convert vibration units in spectrum

[jonhelpmejon]

using lv 6.1 without sound and vibration tool kit. how do you integrat a vibratory spectrum from acceleration to velocity and displacement. I'm using the real fft lv function and displaying with a waveform chart.

[srm@viewpointusa.com]

first convert your raw acceleration units to ft/sec/sec (or whatever unit you want velocity and displacement) Then use the integrate function (twice for displacement). You may use FFT for frequency content but it does not help for velocity or displacement. If you have already tried this, you may have seen undesirable results, you will typically have to filter the accel signal in the range of your mechanical system response. If your system can not vibrate faster than X, you should filter just above this frequency.

[jonhelpmejon]

Thanks Stu, I'll give it a try

[dsb@NI]

Integration typically acts as a lowpass filter. For mechanical vibrations, you should make sure that your acquired signal is AC coupled or that you remove any DC component from the signal to integrate as this DC component will dominate the response and you will see a net displacement after integrating that isn't true of the real system.

You can integrate in the time-domain via filtering or numerical integration, or you can integrate in the frequency domain by simple scaling of the magnitude spectrum and shifting of the phase spectrum. It is easiest to integrate in the frequency domain because you don't have to worry about DC components dominating the velocity or displacement signals as it is easy to null the first bin in the spectrum. On the other hand,
if you need the time-domain displacement, you will have to use a time-domain technique. If you use filtering, be sure to allow enough time for the filter to settle before doing any additional measurements. If you use a numerical integration technique, pay particular attention to the initial conditions you use.

Good luck.

[jonhelpmejon]

Thanks for the info. I can use all the help I can get.

[dsb@NI]

If you only need to integrate the spectrum, you can generate a scaling array to multiply the output of the FFT. Assuming your FFT output is magnitude (as opposed to power) and linearly scaled (as opposed to dB), the scaling array looks like the following for single integration:

[0, 1/w_1, 1/w_2, ..., 1/w_last]

where w_i = 2 * pi * f_i
and f_i = df * i (assuming f0 = f_0 = 0)

You can square each element of the scaling array to perform double integration.

[jonhelpmejon]

Thanks for the help again. I think this will help. Your responses have been great.