Despiking of Surface-NMR Data Using a Model-based Approach

While the applicability of surface NMR for groundwater measurements in noisy areas has increased in recent years due to improved instruments and signal processing there is still a need for improved methods targeting specific noise sources. In surface NMR instruments with hardware bandpass filters, spikes from electric fences appear as several millisecond long features. We present an analysis of the amplitudes of electric fence spikes recorded in a multichannel instrument and show how spikes cluster according to their different sources. The recorded spikes are modeled as an impulsive excitation of the bandpass filter. By averaging fitting parameters over multiple spikes overfitting problems are suppressed.