the capacity of MU-MIMO performance due to feedback. This is that the best-companion PMI can be helpful with MU-MIMO precoding. The advantage of the Zero Forced Precoder technique via the feedback is that there is no interference from the other user at the receivers. Index Terms: Precoding, feedback, channel quantization, zero forced, scheduler and codebook. technology in I. INTRODUCTION M IMO (Multiple Input Multiple Output) technology has become the most popular candidate for multiple antennas used by the signal transmitter and receiver which are key technologies for the future of wireless communications systems. This technique is rapidly developing with the evolution of wireless technology due to the progress in system design for cellular communications in channel transmission such as channel estimation, feedback, quantization, precoding and to the large capacity gains in the network system. Recently, the 3rd Generation Partnership Project (3GPP in the Long Term Evolution-Advanced (LTE-A) system has operated with several MIMO modes in order to achieve high data rates and high system capacity [1]. Many researchers have been attracted to focus on this technology in [2-4] working extensively on the transmit precoding in MU- MIMO or Multi- Input Single-Output (MISO) applications in broadcast channels. Some researchers confirm that reasonable amount of channel feedback for each user in the next generation communication systems [5-7]. 978-1-4673-1356-8/12/$31.00 ©2012 IEEE 1 Fig 1: Configuration of MU-MIMO System Let us consider that N transmission antennas at the base station and K users each receive a signal and a frequency flat fading channel. The received signal can be expressed as = y Hx n ⎡ y 1 ⎢ = ⎢ M ⎢ y ⎣ ⎤ ⎥ ⎥ ⎥ ⎦ n , k + (1) ⎡ n 1 ⎢ = ⎢ M ⎢ n ⎣ k ⎤ ⎥ ⎥ ⎥ ⎦ , H = ⎡ H ⎢ M ⎢ ⎢ H ⎣ 1 k ⎤ ⎥ ⎥ ⎥ ⎦ ∈ C M N × Where y The linearization of precoded transmission signal can be given by 

Where K x = ∑ kW C ×∈ W s k k (2) ∈ and ks C × 1km = k 1 N m are respectively, the k km streams and the linear Precoding Matrix kW should be the linear decoding matrix data vector with for the k th user and for the k th user. The decoding output of the k th user is then given by ˆ s W H x n (3) k k H k kW is generally determined according to the = Where ( H + ) k chosen precoding design criterion. III. SYSTEM PROCESS OF FEEDBACK INFORMATION The eNodeB of this system receives k users’ data which constitute users groups; those users’ groups send the data information for scheduling to the schedule and multiplex section, after scheduling it can be precoded. The feedback information of the users to the eNodeB can be switched by the controller of the user group, schedule and multiplex and precoding. A. Transmit pilot with precoder At the beginning of each frame, eNodeB sends pilots to every UE through a broadcast channel multiplied by the precoder from the codebook; these pilots are used for channel estimation. B. Channel estimation for PMI and CQI The channel estimation for all antennas is made at a common resource element in each Resource Block (RB) based on Channel State Information-Reference Signals (CSI-RS) which can get information due to the Precoding Matrix Indicator (PMI) and rank selection and CQI measurement. The PMI and CQI are both computed and reported per sub- band. IV. PMI/CQI FEEDBACK 1) Multi-rank CQI feedback Fig 2: Configuration of MU-MIMO System 2 For MU-MIMO precoding, the rank restricted (rank-2) information (PMI/CQI/RI) is used. It is obvious that if RI is less than 3, multi-rank feedback rolls back and is equivalent to single-rank. The objective of multi-rank CQI feedback is to provide more accurate CQI corresponding to the most two significant Eigen modes when RI>2. However, is questionable whether scheduling high rank (e.g. RI>2) for MU-MIMO is beneficial. it 2) Best companion PMI/CQI feedback The Channel State Indictor (CSI) is more accurate at UE than eNB. It can be provided by UE and can be used to describe the null-space of the UE and this information could be helpful in generating MU-MIMO precoding at the eNB, so-called best-companion PMI. The corresponding CQI should be also reported associated with the best-companion PMI. If the number of UE is not large enough, it is possible that the selected best-companion PMI of one UE is not selected by the others. In this case, the effectiveness of the best-companion feedback is significantly reduced. interference the multiple access 3) Multiple CQI feedback based on multiple transmission Pairing UEs with orthogonal signaling space can greatly reduce (MAI) and simultaneously achieve good multiplexing and beam-forming gain. As the codebook uses feedback, the limitation of the number of codewords should be considered and the selected codeword (or PMI) of each UE can roughly represent the associated signaling space. Then the codewords which have large chordal distances from the selected codeword, can span the null-space of UE and are called best-companion PMI clusters. To achieve an ideal trade-off between MAI and beam-forming gain, it is reasonable to assume the PMI of paired UEs should belong to each others’ best-companion PMI cluster. With such an assumption, UE is capable of estimating the CQI based on MU-MIMO precoding to some extent, even though the information of paired UEs is unavailable. In this system, we assumed that the channel estimation is used for SINR calculation. 4) Calculate SINR and MCS The calculation of SINR in each transmission mode for the channel characteristic is operated by eNodeB. At the end of this calculation operation, eNodeB uses MCS to transmit the data rate to the SINR. To minimize the feedback in the transmitter, MCS is used for information and PF scheduler. 5) Feedback for best channel information In order to obtain the best channel information, the UE feedback due to the precoder will provide more information bits such as MCSes. This best channel information can be calculated through Transmit Modes between system capacity and feedback information rate. 

