Measuring an AC signal value with sampling when the sampling interval does not exactly divide the AC signal's period

摘要

An AC input signal is sampled with a plurality of sets of equally spaced samples, but whose sample interval between the samples does not exactly divide the period of the input signal. Nevertheless, and error cancellation technique allows ultra accurate measurements to be made. The samples in each set of the plurality of sets are supplied to a computational process that extracts some parameter; e.g., RMS voltage. The extracted parameter is in error, owing to the non aliquot nature of the sampling. The size of the error is related to, among other things, where on the input waveform the associated set began. The error is a period AC function of that starting location. By arranging for n-many sets to start at phase differences of 1/n apart on the input waveform, a series of n-many parameter.sub.i are obtained that are each of the form [result.sub.i +error.sub.i ]. Thus, the error.sub.i are sampled at aliquot locations along an error function, and therefore sum to zero. Thus, averaging the parameter.sub.i causes exact cancellation of the error.sub.i, while at the same time increasing confidence in the result. sub.i. What is more, using inexact sampling even when it might otherwise be avoidable is shown to produce a substantial reduction in errors caused by the aliasing of harmonics within the input waveform.