Design of multi-user hybrid beamforming systems in millimeter wave channel

Abstract

Millimeter wave (mmWave) signals experience a significant path-loss in free space. To overcome this weakness, a large number of antennas are needed to obtain a high beamforming gain. Although a large number of antennas can be implemented in small area due to the short wavelength, the digital beamforming techniques cannot be implemented easily due to the high complexity of hardware. To solve this problem, the hybrid beamforming systems which have smaller number of radio frequency (RF) chains are proposed in the literature. Although the hybrid beamforming systems may achieve the spectral efficiencies of the digital beamforming systems closely, the spectral efficiency cannot be monotonically increased along with the number of data streams even in the case of digital beamformer due to the limited scattering in mmWave channel. This means that the number of spatial degrees of freedom (DOF) in mmWave channel is much smaller than that in rich scattering channel. Therefore, we need only small number of RF chains of hybrid beamformer which can support the spatial DOF of the channel. In this paper, we assume the channel as the clustered channel model which can describe the limited scattering environment. We calculate the theoretical number of spatial DOF of the channel which can be considered as the maximum number of data streams that can be simultaneously transmitted. Based on the spatial DOF, we find the smallest number of RF chains in hybrid beamformer to approach the performance of digital beamformer. In addition, we propose a multi-user precoding method for sparse channel that can be applied to both digital beamformer and hybrid beamformer. The proposed precoder can be constructed basically by the block diagonalization (BD). However, the original BD precoding method cannot make the null space of each user large enough to obtain the full multiplexing gain in sparse channel. Due to this fact, multi-user hybrid beamforming systems which use the original BD method have ...