Sparse Channel Estimation Using Multiple DFT Matrices for Massive MIMO Systems

Abstract

Massive multiple-input multiple-output (MIMO) is a promising technology for 5G systems and is expected to improve the performance of multi-user MIMO (MU-MIMO). When uplink (UL) channel estimation results are used for downlink (DL) MU-MIMO precoding, the UL channel estimation errors degrade the performance of DL MU-MIMO. To reduce the estimation errors, a channel estimation method using the channel sparsity in beam space has been studied. This method, which is called beam space channel estimation (BSCE) in this paper, can reduce the estimation errors by setting the channel estimates of non-dominant beams to zeros. However, when the directions of beams are not close to those of dominant paths, BSCE cannot reduce the estimation errors sufficiently. In this paper, we propose a BSCE using multiple discrete Fourier transform (DFT) matrices which form beams in mutually different directions to increase the probability that the directions of beams are close to those of dominant paths. We also propose a non-zero beam selection method to prevent the directions of nonzero beams from being limited to a specific angular range. Simulation results show that the proposed method using four DFT matrices improves cell throughput performance by 53% compared with not using BSCE and by 24% compared with the conventional BSCE when the signal-to-noise ratio (SNR) of the UL reference signal is 0 dB.