I. Introduction
In Recent years, millimeter wave (mmWave) devices have become the key technology to enable the next generation of wireless communications. These devices operate on an under-utilized mmWave frequency spectrum that is much wider than classic microwave bands. The mmWave communications also bring in several new challenges, such as severe path losses in high frequency bands and limited scattering in both spatial and temporal channels [1]. To combat the severe signal degradation, one practical approach is to use large antenna arrays, also known as “massive MIMO” (multiple-input multiple-output) systems, since the dimension and the half-wavelength spacing of antennas are of the order of millimeters in mmWave frequency bands. Yet, the use of large number of transmit/receive antennas could lead to a crucial cost and power consumption in radio frequency (RF) chains, if the systems were implemented in the classic “fully digital” way—one RF chain per transmit/receive antenna. To reduce both cost and power, one common solution is to use a hybrid architecture, where a few RF amplifiers are shared by transmit (resp. receive) antennas to bridge the interface between a low-dimension digital baseband encoder (resp. decoder) and a high-dimension analog precoder (resp. combiner) consisting of phase shifters.