EMC problem/question

Crew:

I am having a problem with EMC and don't know where to go from here. I am
hoping someone more familiar with EMC can direct me.

I've noticed my mill doesn't come back to zero after a number of moves. If
I repeat

g01 x0.1 f10
g01 x0.0

twenty times in a program the zero is off by 0.010". If I change the
distance moved from 0.1" to 0.2" I get the same error, .010. I can do the
same with moves in the y direction. Double the moves (40) gives double the
error 0.020". I changed the feed rate to F5 and the errors are the
same. I changed the default and maximum accelerations in the .ini file
from 10 to 5 with no effect. Finally I wrote a program with

g02 x0.2 y0.0 i0.1 j0.0 f5
g02 x0.0 y0.0 i-0.1 j0.0

a circle of 0.2" diameter, repeated ten times. This gave a final offset
error of .005" in both x and y. Finally I did this last again using
g03. The errors were the same and in the same direction (+ reading on DRO
in all cases). I did a small move from and to zero to eliminate any
backlash as:

g01 x.05 y.05 f5
g01 x0.0 y0.0

I have a hard time believing this is mechanical since it happens to the
same degree in the x and y directions. Likewise the stepper motors and
drivers are separate for the x and y, leading me to believe the problem is
not there. That leaves the software EMC.

I am running Red Hat linux (2.2.14), rtlinux 2.2 and the Aug 11, 2000 EMC
on a 450 MHz P3 with 64M. The drivers are 5A Camtronics running half step
to steppers of about 750 oz-in. These connect to ballscrews through 2:1
timing belts on a 9x42 knee mill.

Any ideas or suggested tests?

Thanks.

Hugh Currin
Klamath Falls, OR