Sensitivity Accuracy parameter of the MM8452Q Xtrinsic Accelerometer

[dylanpfeifer]

Hi Freescale:

Studying application notes AN4096 and AN4399 closely regarding post-PCB mount gain and offset calibration of the Xtrinsic 8452Q accelerometer, we still have questions regarding the accuracy of each channel over its measurement range in 2G mode.  So can you help us understand the meaning of your parameter "Sensitivity Accuracy" in the datasheet, which says 2.64%?  How does this apply to a calibrated device?

If we apply the 15-parameter calibration algorithm in AN4399, what can we expect the absolute accuracy per channel to be?  Uncalibrated it is +/- 20mg, which results in 1G measurements from 9.65 to 10 m/s^2, way off from the natural variance of 0.7%g. 

Thanks,

Dylan
PSL

[mruckstuhl]

I am comparing accelerometers, and used a modified form of the 12-parameter calibration on the MMA8452Q. I don't have a calibrated incline table, so I can't speak to _accuracy_ and instead I'm evaluating precision. I'm looking at the output variance to indicate how precise the chip really is, instead of depending on the number of bits in the result. I have ODR set to 50Hz, and I use a sliding window average (10 samples) to cut out some of the noise. My most recent test run (3s duration) shows the output stay within a 0.1 degree range at nominal zero tilt, and within a 0.15-degree range at a nominal 45 degree tilt.

Precision doesn't answer your accuracy question, of course.

I'm expecting to derive tilt from arctan(X/Z) instead of arcsin(X) in order to account for the 0.7% variance in gravity that the app note mentions.

Can anyone compare accuracy between the MMA8452Q, the MMA8451, and the MMA9550L? I am expecting the 8452Q and the 8451 to have similar accuracy (but better precision from the 8451), but I have no idea about the 9550L.