How to remove a dynamic offset from an integrated signal?

[Henrik_Volkers]

---

guidout wrote:

First of all thanks for your time.
Maybe my problem is not clear.I am measuring hysteresis loop so I need the total signal. Moreover the offset you can see in the picture of first my post does not make sense, physically.Beacause my excitation centered around zero (shaker), of course sometimes little spikes can occur; laser or shaker spikes and I think that the offset into the integrated signal is just due to these spikes because integration "remembers" all previous hystory. Indeed, some time ago I solved this problem by decreasing the buffer per channel, in this way the offset was almost disappeard, but now I can not do a similar thing because I am performing sweep-sine tests. 

Quote:

Guess why the 100Hz HP is there?

my sweep starts from 10Hz, I do not think i can use this filter. I am not sure, I not so expert...

Guido

---

---

LOL , sorry .. I work a lot with Vibrometers , why do you think that this measured velocity and the calculated displacement makes no sense?

Why do you ignore the facts of your measurement?  Just because you can't explain it? I'm pretty sure it's still there 😉 and it has a phyical reason!

To bad I have only my home PC with no LabVIEW, octave etc so openoffice calc tell me your datafile posted was about a 300Hz with a 20Hz bias, right?  Have you calculated a powerspectrum of that signal?

Can you excite the shaker with white noise, capture the response and calculate the response in the frequency domain?

Can you measure the velocity at the base of your shaker? (with and without your DUT)

If I'm right with the 20Hz 'offset', how do you want to distinguish between a 20Hz signal response and 20Hz 'offset' ??

So you have a shaker, a vibrometer , what kind of DAQ? 

LabVIEW vers? Sound and Vibration suite?

And again: can you tell us (me) what is your setup?  Shaker can fit on a table or simulate a rocket lift-off and shake a hole satellite of some tons 😉  And what are you shaking?

[guidout]

I think the measure does not make sense because, as written, the displacement measured must be aroud the neutral position;
to demonstrate this I tried to perform many measures and then I made the avarage of them and the offset is not stable or
I tried to change the samples per buffer and, again, the offset's behavior is strange, therefore I am quite sure that this offest
is not part of a real phisical phenomenon, indeed, by applying some filters, the signal seems to get better. But I am looking for a method
to remove this offset without the use of filters because I got a similar problem with an accelerometer that has a very dirty signal.

About my measurement chain, I use an NI USB-6216-BNC or a Keithley card KUSB 3116 and LV 2010sp1 professional and yes I have sound and vibration toolkit.
About my test rig, it is composed of a shaft with a conical section in the middle and in vertical position and a disk with the same internal conic profile.
By shaking the disk and by keeping to zero the shaft rotation I want to measure the hystereris loop of this contact and how the friction influences
the dynamics of the system.

PS: I am facing another problem; to perform some sweep-sine tests I am using an external device to excite the disk (VIB-PILOT) because I need to control
the force applied. For instance I want to acquire sweep at 1,2,...N, do you know how to perform a dynamic control like this?
I created a VI that performs a forc control for step-sine tests, but I have no idea how apply the same control to a sweep.

Guido

[guidout]

I forgot these questions:

quote:

tell me your datafile posted was about a 300Hz with a 20Hz bias, right? 

into the file there is a portion of a sweep and it is aroud 300Hz...

Have you calculated a powerspectrum of that signal?

Of course

Can you excite the shaker with white noise, capture the response and calculate the response in the frequency domain?

I did it and I am in a good range for my shaker

Can you measure the velocity at the base of your shaker? (with and without your DUT)

yes, but why? to check if the shaker is subjected to some movements? I do not think so because it is strongly fixed on the table.

If I'm right with the 20Hz 'offset', how do you want to distinguish between a 20Hz signal response and 20Hz 'offset' ??

I do not undestand the question

[Henrik_Volkers]

And now for something completely different...

Are you shure you are doing a continous (buffered) aquisition? Your remark with the buffersize is an indicator for lost data

[guidout]

Yes, in DAQ assistant I set "Continuous samples"

[sunshine1708]

Thanks

I found valuable information on this thread

All the best

[guidout]

I found a way to integrate properly the signal. Practically, first I calculate the spectrum, then I remove the undesired frequency (like a filter), then I integrate every component of the spectrum (simply by dividing by omega^2 and changing the sign), then I rebuild the signal by adding all the components and performing the inverse fft. I found it very useful.

guidout

Ps: sorry for my english 🙂