6) Scheduling for grouping Ues The eNodeB which is the central controller can estimate MU SNR from the reported MCS, one benefit of this technique is to correct the actual number of scheduled layers. When grouping G UEs together for MU transmission, the eNB estimates the SNR by mu ( rG (4) SNR SNR cqi = ) Where SNR (mu-cqi) is the SNR corresponding to the reported MU-CQI. The CQI feedback is utilized for both SU- MIMO and MU-MIMO for signal transmission process. 7) Transmitting and Receiving signal As the eNodeB transmits a signal to a selected user in a selected transmit mode, eNodeB can calculate error-rate and throughput for performance evaluation. The eNodeB updates each user’s PF value and comes to the next frame. Precoding is performed at the Physical Resource Block (PRB), in terms of timing and synchronization during the process, the received signal at the UE for the k th PRB can be represented by: H k E k X k ( ) (5) Where H(k) is the complex channel transmission, E(k) is the Precoding Matrix, X(k) is the transmit vector and N(k) is the Additive White Gaussian Noise (AWGN). Y k ( ) N K ( ) ( ) + = ( ) More specifically, it is assumed that the reported PMI of Aw and its best-companion PMI cluster, UE A is denoted by consisting of n codewords, is given as [ w ,L w . ] , 1A An Firstly, the signal to noise ratio (SINR) based on SU- 2 , where Hw A N MIMO hypothesis is obtained by SINR SU = H is the channel matrix of UE A, N represent the normalized AWGN and inter-cell interferences. With information of best- companion PMI cluster, the MU-MIMO based SINR can be obtained as 2 n ∑ 1 n Hu + 2 , u = } = = 1i MU N SINR , (6) ( )Ai wwf represents a generic MU-MIMO precoder A Hu Ai where MU-MIMO precoder{ u ( )•f generator. For example, if Zero Forced (ZF) precoder is assumed, ( ( ( , norm wwf HHH ]Ai , A wwH Where )1 ) −σ+ I ) = [ = and , . Ai Ai H A , A A 2 The CQI/SINR based on MU-MIMO can be reported in addition to SU-MIMO feedback to facilitate MU-MIMO CQI adjustment. It is noted that with a certain chordal distance threshold assumed, the best-companion PMI cluster can be implicitly known at both UE and eNodeB and, therefore, no feedback is required. For example, when the chordal distance threshold is set to one, the codewords, which are orthogonal to the selected PMI, are defined as the best-companion PMI cluster; the best-companion PMI cluster is pre-determined. V. ZERO FORCED PRECODER TECHNOLOGY FOR FEEDBACK AND THE OPTIMIZATION OF CAPACITY OF MU-MIMO In this section we will mainly focus on the fact that the advantage of the Zero Forced Precoder technique via feedback is that there is no interference from the other user at the receivers, so the received signal is free from multi-user interference. But in some cases of MU, interference can exist due to the channel quantization and feedback delay. The singular value decomposition (SVD) of the channel H = ∑ where the matrices U and V is given by of the are unitary; the SVD decouples the channel into orthogonal directions. H U V H The equivalent channel quantization for k th user is c e k , = ,e kH and computed as denoted by cσν H H H k k arg max Where c is a vector that belongs to the codebook, (7) kσ the first element of the main diagonal of ∑ and ν is denotes the first column of V. A. The ZF precoding technology The ZF precoding technology is very important in MU- MIMO potential precoder design and it can pre-cancel the interference at the transmitter. The ZF precoder can be designed using the Moore- Penrose Pseudo-inverse as H e T T T ,1 , ,2 = ⎡ =⎣ ⎡ = ⎣ e ..., H e H T e N , W W W T ( H H T T e e ,1, ,2, W H H H − ) 1 ⎤ H ⎦ and e (8) W ,..., Where The ZF precoder is obtained when the equivalent channel observed by different users are orthogonal to each other. This technique is key to MU-MIMO improvement between beam-forming gain and the required amount of feedback information [8]. ⎤ ⎦ T N , Additionally the linear precoding matrix applied with the Singular Value Decomposition method (SVD) and the Leakage of MMSE at eNodeB can be also considered. In the case of the LMMSE receiver, the k th user is given by (9) W W H H W H W T Iσ − ) 2 1 n H T k , H k = + ( ( ) H R k 2σ is the noise variance. T k Where SNR can be defined by MMSE receiver combiners and can be formulated as ⎛ g ⎜ ⎝ ⎞ ⎟ ⎠ (10) H t H t k k H t k k β )( k k = + − 1 I ( ) ( ) H H N k The combination of ZF precoder with Leakage of 3 

