Spectral analysis using the Fourier transform (FFT), but dividing the signal into a sequence of overlapping windows, which are analysed individually. This allows development of the spectrum in time, in contrast with the global analysis provided by the other spectral analysis modules. Sample position is shown on the x axis, frequency (in periods per sample) on the y axis, and power on a logarithmic scale as colour or grey scale.
The Short-time Fourier Transform (STFT) can be compared with wavelet analysis, but with a linear frequency scale and with constant time resolution independent of frequency.
The window size controls the trade-off between resolution in time and frequency; small windows give good time resolution but poor frequency resolution. Windows are zero-padded by a factor eight to give a smoother appearance of the diagram along the frequency axis. The window functions (Rectangle, Welch, Hanning, Hamming, Blackman-Harris, multitaper with 3, 4 or 5 tapers) give different trade-offs between frequency resolution and sideband rejection.
Missing values are treated using linear interpolation before analysis.