Hello Carl, I hope you are progressing with this issue and please excuse this interruption.

Jimpinder, your explanation is very interesting. I'm not doubting you, don't get me wrong. I just have some questions for you.
Having the same CV settings, Carl at 10k per and yours at 60k per.
Carl runs 10 cycles and is out .060", You run 60 cycles and you would be out very near the same amount ?
So if my machine is 2k per, 10 cycles could potentially end up being off .300" ? (not that it is 2k, just for an example)

Now as a hypothetical....If a machine with 100k per was to run code similar to the example for a precision part, say +/- .0005", and was set-up to self load and unload parts, the first few parts would be OK. All the rest would be scrap and the table would eventually go to its limits. ? ? ?

Without true position feedback, actual movement at the slides, Mach would have no way of knowing....just like you say ?

I fought a similar situation here with a turned and threaded part in my lathe. The tolerance on the PD is +/- .0007".
This was with 3 tools set up and in Exact Stop mode btw.
The first few were fine, then they were getting smaller.
Re-home the X axis, then they were OK but gradually got undersize again.
So, as a test, I homed and Zeroed the axis, then "UNCHECKED" Auto Zero in homing.
Then ran several cycles, and when homed again, it reflected the drift precisely. 10 cycles = .0023" out. (DIA. mode)
This is a neat trick, it shows exactly what you lose or gain through the cycles. (semi precision switch, always within .0002")
What was really convenient for me was that my home switch is very near to making contact at the end of the program run.
So I put DoButton(22) =Ref X in M777 and put it just before M30 and re-home at every cycle.
Now all of the parts are identical.....well, within tolerance anyway. Otherwise, it would eventually go to the limit.
What's most strange to me is that the Z axis, after over 600 cycles without referencing was only out .0002", and that could have been the poor quality switch.

So basically, if you dont re-reference frequently when doing precision work, you might not get what you ask for from Mach ? ? ? ?
This is a MOST interesting topic.
Thanks Jim and good luck to you Carl,
RC
Hey anybody.....PLEASE correct me if I'm wrong. It wouldn't be the first time and I'm certain it wont be the last.

I have not tested Jims theory out on my mill yet but hopefully will have time today, if not over the weekend.
 However I know for certain that it doesnt happen on my lathe for the following reason. My servo drives are set up for a following error of 20 counts, now I have 1:1 on the motors/ballscrew, the ballscrew is 5mm pitch and the encoder is 2000line (8000 counts) so that means if my position is out more than 0.0125mm (0.000492) the drive will fault with an out of position error. I dont often run vast quantities of the same parts as its usually 1 to 10 parts but I have done some in the 30's and 40's and not had an out of position error yet.
 Hood

Great...we also need to duplicate the CV settings in Carls set-up...to run a valid comparison.
I don't recall seeing them posted...but may have overlooked them.
Anxious to seee what you find.
May get a chance to try it on my lathe too, just change it to the X-Z plane.
Awaiting your results Hood,
RC

With the forum down today I was unable to log in and see the code that had been used so didnt manage to do the test, hopefully tomorrow I will manage.
 Hood

Hi all,
Attached is a modified file of the gear used for some testing. It keeps the four machined quadrants but eliminates the depth repetitions. So you still have the sharp x y moves, lead in's, etc. I ran this on my mill about 50 times total with F at 10,20,30,40. Since I use a SmoothStepper there was no backlash implemented. My mill has a four axis digital readout so comparison to the MACK DRO is easy at different points and at the end. No progressive error was seen. 

I'll waite to see what Hood comes up with, and if you want, can run the program with a different PC not using the SS, and will provide for be a better comparison.
RICH

OK just done lots of tests with Carls two bits of code. If I take out the backlash by making sure the axis are moving in the same direction as they will be going on the last move then I get no error whatsoever. Obviously the extra line in the code means I have to bring the axis to the start position from a different direction to eliminate the backlasdh but again no noticeable errors.
 Measurement was with slip gauges which I know are more accurate than my mill could ever hope to be.
 Unless Carl is making sure that his zeroeing of axis are when moving the same direction that the final move is in then I think it is backlash he is seeing.


Hood

Gentlemen,

I just got back from touring the Rockies on my new motorcycle so am just getting caught up with the replies.  I can't thank everyone enough for thier replies and suggestions so I will be picking this up and working the recommendations.

I am in the process off building a breakout box for the parallel cable so I can hook a scope to it before the breakout board.  The test cables I made did not help nor did the movements of the lines in the box to try and eliminate coupling if that's even a factor?!

I am going to try some tests as Jimpinder recommends with Abosolute Stop mode.

I don't see how this can be backlash as the error is progressive.  Each cycle it gets further off.  I may not understand Hood's point but backlash should be a fixed error in each direction but never more than the total backlash amount I would have thought.

Back to the shop for some test.

Best Regards.

Carl

Yes thats true Carl, forgot it was increasing each time.
Hood

Hood,

Thanks for the clarification.  I'm new to all this so it's easy to overlook something obvious.

Well, I just finished a test using Exact Stop Motion Mode under General Logic Configuration.  I believe this is what Jimpinder recommended.  Anyway, I still have the progressive error but do like the fact that it no longer rounds the corners!

I just ordered a refurbished PC to make sure this isn't the problem.  I'll give it to my daughter if it doesn't work so no loose here to try this variable.  Sometime this week I will have the breakout box made for the parallel cable but will need to wait for my father to return from vacation to get the scope.

The comment about the 201s vs the 203Vs is cerrtainly an interesting one.  Has anyone other than Cobus encountered this problem?  If nothing else works, I may need to drop the coin and get the 203Vs.

This reminds me of a Wang VS65 I was supporting for the Navy which was acting up.  The vendor replaced every board and driver in the computer over a course of a month before realizing the power supply was a couple of volts off and causing all of the other devices to be erratic.  Hopefully I will figure this out before replacing everything.

Best Regards.

Carl

Hopefully I will figure this out before replacing everything.

I hope so too. Thought early on this would be a tuff nutt to crack.

Brett