Asymptotic Performance of BI-GFDE and Unconditional MMSE-SIC Receivers for Large MIMO Systems

This paper studies the asymptotic performance of signal detection for large random multiple-input multiple-output (MIMO) systems. Two iterative receivers are considered: unconditional MMSE receiver with soft interference cancellation (U-MMSE-SIC) and block-iterative generalized decision feedback equalizer (BI-GDFE). We prove that for the limiting case, the output SINRs at each iteration for both receivers converge in probability to their respective deterministic limits. We also show that for U-MMSE-SIC, the limiting SINR is determined by the statistics of the soft estimates; while for BI-GDFE, the limit is determined by the statistical reliability of the hard decisions. Numerical results have shown that the U-MMSE-SIC receiver has faster convergence than the BI-GDFE receiver with the price of higher computational complexity in each iteration due to the use of soft interference cancellation.