New results on perfect channel shortening schemes for MIMO-OFDM systems

In this work, we explore further aspects of the zero-forcing or perfect channel shortening algorithms described in literature for single-input multi-output orthogonal frequency division multiplexing (OFDM) systems. Firstly, we propose a novel perfect channel shortening prefilter (CSP) for multi-input multi-output OFDM that admits a CSP of significantly lower order when compared to the perfect shorteners. Secondly, new analytical results are derived for perfect channel shortening algorithms, namely, their post-shortening signal-to-noise ratio (SNR) and their post-processing SNR. The analytical and simulated results are in close match. Using simulations, we show that the proposed technique outperforms the perfect CSP (zero-forcing), while performing similar to the optimal channel shortening algorithms based on maximum shortened SNR and the linear minimum mean square error. Our results also indicate that the proposed perfect CSP in MIMO-OFDM systems is less sensitive to the choice of decoding delay when compared to the linear minimum mean square error CSP, and it can yield substantial gain in practical scenarios where the delay is chosen in an ad-hoc manner by the CSP.