Beamforming with no instantaneous feedback for mmWave transmission

Abstract

In this paper we consider statistical beamforming for millimeter wave (mmWave) channels. For a commonly used geometry channel model of uniform planar array (UPA), we show that the average channel Gram matrix can be approximated in a closed form, which allows the statistical beamformer to be computed directly. The sparse nature of the mmWave channel lends itself to statistical beamforming as the optimal beamformer that uses full channel information is typically steered in some spatial frequency determined by the statistics. Simulations are given to show that the performance of the proposed statistical beamformer that needs no instantaneous feedback is comparable to a limited feedback system with 1/8 feedback bits per antenna.