Reduced-complexity SISO equalization for Rayleigh fading channels with known statistics

For Rayleigh fading channels with known covariance matrix, reduced-complexity receivers can be derived which approximate the maximum likelihood (ML) receiver by defining additional constraints. The reduced-complexity receiver, which assumes that the current channel output depends on finite previous channel outputs, is termed forced-folding sequence estimator (FSE). We propose another approach, where a long burst is divided into overlapped subbursts and the computations are performed over these short subbursts. This receiver is referred to as overlapped subburst sequence estimator (OSE). We compare the FSE and the OSE and derive their soft-input soft-output counterparts, which are termed as SISO-FSE and SISO-OSE. From the viewpoint of linear estimation, the minimum mean-square error (MMSE) filtering/smoothing approaches are verified to be unsuitable for slowly fading channels in the context of trellis-based equalization. For symmetrical signal constellations, differential encoding is necessary to resolve the phase ambiguity. We show how to exploit the symmetry during the evaluation of branch/path metrics to obtain a reduced-complexity structure without loss of performance.