Re: [CAD_CAM_EDM_DRO] Backlash problem?

> Well, You are in absolute per the G90 command. So it is irrelevant what moves you make when you tell the machine to return to X0Y0 it should return to the same spot every time.

Agreed.

> The only time you can build program end position errors on repitition is in Incremental G91 programming and miscalculating the move to go back to the starting point. Never use Incremental.

As far as what you say about incremental and absolute, you can switch
back and forth as often as you wish. If you start in absolute at (0,0)
and end in absolute at (0,0) then you should end up where you started
with no accumulated error. As you suggest, you cannot accurately move
incrementally to a calculated position.

The thing that can mess you up is if you change your coordinate system,
using G92 for example, then you effectively redefine x0y0 as someplace
new. I do not see that anywhere in the poster's short program so that is
clearly not the problem.

> However, I'm looking at the code more closely. Are you sure Mach2 produced that code exactly like that? You start out in Absolute but your commands are in Incremental then it goes back to absolute??.

I don't see anyplace where G91 sets up incremental mode. As it sits it
is all in absolute mode.

> In absolute the radii are always called out with "I" and "J" values.

Not true. I often use X, Y, and R in absolute mode. The arc must be less
than a semicircle to use R reliably. For arcs greater than a semicircle
you need to use I, J to specify center and not R to specify radius.

In some Gcode interpreters I and J are always incremental values. In
other interpreters I and J are absolute or incremental according to the
G90, G91 mode in effect. In still other interpreters, the user chooses
from a menu to interpret I and J as incremental or absolute. I think
Mach2 allows the user to choose. But there are no I and J values in the
poster's Gcode program so this should be academic.

> If thats not it and you are using servos I'd be looking at possible encoder/motor lead cross-talking. It could be lost steps due to a certain place in the machining process that inducing a heavy load on the axis drive.

Steps are being lost someplace. Could be a software bug. Try sending the
program to the appropriate software folks and see if they can fix their
software or tell you what is wrong with your Gcode.

Try running the Gcode with a different Gcode interpreter like TurboCnc
or EMC.

> If it were a mechanical backlash issue it would still be out the same .033" every time from zero. Not .033" then .066" then .099".

Agreed.

Tom Hubin
thubin@...