Realistic Rician fading channel based optimal linear MIMO precoding evaluation

For multiple-input/multiple-output (MIMO)-based wireless communications systems with precoding, we evaluate the symbol-detection performance for Rician fading. We first consider, as in conventional evaluation approaches, fixed channel statistics, i.e., typical settings for the azimuth spread (AS), Rician K-factor, and deterministic channel matrix component H_d with set rank. Then, to predict performance more realistically, we adopt WINNER II models for AS and K (i.e., lognormally-distributed, correlated, scenario-dependent), and generate H_d based on randomly-selected transmit and receive antenna geometries. Based on our recent work, we consider that the channel matrix, i.e., the instantaneous channel state information (CSI), is available only as an inaccurate estimate at the transmit-side, for precoding. Conventional precoding simply replaces the true instantaneous CSI with its estimate in precoder matrix expressions. Our approach has been to rewrite the true instantaneous CSI in terms of the estimated instantaneous CSI and an error matrix whose statistics depend on the channel statistical CSI, and use this reformulation for an optimal precoding approach. For the numerical results, both instantaneous CSI and statistical CSI are estimated from pilot symbols, unlike in previous work, which generally assumes perfect statistical CSI. Our evaluation reveals the substantial dependence of precoding performance on the various assumptions about AS and K, and H_d. Furthermore, we confirm that the new precoding approach can significantly improve performance.