Design a filter for an acceleration signal

[SergeArmz]

Hello guys,

I have a very good analytical subject here, look I'm generating an acceleration signal at sampling frequency of 48 KHZ, the accelerometer produces a noisy signal by itself, so I perform a FFT analysis to identify where the fundamental frequency lives, after that I select the frequency, then after I select the filtering method applying a low pass filter and the result it's a clear signal, how ever sometimes when you attach the accelerometer to a metalic surface the fundamental frequency is affected by the resonance of all the system and your FFT will produce a variety of peaks, in this case which peak should be selected as the fundamental frequency?, which of the peaks should be disregarded and considered as noise? it is ok that I'm using an eliptic filter to filter the signal?, as I know it has the steepest response at lower order number

I'm attaching two samples with the two conditions I have

I would really appreciate your feedback on this

Regards 

SergeArmz

[johnsold]

The problem is really that you have introduced those resonances. What if the resonance happens to be at the same frequency as your desired signal? Then the resonance may get so large as to destroy the accelerometer! If the frequencies are close but not exactly the same, it becomes difficult to design a fliter which can adequately separate the signals without too much error on the desired signal.

The best solution is to modify the metal surface or the attachment method so that the resonance does not occur or else occurs at a frequency far removed form the desired signal frequency.

If that is not a possibility, consider adding a second accelerometer mounted in such a manner that it does not respond to the resonances but does respond to the frequency of the desired signal. Even if the second accelerometer does not accurately measure the amplitude of the acceleration due to its position, it may permit determination of the signal frequency.  

Remember, the device you are measuring is likely also experiencing accelerations due to the resonances, so you may need to measure them.

Lynn

[SergeArmz]

Yes, I agree to your statements, I'm measuring a metal frame without damping, when applying damping the fundamental moves to a lower frequency value, this is basically the theory of mechanical vibrations, I tried putting the accelerometer in different parts of the system in order to try to identify where the resonance frequencies live. Let me check the other positions to come up with another analysis plots.

If I identify that some resonance frequencies in the other positions are very similar to the ones affecting my original objective spot, can I disregard them as the ones that are producing noise in my signal and start doing my filtering?