Performance of multiple space-time coded MIMO in spatially correlated channels

In Jung-Tao Liu (November 2002), it is shown that multiple space-time coded streams can be transmitted simultaneously by labeling the transmit antennas into different transmission groups. Space-time codes are applied to the antennas within the same groups. A receiver structure using array processing and maximum likelihood (ML) detector was proposed. A multi-code wideband CDMA downlink high-speed packet data channel was simulated assuming the Alamouti code (S.M. Alamouti, October 1998) on all antenna groups. The results show performance gains over the VBLAST at data rates above 10 Mbps for both MMSE and zero-forcing optimized array processors. In this work, we further study the performance of the same scheme in the spatially correlated channels. Performance analysis shows that due to the built-in diversity of the space-time codes, the proposed scheme is robust even in the highly spatially correlated environments. Simulation results are provided only for MMSE under channels with different correlations. It is shown that when transmitting at 14.4 Mbps and at one percent packet error rate, the proposed scheme is 1.5dB worse than the full-blown maximum-likelihood detector in an i.i.d. (independent identically distributed) flat fading channel, and is about 5dB worse in a channel with medium-to-high correlations; while the VBLAST is about 10 dB worse than ML in i.i.d. and is about 17dB and at least 27dB worse than the ML in channels with medium and high correlations, respectively. The proposed scheme strikes a good balance between the complexity and the performance. It should be considered whenever the layered space-time architectures become too cumbersome to implement. Due to its robustness under correlations, it is also recommended for systems that cannot meet the i.i.d. fading assumption that has been imposed by many multi-input, multi-output transceiver algorithms.