Thomas Sanladerer looks like folks are having some success with threadless ball screws.

@Thomas_Sanladerer looks like folks are having some success with threadless ball screws. For z axis anyway.

Obvious caveat that this is early days so who knows how reliable it will be with repeat use.

I don’t imagine it would be unreliable. The bearings are being run on the axis they’re made to have load distributed in and there’s little to no variation in the shaft diameter to make the plastic flex and weaken over time.

Yeah, my thoughts too. If I wasn’t aldeady done with the belt drive Z axis design for the Ingentis, I might have considered this as an alternative…

My main concern with this idea is running it as a positioning system with an open loop control. I’m not confident in the repeatability of a system that relies on tracking perfectly due only to friction. Not that I would mind being proven wrong. The system is elegant. I would just like to see the loop closed.

@Dale_Dunn I think we’d all like to see the loop closed - has anyone come up with a solution for closed loop on a reprap yet? I’ve seen speculations about using micrometers or laser mice but nothing working…

I share @Dale_Dunn 's concerns, and I’m also not confident that the angle of the bearings can be controlled well enough for precise control of the steps/mm value. You could do the oldschool (in that it used to be common, though it was always WRONG) empirical steps/mm calibration assuming you only have one per axis, but it would be no good for the Z axis for the same reason non-metric screws are no good, only more so. Closed-loop would solve the former problem, but not the latter.

@Whosa_whatsis so, this is the rounding errors problem arrising from the fact that even a small error in the empircal measurement results in a regular variation in Z distance moved right?

Yes. You need you layer height to be a multiple of the full-step length for your axis to avoid those errors. If your steps/mm has to be is not controlled by a manufactured part (screw, belt, etc.) with a known pitch, it has to be empirically adjusted, and is almost certain to be an irrational number that doesn’t divide evenly into anything, and the only way to come up with acceptable layer height numbers is to round-off the decimal in firmware and multiply it to come up with layer heights.

Understood. However, what I don’t get is why you can’t work out the full step distance for this empirically measured value and then use multiples of that value (non integer) divided by the steppers microsteps as a layer height, even if that is some 5 digit number?

Technically you can do that, assuming your slicer will retain that many significant digits. Even so, at 5 significant digits, you’re rounding. Also, all your slicer profiles will have to have layer heights with that many decimal places. Give me a 4- or 5-micron full-step length (exactly) any day.

I can see how it would work for Z since that’s the only axis that typically sees no reverse movement. However, one side might end up behaving slightly different than the other, resulting in (ever so slightly) skewed prints and very likely a tilted X-axis over time if the each side isn’t homed individually.