CNC mill conversion - part 11

Backlash reduction and ball-screw inconsistency

In the last CNC conversion post, I measured the backlash to be 0.0042” in X, and 0.0049” in Y. While these values seem minuscule, they are fairly substantial in the CNC world. 

Backlash measurements using the original 0.1250" ball bearings

One pretty common way to reduce this problem is to repack the ball-nuts with oversize ball-bearings. The original were 0.1250”, so I replaced them with 0.1270”

Removing the ball bearings from the ball-nut

100 oversize ball bearings for $12 (times 2)

Y ball-screw being taken apart

Ball-nut repacking video (courtesy of Hossmachine.info)

After spending a couple of days messing with both axes, I did get some improvements out of them.  Unfortunately, they didn’t end up as close to zero as I was hoping, but they did initially go down to 0.0013” and 0.0027”. 

X and Y reduced backlash, after repacking both ball-nuts with bigger ball bearings.

Thinking that I could squeeze a little more out of the X axis, I purchased some 0.1280” ball bearings. 

Going for the "diminishing returns"

I was only able to fit 30 of the 0.1280” ball bearings in the Y ball-nut, before the the ball-screw locked up and stopped turning, so I ended up removing them, and using only 13 of the 0.1280”, and 53 of the 0.1270” ball bearings.

Packing the ball-nut

Ball-nut is filled up

Because there are 11 balls in the return tube at any one time, and I did not want to create a situation where I would have the majority of the big balls disappear in there, and have the ball-nut get loose cyclicly, I spaced the bigger balls at equal increments (1 big for every 5 small ones).

Packing the return tube

Return tube back in position (strap clamp missing)

I really should have bought an intermediate size, but there were none to be had that I could find between the 0.1270” and the 0.1280”.

No intermediate offerings available between the last two sizes

With this latest setup the ball-nut is pretty tight, much tighter than before, but it is still able to rotate consistently, although not freely anymore. Ideally, I’d like to reduce the amount of 0.1280” balls to maybe 7 or 8, in order to reduce the stress on the stepper motor, but I’m not about to take it apart again, at least not yet. The Y backlash is now down to 0.0020”.

Final backlash values

These are values that Mach3's backlash compensation should be able to handle easily, so I’ll call these two axes done.

My new issue is now with the consistency of the Roton ball-screw thread. In other words, the channel in which the ball bearings travel is not wound at exactly the same rate of 5 revolutions per inch throughout the length of the screw. 

This creates a new problem where the CNC controller might command a travel of... say 2.5000”, and while the screw will turn the appropriate amounts of turns, because of the inconsistently wound screw the tool might end up cutting at 2.5025”, for example.

Zero deviation represents the perfect ball-screw

This problem is entirely fixable by purchasing a more expensive screw (as always).

There is however a feature in Mach3 called “ball-screw mapping”, where one can record the actual travel achieved by the commanded movement, and train the CNC controller to undershoot/overshoot by the difference recorded.

I haven’t had much luck with it so far, but I’ll give it another chance to perform once the mill conversion is finished.