Sensing Fast and Distant Targets in Fractional Overspread Channels with Zak-OTFS and Modular Arithmetic

Abstract

Delay-Doppler (DD) modulations, such as Zak-transform-based orthogonal time frequency space (Zak-OTFS) modulation, represent the channel as a sparse set of delay and Doppler shifts, which makes them well suited for radar sensing. Sensing with a single DD frame, however, is unambiguous only when the channel spread factor, i.e., the product of the delay and Doppler spreads, is less then one. Long-range mobility scenarios, including drones, airplanes, low Earth orbit (LEO) satellites, and underwater acoustic links, typically violate this crystallization condition, so that a single frame aliases the target delays, Doppler shifts, or both. Existing methods add a time-frequency pilot to a conventional DD frame, resolve the aliasing in only one dimension, and assume integer delays and Doppler shifts. We propose a frame design that combines multiple DD bases with different delay and Doppler periods, which together cover an overspread channel with fractional delays and Doppler shifts, using the same bandwidth and frame size. We show that modular arithmetic on the peak positions uniquely recovers each target by pairing the peaks through their common projection and their gains, and we derive the condition under which this pairing is ambiguous. We implement the design with two superimposed Zak-OTFS frames and propose two receivers: dual-frame sensing (DS), which performs the joint detection directly, and DS with cross-pilot cancellation (DS-XC), which additionally cancels the cross-pilot interference, re-estimates the gains by joint least squares, and removes ghost targets. Both receivers maintain a detection rate close to one as the spread factor grows, and DS-XC also achieves the lowest false-alarm rate and channel estimation error among all the compared systems. In a 100 km drone scenario, DS-XC outperforms all baselines, with a complexity that grows near-linearly with the time-bandwidth product.