Measurement and data acquisition devices like microphones or sound cards have their own
frequency characteristics which is added to your signal during recording, i.e. corrupts your signal.
By using Custom Filter curve SIGVIEW feature, you
can measure these effects and remove them
automatically from recorded signals.
1. Let us take sound card device as an example and try to measure its frequency response first.
Before you start, you should disconnect all sound input devices like microphones - we want to
measure only internal sound card frequency characteristics in this example. Open "Data
acquisition" dialog ("Data Acquisition/Open data acquisition" menu item), choose DirectSound as
device type and your sound card name as "Device". Leave all other default settings. Press OK
and data acquisition signal window will open.
2. While on data acquisition signal window, press FFT button in toolbar to calculate spectrum of a
signal and then choose "Signal tools/Averager" option on spectrum window. Open properties of
the FFT window (Properties option in Context-Menu) and turn off "Apply Window" option. The
result in Control Window should look like this:
3. Now go to data acquisition signal window and choose "Data acquisition/Start" menu option (or
REC) button in toolbar. Recording from the sound card will start, signal and spectrum will change
rapidly and Averager window will calculate average of all spectrums during this session.
4. Wait for ~30s until Averager window shows stable curve and stop the data acquisition. Since
we do not use logarithmic (dB) spectrum amplitude values and there is no real signal on the input,
do not expect to see too much of the content. It may look like this:
5. This curve is characteristic frequency response of your sound card. Normally, this curve is
shown on log scale, but we need it like this so we can convert it to a Custom Filter curve. As a
next step, go to Averager window and choose "File/Save as filter curve" option from a menu.
6. In Custom filter curve dialog, enter a name of your filter (for example MySoundcard.flt) and
save it to default directory. Choose "Magnitude Values filter" as value type and press OK to save
7. You can use this saved filter for any data analysis using sound card signal to correct your
measurement and remove artifacts caused by the sound card imperfection. To do so, go to your
data acquisition signal from step 1 and apply "Signal tools/Filter..." menu option on it. In the
dialog, choose your saved MySoundcard.flt filter. It means that we remove from the signal exactly
the frequency components introduced by the sound card:
8. To compare results with the original, apply FFT and Averager on the filtered signal as
described above for data acquisition window. The result should look like this in Control Window.
Additionally, open Properties dialog on both FFT windows (Properties option in Context-Menu)
and select "Logarithmic Y-Axis (dB units)" option so you can see results better and turn Apply
Window option off as in the first FFT. Go to both Averager windows and press F3 to reset their
9. Start data acquisition again and stop it after ~30s. The second Averager window (number 6 in
example) will contain pure spectrum of the input signal with removed sound card characteristics.
Since there is no real input signal in this case, we expect this spectrum to be near zero or at
least with much lower amplitude then the original.
10. You can use Overlay option to show both resulting averaged spectrums in one window (select
both in Control window and choose "Show selected as overlay" from Context-menu). The result
will be a window showing both average spectrums. The red one is original and a black one is
filtered. Obviously we achieved to reduce sound card noise for ~130 dB by using custom filter
curve. The result is almost perfectly flat curve as close to zero as it gets considering our
11. You can use this method to calculate frequency characteristics of any other more complex
data acquisition system including microphones, sensors etc.
12. Another method to use Custom filter curve is to apply it to a spectrum instead of using it as
signal filter. In that case, you see how the spectrum would look like if you would apply custom
filter curve as a filter on a signal.