Single-board low-noise fluxgate magnetometer

Low-noise fluxgate magnetometers^1–3 are vector sensors that can operate over a wide dynamic range, e.g., 6 pT–60  $μ$T (140 dB). They are established in several scientific fields, including geomagnetism,⁴ space exploration,^5,6 and navigation.⁷ Looking forward, current sensing presents a potential market growth opportunity, e.g., as a diagnostic tool to optimize the safety, performance, and lifetime of electric vehicle (EV) batteries.^8–10 The current sensing market has recently seen the introduction of a chip-scale fluxgate magnetometer,¹¹ enabling new applications such as current imaging at the mm scale. However, a significant limitation is its high noise level, approximately three orders of magnitude greater than the best commercially available fluxgate magnetometers.

In many fluxgate implementations, magnetically soft ferromagnetic wires are employed as the core material. Optimization of the core properties,¹² together with the electronic circuits, is key to building a low-noise sensor. The sensor head has two main design types. Most popular is the parallel fluxgate,¹³ and state-of-the-art devices exhibit noise densities of $3−4p T r m s/ Hz@1 Hz$⁠.¹⁴ The lowest fluxgate noise levels, $<1p T r m s/ Hz@1 Hz$⁠, have been reported using a fundamental mode open-loop orthogonal design,^15,16 enabling real-time human magnetocardiography (MCG) in an un-shielded environment.¹⁴ This recent research finding indicates the potential for fluxgate magnetometers to compete in bio-medical applications with technologies, such as super-conducting quantum interference devices (SQUIDs) and atomic magnetometers.¹⁷ 

While most fluxgates are operated in a closed-loop feedback mode,¹⁸ open-loop can offer lower noise and crosstalk over a reduced linear range due to the absence of the compensation coil drive.¹⁴ In un-shielded operation, the external field seen by the core can be nulled by orienting the fluxgate axis of sensitivity parallel to the Y (East-West) component of the Earth’s geomagnetic field.

The aim of this work is to develop a single-board low-noise fluxgate magnetometer, containing all components needed to output digital magnetic flux density data, and to use geomagnetic field testing to demonstrate its performance. We use commercially available amorphous wire cores and integrated circuits (ICs) in a mixed-signal printed circuit board (PCB).

Our fluxgate magnetometer uses a single-axis parallel rod Vacquier design^13,18 (Fig. 1). One advantage of this design is the large geometrical anisotropy, which offers high resistance to cross fields. The core elements are CoFeSiB 0.1 mm diameter, 15 mm long amorphous wires (100DC2T from Aichi-Steel¹⁹) with extremely low magnetic hysteresis, coercivity H $c=4$ A/m, and saturation magnetization flux density B $s a t=0.6$ T. To the best of our knowledge, this is the first reported use of this core material in a fluxgate implementation. Excitation and sense coils are formed from 0.1 and 0.2 mm diameter enameled Cu wire, respectively. The number of excitation coil turns N $e=100×2$ and sense coil turns N $s=600$⁠.

We use the conventional method of measuring the fluxgate sense coil second harmonic (2f) signal. A high-level schematic of the open-loop sensor electronics is shown in Fig. 2 and a CAD image of the prototype PCB in Fig. 3.

1f and 2f phase-locked signals are generated by a programmable multi-channel radio frequency (RF) generator/phase-locked loop (PLL). The 1f signal drives the power amplifier to the excitation coils, and the 2f signal is used as the reference input to the synchronous demodulator. A low-pass filter with cut-off frequency (⁠ $−3$ dB) = 100 Hz follows the synchronous demodulator. A phase-shifter enables adjustment of the phase difference between the 1f and 2f signals. The 2f fluxgate sense coil signal is synchronously demodulated back to DC, and this DC signal is then sampled by a bipolar 32-bit $Σ−Δ$ analog-to-digital converter (ADC).

The 32-bit ADC has an adjustable sample rate (5 Hz–10 kHz) and digital filtering options for bandwidth tuning and noise reduction. The effective bit resolution depends on the sample rate and digital filter used. The magnetic flux density noise results presented in this paper use a 200 Hz sample rate and digital filtering for an effective resolution of 24 bits. Geomagnetic field data are taken at a 5 Hz sample rate, digitally filtered, and further averaged to obtain an effective resolution of 28 bits.

The excitation coil circuit is designed to drive the cores deep into saturation, reducing perming effects and noise while maintaining low power consumption.^20,21 The sensor head average power dissipation = 40 mW. A resonant LC circuit provides bandpass filtering and parametric amplification of the 2f fluxgate sense coil signal. The resonant LC circuit is formed by tuning capacitor C $S$ in parallel with the sense coil L $S$⁠. The 2f frequency = 30 kHz is chosen to match the resonant frequency and can be fine tuned in software to maximize parametric amplification thereby reducing amplifier noise. The settling time of the tuned fluxgate output is less than the sampling time of the ADC. The tuned fluxgate output is buffered to avoid loading the resonant circuit.

An on board micro-controller manages the fluxgate magnetometer operation and sends digital magnetic flux density data to a computer USB port. The PCB can also be mains powered using a micro-USB charger or battery powered. Data are then sent by the micro-controller over WiFi or Bluetooth by Message Queue Telemetry Transport (MQTT) to a cloud server. The cloud server provides real-time data plotting options in addition to data storage.

