03-25-2010 05:05 AM
Hello all,
I'm acquiring sound with a NI-USB 9229; then I pass the sample values to a method computing spectrum on them by using the fftw library (http://www.fftw.org/).
The problem is, I am getting non-zero spectrum values even when there is no input! Could noise on the input cause this?
If yes, is there a thumb rule as to what a threshold value should be so that spectrum values less than it can be considered as being generated from noise? Or do I have to compute that myself by trial and error?
Thanks a lot!
--Cristina.
04-06-2010 06:00 AM
Hello Cristina,
the modul USB 9229 has an imput noise from 320µV, so you can´t get directly an zero spectrum.
Which programming language do you use?
You have to programm an range (window) in which the FFT work like from 20 Hz to 20 kHz.
In this window you can apply an signal to get better results.
You can ask the support from www.fftw.org, if they have something implemented.
Here is an picture from the ExpressVI for Spectral Measurements and the options you can select.
I hope i can helped you.
04-06-2010 06:49 AM
Hello Alex,
thank you very much for your comment! I am using C++; in the meantime, I have played around with the FFT window size and I have come to an acceptable set of values which work for me.
I did not know about the 320µV input noise, however -- I think that will be helpful in my project!
--Cristina.
04-06-2010 06:54 AM
Hello Cristian,
here are the specs from the device on page 21 you found the input noise.
NI 9229/9239 Operating Instructions and Specifications
04-06-2010 09:43 AM