Some really good improvements as well as a few surprising results with some of those ideas not actually panning out as expected. Thanks for making this video.
Very clever. Could maybe include some bridging like somehow two arc overhangs reach the corners then bridging connects (scrap that - probably only good on straight edged objects).
I would say reducing the temp as the printer slows down would work better. One thing I've learned with temperature is that it's all relative, you should not print at the same temp when you change layer height for example. I found when I print too slow and thin, even at the lowest PLA temp suggestion, the filament burns, droops, looks bad and jams my nozzle. If I speed it up it doesn't have time to do this, but sometimes you need to print slow on intricate parts. I'm guessing if you lower the temp by 5c when it gets to the slow parts of the print it will help a lot. PS: Is this ever going to come to Cura?
I feel like one thing that needs to be considered more as well is how to prevent subsequent layers from causing warping (as demonstrated by CNCkitchen). But yeah, there are so many different things to try.
Does this method print over the already existing support tower or not? Like the cylinder used just to start the overhang test, does the arcs print on top of that, or is it purely supporting itself? Not sure if this is already the case, but I think it would yield better results if it were to print over the existing support
What about doing double passes, one with extrusion, then a second pass with cooling fan only at 100% to help it solidify? If done on every arc, it may be worse, but maybe every 3 arcs or every other arc would help the part to solidify and give a more solid base?
Although it sounds like a great idea, it might need some refinement. The cooling fans and the nozzle move in unison, which means the fans remove heat with each pass, and the nozzle inputs it back through conduction and radiation, hindering the cooling efficiency. A short pause after each arc, where the print head moves away and lets it cool passively, could work slightly better? Either way, the video proves that things can turn out vastly different than expected, so it might work well enough to make a noticeable difference.
@@vit.budina certainly, but there is nothing stopping him from raising the nozzle on the 2nd pass. You will still get cooling, but not touch the nozzle. This could add benefits as the air from the fan would be a bit more gentle.
Interesting idea! I would usually avoid ideas that involve Z movement but in this case there's no printing, just cooling. Definitely worth a try for the smallest circles that need more cooling.
What's the usual lifecycle of an explosive new feature like this showing up in common slicers? I've had a conundrum with an upside-down bowl-shaped print that I need to be solid with 0% infill, but that prevents Cura from added support structure inside the bowl's ceiling. If I could print the top layers of the bowl as arc overhangs without needing support, it'd be a long-awaited cure to my problem, but I have a feeling it'll take a while to reach that point. Unfortunately, from a glance at the github (which I admittedly am unfamiliar with), it sounds like I can't slice my mesh, convert the ceilings to arc overhangs, and insert that into the original slice. However, the speed with which these kinds of discoveries are made and improved upon is incredible these days.
The lifecycle is explosive! About 1 month after I made the code public, someone added it into a real slicer. Github user rvmn has already implemented this algorithm into SuperSlicer. It's almost ready for testing. Check out the demo here: github.com/stmcculloch/arc-overhang/issues/10#issuecomment-1373436886 I'll be making a video about this soon.
How much are you overlapping successive arcs? I would hypothesize you want the nozzle orifice to overlap at least 30% with the previous arc, and to only extrude the remainder of the width (e.g. 70%) to have something to press against and not droop. I don't think I'll have time to do so soon but I'd really like to test some of this myself now.
I'm using the following print settings, and it seems to work well for 0.4mm nozzles: 0.35mm line width (12.5% overlap) 1.05x flow multiplier 0.4mm layer height 205 degrees hotend temp 2 mm/s print speed 100% fan speed
I'll try this to see how well it works, although it's not clear if it wil still work because at some point a new arc won't have an anchor point that begins on an already printed surface, so it might need a lot of overlap (overextrusion) between the lines to work
I think a whole movile bed would be cool to see. Where every point of the bed can move up to act as a support. We would only need printed supports on parts inside the pieces.
Have you thought about training a defect detection model on endoscope footage of the nozzle. Ai build does this but it sucks that none of it's open source.
Definitely sounds like a great idea, but in terms of making it something that would be widely used, endoscopes would have to become common-place on 3D printers. At this point, most 3D printers don't even come with webcams by default so endoscopes are a bit too far out of reach I think. Webcams are great because they're easy to retrofit onto any 3D printer. Just point it at the printer and plug it in. Endoscopes are more difficult. There would be different mounting solutions for various 3D printer hotends, and then we'd have to ask people to design their own mounts, and that's a pretty large barrier.
Yeah it seems like there's some pressure still in the nozzle, so as soon as there is space available, plastic escapes even though no extrusion command was sent.
@@3DQue Are you not using properly calibrated pressure advance/linear advance? If not you really should fix that first. It may not make too much difference at these low speeds unless you're on bowden, but tuning on top of an unsound base extrusion setup seems like a bad path.
