Steve,

I finished rebuilding my Y axis to be identical to the X, same counts per inch, brackets, etc..  The problem of losing position (thru DMC Smart Terminal) got worse and more erratic!  An electrical engineer I consulted with said he has seen this when there is a floating ground.  SO I plugged in an extension cord and took an Ohm meter from that ground to the motor body, router, frame, etc, all seemed well grounded.  The cable shields connection to my 100 pin break out box were also grounded.  The problem was the cable going from the 100 pin break out to the amps was shielded but not grounded, I grounded that reran the test in DMC terminal cutting the same hole 70 times and the size only grew .001.  Next in Mach3 programmed a a loop to cut a hole measured it, then put it in a loop to cut the hole 50x, again only grew .001. So the ultimate issue was a floating ground.

For somebody in the same boat of losing position here is a list of things I did when trying to fix this problem.
>Try a test cut in DMC ST or Galil Tools  (Problem showed there when I increased AC and SP, which rules out a Mach3 setup). Be sure that you have your Galil setup properly good PID, AC, DC etc and BN.
> Replaced all the cables with shielded and grounded cables.
> Formatted and rebuilt my PC. (probably should have been the last thing I tried, the PC does not have much effect on a Galil)
> Checked and rebuilt my mechanical drive to make it as good as possible.
> Tested the ground of all components.  (Found the shield on the cable going to the amps was not grounded, this caused my position loss issue).

All is well now, now I get to cut some cool stuff, can't wait!!
Thank you so much!
Matt

The electrical engineer also said be sure to only ground your encoder shield on the controller side not the motor side.  Just leave the aux encoder shield loose on the motor side.  Mine was already this way, but maybe that will help somebody in the future.

Matt

Follow up, cut a sign that takes 1/2 hour each last night on 3 different boards, they were all spot on, to the .001 of each other.  This sign in the past would get off from 1/8" to 1/4" over the length of the sign, cutting this same g-code successfully proves the open ground caused this problem.  If you have a position issue, I definitely recommend checking your grounds and use shielded cables with the shields grounded.

Matt

I had similar issues when setting up my Mach-galil machine. Everything was fine when doing 2.5D work, but when I did a 3D surface what should have been a rectangular file was cut as a parallelogram. The drift was about one inch in the Y for every 12 inches of X travel, always in the same direction. After weeks of pulling my hair out trying to eliminate mechanical slip and retuning the servos countless times, I narrowed it down to noise, which I think was being emitted by the z axis motor (hence the problem with 3D files) and adding spurious counts to the Y encoder input. After making sure all proper grounding recommendations were followed, and even replacing the z servo and cables ($$!) the problem was reduced but not eliminated (it is a commercial machine I converted so the wiring/earthing was well laid out anyway.)

I eventually solved it by swapping from an Ethernet controller (2160) to a PCI one (1860). Something to do with the galil having its own ground independent from the machine, I guess?

So 1. noise can indeed cause major problems. And 2. please don't drop PCI support!!