This issue remains me about the issue I had with an electronic compass I tried to use on my motor boat. The compass was from a manufacturer with good reputation, no cheap thing....
The compass had three magnetic sensors, three accelerometer sensors and three gyros so all three directions where covered (same concept as the IMU in the Phantom P4 Pro).
I tested the compass but they never got it working properly, it was sent back for calibration and then I tested again without success, I finally got a refund.
This where my findings:
Driving in small waves, the boat bounce quite hard in the waves but is never airborne. The compass drifts away 10-15 degress while the boat maintains quite steady course.
Changing speed (20-5 knots or 5-20 knots) when travelling east or west, the compass drifts away approx 20 degrees from the actual course and returns to correct course when the speed is steady again.
Changing course when travelling a course close to north. It lags behind the actual course and needs several seconds to stabilize on the new course.
Changing course when travelling a course close to south. It responds well to course changes but it may actually overshoot.
These tests where done in Sweden where the magnetic dip angle is quite high which makes things more sensitive.
This is a known issue on airplanes, check link below:
Acceleration Errors of an Aircraft Magnetic Compass
The solution for me was to use a so called "gps compass" which uses the position difference between two gps antennas to calculate the heading.
Vector V104™ GPS Compass
It is rock solid and always shows the correct heading even in sharp turns and accelerations. The antennas must be separated by at least 250 mm and the cost for such a thing is around 800 Euro with casing and cables, I guess that DJI could get this price down a bit. They have this concept on the Matrice M210 RTK to get good heading but there they also have high accuracy on the position using RTK which cost a lot! Maybe on Phantom 5?
It would be interesting to know if the yaw drift is lower for those flying in lower latitudes (closer to the equator) since the errors from the accelerometers isn't that sensitive with lower magnetic dip angle.
/hakan