CNC Screw Accuracy

[macardoso]

Also I'm very impressed by the zero backlash. Could you share the source of your screws? Mine are Roton and have a ton of backlash from the factory. I've repacked them, but it is a bandaid.

 

[spumco]

Alternatively, is there anyway to use wear compensation to account for dimensional issues in only one axis? If it were consistent in both x and y I could just modify the effective diameter of the tool, however, the y axis is dead nuts on

No. Cutter comp doesn't work this way. It'll add/subtract the comp amount and result in an offset move in both x and y.

If you haven't already sorted this out or bought the glass scales, I have a suggestion for mapping which may help narrow down the culprit.

What you're going to do is use your DTI and a 1-2-3 block or gauge blocks to measure actual distance and compare to the commanded distance. Forgive me in advance if you already know how to do this...

1. Set table in middle of travel with indicator on head.
2. Set a 1-2-3 (or stack of gauge blocks) on the table lengthwise and parallel to the table travel under your indicator. Set another block on edge behind it relative to the X travel direction, and lightly clamp the rear block so the two blocks don't move. You are basically creating two vertical surfaces exactly 3" apart.
3. Adjust the blocks and indicator so the tip is touching the edge of the lower block, zero the indicator and X-axis offset in your controller.
4. Jog up in Z to clear the indicator tip and jog X until the indicator is zeroed on the other block. Check the machine DRO and write down the difference between the reading and your block dimension.
5. Just repeat the process along the X axis over the whole travel every inch or so - just shift the two blocks. You can write the numbers down right on the table in sharpie.

Once you have a bunch of numbers down on the table you shoudl be able to see pretty quickly if the issue is:
1. Ballscrew lead distance varies over the length
2. Ballscrew lead is constant, but not to spec
3. Ballscrew is fine, but your steps per unit of measurement is off.
4. Some other voodoo, like a warped axis or other mechanical gremlin.

If it's #2 or #3 you can compensate in the software. If it's #1, you can average it out in software or write a macro which scales the X axis at various machine coordinate locations.

If it's #4, let us know and we'll try to find a tame priest for you.

-Ralph

 

[benster]

Also I'm very impressed by the zero backlash. Could you share the source of your screws? Mine are Roton and have a ton of backlash from the factory. I've repacked them, but it is a bandaid.

They are the double ballnut versions Chai sells. I think I just got lucky. The first set of screws I ordered from him in 2016 had more backlash. The preload spacer on this new set seems a little tighter. I couldn't be sure on the 2016 set since I had lost some of the balls during assembly and had to repack with new balls, potentially affecting the preload. They were very bent but I was able to work most of it out. With the first set of screws I also didn't have a way to machine the ends accurately so ordered them machined. This go around I was able to machine them myself for a perfect press fit for the bearings and better thread fit on the lock nuts.

Ralph,

I still need to do this. I've been busy on another project but want to use the cnc for a portion of it. Hopefully I'll get to it this weekend. The warped axis is a thought. Even though I've straightened it the x screw is ~26" long. Its supported on both ends and could be getting warped when I clamp the x nut to the saddle. I try to tighten the bearing and nut mounts in such a way to avoid this but its definitely possible. The worst warpage would be at the center of travel (where this error is) since its the farthest supported distance from either end.

 

[spumco]

Even though I've straightened it the x screw is ~26" long. Its supported on both ends and could be getting warped when I clamp the x nut to the saddle.

A warped ballscrew might do it, but if the nut mount plane is far enough out from the screw mount plane(s) it would bind near the ends badly. If you dont' have binding at the travel limits I don't think the screw is causing the travel distance issue. Shimming the nut or screw mounts wouldn't hurt, of course.

I was actually thinking of the table, Y-saddle, or machine base being warped or influenced by however you've mounted it. Or all three.

Once you do the 1-2-3 block tests and map out everything in both (or all three) axis, if you can't adjust for it in software reliably, check out a good video on mill straightness. This is a good starting point for chasing down strange geometry problems:

Cliff has some excellent videos and this one in particular is good for understanding mill geometry.

-Ralph