I have so much respect for people who make a discovery, spend their time to follow up on it and then make their findings accessable to everybody. Thanks =)
THIS IS GENIUS!!!! Fantastic work! With the level of both mecanical and software complexity, non-planar printing is still somewhat far-away from being truly accessible. But THIS!!! This needs to be part of all next main slicer's major revision. Please let us know how we can help you in this journey! Cheers.
The speeds only really need to be slower for the smaller arcs. The speed limit comes from cooling. Maybe if @MirageC can develop a sub-zero cooling fan, we can print arc overhangs much faster!
Honestly the best thing to do is spread awareness and inspire the community to improve upon this idea! There are so many good suggestions out there for improving this algorithm (just see the comments on CNC kitchen's video) and I cannot possibly implement all of those suggestions myself. Since your whole deal is printing fast, maybe you could try testing out how fast these can be printed? I'm sure people would be interested in it :)
Incredible work! This would be a near way to create a quick infil, i imagine a few rods going upward to the surface and printing arcs outeards to support it. You could then print clean surfaces on top
Thank you! Yeah combining arc overhangs with a very minimal amount of support material sounds like a great way to get the benefits of both with minimal material waste.
Ideas: 1) You could replace some of the mini arcs between two existing arcs by standard bridging. That might save some time. And lead to better quality as it seems to me that the smaller the arcs, the worse the quality. Perhaps try to build into the corners first, then bridge between those? 2) It might be interesting to see if ovals would also work. Might be good for certain geometries. Some varying of the flow rate might be needed as the width of each line of plastic would no longer be constant. 3) You have demonstrated 90 degree overhang, bravo! What about slightly less? (eg, instead of the standard 45 degree overhang we can do now, can this technique be used for a 75 degree overhang?)
1. Using bridging whenever possible sounds like a great way to improve the edges of the overhang. Those tiny arcs are the cause of uneven surface finish, so getting rid of them would help a lot. 2. Yes, ovals, or even a shape that slowly grows to mold to the outer geometry. That approach is a lot more complicated mathematically, but it would definitely reduce the total number of arcs required. 3. 100% it can be used for any overhang angle, not just 90°. An 80 degree overhang is just a bunch of 90 degree overhangs one on top of another!
@@3DQue I'm afraid you can't make a traditional (less than 90 degree) overhangs with arcs though. If you try to make a slope (say 45 degrees) with a bunch of arching lines, it's going to make a cone. I don't know if there's a proper way to print rectangular inclines without straight lines, unless you use non-planar printing. I think very high (>80 degree) overhangs are possible if the outer wall is able to overlap the previous layer regardless of the line width and overhang angle. You could make the outer wall zigzag, going over the edge and back onto the previous layer many times like a square wave. I don't know if that been tried before though.
@@3DQue 4. Do the circles really need to be centered at the edges of previous circles? If no, you can eliminate the small fractal nipples. 5. Do the extruder passes really need to begin and end at the edges of adjacent circles? If no, you can actually fill the whole overhang with one circle (or some small number of circles in case of large overhangs). Maybe move the extruder head slightly towards already printed area at the beginning and the end to stick the extrusion. 6. You can start printing the second (third etc.) layer atop of the overhang amidst the process of printing such overhang, so the rigidity of the overhang will increase in order to minimize warping.
@@wormball I've actually just tried offsetting the arc centers so they are slightly inside the previous arc. This does help to reduce the 'nipple' effect! I think it is quite helpful for the lines to start and end on an existing wall. Using just a single arc sounds like an idea worth exploring, although I don't think it will work for every single shape. I agree with your idea to reduce warping, although I think it would be better to just do the next layer really slowly to avoid warping instead of printing multiple layers simultaneously.
@@SoundShunter72 You can actually! You can approximate a rectangle with circle packing. I personally think this algorithm would be only useful for really steep angles 75-90 degrees.
You need a minimum amount of cooling time between each arc. When you start the small arcs you need it to cool before you start the next one. You could go to another area and print it's arc and then come back but you would have to keep track of each arc group.
Yeah the smallest arcs are the most problematic for sure! I definitely agree, and I think there should be as much time as possible between small arcs to maximize quality. You're 100% right with having to keep track of every arc group - that adds a whole extra level of complexity, but I think it is totally doable.
That's so clever. You'd think it would just sag and never be dimensionally accurate, but it's actually really good for what is essentially just printing in mid air.
It's still amazing to me that the overhangs actually tend to curl upward instead of sag downward. Gravity really doesn't seem to make much difference sometimes!
@@3DQue Yeah, I can't figure out how that even happens. It kind of looks like a warping print, but in mid air... I don't think many people are going to use massive overgangs like those in a functional print. The smaller overhangs are probably more than sufficient, which is good, because they look incredibly clean.
Question: What is the biggest radius circle that this works for? The bigger the radius gets, the closer the perimeter gets to a straight line. So if straight lines don't work, then there has to be a limit to what radius circles work.
I believe straight lines would actually work as long as there is sufficient overlap with the nozzle orifice. I don't know what the limit is for the maximum radius - I haven't reached it yet.
You deserve a medal for this. How much landfill are you going to save the planet with this great discovery!? Would triangles be better? You just print two sides of the triangle. Triangles would be easier to compute and may be faster to print.
I saw another video where circular prints can have a perfect 90° overhang. I figured it would only be a matter of time before someone came up with a clever way to add the circles on top of eachother!
