IMPROVED PERFORMANCE OF THE V-BLAST SYSTEM BY MODIFYING THE CHANNEL MATRIX Trung Q. Duong, Een K. Hong, Sung Y. Lee School of Electronics and Information – Kyung Hee University 449-701 Suwon, Republic of Korea dquangtrung79@yahoo.com Abstract - MIMO system (multiple antennas at the transmitter and receiver) is capable of very high theoretical capacities. As an important space-time code, V-BLAST (Vertical-Bell Lab Layered Space-Time) code has been researched recently. The critical research topics of V-BLAST system are to reduce the complexity and to increase the system performance. In this paper, we investigate the effect of modified channel matrix on the system performance. Two nulling criterions Zero-Forcing (ZF) and Minimum Mean Square Error (MMSE) are analyzed. We also perform simulation to verify the analysis. Index Terms – V-BLAST, MIMO, Zero-Forcing and MMSE detection, wireless communication. I. INTRODUCTION It has been shown in recent research that the deployment of multiple antennas on both sides of a transmitter and receiver provides a larger capacity increase compared to single antennas systems [1], [2], [3] ⋅ A multiple-input multiple- output (MIMO) system that employs this trend is the V- BLAST (Vertical Bell Labs Layered Space-Time) architecture proposed in [3]. The structure is designed as a vertically layered coding, where independent code streams (called layers) are assigned with a certain transmit antenna. At the receiver, one way to execute the detection for this system is to use conventional adaptive antenna array (AAA) techniques [3], i.e. linear combination nullling. Nulling is carried out by linearly weighting the received signals in order to meet some relevant performance standard, such as zero-forcing (ZF) or minimum mean square error (MMSE). Zero-forcing was first proposed in [1]. However due to the limitation of pseudo inverse matrix computation when the number of antennas increases, zero-forcing seemed not feasible for real time implementation. To remarkably reduce the computational complexity, a very efficient method utilizing the QR and optimal detection order QR decomposition (called ZF-SQRD) of the channel matrix was proposed in [4]. The problem of noise enhancement through zero-forcing and QR decomposition has been paid attention ⋅ This research was supported by the MIC(Ministry of Information and Communication), Korea, under the ITRC(Information Technology Research Center) support program supervised by the IITA(Institute of Information Technology Assessment) to. A significant improvement can clearly be seen by including the noise term in the design of the linear weighting vector. This can be done by MMSE detection schemes, where we can trade off between noise and interference proposed in [4] and by introducing a lower complexity in [7]. An extension of the ZF-SQRD algorithm to the MMSE solution called MMSE-SQRD was introduced in [6]. In this paper, we propose a modification of the channel matrix in MMSE criterion. After detecting the sub-stream, channel matrix will be deflated by zeroing the corresponding column. This is because in MMSE the received vector is not nulled completely. In every detection step, there still remains the interference. In order to minimize as much interference as possible the corresponding channel matrix must be modified, i.e. corresponding column is zeroed, so that the detected sub-stream will not appear in the receive vector. Otherwise, wrongly detected sub-stream will increase the interference for the next step. The remainder of this paper is organized as follows. In Section II, the system overview is introduced. In section III, ZF and MMSE are reviewed. The effect of the modification of channel matrix is investigated in Section IV. The results are compared in Section V and concluding remark can be seen in Section VI. II. SYSTEM OVERVIEW The system is considered with T n transmit and R T n n ≥ receive antennas. The data is demultiplexed in T n data sub-streams of equal length (called layers). These sub-streams are mapped into M-PSK or M-QAM symbols 1 2 , ,..., T n tt t and simultaneously transmitted over T n antennas. Furthermore, we can use a forward error correction code to encode the data sub-streams before mapping. However, it is not addressed in this paper. We will investigate the application under assumption uncoded symbols. In order to outline the V-BLAST system, one time slot of the time-discrete complex base band model is