FFT (with one time signal) with NI 9232 data

[Schafbo]

I want to do FFT analysis on data gathered with Daqmx hardware - a cDAQ NI 9174 with a NI 9232 Sound and Vibration module. The goal being to find a resonant frequency and graph of the frequency domain data using Diadem. I gathered the data using the sample program IEPE - Continuous Input in LabVIEW and saved it as a TDMS file.

When I go to do the FFT analysis, I select the accelerometer data for the given access and leave all the default options. Is this the correct procedure? Do I need a Time signal, and if so, how do I generate one for any given TDMS data? Data saved in TDMS format should be of 'waveform' type, so I was thinking it shouldn't require it. However, the FFT generated does not match the analysis done by our vibration software running in parallel, and I'm trying to find the source of error.

[c-RADAE]

Hello,

Yes that is correct, you should not need a time channel because the data is of Waveform type. Can you include a picture of what you expect the FFT output to look like and what the one in DIAdem looks like? 

Regards,

Conrad 

Applications Engineer

National Instruments

[gsklyr]

When you go to "FFT with one time signal", it selects a time channel by default.  If you drag a waveform into the "signal channels", you might still have a time channel above that selected by default.  You would have to click on the time channel drop down and select the blank option to get the right FFT.

[Schafbo]

Thanks for confirming that I can use waveform type that way. I've attached outputs of the DIAdem FFT and vibration software FFT for comparison. It seems the peaks are at similar frequencies (minus some DC component), but the magnitudes are all off. I've also attached FFT settings for Time Interval. Is it typical that:

# of intervals: 1

interval length: # of samples

[gsklyr]

Could you right click the 2D plot in DIAdem and select Y-axis [Log.] and then X-axis [Log.]?  Is that still not matching?

[Schafbo]

The frequency peak and general curve of the data seem to be matching, except for some DC component that the DAQ may not be filtering out. I'm still not sure why the data in frequency domain is a few orders of magnitude off. For example, the amplitude in time domain is mostly 1g until it reaches resonant frequencies, then jumping to 7g, 15g, etc. Shouldn't my data in frequency domain be mostly around 1g as well? Both programs have sensor sensitivity set to 100 mV/g.