I have tried some other tests over the last fw days.  adjusting the Motor Tuning step pulse numbers from 0us to 15 us has no effect.  changing the dir between 0 and 5 us has no effect.  I've tried several grounding and shielding schemes but no change in behavior is detected.  Hmmm.

I just ran some more tests. Changing line 5 from G1x+.1 to G1x-.1 does not change the direction of the offset.  it still moves +.
+.005 in 50 repetitions.
+.010 in 100 reps.

That is consistent.

I set the line 5 to read G1X+.00005 and ran the program with the loop set for 100 reps.  No movement was detected at all.  None!.

This is not inconsistent at all. Such a small change may simply leave the Step signal in the opposite state, such that flipping the Dir line does not cause the change in the Step line. In effect, this is working around the bug (if I am right).

I set the line 5 for X+.00010 and 100 reps yields +.010 offset.
Set for x-.00010 yields the same +.010 offset.  Hmmm.

Consistent. But try 150 and 200 cycles as well for really convincing proof.

Now for the wierd one.
I set the line 5 to read G1X+.00015 and ran the 100 reps.  the offset looked like almost .0110"!!!
So I ran it again and again - a total of 10 times.  Offset is .110 total.

Ok, two parts. The first part is that going from 100 cycles to 1,000 cycles gives 10 time the error. This is entirely consistent with the idea of a bug somewhere, either in Mach3 or in the Mach3/Gecko interaction.
What significance there is to the fact that you are micro-stepping the Geckos is something I am not entirely sure about. It may be worth while asking Gecko about this as well - with full data.

When it runs the program, the movement sounds sort of stuttery, unlike when it is actually moving some distance. 

Dunno. Obviously you are way down in the acceleration phase, and that could be tricky. I would check this with both ArtSoft and Gecko.

Bottom line: while cleaning up the earthing etc is good, I do NOT think the real problem lies with your hardware. It is in the ArtSoft and/or Gecko domains. You may be able to produce evidence of the problem, but I do not think you can resolve it yourself.

Cheers

Hi, Roger.  I just uninstalled the Mach software and am cleaning out the computer to make sure it is ALL gone.  Then I'm going to download a new copy and install it.  Without the license file, at first.  If that doesn't fix it, I have one other card in my hat.  I have a couple of Larken Automation Viper 200 servo drives and an Ametek servo motor that I'll try to set up to see if it repeats any differently.  I know it could open another whole can of worms, but It might be worth a try.  Thanks for the help so far.  Don

Thanks for the help so far.  Don

Hey, if the effect is real, I want to know - and to see it fixed! Very selfish motives ... :-)

Cheers

So do I. 
I just completed purging the old Mach3 install.  I installed the Current Lockdown version.  Downloaded onto this computer, burned a CD, took CD to shop, installed it into milling machine.  No license installed yet.  I've configured the axes and run the x-axis test program.  Now the darn thing offsets to the minus direction!!!  Yup, -.005" in 50 repetitions.  I gotta go to dinner.  I want to try the same tests as earlier today ie. short travel, 100 reps, etc to see if the behavior is consistent when I fire it up later tonight.  I feel like I'm in the "Twilight Zone"...  Don 

The strange-ness gets stranger...  I set line 5 of the test program to X-.00015 with the loop set to 100 repetitions, ran the program, and it walked over to -.011...
I set the loop to 500 and ran it again and it walks right over to -.055.
Then I set line 5 to X+.0004 and the x axis buzzed over -.102 total after running the program 10 times.  I set the Loop to 500 reps and ran that twice - offset totaled -.100.
Then I set line 5 to X+.0005 and Loop 100 reps, and the x axis buzzes and moves over +.001 after 10 runs of the program.  After running the program 20 times the offset was +.002.  Yeah.  Positive direction this time.
I set the Line 5 to X+.0006 and in 10 runs of the program, it buzzes with a rapid ticking or pulsing of the sound of the drive system - about 10 pulses in the duration of the program, and offsets about .001 per 100 reps...
I set line 5 to X+.0010.  Very loud vibratory buzz with about 10 cyclical pulses to the sound, but after running this program 10+ times, there is still NO discernable Offset!!!
I set line 5 to X+.002 .  Very loud with 5 initial loud pulses then smoother for the rest of the "run".  No discernable offset despite multiple runs of the program.  Just nasty buzzing.
I changed the loop to 500 reps.
X+.001 buzzes but does not offset.  sound has 9 pulsed and then smooth buzzing...  I ran this multiple times with no offsetting.
X+.00015 machine walks right over to -.055". Aargh!
X+.010 table jiggles like a paint shaker and offsets - get this! - -.050".  
     I am truly bewildered... Don

