Precoding for Multiuser Orthogonal Space-Time Block-Coded OFDM Downlink over Spatially-Correlated Channels

This paper considers precoding for closed-loop multiuser orthogonal space-time block-coded (OSTBC) orthogonal frequency-division multiplexing (OFDM) downlink over multiple-input multiple-output (MIMO) channels. The authors present a general transmit-antenna-correlated, frequency-selective fading MIMO channel model with imperfect channel estimates and feedback delay, and derive the conditional means of the channel response. Exploiting the channel statistics, the authors develop new linear precoding and non-linear Tomlinson-Harashima precoding (THP) to maximize the signal-to-noise power ratio (SNR). By considering the conditional mean as the equivalent channel matrix, our precoding takes into account the estimation errors and channel time variations over feedback delay. The authors confirm the intuition that when channel state information at the transmitter (CSIT) becomes accurate, the full-CSIT precoder outperforms the covariance-feedback precoder. The proposed precoder has a bit error rate (BER) gain over unprecoded systems. As well, non-linear THP is shown to outperform linear precoding.