Continuous Layer shift - Mid-band resonance? Nema 23 + long belts



  • Hi

    A curious layer shift is driving me crazy. It makes the parts neatly skewed like the tower of Pisa. No steps or randomness in the error, every “vertical” wall is angled the same way through all the print. The error changes it’s nature when I change the micro stepping and / or “jerk” and acceleration settings. The error is not mechanical (the typical way at least, no grease, loose parts or so on) in another ways but maybe some belt resonance has a role in it. I now have a theory of motor mid-band resonances causing this since I use nema 23s and have a quite heavy gantry and long belts to add weirdness into the picture. Maybe I am far from the real reason.

    I printed 9 test parts today and got the problem tuned almost out. The thing is, if I want to tune it completely out I would have to tune the acceleration and jerk into 0 as well, ha ha ha. Or so it looks at the moment.

    The couple of first test prints had this shift so that it was in the directions Y+ and X-, so the new layers always jump the same amount to the Y+ and X- directions compared to the previous ones. The funky part in this problem is that when I doubled the steps per unit on both axis (M92 X74.4 Y206, was half of those)
    , the X shift changed on the other side (X+) and the Y error got better (measured by eye it looks like the angle of error dropped in half) but didn’t disappear. Pardon my french, but wtf?

    Weirdly enough dropping the print speed into 70% (from 45-60mm/s for the walls, infills 70) did not have any effect on the error.

    Now The layer shift on X is very small to the eye, but of course too much since it has to be 0, and on Y the walls still lean maybe 4 degrees to the “north”.

    This is not a new printer, I had it running very nice for a while with previous Duet. This problem did not occur although hundreds of others had to be fixed. (I am not blaming the controller)

    Anybody having anything similar? Or having a good guess what it might be? I start to be out of ideas…



  • I have not had this issue with any surface mounted drivers for Duet/RAMPS/etc but I did observe this issue when trying to use some independent Gecko drivers. I never resolved it but I suspect that it was a signal timing issue. I doubt that is the cause of your problem. Sorry I can't be of more help.



  • I am using separate drivers, DM542A is the model. I just read of somebody having similar layer shift issue (although only with one file to print) but from there found the M569 T-parameter and just set the pulses into 2,5ms.

    Is this the correct way: M569 T2? Or do I need to name the drivers?



  • Ok M569 T2.5 was not the solution, neither T3.

    With these same drivers I have been able to print worryfree before. Is the signal from Duet Ethernet (through the breakout board) somehow different compared to Duet 0.8.5?


  • administrators

    You need a separate M569 command for each driver that needs extended timing. You probably have one M569 command per driver already, to set the motor directions. Just add the T parameter to the M569 commands whose P parameter refers to an external driver, which will be P5 through P9.



  • All right! Now the test piece came out straight!

    So great to get fast solutions from here 🙂 If only I would'we cried here earlier…

    Thanks!


Locked
 

Looks like your connection to Duet3D was lost, please wait while we try to reconnect.