A previous experiment on AZ/EL linear trajectories shows AZ behaving differently as EL does (see that post). Resolvers are located at the interface between the mechanical and electrical parts of the return chain. If something goes wrong somewhere, the estimated position will be inaccurate.
AZ is our bugbear. To figure out what’s going on we ran today an experiment with EL constant and AZ moving slowly at a constant speed (0.002°/s).
The next picture shows 3 curves. In blue successive AZ positions of the antenna during a constant speed trajectory (scale on the left). Time is on the x-axis. The green curve shows the deltas between two consecutive outputs of the AZ resolver. The red curve (scale on the right) is an estimate of the AZ speed (Savitzky–Golay filter).
Update: André added a mechanical system to log the AZ movment in parallel with the resolver. Verdict, PB8 is moving, as expected… We have to give more effort on this issue.