Uncertainties of a Spherical Magnetic Field Camera
Fynn Foerger 1 2 3 , Philip Suskin 1 2 , Marija Boberg 1 2 , Jonas Faltinath 1 2 , Tobias Knopp 1 2 3 , and Martin Möddel 1 2 · 2025
Read Uncertainties of a Spherical Magnetic Field Camera by Fynn Foerger 1 2 3 , Philip Suskin 1 2 , Marija Boberg 1 2 , Jonas Faltinath 1 2 , Tobias Knopp 1 2 3 , and Martin Möddel 1 2 (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-2506-17359.txt.
Introduction
Fynn Foerger and co-authors study the errors that creep into a magnetic field camera, in work from 2025. Such cameras estimate a magnetic field from readings on a spherical shell of sensors, a method used in medical imaging, geomagnetism, and brain studies.
While the underlying spherical-harmonic math is well understood, real sensors carry calibration errors and small position mistakes whose effect on the field estimate had not been mapped. Using Monte Carlo simulation, the team tracked how these flaws spread through the estimate. The results show how robust the method is in practice and point to which error sources dominate, guiding better camera design.
Catalog record
| CREATOR | Fynn Foerger 1 2 3 , Philip Suskin 1 2 , Marija Boberg 1 2 , Jonas Faltinath 1 2 , Tobias Knopp 1 2 3 , and Martin Möddel 1 2 |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 15 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-17359
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Uncertainties of a Spherical Magnetic Field Camera Fynn Foerger 1 2 3 , Philip Suskin 1 2 , Marija Boberg 1 2 , Jonas Faltinath 1 2 , Tobias Knopp 1 2 3 , and Martin Möddel 1 2 Abstract Spherical harmonic expansions are well-established tools for estimating magnetic fields from surface measurements and are widely used in applications such as tomographic imaging, geomagnetism, and biomagnetism. Although the mathematical foundations of these expansions are well understood, the impact of real-world imperfections, on the uncertainty of the field model has received little attention. In this work, we present a systematic uncertainty propagation analysis for a magnetic field camera that estimates the field from surface measurements using a spherical array of Hall magnetometers arranged in a spherical t t -design. A Monte Carlo–based approach is employed to quantify how sensor-related uncertainties, such as calibration errors and positioning inaccuracies, affect the spatial distribution of the estimated field’s uncertainty. The results offer insights into the robustness of spherical harmonic methods and help identify the dominant sources of uncertainty in practical implementations. I Introduction Magnetic field measurements play a crucial role in a wide range of industrial and scientific applications. Depending on the application, magnetic field sensing systems must be tailored to capture static or dynamic field distributions with specific requirements regarding precision, dynamic range, and measurement volume. To measure magnetic fields within a three-dimensional volume, various strategies have been developed. A traditional approach involves moving a single magnetometer through a predefined set of positions, recording the field sequentially at each point [ 1 ] . While this method can offer high spatial resolution, it is highly time-consuming. To overcome these limitations, array-based systems have been proposed that employ multiple sensors distributed throughout or around the measurement volume. By acquiring data simultaneously at all sensor locations, these systems offer significantly increased measurement speed [ 2 , 3 , 4 ] . Depending on the required field range, and sensitivity, a variety of magnetometer technologies can be used, including Hall sensors, fluxgates, and magnetoresistive elements [ 5 ] . A particularly effective technique for estimating magnetic fields in a volume is to place sensors on a surface of a predefined volume enclosing the region of interest. In regions with zero current density, the magnetic field satisfies the Laplace equation, allowing each field component to be expressed as a harmonic series. By sampling the field on the surface and solving for the expansion coefficients, the magnetic field throughout the enclosed volume can be accurately reconstructed using a relatively small number of measurement points [ 6 , 7 , 8 , 9 ] . In this work, we build upon a previously developed spherical magnetometer array, which provides surface measurements used to construct a surrogate model of the magnetic field via a spherical harmonic expansion [ 10 ] . The array consists of N=86 Hall sensors placed on a 3D-printed sphere with a diameter of 9\text{\,}\mathrm{cm} , following a spherical t -design with t=12 [ 11 , 12 ] . It enables the simultaneous acquisition of full 3D vector field data at a temporal resolution of 10\text{\,}\mathrm{Hz} , with field strengths covered ranging from \pm 266\text{\,}\mathrm{mT} . The reconstructed field is represented by a sixth-degree