X+.001 buzzes but does not offset.  sound has 9 pulsed and then smooth buzzing...  I ran this multiple times with no offsetting.
X+.00015 machine walks right over to -.055". Aargh!
X+.010 table jiggles like a paint shaker and offsets - get this! - -.050".

Some of the results are consistent with the idea of a bug at the change of direction. However, this buzzing puzzles the hell out of me. Jiggling like a paint shaker ... ? That's really weird!
Thing is, until we can explain everything, we do not know what is going on.

Hum - I don't suppose you can find a digital storage CRO and LOOK at the signals by any chance?

Can I suggest ... putting a pause between each reversal. Yes, that will make the program much slower, but it should let the first motion stop before the second (reverse) motion starts.
Alternately, do you have Mach in Constant Velocity mode? If so, switch to Exact Stop.  (Ho hum - or vice versa...)

Running experiments is all very well, but they can be bewildering if you don't have a theory to test. (Sorry about the preaching: my background is a PhD in Physics.) We seem to have some consistency with the number of cycles and the offset, and some consistency with the idea of very small increments producing variable results, so what's the buzzing? Is it that the acceleration parameters are just not handling the mass of the table when you combine very small steps with frequent reversals? Or is there a slight overshoot making the Geckos get very shirty about the reversal? Hence the pause idea.

Cheers

Hi Don,

You've been getting very strange results... The exact stop test that Roger suggests is interesting. Following the nature of the problem (its consistency and inconsistency) it looks to me that it is a combination of several things which causes it. Like both the particular mechanics and Gecko's/Mach3...

Dan

Hi, Roger.  Shaking back and forth like the program calls for, basically.  But with the added offsetting.
A storage scope and a way to count the steps would be way cool, but I don't have access to one.  Oh, well.
I have been using constant velocity mode.  I'll check that out tomorrow, along with the pause idea.
I did, however, get the Larken Viper 200 servo drive and Ametek motor with a 500 line (2000 count) encoder hooked up.  It only accepts +5V common signals, so I had to use the PMDX breakout board.  I put an index mark on the sproket that is on this motor and an indicator so I can see if it offsets.  Yup.  It seems to be doing the same things the stepper system does.  The PMDX BOB has an activity LED that flashes when pulses are flowing.  It seems to flicker in time with the stuttering/pulsing of the tests I ran.  I haven't gotten to run all the tests but at .00015 it walked over to its new offset just like the stepper motor system did. 
I'll report back soon.  Curiosity...  OCX driver test - what should I be looking for when running it.  It had very small vertical lines in the graphical display when I ran it when I put the computer into service... 
Don

Dan, Good to hear from you.  It might be a combination of things, but I'm starting to wonder about software/pc issues...  Changing out the entire Drive system with the exception of the power supply seems not to cure anything.
Don

I did, however, get the Larken Viper 200 servo drive and Ametek motor with a 500 line (2000 count) encoder hooked up.  It only accepts +5V common signals, so I had to use the PMDX breakout board.  I put an index mark on the sproket that is on this motor and an indicator so I can see if it offsets.  Yup.  It seems to be doing the same things the stepper system does.

Getting harder to say the problem is random noise, isn't it?
Ahhhh... was the motor hooked up to the hardware, or just sitting bare? This affects the acceleration tuning etc.

Curiosity...  OCX driver test - what should I be looking for when running it.  It had very small vertical lines in the graphical display when I ran it when I put the computer into service... 

Mine does too.

Cheers