G2 move has too much error

[SGoldthwaite]

I'm making a spur gear on my Shapeoko which uses a gShield board with GRBL firmware.  GRBL generates an error if a G2 or G3 command has more then 0.005" error.  GRBL is giving me this error for some gcode that Fusion 360 is generating.  You can find the Gcode here: http://pastebin.com/G6Ep09Ry.  Line 56 is the one causing a problem.
The error for this move is: 0.005007"
The post processor I'm using is a modification of generic grbl.  The modifications I made are not related to how it calculates G2 or G3 commands.

I attached my Fusion360 file.  The operation that generates this gcode is called "Teeth Finish"
I also attached the spreadsheet I used to calculate the error.

Are there some settings I can change to reduce the calculated error so it's less then 0.005"?

[SGoldthwaite]

Achim

I saw you posted a solution, but now it's gone.  Did you delete your post?

[AchimN]

Jep, the code i posted wasn´t working for your issue, working on it now  😉

[AchimN]

try this code in onCircular please:

  var radius= Math.sqrt(Math.pow(xyzFormat.format(cx - start.x), 2) + Math.pow(xyzFormat.format(cy - start.y), 2));
  var error= Math.abs(Math.sqrt(Math.pow(xyzFormat.format(x) - xyzFormat.format(cx), 2) + Math.pow(xyzFormat.format(y) - xyzFormat.format(cy), 2)) - radius) /radius;
  
  if (getCircularPlane() == PLANE_XY) {
    if (error > toPreciseUnit(0.0049, IN)) {
      linearize(tolerance); 
      return;
    }
  }

This will only work for XY arcs, what about YZ, ZX arcs?

Let me know if it is working.

[SGoldthwaite]

The post failed (see attached). 

I'm only doing x-y arcs right now, but I do want to do some 3D stuff in the future.

[AchimN]

The code was only into the wrong place into your post, thats why it failed, try this one please.

[Laurens-3DTechDraw]

Achim,
Where does this error come from?
Is the data from the system wrong or is the post just not using the info correctly?

[AchimN]

What i´ve seen yesterday is that this is only because of rounding. When I use the values without any formats the error is something like 0,000004532, so nearly nothing. Another question, linuxcnc also supports radius format for G2/3 which is often the better choise to avoid exactly those issues. Did anyone try this on a linuxcnc?

[Laurens-3DTechDraw]

Achim wrote:
What i´ve seen yesterday is that this is only because of rounding. When I use the values without any formats the error is something like 0,000004532, so nearly nothing. Another question, linuxcnc also supports radius format for G2/3 which is often the better choise to avoid exactly those issues. Did anyone try this on a linuxcnc?

Got ya!

[SGoldthwaite]

Achim wrote:
The code was only into the wrong place into your post, thats why it failed, try this one please.

It's working now.  I even generated some gcode that was causing an error on GRBL and it's ran fine.  Thanks!

[AchimN]

Ok perfect, thanks for the feedback. This code will be included in the generic post for future releases.

[fonsecr]

I just looked at this example. And the error for the move you list is only 0.0002" which is well below then your 0.005" limit. So not sure why your CNC control would error out for this case. Since 4 decimals are used you can expect a maximum error around 0.0002" when coordinates are rounded up and down, respectively.

Maybe the CNC is actually failing for a different move?

René

[SGoldthwaite]

René Fonseca wrote:
I just looked at this example. And the error for the move you list is only 0.0002" which is well below then your 0.005" limit. So not sure why your CNC control would error out for this case. Since 4 decimals are used you can expect a maximum error around 0.0002" when coordinates are rounded up and down, respectively. Maybe the CNC is actually failing for a different move?

Did you look at the spreadsheet with the calculation that's in the original post?

[Laurens-3DTechDraw]

SGoldthwaite wrote:
Did you look at the spreadsheet with the calculation that's in the original post?

I just put it all in a SW sketch and the biggest error in those value's I can find is 0.0006.

[fonsecr]

Yes, I used your spreadsheet. Just removed your divide by radius to get the error in inches.

René