Unlocking Off-the-Grid Sparse Recovery with Unlimited Sensing: Simultaneous Super-Resolution in Time and Amplitude

Ruiming Guo, Ayush Bhandari · 2025

Read Unlocking Off-the-Grid Sparse Recovery with Unlimited Sensing: Simultaneous Super-Resolution in Time and Amplitude by Ruiming Guo, Ayush Bhandari (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-16948.txt.

Introduction

Ruiming Guo and Ayush Bhandari attack a signal recovery problem in a 2025 paper. The task is finding sharp spikes hidden inside filtered measurements. Spikes sit at exact real-valued times, not on a neat grid. Strong and weak spikes together break ordinary digital recording.

Strong spikes clip, while weak ones sink below recording noise. The authors use a recording method called unlimited sensing. It folds oversized signals instead of clipping them outright. That folding preserves detail that ordinary recording would destroy. New theory then recovers spike times and sizes together. The results extend to common non-ideal filter shapes as well. Better spike recovery helps radar, imaging, and communications systems alike.

Catalog record

CREATORRuiming Guo, Ayush Bhandari
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE78 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2510-16948

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Unlocking Off-the-Grid Sparse Recovery with Unlimited Sensing: Simultaneous Super-Resolution in Time and Amplitude Ruiming Guo, Ayush Bhandari Abstract The recovery of Dirac impulses, or spikes, from filtered measurements is a classical problem in signal processing. As the spikes lie in the continuous domain while measurements are discrete, this task is known as super-resolution or off-the-grid sparse recovery. Despite significant theoretical and algorithmic advances over the past decade, these developments often overlook critical challenges at the analog–digital interface. In particular, when spikes exhibit strong-weak amplitude disparity, conventional digital acquisition may result in clipping of strong components or loss of weak ones beneath the quantization noise floor. This motivates a broader perspective: super-resolution must simultaneously resolve both amplitude and temporal structure. Under a fixed bit budget, such information loss is unavoidable. In contrast, the emerging theory and practice of the Unlimited Sensing Framework (USF) demonstrate that these fundamental limitations can be overcome. Building on this foundation, we demonstrate that modulo encoding within USF enables digital super-resolution by enhancing measurement precision, thereby unlocking temporal super-resolution beyond conventional limits. We develop new theoretical results that extend to non-bandlimited kernels commonly encountered in practice and introduce a robust algorithm for off-the-grid sparse recovery. To demonstrate practical impact, we instantiate our framework in the context of time-of-flight imaging. Both numerical simulations and hardware experiments validate the effectiveness of our approach under low-bit quantization, enabling super-resolution in amplitude and time. IEEE Journal of Selected Topics in Signal Processing (to appear). I Introduction Consider the problem of recovering echoes of backscattered light, where each pulse-echo corresponds to an object at a specific depth. Since light travels roughly 30 centimeters (cms) in one nanosecond (ns), resolving objects separated by just a cm requires an analog-to-digital converter (ADC) capable of sub-ns temporal resolution; a requirement that is technologically prohibitive. Although this particular example relates to light-based 3D imaging [ 2 ] —a scenario discussed later in this paper—similar challenges also arise in widespread areas such as radar systems [ 3 ] , DNA profiling [ 4 ] , optical coherence tomography [ 5 ] , terahertz sensing [ 6 ] , super-resolution microscopy [ 7 ] , and radio-astronomy [ 8 ] . Fig. 1: Conventional systems with fixed digital resolution cannot achieve simultaneous amplitude and temporal super-resolution because weak signals are obscured by coarse quantization and rendered unresolvable. The USF addresses this limitation by enabling digital super-resolution, which enhances measurement precision and facilitates super-resolution in both amplitude and time (see Fig. 2 ). In scientific and engineering disciplines, this challenge is frequently termed the super-resolution (SRes) problem, aptly named because it involves recovering fine-scale features from coarse-scale measurements. Mathematically, the problem can be abstracted as recovering sparse spikes (or Dirac masses), s_{K}\left(t\right)=\sum\limits_{k=0}^{K-1}{\Gamma\left[k\right]\delta\left({t-\tau\left[k\right]}\right)},\quad\{\Gamma,\tau\}\in\mathbb{R} (1) from filtered measurements, \displaystyle g[n] \displaystyle=g(nT),\qquad n=0,\ \ldots,N-1,\quad T>0 \displaystyle