Does it work for circular objects only? What about just extending contours of the object instead of creating arcs? Could be easier and would look better. (I would try it myself, but I don't 3D print yet, I'm just a fan)
that's the fullcontrol.xyz demo. Fullcontrol gcode is going to be a huge game-changer for 3D slicing innovation, since it'll let many people experiment outside the boundaries of what normal 3D slicers allow.
Thanks! CNC kitchen's video helped give this idea a huge boost :) Hopefully this inspires the community to make their own improvements on the algorithm.
Also: I have a decent bit of experience hacking on Cura's codebase. I won't have occasion to do much for the next couple months but I could probably help you with it when I'm available again or work on integrating some of your code/algorithm myself. The basic idea, which Cura already has abstractions to do, would be subtracting the previous layer offset outward by a maximum ordinary overhang distance from current layer, and applying your algorithm to that region to print at the end of the layer (and subtracting that region from what Cura attempts to print normally).
Thanks! I tried working on this for Cura initially, but developing for Cura on Windows is a real pain. Prusaslicer was a lot easier to build. If it still isn't developed for Cura in a few months, it would be amazing if you would help me get that started :)
It's been said already but thank you for your time and thought on this algorithm. Literally two weeks ago I thought of something similar in a failed print. But I do not have any coding knowledge to make anything tangible. Thanks for your expertise and diligent work with this project. It's amazing as is but I can't wait to watch this continue to evolve!
Great stuff... arrived here from CNC Kitchen, and Subscribed. Sometimes accidents lead to a Genius idea! One way to improve the speed would be to slow the printing head for the first layer of the overhang, but subsequent vertical layers could have the print velocity increased since the underlying layer of support is already there. It might have to be a gradual increase depending on the material. I also like the idea by Dallas Prints (below) who suggested "overlapping the nozzle orifice with the previous line when building this outward"... it might be worth looking at too, especially for the first layer of the over hang.
Yeah the arcs need to be printed slowly but the rest of the print could be normal speed. I expect that some special slower print settings could be used for the layer directly after the arcs that would help reduce warping.
when the arcs reach a previously defined minimum diameter, don't continue adding smaller arcs instead, blend over from the current outline to the outline of the overhang you could also start with non circular arcs, like an offset to the inside of the overhang outline genious method! i hope it makes it into slicers
That's a really good idea. I agree that the micro-arcs on the outer edge are kind of unnecessary and ruin the final surface finish. I hope to be able to do some kind of 'interpolation" between the outer perimeter and the arcs to eliminate those micro arcs.
That is absolutely amazing. This is really useful and I do hope one day it would be an option in slicing programs. Keep up the good work 👍🏼 And thanks for sharing :)
Very Very cool, discoverd this one after CNC kitchen showed it. Hope it can be implemented to Cura Slicer. As the 2nd layer is the main issue for sagging momentarily, i wonder if we can use overlapping arcs on the 2nd layer aswell to minimize the issue. so a 3rd layer can be printed regularly. Most important: thanks to youtube and this Prior Art, we can keep the tech out of Patents of commercial company's and inside the Open Source & community Space. thanks for sharing and hope we can sponsor/credit the inventor.
Definitely looks promising. I remember one time I forgot to ture on supports for a capital i that was upright. I was surprised how well it still came out. Your 90 degree overhangs looks really good and I bet it could be used for a lot of prints.
Are you overlapping the nozzle orifice with the previous line when building this outward? Doing that should give much better surface quality, reliability, and dimensional accuracy (avoiding sag from not having something to squish against). I've basically had this concept - minus your recursive circle approach which I think is really smart because it takes advantage of convexity and how curving inward makes the overhang succeed where it would otherwise fail - on back burner for a couple years now and would love to see it become reality.
Very cool. Thanks for your hard work.... This will definitely expand the capabilities of 3D printers if popular slicers would adopt your method. I've always hated supports especially on PETG. Sometimes, it's almost impossible to remove.
Yeah, also TPU supports are an absolute nightmare. I haven't tried this overhang algorithm with anything except PLA, but that might be something to try in an upcoming video!
In CNC Kitchen Channel I leaved a comment with an idea about that, if you are interested you can check. Another idea is: the property of stitching in arc shapes used in that ways always work, so you can use elliptical paths that allow a lot more flexibility in reaching the limit of geometry that is in printing. Also with elliptical shapes the other idea on CNC Channel can be used more efficiently, due to the shapes of primary overhang printed.
There might be an in-between solution that offers the superior finish quality of a supported overhang and the filament efficiency of this. Could you build a sort of sparse arc "web" structure that attaches to the main body with a break-away pattern, and then use that as a basis for a mid-air support structure? That ways the "finish surface" still gets fully supported, and support waste is drastically reduced.
Yeah that is a really promising idea. Using Arcs as a mid-air platform to build normal support material. I think this would be really good for tall prints. Instead of 200mm tall support structures, we could just have 2mm tall supports that are just arc overhangs (gently connected to the main part) + a few interface layers for smoother bottom surfaces.
Next you can do negative overhang by ironing a seam where it can bend and then use the print head to push the flange over a bit and make it droop, then print on the bent flange.
Exactly what I was thinking. the overhang ‘hinge’ area could be printed very thin/ not much infill so the cantilevered area would droop down and possibly be pushed down by the nozzle or the nozzle assembly (so nozzle doesn’t melt part)
It would be nice to add ribs similar to the structure on leaves. It could be only a mm high to avoid hitting the gentry but would probably help reduce sagging
Yes for sure. I think some special settings for the layer right after the arcs would be needed to reduce warping. I want to explore that first before incorporating normal support structures.
