Another problem with circles

Discussion about the arduino based g-code interpreter, grbl
Post Reply
Posts: 8
Joined: Fri Jun 03, 2016 12:43 am
Location: Australia

Another problem with circles

Post by NightRoad » Tue Jun 07, 2016 11:38 am

I searched the forum but I couldn't find a solution, so I decided to ask if anybody has seen and solved this peculiar problem before.

I have a small milling machine controlled by GRBL v0.9j. It carves lines at any angle quite well (using G1 command). Precision is also very good - the control point always returns to the same spot. But as soon as I start using G2/G3 commands there are aways visible problems.

The developers claim that as long as the arc tolerance parameter ($12) is set to a small enough value the inconsistencies should be unnoticeable. I ended up using extremely low value $12=0.0001mm and in most places the circumference of the circle is indeed very good. But look at the picture:

While the inner circle is perfect, the outer has flat regions parallel to main axes. To prove that it is not an optical illusion I measured the diameter and it is indeed smaller in vertical or horizontal directions. How this is possible without quite complex trajectory involving Z-axis, I don't understand. I'm using a 2.4mm engraving cutter, but I think it shouldn't matter. The material is solid timber and the speed is about 5000RPM. G-code lines went directly into GRBL using a simple serial port sender (command line, no GUI, and I'm sure it sends only what's submitted). I fed commands line by line waiting until the previous finishes executing before sending the next.

These are the commands:

Code: Select all

G17 G21 G90 ( use mm)
G00 Z2 (above the surface)
G00 X-8 Y0 (move to the start point)
G01 Z-0.3 F30 (down to cutting, slow feed)
G02 X8 Y0 R8 F30 (first CW semi-circle)
G02 X-8 Y0 R8 (second CW semi-circle)
G03 X8 Y0 R8 (first CCW semi-circle back)
G03 X-8 Y0 R8 (second CCW semi-circle back)
And I did several runs for deeper cut, but they all followed the same path.

I wonder if anybody can explain what could be wrong and what else I may look for? Thanks!

Posts: 376
Joined: Tue Aug 12, 2014 2:16 pm
Location: Albuquerque NM, USA

Re: Another problem with circles

Post by chamnit » Tue Jun 07, 2016 2:10 pm

Lead developer of the Grbl CNC project here. So, the arc tolerance setting configures how short to make line segments that traces out the arc. Since arc radii can be different, the line segments need to scale with radius to keep the arc accurate. This is what the arc tolerance setting does. I would recommend that you keep it relatively close to the default value. If you make it really small, you'll start to effect performance through the arc, because Grbl will be generating thousands of tiny line segments to achieve your desired tolerance value.

That said, the arc tolerance setting may make an arc "look" faceted, but the arc dimensions will be traced to within the value. So no part of the arc should be out of spec. The reason it looks faceted is because the human eye can easily see sub-micron-level (<0.00004") surface details, while a CNC machine can only accurately machine down to a handful of microns (0.0001") at best. If "look" is more important than dimensions, then you can reduce the arc tolerance setting. However, I wouldn't go much lower than the default of 0.002mm.

Back to your issue. Your problem looks like backlash in your machine. Whenever the machine changes direction, you have some slop that the steppers have to rotate through before it starts moving the tool. Check the tightness of your belts (if you have a belt-driven CNC). If you have a leadscrew style machine, then you'll need anti-backlash nuts.

Posts: 4442
Joined: Thu Jul 19, 2012 6:50 pm
Location: Camberley, UK

Re: Another problem with circles

Post by cvoinescu » Tue Jun 07, 2016 2:22 pm

With @chamnit on this one. Just adding that, if this is a belt-driven machine, very often we see loose set screws on the belt pulleys, and occasionally loose idler wheels, too.
Proud owner of ShapeOko #709, eShapeOko #0, and of

Posts: 148
Joined: Fri Jun 05, 2015 4:19 am

Re: Another problem with circles

Post by DRobs86 » Tue Jun 07, 2016 4:53 pm

I will throw another vote on machine problem as chamnit described.

Posts: 8
Joined: Fri Jun 03, 2016 12:43 am
Location: Australia

Re: Another problem with circles

Post by NightRoad » Wed Jun 08, 2016 9:57 am

Hi chamnit,

Right,that makes perfect sense. And explains some other irregularities I got earlier. I made a search on internet and there are several reports that the CNC miller that i have (Proxxon MF70) has problems with backlash, which can be as high as 0.1mm in the worst case. I think I would be able to prove it by cutting straight lines (using just G1) with very shallow angle to the main axes, back and forth.

Well, while I know a bit of programming there is still lot to learn in mechanics...

Thanks to everybody who replied!

Post Reply