An improved matrix inversion approximation method for massive MIMO systems with transmit antenna correlation

Abstract

Zero-forcing (ZF) precoding is known to achieve near-optimal performance at high signal-to-noise-ratio (SNR) in a massive MIMO system. While ZF precoding can be implemented by exact matrix inversion, this scheme is often prohibited due to stability and complexity issues. Hence a few matrix inverse approximation (MIA) methods were introduced to address these concerns. An eigen-based algorithm was recently proposed to enhance MIA accuracy for uncorrelated Rayleigh fading channel by estimating asymptotic eigenvalues of a Wishart matrix following the random matrix theory. In this paper, the eigen-based MIA concept is extended to the one-sided Kronecker model, (i.e., Rayleigh fading channel with transmit antenna correlation). Asymptotic eigenvalue probability density function of the non-commutative Kronecker channel is constructed from free probability theory and applied in eigen-based MIA. Simulation shows that MIA is degraded by transmit antenna correlation. However, compared to other Neumann expansion based methods, the proposed eigen-based MIA precoding scheme is still less susceptible to transmit antenna correlation especially for low to mid correlation values.