Yes, material cost is a concern. This is a sort of roto-moulding alternative that is tool-less/on-demand, so it's not going to be the first choice for low-cost items
filament doesn't need to be expensive though... they just make it that way so you can't compete in the market using only a 3d printer. Only the ones with money get to make more in a pyramid society.
when i first saw your channel i knew you were a gifted engineer who was gonna bring something meaningful to the space (im sure my earlier comments reflect this assumption). this printer is in a league of it's own. i keep rewinding the video because watching it move is magic. thank you for sharing this with the world.
A closed loop computer vision or LIDAR alignment system is the only real solution I can think of. Maybe you could print a special test print and use a 3D scanner to calibrate it.
Maybe something simpler, All we care about is the distance of the toolhead from the center the machine, you could use a lot of systems to achieve that (a 4 way tensioned rope that is hooked up to a potentiometer is the jankest of solutions but you could make it more fancy to get accuracy)
@@arturdobrzynski6531 well, its not really LIDAR. that's the issue. literally a cheap interferometer. But because they use a camera, lasers and an algorithm, they call it "lidar". Thing is the camera they use is a camera, not a LIDAR sensor. True LIDAR doesn't use cameras. the MicroLIDAR they use reads underpowered laser lines VISUALLY, then makes a guess at how long the line is by LOOKING at it like a human would. Looking at angles, ect. True LIDAR reads reflections off of an object from a high powered laser and the sensors cant see the world like a camera can. LIDAR was also a thing long before camera systems were small and good like they are today and long before algorithms were a thing. Also, true LIDAR systems cost A LOT. Like 10s of thousands for highly accurate systems. Most of those highly accurate systems are only accurate to about 0.5mm. BAMBU is out here claiming 7 micros. which is impossible. The machine itself is no where near 7 microns accurate let alone that 1080p camera the company is using. So does it work? sure but not like real LIDAR would and nowhere near the accuracy the company claims. Thats what makes it a gimmick. Might as well bed level with a real micron-level feedback sensor and a traditional bed mesh.
WOW, this is a whole other level. I recently jumped on the RatRig IDEX wagon and while they have got great control over precision alignment with their visual assisted alignment tool (VAOC) this has so many more degrees of freedom! Exciting to see how far you got this already.
This isn't at the level where it would be plug and play like a Ratrig, but in a few revisions it should be! Just need to fine tune the mechanical build, then overhaul the electronics and firmware.
This project is really cool honestly. It is a very unique rig and an out of the box thinking for solving printing speeds. I have too many questions about your design choices but one of them is.. to get a precise frame and gantry alignment and a stiffer frame, why not make the frame square instead of rectangle? This way, the gantries form a perfect 90° angle X shape and can be connected in the middle.
I’ve been considering making the gantry a solid piece of aluminum plate with all 4 linear rails attached to it. That would make them all more precise relative to eachother, but it also means I the z heights will be coupled together. I went with the current design to keep things mechanically simple, and be easier to program in firmware, but there are many other design alternatives
Relative positional accuracy is much easier than absolute positional accuracy. A normal 3D printer has great relative positional accuracy, but can get away with pretty poor absolute positional accuracy.
@brandonhicks7549 Hmm, I think you've got a slight misconception about CNC machines. Saying "absolute" I think you are referring to a closed-loop system, where there's some coupling of the motion axis to a measuring scale, be it a rotary encoder on the drive shaft, or better, a linear scale on the rail. (most) 3D printers have no feedback circuit. They run simple open-loop motor commands, because unlike metal-cutting CNCs , there's no forces deflecting the toolhead. So you can generally assume that G90 absolute positioning command will be accurate, given the accuracy of your machine's kinematic construction, rigidity, so on...
@@NathanBuildsRobots👈THIS You explain it very well in your video. Trying to align 4 semi-independent coordinate systems and kinematics is a NIGHTMARE . Massive respect to you for undertaking this epic engineering task ! And you really did pull it off! Those are some functional and efficient prints!
@@Metapharsical I was referring to construction more generally not really 3D printing specifically. I guess what I meant was that the frame of a normal 3D printer doesn’t have to be all that square to get reasonably accurate prints. The printer shown has to have a high degree of parallelism and perpendicularity between the axes of the different segments.
@@Metapharsical That's very interesting, I didn't think about how the term is used in CNC machining. In the context of my comment, I am referring to the fact that all 3D printers have error when compared to an ideal, or absolute, coordinate frame. The best way to unite the printers onto the same coordinate frame is either to 1) Synchronize them all to an absolute/ideal coordinate frame, as in, reducing all positioning error to zero or 2) Couple all the printheads together in such a way that all printheads will have the same error, so relative to eachother they will all be off the same amount and still just work in the way that single printheads just work.
