Y-Axis Shift - Very Puzzling

Hey everybody, long time reader first-time poster and I love how insightful everybody is. I have a very custom Prusa i3 built on a P3 Steel frame, using Ramps 1.4 controller on a Mega 2560. I've been having layer shifts for a long time now specifically on my y-axis. I've been searching for months and have been unsuccessful in coming up with a solution. I originally used Marlin with Slic3r and the shift would only show up on larger prints, then I started using Repetier because I found it to be much more user-friendly and convenient with the inbuilt Slic3r slicer and the problem got much more severe. I've tried everything I can think of to fix the problem some of which are listed below.

Adjusted stepper alignment to ensure it was perpendicular to the belt
Adjusted A4988 current up and down
Adjusted acceleration and jerk to extremely low levels (250mm/s^2 and 5mm/s respectively)
Changed stepper driver
Increased cooling to stepper driver
Changed steppers to higher torque models

If anybody could help you would be a lifesaver! Thank you!

Have you tried moving the axis by hand? Does it bind up or does it move freely? Otherwise you seem to have covered most of the other bases with checking pulley alignment, tuning accel/jerk and the likes.

Also, what sort of speeds are you printing at?

EDIT: If you switch back to marlin and slic3r (and keep your jerk/accel settings), does the problem get any better again? If it's a mechanical issue like the bearings binding in the axis, it may have just worsened over time (as they tend to do) and happened to coincide with your software shift. Check your rods for excessive wear, like grooves dug in by the ball bearings and check the bearings for free movement and see if they make excessive noise.

Edited 1 time(s). Last edit at 04/12/2018 11:21PM by Trakyan.

There is one point missing on your checklist:
calibrating e-steps to ensure, overextrusion is not the cause for the nozzle to dig into the part.

Probably already covered during the switching of steppers, but for completeness... Have you checked the tension on your belts? And that the pulley on the stepper is locked down properly with the grub screw?

Also, is the stepper itself overheating?

Do you use auto bed-levelling? If you do, then play in the Z axis (e.g. a loose Z-coupler) can cause discrepancies in the X & or Y axes.

Thanks for all the feedback everybody! So to answer questions in order (mostly) I have checked the bed and it does seem to roll freely with every little friction. I keep it lubricated pretty well but it does have those cheap linear bearings. Not really sure how to check belt tension but it seems like the tension is good to me. You don't have to work to pull it in any direction but it snaps back to taught when you let go, also has a spring loaded belt tensioner to keep things consistent. Also not 100% sure how to check for overextusion, but the parts it does make seem to have pretty standard extrusion lines (not like it was really being packed on there) but I'll tone down the extrusion multiplier by a bit to see if it helps. I do not use auto be leveling but I will check and make sure that the pulley gear on the stepper is nice and tight on there.

A spring-loaded belt tensioner is IMHO a bad thing. It essentially allows play in the mechanism. If you're accelerating the bed, the spring will compress/expand so that the motion of the bed won't correspond exactly to the motion of the stepper. A screw-adjusted tensioner is much better.

The effect of any play in the mechanism on your output will also depend on what motion your slicer generates. e.g. If you print a circle which is (say) 50mm in diameter and two perimeters, your slicer may choose to print both perimeters in the same direction (say clockwise), in which case the same errors would apply to both traces; the perimeters would be adjacent, but not exactly circular. Or it could print one perimeter clockwise and the other counterclockwise, in which case opposite errors would apply, and you would see gaps between the perimeters. Similarly, squares become rectangular, if aligned with the axes. If you print a cylinder or cone or cube, then it may print some layers in one direction, and other layers in the opposite direction, in which case you would see the errors as layer shifts. If you have more complex models, then the effect becomes more difficult to predict. It may be worth experimenting with printing something in vase mode, in which case errors would be consistent from one layer to the next.

A bit off topic I know, but I've always been curious. People bash spring tensioners in belt systems, but how come everyone seems to be quite happily using spring loaded anti backlash nuts? Their effect is exactly the same, when you change direction and the other half of the nut is doing the driving, the spring can compress and you get the same inaccuracy. That and when you use them on a CNC mill (as many people do), the cutting forces themselves can compress the spring.

