Noise Rejection in High-EMI Mining Environments: Passive Shielding vs Active Compensation

Underground mining environments produce electromagnetic interference at levels that would overwhelm a precision magnetometer if the instrument were not designed specifically for rejection. The interference sources are numerous and varied: variable-frequency drives on ventilation fans and conveyor systems producing broadband power-frequency harmonics, DC welding equipment producing baseband magnetic transients, ground-return currents from trolley and cable systems producing low-frequency wandering fields, and the mechanical motion of steel equipment in the geomagnetic field producing slow drift at the nanotesla level. For a geological gradient instrument targeting sub-nanotesla signals, none of these can be ignored.

Characterising the underground EMI environment

Before designing a rejection strategy, we characterised the actual EMI spectrum at three different underground sites: a gold mine in Western Australia, a copper mine in Queensland, and a nickel mine in Western Australia. We deployed a broadband magnetic field monitor (2 Hz to 100 kHz) at each site for 48-hour periods at measurement points 5 to 50 metres from active equipment.

The dominant interference at all three sites was in two frequency bands: 1 to 100 Hz from variable-frequency drives and cable systems, with amplitudes of 1 to 200 nT at 10 metres distance from active drives, and 300 Hz to 20 kHz from switching converters and welding equipment, with amplitudes of 0.01 to 1 nT at 10 metres. The 1 to 100 Hz band is directly coincident with the geological gradient signal band, which makes it the critical interference regime. The high-frequency band is above the geological signal bandwidth and can be addressed by filtering in the signal processing chain.

The specific spectral structure matters for designing the rejection approach. Variable-frequency drives produce interference at the drive output frequency and harmonics, which drift as the drive speed changes. This drift means narrowband notch filters are ineffective; the interference occupies a moving frequency, and the notch would need to track it. Broadband rejection across the full 1 to 100 Hz range is required.

Passive magnetic shielding: characterisation

A single-shell cylindrical enclosure of 1-mm-thick mu-metal alloy (permeability 60,000 to 100,000 at low field) with inner diameter of 80 mm, enclosing the NV-center sensor optics assembly, provides approximately 25 to 35 dB of rejection across the 1 to 100 Hz band. At the 1-nT interference level measured at typical underground survey distances, this reduces the interference at the sensor to approximately 20 to 50 pT, which is at the sensor noise floor level and still affects measurement quality for the lowest-amplitude geological signals.

A double-shell configuration (two concentric cylinders with a 5-mm air gap between shells) achieves 45 to 60 dB rejection in the same frequency range, reducing the 1-nT interference to below 1 pT. This exceeds what is needed for the current sensor noise floor. The double-shell enclosure adds approximately 1.5 kg to the sensor assembly and requires degaussing every few weeks to restore the mu-metal permeability after magnetisation by handling in the geomagnetic field.

Passive shielding is most effective at frequencies where the field penetrates the shield by induction (below the shield's geometric skin depth). At frequencies above approximately 1 kHz, shielding efficiency drops because the shield wall becomes thin relative to the skin depth, and eddy currents can no longer fully exclude the alternating field.

Active gradient compensation: principle and performance

Active compensation uses a reference magnetometer, positioned away from the geological gradient and oriented to preferentially sample the ambient interference, to generate a correction signal. The primary measurement channel subtracts this correction in real time, reducing the interference seen by the geological channel. The reference channel must have sufficient sensitivity to measure the interference accurately (an NV-center reference channel works), and must be positioned to have minimal sensitivity to the local geological gradient being measured.

In practice, the reference sensor is mounted 0.8 metres from the primary sensor in the direction perpendicular to the survey traverse, where geological gradient contributions are minimised. An adaptive subtraction filter updates the correction coefficients every 200 milliseconds, tracking the changing spectral structure of the interference as equipment speeds change. In the 1 to 50 Hz band, active compensation achieves 20 to 30 dB of additional rejection beyond the passive shield baseline, with the most effective performance in the 5 to 30 Hz range where the adaptive filter update rate is adequate to track interference variations.

The combined passive-plus-active system in our field instrument configuration achieves 45 to 60 dB total rejection from 1 to 50 Hz, bringing interference from the 1 to 200 nT source amplitude range down to below 0.3 nT at the geological measurement channel. This is sufficient for gradient measurement at the 0.3 to 0.5 nT/m noise floor level we achieve in the high-interference environments we have surveyed.

Limits of the combined approach

The active compensation approach degrades when the interference field has strong spatial structure at the scale of the sensor baseline (0.8 metres). If an interference source is very close to the survey position (within 3 to 5 metres), the field gradient of the source itself is large enough that the reference channel does not accurately represent the interference at the primary channel position. In these cases, we cannot survey closer than 5 to 8 metres from operating equipment. This is a hard operational constraint that passive shielding alone also cannot overcome.

Additionally, the reference sensor approach assumes that the interference field is spatially coherent over the 0.8-metre sensor separation distance. This assumption fails for high-frequency (above 100 Hz) interference from sources very close to the instrument, where the field decays sharply over short distances. The adaptive filter does not reliably correct for this source type, which is why we separately filter the high-frequency band before the geological measurement stage.

For surface surveys in rural areas with minimal industrial infrastructure, neither passive shielding nor active compensation is necessary at the current sensor noise floor. The ambient magnetic noise in a remote hard-rock terrain away from powerlines is typically below 10 pT from 1 to 100 Hz, which is at or below the sensor noise floor. The interference rejection system adds unnecessary complexity for surveys in quiet environments, and we offer a simplified field configuration without the active compensation system for those cases.