This is a great project. There are plenty of people building and refurbing cars with 3D printed parts, this could be a game changer for them. First video I have seen of this channel, will be back!!
Honestly just having a rotating build plate is something I've never seen before. Me coming from a machinist's background, you can make that build plate very stable and very flat like on a vertical lathe. Just have to have the bearings supporting the right places. What you were saying about machine rigidity is absolutely true. No machine is truly square. It's as close as it can be and then you put in the right offsets to counteract its imperfections.
Wow! I knew you were working on this, but I had no idea you designed and built it yourself. I really like the idea of being able to use one quadrant of the bed and treating it like a 4 head tool changer. It would be just like having 4 printheads on a Prusa XL, but you rotate the bed under each head to get the different material or color. If you can just get independent or switching extruder stepper control, you can already accomplish this. To fix a lot of your alignment issues, you can still use bed probing. Provided your rotating bed is stable enough, you can actually probe the surface with each printhead for compensation values. Anyways, super cool! Look forward to more.
The Duet 3 control board allows for parallel processes. With the X5 expansion board you can control about 10 axes. Firmware also allows for parallel printing. I don't think it's still enough for your application though. How it works in practice is that two printheads will both execute their particular task until they receive a command to sync them up. At that point I think the fastest printhead will wait for the slowest printhead to complete its gcode.
Would be cool to explore the complicated topic of using these multiple heads to print _different_ models. Here we see them either making the same model per head or one big symmetrical model divided into 4 quadrants.
You're onto something very special here! Fantastic work.... Side note - Never seen you smile so much. Can tell you're truly locked in with this concept.
That’s impressive work you did. I see one flaw in the polar system if printing non-symmetrical parts. Different heads will have to deal with different tangential speeds. I’m sure it’s negligible in most cases and could also park the suboptimal head and let the other heads complete the last troubling extrusion. I can’t imagine the slicer headache
You’re absolutely right, there will be an inefficiency when the Printheads are at different radii where it will have to spin as slow as the Printhead that is furthest away from the center. With 4 SCARA type arms, the arms could advance or delay their phase so maximum speed would always be achieved
@@NathanBuildsRobots Have you think about a slotted nozzle, let’s say 0.4mm x 1.6mm, that you can dynamically orient. Have seen it in those huge concrete 3D printer. Or event even a nozzle with a row of 4x 0.4mm holes all independently feed by 4 filaments and dynamically orientable. That way, you can still print small features with one hole and then print thick wall by feeding all holes. An Arachne on steroids
Cantilevered gantrys.. scary lol, amazing concept and build, so this thing should be able to print perimeters and infill at the same time once software in scaled
This may be a lot of the rigidity issue. linking the 4 cantilevered gantries into hub to effectively have a single gantry plate - requires a change in operating procedures but is way more rigid. might be tricky to calibrate a 4 z axis print gantry warp to match any build plate plane deformations
Yes, it could do a dual spiral infill pattern while simultaneously printing 2 perimeter walls. Or quads symmetric infill and 4 perimeters simultaneously
This is exactly the approach we need for faster printing. Unless we find some materials science miracle, we can only melt and extrude plastic so fast with one printhead. More printheads is the answer!
Dude, you're a legend. This is actually, seriously (a part of) the future of 3D printing. I really hope others follow in these footsteps, or you can just release it as a product and make millions, then the posers will follow ;) btw I love you 😘💕
Sharks with friggen lasers. The beauty of this arrangement of toolheads is you can basically add unlimited tools around the perimeter. Specialized nut installation tools, milling heads, lasers, ironing heads, whatever you want, there's plenty of room.
@@NathanBuildsRobots That is what i was thinking right away, the potential for secondary operations after the print is over or ever while it is printing is insane.
I’m reminded of The DED (Directed energy deposition) Metal 3D printer in Avatar: The Way Of Water. That thing was just RIPPING metal and the huge structure it was making seemingly grew in real time.
Amazing idea. I love the whole concept. Definitely has some limitations like you mentioned, but I can see a lot of positive development on these designs in the future. Excellent work!
Increased the print speed but also the amount of moving parts and possible errors from clogs on the nozzles to issues with offsets and belts, increased the misprints by 4x as well
Compact and elegant design for such massive capacity! As you said, there's certainly more ridgid kinematic configurations, but you've got firmware limitations. Combining Cartesian and Polar motion 🤩 It's absolutely fantastic what you achieved here!!
