Nonparametric channel identification and multichannel equalization for narrowband digital communications

We previously presented a novel multichannel linear equalization scheme for a TDMA based cellular/PCS communication system commensurate with the US IS-136 standard. This scheme assumes at least two different channels with oversampling at each channel. For a given channel, it was shown that the frequency response of the associated RF multipath channel may be well approximated over the raised cosine spectral bandwidth by a small number of sampled-spaced (not symbol-spaced) filter coefficients encompassing the multipath delay spread. As a result, multichannel (zero forcing) linear equalization may be effected via a small number of sample-spaced taps at each channel encompassing the multipath delay spread which, in a typical urban cellular environment, is a fraction of the "mainlobe" of the square-root spectral raised cosine pulse employed in the IS-136 standard. This paper shows how a decision feedback equalizer (DFE) may be constructed from the small number of sampled-spaced filter coefficients used to match the frequency response of the RF multipath channel associated with each channel. Simulations involving two beam channels are presented comparing the performance of the sampled-spaced multichannel linear equalizer with the resulting DFE. Averaging over an ensemble of multipath realizations, it is shown that the two schemes have roughly the same performance with respect to the mean square error of the output signal constellation.