Model-based subtraction of spikes from surface nuclear magnetic resonance data

Abstract

Surface nuclear magnetic resonance (surface NMR) has progressed significantly in recent years due to advances in instrumentation. In particular, the introduction of multichannel surface NMR instruments has been effective in improving the signal-to-noise ratio. The current methodology for noise reduction with multichannel instruments is, however, inadequate in complex noise environments, and there is a need for improved signal processing. We have evaluated a study of impulsive noise (spikes) in surface NMR data acquired with a Numis Poly instrument. We have determined how the spectral content can be used to classify spikes as originating from electric fences or sferics. Measurements of spikes from two electric fences were evaluated. The spikes were highly deterministic and can be modeled as impulsive excitations of the band-pass filter in the surface NMR receiver system. We investigated the feasibility of a model-based approach for subtraction of electric fence spikes. Model-based subtraction was shown to be possible, but it is limited by accidental fitting of the NMR signal in its current embodiment. We evaluated an example of a surface NMR data set in which subtraction of powerline harmonic noise and electric fence spike noise removed all coherence in the multichannel data, and the consequences for further noise reduction using multichannel methods were developed.