Machine off at the end of the programm. Repeatable!

Hey, I have a weird issue. Not sure where it comes from or what the cause is, but I don’t know what to look for anymore.My machine is off position at the end of the program. For troubleshooting: I put a work piece in my machine and went to X10 Y20 and made manually a mark on the workpiece.Then I run a program and at the end I go back to X10 and Y20, but the machine is off in X and Y for a few millimeter. This is happening again and again and it is always the same distance off. First I thought it had something to do with speed and acceleration. I tried lowering it. Same error. I let the machine then run in a higher Z position without any load. In the air so to say. Same error!Then I tried another NC program, same thing here. At the end it’s repeatable a little bit off.I lowered the acceleration (75mm/s2) and feed speed to 1/10 (600mm/min) of its original value. Same error!I had deadband at 0.01 and lowered it to 0.007. Same error.I tried changing the Ferror from 2.0 to 1.0. Same error.I tried different programs and it feels like the offset is larger if the travel within the program is larger.I checked the mechanics and I am sure it’s not a mechanical error.I don’t know what else to test.

The machine is an open loop stepper system with DM322T Driver and Nema 23 motors.

Have you checked for backlash with a Dial Indicator, as opposed to a dial test indicator which can be difficult to setup to negate any cosine error. T me this sounds like a backlash/leadscrew related issue.
Many times an similar issue has come up similar to this, even tho the user had checked the mechanical side, it turns out upon closer inspection it is indeed a mechanical issue. Coil type couplers aren't worth much as they can wind up. Don't rely on the stepper motor to provide any fixed bearing support (as many 3d printers and cheap chinese mills do), use angular contact bearings, the non driven end of a leadscrew should be able to "float" laterally to allow for thermal expansion.
Even ballscrews have backlash, unless you are paying alot and I mean alot for ballsrews expect some backlash.
Missing steps could also be a cause. But if I were a betting man I would put a wager on a mechanical issue. The thing regarding mechanical issues is they can add up, a little here and little there and another bit there can end up as a nightmare.

Here's a document far more eloquent than I could describe.
rick.sparber.org/mt.pdf

I am using LinuxCNC Rio. The machine came from a parallel port, but it's converted. I don't think it has anything to do with Rio, I think it's some sort of configuration error in the ini file. It is all much to repeatable for a mechanical issue.

unknown wrote:

Have you checked for backlash with a Dial Indicator, as opposed to a dial test indicator which can be difficult to setup to negate any cosine error. T me this sounds like a backlash/leadscrew related issue.
Many times an similar issue has come up similar to this, even tho the user had checked the mechanical side, it turns out upon closer inspection it is indeed a mechanical issue. Coil type couplers aren't worth much as they can wind up. Don't rely on the stepper motor to provide any fixed bearing support (as many 3d printers and cheap chinese mills do), use angular contact bearings, the non driven end of a leadscrew should be able to "float" laterally to allow for thermal expansion.
Even ballscrews have backlash, unless you are paying alot and I mean alot for ballsrews expect some backlash.
Missing steps could also be a cause. But if I were a betting man I would put a wager on a mechanical issue. The thing regarding mechanical issues is they can add up, a little here and little there and another bit there can end up as a nightmare.

Here's a document far more eloquent than I could describe.
rick.sparber.org/mt.pd

I am almost certain that this is not a mechanical issue. The fixed and floating bearing design for the ballscrews is what my machine has. I also don't use coil couplers. Not sure what the right english word is, but blieve mine are called spider couplers. But you are right, I first thought it was a mechanical error or broken cable, but when I started narrowing down the error I am more and more convinced that it's something in the machine set up. Ferror, Fmin, Deadband,... something in there... I think. 
 

Some of the cheaper fixed bearing blocks are sold with non angular contact bearings, this is something to check.
Until you perform a backlash test, and yes this can manifest as repeatable, you can not rule out a mechanical issue.

Consistent position errors after a long program could indicate a step timing
or polarity issue. Too low direction setup or hold times could cause a lost step
at reversals. Wrong step polarity can also cause this issue as stepping on the
trailing edge of the step pulse (rather then the expected leading edge) will
decrease the expected hold time.

It might be worth either inverting the step polarity or increasing the setup/hold times,
at least to see if it changes anything.