A New Adaptive Window for Improving the Mean Computing Accuracy in Non-Coherent Sampling

In numerous applications related to both signal processing for the purpose of finding out useful information, as well as in power supply applications, it is necessary to know the mean (i.e. the DC component) and the root mean square (rms) values of alternating voltages and currents. If the signal is digitized, then these values are obtained by calculating linear or quadratic averaging formulas. It is known that, in order to obtain correct results of the two values, it is necessary to take into account a whole number of periods of the signal, which is the case of coherent sampling. This condition is, however, difficult to fulfill practically, so, as a rule, for signal processing aims, non-coherent sampling is performed, i.e. the number of samples acquired from the signal is not necessarily related to the number of samples per period of the signal. In this case, the result of calculating the mean value is very imprecise, and to a certain extent also the rms value. The paper analyzes the calculation errors of the two quantities according to a series of parameters and presents a new method of substantial reduction of the calculation error, especially of the mean value, based on an asymmetrical adaptive window (AAS). This new window, applied to the signal, reduces the contribution of the residual component given by the fraction of period acquired in addition, significantly improving the calculation accuracy compared to the known methods based on the application of fixed windows.