Massive MU-MIMO Downlink Training with Orthogonal Pilots Allocation

Abstract

This paper considers a multi-cell massive multiple-input multiple-output (MIMO) downlink system. To detect the signals transmitted from BS, users should have their channel information. Instead of assuming that each user can perfectly know the expected value of its effective channel gain, users use least-square (LS) estimation to estimate their effective channel gains through downlink training. However, the estimated effective channel gains will be polluted by interference from other cells due to the limited pilots, which will degrade the system performance if it is not treated well. To alleviate it, a novel pilots allocation scheme is proposed in which users who use the same uplink pilot sequence are allocated with mutual orthogonal downlink pilots. With this scheme, we prove for maximum-ratio transmission (MRT) and zero-forcing (ZF) precoding that the relative estimation error of the effective channel gain tends to zero when the number of BS antenna goes to infinity. Simulation results validate the efficiency of our proposal.