create an order spectrum from scratch (i.e. getting a position-based fft from even-time-sampled data)

[chassan]

Hello Folks,

WHAT THE QUESTION PERTAINS TO: 

I am performing time-based sampling on 2 parameters of a system: rotary position and vibration (accelerometer in units of g).  I desire to take a position-based fft to create a magnitude-phase order spectrum of velocity in in/s.  To do this I must perform the following two operations:

1. integrate (and scale) the vibration signal from g to in/s (SVT Integration.vi) 

2. resample the even-time sampled vibration signal to an even-angle sampled signal (ma-resample unevenly sampled input (linear interpolation).vi)

THE QUESTION:

Which order should the operations be performed in, integrate then resample or vice versa?  I didn't think that the order would matter, but using the same set of data, the results are drastically different.  

NI ORDER ANALYSIS TOOLSET 2.0: 

I have the NI Order Analysis Toolset 2.0, but I haven't been able to figure out how to get the speed profile generation vis to work with DAQmx (via pxi-6602) quadrature encoder position signals.  Furthermore, it appears that I have to specify every order that I'm interested in looking at, which I don't really know at this point (I want to see all available orders) so I just decided to do my own position-based fft to get an order spectrum.

Any help is greatly appreciated.

Chris

[Mr.F]

Hi Chassan,

 

I know the SVT Integration will make certain assumptions about your input signal, so order could be a factor. Drastically different? Did you look at the Sound and Vibration Examples. How are you resampling your data? Can you post your VI and data?

[chassan]

Hi Mr. F,

Thanks for the reply.  I think I can post some data and vis, but it may take a little while as I have had to change my focus to a higher priority task.

Best Regards,

Chris

[preston_johnson · Accepted solution]

The proper order is to integrate the time domain first - creating a velocity channel.  Now you have a new channel of data.  Typically I would put this into the same waveform array along with the acceleration time waveforms. 

Then resample your accleration and/or your velocity waveforms, and then you can compute the order spectrum. 

[chassan]

Hello Preston,

Thanks for the good information.  That was the conclusion that I came to; that is, integrating an even-time sampled acceleration using time as the variable of integration would yield an even-time sampled velocity whereas integrating an even-angle sampled accleration using rotary position as the variable of integration would yield something, but I'm not sure what.

Best Regards and thanks again (Preston and Mr. F.).

Chris