Circle cutting problem

Post technical questions about the Buildbotic Controller here.
Forum rules
No profanity, no gambling, no illegal activity, so sexual or pornographic material.

Posts not related to the Buildbotic CNC Controller are likely to be moved or deleted.
Post Reply
retracile
Posts: 3
Joined: Tue Apr 27, 2021 3:25 pm

Circle cutting problem

Post by retracile » Thu May 06, 2021 6:30 pm

I'm attempting to cut some circles, and winding up with a shape that is not-quite-a-circle; more of a skewed, mostly-roundish shape. With a bit of imagination you might call it a skewed lemon shape.

Total newbie here; feel free to ask me the dumb questions, 'cause I may be doing something utterly stupid here.

Ran into the problem on firmware version 0.14.17. Still see the problem on 1.0.0.

The cut not-quite-a-circle appears to be rotationally symetric; cut it in half, rotate one half 180-degrees around, and the halves would match up.
It appears to be repeatable. Cutting the hole by helixing down, and cutting the bottom flat yield the same not-quite-a-circle shape. Cutting circles at a couple different sizes yields the same not-quite-a-circle shape.

The not-quite-a-circle seems to be elongated along a line parallel to Y=2X or something. It's not elongated along a single axis, so I don't think it's an incorrect thread-pitch in the config.

I ran the machine through a circle without cutting anything and it just looked "jerky" to me. It didn't seem like a smooth circular motion with both axis in constant motion; it felt like each axis would stop too abruptly in their motion instead of making that smooth sinusoidal swaying motion I would expect.

Cutting a circle with outside radius (including the 5/16" endmill size) of 0.75", the outside diameter varied from about 1.52" to about 1.47". (Inexpertly measured, but it's enough you can see it clearly.)

Very light, slow cuts in aluminum with a HSS 5/16" endmill.

Code: Select all

G20 ( imperial )
G90 ( absolute coords )
F2.500 ( set feed rate )
G0 Z9.700 ( move to safe Z )
G0 X2.2083 Y2.237
G0 Z4.340
G1 Z4.090
( Helix down )
G2 Z4.060 I0.4437 J0 P2
( Flat bottom )
G2 I0.4437 J0
( done )
G0 Z9.700 ( move to safe Z )
Changing the Path Accuracy from 0.0039" to 0.001" yielded no apparent change; it seemed to cut the exact same path.

This is on a Grizzly G8689 converted to CNC using the HeavyMetal CNC X2 kit, with 570oz-in 3.5A NEMA23 steppers on all 3 axis. My Y axis is my shortest, with only 3.0" of travel after the conversion.

If there are some diagnostic shapes I should cut to help narrow down what's working and what isn't, that might be helpful.

Any ideas?

retracile
Posts: 3
Joined: Tue Apr 27, 2021 3:25 pm

Re: Circle cutting problem

Post by retracile » Thu May 06, 2021 8:53 pm

Hm... I've stumbled across some articles on backlash and how that can affect circles. The irregularities in my circles don't line up with the X/Y axis, but I would not be surprised to find that's the cause.

So that looks like a promising lead to me. Other pointers also welcome, but I plan to start there.

User avatar
Doug
Posts: 343
Joined: Fri Feb 02, 2018 4:56 pm

Re: Circle cutting problem

Post by Doug » Fri May 07, 2021 3:19 pm

I looked at the kit you purchased on line and it comes with ball screws and nuts with backlash compensation so I doubt that is a source of backlash. If a motor coupler is loose, it could result in a problem like you described. You may also want to check that no other hardware is loose on your machine.

This is not likely to be the problem, but you should specify the XY plane with a G17 command and specify G91.1 for incremental arc distance mode prior to issuing your G2 arc moves. G17 is the default, but I'm not sure about G91.1.

You said that you wanted circles that are .75" in diameter and, by my figures, the command you issued 'G2 Z4.060 I0.4437 J0 P2' with a 5/16" diameter tool will result in a circle that is approximately 1.2" in diameter. According to the LinuxCNC documentation, the 'I' argument is the 'X' offset to the center of the arc.

The documentation for the conversion kit says that the pitch of the ball screws is 0.2", so the 'travel-per-revolution' fields on the Motor Configuration pages should be 0.2" or 5.08mm.

Finally, you may want to confirm that the drive-current is set correctly for your motors.

retracile
Posts: 3
Joined: Tue Apr 27, 2021 3:25 pm

Re: Circle cutting problem

Post by retracile » Fri May 07, 2021 8:10 pm

Thanks, Doug.

I measured the backlash on my Y-axis; it was a whopping 0.0265". I had a clamp that hadn't been tightened, and the nuts on the end of the ball screw were not tightened down. (Sloppy work installing the conversion, what idiot forgets to tighten these things... oh, wait, I did that install. *cough* ;) )

After addressing that, Y-axis backlash was about 0.006", and X-axis about 0.002". A test circle was slightly out of round by single-digit thousandths. Just looking at it, I can see a huge improvement. I don't know how much more accuracy is reasonable to expect, or how much it would take to get there.

So looks like that was the primary issue I had.

Looking at the g-code, yes, that'd do a 1.2" circle. I had run a few different sizes, and I clearly provided the g-code for one, and the measurements I took for a _different_ one. My apologies.

Thank you for the G17 and G91.1 pointers.

I'll also review my motor settings.

Thanks again!

Post Reply