g\left(t\right) \displaystyle\stackrel{{\scriptstyle\eqref{eq:SRF}}}{{=}}\left(s_{K}*\psi\right)\left(t\right)\equiv\sum\limits_{k=0}^{K-1}{\Gamma\left[k\right]\psi\left({t-\tau\left[k\right]}\right)} (2) where \psi represents a smooth kernel and T is ADC’s temporal resolution. The goal of SRes is to recover the off-the-grid , continuous-time signal s_{K}(t) from the discrete samples g[n] , which lie on a coarse temporal grid determined by T . One of the earliest solutions to this problem dates back to the seminal work of de Figueiredo & Hu (1982) [ 3 ] , who showed that 2K measurements suffice to recover a continuous-time, K -sparse signal, s_{K}(t) using Prony’s method—a foundational approach in spectral estimation. This method has since underpinned key developments in areas such as spike deconvolution [ 4 , 9 ] , finite-rate-of-innovation (FRI) sampling [ 10 ] and sub-Nyquist sampling [ 11 ] . Over the past decade, there has been a renewed interest in bridging finite-dimensional measurements with infinite-dimensional solution space . Notable contributions in this direction include the work of Bredies et al. [ 12 ] , Candès & Fernandez-Granda [ 13 ] , Tang et al. on compressed sensing off the grid [ 14 ] , and the total variation (TV) regularization approach by Duval & Peyré [ 15 ] , among others. For a comprehensive overview, we refer the reader to the excellent survey by Chi & Ferreira Da Costa [ 16 ] . What’s Missing in Previous Art? In practice, ADCs bridge analog and digital domains but are fundamentally constrained by a trade-off: for a fixed bit-budget, one cannot simultaneously achieve high dynamic range (DR) and high digital resolution (DRes). Improving one typically compromises the other. Prioritizing DR results in coarse quantization, causing (i) weak signals to be buried beneath the noise floor in the presence of strong components, as illustrated in Fig. 1 , and (ii) degradation in algorithmic SRes performance due to elevated quantization noise. Conversely, optimizing for DRes leads to clipping of high DR or HDR signals, causing irreversible information loss and breakdown of SRes methods [ 17 ] . Although SRes is well explored, its hardware validation is limited, and the information loss at the analog–digital interface remains under-explored due to high DRes assumptions. For current ADCs, capturing HDR signals at high-DRes (HDRes) is fundamentally impossible ; a key limitation for practical SRes. This motivates a wider perspective: super-resolution must simultaneously resolve both amplitude and temporal structure –an ability that conventional systems inherently lack. Increasing bit depth is unsustainable, as ADC power consumption grows exponentially with the number of bits, but only linearly with the sampling rate [ 18 ] . As a result, oversampling emerges as a more power-efficient alternative. This trade-off becomes especially critical in high-dimensional imaging, where the data volume per pixel/voxel is large, making high bit-budgets impractical due to power and storage constraints, e.g. data tensors in Time-of-Flight (ToF) imaging [ 19 , 20 , 21 ] . The USF Breakthrough. The Unlimited Sensing Framework (USF) introduced in [ 22 , 23 , 24 ] breaks the fundamental trade-off between dynamic range (DR) and digital resolution (DRes) in conventional ADCs by demonstrating that, for a fixed bit-budget, one can simultaneously achieve HDR signal capture with HDRes. USF is built on the mathematical insight that, for smooth signals [ 23 , 25 , 26 , 27 ] , the integer part can be recovered from the fractional part . In engineering terms, the integer part corresponds to the conventionally quantized signal, while the fractional part represents the quantization noise, which USF treats as an informative signal representation rather than a distortion artifact. The strength of USF arises from hardware–algorithm co-design. With folding parameter, \lambda>0 , the hardware employs an analog-domain modulo non-linearity [ 24 ] , defined as: \mathscr{M}_{\lambda}:g\mapsto 2\lambda\left(\left[\kern-3.00003pt\left[\frac{g}{2\lambda}+\frac{1}{2}\right]\kern-3.00003pt\right]-\frac{1}{2}\right),\quad\left[\kern-1.49994pt\left[g\right]\kern-1.49994pt\right]=g-\left\lfloor g\right\rfloor, (3) where \left\lfloor g\right\rfloor=\sup\{k\in\mathbb{Z}\mid k\leqslant g\} denotes the integer part of g . Since |\mathscr{M}_{\lambda}(g)|\leqslant\lambda , sampling and quantizing \mathscr{M}_{\lambda}(g) avoids clipping and yields “digital super-resolution” or HDRes (see Fig. 1 ). Combined with advanced recovery algorithms, this enables new capabilities. Hardware validation of USF has verified a 60-fold DR extension in practice [ 28 ] , along with at least 10 dB enhancement in DRes for modulo tomography [ 27 , 26 ] and radars [ 29 ] . Such improvements are critical for advanced modulation formats like 1024-QAM [ 30 ] , full-duplex communication [ 31 ] and SRes spectral estimation [ 32 , 33 ] . Fig. 2: Performance evaluation across varying dynamic ranges (DR) and quantization bits. The measurements follow

Read the complete text