Innovative surface NMR signal processing to significantly improve data quality
Abstract
Surface Nuclear Magnetic Resonance (SNMR) is a relatively new geophysical technique primarily used for water detection in the shallow subsurface. Magnetic fields arising from current pulses in a surface loop antenna penetrate the subsurface and interact with the hydrogen protons of liquid water. Among the various geophysical methods, surface NMR is unique in that it is directly sensitive to water molecules. Hence it has the powerful potential to quantitatively map the water distribution with depth. The signal measurement relies on the principle of induction that creates a weak voltage in the range of nV to a few μV in the surface receiver loop. However, the record is obscured by (i) man-made, industrial, and cultural (harmonic) noise such as power-lines and railway tracks, (ii) spike events (incoherent noise), and (iii) atmospheric background noise (random). Extreme hardware requirements and the weakness of the signal cause the records to be heavily noise contaminated in general. As a consequence, efficient noise suppression techniques are required to extract the weak surface NMR signal, i.e. stacking, loop design, and digital post-processing. In this study, we present a state-of-the-art workflow for full time series NMR data processing. As a first step, random spike events are removed from all records. Reference channels are further used to create a shaping filter by which the noise component in signal record is largely reduced. In the latter stage, signal extraction is performed using digital quadrature detection with an additional phase correction. The filter design is based on a least-squares approach using different input channels. This multi-dimensional Wiener filter method allows for a multi-channel noise reduction. Today, state-of-the-art full bandwidth multi-channel recording systems offer the possibility to record four channels simultaneously. Therefore, it is possible to use up to three reference channels for noise attenuation. By analyzing the optimal filter length and reference receiver combinations, we were able to to extract the NMR signal from highly noise contaminated records. In the case where one reference channel for noise suppression fails, the NMR signal can be successfully extracted using the multi-dimensional Wiener filter.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2010
- Bibcode:
- 2010AGUFMNS11A1161N
- Keywords:
-
- 0910 EXPLORATION GEOPHYSICS / Data processing;
- 0925 EXPLORATION GEOPHYSICS / Magnetic and electrical methods;
- 1894 HYDROLOGY / Instruments and techniques: modeling