After finally getting some long enough sleep I thought again on my failed attempts with a Melexis sensor. The sensor itself is quite precise and has some DSP onboard, so I tried it again with a spherical gimbal and noticed that depending on how close a magnet is to the sensor, it outputs not a circle, but a square, inclined to 45 degrees. I thought, “- what if I simply cut excess values and make it a square?”, and did just that. The idea behind it was that jitter of the point in a joystick tester sketch seemed minimal, so the physical precision of the sensor looked like being enough to work well. As long as we have a 15 bit external ADC that powers our gimbal, we should still have something like 4096 points per axis even with this reduced range, which is plenty.
Now, if we use a disk magnet instead of a square one, unarrested twist axis is not an issue anymore, it doesn’t affect X and Y positions in a big way. So this design seems to be perfectly fine for vertical sticks, or sidesticks, and I think with a pneumatic mod it can be as good for everything else. Its beauty lies in its mechanical simplicity, only a few parts are needed, it can be quite small in size.
I think I will make, test and publish both versions so everyone will be able to choose something for himself =) There’s also a third version, the 608 and MLX based one, its fate will depend on a degree of success of the spherical bearing based one =)
Meanwhile, some progress on the 608 and SS495A based version: