Study of Windowing in Spectral Analysis of Resonance Regimes

PurposeTo evaluate the influence of different types of data window selection, applied to time series recording resonance regimes.MethodSpectra were calculated by windowing vibration signals. Four window types are used: Rectangular, Hanning, Hamming and Blackman. For all of them, two parameters: window type/shape and length L, which influence its spectrum and consequently the spectral analysis, were assessed.ResultsRegarding the windows shape, no important variations by using Hanning, Blackman or Hamming were noted. For signal with the presence of resonance (chirp-like signal), the Rectangular window is not recommended because of the larger sidelobes and do not allow the analysis of the isolated resonance frequency. For smaller window length L, the measured resonance is in smaller frequencies, due to the spectral leakage effect. On the other side, for larger values of L, more frequencies are analyzed, what can prevent the definition the frequency of resonance. For the sampling rate used to collect the source signal, values of L around 1000 is the best choice.ConclusionsThe Hanning and Blackman windows present very similar results and, despite the wider mainlobe when compared to the Rectangular window, the smaller levels and faster decay sidelobes contribute to more accurate determination of the resonance frequencies. Due to the chirp-like excitation signal and resonance regime, the window length must be determined according to the sampling rate and rate of change of the chirp signal. Small window lengths have the problem of wide mainlobes, and for larger window lengths, more frequencies are windowed, and a trade-off must be determined for each system.