A Practical Pilot for Channel Estimation of OTFS

Abstract

The widely used embedded impulse pilot for channel estimation of orthogonal time frequency space modulation (OTFS) has a prohibitively large peak to average power ratio (PAPR). Hence, in this paper, we propose a novel embedded pilot with cyclic prefix (PCP) that has a significantly reduced PAPR compared to the impulse pilot. This is achieved by spreading the pilot power along the delay dimension using a constant amplitude Zadoff-Chu (ZC) sequence with a cyclic prefix (CP). We analytically derive upper bound PAPR expressions for the impulse pilot and the proposed PCP. Together with our numerical results, these upper bounds attest the significant PAPR improvement that is achieved by PCP. We also develop a two- stage channel estimation technique with a superior performance to the threshold-based channel estimation for the impulse pilot. At the first stage, the channel is estimated by a linear estimator under the assumption of the channel being locally linear time invariant over each time-slot within the OTFS block. Taking advantage of the benefits that are offered by the CP in our proposed pilot structure, we develop a low complexity least squares based estimator for implementation of the first stage. At the second stage, we use the channel estimate from the first stage and the generalized complex exponential basis expansion model (GCE-BEM) to accurately estimate the full channel. Finally, we numerically analyse and show the superior estimation performance of our proposed channel estimator for PCP to the threshold-based estimator for the impulse pilot.