Surface nuclear magnetic resonance (surface-NMR) is a promising technique for exploring shallow subsurface aquifer structures. Surface-NMR can be applied in environments that are characterized as a 1-D layered Earth. The technique utilizes a single loop and is referred to as magnetic resonance sounding. The technique referred to as magnetic resonance tomography (MRT) allows complex 2-D aquifer structures to be explored. Currently, MRT requires multiple loops and a roll along measurement scheme, which causes long survey time. We propose a loop layout using an elongated transmitter and an in-loop receiver arrays (ETRA) to conduct a 2-D survey with just one measurement. We present a comprehensive comparison between the new layout and the common approaches based on sensitivity and resolution analyses and show synthetic and field data. The results show that ETRA generates subsurface images at sufficient resolution with significantly lower survey times than other loop layouts.

Surface nuclear magnetic resonance (SNMR) data typically suffers from very low signal-to-noise ratio. Two of the main sources of noise that generate low signal-to-noise ratios are powerlines and railways that create harmonic noise. Some noise cancellation strategies for mitigating this harmonic noise have been presented. However, when the frequency of one source of harmonics is close to the SNMR (Larmor) frequency value and thus when co-frequency harmonics are present, either the SNMR signal is significantly affected or harmonic noise cannot be effectively cancelled. Recently, a new approach to handling the issue of co-frequency harmonics that utilizes additional reference loops has been presented. In this paper, we propose an alternative approach to the elimination of co-frequency harmonics that involves using a processing-based, data-driven fitting method to avoid using additional reference loops. We investigate the accuracy and stability of the proposed fitting-based method and compare our results to those of other processing approaches using synthetic data. We find that the fitting-based approach to preserving the SNMR signal is superior to all other processing-based algorithms while effectively cancelling co-frequency harmonics. We apply the proposed fitting method to field data and demonstrate that even two different noise sources that both generate co-frequency harmonics can be handled. Finally, an efficient approach to determine the fundamental frequency of harmonics that allows acceleration of the processing of large volumes of data is given.

Surface nuclear magnetic resonance (surface NMR) is a geophysical technique for direct and quantitative detection of underground water sources. In recent years, with the application of lightweight instruments in tunnels and mines, coils with small dimensions and multi-turn have been used for surface NMR. Increasing the coil turns increases the inductance and causes problems such as severe ringing, long dead time and significant loss of early stage surface NMR signal. To address these problems, the cause of ringing in surface NMR instruments and the effect of ringing on the dead time are analyzed in this paper. The bipolar phase pulse (BPP) method is introduced to reduce the ringing based on the positive and negative superposition principle and therefore shorten the dead time. A surface NMR signal simulation test and a free radical solution test are performed to verify the effectiveness of the BPP method. A comparison with technologies such as Q-SWITCH shows that the dead time of the surface NMR instrument decreases from 35 ms to 5 ms and that the signal-to-noise ratio (SNR) of the collected signal improves by 5 dB.