The optimal diversity combining technique is investigated for multipath Rayleigh and Ricean fading channel with additive white Gaussian noise where only imperfect channel knowledge is available at the receiver. The non-observable estimation error contributes as an additive source of noise which is not white. Therefore, the optimal combining weight is derived taking into consideration the imperfect channel knowledge.

The bit error rate for BPSK modulation over correlated Rayleigh and Ricean fading channel is derived for minimum mean square channel estimation using pilot symbol assisted modulation. Analytical result and Monte-Carlo simulation are presented for specific channel and estimation models to demonstrate the effect of diversity combining with imperfect channel estimation on error performance in comparison with the case when perfect channel knowledge is available at the receiver. The trade-off between the channel estimation accuracy and the effective bit SNR is also discussed. The Pilot-to-Data power ratio is studied for different Rice K factors for optimizing the bit error performance.