Multiuser Hybrid Analog/Digital Beamforming for Massive MIMO

We propose a generalized architecture for the phase shifter network in hybrid analog/digital beamforming (BF), which can provide flexible tradeoff between the performance and implementation complexity. The proposed architecture has a multi-layer structure. By changing the layer number, we can adjust the complexity of the phase shifter network, as well as the achievable performance. A two-stage BF algorithm is adopted, in which the analog and digital BF matrices are designed separately. The analog BF is adaptive to the second-order statistics of the channel. A simple iterative analog BF algorithm is proposed to maximize the average achievable rate of the effective channel after analog BF. The digital BF is adaptive to the effective channel (including the physical channel and analog BF matrix) for multiuser interference cancellation. Based on the proposed two-stage BF algorithm, we analyse the performance gap between the hybrid and full digital BF, as well as the impact of the layer number on the performance of hybrid BF. Simulation results show that, with the proposed hybrid BF algorithm, the hybrid BF technique can perform closely to the full digital BF. Further, the simulation results also demonstrate the tradeoff flexibility between the performance and complexity of the proposed hybrid structure. For example, in a 64-antenna system, the proposed technique performs very close to that of the full digital BF with only 16 RF chains and a fully connected-phase shifter network compared with 64 RF chains for the full digital BF case. For the same system, we still can achieve 70%-80% performance of that of the full digital BF with 16 RF chains and simple phase shifter networks composed of 16 1-by-4 sub-networks. The second one has much lower implementation complexity. Therefore, the proposed generalized structure enables us to select the proper performance and implementation complexity according to the requirement and capability of real systems.