The fluxgate magnetometer has sensitivity = 80 kV/T and bandwidth = 100 Hz. Calibration was made by rotating the axis of sensitivity in the Z component of the Earth’s local geomagnetic field (47.5  $μ$T). We measured noise performance in open-loop inside a three-layer magnetic shield (Fig. 4). Also measured is the noise contribution from the sense electronics (fluxgate connected but with excitation switched off in software) and from the ADC only (ADC inputs shorted to ground). Noise spectral density is calculated using Welch’s method²² from a 5 min data sample at an ADC sample rate of 200 Hz. From Fig. 4, the fluxgate magnetometer noise spectral density $=5p T r m s/ Hz@1 Hz$⁠. The equivalent magnetic flux noise spectral density of the sense electronics $=1.5p T r m s/ Hz@1 Hz$ and for the 32-bit ADC $=0.4p T r m s/ Hz@1 Hz$⁠.

Low-noise fluxgate magnetometers are used to monitor the Earth’s geomagnetic field at approximately 200 INTERMAGNET standard observatories around the world.^23–25 In addition to their remote locations, selected to minimize anthropogenic magnetic noise, geomagnetic observatories are pre-screened for magnetic anomalies and include passive or active temperature compensation.

Making geomagnetic measurements in the presence of anthropogenic magnetic noise presents a challenge; however, it has been demonstrated that useful measurements are possible from a suburban location.^23,26 Such capability may lead to new understanding about how the Earth’s geomagnetic field changes over regions smaller than currently covered by dedicated magnetic observatories.²³ 

Our suburban test location is 10 m from housing and contains mains power magnetic noise. The fluxgate magnetometer illustrated in Fig. 3 is located at ground level on a concrete floor of high thermal mass, alignment is made using a compass and spirit levels. The fluxgate magnetometer is mains powered using a micro-USB mobile phone charger. Magnetic flux density data are recorded and sent over WiFi by the micro-controller at a cadence of 0.45 Hz to a cloud server. The data are uncorrected for any local temperature changes.

Results are compared (Fig. 5) over a 24 h timeframe with data from the nearest British Geological Survey (BGS) observatory at Eskdalemuir, 120 km distant. Our suburban test location results in higher anthropogenic magnetic noise than BGS Eskdalemuir, including variable amplitude fast spikes due to moving vehicles. However, the ionospheric diurnal variation of the Earth’s geomagnetic field is clearly visible along with faster geomagnetic field changes at the nT level, which may be due to local variations from the magnetosphere.

From a comparison of power spectral density spectrograms (Fig. 6), our data from a suburban location show broadband daytime noise approximately 20 dB higher than BGS Eskdalemuir. It also highlights short time windows overnight where the anthropogenic magnetic noise at our suburban test site becomes comparable with BGS Eskdalemuir. A dataset comparison during one such time window is shown in Fig. 7, where our prototype fluxgate magnetometer successfully tracks the BGS Eskdalemuir observatory fluxgate magnetometer (data truncated to 0.1 nT by BGS for historical reasons) with sub-nT resolution.

We have developed a parallel rod fluxgate magnetometer in a single PCB of size $5×12$ cm, weight 45 g, and sensor head average power dissipation = 40 mW. The magnetic flux noise spectral density in open-loop $=5p T r m s/ Hz@1 Hz$⁠, sensitivity = 80 kV/T, and bandwidth = 100 Hz.

Our 0.1 Hz noise (Fig. 4) is approximately 2 $×$ higher than the INTERMAGNET 1 s standard.²⁴ However, the fluxgate magnetometer performance level is sufficient to demonstrate sub-nT tracking of the Y component of the Earth’s geomagnetic field from a suburban test location.

In addition to geomagnetic applications, there is demand for such a low-noise sensor in industrial applications, e.g., as a current mapping diagnostic tool to optimize the safety, performance, and lifetime of EV batteries.

Future work will focus on the development of fully integrated open and closed-loop fluxgate magnetometers with noise spectral density $<1p T r m s/ Hz@1 Hz$ that also meet the INTERMAGNET 1 s standard. Testing of multiple sensor heads will be undertaken to determine the performance spread associated with the Aichi-Steel amorphous wire cores.

The authors would like to thank Gary Kendall, Ciarán Beggan, John McFadden, Iain Chalmers, and Marcin Mrozowski for useful discussions. The results presented in this paper rely on the data collected at Eskdalemuir observatory. We thank British Geological Survey for supporting its operation and INTERMAGNET for promoting high standards of magnetic observatory practice (www.intermagnet.org). This work was funded in part by a Faraday Institution Industrial Fellowship for T. Dyer in partnership with CDO $2$ Ltd, Sussex Innovation Centre, Science Park Square, Brighton BN1 9SB, UK.

The authors have no conflicts to disclose.

T. Dyer: Conceptualization (lead); Data curation (lead); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Writing – original draft (lead); Writing – review & editing (equal). P. F. Griffin: Conceptualization (equal); Funding acquisition (equal); Project administration (equal); Resources (equal); Supervision (equal); Writing – review & editing (equal). E. Riis: Conceptualization (equal); Formal analysis (equal); Funding acquisition (equal); Project administration (lead); Resources (lead); Supervision (lead); Validation (equal); Writing – review & editing (equal).

The data that support the findings of this study are openly available from the University of Strathclyde KnowledgeBase at https://doi.org/10.15129/f2c59f69-a82a-41f2-a068-77b07e2607c8, Ref. 27.