Accumulating position error with servo system

[Kurtis Martineau <klmartineau-at-shaw.ca>]

Hello all,

Anyone have a "Not a problem" response to the situation below?

The problem:
It was discovered that the Y axis (axis 1 on a 3 axis system) drifted out of
position by as much a tenth of an inch.
When I checked the axis with a micrometer,  I found that when it was driven
out 12 inches and brought back 12 inches,
it always came up short, but by a vaying amount.  When the test was repeated
with varying velocities and varying
distaances, the error accumulated.

The system:
1. Is a large cnc drilling machine - X, Y, Z - 144, 48, 12 inches
2. Uses resolvers and resolver to encoder transmiter modules for each of the
three axis.
3. Resolution is 40960 pulses per linear inch for each axis.
4. Axis X and Z display acurracies of +/- 1.5/1000 inches.
5. PC is PIII 667MHz
6. Servo-To-Go model II
7. BDI2-18 with cvs updates from 04 Feb 03.
7. EMCMOT = newstgmod.o  (ini file attached)

Checks, repairs, and part replacements:
1. The coupling between the servo motor and the lead screw was re-keyed and
re-tightned up.
2. Some of the 400 ball bearings in the bearing block were measured and were
within limits.
3. The tachometer on the servo motor was cleaned and checked with  a 200MHz
Fluke scope.  Only  the slightest amount of noise was noticed, but this is
consistent with the X and Z axis tach feedback signals.
    NB:  At this point, I can push on the carrage and see that a measurent
of a few ten-thousandths of an inch on both a rigidly mounted micrometer and
the Tkemcex gui.
4. The resolvers from the Z and Y axis were swapped.
5. The resolver/encoder modules from the X and Y axis were swapped.
6. All resolvers cables have individually shielded conductors within factory
installed shielded cables.  All cables were visually inspected,meggered, and
capactive checked.
7. The STGII 8 axis board was replaced with a 4 axis board.

8. The last thing I did was swap the X and Y axis wiring to the STGII board.
ie: The X axis on the Gui now controlled the position of the real world Y
axis, and vice-versa.
    At this point the position error moved to the real world X axis and the
actual position control of the Y axis was +/- 1.5/1000 inches.

Discussion:
1. I recall that a few months back when Paul fixed the
homing-off-encoder-pulse for the STG boards, he mentioned not to use
newstgmod.o.  This is the only driver however that will allow me to home off
the index pulse out of the choices that I have in the /plat/realtime/lib
folder.  I never noticed any other fault, and I always assumed the one
mentioned here on the Y axis was going to be mechanical.
2. After step [8] above,  I feel convinced that I have eliminated the
possibilty of a mechanical fault.
3. Also, I noticed that when I home the three axis, the Y position display
is either -0.0002 or -0.0003, while the X and Z is always 0.0000.  I checked
for typos in the ini file, but I can't find any reason for what appears to
be number stuffing.  I have tried homing this axis at 0.24 IPM, but the X
and Z have no problem at 24 IPM. (my default homing velocitiy at 100% feed
overide is 0.4 IPS)

Kurtis

burgmaster.ini