Pulse-Shaped OTFS over Doubly-Dispersive Channels: One-Tap vs. Full LMMSE Equalizers

Orthogonal time frequency and space (OTFS) is a modulation technique combining Gabor structure with additional time-frequency spreading. It promises significant improvements of the wireless transmission in terms of robustness and efficiency for high mobility users. However, it requires sufficiently accurate channel state information (CSI) and an appropriate equalizer. In particular, full CSI is often assumed to mitigate self-interference. This is particularly challenging in highly dynamic vehicular scenarios where the channel is truly doubly-dispersive. Self-interference is caused by the channel cross-term coefficients coming from pulse and grid mismatch of the OTFS system with the channel scattering function. The estimation of the channel cross-term coefficients is a tedious task which is not always feasible. Focusing on practically feasible channel main diagonal estimation and equalization techniques, we propose a tuned one-tap minimum mean square error equalizer (MMSE-EQ). We consider an additional variance parameter including the power of the channel estimation error and self-interference. We determine it by minimizing an error metric between the transmitted and received pilot and guard symbols by using gradient decent with reasonable initial guess. In addition, we numerically compare the proposed one-tap MMSE-EQ with full linear MMSE-EQ with ideal CSI, and orthogonal frequency-division multiplexing (OFDM) in terms of uncoded performance. Our results indicate that, with our proposed equalizer, OTFS significantly outperforms OFDM with low-complexity one-tap equalization over doubly dispersive channels.