The Influence of Windowing on Bias and Variance of DFT-Based Frequency and Phase Estimation

In sinusoidal parameter-estimation problems for measurement applications, discrete spectra with high dynamic range are often under investigation. To suppress unwanted interference effects (leakage) in the discrete Fourier transform (DFT) spectrum, the application of data windows is common practice. In this paper, the influence of windowing on the bias and variance of the frequency and phase estimates, calculated via the DFT, is investigated. Furthermore, the effect of choosing different sampling start times on both the bias and the variance is discussed.