It seems like the biggest droops appear at the center of the circle so maybe it can be a flatter bottom if it chooses circles so the center is slightly inside the previous one rather than having the center on the edge itself. Also, ellipses might be able to fit the shape with fewer small circles needed.
This 100%. The arc centers should definitely be slightly "inside" the previous arc to minimize the "micro-arcs". Then maybe people would stop calling it nip-overhangs LOL
I think it would be useful on extended overhangs to have a minimal support added if it is likely of too much sag from the subsequent layers. Maybe this can be added manually, picking a spot by the user.
From CNCkitchen's video, it seems the problem is the opposite of sag. The plastic contracts, causing it to bend upwards. Still, some "anchors" along the edges might help.
@0:30 is reminiscent of trochoidal CNC milling cutpaths. If you implemented a proper medial-axis calculating algo the next step to trochoidal print motions are pretty straightforward and would be quite a bit more efficient in terms of the extraneous separate moves involved with your existing algorithm where many disparate separate print moves come in to fill in the leftover space. Following the medial-axis of the 2D shape to print will be able to produce effectively the same supportless strategy but with less moves between cuts because more of them will be in a series of concentric arcs rather than becoming sprinkled around the perimeter. Good work!
Interesting! Yeah this algorithm is quite simplistic, and is more of a "brute force greedy algorithm" . It is the proof of concept that is easiest to implement. I'd love to see more complex algorithms used to simplify the toolpath and improve the surface finish.
Saw a video from CNC kitchen on this, then just now found this video. Made a comment there, but you invited comments here, so I'll repeat it. The other video shows that the bottom layer of the overhang can warp when subsequent layers are printed on top of it. It seems to me, if the overhang were constructed thicker, it may tolerate layers on top better without warping so much. Say the overhang is two or more layers thick. Inspired by conical slicing, if the extruder has the necessary clearance, you might be able to get away with a pattern like this: Print the first arc as normal. Print the second arc, then the first arc on the layer above. Print the third arc of the overhang, then the second arc of the layer above; the fourth arc then the third arc above, and so on. If that angle is too steep, you could have the top layer 'lag' by two arcs. Layer 1 arc 1, L1a2, L1a3, L2a1, L1a4, L2a2, ... L1a(n) L2a(n-2) ... The Z endstop may make this approach unworkable, though... The most sophisticated manifestation of this algorithm would calculate an appropriate 'arc lag' of the second layer (and more? third layer, fourth layer?) from the layer height and extruder nozzle diameter.
Interesting idea! My hope is to incorporate this into current slicers, so having additional Z-moves is something I would hope to avoid unless absolutely necessary. I want to first try printing the layer directly above the arc much slower to avoid the issue of plastic contraction and warping. Or maybe repeating the exact same arc-overhang pattern, but printed faster because it has support underneath.
Amazing work! I think that one way to improve the micro arches would be to try to fill that space with more elongated arches to gain quality and of course time ex: if 3 mini arches are in series it makes a curved line from point to point like a bridge but curved.
Even if you can not get the surface quality well under control, this may allow some larger bridging or - at least less material usage for supports. A small support can now grow a lot larger further up. Definitely a VERY good demonstration that now need some slicer makers to properly start implementing it into their slicers to see how far that can be pushed. Impressive.
Thanks! I think this method is definitely best for smaller overhangs. It is fun to push it to the limit, but I can see the most practical benefits coming from small overhangs that would otherwise require those tall, skinny, tower supports that end up being kind of useless.
FINALLY! I hate watching my cura overhangs just decide to put a line straight out in the middle of nowhere with a fast angle to make sure it tears off and ruins the print.
thought on strength and minimizing warping from subsequent layers: if this program could create 'ribs' that extend outward by over-extruding at the same point for each pass on the arc, it might increase the stiffness, and stop the layer from bending up or down. it would also make a cool radial pattern :p this idea also extends to other patterns besides ribs, a hexagonal rib pattern maybe? a grid?
Definitely a good option for support structure. You could do a 0% density support and us this to start a support interface. Just need one small pillar on the furthest point to prevent dropping instead of a whole support structure or a wall far an entire bridge to form.
It's been a long, long time but iirc lines joining circle centres define the voronoi medial axis, derived from the voronoi diagram for a 2D shape. Didn't see this mentioned in the video. Perhaps worth looking into as Voronoi algorithms are well established and might make the slicing job easier and more robust. Anyway really nice idea. Cheers
I've explored the idea of using algorithms like voronoi, and polygon decomposition to split shapes into much simpler convex shapes. See this algorithm for details: mpen.ca/406/bayazit
Interesting concept, however the overhangs at this point are not usable on any real world print where accuracy and flatness of the overhang is required. It will be interesting to see where thus goes and how it develops.
Amazing work! I wonder if it's possible to have the arcs start with less curvature? Perhaps by having centers of arcs slightly recessed or arcs that intersect eachother at some fixed angle? This could be really groundbreaking!
Yep that's probably one of the first things I'll work on - reducing the frequency of the micro-arcs since they are the main cause of surface imperfections.
Wonderful discovery! Could it be improved by adding a small perpendicular wavy pattern along the direction of circumference of the circles? Just a thought.
I think there needs to be further testing to see what exactly is causing this algorithm to work in the way it does to the print, what is the factor that actually allows for the overhang of the part…
Very nice. At 1:52 we can see the nozzle slinging the filament around the circle, rather than depositing it directly where it needs to be(the nozzle makes a wide circle, then "rolls" the new row of filament on the circle), which seems to work quite well. Is this intentional or accidental?
It was intentional. Using the curvature of the arcs means the filament is always pulled toward the center of the arc, improving adhesion to the previous arc.
