Missing Steps - Creeping Values

My old iteration of the Monster mill used DC motors, 500-line encoders, and Geckodrives. I could run it all day and tell it to go to HOME... and it would be there +/- 0.0005 inch (0.0127mm). Because I blew out Geckodrives more than once... I realized that the Geckodrives could not handle the current... so I completely retrofit the machine with 1841 in/oz (13n) steppers.

The steppers are open-loop... and i am having some kind of issue with the Y-Axis that I can't figure out.
This is what I used: www.amazon.com/gp/product/B011E09GRW

So I was running a job today where I made a holding fixture... very accurate... had registration holes at X+230mm, X-230mm, Y+50mm, and y-50mm. This way I could align the 2mm endmill tot he holes and know that the unit was set up perfectly. (Within 0.1mm at least.)

So I ran the job... and I noticed the Y-axis was off 1.3mm. I re-calibrated... checked against all the holes... ran the job... again it was off about 1.3mm... it was creeping with every Y-Axis move.
(The X-axis and Z-axis appeared to be steady... only the Y-axis seemed to bee showing this error... need to verify.)

At this point, I went into STEPCONFIG and scrolled to the Y-Axis... then connected a dial indicator and told it to go back and forth 15mm... 10 cycles... when I stopped it, 0.0035 off! (Only 10 cycles!) I dropped the acceleration from 100mm to 5mm... ran it again... still going off noticably. Dropped the max speed from 45mm to 5mm... really slow... still off. I don't think the Z or the X are doing that... I'll have to verify... what is going on?

Do I need to drop to a smaller microstep? Bad controller?

This is so disappointing... almost $1000 in upgrades... and it won't hold tolerance.
Meanwhile... the little Frankenlabs mill I built from scratch for $800 holds perfectly.
RE Frankenlabs: www.youtube.com/playlist?list=PLa_L8q5ve...zP_yOKBXPB8zKn-iWCsX


Suggestions?
Jerry

I have struggled with someting similar for almost a year. It was on one XYZ mill with Mesa 6i25 driving three G213V controlleres. I was losing steps and drifting slowly on all three axes. Seems to be related to direction changes.
If i run a series of +1mm -1mm moves I get the same absolute error as if I run +3mm -3mm moves, not three times as large. The motor loses on an average 20 microsteps in one direction and 16 microsteps in the other direction.
The Z axis creeps upwards so it is not likely a mechanical issue.

After much testing by chance I changed the G213V setup from 10 microsteps to 5 microsteps and the errors disappeared!

It is repeatable, machine runs 6000 mm/min flawlessly at 5 microsteps with a 7 kg chunk of iron clamped to the table.
Changing back to 10 microsteps it misses steps as above even at 100 mm/min without any load on the table.

FWIW, I have no explanation to offer.

Wirewrap:

I can believe that. I didn't notice it until I did a pocket that required about 100 back and forth passes... at the end I was almost 3mm off. Then I ran a test running the machine +100mm and back 100mm... after 20 cycles, it was off a few thousandths. Then I ran the test in Stepconfig at the default 15mm +/- and got the same thing. Ever reversing move... fail.

This morning I'm going to verify the X and Z axis... and if they are drifting... try another microstepping setting. Not happy about that as lower settings can cause missed steps because of high speed... and lower settings will slow down the machine. I miss the absolute positioning of the servo with quadrature encoder... I really wish there was something like the Geckodrive... but capable of running 4-6 amps.

Too late now... money spent... gotta get this to work.

Well... I tried 1000 steps, 1600 steps... same thing only slower... and not quite as much creep.

I noticed that the creep was ALWAYS in one direction.

Mechanical???

The motor is connected to a shaft... two flats, two set screws... tight.
There is a 24 to 60 tooth pulley arrangement... also two flats, two setscrews... tight.
The upper pulley connects to the main shaft bay a ... what's this? Compression fitting??

So... it was tightened this much Aggg... but it needed this much Arrrrgggg... so now I have it tightened to an exact setting of Arrrrrggggggg.

Ran 50 passes +/- 10mm... showed 0.005" off (0.127mm). I didn't change anything...
told it to run 150 more iterations... 8 minutes...

Difference... unmeasurable.

Spot on... perfect.

I went to the X-Axis... and also adjusted it to a setting of Arrrrrggggggg... just to be sure.

Now I know why some machines are drilled and pinned.

Thankfully... the drivers are dead on... yeay!

Jerry