Noise suppression method of rotating accelerometer gravity gradiometer instrument based on oversampling

Abstract

In this paper, a combined method analyzing both Allan variance and power spectral density was used to characterize the original output signal and the demodulated gravity gradient signal in the rotating accelerometer gravity gradiometer instrument (GGI). the purpose is to solve the weakness of the low recognition of noise in applying simple frequency domain analyze and to accurately identify the noise pattern. The result showed that, The noise intensity near the modulation frequency point of the original output signal of gravity gradiometer is $8\times 10^{-9}g/\sqrt{Hz} l$ The noise modes are mainly white noise, pink noise and random walk noise. The means, it is very hard to effectively make gravity gradient measurement by using conventional processing method. Over sampling method had been chosen in the study to make use of the dither of white noise to lift the useful signal in original output from GGI, which caused little modulation on the latter. Then the digital filter could help eliminating the white noises. So the researcher got a better SNR. It was estimated that, through this over sampling method in the background of white noise modulation, the single frequency noise amplitude decreased 3 dB for each doubling of sampling frequency. This finding revealed a way to enhance GGI by picking up the sampling frequency from 8Hz to 9kHz and it must be supported with down-sampling-anti-aliasing filter to reduce the transmission pressure on data line. The strict linear relationship between the original signal and the modulation phase in GGI was considered together with the ARMA model of time series to calculated the original gravity gradient signal under high frequency output. With the enhanced method the outputting frequency of GGI can be put up from 0. 5Hz to 9kHz without physical updating, which permits the possibility to further suppress the white noise. According to the test in the study, after adopting the both above enhancements, the demodulated noise amplitude of the measured signal decreases by 50% in valid field from 0.01 Hz to 0.1 Hz of GGI.