Novel zero-forcing, MMSE, and DFE equalizer structures employing oversampling and multiple receiver antennas

Novel multichannel equalizer structures are developed employing decision-directed updates of initial channel estimates formed from processing a short training sequence. For the case where (1) the delay spread is of the order of T, where 1/T is the symbol rate, (2) at least two spatially separated antennas are available at the receiver, and (3) each baseband antenna output is oversampled by a factor of 3 or 4, we show that equalization may be effected via either the zero forcing, MMSE, or DFE criteria via a small set of "sample-spaced" taps encompassing the delay spread which, again, is assumed to be a fraction of T. For the DFE equalizer, the feed-forward filter is sample-spaced while the feedback filter is symbol-spaced. This is in contrast to conventional equalization which employs symbol-spaced taps encompassing roughly the duration of the pulse symbol waveform, which is on the order of 10T when the excess bandwidth parameter is less than 0.35, for example. The comparative BER performance of the multichannel zero forcing, MMSE, and DFE equalizers using sample-spaced taps is presented for both a time-invariant channel and a case where the channels vary substantially over the burst due to the mobile motion. In the latter case, the new sample-spaced zero-forcing equalizer is shown to dramatically outperform the MMSE and DFE equalizers.