Low-Complexity Hybrid Beamforming for Massive MIMO Systems in Frequency-Selective Channels

Hybrid beamforming for frequency-selective channels is a challenging problem, as the phase shifters provide the same phase shift to all the subcarriers. The existing approaches solely rely on the channel's frequency response, and the hybrid beamformers maximize the average spectral efficiency over the whole frequency band. Compared to state-of-the-art, we show that substantial sum-rate gains can be achieved, both for rich and sparse scattering channels, by jointly exploiting the frequency- and time-domain characteristics of the massive multiple-input multiple-output (MIMO) channels. In our proposed approach, the radio frequency (RF) beamformer coherently combines the received symbols in the time domain and, thus, it concentrates the signal's power on a specific time sample. As a result, the RF beamformer flattens the frequency response of the 'effective' transmission channel and reduces its root-mean-square delay spread. Then, a baseband combiner mitigates the residual interference in the frequency domain. We present the closed-form expressions of the proposed beamformer and its performance by leveraging the favorable propagation condition of massive MIMO channels, and we prove that our proposed scheme can achieve the performance of fully digital zero-forcing when the number of employed phases shifter networks is twice the resolvable multipath components in the time domain.