A Comparison of MIMO Techniques in Downlink Millimeter Wave Cellular Networks with Hybrid Beamforming

TL;DR

This paper compares hybrid beamforming MIMO techniques in mmWave cellular networks, analyzing coverage, rate, and power tradeoffs through a stochastic geometry model, and finds MU-MIMO generally outperforms other methods under ideal conditions.

Contribution

It introduces a stochastic geometry model for analyzing hybrid beamforming MIMO techniques in mmWave networks, considering hardware constraints and blockage effects, and evaluates their performance tradeoffs.

Findings

MU-MIMO generally outperforms SM and SU-BF with perfect CSI and round robin scheduling.

Coverage and rate depend on hardware constraints and channel conditions.

Overheads in channel acquisition affect the practical advantage of MU-MIMO.

Abstract

Large antenna arrays will be needed in future millimeter wave (mmWave) cellular networks, enabling a large number of different possible antenna architectures and multiple-input multiple-output (MIMO) techniques. It is still unclear which MIMO technique is most desirable as a function of different network parameters. This paper, therefore, compares the coverage and rate performance of hybrid beamforming enabled multi-user (MU) MIMO and single-user spatial multiplexing (SM) with single-user analog beamforming (SU-BF). A stochastic geometry model for coverage and rate analysis is proposed for MU-MIMO mmWave cellular networks, taking into account important mmWave-specific hardware constraints for hybrid analog/digital precoders and combiners, and a blockage-dependent channel model which is sparse in angular domain. The analytical results highlight the coverage, rate and power consumption tradeoffs in multiuser mmWave networks. With perfect channel state information at the transmitter and round robin scheduling, MU-MIMO is usually a better choice than SM or SU-BF in mmWave cellular networks. This observation, however, neglects any overhead due to channel acquisition or computational complexity. Incorporating the impact of such overheads, our results can be re-interpreted so as to quantify the minimum allowable efficiency of MU-MIMO to provide higher rates than SM or SU-BF.