HYBRID BEAMFORMING FOR MASSIVE MIMO ANTENNA STRUCTURE

Abstract

Massive multiple-input multiple-output (MIMO) systems, in which a base station is equipped with tens to hundreds of antenna elements, are a cornerstone technology for meeting the throughput, reliability and spectral-efficiency demands of fifth-generation (5G) and beyond5G wireless networks. A fully digital implementation, in which every antenna element is driven by an independent radio-frequency (RF) chain, achieves the best attainable performance but is impractical at millimeter-wave (mmWave) and sub-6 GHz massive-MIMO scale because of the prohibitive hardware cost, circuit-board area and power consumption of large numbers of RF chains, mixers, analog-to-digital converters (ADCs) and power amplifiers. This paper presents a hybrid analog–digital beamforming architecture for a massive MIMO antenna structure that combines a low-dimensional digital baseband precoder with a large-dimensional analog RF precoder realised through a network of phase shifters, so that a small number of RF chains can drive a much larger number of antenna elements. The existing fully digital and pure analog beamforming schemes are reviewed and their architectural limitations are analysed. A fullyconnected hybrid precoding architecture is then proposed together with an orthogonal-matchingpursuit based precoder design that exploits the sparse, low-rank structure of massive-MIMO / mmWave channels to approach the performance of the fully digital solution while using an order-of-magnitude fewer RF chains. System-level simulation results show that the proposed hybrid architecture attains spectral efficiency within 1–2 bits/s/Hz of the fully digital upper bound across the practical SNR range of −10 dB to 20 dB, while reducing the number of active RF chains, and hence hardware cost and power consumption, by more than 60%. The results confirm that hybrid beamforming is an effective and hardware-efficient candidate for realising massive MIMO antenna structures in next-generation wireless systems.