Can't mill a round circle to save my life

[dhanger]

Okay, I'm throwing in the towel and looking for help on this one. I have two mills running UCCNC 1.2113 with UC300ETH, and I cannot mill a round circle on either one. I have been programming complex mold surfaces for 20 years so I'm no rookie programmer. I've built my own cnc router from the ground up and refurbished 2 mills with new motors and electronics, so I'm no rookie in that department either. But....I must be making some kind of rookie mistake that I just can't locate. I've been inclined at times to suspect the software but I have to assume something this elementary would have already been identified and fixed. I even tried downgrading to 2 different versions and got the same results.

I've attached the program I'm running (about as simple as it gets) and my profile. It's a 1.500" circle using cutter comp (I've tried it without comp. and still get the same results). The circle comes out exact dimension in line with the axes but over and under size on the diagonals by approximately .010", so I'm getting an ellipse with the axes rotated 45 degrees. I can mill a square/rectangle all day long and get accurate dimensions. I've tried different size circles and the error seems to scale with the dimension (the smaller the circle the less the error). One of my tests was to halt the program on one of the diagonal locations, then back plot in CAM by a couple thousandths at a time to the same location and compared the DRO with the CAM coordinates and it was perfect.

I tore apart one of the mills to check for anything loose on the motor couplings, ballscrew, bearings, you name it. Nothing. Tight as a drum.

Can someone please humor me and run my program and just confirm that you're getting a round circle or not?

I know I'm going to end up with egg on my face and embarrassed, but I've been chasing this around for several days now and just have to give up. What's weird is that I've been running one of the mills with this configuration for over 2 years now and never noticed this problem before, which is one of the reasons I downgraded the software to see if a recent change introduced a bug, but as I said that didn't clear up the problem.

Thanks,
Dan

[cncdrive]

If the circle ovalness is proportional to the size of the circle then it is probably that one or both XY are not calibrated properly.
So, I would verify the steps per calibration with measuring running different distances.
For example run 100mm measure if it is 100mm or 100.1mm or 99.9mm or ... ?
Do the same with larger distances too.
And do the same on both axis.
If you find a misscalibration then adjust the steps per value.

You can also use the axis calibration button, tand try to calibrate it on an as long distance as possible to have the best accuracy.

[dhanger]

Yup, I've done all that already and haven't found any issues with my hardware setup. Anyway, as I said I can get perfect (at least within tolerance) dimensions in line with the axes at any place on the mill table, so there can't be any issues with stepping, etc. I'm running closed loop steppers on one mill and a combination of one servo, one closed loop and one open loop stepper on one (my current project, that's why I'm trying out different motors). The mill with the closed loop steppers in the one I mentioned running for over 2 years now, so I know I'm not having any problems with lost steps.

[dhanger]

The '2 year' mill is one that I actually sold to my current employer and I use it daily to surface mold cavities that use up nearly all the travel on X and Y axes, and it is very accurate over the entire travel (except for simple round holes, LOL).

[spillage]

Could backlash compensation be playing into the error?

[dhanger]

Could backlash compensation be playing into the error?

It's a good point, but I tried that. I re-balled all the screws on both mills and got backlash to under 0.0005" or better. Even so I had backlash compensation turned on for both X and Y, so I turned them off to test it since the amount of error was far greater than the backlash. No difference.

[Dawid92]

Did You looked in closed loop stepper config? Maybe there are some way out of spec values for accel and decel??

[dhanger]

Did You looked in closed loop stepper config? Maybe there are some way out of spec values for accel and decel??

Interesting thought. I've been using the closed loop steppers for a couple years but never connected the tuning software as everything seemed to work fine out of the box. Maybe there's an issue with interpolating XY motions that doesn't show up on small scales but does on larger moves? Do you know what parameters I would need to study related to that?

Your suggestion gives me a thought for my next trial (besides the tuning software): I'm going to re-install the original stepper motors from the OEM and run it in open loop mode. The original motors have encoders and they were implemented in the original electronics, but I can just run them in open loop which would show one way or the other if the closed loop motors have an issue.

For a while I was really suspicious that there might be a bug in the G2/G3 moves in UCCNC so I tried an experiment on that line--in CAM I converted the circle to a spline and then re-posted using point to point moves with a high tolerance to mimic a round circle, and surprisingly I got the same results--an ellipse, not a circle. That pretty much rules out a software problem and points to hardware/electronics, but so far I've had nothing but dead ends in my investigations.

[dhanger]

Wouldn't I be getting a position following error from the closed loop steppers? The following error should be checking on a far smaller scale than the error I'm getting, which is in the range of +/- 0.010" on the 1.500" diameter circle. I don't see how the driver could allow it to get that far out without an alarm.

[Dawid92]

There is a option you can set error in pulses, from factory in case of mine drivers it was set for 4000, which is one full revolution of motor (encoder 1000 ppr x 4 ), so You would have to crash hard to stop the motor .
Here is some stuff that may help https://www.haascnc.com/service/trouble ... mills.html