Computational Complexity Reduction of Receive Weight Matrices in Massive MIMO SC-FDF†

Abstract

MIMO single-carrier frequency domain equalization (SC-FDE) can achieve high data-rate uplink transmission. We need, however, weight matrices for all the frequency points which separate spatially multiplexed signals and equalize channel distortion. In this paper, we deal with zero-forcing (ZF) matrices. To obtain them, we need matrix inverse calculations, and the computational complexity is heavy when the number of frequency points is large. It is possible to reduce the amount of calculations using an interpolation technique. We calculate ZF weight matrices for a part of frequency points, and for the remaining ones, we obtain the matrices by interpolation. This technique has been investigated for MIMO OFDM systems so far. In this paper, we consider computational complexity reduction for massive MIMO SC-FDE. We examine bit error rate (BER) performance of the linear and DFT interpolations. It is shown that the interpolation techniques achieve satisfactory BER with lower computational complexity. Also, we show that the ZF matrix is approximated by a maximum ratio combining (MRC) one in a massive MIMO case, and that the MRC matrix requires almost no computations and the degradation of the BER performance is small.