spherical harmonic expansion, resulting in a compact and smooth volumetric field description. A picture and CAD rendering of the sensor is shown in Fig. 1 . While previous work focused on system architecture and calibration method, this study investigates how real-world uncertainties, such as sensor calibration uncertainties and placement inaccuracies, propagate into the reconstructed field. We present a systematic analysis of key uncertainty sources and quantify their individual and combined influence on the reconstruction accuracy. II Methods Fig. 1: Left: CAD rendering of the spherical magnetometer array. Colored brackets hold the individual sensors in place and define their orientation. The hexagonal openings enable convenient routing of the flexible circuit connections without compromising the mechanical stability. Right: Detailed view of a flexible printed circuit board with integrated Hall sensors. Reprinted from [ 13 ] . II-A Field Expansion The field expansion for a discrete measurement on the surface of a Ball \mathbb{B}_{R} with radius R according to a t -design can be expressed as \displaystyle B^{i}(\bm{r})\approx\sum_{l=0}^{\lfloor t/2\rfloor}\sum_{m=-l}^{l}\gamma_{lm}^{i}Z_{l}^{m}(\bm{r})\quad\forall\bm{r}\in\mathbb{B}_{R}, (1) for each field component i\in\{x,y,z\} , where \gamma_{lm}^{i}\in\mathbb{R} are the expansion coefficients, and Z_{l}^{m}(\bm{r}) are the solid spherical harmonics [ 11 ] . With the t -design sampling points \{\bm{r}_{1},\dots,\bm{r}_{N}\}\subseteq\partial\mathbb{B}_{R} the expansion coefficients are given by \displaystyle\gamma_{lm}^{i}=\frac{2l+1}{R^{l}N}\sum_{k=1}^{N}B^{i}_{k}Z_{l}^{m}\left(\frac{\bm{r}_{k}}{R}\right), (2) where B^{i}_{k} are the measured components of the magnetic field at \{\bm{r}_{1},\dots,\bm{r}_{N}\} [ 13 ] . For a more detailed analysis, see [ 11 ] . II-B Uncertainty-Prone Variables If the magnetic field can be accurately described by the truncated expansion in ( 1 ), then any reconstruction uncertainty in the field B^{i} originates from uncertainties in the expansion coefficients \gamma_{lm}^{i} as defined in ( 2 ). These coefficients are influenced by several sources of measurement uncertainty. In our setup, we apply a linear calibration model to the measured field vectors, given by \bm{B}_{k}=\bm{R}_{k}\bm{b}_{k}+\bm{O}_{k} , where \bm{R}_{k}\in\mathbb{R}^{3\times 3} represents the rotation and scaling matrix, \bm{O}_{k}\in\mathbb{R}^{3} the offset correction, and \bm{b}_{k}\in\mathbb{R}^{3} the raw sensor reading [ 13 ] . Substituting this into the expression for the expansion coefficients yields \displaystyle\gamma_{lm}^{i}=\frac{2l+1}{R^{l}N}\sum_{k=1}^{N}(\bm{R}_{k}\bm{b}_{k}+\bm{O}_{k})^{i}Z_{l}^{m}\left(\frac{\bm{r}_{k}}{R}\right). (3) The key variables contributing to uncertainty in this equation are: • \bm{b}_{k} : The raw sensor readings are subject to various sensor-specific uncertainties. For the TMAG5273x2 (Texas Instruments) used in our field camera, this includes a sensitivity drift of up to 5%, an offset drift of approximately 3\text{\,}\mathrm{\SIUnitSymbolMicro T} , and a 1 sigma magnetic noise of 24\text{\,}\mathrm{\SIUnitSymbolMicro T} . • \bm{R}_{k} , \bm{O}_{k} : The rotation matrices and offset vectors are determined by exposing each sensor to six orthogonal magnetic fields of 20\text{\,}\mathrm{mT} , generated by a dedicated calibration setup [ 13 ] . A total of 3000 measurements were acquired, and the calibration parameters were obtained by solving an optimization problem [ 13 ] . Uncertainties in \bm{R}_{k} and \bm{O}_{k} arise primarily from spatial inhomogeneities in the calibration field and sensor noise. The field exhibits a maximum norm inhomogeneity of 1.1\cdot 10^{-3} and a maximum undesired orthogonal component of 1.2\% . Additionally, the Earth’s magnetic field was not compensated during calibration, introducing an offset uncertainty of approximately 50\text{\,}\mathrm{\SIUnitSymbolMicro T} . • \bm{r}_{k} : Sensor positions are inferred from the rotation matrices \bm{R}_{k} using an optimization procedure. Each sensor is assumed to be mounted tangentially to the surface of the sphere, allowing the position to be deduced from its local z axis [ 13 ] . In practice, however, the manual placement introduces an estimated angular uncertainty of about 2°. However, since \bm{R}_{k} is calculated from the calibration measurements, the inhomogeneity of the calibration field further contributes to the uncertainty. • R : The sensors are mounted on a 3D-printed spherical support structure. Due to limited printing accuracy and possible thermal deformations, the actual radius may deviate from the design value. To account for this, we treat R as a free parameter within a tolerance of 0.5\text{\,}\mathrm{mm} . The optimal radius is determined in a subsequent fitting procedure by aligning the reconstructed field with an independent reference measurement. For this reason, we neglect the uncertainty in R