lothman wrote: ↑Tue Jul 07, 2020 9:07 amI think in the sub µm range the problem is no longer the motor but stick-slip phenomena of the spindel/nut. May be the ball screw of the THK KR series has an edge over all the thread screw spindels. But probably such small movements with high accuracy and repeatability are the domain of piezo positioners or thermal positioners.
Yes, if the encoder is attached to the motor shaft as in common rotary types (not linear types), then the encoder can't detect any of the stage movement, only the motor shaft rotation. So almost completely useless in reducing any movements by the stage and certainly no help on reducing backlash effects caused by the screw to stage interface (nut), which is likely the major source of backlash as you mention.
The THK KR types are probably some of the best available screw based rails, and use an elegant method of screw to stage engaging which produces little if any measurable backlash in typical systems at a few microns resolution, but no match for the precision of a properly designed closed loop piezo stage utilizing continuous strain gauge or capacitive feedback sensors. Of course the screw based stages can travel many mm and the traditional piezo stages are limited to much less travel, and the load capability of the screw based stages is much higher, so there's always a tradeoff.
An engineering way of viewing the encoder feedback location, if the position disturbance (backlash or other) is outside the feedback loop then it can't be detected nor corrected since the loop doesn't "see" the disturbance and thus can't compensate for it. With the encoder on the motor shaft it doesn't "see" the stage backlash, and can't provide feedback for correction by the closed loop system.