An Efficient Precoding Method for Improved Downlink Massive MIMO System

This paper proposes an efficient precoding method based on Neumann series (NS) to improve error performances for already developed diagonal NS (DNS) in massive MIMO system. The conventional DNS has very low complexity compared to a zero-forcing (ZF) and any other NS based ZF precoding methods such as tri-diagonal NS (TNS) and eigenvalue based NS. However, the conventional DNS has very poor error performance in highly correlated massive MIMO system and performance degradations are severe as the used modulation is large. For an efficient precoding, diagonal plus columns NS (DCNS)-1, DCNS-k, and hybrid DCNS-1 and DCNS-U are proposed, and these precoding schemes are adaptively used according to the number of total diagonal dominant active users. The DCNS-1 is applied to fully diagonal dominant users and DCNS-k is applied to non diagonal dominant users, and finally, hybrid DCNS-1 and DCNS-U is applied to partially diagonal dominant systems. For improved DCNS-k, k active users which cause the largest post interference power are selected by simple calculations. The simulation results show that bit error rate (BER) performance for the proposed scheme is very higher than the conventional DNS and is nearly approximated to the BER performance for the conventional optimal ZF with very low complexity.