Hybrid Training Signal Design for Multiuser OFDM Channel Estimation in mmWave Bands

Hybrid training signal design is conceived for the estimation of multi-user frequency-selective mmWave channels, where hybrid architectures are adopted due to the limited number of available radio frequency (RF) chains. In this setting, the training signal has a hybrid structure, expressed as the product of a high-dimensional, low-resolution analog beamformer (ABF) and low-dimensional subcarrier-wise digital pilots (SDPs). All pilot subcarriers share the same analog beamformer. Consequently, the joint design of ABF and SDPs to minimize the minimum mean square error (MMSE) of channel estimation leads to a challenging high-dimensional mixed continuous-discrete optimization problem. The main contribution is the development of a new path-following algorithm that leverages closed-form updates of scalable complexity for both ABF and SDPs. We further examine the impact of channel estimation accuracy on the capacity of multi-user wideband mmWave communication systems.