Approximative matrix inversion (MIA) has been extensively researched with the aim of addressing the complexity issues encountered by Zero-forcing (ZF) precoder design in massive multiple-input multiple-output (MIMO) systems. The eigen-based MIA scheme with diagonal matrix splitting is presented recently where additional coefficients can be designed and optimized to improve MIA accuracy and bit-error-rate (BER) compared to the well-known Neumann series expansion (NS) or Newton iteration (NI) framework for Rayleigh flat fading channel, regardless the presence of transmit antenna correlation or not. In this paper, stair matrix splitting is proposed for the eigen-based MIA method for uncorrelated Rayleigh flat fading channel. Simulation result shows that, with additional information in the inverse of stair matrix, both the ZF precoder approximation precision and BER can be enhanced compared to the diagonal matrix splitting scenarios without sacrificing complexity.
2018 International Symposium on Intelligent Signal Processing and Communication Systems (ISPACS)