Acceleration data

[suviuuno]

Hi,

To get velocity from acceleration data we integrate once. Integrating velocity we get displacement.
Acceleration data is in time domain.
Velocity is given by integrating acceleration in time domain and same for displacement or integrating acceleration in frequency domain to get velocity and displacement?
My experiment does not make sense, and I am trying to make sense out of it. As you may guess, I am new in the field.

Thanks

 

[Skogsgurra]

There is a lot more to it than the books are telling you.

Any zero offset in your acceleration signal will cause an ever increasing speed signal. So even if your accelerometer is at standstill, you will get an increasing speed signal.

The zero offset can come from the transducer electronics or from the simple fact that your vertical is a bit off. That will generate a zero error that is proportional to sin(angle). Where angle is the offset between transducer vertical and true vertical.

The offset can also come from the integrator or A/D (if digital integrator). Many real-world integrators also leak. That means that the output strives to zero when no input present.

Your next integrator (for position) will amplify the errors in the speed signal. And then add a few more. So, if you are planning on an inertial guidance system, you are in for a tough challenge.

But it can be done. And it has been done.

Gunnar Englund

http://www.gke.org

 

[ibrahimucak]

Integration in time domain causes high frequency noise signals. If possible use a displacement transducer to measure the displacement instead of double integrating the acceleration data. If this is not possible, then try to use some digital filters during integration. However if you conduct the integration process in frequency domain, you would get rid of the high frequency noise signals due to integration.

Regards,
?brahim Ucak

 

[Skogsgurra]

No. Integration does not cause HF noise. Quite the opposite. Derivation does.

Gunnar Englund

http://www.gke.org

 

[suviuuno]

Thanks for the replies. I do not see how integration in frequency domain is possible. Can you share some light, ibrahimucak?

 

[fetterlabs]

If the data is all ready in the frequency domain, just divide each frequency line by jw. I suspect, if you were able to get the data into the frequency domain, you were using a program, and this program should have the ability to integrate in the frequency domain. Many even do the double integration in on step.