How tough is the MotionFit SDK?

Hi Everyone,

So I finally got to strap a MotionFit SDK board to a sports instrument and proceeded to whack a sphere with it ;) All this whilst letting the MotionFit happily send off Bluetooth data to a recording device. Everything seemed to go well. The MotionFit kept sending data and I managed to not accidentally destroy it (well, not that I could see anyway).

Back home, I extracted the recorded data and yes, now I am worried that I may perhaps have damaged the MotionFit SDK, or specifically the 9150 chip. Which leads me to the question of this topic : How much punishment can the 9150 chip (or the Motionfit SDK) withstand. In essence, I would like to invite readers to please look at my recorded data and give their opinions on if what I see is actually valid data, or if I have damaged my SDK :(

You can view the data with a modern web-browser (I used Chrome) by going to http://bit.ly/10ZdeRwl
The page may take quite a while to load because it contains a Javascript graphing library and then about 3Mbyte of data. Also because this is on a home ADSL connection ;) Once loaded, the page shows a graph that can be zoomed (just drag inside the graph to zoom a section) or panned (hold shift while dragging over the graph). The graph can be reset to original size by double-clicking on it.

Now what we are looking at is raw accelerometer readings as sent by the default MotionFit Firmware (by sending inva). The frequency seems to have been 50Hz.
If you zoom right at the beginning of the graph, I pretty much see what I expect - the z axis accelerometer reading of a pretty steady -1, and the x and y axis fluctuating around 0. Next comes impact and one sees the accelerometer going nuts. What bugs me, is that it KEEPS going nuts for another 20 seconds before things kind of stabilise. Sure, so maybe I was moving the sports instrument around a bit, but when it does finally stabilise, it now seems to have much higher noise values than before :(

So, did I break my SDK or am I just reading the data incorrectly? It has been recommended to me that I should be rotating these raw accelerometer readings with the quaternion that the MotionFit SDK is sending as well. I would imagine that would leave the accelerometer readings relative to world coordinates that would make my graphs a birt more sensible, but yes, I am more worried about why there is so much more noise after impacy. And that that is the case because I overstressed the 9150 chip? Is the 9150 the wrong device to take a bit of punishment?

Your input very much appreciated?