Odd slicing bug?

[jerolee 0]

I experience some odd slicing happening. Cura invents how to alter a sharp corner. it also changing with rotating the object. What makes it even more of a odd behaviour is when i multiply the item, it aint consequently slicing them all the same. The slicing also changes when u move the objects a bit like  rearranging them neatly, it just changes the slicing.
Works as intended?  😛   or is it a bug?
I strive to get all items dimensions exactly the same and this does make me scratching my head, wtf is goin on here?

Anyone?  i havent found a solution.
I add a piccie to make it more clear.

I think im goin back to an older version. im using 4.10,   4.11 does the same i didnt have this behaviour before, i reckon i would have noticed this 🙂
Thanks all.

 

Edited October 13, 2021 by jerolee

[fvrmr 73]

Hi @jerolee could you share your project file? (file - save project)

It makes it easier for me to reproduce your issue. 

[jerolee 0]

Hi, @fvrmr  file in de bijlage. bedankt   lol i accidentily wrote in Dutch 😛  i meant to say that the file is attached!

Cura.3mf

Edited November 20, 2021 by jerolee

[GregValiant 980]

Sorry about my answer in English.

When a part slices differently dependent on where or how it's located on the build plate then my experience is that the problem is always in the part.  I used both MS 3D Builder and https://formware.co/OnlineStlRepair for repairs and both repaired parts sliced OK.  Attached is the one from the OnlineStlRepair site.

Here is the report from the repair site:

-> Analysed your file:
--> 0 Naked edges (?)
--> 0 Planar holes (?)
--> 0 Non-planar holes (?)
--> 0 Non-manifold edges (?)
--> 0 Inverted faces (?)
--> 0 Degenerate faces (?)
--> 0 Duplicate faces (?)
--> 0 Disjoint shells (?)
-> Repairing: 100.00%
----- Repair completed in 403ms ------
-> Vertex count changed from 1447 to 1512 (+65)
-> Triangle count changed from 2890 to 3020 (+130)

 

CuraBadPart_fixed.stl

[jerolee 0]

No worries about your English Greg, it was my answer in half English/half Dutch 🙂.  i dont agree with you mate. when i slice the design in Prusa copied it there 3 times, all the 3 items get sliced exactly the same, where ever i put them on the bed or rotate. Ill show you in piccies.
Your repaired file shows an artifact which wasnt there in original, thats odd, it is also cutting corners!

I am sure that my design has no defects, its done in fusion360, a piece sliced from a circle etc etc. the corners are exact same 🙂
I love cura slicer, thats why i would like a developer look into it.
Cya!

Edited November 20, 2021 by jerolee

[GregValiant 980]

I'm still using my old Mechanical Desktop (father of Inventor) and it also does a fine job creating STL's.

Looking closely I see that Cura has decided that the path that creates the apex of the point should be a short line segment.  It builds on that segment and that's what we see regarding the inner walls.  This makes the "Wall Thickness" nearly correct off the point.

PrusaSlicer is extending the walls rather than truncating it's movement.  If you were to decrease the angle, you would see the PrusaSlicer paths extend further and further from the point.  So the Wall Thickness off the point would increase.  I think Cura is being more accurate.

 

If we view the Prusaslicer paths we see this:

 

The dimensions indicate "Wall Thickness".  I think it is clear that the black dimensions indicate a true wall thickness and that the yellow dimension is a trig function of the angle.

 

Looking at Cura paths they appear to maintain the Wall Thickness about the point.

 

Doing a line study on a 15° angle I get this.

 

The white lines to the right are the part and the white lines extending left indicate walls 1.2mm thick.  The Cura path (red lines) leaves a wall off the point of 1.77mm where the (suspected) Prusa paths (cyan lines) leave a wall 9.19 thick off the point.

 

Thinking about that - I decided to go a bit further.  Here is a part with the angles lowered to 15° and sliced with Prusaslicer.  It looks much like what your larger angle sliced like in Cura.  My guess is that the angle has gone below some threshold in PrusaSlicer maybe to maintain the accuracy of the wall thickness off the point.

 

Since I'm just a user and unfamiliar with what actually goes on inside Cura I can't tell if my wonderful logic is actually true.  I don't generally speculate, but in this case I thought I'd throw my thoughts out there.

 

@ghostkeeper and @ahoeben are two of the real "nuts and bolts" guys of Cura.  Maybe they would have a take on this?

 

[tinkergnome 924]

8 hours ago, jerolee said:

I am sure that my design has no defects, its done in fusion360, a piece sliced from a circle etc etc. the corners are exact same 🙂

 

The Prusa slicer does a lot of automatic repairs, that hides the errors.

 

This is what Meshmixer thinks about your file... (every "pin" marks an identified mesh error):

 

[jerolee 0]

Hi guys,  I gotta blame fusion360 for this then 🙂   the "cut" shape is done by combine and cut in fusion, not drawn and extracted or something.
Its not that biggie, i was just wondering  wth is happening here, dragging  rotating the pieces and every time it decided to handle that sharp corner differently  🙂
I prefer Cura anyways, im used to using it so yeah...
Thanks all.

[ghostkeeper 105]

This discussion is about a geometric term called the "miter limit". If you perform an offset or inset of a shape, very sharp corners may move a long distance, so the sharp inner corner of your layer would move a long way inside. The miter limit adds a cap on that. If the corner is further away than a certain number of insets, it'll get capped off and turns into two shallower corners.

I think the original model was just on the edge of this miter limit, and got capped off on one side due to rounding errors.

 

MDN has a nice explanation of it here (since it's also a CSS property): https://developer.mozilla.org/en-US/docs/Web/API/Canvas_API/Tutorial/Applying_styles_and_colors#a_demo_of_the_miterlimit_property

 

Cura uses a miter limit of 1.2 for most things (including wall insets), meaning that corners get cut if the inset vertex would go beyond 1.2 times the offset distance: https://github.com/Ultimaker/CuraEngine/blob/6fc5f512f2d0dfe602a3d1b71d61853578cb8b42/src/utils/polygon.h#L128

 

A miter limit of 1.0 would mean that corners are exactly as wide as the normal wall. This is not always desirable though, since corners are also weak points of the print. We went for 1.2 as a balance.

 

Arachne produces round joints instead of miter joints, by the way. The miter limit is removed from the walls there.