Centroid CNC12 Screw Mapping on a RF-30 Mill/Drill

[rgsparber]

The intent of Screw Mapping is to take near perfect ball screws and make them even closer to perfect. What happens when Screw Mapping is applied to the lowly leadscrew on a RF30 Mill/Drill? Leadscrews are far from perfect yet do benefit. The advantages and limitations of trying to turn a sow's ear into a silk purse are presented.

If you are interested, please see

http://rick.sparber.org/CSMX.pdf


Your comments are welcome. All of us are smarter than any one of us.


Thanks,

Rick

[tonyfoale]

Rick,
That's very interesting.
I note from the graph on p13 that if you remove the "random" errors there is a well defined increasing error which looks to be roughly linear at 3.5 thou/ft. That would seem to indicate a similar error in the machine used to cut your lead screw. If the screw was cut on a geared lathe that could be explained by either; the lead screw on that lathe was far from perfect or maybe it was of the wrong units (inch to metric or vice versa) and compound gearing was used in place of a 127 tooth gear. It might be interesting to compare the error that would occur with the commonly used compound ratios to the 3.5 thou/ft value.

BTW I am thinking of getting an Acorn for a lathe conversion. Also the Oak appeals as a conversion for my mill which has servos and linear optical slides. I currently use a Galil card for control. It is a very capable controller but it confines me to using Mach3 which I hate with a vengence, despite its popularity.
I'd fit ball screws and open loop steppers to the lathe and I'd follow your example and map those. The mill doesn't need the ball screws mapping but maybe I should map the slides?

[rgsparber]

Rick,
That's very interesting.
I note from the graph on p13 that if you remove the "random" errors there is a well defined increasing error which looks to be roughly linear at 3.5 thou/ft. That would seem to indicate a similar error in the machine used to cut your lead screw. If the screw was cut on a geared lathe that could be explained by either; the lead screw on that lathe was far from perfect or maybe it was of the wrong units (inch to metric or vice versa) and compound gearing was used in place of a 127 tooth gear. It might be interesting to compare the error that would occur with the commonly used compound ratios to the 3.5 thou/ft value.

BTW I am thinking of getting an Acorn for a lathe conversion. Also the Oak appeals as a conversion for my mill which has servos and linear optical slides. I currently use a Galil card for control. It is a very capable controller but it confines me to using Mach3 which I hate with a vengence, despite its popularity.
I'd fit ball screws and open loop steppers to the lathe and I'd follow your example and map those. The mill doesn't need the ball screws mapping but maybe I should map the slides?

Tony,

It never occurred to me to think about the source of the leadscrew's constant error. See the bottom of page 4 where my uncorrected leadscrews measured 20.02 rev/inch including my 2:1 gearbox. so just the leadscrew is 10.01 rev/inch or .39409 rev/mm. Multiply this by (13 + 0.0043)/13 for the residual error and I get 0.39436. A ratio of 101:256 yields 0.39453. So it is likely the constant error is due to re-gearing a metric lathe to cut imperial.

When I used Mach3 for small jobs, it worked good enough as long as I avoided those functions that were not reliable. But my last project had a 12 hour run time and Mach3 went nuts near the end and ruined the part. My theory, therefore, is that people's satisfaction with Mach3 is a function of the complexity of their part. Mach3 is a very impressive hobby quality CNC program. Centroid's CNC12 is commercial grade and has run multi-million dollar machines for decades. It is not fair to compare the two.

If you mapped the linear optical slides, it would certainly be interesting to see how they compare to gage blocks and a finger DTI.

Rick

[rgsparber]

I have reworked this article with more measured data and more theory.

Rick

[tonyfoale]

I have reworked this article with more measured data and more theory.

great stuff, very interesting.

[rgsparber]

Thanks Tony. You are looking at 8 weeks for work. I spent a lot of time stumbling around.

[tonyfoale]

Thanks Tony. You are looking at 8 weeks for work. I spent a lot of time stumbling around.

I am an inveterate tester and measurer so I understand only too well the effort that your work required.

[Responsible]

Hi Rick;

That is interesting. I wonder if a quick note about cosine error with DTI's might be a good addition. It could be a problem with larger error measurements, or if someone doesn't carefully implement their DTI in the setup.

Pete

[rgsparber]

If I understand you correctly, I only use the DTI to indicate 0 so there is no cosine error.

[Responsible]

Ah Ok! I initially misunderstood the section I had read where I thought it might be a concern. My mistake, you're right, it should be no problem.

Pete