I think that this algorithm works based on fractals, so this makes me curious if other shapes in a fractal could work better, could you perhaps modify the algorithm to use triangles instead of circles?
Here's a few thoughts I had: would printing a perimeter after all the arcs improve quality assuming that the arcs stop just shy of the true edge would printing a second layer of arcs starting on the other side improve rigidity for further prints above the primary overhangs?
I'm 100% sure it can be used to print negative overhangs. The only problem is that will involve non-planar slicing which is fundamentally impossible without a complete reworking of how current slicers work.
Thanks! I definitely want to try straight lines, however it actually isn't so easy to implement a recursive algorithm using straight lines. Arcs have a nice way of filling the space of any arbitrary shape. But I agree that straight lines would be better in the cases of simpler geometry.
@@3DQue - yes. I love ur algorithm. Can’t wait to see more of this! Just want to add. If straight lines woks there would be many applications that would make it useful. Bridging over holes (even very large holes) or square or triangular pockets. You could write an algorithm to deal with all those scenarios. U have probably thought of this - I’m just excited haha
I assume this uses a slight overlap with the previous pass to get it to self support could this be more generalized to say improve angled overhangs? In more organic objects like figurines? Or like a ball
@@3DQue ok but while your arch is a break though I don't know if as demonstrated it would leave a good finish on the overhang on a ball where I would think something more concentric would possibly be a better solution. Like a mash up between your work and nonplaner printing which has improved the looks of the tops of balls and other curved surfaces. I unfortunately can't code, to help or develop my own but hopefully you find my comment useful.
Wow. What a great idea. I hope you get it into Prusa Slicer. It would save so much! The under sides look fairly ugly. Can that be helped? (or is it just a result of close ups)
The underside quality can be improved so much with improvements to the code, I'm sure of it. Just see CNC kitchen's video - his overhangs look really good compared to mind. This initial version is just the proof of concept rather than the final version.
Have you tried to use this as a type of support? having a little disks vs. a whole column from the build plate up to the supported part. Have you tried other shapes besides an arc? can you do longer more straight arcs to fill in some of the spots? also it shows extrusion inconsistency you can see the pattern its almost like it grabs and skips a tad or its the meshing of the gears
my 2 questions are: does it work on budget printers like the anet a8? can we get a tutorial on how to install the algorithm for testing ourselves in cura and prusa slicers?
1. It works on any FDM printer, the limitation is that you need good part cooling fans, otherwise the overhangs will look bad. 2. It currently doesn't exist in any slicer. It has to be run directly from a python script. I'll definitely make a video about it once it is available within a real slicer
I'm curious, why use arcs, instead of a concentric pattern? I saw you mention that type of pattern would be tricky mathematically, but it should be already implemented elsewhere, right? Like in the top/bottom layer fill pattern?
A concentric pattern will work with simple shapes, but not for complex shapes. There are some ways to perhaps combine recursive space-filling with concentric patterns, but the code will become very complicated.
![HIWWHEE | OHM [Official MV]](http://i.ytimg.com/vi/OJJLtTLqLag/mqdefault.jpg)
I have so much respect for people who make a discovery, spend their time to follow up on it and then make their findings accessable to everybody. Thanks =)
THIS IS GENIUS!!!! Fantastic work! With the level of both mecanical and software complexity, non-planar printing is still somewhat far-away from being truly accessible. But THIS!!! This needs to be part of all next main slicer's major revision. Please let us know how we can help you in this journey! Cheers.
Oh! Looks like CNC kitchen is covering the topic in a video released today :)
@@MirageC unfortunately for you, the print speeds are quite slow for this so you're pretty much excluded from this haha
The speeds only really need to be slower for the smaller arcs. The speed limit comes from cooling. Maybe if @MirageC can develop a sub-zero cooling fan, we can print arc overhangs much faster!
Honestly the best thing to do is spread awareness and inspire the community to improve upon this idea! There are so many good suggestions out there for improving this algorithm (just see the comments on CNC kitchen's video) and I cannot possibly implement all of those suggestions myself.
Since your whole deal is printing fast, maybe you could try testing out how fast these can be printed? I'm sure people would be interested in it :)
Wow, just wow!
Incredible work! This would be a near way to create a quick infil, i imagine a few rods going upward to the surface and printing arcs outeards to support it. You could then print clean surfaces on top
Thank you! Yeah combining arc overhangs with a very minimal amount of support material sounds like a great way to get the benefits of both with minimal material waste.
@@3DQue totally?
Ideas:
1) You could replace some of the mini arcs between two existing arcs by standard bridging. That might save some time. And lead to better quality as it seems to me that the smaller the arcs, the worse the quality. Perhaps try to build into the corners first, then bridge between those?
2) It might be interesting to see if ovals would also work. Might be good for certain geometries. Some varying of the flow rate might be needed as the width of each line of plastic would no longer be constant.
3) You have demonstrated 90 degree overhang, bravo! What about slightly less? (eg, instead of the standard 45 degree overhang we can do now, can this technique be used for a 75 degree overhang?)
1. Using bridging whenever possible sounds like a great way to improve the edges of the overhang. Those tiny arcs are the cause of uneven surface finish, so getting rid of them would help a lot.
2. Yes, ovals, or even a shape that slowly grows to mold to the outer geometry. That approach is a lot more complicated mathematically, but it would definitely reduce the total number of arcs required.
3. 100% it can be used for any overhang angle, not just 90°. An 80 degree overhang is just a bunch of 90 degree overhangs one on top of another!
