Date: May 8, 2013 1:04 AM
Author: Derek Goring
Subject: Re: Filtering Signal from Accelerometer Impact Data

On Wednesday, May 8, 2013 10:28:10 AM UTC+12, Jonathan  wrote:
> Hello,
>
>
>
> I have conducted a series of impact tests using an accelerometer mounted to the hammer of a Charpy impact tester. Instead of swinging through at the bottom of the swing arc, I installed a flat plate as an impact "target" to stop the swing.
>
>
>
> From each of the tests, I acquired various sets of data from the impact. I aquired the following type of plot:
>
>
>
> https://docs.google.com/file/d/0BxFCEJ8KYcAJMENERkltckp3Rms/edit?usp=sharing
>
>
>
> After the first impact, the hammer bounced several times, explaining the repeated "bump" sections.
>
>
>
> The objective of the tests were to determine the dampening effect of adding certain material cross sections to the hammer impact face.
>
>
>
> This appears to be a sinusoidal waveform, which is expected. My question is, what would be the best way to filter this data in order to remove the noise associated with the accelerometer?
>
>
>
> Any help is appreciated. Thank you


What noise?
Compared with what my accelerometers measure, your signal looks pretty clean to me.
As far as filtering such a signal goes, it's non-stationarity suggests using wavelets. If you have the Wavelet Toolbox, you could try wavedec which will decompose the signal into wavelet space. Then you can either low-pass or high-pass filter by combining details and approximations appropriately using waverec.