Removal of power-line harmonics from proton magnetic resonance measurements - BRGM - Bureau de recherches géologiques et minières Access content directly
Journal Articles Journal of Applied Geophysics Year : 2003

Removal of power-line harmonics from proton magnetic resonance measurements


The Magnetic Resonance Sounding (MRS) method is based on the resonance behaviour of proton magnetic moments in the geomagnetic field. The main distinction between MRS and other geophysical methods is that it measures the magnetic resonance signal directly from groundwater molecules, making it a selective tool sensitive to groundwater. As the signal generated by the protons is very small, the method is also sensitive to electromagnetic interference (noise) and this is one of the major limitations for practical application. The frequency of the magnetic resonance signal (the Larmor frequency) is directly proportional to the magnitude of the geomagnetic field and varies between 800 and 2800 Hz around the globe. Whilst natural noise within this frequency range is generally not very large (excepting magnetic storms or other temporary disturbances), the level of cultural noise (electrical power lines, generators, etc.) may be very high. In order to improve performance, three existing filtering techniques were adapted to processing MRS measurements: block subtraction, sinusoid subtraction and notch filtering. The first two are subtraction techniques capable of suppressing stationary power-line noise without distorting or attenuating the signal of interest, both involve subtracting an estimate of the harmonic component but differ in the way the component is estimated. The block subtraction method consists of ascertaining the power-line noise (or “noise block”) from a record of the noise alone, and then subtracting this block from a record containing both the noise and the signal. The sinusoid subtraction method is based on the calculation of the amplitude, frequency and phase of power-line harmonics using noise records. The notch filtering method does not require knowledge of the power-line harmonic parameters but it may cause distortion of the measured signal. During the study, it was found that, in the investigated frequency range, the electromagnetic noise produced by electrical power lines was much less stable and regular than expected. The proportion of 50 Hz harmonics (regular part) in the noise energy is site-dependent and may vary between 20% and 50%. Whilst the power-line harmonics are seen clearly on the noise spectra, the amplitude and frequency of each harmonic may vary significantly from one record to another. Under these conditions, any block subtraction scheme based on a high noise regularity cannot be used systematically. The sinusoid subtraction is generally more efficient than the block subtraction and its application allows noise reduction by a factor of up to nearly 5. The notch filtering technique was studied further using a synthetic signal mixed with real noise and the results show that the noise can be reduced by factors of 2–10. The efficiency of the investigated filtering techniques is site-dependent. Three important factors define how successfully noise can be filtered from the signal record: the difference between the Larmor frequency and the nearest power-line harmonic frequency; the relaxation time of the magnetic resonance signal; and the proportion of the regular part of the noise spectrum. In most cases though, the improvement achieved in the critical signal-to-noise ratio (S/N) (5- to 8-fold on average) enables MRS application to be extended towards more noisy areas. The efficiency of the notch filtering technique applied to MRS measurements is demonstrated by a field example from France.


Earth Sciences
Not file

Dates and versions

hal-04083332 , version 1 (27-04-2023)



A Legchenko, P Valla. Removal of power-line harmonics from proton magnetic resonance measurements. Journal of Applied Geophysics, 2003, 53 (2-3), pp.103-120. ⟨10.1016/S0926-9851(03)00041-7⟩. ⟨hal-04083332⟩


1 View
0 Download



Gmail Facebook Twitter LinkedIn More