Doing some test, I found that at least one channel seamed not to work (Sensor 2, blue). Also by moving the whole know in the earth magnetic field, the signal did not

Then I printed a fast fixtured I had ten to tighten with tape. The fixture holds a 60m diameter 10mm high metal block. Fixture was not too bad, maybe a bit to stiff. I will have to design a better housing and then test again.

Then I did some more test, and contradictional to my first findings, there was significant less signal when I displaced the complete knob in x/y direction on the table. Only by rotating the complete knob I got some signals:
What still does not seam to be wrong is the transformation of the units. Convertig the rotation in all 3 axis I would assume to have about 30 to 60mT amplitude (the earth magnetic field. Values seamed to be much less, I might need to debug my test software here a bit more.