@@3DQue I'm afraid you can't make a traditional (less than 90 degree) overhangs with arcs though. If you try to make a slope (say 45 degrees) with a bunch of arching lines, it's going to make a cone. I don't know if there's a proper way to print rectangular inclines without straight lines, unless you use non-planar printing. I think very high (>80 degree) overhangs are possible if the outer wall is able to overlap the previous layer regardless of the line width and overhang angle. You could make the outer wall zigzag, going over the edge and back onto the previous layer many times like a square wave. I don't know if that been tried before though.
@@3DQue
4. Do the circles really need to be centered at the edges of previous circles? If no, you can eliminate the small fractal nipples.
5. Do the extruder passes really need to begin and end at the edges of adjacent circles? If no, you can actually fill the whole overhang with one circle (or some small number of circles in case of large overhangs). Maybe move the extruder head slightly towards already printed area at the beginning and the end to stick the extrusion.
6. You can start printing the second (third etc.) layer atop of the overhang amidst the process of printing such overhang, so the rigidity of the overhang will increase in order to minimize warping.
@@wormball I've actually just tried offsetting the arc centers so they are slightly inside the previous arc. This does help to reduce the 'nipple' effect!
I think it is quite helpful for the lines to start and end on an existing wall. Using just a single arc sounds like an idea worth exploring, although I don't think it will work for every single shape.
I agree with your idea to reduce warping, although I think it would be better to just do the next layer really slowly to avoid warping instead of printing multiple layers simultaneously.
@@SoundShunter72 You can actually! You can approximate a rectangle with circle packing. I personally think this algorithm would be only useful for really steep angles 75-90 degrees.
someone pay this man a fortune right now.
You need a minimum amount of cooling time between each arc. When you start the small arcs you need it to cool before you start the next one. You could go to another area and print it's arc and then come back but you would have to keep track of each arc group.
Yeah the smallest arcs are the most problematic for sure! I definitely agree, and I think there should be as much time as possible between small arcs to maximize quality. You're 100% right with having to keep track of every arc group - that adds a whole extra level of complexity, but I think it is totally doable.
Could you use regular line printing for smaller arcs? Almost treat it as infill?
Or print along (around?) the arcs to reinforce them?
@@francisdoherty5580 This sounds very similar to the changes I made in the update video I just uploaded. Check out the newest video on this channel
That's so clever. You'd think it would just sag and never be dimensionally accurate, but it's actually really good for what is essentially just printing in mid air.
It's still amazing to me that the overhangs actually tend to curl upward instead of sag downward. Gravity really doesn't seem to make much difference sometimes!
@@3DQue Yeah, I can't figure out how that even happens. It kind of looks like a warping print, but in mid air... I don't think many people are going to use massive overgangs like those in a functional print. The smaller overhangs are probably more than sufficient, which is good, because they look incredibly clean.
Question:
What is the biggest radius circle that this works for? The bigger the radius gets, the closer the perimeter gets to a straight line. So if straight lines don't work, then there has to be a limit to what radius circles work.
I believe straight lines would actually work as long as there is sufficient overlap with the nozzle orifice. I don't know what the limit is for the maximum radius - I haven't reached it yet.
You deserve a medal for this. How much landfill are you going to save the planet with this great discovery!?
Would triangles be better? You just print two sides of the triangle. Triangles would be easier to compute and may be faster to print.
I saw another video where circular prints can have a perfect 90° overhang. I figured it would only be a matter of time before someone came up with a clever way to add the circles on top of eachother!
Does it work for circular objects only? What about just extending contours of the object instead of creating arcs? Could be easier and would look better. (I would try it myself, but I don't 3D print yet, I'm just a fan)
that's the fullcontrol.xyz demo. Fullcontrol gcode is going to be a huge game-changer for 3D slicing innovation, since it'll let many people experiment outside the boundaries of what normal 3D slicers allow.
@@az09haz Stay tuned for the next video. I'm combining your idea of offsetting the contour and connecting it to the arcs to make the best of both.
I really hope this becomes a Cura plugin one day ! Love this stuff.
Saw this on reddit, so excited you are continuing to push this forward!
Thanks! CNC kitchen's video helped give this idea a huge boost :) Hopefully this inspires the community to make their own improvements on the algorithm.
Also: I have a decent bit of experience hacking on Cura's codebase. I won't have occasion to do much for the next couple months but I could probably help you with it when I'm available again or work on integrating some of your code/algorithm myself. The basic idea, which Cura already has abstractions to do, would be subtracting the previous layer offset outward by a maximum ordinary overhang distance from current layer, and applying your algorithm to that region to print at the end of the layer (and subtracting that region from what Cura attempts to print normally).
Thanks! I tried working on this for Cura initially, but developing for Cura on Windows is a real pain. Prusaslicer was a lot easier to build. If it still isn't developed for Cura in a few months, it would be amazing if you would help me get that started :)
This technique allows for sealed, large hollows in objects, so you can e.g. print a sealed box that you can then let swim or such.
This kinda reminds me of an origami design technique called „circle packing“. I believe it was described thoroughly by Robert Lang?
You're right; the world does need to see this 😎
It's been said already but thank you for your time and thought on this algorithm. Literally two weeks ago I thought of something similar in a failed print. But I do not have any coding knowledge to make anything tangible. Thanks for your expertise and diligent work with this project. It's amazing as is but I can't wait to watch this continue to evolve!
Great stuff... arrived here from CNC Kitchen, and Subscribed. Sometimes accidents lead to a Genius idea!
