Compressed sensing enabled high-bandwidth and large dynamic range magnetic sensing

Galya Haim, Chris Mullarkey, John Howell, Nir Bar-Gill · 2025

Read Compressed sensing enabled high-bandwidth and large dynamic range magnetic sensing by Galya Haim, Chris Mullarkey, John Howell, Nir Bar-Gill (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-2502-06070.txt.

Introduction

Galya Haim and co-authors apply compressed sensing to magnetic sensing with diamond defects in work from 2025. Electron spin resonance with nitrogen-vacancy centers can sense magnetic fields, but scans need many points and heavy averaging, which slows them down.

Compressed sensing rebuilds the signal from far fewer samples. Against standard raster scanning, it tripled accuracy on noisy data, or matched accuracy with only fifteen percent of the points. Simulations back the lab results. The method is not limited to diamond centers and can extend to other resonance sensing systems, the authors note.

Catalog record

CREATORGalya Haim, Chris Mullarkey, John Howell, Nir Bar-Gill
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE25 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-06070

Browse the full library · About the Tower of Babel

Full text

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

Compressed sensing enabled high-bandwidth and large dynamic range magnetic sensing Galya Haim, Chris Mullarkey, John Howell, Nir Bar-Gill Abstract Electron Spin Resonance (ESR) is a widely common method in the field of quantum sensing. Specifically with the Nitrogen-Vacancy (NV) center in diamond, used for sensing magnetic and electric fields, strain and temperature. However, ESR measurements are limited in temporal resolution, primarily due to the large number of data points required—especially in high dynamic range regimes—and the need for extensive averaging caused by low signal-to-noise ratio (SNR). This study introduces a novel application of compressed sensing (CS) for magnetic sensing using NV centers. By comparing CS with conventional raster scanning, we demonstrate the potential of CS to enhance sensing applications. Experimental results, supported by simulations, show an improvement of factor 3 in measurement accuracy in low SNR data, which also translates to achieving the same accuracy with only 15 % 15\% of the data points. Moreover, the proposed approach is not confined to NV centers but can be extended to ESR measurements in other systems, broadening its applicability in quantum sensing. I Introduction Compressed Sensing (CS) has been established as a powerful technique, allowing for the accurate reconstruction of signals with significantly fewer measurements than traditional methods in a broad range of applications Baraniuk et al. 2010 ; Eldar 2015 ; et al. 2019 . This method leverages the principle that a signal, if sparse in a particular basis, can be effectively reconstructed from a limited number of measurements, provided these measurements are orthogonal or at least incoherent with respect to the sparsifying basis. By focusing on solutions that maximize sparsity, CS ensures high fidelity reconstruction even in noisy environments Candes and Wakin 2008 ; Duarte and Eldar 2011 . The technique’s ability to reduce measurement time, cost, and complexity has led to its widespread application across diverse fields including bio-sensing Craven et al. 2015 , Magnetic Resonance Imaging Jaspan et al. 2015 , and LIDAR Howland et al. 2013 ; Sher et al. 2019 Additionally, CS is making significant strides in quantum sensing and state tomography Gross et al. 2010 . Standard formulations of CS and their corresponding proofs of accuracy and convergence often assume that the sparse-representation of an underlying signal is in an orthonormal basis, for instance in a Fourier or wavelet basis. Recently increased attention is being paid to CS techniques where the sparse representation is neither normalized or orthogonal, and may even be highly overcomplete Sadeghi et al. 2014 ; Candès et al. 2011 . Here we present CS for quantum sensing of magnetic fields based on nitrogen-vacancy (NV) centers in diamond Pham 2013 , using a highly overcomplete sparse basis in the frequency space. We demonstrate that a CS - based measurement benefits both in terms of measurement time and accuracy in low SNR data, where CS achieved the same error as raster scanning with as few as 15\% of the data points, and a lower error overall by a factor 2 and up to 5. NV magnetic sensing has been established as a useful and diverse technique Barry et al. 2020 ; Meirzada et al. 2021 ; Ninio et al. 2021 ; Grant et al. 2023 ; McCloskey et al. 2022 ; Huxter et al. 2022 ; Li et al. 2023 , leveraging the robustness of the diamond sensor, along with the high-resolution and high sensitivity afforded by the optical readout of the signal and the favorable coherence properties of the NV defects Pham 2013 ; Doherty et al. 2013 ; Barry et al. 2020 . Standard NV magnetic sensing and imaging techniques are predicated on the accurate determination of ground-state spin energy levels positions (figure 1 (b)). This is achieved through Electron Spin Resonance (ESR) measurement (figure 1 (d)) Barry et al. 2020 ; Doherty et al. 2013 ; Maertz et al. 2010 ; Clevenson et al. 2015 . The spin state of the NV center is optically initialized, followed by the detection of fluorescence while varying microwave frequencies. Resonances are identified by a decrease in recorded fluorescence counts, as the NV center’s fluorescence is dependent on its spin state. For a detailed description of NV ESR and the experimental setup, please refer to the methods section. Figure 1: (a) Experimental setup schematic, NV’s are excited with green laser, manipulated with MW, and red fluorescence is imaged onto a sCMOS camera, in an Epi-illumination scheme. Helmholtz coils generate a magnetic field to distinguish between the resonances of the different orientations. (b) NV center’s electron energy levels, ground state degeneracy is lifted due to Zeeman shift. (c) Four possible NV orientations in the diamond lattice. (d) ESR spectrum of NV ensemble, 8 resonances correspond to m_{s}=\pm 1 of each orientation. Although this method is effective and well-established, it can be time-consuming to sample a large number of microwave frequency points, along with additional averaging, to accurately identify resonance locations with high precision Barry et al. 2020 . This challenge is particularly pronounced in high magnetic field environments that require a broad frequency range, leading to increased measurement time. This issue is well recognized in the field of high dynamic range magnetic sensing Barry et al. 2020 ; Marblestone* et al. 2013 ; McCloskey et al. 2022 , where extensive sampling can result in diminished sensitivity, and bandwidth. Addressing this challenge is crucial for enhancing the efficiency and accuracy of magnetic sensing. We note that in most scenarios and in the values quoted for magnetic sensitivity in the literature, it is assumed that a small signal regime is relevant, such that the spectral locations of the resonances are roughly known, and only measuring small deviations from them is required, such that only a couple of measurements on the slope of each resonance are needed. Here we focus on the large signal regime, which is relevant for unknown fields with potentially non-negligible gradients, and for which the standard approach inevitably leads to large frequency sweeps that translate into the limitations mentioned above. In such cases, a CS approach can significantly enhance magnetic sensing. II main A conventional ESR raster scan involves sweeping across a complete frequency window, determined by a bias magnetic field and the desired dynamic range, along with averaging to achieve the desired SNR, and minimize the measurement’s error. After the measurement is complete, the data is analyzed by fitting the results, identifying the resonance locations, and calculating the vectorial magnetic field based on these resonance frequencies Maertz et al. 2010 . We introduced the CS algorithm (Alg. 1 ) into the NV magnetic sensing system as follows: Initially, a limited number of randomly selected frequency points are measured. This preliminary phase serves both to establish a baseline for fluorescence counts and their deviations, and to expedite the process, as the algorithm requires some initial data to generate meaningful results. Once this stage is concluded, the measurement protocol shifts to operate within a “while” loop. Result: Estimate of where Lorentzian Peaks Occur Measure user-defined (small) number of initial projections at randomly chosen frequencies; Get initial estimate of mean reference power and peak locations; while fit has NOT converged do choose random frequencies within window given frequency window; measure a new projection at those frequencies;

Read the complete text