Channel Estimation for OTFS using Pilots with Cyclic Prefix in Fractional Doppler Channels

We propose a novel method for performing channel estimation (CE) for an orthogonal time frequency space (OTFS) modulated waveform that exploits the recently introduced pilot structure with a cyclic prefix (PCP). This is done by analyzing the received symbols in terms of a superposition of multiple delay-spread PCP symbols. In contrast to the original single impulse pilot (SIP), PCP reduces the PAPR by distributing the same power among multiple pilot symbols that carry a Zadoff-Chu sequence. We also present an extension of the PCP-based channel estimator in the delay-Doppler (DD) domain by considering the phase rotations, a result of using non-orthogonal pulse shaping filters, which improves performance for practical implementations. Finally, we introduce a convolution-based filter algorithm for fractional Doppler estimation in the DD domain. We conduct extensive simulations and compare OTFS against 5G NR OFDM using a hardware implementation to assess the performance of the proposed approaches in terms of the normalized mean square error (NMSE) of the estimated channel impulse response and the error vector magnitude (EVM).