One way to improve the speed would be to slow the printing head for the first layer of the overhang, but subsequent vertical layers could have the print velocity increased since the underlying layer of support is already there. It might have to be a gradual increase depending on the material. I also like the idea by Dallas Prints (below) who suggested "overlapping the nozzle orifice with the previous line when building this outward"... it might be worth looking at too, especially for the first layer of the over hang.
Yeah the arcs need to be printed slowly but the rest of the print could be normal speed.
I expect that some special slower print settings could be used for the layer directly after the arcs that would help reduce warping.
when the arcs reach a previously defined minimum diameter, don't continue adding smaller arcs
instead, blend over from the current outline to the outline of the overhang
you could also start with non circular arcs, like an offset to the inside of the overhang outline
genious method! i hope it makes it into slicers
That's a really good idea. I agree that the micro-arcs on the outer edge are kind of unnecessary and ruin the final surface finish.
I hope to be able to do some kind of 'interpolation" between the outer perimeter and the arcs to eliminate those micro arcs.
This is sorcery of the highest order!
That is absolutely amazing. This is really useful and I do hope one day it would be an option in slicing programs.
Keep up the good work 👍🏼
And thanks for sharing :)
This shows a lot of promise, if not for removing support material, certainly for minimising it!
Who knows where it will lead?
This is a serious game changer man, so much saved filament
Very Very cool, discoverd this one after CNC kitchen showed it.
Hope it can be implemented to Cura Slicer.
As the 2nd layer is the main issue for sagging momentarily, i wonder if we can use overlapping arcs on the 2nd layer aswell to minimize the issue. so a 3rd layer can be printed regularly.
Most important: thanks to youtube and this Prior Art, we can keep the tech out of Patents of commercial company's and inside the Open Source & community Space.
thanks for sharing and hope we can sponsor/credit the inventor.
Definitely looks promising. I remember one time I forgot to ture on supports for a capital i that was upright. I was surprised how well it still came out. Your 90 degree overhangs looks really good and I bet it could be used for a lot of prints.
Yeah It's crazy how prints find a way of succeeding sometimes even when they really shouldn't.
Saw it on Stephan's channel. Glad I stumbled upon the original as well. Great job!
Are you overlapping the nozzle orifice with the previous line when building this outward? Doing that should give much better surface quality, reliability, and dimensional accuracy (avoiding sag from not having something to squish against). I've basically had this concept - minus your recursive circle approach which I think is really smart because it takes advantage of convexity and how curving inward makes the overhang succeed where it would otherwise fail - on back burner for a couple years now and would love to see it become reality.
Very cool. Thanks for your hard work.... This will definitely expand the capabilities of 3D printers if popular slicers would adopt your method. I've always hated supports especially on PETG. Sometimes, it's almost impossible to remove.
Yeah, also TPU supports are an absolute nightmare. I haven't tried this overhang algorithm with anything except PLA, but that might be something to try in an upcoming video!
Oh, that practical application at 5:23 looks *really* nice.
Thanks! This is the kind of use case that I would actually expect to see most often. The ridiculous random shapes are more of a stress test.
Thats a right drop on you should try getting in touch with slicer makers
In CNC Kitchen Channel I leaved a comment with an idea about that, if you are interested you can check. Another idea is: the property of stitching in arc shapes used in that ways always work, so you can use elliptical paths that allow a lot more flexibility in reaching the limit of geometry that is in printing. Also with elliptical shapes the other idea on CNC Channel can be used more efficiently, due to the shapes of primary overhang printed.
This is nuts man. I suspected something like this might be possible, because ive had some weird failures before. But sheesh
There might be an in-between solution that offers the superior finish quality of a supported overhang and the filament efficiency of this. Could you build a sort of sparse arc "web" structure that attaches to the main body with a break-away pattern, and then use that as a basis for a mid-air support structure? That ways the "finish surface" still gets fully supported, and support waste is drastically reduced.
Yeah that is a really promising idea. Using Arcs as a mid-air platform to build normal support material. I think this would be really good for tall prints. Instead of 200mm tall support structures, we could just have 2mm tall supports that are just arc overhangs (gently connected to the main part) + a few interface layers for smoother bottom surfaces.
Next you can do negative overhang by ironing a seam where it can bend and then use the print head to push the flange over a bit and make it droop, then print on the bent flange.
Negative overhangs are definitely possible with this method, I'm sure of it. It just requires non-planar slicing which is an entirely different beast.
Exactly what I was thinking. the overhang ‘hinge’ area could be printed very thin/ not much infill so the cantilevered area would droop down and possibly be pushed down by the nozzle or the nozzle assembly (so nozzle doesn’t melt part)
It would be nice to add ribs similar to the structure on leaves. It could be only a mm high to avoid hitting the gentry but would probably help reduce sagging
Yes for sure. I think some special settings for the layer right after the arcs would be needed to reduce warping. I want to explore that first before incorporating normal support structures.
It seems like the biggest droops appear at the center of the circle so maybe it can be a flatter bottom if it chooses circles so the center is slightly inside the previous one rather than having the center on the edge itself. Also, ellipses might be able to fit the shape with fewer small circles needed.
This 100%. The arc centers should definitely be slightly "inside" the previous arc to minimize the "micro-arcs". Then maybe people would stop calling it nip-overhangs LOL
I think it would be useful on extended overhangs to have a minimal support added if it is likely of too much sag from the subsequent layers. Maybe this can be added manually, picking a spot by the user.
From CNCkitchen's video, it seems the problem is the opposite of sag. The plastic contracts, causing it to bend upwards. Still, some "anchors" along the edges might help.
