The G540, is designed to run Active high , not Active low. So if I follow UCCNC instructions and set Direction to Active low, it means while the Direction may have reversed, the G540 is telling the Z axis to stop / go on the wrong side of the pulse from UCCNC software - as someone else said, it could create missed steps at the end of a direction change, and lots of them = lots of errors.

I ended up re-soldering the motor plugs, so that I can select active High, and still have motors going in correct direction. I believe - this + lowering my Kernal frequency from 200Hz to 50Hz, has fixed my problem, but cannot test until tomorrow.

On another note, I have been using Fusion 360 for the Cam work, and have a question about tolerance and how to get the fastest movement withing tolerance settings.

Fusion 360 has a pass tolerance, which basically makes the tool stay within tolerance distance of the finished product at all times when cutting. Then they have a smoothing feature which acts like constant velocity mode, and looks ahead to reduce the number of points, as long as it stays withing the smoothing tolerance, of the original tolerance. On my workpiece, this reduced the G-code length by about 40-50%.

Then - when I open the Post Processor to get the code out, it has it's own Built In Tolerance stated. No idea where it comes from or what it does. Is this the third layer of error? Ie it will now plot paths within say 0.01, of the paths I just calculated in fusion 360, and add to my other pass tolerances? Eg (Post processor)0.01 + (pass tolerance )0.01 + (smoothing tolerance) 0.01 = max error of 0.03 from the actual part surface?

Then I open UCCNC and it has velocity mode, so if I set that to a tolerance of 0.03, then now I could have 0.06 total error from the surface, or up to 0.12 error between two parallel passes if one has a low error ( into the part ) and the other a high error above the surface?

Am I better off turning smoothing off in Fusion360, - and using the UCCNC velocity mode to do the smoothing so only 3 layers of error instead of 4?

Nick