@@NathanBuildsRobots sure, there are many different sizes. From the Wikipedia article "Sizes range from 16 to 28 inches (41 to 71 centimetres) in diameter while depths range from 12 to 22 inches (30 to 56 centimetres), with 14 to 18 inches (36 to 46 centimetres) being normal." en.wikipedia.org/wiki/Bass_drum
It would just be a matter of printing in one of the standard sizes and then adding the appropriate hardware. Typical drum shell materials are wood, metal or plexiglass, so 3D printing should work
I gotta say I had my doubts when you first showed us the parts of the machine and said you're gonna build something out of them, but I was indeed surprised how quickly you got it built. Or to be honest, that you even got a working machine out of them :D Well done! What's next? More development on the firmware side? Will we get more videos about the progress? How are the shoes holding up?
Shoes are falling apart but still functional. I'm split between spending more time working more on the firmware or hardware. The firmware change will require new electronics, so either way it will require a partial rebuild.
Very cool. My brain is on fire with ideas about alternative ways to arrange those X gantries and produce a kinematic where all the print heads could reach across the middle, or at least 100% of the platform when it's rotated. As for the frame alignment, you could definitely add a frame warp compensation to the kinematics to account for the irregularities in the frame. The tricky part would just be taking those measurements, so maybe lasers (passing through guide holes) could be involved. Keep up the good work!
In terms of axis control, you're better off with a "bedslinger" in the same style as milling machines. Then you can use a circular bed with a rotational axis to use any number of print heads you like with any number of subdivisions/symmetries
AWESOME MACHINE! Ok I think I've got some ideas you might not have considered for axis alignment. REAL IDEA: 1- put load cells in each of the print heads. 2- Use one print head as a "master" print head, and have it print a large circle. 3- use the load cells in all of the print heads to locate the edge of the circle 4- use this at multiple locations to figure out where each head is in space and adjust the hardware positioning in software accordingly. The main benefit is that it doesn't use any external hardware to register alignment but instead the actual position of one of the other heads to calibrate them all. At the end of the alignment protocol, have each print head print a circle on top of the others to check and see how close they are to each other. Idea other - do the opposite axes need to be independently cantilevered? Or could you effectively make an "X" (not exactly, but with 2 overlapping cross bars). That would help greatly with the parallelism/perpendicularity alignment of the heads across from each other. Also you could have a single print head extend further than the others in order to reach the middle of the bed. COMBO BONUS IDEA: 1 - use "X" idea from idea "other" above, and the load cells to perform regular bl touch style bed leveling, but then use the lead screws to actively "torque" the frame into the correct shape. for the first few layers at least. But that's another software solution. The perpendicularity and parallelism of this machine is crazy important but it should be achievable mechanically with enough thinking about it... And Blanchard ground 1 inch thick pieces of tool steel LMAO. BUY A DUET MAIN BOARD + EXPANSION CARD TO CONTROL THIS MACHINE AND THANK ME LATER
I think I'd start with one IDEX arrangement that crosses the center of rotation, so you can print in the center of the bed, then add more radial axes. For alignment, best option would be a large surface plate to get the frame planar a straight edge to get one rail true to itself, and then a test indicator to let you adjust the other rails parallel to that one.
Great work Nathan! It has been a lot of fun following your project. What about having different size nozzles in the same print? E.g. 1mm for large areas with less precession, but more speed and 0.4 when you need those details and precision. Do you think the extra print heads will add much to the slicing time? I been printing some lattice structures lately and the slicing takes forever. I hope GPU slicing will be a thing soon.
Probably the easiest way to ensure each rail is parallel is mounting a dial indicator to one and measuring the distance to the other as you move the rail up and down, this is very common when making or preparing machine tools. Just need to be careful with how rigid the mounting for the indicator is and what surfaces you are actually measuring. For example doing this will not guarantee squareness to the base, and you'll need to measure the front and side surface of each rail for both axis of the rail that can be out of parallel. Use the same rail to measure the adjacent rails to avoid compounding errors, and adjust the opposite corner rail with the dial mounted to it while doing your best to make it agree with both newly adjusted rails if you don't have a dial stand long enough to reach it from your reference rail.
That’s an amazing project! I think, the bed adhesion could be solved by using some pressure measuring sensors like the prusa mk4. I also think, that the machine is more suitable for pellet printing. Also, once it is possible to control each extruder by itself (so 6 axis), it would make sense to build in a gyro on each printhead to compensate for any misalignment of a particularly axis
It would be interesting to translate this to a linear configuration. Multiple print heads along a single belt or having the printheads arranged in a grid above a larger bed could accomplish something similar with potentially less mobing parts and less concern for the resolution changing towards the center of the rotating bed
I think a belt style 3d printer would be a good candidate for true multi head printing. You could put the heads in a pipeline, each responsible for e.g. a third of the total height by mass
I forsee two major issues: 1.) you can never reach the center of the buildplate. 2.)when you want to print asymmterical shapes, the accelerations of your "X" axis' will be very geometry dependant ans eratic. This will cause a lot of artifacts and oozing defects.
