Self-noise modeling of electromagnetic sensors using M-estimation and multi-channel correlation analysis

Self-noise is a critical metric for evaluating the performance of geophysical instruments and ensuring the reliability of observed signals. This study introduces a multi-channel correlation analysis method for self-noise estimation based on M-estimation. By calculating the power spectral probability density functions (PDFs) of observed electromagnetic data, the amplitude-frequency characteristics of sensor self-noise can be effectively estimated using the M-estimation algorithm. This method not only makes full use of the observed samples but also reduces the bias caused by outliers from noisy channels and/or windows, thereby improving the reliability and robustness of the self-noise estimation. We applied the method to underground data recorded from typical electromagnetic sensors, including induction coils, fluxgate magnetometers, and Pb-PbCl2 non-polarizable electrodes. The results demonstrate that robust self-noise models for these electromagnetic sensors can be obtained under real-world noise conditions. The proposed method shows significant practical value for rapid self-noise detection and effective analysis of weak signals in applied geophysical studies.