Design of a Hum Filter for Suppressing Power-line Noise in Seismic Data

Abstract

A unique filtering approach designed to eliminate power-line noise on shallow seismic data without affecting the frequency content of signal provides a powerful harmonic noise suppression tool for data acquired with modern large dynamic range recording systems. The filter is determined by the Levenberg-Marquardt (L-M) method using records present before the first arrivals. Initial amplitudes of sinusoids (functions of power-line noise) are determined in the frequency domain using fast Fourier transform methods while initial phases are obtained by time domain correlation. Modeling results suggest the relative error of these initial estimates is less than 50 percent. Well-defined initial values guarantee convergence of the L-M method. Calculation efficiency is achieved by simplifying the L-M solution using the singular value decomposition technique. The approach can handle cases where power-line noise with frequencies of [Formula: see text] and∕or its multiples ([Formula: see text], [Formula: see text], and [Formula: see text], etc.) exist simultaneously. Once determined, the amplitudes and phases of sinusoids can be directly subtracted from the raw data. Recorded frequencies of high-resolution shallow seismic surveys generally range from [Formula: see text]. Power-line noise and its multiples are within the optimum frequency range of shallow seismic surveys. This filtering technique only removes harmonic noise and does not alter the spectra of signals. Real data examples demonstrate the efficiency and accuracy of this method when implemented on a normal shallow seismic data processing flow. The stability and efficiency of the filter were also tested by applying it to more than 8, 000 shots of shallow seismic reflection data.