Low complexity massive MIMO detection architecture based on Neumann method

Massive (or large-scale) multi-input multi-output (MIMO) technology becomes one of the most promising concept in 5th generation wireless system. In the uplink of a massive MIMO system, large dimensional channel matrix inversion becomes the most challenging issue of hardware implementation. In the previous work, Neumann series based matrix inversion method (Neumann method) was proposed as an approximated inversion method. However, the conventional Neumann method shows a considerable performance degradation compared to the exact matrix inversion. In this paper, we propose a 3-order Neumann method based architecture to achieve a reasonable tradeoff between detection performance and hardware complexity. In the proposed approach, only numerically dominant (diagonal and near diagonal) elements of approximated inversion matrix are judiciously used for a low complexity matrix inversion. The proposed matrix inversion hardware was implemented using 65-nm CMOS process. The experimental results show that our matrix inversion hardware shows 45% area reduction while maintaining a reasonable BER performance.