Wow! I never even thought to check the tensioner! Thank you! I'm gonna take a look at that ASAP!

Quote
Trakyan
A bit off topic I know, but I've always been curious. People bash spring tensioners in belt systems, but how come everyone seems to be quite happily using spring loaded anti backlash nuts? Their effect is exactly the same, when you change direction and the other half of the nut is doing the driving, the spring can compress and you get the same inaccuracy. That and when you use them on a CNC mill (as many people do), the cutting forces themselves can compress the spring.

I guess, when the spring tension is higher than the acceleration forces, it doesn't matter much. Same for the belt tensioner springs.

Quote
DTolbert
Hey everybody, long time reader first-time poster and I love how insightful everybody is. I have a very custom Prusa i3 built on a P3 Steel frame, using Ramps 1.4 controller on a Mega 2560. I've been having layer shifts for a long time now specifically on my y-axis. I've been searching for months and have been unsuccessful in coming up with a solution. I originally used Marlin with Slic3r and the shift would only show up on larger prints, then I started using Repetier because I found it to be much more user-friendly and convenient with the inbuilt Slic3r slicer and the problem got much more severe. I've tried everything I can think of to fix the problem some of which are listed below.

Adjusted stepper alignment to ensure it was perpendicular to the belt
Adjusted A4988 current up and down
Adjusted acceleration and jerk to extremely low levels (250mm/s^2 and 5mm/s respectively)
Changed stepper driver
Increased cooling to stepper driver
Changed steppers to higher torque models

If anybody could help you would be a lifesaver! Thank you!

I had this very same head scratcher with a custom built delta of mine. I made a very small desk top whose volume was about 70mm diameter and 90mm high. It continually had either X or Y offsets on certain layers. Watching it, I saw that the head would appear to speed up, or slow down in certain areas. After puzzling over this I did some experiments and discovered that if the radius from the nozzle to the bolts holding the arms on (delta rod offset I think Marlin calls it) is too high compared to the length of the arms, you have problems. I discovered that the "80%" rule appeared to hold here. The rod length to "nozzle-to-rod offset" radius ratio can be no higher than 80% or you run into trouble.

All that said, my problem was with a delta, not a Prusa variant. Is there some kind of similar offset that is comparable on a Prusa as I saw on my delta?

Another issue that I had with my "microdelta" is that the small base meant that there was little room under the bed for anything but the NEMA14 steppers. They would warm up, then REALLY warm up and I would have a motor skip, or just plain stop and rattle. Very similar to a driver overheating. So, check your steppers for overheating, that might be a problem.

DLC

Edited 1 time(s). Last edit at 04/27/2018 05:21PM by dlc60.

You came the problem the opposite of me. I was convinced it was a dragging bearing, in my case lowering the jerk and accell solved my problems.

But you can try this. unhook your drive components and try try moving the yaxis by hand, feeling for any movement that feels 'gritty'. This is your bearings doing a jitter.
My guess is that you see it on larger prints because the jitter/drag is at the end of the travel, not the center.

Also try this with small back and forth moves of about 5mm in various areas. As the balls roll around in the bearing you get spots where small reversal moves drag alot.

I would remove the bearings and drop them in a solo cup with some mineral spirits, swirl and jostle them about for a few minutes....you will be stunned how much flakes come out.
Repeat with fresh MS a few times.
Re oil and reinstall. Then recheck the travels.

I made a 'test bar', 5mm wide x 10mm high x 80mm long that I could test print in several directions to isolate the drag, and to prove that layer shift was fixed.

In future if someone else has this issue.
Also check the nuts of your bed leveling kit , sometimes they slightly rub on the frame , but is still able to go past.
I filed my frame down by about 2mm for the hex bolt to pass without touching.

Edited 1 time(s). Last edit at 06/04/2018 04:12AM by Techtinker.