On Diversity Combining with Unknown Channel State Information and Unknown Noise Variance

We derive detection metrics for soft-output diversity combining for the case of imperfect channel state information at the receiver. We treat in particular the case when the noise variance at the receiver is unknown. We contrast conventional training-based methods to a detector based on the generalized likelihood-ratio (GLR) test paradigm. We study the performance of the detectors via EXIT chart analysis and via simulations of LDPC coded transmission over a fast Rayleigh fading channel. The results show that the GLR receivers can significantly outperform the conventional detectors.