Beamspace Local LMMSE: An Efficient Digital Backend for mmWave Massive MIMO

We explore an all-digital architecture for a mmWave massive MIMO cellular uplink in which the number of users scales with the number of antenna elements at the base station. We consider the design of multiuser detection strategies after a spatial DFT, which concentrates the energy of each user onto a few DFT bins in “beamspace.” In this paper, we propose and investigate a local LMMSE receiver that exploits this property, using a small window in beamspace to demodulate each user. The proposed architecture is computationally efficient: the required window size depends on load factor (the number of users divided by the number of antenna elements) and does not scale with the number of elements. We also show that adaptive implementations of such local LMMSE receivers naturally extend to provide implicit channel estimation.