Back in the US-SR: Unlimited Sampling and Sparse Super-Resolution with its Hardware Validation

TL;DR

This paper introduces a novel super-resolution recovery algorithm for high dynamic range signals acquired via modulo sampling, validated through hardware experiments demonstrating robustness in noisy conditions.

Contribution

It presents a new end-to-end super-resolution method leveraging sparse structures in modulo samples, extending USF to non-bandlimited signals with hardware validation.

Findings

The US-SR algorithm effectively recovers sparse signals from modulo samples.

Hardware experiments confirm robustness in noisy environments.

Sampling criteria ensure reliable super-resolution reconstruction.

Abstract

The Unlimited Sensing Framework (USF) is a digital acquisition protocol that allows for sampling and reconstruction of high dynamic range signals. By acquiring modulo samples, the USF circumvents the clipping or saturation problem that is a fundamental bottleneck in conventional analog-to-digital converters (ADCs). In the context of the USF, several works have focused on bandlimited function classes and recently, a hardware validation of the modulo sampling approach has been presented. In a different direction, in this paper we focus on non-bandlimited function classes and consider the well-known super-resolution problem; we study the recovery of sparse signals (Dirac impulses) from low-pass filtered, modulo samples. Taking an end-to-end approach to USF based super-resolution, we present a novel recovery algorithm (US-SR) that leverages a doubly sparse structure of the modulo samples. We derive a sampling criterion for the US-SR method. A hardware experiment with the modulo ADC demonstrates the empirical robustness of our method in a realistic, noisy setting, thus validating its practical utility.