Noise Cancellation for Surface NMR: A Comparison of Time and Frequency Domain Approaches

Abstract

Even though surface NMR is the only geophysical technique that provides hydrogeophysical rock properties on the base of direct sensitivity to subsurface water, in many cases surface NMR measurements suffer from bad signal‐to‐noise ratio, and measurements can be carried out only far from sources of electromagnetic noise. To overcome these restriction some approaches using a reference loop based system were developed during the last years (Mueller‐Petke and Yaramanci (2010), Radic (2006), Walsh (2008)). All approaches have demonstrated useful capability to improve S/N. But comparison that allows for determining properties, i.e., pro and contra of each approach is missing. The scheme of Mueller‐Petke and Yaramanci (2010) calculates a transfer function between detection loop and reference loop based on an optimal filter, i.e., working in the time domain. However, calculating transfer functions in the frequency domain (as proposed by Radic (2006) but not published as replicable and verifible) is close to the time domain approach. In order to test the frequency approach explicetely, we developed our own frequency domain code. This now allows for evaluating and comparing important properties of the schemes such as (i) factor of improving S/N, (ii) appropriate filter length, (iii) accuracy and stability of the transfer function and (iv) numerical costs. Thus, for the first time we present a comprehensive comparison of the mathematical basis of the schemes and application of the time and frequency domain approach on synthetic and field data. Finally we conclude that the frequency domain approach is to be preferred usually provided that the noise is stationary.