Performance analysis of downlink MMSE beamforming training in TDD MU-massive-MIMO

Abstract

We consider a time-division duplexing multi-user massive multiple-input multiple-output system, where a so-called beamforming training (BFT) occurs prior to the downlink (DL) data transmission. The BFT process helps the users to obtain the estimates of effective channels, and thus facilitates a more reliable DL data detection compared to the conventional case, where the users have only the statistic channel knowledge. We first derive the closed-form lower bounds on the achievable DL data rates in the conventional and BFT schemes, using the minimum mean-square-error (MMSE) precoder. This enables us to further analyze the spectral efficiency, which takes the loss due to the training overhead into account. Moreover, we also consider a realistic scenario where the channel aging is present: it may destroy the training gain and lead to an efficiency loss. The effects of channel aging in both schemes are evaluated. In simulations, the performance of MMSE precoder is compared to the maximum-ratio transmission and zero-forcing precoding methods for both schemes in the static and aging channels.