That's great! I hate supports, they always mess up my prints, look, and take forever
Great stuff. There's so much room for improvement in 3d printing on the software side. Great to see someone doing development. Cheers
Software is the next frontier for FDM 3D printing. There is still so much to discover on the software side.
@0:30 is reminiscent of trochoidal CNC milling cutpaths. If you implemented a proper medial-axis calculating algo the next step to trochoidal print motions are pretty straightforward and would be quite a bit more efficient in terms of the extraneous separate moves involved with your existing algorithm where many disparate separate print moves come in to fill in the leftover space. Following the medial-axis of the 2D shape to print will be able to produce effectively the same supportless strategy but with less moves between cuts because more of them will be in a series of concentric arcs rather than becoming sprinkled around the perimeter. Good work!
Interesting! Yeah this algorithm is quite simplistic, and is more of a "brute force greedy algorithm" . It is the proof of concept that is easiest to implement. I'd love to see more complex algorithms used to simplify the toolpath and improve the surface finish.
I don't know much about programming but 1 suggestion for easier calculations.. A honeycomb approach.. Brilliant work!
This is sick, I love how this came from a mistake, as most great things usually do. Great work!
If you keep trying to do things that are "impossible", eventually good ideas will emerge :)
Incredibly Impressive this
Incredible!!!!! Nice work
This amazing, nice work mate!
Saw a video from CNC kitchen on this, then just now found this video. Made a comment there, but you invited comments here, so I'll repeat it.
The other video shows that the bottom layer of the overhang can warp when subsequent layers are printed on top of it. It seems to me, if the overhang were constructed thicker, it may tolerate layers on top better without warping so much.
Say the overhang is two or more layers thick. Inspired by conical slicing, if the extruder has the necessary clearance, you might be able to get away with a pattern like this: Print the first arc as normal. Print the second arc, then the first arc on the layer above. Print the third arc of the overhang, then the second arc of the layer above; the fourth arc then the third arc above, and so on.
If that angle is too steep, you could have the top layer 'lag' by two arcs. Layer 1 arc 1, L1a2, L1a3, L2a1, L1a4, L2a2, ... L1a(n) L2a(n-2) ...
The Z endstop may make this approach unworkable, though...
The most sophisticated manifestation of this algorithm would calculate an appropriate 'arc lag' of the second layer (and more? third layer, fourth layer?) from the layer height and extruder nozzle diameter.
Interesting idea! My hope is to incorporate this into current slicers, so having additional Z-moves is something I would hope to avoid unless absolutely necessary.
I want to first try printing the layer directly above the arc much slower to avoid the issue of plastic contraction and warping. Or maybe repeating the exact same arc-overhang pattern, but printed faster because it has support underneath.
You can have a single arc that just starts and ends at different points to cover the whole overhang. I think it could have better surface quality
Thanks for sharing!
Amazing work! I think that one way to improve the micro arches would be to try to fill that space with more elongated arches to gain quality and of course time ex: if 3 mini arches are in series it makes a curved line from point to point like a bridge but curved.
Yeah, good idea. The micro-arches are definitely the cause of poor surface quality, and if we can eliminate them, the arcs would look a lot better.
you should get the Nobel Prize of 3d Printing
Even if you can not get the surface quality well under control, this may allow some larger bridging or - at least less material usage for supports. A small support can now grow a lot larger further up. Definitely a VERY good demonstration that now need some slicer makers to properly start implementing it into their slicers to see how far that can be pushed. Impressive.
Thanks! I think this method is definitely best for smaller overhangs. It is fun to push it to the limit, but I can see the most practical benefits coming from small overhangs that would otherwise require those tall, skinny, tower supports that end up being kind of useless.
FINALLY! I hate watching my cura overhangs just decide to put a line straight out in the middle of nowhere with a fast angle to make sure it tears off and ruins the print.
Awesome idea. Thanks for sharing!
Very cool and well thought, cooling is a major factor and will be great if slicer software implements it.
Crazy !! Wow .. glad we have people like you with incredible talent and knowledge and skills!! Thanks man
Thanks! You have your own special way of pushing the limits of 3D printing! Maybe one day we'll see 2000 mm/s arc overhangs :)
Amazing, thank you for sharing
thought on strength and minimizing warping from subsequent layers:
if this program could create 'ribs' that extend outward by over-extruding at the same point for each pass on the arc, it might increase the stiffness, and stop the layer from bending up or down. it would also make a cool radial pattern :p
this idea also extends to other patterns besides ribs, a hexagonal rib pattern maybe? a grid?
Amazing! This is transformational. Can I nominate you for the Nobel Prize in 3d printing?
Amazing! Great job👏🏿
Commenting for the algorithm this idea has to spread!
Cool work!
this is cutting edge technology!
Awsome -> directly subscribed 😅👍🏻
Definitely a good option for support structure. You could do a 0% density support and us this to start a support interface. Just need one small pillar on the furthest point to prevent dropping instead of a whole support structure or a wall far an entire bridge to form.
It's been a long, long time but iirc lines joining circle centres define the voronoi medial axis, derived from the voronoi diagram for a 2D shape. Didn't see this mentioned in the video. Perhaps worth looking into as Voronoi algorithms are well established and might make the slicing job easier and more robust. Anyway really nice idea. Cheers
I've explored the idea of using algorithms like voronoi, and polygon decomposition to split shapes into much simpler convex shapes. See this algorithm for details: mpen.ca/406/bayazit
Hey cura. CURA! GIVE HIM MONEY. Cura 5.4
Great discovery/invention! I love it!
