Flat-top smoothing of periodograms for improvement of accuracy in estimating power spectra with the Welch's method

The main problem with the test engineer is often involved when approaching spectral estimates through FFT-based measuring systems is to control the bias and variance in the periodogram. This trade-off is particularly difficult to be reached as different parameters (type of taper for spectral smoothing, the length of the observation interval, noise level and so on) can bring to opposite results: usually, a decreasing of bias yields an increasing of variance of the estimate. The authors obtained, in previous papers, quantitative expressions for the normalized bias and standard errors when flat-top tapering and frequency averagings are employed. In this paper the experimental validations of such results are shown and the performance of a Virtual Instrument (VI) carrying out spectral analyses according to the explained techniques and realized in OOP (Object Oriented Programming language) is highlighted.