Lathe G2/G3 arc move path inaccuracy

I am running out of ideas.

Maybe you could log the XZ position with HALscope, then plot the points to see what the actual path is? Or have you done something equivalent to this?

I took a slow motion video of the X,Z coordinates on the preview tab while running the program.  I made a X,Z point cloud from the video and applied a best fit arc in CAD.  The arc radius measured .734".  This with a .016" insert radius should yield .750".  I zipped and attached point cloud data with both G64 and G61 evoked.

Are the coordinates displayed in linuxcnc a real time representation of the TCP?

Thanks! 

agdeutschman wrote:

Are the coordinates displayed in linuxcnc a real time representation of the TCP?

 

They can be, you have the choice of displaying commanded or actual position. 

Maybe try repeating the procedure with actual / feedback position displayed? 

Are you 100% sure that your Z scaling is correct? (I imagine that if it wasn't you would have noticed before now) but I have been caught out once by a 5tpi screw that I thought was 5mm pitch. 
(Though I just tried dividing the Z positions by 1.016 in the sketch, and it made the radius .7226, which is even smaller than you are seeing?)
 

Thanks, I will re-attempt using feedback position.

I feel confident in the Z scaling because my Z-lengths are all well within tolerance over dozens of parts.  It wouldn't hurt to put a travel dial indicator on the Z axis and test though so I will try it.