Parameter Estimation for Block Diagonalization Based Hybrid Beamforming in Massive MIMO Communications

Massive multiple-input multiple-output (MIMO) will be employed in the 5th generation mobile communications (5G). To implement the massive MIMO, hybrid beamforming (HB) is one of the most promising techniques, because HB is composed of the analog beamforming (AB) and the digital beamforming (DB) and thus can decrease the number of baseband and radio frequency (RF) circuits significantly. Furthermore, HB can reduce a much larger amount of cost and power consumption than the full DB (FDB) configuration, which requires one baseband and RF circuit for each antenna element. This paper proposes a scheme to estimate both phases for phase rotators in AB and a precoding matrix for DB so that HB can approximate the block diagonalization (BD), which can nullify interference from the undesired users and form efficient transmit beams. For time division duplex (TDD) multiuser MIMO communications, the proposed scheme estimates the FDB channel matrix in the uplink (UL) and the phase for the phase rotators of AB in the downlink (DL). After obtaining an equivalent channel matrix from both the estimated FDB channel matrix and the AB precoding matrix, the scheme estimates the DB precoding matrix in DL from the equivalent channel matrix so that HB can approximate BD in DL. Computer simulations demonstrate that HB employing the proposed scheme can achieve DL average bit error rate (BER) performances that are little inferior to those of FDB.