So, first off, this isn't a problem with backlash.  My circles are still round, they're just ... not where they're supposed to be.  They're consistently offset by about .08 inches in -X and +Y.

For instance, in the part whose g-code I've attached below, I have zero set at the inside left corner of the fixed jaw on my vise.  I've got a piece of 1/2x6 hot rolled steel clamped up.  

I drill out the center with a 1" bit, then a 5-bolt pattern around it with a 1/2" bit.  This goes as expected.

Then things go a little abnormal.  

The next operation takes a 5/16 end mill and bores the holes out by a few thousandths for clearance, then cuts out the whole thing.  The boring operation is offset away from the center of the actual hole.  So's the cutout.  Everything is round ... it's just not centered.

This has been a problem for a while now, but I've never really done anything where the precision mattered as much as it does here so it wasn't quite as obvious.  When I saw it happening with thread milling, I figured I was stuck on stupid with something in my operation and never put any more thought into it since I was just experimenting, but it's showing up everywhere else.  If I am just milling my whole part everything's fine - the offset takes care of itself because everything is shifted ... but when I'm working with something where I have to index off an existing hole, or where I have tool changes, then there's issues.

Anyone got any ideas what's going on?

Ok here is teh test. Take the first program and add teh second program to the end of it  then do the 3rd. Next take OUT the m30s Then load the total program Do you see the offsetting in the toolpath ?? NO then it your machine loosing or gaining steps.

(;-) TP

Ok here is teh test. Take the first program and add teh second program to the end of it  then do the 3rd. Next take OUT the m30s Then load the total program Do you see the offsetting in the toolpath ?? NO then it your machine loosing or gaining steps.

(;-) TP

The machine can't be gaining or losing steps - when I do the final cutout on the part, I'm cutting that circle out in about 50 passes.  If I was gaining or losing steps, then I'd be seeing a stairstep effect on the walls of the part and the remaining stock in the direction of the lost steps, and that's just not happening.  The walls are completely vertical.  This behavior is consistent through all the parts where I've seen this happen.

(;-) Those are famous last words it can't be the machine (;-)..  Did you run the test? does it show ANY offsetting in the the Gcode  toolpathing ??. You could be losing steps ONLY on a rapid not on G1/2/3 move.  There are a HOST of things that can cause a machine to gain or loose steps only in certain circustances. it seems you have found one of those circumstances (;-)

Just a thought, (;-) TP

(;-) TP

(;-) Those are famous last words it can't be the machine (;-)..  Did you run the test? does it show ANY offsetting in the the Gcode  toolpathing ??. You could be losing steps ONLY on a rapid not on G1/2/3 move.  There are a HOST of things that can cause a machine to gain or loose steps only in certain circustances. it seems you have found one of those circumstances (;-)

Just a thought, (;-) TP

(;-) TP

Good point.  No, I haven't had time yet to run the test.  I'm actually trying to figure out how to do it without needing a tool change.  (I don't have a tool changer or enough toolholders for it to be worth figuring out how to set up offsets.)  I guess a piece of wood would be okay - I can drill through that with an end mill.

Actually, waitaminute.  I've got a perfect example of a situation where the circle was offset and there were no rapids, no tool changes, and only a single operation / single program.

The other day I had a sprocket that I needed to cut the teeth off in order to use it as a flange instead of a sprocket.  (Taper lock flanges are stupid expensive, sprockets far less so.)

Zero on my program was center of the sprocket bore, top flat of the thing.  Only operation was a circle cutting the teeth off, no rapids or anything.  It was offset in the same way.  Part is round, but not concentric with the bore.  At the time I thought I'd made an error with the center finder.

You have to know I already checked your code it was fine there is NO coded offsetting from the G code.  That really only leaves the machine as the problem.

(;-) TP

Yes, I understand that.  I checked the code as well.

What I was hoping was that there was something in Mach that might be causing this.  IJK absolute / relative or something.

I simply can't see anything in the machine that would make drilling operations go to the right spot but make circles go to the wrong place.

Unless you verify the actuall positions from the point of origin you don't really KNOW if  any of it is going to the right spot it just appears that the 2 do not line up. The first operation COULD be wrong as well.

(;-) TP