Optymalizacja ilości danych w pomiarach charakterystyk częstotliwościowych

W artykule przedstawiono metody ograniczenia ilości danych podczas zautomatyzowanego pomiaru charakterystyk częstotliwościowych układów elektronicznych poprzez optymalizację rozdzielczości w dziedzinie częstotliwości oraz grupowanie i uśrednianie wyników w ramach ustalonych przedziałów. Efektem przedstawionej metodyki jest redukcja danych pomiarowych i udokładnienie charakterystyk w obszarach o podwyższonej niepewności.

---

In this paper, there are presented two methods limiting the amount of data during an automated measurement of the frequency response characteristics of linear electric circuits. The first method involves the optimization of the resolution in the frequency domain. It consists in the usage of a changeable frequency of measurements dependent on the gradient of the characteristics of the measured circuit. The frequency of measurements is automatically regulated so that the absolute difference between the values of the subsequent measurements is approximately constant. The second method involves the reduction of data in the areas with the increased measurement uncertainty, with the standard method of increasing the number of measurements. The method requires division of the frequency range with the increased measurement number into intervals, grouping and averaging the data in these intervals. The aforementioned techniques can be applied in parallel, integrating them into a single system. The result of the described integrated methodology is the decrease in the number of measurement data files and frequently decrease in the overall experiment time without significant decrease in the quality of the frequency characteristics reconstruction. Depending on the assumed quality and characteristic of the measurement, the amount of data can be reduced two to ten times. Moreover, the accuracy of the characteristic areas with the increased measurement uncertainty can be increased with the averaging method without the increase in the number of the data. The presented methodology can be implemented in computer measurement systems.