UM S5 Dimension Accuracy - Uneven X and Y Dimensions

[gchiu 0]

Printer: Ultimaker S5
Filament: Ultimaker Tough PLA
Slicer: Cura 4.12.1

Hi everyone, I've recently started experiencing issues with my parts dimensions. I printed a 20mm calibration cube using UM Tough PLA and my results are as follows: X (20.15mm) Y (20.5mm) Z (20.1mm). What could be causing such a large deviation in the Y axis? I printed the cube at 0.2 layer height and 10mm/s print speed, 215 degrees C.

I replaced the glass bed with a brand new one and manually leveled. I've re-tightened both the short belts. There was some play in my left Y-axle so I've loosened the pulley and re-tightened the axle. The result is still the same - I'm getting about +0.4mm on the Y axis of my prints. Any idea what I should be checking for next?
 

[gr5 2,266]

10mm/sec is crazy slow.  Try at least 20mm/sec although that won't help.  Also if you prefer dimensional accuracy over how nice the part looks then disable acceleration and jerk control in cura.  You will get an increase in ringing which causes "ugly" ripples near sharp angle changes such as text or edges but will improve dimensional accuracy for example when using a caliper to measure things.

 

Are you measuring the whole cube or only the face because corners tend to bulge out.  So I'm wondering if the Y issue is at the corner or across the face?  Please answer this question.  So many times people don't answer critical questions and then 5 days and 30 posts later I get the answer and it changes everything.

 

So I think the most likely issue is that your pulleys are wobbly - meaning the hole through one of the 11 pulleys (I think there are 11?) is not centered.  This means the belt speeds up and slows down with each rotation.  The problem could be with the pulley on the stepper motor or one of the 4 on the long belts.  I could be wrong but if you stick with this hypothesis you can test it.  If you print a cube that is exactly the same dimensions that is one pulley circumference then the problem should go away.  By the way I'm thinking possibly the problem could be with the X axis and if you didn't have this issue then both axes would be 20.5.  Or it could be on the Y axis.

 

Another possibility is play/backlash.  Push on the nozzle in both directions for the X axis and again in the Y axis.  Push with less force needed to actually move the steppers. Typically play will make the part smaller in the axis that has play.  If you don't know what this word means, read about it on wikipedia.  So you might have X axis play making that axis shorter.  And some other effect is making both axes larger in equal amounts.

[3DRocketry 1]

Interesting, so if acceleration control and jerk control are disabled - the the accuracy should improve? What does this sat about Cura? If those controls are disabled, then the printer itself makes the decisions how to behave with regards to this those two controls. 

Edited February 1, 2022 by 3DRocketry

[gchiu 0]

I was printing at 40mm/sec and slowed down to 10 to see if it was an issue with loose belts. But got the same results.

 

I measured across the face only and avoided the corner bulges. I also measured the upper layer only to avoid the bottom elephants foot.

 

Do you think I should loosen and realign all the axles to be safe? I suspect the problem is on the Y-axis because I noticed some layer shifts in some previous prints (Y direction only). Which is why I tightened the belts and noticed some slight play in the left Y axle.

 

Actually I just pushed on the printhead like you suggested. There is no play in the X axis. However, on the Y axis there is a fair bit of movement. The print head tilts up and down when I push it slightly. It looks to me like there is too much room between the center Y axle (where the print head is mounted) and the print head thru hole.

[gr5 2,266]

Ultimaker printers have very high acceleration and jerk because the print head is light weight compared to most printers.  This reduces overextrusion on corners.  Corners can bump out if you have low acceleration because as the head slows down at the corners, the pressure in the nozzle doesn't reduce fast enough so you get bumped out corners.  If you measured the bulged corners and that was the issue than speed up accel and jerk by unchecking and the printer will use the defaults which are much faster.

 

The downside is now you get lots of ringing as I said.

 

Anyway you weren't measuring the corners.  So I'm still sticking with my pulley theory.  That is a very common problem.  If the pulley is drilled in the center within 0.3mm then that could easily cause errors of 0.3mm.  0.3mm is pretty damn tiny error for a pulley that is being drilled out.

[gchiu 0]

I just re-tightened the print head down and noticed better results. The new cube measures as follows: X (20.15) Y (20.23) Z (20.1). I'm gonna re-check the pulleys to see if I can get a closer result for both X and Y.

 

For those wondering about what I did...there are four 2mm internal hex bolts located above the print head. I just tightened those four bolts down as tight as I reasonably can (without stripping it) to reduce print head play on the Y axle.

[gr5 2,266]

That's great.  I don't think you can do better than that.  Now are you using the "engineering" profile to print with?  This profile gives you some pretty good improvements (at the cost of beauty as you will get more ringing but it won't be measureable with a caliper so -- better dimensional accuracy).

 

If so then I would stop messing with the printer and correct remaining errors in CAD.  The errors you see may scale or they may not.  Probably both - some of that 20.23 error is a scaling error and will be 40.46 (exactly double) if you print a 40mm cube.  Some of that error will not scale so it won't be exactly 40.46.  It could be that most of the error doesn't scale.  Unless you do at least 2 different sized prints you won't know if the error scales or not.

 

So if you really need accuracies better than 0.2mm then you really need to fix things in CAD.  This is how it is done with injection molding.  The engineer hands the injection molded company a "perfect" cad model and then they are experts at adjusting the model to work with injection molding.  They even have to change crazy things like the angles of surfaces.  90 degree angles in the original cad need to be changed to a different angle so that when the part pops out of the mold the edge is 90 degrees again.

 

Unfortunately this is true with 3d printing as well.  If you want that kind of accuracy you need to fix it in cad. No one likes to hear this but it's the best solution.  Trying to adjust the scale by changing "steps/mm" on the printer or trying to scale the part in Cura will result in more problems then it is worth.

 

For example all vertical holes in all your parts need to be enlarged by about 0.4mm.  This will be your next frustration with dimensional accuracy as it's a huge discrepency between CAD and what you get out of the printer. This is true with all FFF printers (printers that melt plastic and squirt out of a nozzle).  It has to do with how 3d printing works and the properties of thermoplastics.

[gchiu 0]

@gr5 thanks for your help