User manual

Reference174
Copyright © 2007-2013 Pico Technology Ltd. All rights reserved.psw.en r32
9.3
Spectrum window functions
To create a spectrum view, PicoScope captures a block of sampled data over a finite
time interval and then uses a Fast Fourier Transform to compute its spectrum. The
algorithm assumes a signal level of zero at all times outside the captured time
interval. Typically, this assumption causes sharp transitions to zero at either end of
the data, and these transitions have an effect on the computed spectrum, creating
unwanted artefacts such as ripple and gain errors. To reduce these artefacts, the
signal can be faded in and out at the start and end of the block. There are several
commonly used "window functions" that can be convolved with the data to effect this
fading, and which are chosen according to the type of signal and the purpose of the
measurement.
The Window Functions control in the Spectrum Options dialog lets you select one of
the standard window functions for spectrum analysis. The following table shows some
of the figures of merit used to compare the functions.
Window
Main peak
width (bins
@ -3 dB)
Highest
side lobe
(dB)
Side lobe
roll-off
(dB/octave)
Notes
Blackman
1.68
-58
18
often used for audio work
Gaussian
1.33 to 1.79
-42 to -69
6
gives minimal time and
frequency errors
Triangular
1.28
-27
12
also called Bartlett window
Hamming
1.30
-41.9
6
also called raised sine-squared;
used in speech analysis
Hann
1.20 to 1.86
-23 to -47
12 to 30
also called sine-squared; used
for audio & vibration
Blackman-Harris
1.90
-92
6
general-purpose
Flat-top
2.94
-44
6
negligible pass-band ripple;
used mainly for calibration
Rectangular
0.89
-13.2
6
no fading; maximal sharpness;
used for short transients