High speed near-ML MIMO detection for the next generation wireless systems

The Multiple-Input-Multiple-Output (MIMO) has become an attractive technique due to its capability of providing high data rates and link reliability. For next generation wireless systems, for example, WiMAX, LTE and other Orthogonal Frequency Division Multiplexing (OFDM) based communication schemes, system performance greatly depends on the quality detection. Optimal solution for hard decision detection in MIMO wireless systems is the Maximum Likelihood (ML) detector. However, no matter how superior the ML is, its complexity is what limits the implementation of the system with high antenna count. Sphere Decoding (SD) is widely considered the most promising approach to obtain optimal ML performance. The landscape of possibilities for architecture and algorithm optimizations is very rich if the SD method is used. This paper proposes efficient Very Large Scale Integration (VLSI) implementation of the near-ML sphere detector and channel matrix preprocessor for MIMO 802.16e broadband wireless systems. The implementation results show that close to ML performance, high detection throughput, optimal Bit Error Rate (BER) performance and reasonable complexity are achievable.