Errors Introduced by Fluctuations in the Sampling Rate of Automatically Recording Instruments: Experimental and Theoretical Approach

Abstract

The sampling rate in automatically recording instruments is usually assumed stable. Small fluctuations in this rate have practically little influence on low recording rates, but they might be important in high rates for three reasons. First, they lead to wrong estimation of certain parameters, equations of motions, etc. Second, errors accumulate and become very important in certain cases, for instance, displacements deduced from double numerical integration in instruments such as the accelerographs. Finally, they lead to noisy or wrong spectral characteristics in periodic functions. Fluctuations in the sampling rate were studied on the basis of experiments with a robotic total station, the built-in software of which was configured to display time of recordings with centisec resolution. The conclusion of these experiments is that at high sampling rates, the signal-to-noise ratio decreases, introducing additional noise in final results. Such instabilities in the sampling rate are not easy to identify in most automatically recording instruments, but if modeled, they have the great advantage that they permit high-rate aliasing-free estimates.