MPU9150 magnetometer/compass accuracy

By jhinkle , 15 September 2013

Hello,

I am wondering what the expected accuracy for the magnetometer on the MPU9150 is? I am getting significant noise of around 10% in the raw values, which after calibration translates to a 8-10 degree variation in the computed compass heading. Also there appear to be significant non-linear effects. If the modules is rotated by exactly 90 degrees, after calibration and tilt-compensation the difference in the average compass heading can be as high as 100 degrees, i.e., 10 degrees off.

Is this expected behavior?

I am using a full ellipsoidal least-squares fitting calibration. This is a plot of the calibration data points after applying the calibration:

Image removed.

The color encodes then norm, which ideally should be 1.

Side view:

Image removed.

The four clusters correspond to four stops where the sensor PCB was held flush at rest against a (non-magnetic) wall for a few seconds, and then rotated by 90 degrees around the vertical axis.

In particular the z axis seems very noisy, see the four clusters at the bottom. It can have a +/- noise range of 10-20%. Top view:

Image removed.

phpbb Topic ID
15600