http://www.freescale.com/files/senso...MMA7260QFS.pdf
Above is the device, I have post processed the outputs thus my input data is calibrated to +- 1g on each of is axes. Thus the device is on my desk and I can rotate it around to axis and read plus or minus g from that axis.

If as in the example I have rotated the device at around the y axis at 45 degrees, I get 0.7g from the x and z axis and 0 g from the y axis. If I sum these I get 1 g as expected, now if I take the cos-1 of 0.7 I get 45 degress. Which compare to the device actual orientation.

If however I slide the device across my desk I get a set of dynamic values these sum to 1.28g in total as the device is accelerating, however the cos-1 of the value does not come to 45 degress. So this simple method of pridicting the device orientation does not work.

What I want to know is what is the device orientation when it is moving