Distance from acceleration: data does not look right

[Agile]

We have run a test that derives distance from acceleration data. The accelerometer is attached to a cantilevered plastic subject (rectangular) that is "plucked" -- sort of like a ruler over the edge of a table sprung like a diving board. We used the integration VIs from the sound and vibration toolkit.  Here's snapshot of the block diagram.

With this, we were able to get the following graphs:

BUT when the customer ran the test, they got the following graphs:

Please note that the accleration graph "wobbles" (a lower freqeucny sine wave is superimposed on the expected decaying sinusodial).

Why, why, why?  I have no hair left.

For more information, please refer to the following previous post: http://forums.ni.com/t5/Multifunction-DAQ/How-do-you-derive-velocity-and-position-from-acceleration-...

[johnsold]

The customer's support mechanism vibrated at ~ 1Hz?

Lynn

[josborne]

Is the customer sampling fast enough? 

[TCPlomp]

What happens when you run the data from the client through your VI?

Especiallly the accelaration data and see if you can recreate the low frequency component. Try to determine the low frequency.

You could set a higher/lower low-pass frequency than 1Hz.

Ton

[Agile]

---

@johnsold wrote:

The customer's support mechanism vibrated at ~ 1Hz?

Lynn

---

Thank you for your response.  I apologize for the delay, but Iwanted to get as many responses as possible before we muddy the water with my opinion.

Yes, it could be because the DUT (device under test) is not securely fastened to the support mechanism or the support was also moving.  Because of the holiday break, we do not have access to their lab to verify it.  We plan on checking next week after New Year's.

See my oppinion below.

[Agile]

---

@josborne wrote:

Is the customer sampling fast enough? 

---

Thank you for your response.  I apologize for the delay, but Iwanted to get as many responses as possible before we muddy the water with my opinion.

Yes, it could be because the system was sampling too slow, producing aliased data.  The program allows the operator to change the MAX settings which includes acquisition rate.  We plan on checking next week after New Year's.

Please see my opinion below.

[Agile]

---

@TCPlomp wrote:

What happens when you run the data from the client through your VI?

Especiallly the accelaration data and see if you can recreate the low frequency component. Try to determine the low frequency.

You could set a higher/lower low-pass frequency than 1Hz.

Ton

---

Thank you for your response.  I apologize for the delay, but I wanted to get as many responses as possible before we muddy the water with my opinion.

Both graphs are based on data from our VI.  The first is a screen shot, and the second is a chart created in Excel based on the data from our VI.  Our concern is that if we apply a filter we will lose data that is characteristic of the material.

My opinion:  It might also be:

1.  A characteristic of the material either because of its shape or harmonic vibrations on a molecular level.  I'm a EE and not a polymer or material scientist, so I'm not sure of this is possible. 

2.  The accelerometer is not securely fastened.  It could be rocking back and forth at about 1 Hz.

3.  The motion is not in only one axis of motion (up and down), but wobbling side to side.  The accelerometer only measures in one degree of freedom.  Could there be a yaw/pitch that is registering as additional movement up and down.

So if

     a.  the material is properly secured on a "immovable" fixture, and

     b.  the accelerometer is securely fastened, and

     c.  the motion is in line with the sensor, and

     d.  the system is sampling fast enough, and

     e.  a meteor does not strike the lab during the test,

then

     it is likely a characteristic of the material.

Comments?

[johnsold]

I think you have covered the most important possibilities.  Careful evaluation of those should get you a long ways toward determining what is going on.

Please keep the Forum posted on what you learn.  And more questions with the new data are always welcome.

Lynn