The y-axis numbers in the plot are actual wall measurements (about 10-12ft away). I advanced the axis in 0.1mm steps, for three periods, and then used my calipers to measure each period twice (the legend Dn/m stands for delta-step on n-th period, m-th measurement, and last two curves are mean and median).
As you can see, with a freshly installed well-nut, the error is on the order of a few (~5) microns, so it's indeed below the accuracy of the dial gauge + mount (perhaps the wiki instructions should point out that dial gauge may not be accurate enough?). Also, it's not perfectly repeating, but taking avg/median of a few periods and measurements gives a clear pattern.
Still, even though it's a few microns, the z-wobble fix helps very noticeably. Here is the proof (full size):
Right is w/o fix, left is with fix, back-to-back prints, 0.2mm layers. FWIW, I used the cumulative sum of the above median values in my start gcode.
I hope this info is useful to others, and thanks for the great work on the firmware fix!
]]>
