Efficient Beamspace Downlink Precoding for mmWave Massive MIMO
We investigate efficient downlink precoding for all-digital downlink mmWave massive MIMO, with the number of users scaling with the number of antennas. The iterative computations required for optimal linear precoding are a severe bottleneck as the number of antennas increases, with the computational complexity per iteration scaling cubically with the number of antennas. In this paper, we propose a near-optimal linear precoding algorithm that exploits the sparsity of mmWave channels, employing a beamspace decomposition which limits the spatial channel seen by each user to a small window which does not scale with the number of antennas. This drastically reduces the complexity of computing the precoder, with complexity per iteration scaling linearly with the number of users, and makes it feasible to scale the system up to hundreds of antennas as considered in this paper.