Well, I'm not an electronic engineer. But I have following thoughts:
1. For me the yaw movements happen at extremely high pitch points. I.e. I need to look mostly at ceiling to make it happen.
2. Something makes me think that this is not magnetometer's issue but gyroscope's one. Because the magnet fields are constant (especially from headphones) so their influence expected to be constant but we have them at certain points.
Logic would make think the same, but actual testing definitely points to magnetometer shenanigans. I just did some additional tinkering trying to make the tracker behave with the headphones.
Tried a calibration at medium sensitivity with the tracker strapped as in the photo above. Took an eternity to complete, the resulting graph gave a generally round green sphere, but with the red one (should be pre-calibration data) totally off and away from it. Tried in game and there was extremely annoying yaw drift at high pitch angles (workaround: if you look up for 5 seconds, remember to look down for another 5 and bam, centering again spot-on XD).
Then I tried a new calibration, with tracker detached from headphones but in the same spot relative to it, by rotating it on its axis. Calibration bar very fast, resulting graph a total mess. Doing the same just 5-6 cm higher, or in front of the headphones gave instead a very round sphere. Those K712 drives must have a hell of a magnet inside.
Then I attached the tracker to headphones again and tried lowering the sensitivity to a minimum; progress bar not even advancing. Scratch that.
Then I said "whatever" and did the opposite, sensitivity to the max. The progress bar completed in a matter of seconds without even moving the headphones! The resulting graph was no graph at all, just a simple green point at the center. I thought "this must be fun in game" and so I tried. Lo and behold, yaw drift at high pitch practically absent. Took off and did a bit of FA Off flying around the station to see if there were any spikes or weird behaviour, and noticed some things:
- it was fun
- damn, it was very fun
- the tracking was just about good, with only a bit of center misalignment once I got back inside, but I also did several exaggerated movements on purpose.
Need additional testing, but the more I do, the less I seem to understand how it all works.