Hum Filter: Power-Line Noise Eliminator For Shallow Seismic Data

A unique filtering approach designed to eliminating power-line noise on shallow seismic<br>data without affecting the frequency content of signal provides a powerful harmonic noise<br>suppression tool for data acquired with modern large dynamic range recording systems.<br>Amplitudes and phases of sinusoids (functions of power-line noise) present before the first<br>arrivals can be estimated using the Levenberg-Marquardt (L-M) method. Initial amplitudes of<br>sinusoids are determined in the frequency domain using fast Fourier transform (FFT) methods<br>while initial phases are obtained by time domain correlation. Well-defined initial values<br>guarantee convergence of the L-M method. Modeling results suggest the relative error of initial<br>estimates are less than 50 percent. Calculation efficiency is achieved by simplifying the L-M<br>solution using the singular value decomposition (SVD) technique. The approach can handle<br>cases where power-line noise with frequencies of 60 Hz and/or its multiples exist simultaneously.<br>Once determined, the amplitudes and phases of sinusoids can be directly subtracted from the raw<br>data. Recorded frequencies of high-resolution shallow seismic surveys generally range from 30<br>to 300 Hz. Power-line noise (60 Hz) and its multiples (120 Hz, 180 Hz, and 240 Hz, etc.) are<br>within the optimum frequency range. This filtering technique only removes harmonic noise and<br>does not alter the spectra of signal. Real data examples demonstrate the efficiency and accuracy<br>of this method when implemented on a normal shallow seismic data processing flow.