Reduction of the measurement time required for high accuracy waveform acquisition by employing digital filtering instead of averaging

Abstract

High accuracy data acquisition frequently employs averaging in order to improve the signal-to-noise ratio (SNR) of the output signal. Averaging requires acquisition of several copies of a noisy input signal that results in an increase in the measurement time, and is applicable for the periodic processes only. The SNR can alternatively be improved by digital filtering of out-of-the-signal-band noise, especially when the sampling frequency was selected significantly higher than the notional Nyquist frequency (oversampling). A quantitative assessment of the applicability of this approach to high accuracy data acquisition was tested using an ultrasonic nondestructive evaluation (NDE) dataset of 1000 noisy records. Five different FIR filters were designed including two ideal filters defined in the frequency domain, and three realizable filters. It was found that the relative total mean square error filter exhibited the performance closest to that of the ideal Wiener filter. The achieved improvements for two realizable filters were above 6 dB meaning that by using FIR filtering the required measurement time could be reduced by a factor of 4 in the considering case.