MMSE at eNodeB and (SVD) can reduce the interference and improve system performance. B. The optimization of the capacity MU-MIMO The optimization of the capacity MU-MIMO can be obtained by the codebook and feedback with index. The UE feeds back a Channel Quality Indicator (CQI) value for every matrix in the codebook, which gives more flexibility and accurate CQI information for scheduling. Based on the feedback, the scheduler which provides more accurate CQI information at eNodeB chooses the precoding matrix with the highest sum capacity and applies it to the PRB. The SINR for each data stream can be calculated as: SINR MMSE m = ′ ′ ε E k H k ( ) ( ) s × MN G MMSE k ( o ) The achievable data rate for PRB k is given by: r k log (1 2 + SINR m ) (12) rN ∑ = m This highest sum capacity and application to the PRB produce considerable improvements in error performance. VI. SIMULATION In this section, we will compare the performance of SU- MIMO/MU-MIMO through Channel Quality Indicator. In order to operate on the same simulation platform for the performance of both SU-MISO and MU-MIMO at high SNR, with the multiplexing gain of the MU-MIMO system is twofold while it is limited to one for the SU-MISO. We assume that for both the ideal and the measured channels, MU-MIMO yields a higher sum rate than SU-MISO, and all frames and all subcarriers are subsequently normalized to bps/Hz. TABLE SIMULATION PARAMETER Parameter Center Frequency Bandwidth eNB transmission power Number of UEs Number of subcarriers Downlink transmission scheme Downlink scheduler Feedback assumptions Value 1917.5 Mhz 4.8 Mhz 30dBm 2 160 Dynamic SU/MU-MIMO scheduling:MU-MIMO pairing: Max 2 users/RB; PF in time and frequency 5ms periodicity and 4ms delay; Sub-band CQI and PMI feedback without errors. We will compare the capacity of SU-MISO TDMA and MU-MIMO with two UEs as shown below: − 1 (11) 0 0 5 10 15 SNR(dB) 20 25 30 Comparison of SU-MISO and MU-MIMO capacitry SU-MISO ideal SU-MISO mesured MU-MIMO ZF ideal MU-MIMO ZF mesured 18 16 14 12 10 8 6 4 2 ) s p b ( t u p h g u o r h t Fig 3: Ergodic sum-rate capacity of SU-MISO TDMA and MU-MIMO with two UEs. In this result, since these correlation effects result in a rank-deficient channel matrix in quantized beam-forming, it shows that MU-MIMO with ZF filter can be used to the minimum-correlation precoder so MU-MIMO with ZF measured channel is much higher than in the single-user SU- MISO. In that case, MU-MIMO power normalization is performed. It is seen that the proposed optimized linear precoder offers a significant gain over the other schemes. The random beam-forming for MU-MIMO performance becomes a challenge, which is not surprising, since random beam-forming is only useful for a large number of users as it realizes its gain through scheduling. QLP for MU-MIMO with 128 quantization vectors 4x1 MU-MIMO 4x2 MU-MIMO 4x3 MU-MIMO 18 16 14 12 10 8 6 4 ) s p b ( t u p h g u o r h t 2 0 5 10 15 SNR(dB) 20 25 30 Fig 4: MU-MIMO with quantized beamforming, VII. CONCLUSION Our research on the feedback technique through channel quality information (CQI) is applicable due to its support from other technology such as best-companion PMI , the Zero Forced Precoder technique through the feedback, 4 

