On Thursday, March 14, 2013 3:47:30 PM UTC-7, wrote:
8 Gs seems like a lot. And position error will indeed maginfy second derivatives. I'd first fit a smoothed flight path through 10 points or so, then take second derivatives of the smoothed function. Just run a regression with a quadratic function for the x y and z coordinates separately +/- say 5 points, and use the coefficient on the quadratic term to get local acceleration.

John Cochrane

I think 4 second intervals is pretty tough for trying to generate a view of acceleration in the vertical plane. You need three points to fit a quadratic which is what you need to measure acceleration. 3x4 =12 seconds. Take a look at this to see what happens when you pull just 3 Gs for 12 seconds.

http://www.youtube.com/watch?v=9aDJLDQ-5QU

I think you need a higher sample rate to come up with anything for normal maneuvering. Certainly measuring a circle will give you horizontal centripetal acceleration, but circling gives you a fair number of data points without getting inverted.

9B