Niepewność pomiaru impedancji metodą dopasowania do elipsy

W artykule przedstawiono porównawczą ocenę niepewności pomiaru impedancji, wyznaczonej za pomocą algorytmu dopasowania do elipsy oraz algorytmu DFT z oknem Hanninga. Wykorzystując metodę Monte Carlo, przeanalizowano wpływ niekoherentnego próbkowania na rozkład prawdopodobieństwa błędu składowych impedancji w układzie współrzędnych biegunowych.

In this paper there is presented comparative evaluation of the result uncertainty of impedance component measurement with use of the ellipse-fitting algorithm and DFT algorithm with Hanning's window under the non-coherent sampling conditions. Impedance compo-nents in both cases are determined on the grounds of pairs of signal samples collected simultaneously, in accordance with the model described by equation (1). After elimination of time, it can be presented as conic curve equation (2). Under asynchronous sampling conditions, the dependence between sampling period Ts and unknown signal period T can be described as (5), in which is a window desinchronisation factor. Then, in order to decrease the influence of the spectral leakage effect, time window w(n) should be used and the values of complex spectrum components should be determined from equation (7), while the unknown impedance components from equation (4). The ellipse-fitting algorithm determines the values of ellipse equation coefficients (8) with use of the least squares method, calculating the eigenvector a corresponding to the least positive eigenvalue . On the basis of the known values of vector a elements, the impedance component values are calculated from equation (12). Characteristics of the compared algorithms have been examined with use of the Monte Carlo method, analysing the influence of non-coherent sampling on the probability distribution of the impedance component error, for two impedances Z1 and Z2 with different values of phase angle. The results of this experiment in the form of bivariate histograms of the resultant relative measurement error of impedance components ?|Z| and ? are presented in Figs. 2-4. The influence of the desinchronisation factor value on random characteristics of the impedance relative measurement error in the form of empirical distribution curves are shown in Figs. 5 and 6.