Yesterday we ran an new experiment, moving PB8 at constant speed on AZ, and EL constant. We collected the positions out of the AZ resolver and stored the difference between the current positions and its target (it should be a straight line).
On the next picture, we show, in blue, the errors of the AZ trajectory and in red the signal logged during the last moon tracking:
The two curves fit perfectly ! That means that the periodic fluctuations of the received signal (in red) are exactly correlated to the errors on AZ. Second point, the blue curve is periodic, we can estimate the period to 15°. And 15° is exactly one tooth of the pinion that drives the resolver !
We know the culprit know, but we still have no idea how to fix it…