So I finally purchased a UM2 and I should be receiving it this Thursday. Reading this thread has me worried though - is the print quality really going to be that bad? Just looking at the cylinder print on this page makes me think my 2 1/2 year old Thing-O-Matic would do better. Should I try to find a copy of 14.01?
I tested Cura 14.05-RC2 vs Slic3r.
Left Cura, right Slic3r.
As you can see, there are still a lot of blobs on the surface with Cura.
Cura has a user friendly interface and nice features (like slice preview and object rotating).
It's a pity the engine has these flaws.
Yup I have the same issue too. Cura seems to be is very poor on these curved surfaces unless you slow the speed right down (using 14.01). If it wasn't for its garbage retraction I probably would use Slic3r more again - conversely Cura's retraction is awesome in comparison.
Yeah, the latest 14.05 RC2 is still giving me bumpy surfaces on medium resolution models. Went back to 14.01 again.
I don't have any comparison pictures yet, but I want to + 1 this effect.
I'm using 14.03 and while not as pronouced as some of your folk's bumpyness it's noticeably not as smooth as 13.11.
Watching the head, it does some slowdowns in areas where it should just be a constant speed making it slightly bulgier and is inconsistent from layer to layer.
On mine it is the movement for multiple walls that causes it. As the nozzle moves from the current wall to the next outer wall it drops a small bump of filament. So you end up with a small line at right angles to the flow of the filament. On Slic3r I tried it with option of printing the outside wall first and moving inwards instead of outwards: no bumps
yellowshark: Could it be, that this is the same problem as in following thread?
http://umforum.ultimaker.com/index.php?/topic/5581-printingtool-paths-slicer-issue/
There I have tried to convince Daid, that the additional material seems to be a problem.
Try to switch off combing. With combing on, there is no retraction on this pathes.
Hi Conz thanks for that. I have had a quick look but will read thoroughly over the weekend and see if there is anything there I can use - I do not use combing. Having read it I do wonder if setting the minimum travel distance on retraction to 0 might help but for me that causes other issues as I tried it before and got multiple retractions per second which chewed up the filament and blocked the extrusion path. I know if I drop the speed to 20mm/s it does at least for the most part disappear but why would I want to that when I can run it in Slic3r at 60mm/s without a problem!
it is amazing... even printing just a box with rounded corners... comparison 14.01 and 14.03...
14.01, when printing the outer shell, just it prints it smoothly.
14.03... in the round corners tr takes more time drawing the corner and producing some artefacts in the corner...ç
this is just a rectangle with round corners!!!...
please.. solve that BIG ISSUE
Could you send me the STL of this "box with rounded corners" please? I want to do some tests. I want to use your exact box that has problems. You can email it to - lets see - here's a temporary email:
gr-tempjan06 _at_ spamarrest.com
Change the _at_ to the @ symbol. And remove spaces.
Could you send me the STL of this "box with rounded corners" please? I want to do some tests. I want to use your exact box that has problems. You can email it to - lets see - here's a temporary email:
gr-tempjan06 _at_ spamarrest.com
Change the _at_ to the @ symbol. And remove spaces.
Thanks for your interest. Mail sent.
Follows a picture of both parts:
Cra 14.01 left, cura 14.03 right.
we can see a "bumpy" corner on the right. An the noozle follows that artifact very slowly, incrreasing printing time.
I have found in many parts that 14.03 produces worse prints and takes quite a lor of more time to print it.
I decided it was time to print braddock`s great Ieyo bust yesterday, i will post pics in the latest print thread in a minute so you can have a look.
Anyway... loaded up the model into both 14.03 and the newer 14.05 RC2. Exactly the same settings on both. 14.03 gave an estimated time of 9 hours. 14.05RC2 gave estimated time of 7 hours. So i picked 14.05.
Print came out great by the way. No problems using the 14.05 so far.
Okay - so the problem was kind of obvious once I sliced this file using both Curas and then opened the gcode in repetier host. The STL file has about 60 edges along the curved corners. Some of these edges are about 1/10 as wide as others in a seemingly random pattern. The older Cura had a feature where it would just skip over line segments that were shorter than a particular length - probably around .1mm, but the new Cura lowered that threshold by quite a bit due to demand from users like us because of complicated results related to rounding errors. I think the original threshold was better.
The reason 60 edges on a corner is a problem is that Marlin only has enough computing power to plan out about 10 edges in advance (or maybe it's 20?) and so it has to be ready to come to a complete stop after 10 edges because it doesn't know what's next so it starts decelerating maybe 10mm before then but with 60 edges taking up 3mm Marlin slows the hell down - probably to around 10mm/sec around those corners. Slowing down from a typical printing speed to 10 or 20mm/sec causes some overextrusion on those corners.
