Impact of modeling errors on the performance of MIMO receivers with APP and PIC detection

Turbo-encoded multiple-input multiple-output (MIMO) radio links have been recently proposed for high-speed downlink packet access (HSDPA) in UMTS, where the reuse of spreading codes across the transmitter antennas results in high levels of interference. A state-of-the-art receiver chain for such a link incorporates space-time channel equalization, despreading, prewhitening, and detection. In this paper, the impact of modeling errors at the equalizer output on the receiver performance is investigated. Both a precise model and a more pragmatic approximate model are considered. Degradations resulting from imperfect knowledge of the channel state information at the receiver are also evaluated. The a posteriori probability (APP) detector and its max-log variant as well as the multistage partial parallel interference canceller (MS-PPIC) are examined as detection candidates in 1/3-rate and 1/2-rate coded 4/spl times/4 MIMO links. It is shown that, surprisingly, the MS-PPIC can provide superior performance compared to max-log-APP.