During the flight tests of our quadcopter (Tarot 650 frame, Tarot 4114 / 320 motors, pixhack FC) we've encountered quite strange behaviour with regard to yaw control: in position control mode the 'copter at some point yaws approximately 90 degrees in counterclockwise direction and then rotates back into original position, all by itself (sticks are centered, etc). The rotation takes about 1-1.5 seconds and looks quite smooth. Here's the flight log of such a flight: http://logs.px4.io/plot_app?log=724f8712-13cc-41a6-b8c1-fcc349a431a9
We've tried doing a short autonomous flight in mission mode and this behaviour manifested itself like this: while in flight the 'copter kept rotating in counterclockwise direction at a steady pace. It went through all waypoints and landed itself with acceptable accuracy, though. Here's the flight log of the flight in mission mode: http://logs.px4.io/plot_app?log=dc7493ea-d359-405d-abd7-7e5c4019954e
I've tried flying the 'copter in stabilize mode and I've never noticed any problems with spontaneous yaw. Flight log of the flight in stabilize mode: http://logs.px4.io/plot_app?log=540e6335-a755-4707-9538-229c2d5bc823
We've calibrated all sensors (including magnetometers) a couple of times with no effect whatsoever.
What's even more strange, we actually did tests on two different locations on that day (separated by about 10km) and this yaw problem only manifested itself at the second location. The flight log from the first location, when there were no yaw problems, looks like this: http://logs.px4.io/plot_app?log=25f46bfb-4389-40e7-b06d-07f96094121e
What might be the cause of such behaviour and which actions could be taken by us to fix it? It seems, it might be something connected to the magnetometer(s)? We haven't tried swapping out the GPS/compass yet, could that help?