I wander If it's possible to not have abrupt transition from arcs to concentric but gradualy warp from one to another, probably it'll need variable extrusion width
@@3DQue I don't use python often, but I'll try forking and implementing this approach. The most difficult aspect seems to be to decide when to stop drawing arcs and have a shape of completed outline, after that it's just bunch of lerps between completed shape and target shape
@@AlexServirog Yeah, the 2 things are the linear interpolation and the variable line width extrusion calculations. If you figure that out it would be amazing! Also, last night I made some more progress that greatly simplifies the shape even further. Will post a short video soon about it.
@@3DQue short intro would be great so all people willing to contribute are on the same page. I'll move to github issues so that discussion is somewhat more organized/permanent than youtube comments =)
Why not just extend the arcs to the full width and keep on adding arcs without restarting the arcs with 0 radius. Keep the radius constant when it reaches the edge and offset it from the previous arc until you reach the far edge. Varying the extrusion rate could also allow you to add a transformation that when you reach the edge, the top and bottom corners are filled as well. It seems to me the smaller the arcs are the more droop there is. I'd try to find an algorithm that tries to fit the area with larger arcs aside from the starting arc from the first edge. I'd also try arcs that are the width of the edge but with such a large radius it's almost a straight line then keep making it rounder or something. Maybe add some padding layers at the far end of the overhang to level out the area for the next layer.
Cura, prusaslicer, and superslicer are all aware of this, and they all have full permission to incorporate it. Superslicer will probably be the first to implement it because SS people are actively working on it now!
Some really good improvements as well as a few surprising results with some of those ideas not actually panning out as expected.
Thanks for making this video.
Thanks!
Very clever. Could maybe include some bridging like somehow two arc overhangs reach the corners then bridging connects (scrap that - probably only good on straight edged objects).
I would say reducing the temp as the printer slows down would work better. One thing I've learned with temperature is that it's all relative, you should not print at the same temp when you change layer height for example. I found when I print too slow and thin, even at the lowest PLA temp suggestion, the filament burns, droops, looks bad and jams my nozzle. If I speed it up it doesn't have time to do this, but sometimes you need to print slow on intricate parts. I'm guessing if you lower the temp by 5c when it gets to the slow parts of the print it will help a lot.
PS: Is this ever going to come to Cura?
If you try this, let us know how it works out!
@@3DQue If it ever comes to Cura then I will, since that's my weapon of choice.
@@ArcanePath360 Fingers crossed!
I feel like one thing that needs to be considered more as well is how to prevent subsequent layers from causing warping (as demonstrated by CNCkitchen). But yeah, there are so many different things to try.
Yeah this is probably what I'll explore next
I think if the above arc Part is printed veeery slow it might solve the warping as it does with warping at conventional strong overhangs
@@Kaliumcyanidful Good idea, I think I will try that soon!
Does this method print over the already existing support tower or not? Like the cylinder used just to start the overhang test, does the arcs print on top of that, or is it purely supporting itself? Not sure if this is already the case, but I think it would yield better results if it were to print over the existing support
what if instead of using "circles" you use elipses? so that the smaller arcs are bigger?
Hmmmm
What about doing double passes, one with extrusion, then a second pass with cooling fan only at 100% to help it solidify? If done on every arc, it may be worse, but maybe every 3 arcs or every other arc would help the part to solidify and give a more solid base?
Although it sounds like a great idea, it might need some refinement. The cooling fans and the nozzle move in unison, which means the fans remove heat with each pass, and the nozzle inputs it back through conduction and radiation, hindering the cooling efficiency. A short pause after each arc, where the print head moves away and lets it cool passively, could work slightly better? Either way, the video proves that things can turn out vastly different than expected, so it might work well enough to make a noticeable difference.
@@vit.budina certainly, but there is nothing stopping him from raising the nozzle on the 2nd pass. You will still get cooling, but not touch the nozzle. This could add benefits as the air from the fan would be a bit more gentle.
Interesting idea! I would usually avoid ideas that involve Z movement but in this case there's no printing, just cooling. Definitely worth a try for the smallest circles that need more cooling.
What's the usual lifecycle of an explosive new feature like this showing up in common slicers? I've had a conundrum with an upside-down bowl-shaped print that I need to be solid with 0% infill, but that prevents Cura from added support structure inside the bowl's ceiling. If I could print the top layers of the bowl as arc overhangs without needing support, it'd be a long-awaited cure to my problem, but I have a feeling it'll take a while to reach that point.
