Its noise generated by the spindle motor, if I physically disconnect the spindle, it doesnt do it. It was fine with my 2 previous motors too. If you hold onto the screw it is intermittently jumping a few steps a second.
I will knock up some noise suppresion for the motor, probably just needs a little ceramic cap between the terminals and maybe a ferrite on the wires.
I seem to have this issue unde control too.
I added an additional 1uF ceramic to the Vlogic rail of my extra pololu, I figured adding a cap on the motor rail here could feed more noise from the 12v rail to ground and make the issue worse, it already has filtering on this pin whereas the logic voltage did not have as much.
Added a 100nF across the spindle motor terminals.
It helped but the z was still randomly stepping a bit, nowhere near as much but still twitching.
So attacked it from a second direction, added 6.8nF caps between step and direction input pins and ground on the Z pololu, problem seems to have gone away.
No idea why I chose 6.8nF .. it just seemed about right to me.
So here is my first milling test with a resharpened snapped stub of a 3mm cylindrical cutter on balsa with the chilipeppr test gcode. You can see the Z stayed uniform throughout.
Rapidly followed by yet another fail, on right I tried to go a but deeper, balsa shifted on my blutak 'clamp' then the chuck taper released and dropped into the work piece.
Also in that area but I may have missed with the photo is a 30mm long groove that got progressivly deeper along ots length, that was just before I added the caps to the step/dir logic inputs.