Issue with step servo setup - x-axis cutting skewd

More grasping at straws:

I would almost think a loose coupling could do this but it seems so repeatable which makes this not too likely.

Since this is always on the X axis, maybe it would be worth the trouble to bring the X axis encoder into EMC and see how well the step/dir drives are following the commanded position. Also what is the value of the following error limit in the drives?

Dynamic (acceleration dependent for example) errors could be tuning related. Any amount of integral term in the drives control loop will zero out static errors (and thats all you measure manually), but the drives PID loop must be properly tuned for dynamic performance. Monitoring some x axis moves in HALScope, looking at commanded position and encoder position (scaled so they are the same) would allow checking the dynamic drive performance.

Could this be the result of a tilted Z Axis and slightly different heights for the various cuts?

Does anyone have an example configuration using encoders for position feedback when driving with step/dir signals? I want to continue to run this in position mode and let the servo drives handle the auto tuning and position, but I want the control to error out if the drives lose position as they have been with these mechanical issues. The encoders are 2000pp/r, and I would be hooking them up through the mesa board.

As for our original problem - we think we've solved it by finding all the mechanical bug-a-boos that were hiding in the axis - and she's purring now - no more squaking, knocking, banging, etc. 7 of 9 bearings in the drive system (it's a centrally mounted drive that turns drive shafts to each side of the machine) were bad. We are planning on test cutting monday.

Theres at least one example of a closed loop step/dir system in this forum (step motors driving an axis with a linear scale for feedback)
Unfortunately the forum search does not work for me or I might be able to find it
(first page is OK all subsequent pages are blank)

I think the trick was setting up the hardware stepgen in velocity mode. and using a PID loop with FF1 = 1
and a small P term to correct timebase and DIR setup/hold delays.

Guys, we set BACKLASH = 0.0047 in the .ini file in the x-axis config area, but it doesn't seem to be taking it up. No matter what we set the backlash to, it's consistently reading 0.0047. Any ideas?