Unfortunately, from a glance at the github (which I admittedly am unfamiliar with), it sounds like I can't slice my mesh, convert the ceilings to arc overhangs, and insert that into the original slice. However, the speed with which these kinds of discoveries are made and improved upon is incredible these days.
The lifecycle is explosive! About 1 month after I made the code public, someone added it into a real slicer. Github user rvmn has already implemented this algorithm into SuperSlicer. It's almost ready for testing. Check out the demo here: github.com/stmcculloch/arc-overhang/issues/10#issuecomment-1373436886
I'll be making a video about this soon.
How much are you overlapping successive arcs? I would hypothesize you want the nozzle orifice to overlap at least 30% with the previous arc, and to only extrude the remainder of the width (e.g. 70%) to have something to press against and not droop. I don't think I'll have time to do so soon but I'd really like to test some of this myself now.
I'm using the following print settings, and it seems to work well for 0.4mm nozzles:
0.35mm line width (12.5% overlap)
1.05x flow multiplier
0.4mm layer height
205 degrees hotend temp
2 mm/s print speed
100% fan speed
great stuff
you could could eliminate some more of the small arcs by not starting new ones, just shift the biggest loop forward
I'll try this to see how well it works, although it's not clear if it wil still work because at some point a new arc won't have an anchor point that begins on an already printed surface, so it might need a lot of overlap (overextrusion) between the lines to work
I think a whole movile bed would be cool to see. Where every point of the bed can move up to act as a support. We would only need printed supports on parts inside the pieces.
Interesting
Have you thought about training a defect detection model on endoscope footage of the nozzle. Ai build does this but it sucks that none of it's open source.
Definitely sounds like a great idea, but in terms of making it something that would be widely used, endoscopes would have to become common-place on 3D printers. At this point, most 3D printers don't even come with webcams by default so endoscopes are a bit too far out of reach I think.
Webcams are great because they're easy to retrofit onto any 3D printer. Just point it at the printer and plug it in. Endoscopes are more difficult. There would be different mounting solutions for various 3D printer hotends, and then we'd have to ask people to design their own mounts, and that's a pretty large barrier.
Under-extrude when it first starts making a new circle. The blob is probably because there's just too much plastic there.
Yeah it seems like there's some pressure still in the nozzle, so as soon as there is space available, plastic escapes even though no extrusion command was sent.
@@3DQue Are you not using properly calibrated pressure advance/linear advance? If not you really should fix that first. It may not make too much difference at these low speeds unless you're on bowden, but tuning on top of an unsound base extrusion setup seems like a bad path.
我自己也在装一台VORON 2.4看到这些觉得这个真的好难，不知道这个是什么切片软件做出来，很厉害！
I wander If it's possible to not have abrupt transition from arcs to concentric but gradualy warp from one to another, probably it'll need variable extrusion width
I wanted to do that initially, just programming it is more difficult. I think that idea it is worth exploring though for sure.
@@3DQue I don't use python often, but I'll try forking and implementing this approach. The most difficult aspect seems to be to decide when to stop drawing arcs and have a shape of completed outline, after that it's just bunch of lerps between completed shape and target shape
@@AlexServirog Yeah, the 2 things are the linear interpolation and the variable line width extrusion calculations. If you figure that out it would be amazing! Also, last night I made some more progress that greatly simplifies the shape even further. Will post a short video soon about it.
@@3DQue short intro would be great so all people willing to contribute are on the same page. I'll move to github issues so that discussion is somewhat more organized/permanent than youtube comments =)
how can i deactivate arc overhangs on pleccer?
Srsly?
So 2024, where can I use this?
Great question ... people are working on getting it into slicers
@@3DQueAnywhere I can lurk and read about any progress being made?
@@D4RKFiB3R Github - link in description. Haven't heard much recently but the slicer teams are pretty busy with new features.
Why not just extend the arcs to the full width and keep on adding arcs without restarting the arcs with 0 radius. Keep the radius constant when it reaches the edge and offset it from the previous arc until you reach the far edge. Varying the extrusion rate could also allow you to add a transformation that when you reach the edge, the top and bottom corners are filled as well.
It seems to me the smaller the arcs are the more droop there is. I'd try to find an algorithm that tries to fit the area with larger arcs aside from the starting arc from the first edge. I'd also try arcs that are the width of the edge but with such a large radius it's almost a straight line then keep making it rounder or something.
Maybe add some padding layers at the far end of the overhang to level out the area for the next layer.
If you try this, let us know how it works!
Please tell me CURA already reached out to you, either to hire you or component your skill and ask for permission to incorporate your idea
Cura, prusaslicer, and superslicer are all aware of this, and they all have full permission to incorporate it. Superslicer will probably be the first to implement it because SS people are actively working on it now!
Can we please not call them nip overhangs?
I think we can safely assume that won't be their official name