Fractional Spaced Channel Estimation and Shortening for Joint Delayed Decision Feedback Sequence Estimation

A single antenna GSM/EDGE receiver design with fractionally spaced (FS) channel estimation and shortening is presented using trellis based joint detection for co-channel interference cancellation. For the combination of joint trellis equalization and DFE, we propose FS prefiltering to shorten both the impulse responses of a desired user and a dominant co-channel interferer jointly, followed by an optional whitening filter of low order to mitigate the coloration induced by the shortening prefilter. For the whitening filter, an auto-regressive (AR) model is assumed. We compare two optimization criteria: the maximum shortening SNR (MSSNR) and a novel criterion related to the minimum mean square error (MMSE) criterion, henceforth called MMSEr, by simulating the resulting SSNRs, the BERs, and assessing their numerical stability. In addition, we compare fractional with symbol spaced prefiltering. We address the problem of detecting shifted users and show results in mixed modulation environments. The channel estimators based on joint least squares (JLS) and maximum likelihood (ML) are compared to the Cramer-Rao bound (CRB) for FS processing. Finally, computational issues are discussed