Cura generates geometry not present in the original file

[baltic 1]

Cura generates some weird paths at layer 424 (0.2 layer size) Not present in the original model

горшок балкон пол низ.STL

[GregValiant 1,144]

I couldn't duplicate that so I wonder if that is just an artifact in the preview display?

I checked the model for errors using the Cura mesh analyzer and MS 3D Builder and no errors were found.  I uploaded it to Service.Netfabb.Com for repair anyway because sometimes it finds things that other programs miss..  Your file is 39kb and the repaired file I got back is 98kb.  I have no idea why that is but I'll pass along the altered model.  Maybe somebody else here has an idea, but I'm leaning toward a glitch in the preview.  If you save the gcode file and then open it in Cura is that strange pattern still there?

1853057440__fixed.STL

[baltic 1]

Nah, it's not a preview. I have checked the gcode.

The fixed model doesn't have it

Here are the settings i am using

petg.curaprofile

Edited April 5, 2021 by baltic

[GregValiant 1,144]

"The fixed model doesn't have it" 

That makes it sound like it was indeed something in the model.  If that's true (and it looks to be so) it was very subtle.  I didn't see anything in X-Ray view, Cura couldn't find anything, and 3d Builder didn't find anything.  The only real reason I sent it to Netfabb is that those types of artifacts are often in the model, and I'm hard-headed.  The fact that the file expanded from 39 to 98kb is telling us something but I have no idea what the message could be.  Subtle.

 

Your settings look fine.  There are things I would change as a matter of personal preference, but you know your machine.

[gr5 2,097]

It's better to post a project file than an stl or a profile.  The project file has everything you chose including your printer, your model, how you positioned your model and much more.  To save a project file do menu "file" "save project".

[baltic 1]

Here is the project file, sir.

bug.3mf

[GregValiant 1,144]

Once again I can't duplicate that problem.  Going back to your first image it does have the look of a model problem though.  Unfortunately the repair utility doesn't tell me what problem was encountered.  Since the "fixed" model doesn't have the problem I think you should be good to print.

I still wonder what caused the file to nearly triple in size.

Service.Netfabb.Com is an AutoDesk site.  You need to create a free account but it's a pretty good repair utility. 

[baltic 1]

Yeah thnx for the fix.

But its weird that the bug is not reproducible. Its 100% reproducible for me. At exactly the same layer.

I have 4.8.0 under linux

 

P.S. Prusa slicer eats the original stl just fine.

Edited April 6, 2021 by baltic

[baltic 1]

Huh, apparently when i check "use adaptive layers" the problem reappears even on the fixed model.

Maybe its just luck that you didn't hit it, because you have used different layer thickness, and so the slicer didn't hit the sweet spot while was cutting it to layers?

[GregValiant 1,144]

I opened the original file in IdeaMaker.  It tells me that there are 14 non-manifold edges.  So it isn't watertight and Cura didn't issue a warning.

I don't know what the difference is between Cura and Prusaslicer in this regard.  I suppose it's possible that Prusaslicer is seeing the errors and correcting them and then it moves on to slice.  I sliced it in IdeaMaker without repairing the model and the slice looked OK.

Adaptive layers is good when you are slicing something that has a ball or arch shape.  On things that are mostly planer it can cause odd effects.  That's a reason it is in the experimental section.

[baltic 1]

A-ha! I have changed the layer thickness to the ridiculous 0.01 and i now enjoy a bunch of them, even on the fixed model! 

[baltic 1]

2 minutes ago, GregValiant said:

Adaptive layers is good when you are slicing something that has a ball or arch shape. 

You don't get it. The point is not in the adaptive layers. The point is you were trying to reproduce it with different layer size, and that's the likely reason you didn't hit the bug.

Change it to something like 0.01 and you will get tonns of them all around

[GregValiant 1,144]

I couldn't get down to .01  Here is .02 with the un-repaired model.  At 18 days to print I think I could cut it out of a piece of granite faster.

 

 

[baltic 1]

6 minutes ago, GregValiant said:

At 18 days to print I think I could cut it out of a piece of granite faster.

True, but the point was to try to reproduce it. Does the 3mf store all the parameters? Or only the model configuration?

 

Or does it make sense to reproduce it? Are they going to fix it, even if the problem is in stl?

Edited April 6, 2021 by baltic

[gr5 2,097]

The point of the project file is so that we would have the exact same settings (adaptive layers versus not, bad model versus good, .01 layer versus whatever).  And also same placement of the model in the same location (in case you rotated it or moved it) and so on.

 

Also there is "fixed" and there is truly fixed.  Did you create the model yourself?  If so what CAD software?

 

Anyway this is a model problem and not something Cura team will fix.  You are correct about that.  It would be good though for you to know how to avoid this in the future.

 

This particular issue is impossible in most CAD software but is common in a few such as blender and sketchup.  If you use one of those I have suggestions on how to fix next time.

[baltic 1]

1 hour ago, gr5 said:

If so what CAD software?

Solidworks 2014

 

Strange that it's not 100% reproducible on other machine though, even with 3mf.