the channel quantization and feedback, the Singular Value Decomposition method (SVD) and the Leakage of Minimum Mean Square Error (MMSE) at eNodeB, which all contribute to optimize the capacity of MU-MIMO channel transmissions performance. The simulation results show clearly that the capacity of MU-MIMO channel transmissions performance using the ideal and the measured transmission channels compare to Single User (SU)-MISO is superior. Therefore MU-MIMO channel transmission is a beneficial technology the capacity of transmission performance using the feedback and there is no interference from the other user at the receivers. terms of optimizing in Our future work includes incorporating the proposed Pairing UEs with orthogonal signaling space systems. ACKNOWLEDGMENT Thanks to Mr Thomas Maj for editorial assistance on this paper, special thank to China Post Doctorate project for sponsor. REFERENCE [1] 3GPP TR 36.211 V8.1.0, “Evolved universal terrestrial radio access (EUTRA); physical channels and modulation,” http://www.3gpp.org, Nov.2007. [2] D. Bartolome, A. Pascual-Iserte and A. I. Perez-Neira, “Spatial scheduling algorithms for wireless systems,” P ICASSP’03, April 2003. [3] M. Fuchs, , G. Del Galdo, and M. Haardt, “Low-complexity space- timefrequency scheduling for MIMO systems with SDMA,” IEEE Tr. Vehic.Tech., Vol. 56, No. 5, pp. 2775-2784, Sept. 2005. [4] E. Viterbo and A. Hottinen, “Optimal user pairing for multiuser MIMO,” submitted [5] Philips  “Comparison between MU-MIMO codebook-based channel in 3GPP TSG RAN for LTE downlink,” reporting WG1#46bis/R1-062483. techniques [6] Samsung Electronics, “Downlink MIMO for EUTRA,” in 3GPP TSG [7] Samsung Electronics, “Downlink MIMO Schemes for IEEE802.16m,” in [8] Performance of Linear Multi-User MIMO Precoding in LTE System, 2008 RAN WG1 #44/R1-060335. IEEE C802.16m-08/285. IEEE BIOGRAGHIES Jean-Baptiste YAMINDI born in Central African Republic, received his B.E. degree in Math-Physics in 2001, Learned Chinese language from 2002-2003 at Beijing University of Language and Culture (BLCU), got M.E. degree in System of Information and Telecommunication and PhD degree in BUPT research interests area WiMAX Physical Layer from Beijing and Telecommunications(BUPT) in 2006, 2010 CHINA. 2011 Algorithm Conception Engineer the Downlink system for the Long Term Evolution- Advanced (LTE-A)technology, System Planning Division-Radio Access Technology Division, Potevio Institute of Technology Co. Ltd, Beijing, CHINA. University Posts of in of Posts Beijing University Wu Mu-Qing, born in July 1963, Ph.D, professor of and Telecommunications (BUPT), senior membership of China institute of communications. Now, he was interested and researching in Ad Hoc wireless network, UWB, high-speed network traffic control and performance analysis, GPS locating and services, teaching basic theories of telecom networks in and information Hong Ji (SM’09) received the B.S. degree in communications engineering and the M.S. and Ph. D degrees communications engineering from the Beijing university of Posts and Telecommunications (BUPT), Beijing, China, in 1989, 1992, and 2002, respectively. From June to December 2006, she was a Visiting Scholar with the University of British Columbia, Vancouver, BC, Canada. She is currently a Professor with BUPT. She is also engaged with national science research projects, including Hi-Tech Research and Development Program of China (863 program) and the National Natural Science Foundation of China. Her research interests include heterogeneous networks, peer-to-peer protocols, cognitive radio networks, and cooperative communication. Prof. Ji serves on the editorial boards of several journals, including the Wiley International Journal of Wireless Communications and Mobile Computing, the Wiley International Journal of Communication Systems. She has also served on the Technical Program Committee (TPC) of numerous conferences, e.g., as a TPC of the 2010/2011 IEEE International Conference on Communications, the 2010/2011 IEEE Global Communications Conference, the 2010 /2011 IEEE INFOCOM workshop, the 2007 IEEE VTC, and as a TPC Cochairs of the 2011 CHINACOM Workshops. 5