Interesting concept, however the overhangs at this point are not usable on any real world print where accuracy and flatness of the overhang is required. It will be interesting to see where thus goes and how it develops.
Your brain is handsome.. you da man
Amazing work! I wonder if it's possible to have the arcs start with less curvature? Perhaps by having centers of arcs slightly recessed or arcs that intersect eachother at some fixed angle? This could be really groundbreaking!
Yep that's probably one of the first things I'll work on - reducing the frequency of the micro-arcs since they are the main cause of surface imperfections.
Wonderful discovery! Could it be improved by adding a small perpendicular wavy pattern along the direction of circumference of the circles? Just a thought.
Can we please get prusa to see this? we need it asap lol
Bravo!
Wow! Hats off to you sir
Thanks!
Fantastic work!
Yea this is incredible.
There were some cool ideas for improvements in the comments to CNCkitchen's video about this.
I think there needs to be further testing to see what exactly is causing this algorithm to work in the way it does to the print, what is the factor that actually allows for the overhang of the part…
How it's done. Take advantage of failure, mistakes and accidental discoveries and make the best of it.
You bet!
Very nice. At 1:52 we can see the nozzle slinging the filament around the circle, rather than depositing it directly where it needs to be(the nozzle makes a wide circle, then "rolls" the new row of filament on the circle), which seems to work quite well. Is this intentional or accidental?
It was intentional. Using the curvature of the arcs means the filament is always pulled toward the center of the arc, improving adhesion to the previous arc.
Amazing, but why does it work?
This needs to be implemented in a slicer
I know all the slicer devs are now aware of this, so hopefully it's just a matter of time :)
I think that this algorithm works based on fractals, so this makes me curious if other shapes in a fractal could work better, could you perhaps modify the algorithm to use triangles instead of circles?
Here's a few thoughts I had:
would printing a perimeter after all the arcs improve quality assuming that the arcs stop just shy of the true edge
would printing a second layer of arcs starting on the other side improve rigidity for further prints above the primary overhangs?
Hopefully somebody tries it out and lets us know.
cool
...Like a fractal structure imposed on the circles
I love fractals! So cool :)
That's great at 90 degrees. What about say, 100 degrees?
Depends on the nozzle and cooling placement, and once we're playing with those the complexity grows insanely fast. As do the capabilities.
I'm 100% sure it can be used to print negative overhangs. The only problem is that will involve non-planar slicing which is fundamentally impossible without a complete reworking of how current slicers work.
Dude, that looks awesome!!! Can't wait to try it myself, so many opportunities.
Did anyone try it on Anycubic i3 Mega S?
Surely if arcs works - then straight lines could work too? Can you try this for straight over-hangs like in ur part at the end?
Great stuff btw!
Thanks! I definitely want to try straight lines, however it actually isn't so easy to implement a recursive algorithm using straight lines. Arcs have a nice way of filling the space of any arbitrary shape. But I agree that straight lines would be better in the cases of simpler geometry.
@@3DQue - yes. I love ur algorithm. Can’t wait to see more of this!
Just want to add. If straight lines woks there would be many applications that would make it useful. Bridging over holes (even very large holes) or square or triangular pockets. You could write an algorithm to deal with all those scenarios. U have probably thought of this - I’m just excited haha
Has anybody seen this? I mean "Helloooo, Error" Cura community, please do your thing. I could use this.
What about print first layers with this technique...? It could maybe avoid warping. Or?
Brilliant!
I assume this uses a slight overlap with the previous pass to get it to self support could this be more generalized to say improve angled overhangs? In more organic objects like figurines? Or like a ball
This also works on angles less than 90 degrees. Once an object is sliced, any overhang is technically a 90 degree overhang because layers are flat.
@@3DQue ok but while your arch is a break though I don't know if as demonstrated it would leave a good finish on the overhang on a ball where I would think something more concentric would possibly be a better solution.
Like a mash up between your work and nonplaner printing which has improved the looks of the tops of balls and other curved surfaces.
I unfortunately can't code, to help or develop my own but hopefully you find my comment useful.
@@kevinbruce5981 Good news! Steven is also a master of nonplanar and working on cleaning up the overhangs.
Wow. What a great idea. I hope you get it into Prusa Slicer. It would save so much! The under sides look fairly ugly. Can that be helped? (or is it just a result of close ups)
The underside quality can be improved so much with improvements to the code, I'm sure of it. Just see CNC kitchen's video - his overhangs look really good compared to mind. This initial version is just the proof of concept rather than the final version.
One thousandth like!!!
Have you tried to use this as a type of support? having a little disks vs. a whole column from the build plate up to the supported part. Have you tried other shapes besides an arc? can you do longer more straight arcs to fill in some of the spots? also it shows extrusion inconsistency you can see the pattern its almost like it grabs and skips a tad or its the meshing of the gears
my 2 questions are:
does it work on budget printers like the anet a8?
can we get a tutorial on how to install the algorithm for testing ourselves in cura and prusa slicers?
1. It works on any FDM printer, the limitation is that you need good part cooling fans, otherwise the overhangs will look bad.
2. It currently doesn't exist in any slicer. It has to be run directly from a python script. I'll definitely make a video about it once it is available within a real slicer
I'm curious, why use arcs, instead of a concentric pattern? I saw you mention that type of pattern would be tricky mathematically, but it should be already implemented elsewhere, right? Like in the top/bottom layer fill pattern?
A concentric pattern will work with simple shapes, but not for complex shapes. There are some ways to perhaps combine recursive space-filling with concentric patterns, but the code will become very complicated.
Freaking rad!
amazing!!