Adaptive Backlash Reduction (well... almost)

[rgsparber]

This article presents a partially worked out means of automatically reducing variable backlash in a CNC system. All was going just fine until I ran right into a bit of reality. However, others will hopefully see a good solution to the problem that blocked me and the scheme can be reduced to practice.

If you are interested, please see

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

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

[Workshopshed]

Hi Rick,
an interesting idea and definitely worth investigating. I'd be concerned that dirt and other containments on the screw might affect the reliability.

I've seen variations on this using a digital readout to get the absolute position of the moving parts. You might also be able to look at the current flowing through the motor, with a low load that could be less than when the motor is actually moving the parts.

Also perhaps a load sensor in the thrust bearings of the lead screw might work?

Good luck with your experiments.

Andy

[C-Bag]

Hi Rick,
an interesting idea and definitely worth investigating. I'd be concerned that dirt and other containments on the screw might affect the reliability.

I've seen variations on this using a digital readout to get the absolute position of the moving parts. You might also be able to look at the current flowing through the motor, with a low load that could be less than when the motor is actually moving the parts.

Also perhaps a load sensor in the thrust bearings of the lead screw might work?

Good luck with your experiments.

Andy

I'm glad you jumped in here Andy. This like so many of Rick's journey's into realms I know little about I don't feel qualified to respond to. But my take on it was similar. I've contemplated how to get exact coordinates in the x and y axis with the sloppy lead screws on my mill and it all comes down to a DRO. Now how that can be interfaced with a program like Mach3 is beyond my pay grade. That would seem to be the only way to truly know exactly where you are irrelevant to the slop in the leadscrews. Does anybody use DRO output instead of motor pulses?

But all this is why I backed off on getting a CNC plasma table. When you started getting into the servo vs stepper etc etc my brain started to swim. Also looking into people converting existing mills to CNC and the expense of ball screws, adapting them and then the motors and interfaces makes me salute them for their tenacity.

[Workshopshed]

I found this comment about using a DRO with Mach3, it's only used for display not for feedback

can DRO be used to make closed loop system?

This is seconded by another respondent

http://www.model-engineer.co.uk/foru...s.asp?th=96662

[C-Bag]

that really explained a lot. The two systems are really speaking different language essentially. The Mach3 generates everything internally and outputs the #'s to the steppers. And that's why there can't be any play and why ball screws are used. But it's also why there can be so much problem with precision because this stuff runs in a hostile environment.

I can also see why at the speeds CNC wants to run outside loop as in a DRO would be slower. But it makes you wonder at the speeds the outputs on the DRO and the inputs to the software are running now how slow it would be in reality?

[PJs]

Hi Rick, I've followed your great work () for some time now and this post has inspired me to chime in my 2 cents worth. The idea you have for the adaptive system is quite good but has its limitations as you point out. The other thing that came to mind is if you needed to perform an operation at a float point then your system would automatically push or pull to the no backlash point although by your measuring seems to rest in the bottom half. Also steppers are kind of a pain in that they tend to fall off position under load, which can lead to other issues, so I tend to lean toward servos. Using some kind of quadrature add-on system might give some advantage but adds to the complexity and interface although I believe M3 has a quadrature interface option. The real issue to me is all the variables in the screw, nut, end play, concentricity, bow, etc. and it occurred to me that maybe approaching this more from the mechanical aspects (as you have been) would give you more leverage on the tolerance stack.

If you go to the trouble of making a 2 material split nut, why not build or buy a ball screw nut that would fit the pitch of the screw. They still have an inherent slop but much less than the standard nut. If the balls could match the pitch of the screw then possibly it could be split horizontally (instead of vertically) and spring load the top half (not sure if that's possible in your nut mounting) to be more like half nuts that force the balls into the pitch of the screw. I've seen split ball nuts somewhere in my travels but slept since then...and have CRS sometimes. ~¿@ This I think would give you constant contact with the balls to the pitch over the entire run, except for excessive bow slightly larger than the width of the nut. Also not sure how it might be affected by high travel speeds. Hemingway, my 2cents. Great post and got me thinking too!! Thanks, ~PJ

[Jon]

Thanks Rick! I've added your Adaptive Backlash Reduction Method to our CNC category, as well as to your builder page: Rick Sparber's Homemade Tools. Your receipt:

Adaptive Backlash Reduction by Rick Sparber

tags: CNC

[DIYer]

An interesting idea, Rick. I'm waiting to see if you can come up with a definitive solution for your problem, or adopt a different method.

[rgsparber]

The method I am using right now is to precisely align my X and Y lead screws and use modified bronze take up nuts. It gets me a backlash of 2 thou on each axis with no binding. I'm releasing articles describing this work as time permits. The Y axis is out now. Will release the X axis nut mod tomorrow morning and then only have the X axis leadscrew alignment description to do.