| Abstract |
The paper presents an efficient digital method of measurement, suitable to computer-processing of pulse width modulated (PWM) waveforms. The aim is to provide a technique which can be applied both to simulated and actual PWM inverter outputs, in order to obtain precise information about their harmonic structure. The analysis is performed implementing the fast Fourier transform (FFT) algorithm to the sequence deriving from a sampling of the analog signal. Particularly, the methods to reduce the negative effects due to aliasing and spectral leakage are emphasized, namely, it is shown how the aliasing can be averted through a non-recursive realization of a suitable low-pass digital filter, having a Finite Impulse Response (FIR). Spectral leakage, caused by a non-synchronized sampling, is considerably reduced performing the convolution between the sequence returned from the FFT and the spectrum of a suitable window, in the frequency domain. Successively, simple interpolation formulas must be applied to the windowed spectral lines in order to find the frequency of the PWM signal, with a high degree of accuracy. Several test results, performed both on simulated models and on actual inverter outputs, are presented. They confirm the effectiveness of the proposed method, encouraging future applications in microprocessor-based digital instrumentations. |