In the photos below the yellow portion on the left is the same as the highlighted portion on the right - the 14.03 photo (lower photo) has 32 line segments in that yellow portion and some are as short as .05mm!! The 14.01 photo has many fewer line segments as cura tossed out the short ones:
The obvious fix is to go to your CAD software and in the STL export settings there should be a way to specify the resolution on curves - something like that. Tell it not to output triangles smaller than say .2mm across (that's the radius of the nozzle and in one sense you can't really print finer than that anyway without getting a smaller nozzle).
There's probably some "max resolution" or "min resolution" or something that affects this.
Also maybe Daid should set the limit back to where it was. Also there will be future Marlins that run on faster processors - actually Eric Zalm is working on this now and so the UM3 will hopefully have an ARM processor which can look out 100 line segments instead of 10 and maybe there will be a UM2 and/or UM Original upgrade. But this is a ways off.
After thinking more about why Daid did this in the first place, I have a cura fix idea. The original problem was that combining these short segments (removing intermediate move points) into slightly longer moves on corners is that when you do the long edge move (the side of the square) sometimes it starts .05mm off from the layer above or below and you get a slight indent along that line - looks kind of like under extrusion. .05mm is more obvious when it affects 20mm of a print line.
The fix is to go back to the old threshold but only combine line segments (only drop intermediate travel points) if all of the line segments are each shorter than say 3mm. That way you won't combine a short move with a long move thus causing the problem that was occurring in cura 14.01.
Thanks Gr5 for finding the origin of the problems.
Some of us are using free CAD software where we cannot manipulate that settings, tinkercad in my case.
So I am worried about future versions of Cura having the same issue and with no possible solution.
Fortunately Daid is really good about fixing stuff like this. Unfortunately he is very busy working on the GUI part right now. Fortunately this is open source so it might get fixed by someone else within a few months but I doubt that. More likely Daid will come up with something.
Workarounds:
print slower
20mm/sec should be plenty as that is the "jerk" speed which means I think it is willing to instantaneously change speed by 20mm/sec from line segment to the next so no deceleration on corners necessary. 30mm/sec might also be slow enough.
Use 14.01 for now
Decimate
You can reduce polygons (I've done it for artistic things that have millions of polygons but never for basic shapes like this) here:
http://www.shapeways.com/blog/archives/226-polygon-count-reduction-with-meshlab.html
Other free CAD
I only use free CAD. Daid, and Illuminarti, and myself like: DesignSparkMechanical. Very nice. My recommendation for most CAD needs (not for people making sculpture though). I also use google sketchup. Neither of these create excess polygons. openSCAD lets you set the resolution on arcs. Isn't tinkercad based on openscad? I don't know tinkercad - never tried it.
Learning a new CAD sucks. It really sucks. I'm so damn fast at google sketchup and it's frustrating to use designSparkMechanical even though it is clearly superior. I have about 20 hours experience with DSM but 200 hours with sketchup. Also DSM always gives you manifold STL files (it isn't capable of not doing that - it's like solidWorks that way) whereas sketchup is always leaving my CAD with interior walls and such when I put two parts together. And holes. So I have to go back and clean up the model all the time if I want to print it.
Can someone send me the linke where I can get the TEST2 version?
After thinking more about why Daid did this in the first place, I have a cura fix idea. The original problem was that combining these short segments (removing intermediate move points) into slightly longer moves on corners is that when you do the long edge move (the side of the square) sometimes it starts .05mm off from the layer above or below and you get a slight indent along that line - looks kind of like under extrusion. .05mm is more obvious when it affects 20mm of a print line.
The fix is to go back to the old threshold but only combine line segments (only drop intermediate travel points) if all of the line segments are each shorter than say 3mm. That way you won't combine a short move with a long move thus causing the problem that was occurring in cura 14.01.
Actually, the problem is a complex combination of older "combing" code, with a bug in the Clipper library. The end result was the "unaligned" layers that we had a month and a half ago. So I fixed that, which introduced this new problem. Which I did my best of implementing just exactly the above (only combine if the distance is short).
I'm using 0.5mm right now, as you can see here: https://github.com/Ultimaker/CuraEngine/blob/master/src/polygonOptimizer.cpp#L16
ArunC posted a topic in UltiMaker Cura,
.thumb.jpeg.0b7a05eafc09add17b8338efde5852e9.jpeg)
Dustin posted a topic in Firmware,
Recommended Posts
Top Posters In This Topic
16
9
7
6
Popular Days
Mar 30
12
May 23
7
Apr 23
6
Apr 3
6
Top Posters In This Topic
Daid 16 posts
cor-kiekens 9 posts
gr5 7 posts
yellowshark 6 posts
Popular Days
Mar 30 2014
12 posts
May 23 2014
7 posts
Apr 23 2014
6 posts
Apr 3 2014
6 posts
profepaco 0
2 hours 40 minutes
7 cm tall, 7 cm diameter with a wall of 3 mm. 15% infill. 90 mm/s. 3 mm bottom.
do you still think is too slow?
Link to post
Share on other sites