Newbie- Debugging axis drift

[klk64]

Hi all, been lurking on and off for several months now. Finally got around to building my first machine with the "just do it already" method (resulting in the machine having little idea about what it wants to be), and have gotten to the point of cutting some 1/8" to 1/2" plastic.

My problem is that I have noticeable drift in my X axis (but basically none in Y or Z) and I can't quite figure out why. For example, with 1/64" passes, after cutting to the bottom of a 3/16" piece of plastic, it's typically off 2-3 mm, not exactly repeatable. Can't remember a time when the drift wasn't there, either, although it hasn't always been in the same direction, and sometimes it wasn't consistently in one direction during a single build.

Some info about my machine if it helps:
- X axis, as with Y and Z, runs on bushings on unsupported hardened steel shafts. Specifically, the X axis has 2 bushings on each side on a 3/4" dia. by 24" long case hardened shaft.
- The stepper motors on all axes are Keling KL23H276-30-8B, with 282 oz/in torque, wired in the series bipolar configuration
- The leadscrews are black-oxide coated 1/2-10 Acme from McMaster. The leadnuts are the matching ones from DumpsterCNC.
- The motor is attached to the leadscrew by a helical shaft coupler. There are thrust bearings and radial bearings involved (one on each side of the leadscrew)
- The motor drivers and controller are custom hardware and software (for the learning experience, I'm an EE/CS student), Y/Z driven by A3982 chips, X was also but now has a A3977 based driver (no difference in drift)
- Here's a pic of the whole thing, I can take more pictures if needed


I've already done a few quick tests:
- I switched up the X and Y axes in software - that is, the logical X was the physical Y and vice versa - drift was still only in X
- I've loosened and tightened the shaft collars that push up against the thrust bearings and guard against backlash. Drift was basically the same.
- I've loosened the bolts that hold down the rails and the bushings to check if it was an alignment problem, with no change that I could see.
- Turning up the current on the motor drivers, switching out the motor drivers, neither changed anything.
- Running it without cutting anything gives pretty much the same amount of drift as far as I can tell

So where I am now is: I know of several tests that I could do, but rather than run around like a headless chicken trying to interpret a handful of data points, I figured I ought to ask some people who know what they're talking about. So, what kinds of tests would you suggest I run (or re-run) first?

[wendtmk]

From just a quick gander, I would suggest maybe racking is taking place. Your gantry doesn't have much in the way of cross bracing to stiffen it up and may be twisting as you move it along the X axis.

Mark

[klk64]

I rebuilt the machine last night as a fixed gantry, moving X machine, and the problem still manifests in the same way.

I've further narrowed it down to some sort of mechanical issue related to the segmentation of motion into steps: gcode to draw a square of side length 1" drifts 4 times slower than if the side length is 0.25". I was under the impression that backlash wouldn't cause this; am I wrong?

[jsheerin]

No, backlash should not cause drift. I'm not totally clear on your last comment. It makes me think that your axes are not calibrated correctly, but I don't think that's what you mean... Can you verify that a commanded move of 1" repeatably moves 1"?

To check backlash you could move 1", then back to zero. Before the move zero a dial indicator and see what value it returns to. If you don't return to zero, that is typically backlash, but it will be repeatable. If you go past zero and then return to zero in the same direction, you should end up exactly at zero. Anyway, go back and forth for a while and see if you drift away from zero or keep the same amount of backlash. Then do the same thing but travel a larger distance and see if things change. From what you're saying, the results would change?

Also, if you only changed the pin assignments in software to swap axes, try trouble shooting that more. Swap the cables to the x and y motor drives and see if drift stays on the x axis. Then try swapping motors. Trouble shooting problems like these is generally best accomplished by dividing the system into halves, figuring out which half is causing the problem, and then breaking that system into halves, etc. You'll keep narrowing in on what the problem is if you do it right. Although a lot of the time you trick yourself into thinking something is fine when it's not...

[klk64]

The axes could probably be aligned better but moving things around doesn't seem to change the magnitude of the effect.

I'm not in a position to properly measure backlash (don't have a dial indicator or calipers that are at all accurate), I'm certain that there is some and I am not correcting for it at all. I doubt that backlash itself is really the problem.

I've done the back and forth test in the X axis, and it doesn't seem to be missing steps over the duration of the distance - the drift is the same (to the naked eye) after the same number of iterations whether the distance is 1" or 0.1".

I'll try switching the motors and/or cables when I find the time.

[PaulRowntree]

The X drive is also the one pushing the greatest mass about. Is it possible that the acceleration is too great on this axis? What happens if you slow it all down?
IMHO a dial indicator is absolutely required for this type of problem solving.