I am observing heading (yaw) drift with a sensorfusion configuration and sensitivity loss. My application is a simple compass, based on the Teensy 3.2 platform with 9DOF NXP sensors. I see high performance tracking orientation during the first few minutes, but inevitably the heading starts to drift after a few minute of operation, and doesn't reset even when the test rig is moved around (as suggested by other threads.) I often see the system settled into behavior of slowly increasing or decreasing heading.
Based on the other threads here, I see others have seen yaw issues although I don't understand the explanation given (which seems to range from gyro noise to lack of gravity correction factors on rotation perpendicular to the direction of gravity.)
Is this really the expected behavior? Shouldn't the magnetometer generally act to guide back any gyro drift?
I'm happy to provide source code but I'm pretty much using raw examples with very little changed.
Calibration has no effect on this.