Performance Comparison of Multiuser Detectors with Channel Estimation for Flat Rayleigh Fading CDMA Channels

Since the conventional detector performs poorly in bandwidth-efficient Code Division Multiple Access (CDMA) systems, several multiuser detectors were proposed recently. In this paper, we compare performance of the decorrelator, the two-stage detector (2S), and the decision-feedback detector (DF) for the Rayleigh flat fading synchronous CDMA channel. First, assuming perfect channel estimation, we show that the 2S and the DF have much lower bit error rate (BER) than the decorrelator, and prove that the ideal DF has unity Asymptotic Multiuser Efficiency (AME). Furthermore, we evaluate the performance of these detectors in the presence of channel mismatch. We model the Rayleigh flat fading channel as the second order Auto Regressive (AR) process, and use the Kalman filter as the channel estimator. The lower bounds on the BER of the 2S and the DF are derived. The analytical results and the simulations show that the estimation error accumulated due to cancellation of other users limits the performance of the 2S and the DF. Therefore, in the presence of channel mismatch, the decorrelator offers comparable or even better performance than more complex decision-feedback and two-stage detectors.