A reduced quantization range feedback method for time-varying MIMO systems

Abstract

MIMO systems benefit from spatial multiplexing when channel state information (CSI) is available at the transmitter. Givens decomposition of unitary matrices can be used to feedback the eigenvectors of the channel in an efficient way. Several Givens parametrization based quantization methods had been suggested in recent years. In this paper, a simple feedback method that exploits the temporal correlation of the channel is proposed for slowly time-varying MIMO channels based on cyclic Jacobi tracking. The proposed method applies only one Jacobi sweep thus avoiding the need for Givens decomposition. Jacobi solutions are then combined with a novel sign shifting method that reduces the quantization range of one of the parameters by 70%. For uniform quantization and an N × M MIMO channel, the proposed method saves approximately M 2 − M feedback bits as compared to the state of the art Givens decomposition based methods without a significant decrease in performance. This is shown by a simulation on 4 × 4 time-varying MIMO channel. It is observed that the capacity penalty of using only one Jacobi sweep is insignificant for highly correlated channels.

Linear prediction of time-varying MIMO systems using Givens rotations

“…A partial sweep of a Jacobi eigenvalue decomposition algorithm is suggested in [9] for the purpose of limited feedback. A new scheme for reducing feedback by using a single sweep Jacobi algorithm in slowly varying MIMO channels is also proposed in [10].…”