Worm gear mounts, even premium mounts, have backlash, because you always have some tiny gap between the trailing edge profile of the worm wheel tooth relative to the trailing edge of the worm thread. If these were always in contact then the worm could not turn if there was any dimensional error at all. You can't get out of that without using a split wheel or a deformable worm or a torque preload motor which none of the "premium" mounts have. Larger observatory mounts often use a large DC torque motor to create preload against the worm for exactly this reason. Direct drive and a couple other drive designs eliminate backlash. In a worm gear, the stiffness of the drive is ultimately limited by the gear tooth profile and width, but a direct drive mount has "active" stiffness. Encoders on a worm gear driven mount only partially improve the situation because while they eventually provide backlash compensation, there is a time lag between when a disturbance force occurs and when it is possible for the mount to react, if the disturbance is reversing load. The drive system has to see the upset, then reverse the motor, and in doing so unwind all of the backlash, and then start the compensating motion. In direct drive the compensating motion is happening all the time at the same rate as the servo loop is closed.
The plot below shows what a direct drive can achieve in terms of dynamic performance (during slewing no less) with a reversing load.
In the plot, the blue line is the commanded profile which is a rate-limited trapezoid which is why it has rounded corners. The vertical axis on the left side is encoder tics for a 26 bit encoder and the right side, which corresponds to the green line, is the position tracking error in tics. There's about 52 tics per arc second. The profile produces about 13 degrees of movement in 1.7 seconds, and reverses, and repeats continuously. The mount shows that it only has 15 arc seconds of following error when the direction reverses, and that is only for a small fraction of a second; the rest of the time the tracking error (this is while slewing mind you) in the arc second range. You can't achieve that kind of performance except with a direct drive mount.
Edited by 555aaa, 15 October 2019 - 01:31 AM.