Spectrum windowing options in PicoScope 5
PicoScope’s built-in spectrum analyser produces accurate results, especially when used with a high-resolution converter, and can show them in great detail on your PC’s display. However, no spectrum analyser is perfect, and one type of error common to all FFT-based spectrum analysers is the generation of spurious harmonics. These occur because the FFT operates on a time-limited set of data called a window, which spreads out the spectrum of the original signal and causes ‘spectral leakage’. To minimise this, PicoScope can apply a windowing function to the signal, which smooths out the endpoints of the data. To get the best out of your measurements, we suggest that you experiment with the range of windowing functions that PicoScope offers. Your choice will depend on the balance that you require between frequency resolution and amplitude accuracy.
To choose the windowing function, open a spectrum window, then then go to ‘Settings | Options’ and change the value in the ‘Windowing’ drop-down box. Here is a summary of the window types:
- Rectangle. Simply truncates the data at the start and end of the window. Suitable only for transient signals, but provides maximum amplitude accuracy.
- Triangle. Also called Bartlett window. Fades the signal in and out linearly. Suitable for continuous, non-periodic signals. Greatly reduces spectral leakage but has poorer frequency resolution.
- Gaussian. Very low spectral leakage. Produces minimum possible time and frequency uncertainty, but slow to compute.
- Hamming. A type of Gaussian window with almost zero first and last samples. Suitable for continuous waveforms. Commonly used in speech analysis.
- Blackman. Gives maximum frequency resolution. Very close to the ‘ideal’ window, with minimum possible leakage.
- Parzen. Superior version of triangle window, with less leakage. Guarantees no negative leakage peaks.
- Hanning. Also called Hann and von Hann. Uses one cycle of a cosine wave. Simple to calculate, but accurate only for continuous waveforms.
12 Oct 2007