This function allows the removal of regular and/or high frequency noise from the voltammograms.
The program allows you to change the cutoff (the highest relevant frequency, or the lowest relevant wavelength) and to see at the same time the raw (yellow) and FFTed (blue) data in the main box, and either the residue or the power spectra (of the raw and filtered data) in the lower box.
If you choose to monitor the residue (the difference between raw and ftted data), you should increase the filter size until the amplitude of the residue becomes greater than the noise (this means that you are over-filtering).
Alternatively, if the power spectrum is displayed, the cutoff can be directly selected using the mouse.
dto Decrease the filter size (-1), (the filter size is given in real units = intervals between points.)
ito Increase the filter size (+1),
2to decrease/increase quickly (+/-10) the filter size,
soasguess the filter size and shape (they can be further modified),
rto turn off and on again the display of the raw data,
qto Quit and filter.
sto switch between zeroth, first and second order derivative. (expert mode only)
c: the shape of the windowing function in fourier space is set to Gaussian, but can be changed to a box function typing c. In the latter case, all the frequencies above a threshold are set to zero; it gives better results (allowing a harder filtering) when there is no discontinuity (= relevant high frequency) in your data (this is usually the case when one analyses a single anodic or cathodic scan). Type
cagain to go back to the Gaussian shape.
pto display the power spectra (raw data in yellow, FFTed data overlayed in blue) instead of the residue. When they are displayed, the left button of the mouse can be used to select the cutoff frequency. Type
pagain to go back to the display of the residue.
b(when the power spectra are displayed): you can use a bandpass filter to delete a range of frequencies: press
b, then left click twice to select the band, or right-click to resume.
The screen capture below illustrates the effect of removing a band of
intermediate frequencies on synthetic data set (y=sin(1/x))
The cutoff, the shape of the windowing function and the choice of
viewing the power spectra instead of the residue are saved
for the next time you use a FFT routine (
The option to let
soas guess the filter size/shape is also
saved. The next time you use the Fourier transform, the starting
parameters of the filter will be set automatically. To disable this,
This command uses the public domain, real, fast Fourier transform from the NETLIB numerical library (http://www.netlib.org/fftpack/)
Christophe Leger 2009-02-24