Model problem suppose to cause the same effect on any slicer instance.

[GregValiant 1,144]

When there are errors in a model any slicer will get "twitchy" or "flaky" or "stupid" or something besides "perfect".  The resolution that the STL was created with has an effect.  Sometimes just rotating a model will make a difference in how errors show up in a slice.  Trying to relate a fix in Cura to a model problem just isn't doable.  It does a fair job of finding some sorts of errors, but in the case of your model, It wasn't until the 3rd repair utility that the errors were actually found and fixed.  There have been some "supposed" model errors that were simply generated in the preview because of issues with compatibility between a graphics card and the computer processor and the model was actually OK.  So I think in the end that notifying a user that the model has problems is the best way to go.  "Error Descrimination" could be tweaked, but making changes in Cura so poorly constructed 3d models could be sliced would require the software to GUESS at what the model geometry really is.  In my experience that is never a good thing.  We don't have a HAL9000 and that is just as well.

[baltic 1]

4 minutes ago, GregValiant said:

When there are errors in a model any slicer will get "twitchy" or "flaky" or "stupid"

But apparently yours doesn't

Thats what i was talking about.

No matter how you have tried, you wasn't able to reproduce it. Thats rather strange.

Edited April 9, 2021 by baltic

[GregValiant 1,144]

What computer operating system are you using?

[baltic 1]

Do you believe the slicing code works somehow different on different OS? 😉

I'm afraid its not that much platform specific.

I have suspicion that it's some setting, which leads to this.

Especially coz you clearly have different settings then mine, per your screenshot. So .3mf doesn't seem to save them at all, or maybe just doesn't save all of them.

 

Besides, what makes you so sure its the .stl error? No software had reported one on the .stl so far. Yet you quickly jumped to the conclusion that it's .stl error merely based on the fact, that some web service returned you a different .stl. And the fact that you can't reproduce it on your slicer, with the same "erroneous" .stl doesn't bother you.

 

[GregValiant 1,144]

MAC systems have problems with Cura.  Some video sub-systems have trouble with Cura.

Cura is designed to work with Mac's, Windows, and Linux.  There is an old saying that says "Ambidextrous means it won't work either way" and another that says "Universal means you'll need a big hammer and a cutting torch".

 

The extraneous surfaces you show are consistent with errors in a model file.

IdeaMaker is another slicing program.  It returned that there were 14 disconnected faces.  NetFabb is an AutoDesk site designed expressly to fix STL files.  It returned a file that was almost 3 times as large as the one submitted.  There was no explanation but I view the size change as significant.

 

I used the your 3mf file and it did have your settings in it.  I then changed some settings and re-sliced.  Then I switched to my printer and my main profile.  I sliced, changed, sliced, trying to get the extra surfaces to appear.

 

What broke?  Why did it break?  What needs to be done to keep it from happening again?  I'm a retired forensic engineer.  This is what I do.  I hang around here because it allows me to keep my hand in.

 

So I asked - What operating system are you using?  It goes under my heading of "Why did it break?".  The answer may or may not provide insight.

 

 

[baltic 1]

Just now, GregValiant said:

MAC systems have problems with Cura.  Some video sub-systems have trouble with Cura.

Dude it's clearly has nothing to do with the video. Since it's in the .gcode file. It's slicing issue.

 

8 minutes ago, GregValiant said:

It returned that there were 14 disconnected faces.

Ok that explains the inferences. The only weird thing about it, is why this bug feature is not reproducible with another Cura of the same version?

 

8 minutes ago, GregValiant said:

I used the your 3mf file and it did have your settings in it.

Did you try to slice it? Did it produce the same result as for me?

 

11 minutes ago, GregValiant said:

What operating system are you using?

Linux x64, it is.

[gr5 2,097]

On 4/9/2021 at 12:57 PM, baltic said:

Did you try to slice it? Did it produce the same result as for me?

That's my question as well.  @GregValiant I thought you implied that you didn't get the problem with his project file but I'm not certain.  I don't think you were explicit.

[gr5 2,097]

I just loaded your project and sliced on Ubuntu and couldn't get the error.  It slices fine.  HOWEVER I had to reduce the part size to 98% to get it to fit on the bed.  Usually project files are supposed to store everything including scaling of the part.  It said you had an Ender 3.  Is that correct?  anyway I couldn't get it to fail.  I changed the layer height from 0.2 to 0.11 and 0.003 and I don't get those triangular shelves.

 

So either you included the wrong project file or something didn't get saved into the project file (I'm guessing something about the printer aka "machine").

 

When I opened it a second time it asked if I wanted to update my "machine" to match the machine in the project file and I said yes.  It still won't slice as the model is ever-so-slightly too large.

[GregValiant 1,144]

Hi @gr5.   I guess I was being round-a-bout.  I couldn't duplicate the error in either the first file posted, or in the one I had repaired.  I'm starting to think it might be his Cura installation.  He might need a re-install.  That's been known to happen.  I haven't seen many complaints regarding Linux systems though (unlike the MAC's running BigSur).