A Sliding-Window LMMSE Turbo Equalization Scheme for OTFS

This letter discusses a recently proposed linear minimum mean square error (LMMSE) turbo equalization scheme for OTFS. The complexity of the LMMSE-based equalizer, measured by the number of complex multiply-and-addition operations, scales with the cube of the number of transmitted symbols in an OTFS frame. To lower the complexity, we propose to limit the number of symbols in the equalization process by introducing a sliding window and exploiting the periodicity properties of the delay-Doppler domain. We demonstrate that for a practical channel model, which introduces fractional delay and Doppler shifts, even a narrow sliding window provides a bit error rate (BER) performance within
$ {\mathrm {0.17~ \text {dB}}}$
compared to the unconstrained LMMSE-based turbo equalizer. At the same time, the complexity of the sliding window LMMSE turbo equalizer is reduced by a factor of 10. The BER gap becomes even smaller with increasing window size, allowing for a complexity/performance trade-off.