06-18-2012 04:19 PM - edited 06-18-2012 04:20 PM
In the Sound and Vibration Toolkit there is a VI that takes acceleration data in the frequency domain and integrates it to velocity (I am converting to IPS-peak). I would then like to apply a calibration slope and offset to the IPS values (which I can do) but then I need to convert these scaled IPS back to Gs-peak. Is there some sort of derivative function in the SVTK for data in the frequency domain? I thought there may be, seeing as there is an integration function.
06-18-2012 04:26 PM
To answer my own question, the SVFA Unit Coversion VI may be a solution...
06-19-2012 04:56 PM
Hi Greg,
Yea the SVFA Unit Conversion may be the way to go. Off hand I don't see/know of any specific VIs for this. Alternatively, will the Derivative x(t) do it? Not S&V, but you may be able to leverage it to do a raw derivative on the data. Just a thought.
Chris G
06-19-2012 05:43 PM - edited 06-19-2012 05:45 PM
@Chris-G wrote:
Hi Greg,
Yea the SVFA Unit Conversion may be the way to go. Off hand I don't see/know of any specific VIs for this. Alternatively, will the Derivative x(t) do it? Not S&V, but you may be able to leverage it to do a raw derivative on the data. Just a thought.
Chris G
I think this is what we're going to end up doing, tell me if it makes sense or there is a flaw in my logic. The accelerometer returns g's in the time domain. We will convert that to g's in the frequency domain and calibrate at some set frequency. From there, we can integrate our calibrated g's peak data in the frequency domain to IPS-peak in the frequency domain. We can also apply our calibration slope/offset from g's in the frequency domain to g's in the time domain.
The thing I am concerned about is my last sentence, can a calibration slope and offset in the frequency domain be used in the time domain? I would assume so because x g's is x g's, whether or not you're saying x g's at y hertz, or x g's at time t. Sorry, lots of variables there! Hopefully that makes sense, but please tell me if I'm incorrect in this assumption.
06-20-2012 08:36 PM
Hi Greg,
Hmmm. I must admit I'm not sure.
My math's a bit rusty, but I would imagine that the scaled signal in the frequency spectrum could be converted back to the time spectrum with that scaling taken into account. From there, you could use the standard Inverse FFT to bring the data back to the time domain. This may not be terribly straight-forward though because you can't pass all the signal information through this generic VI. A quick search also came up with this accel_to_vel_&_disp library. Pretty basic, but maybe it can give you some ideas.
As a side note, I'm assuming you are using DAQmx to acquire the signal, correct? If so, there is a polymorphic instance of the create virtual channel VI that measures velocity rather than acceleration from an IEPE sensor. You can configure the units for this to be in/s. Just a thought.
Hopefully this gives you a start.
Chris G