On the Convergence and Performance of MF Precoding in Distributed Massive MU-MIMO Systems

TL;DR

This paper investigates how distributing antennas into clusters affects the convergence rate and performance of matched-filter precoding in massive MU-MIMO systems, especially under spatial correlation and imperfect CSI.

Contribution

It derives a SINR limit for MF precoding in large systems and analyzes the impact of antenna distribution and correlation on system performance.

Findings

Distributed antennas improve performance under spatial correlation.

Co-located antennas perform better in uncorrelated scenarios.

Performance depends heavily on spatial correlation and antenna distribution.

Abstract

In this paper, we analyze both the rate of convergence and the performance of a matched-filter (MF) precoder in a massive multi-user (MU) multiple-input-multiple-output (MIMO) system, with the aim of determining the impact of distributing the transmit antennas into multiple clusters. We consider cases of transmit spatial correlation, unequal link gains and imperfect channel state information (CSI). Furthermore, we derive a MF signal-to-interference-plus-noise-ratio (SINR) limit as both the number of transmit antennas and the number of users tend to infinity. In our results, we show that both the rate of convergence and performance is strongly dependent on spatial correlation. In the presence of spatial correlation, distributing the antennas into multiple clusters renders significant gains over a co-located antenna array scenario. In uncorrelated scenarios, a co-located antenna cluster has a marginally better mean per-user SINR performance due to its superior single-user signal-to-noise-ratio (SNR) regime, i.e., when a user is close to the base station (BS), the links between the user and all transmit antennas becomes strong.