Do freescale sensor fusion librarty compasate temperature drift in raw data?How is it compensated?

[harshpatel]

Hello I am working on freescale 9 axis sensor and using freescale sensor fusion library to get fussed data. In some case we found that, sometimes board temperature increases too much hence some drift in sensor performance.

So let me know is there any way to overcome temperature drift in sensor output? 

Thanks & regards

Harsh

[harshpatel]

Hello Mike

Thank you for inputs..

I have found that when sensors are steady then due to temperature there is +/- 2 deg  drift in Gyro data. so may YAW angle is fluctuating at same position.

So i want to know formula of YAW in our freescale sensor fusion library:

1] 9 dof mode 

2] 6 dof mode

- Main intense is to know how much effect of gyro in YAW angle calculation in both modes. Is  +/- 2 deg. variation in gyro can cause major variation in YAW? 

Thanks & regards

Harsh Patel

[michaelestanley]

Harsh,

If the sensors are stationary, then over the long term (many seconds) the gyro has no effect on the 9-axis yaw computation.  The reason for this is that the accelerometer and magnetometer provide the only steady state orientation information.  The 9-axis algorithm will always converge back to the eCompass solution if the sensors are not moving.

For 6-axis accel+gyro, gyro offset WILL affect yaw, especially in the case of stationary sensors.  That is because there is no steady state truth to keep the algorithm properly aligned about the Z-axis.  That shortcoming in the 6-axis solution is why the 9-axis one exists.

As to equations ...  It's not that simple.  Both Kalman filters are optimized to compute device orientation relative to the global frame.  All calculations are done using quaternion math.   Rotation matrices and roll/pitch/yaw are extracted from the orientation quaternion.  If you really want to know that math, it is fully described in the app notes packaged with the sensor fusion library.  You will quickly see it is far beyond the scope of a community posting.

Regards,

Mike

[michaelestanley]

The library does not include temperature compensation.  The magnetometer is (I think) the most sensitive to temperature.  This is one reason we do continuous magnetic calibration in the background.  By continually recomputing the trim coefficients, we sidestep the temperature problem.  But to do so requires that the sensors be in motion.  If you have an application where the sensors are stationary, you'll see drift.  Sorry, but I have no solution to offer for that.

Mike