12:00 - The multitude of out of tram conditions in a 3D printer of this design could largely be calibrated in software, once you have all four print heads completely independent of each other, similar to Z axis offset after creating a print surface map. 13:00 - Never mind. 😀
Just a thought but to get coverage in the center, maybe putting a pillar down from above between the gantries. Maybe an upside down T gantry that can rotate and allow a 5th extruded to move around and compensate for the blank spot.
Pretty cool idea and design. Come on now, the 3d printing community is interested in knowing can you change out the massive 1.8mm nozzles to a 0.4 or 0.5mm nozzle and print out a Benchy!! Kidding - very interesting and well done.
To solve problem with many many axes, I think you need take 4 controllers - one to every print head(or more if you need more print heads), and of course will need rebuild Klipper firmware on this case of printing.
I think for your project reprap firmware is better suited. Proper printing has built a dual toolhead machine that has all axis independent, so a separate xyz for each of the extruders. Btw klipper had for a long time support for idex printers (dual carriage), copy/mirror has been added this year in mainline klipper. So no changing pin assignments, you just park one tool, activate the other and continue.
just as an Idea: with independent z axis control it might be possible to do vase mode with 4 layers done in one rotation of the build plate (not exactly 4 layers in one rotation, but it should cut the print time of vase mode prints down to 1/4 of the original time) each print head does one rotation on one level and when the newly printed layers from the other extruders arrive you could move up the extruder whilst printing up 3 layers. If you stagger the end of the ends of the circles of the different extrusions you also don't need to move up 3 layers immediately but a bit more diagonally up. not sure if it makes complete sense but this could be a fun application of this design edit: for g-code generation you could essentially run the same g-code on each extruder, but starting out on different layers and also staggered in phase of the rotation of the buildplate, this could ease up on the software implementation of this.
Very interesting project! I’d look into pellet printing. I work with some very big companies doing large scale high flow pellet and filament using gantry and nonplanar printing. But pellet is definitely the future of 3D printing…
This is a cool printer concept! We're Tesseract Technologies, and you may have seen some of our previous comments here on YT. We're currently developing a high-speed 3D printer using custom-made BLDC linear motors. However, due to budget constraints, we've decided to start by offering an affordable, high-speed linear motor that is plug-and-play compatible with existing control boards using step and direction signals. We're also planning to release a video soon, where we'll explain a unique approach using an H-bot configuration. In this setup, multiple X-axes can share a Y-axis, driven by several linear motors, allowing for overlapping print heads. This system is simpler and more efficient than complex belt-driven mechanisms, enabling extremely high-speed 3D printing with multiple print heads. You can check out an image of this concept on our channel. We're excited to see how your design evolves and to share more about our own journey!
Fascinating, ambitious, innovative. Regarding vertical alignment issues of the four arms. Did you consider moving the print bed vertically rather than all the individual arms? Kind of like Bambu style CoreXY. Another strategy would be to lock the four linear rails together in the center with a strong x connection. Alternately they could all be tied together with a periphery frame. Like how a Voron style CoreXY moves the entire gantry as a single unit tied together.
Super cool looking forward to seeing more developments. Is there a reason for the wall thickness being so high? I'm assuming 1-2 walls doesnt work well due to alignment issues tku mentioned.
I feel like the squareness of the frame could be fixed/helped, if diagonal steel cables could be put on the sides, back, top and bottom, and tensioning them in order to callibrate the whole structure I'm recommending steel cables, or threaded metal rods so that it can more easily be callibrated, as opposed to having to presicely machine the frame/diagonal pieces, but then again, another issue that could arrive with that is the frame bowing out in the middle and to allow for the printer to print in the middlle, you might consider allowing 1 of the print heads to go a bit past the middle so that that spot can be reached, however that would make the print heads be asymetrical, which would require them to have individual control (unless if that is actually already possible, just not technically due to the print heads having the same g-code)
I guess the "easiest" way to fix almost all misalignement issues would be to have a constant calibration routine with something similar to vive trackers (I wonder if just using stock trackers would give a good result), you have basically a barcode scanner (sweeping line laser) and multiple light sensors that have a very high speed clock, and you time a flash then a sweep in 2 axis, since there would be a single reference sweep, the relative position of each arm should be fairly precise. you can do a few sweeps on each layer with static positions (e.g. min middle max) if you aren't doing fully 3d printing. that would require painting allmost all of the printer matte black, because reflections are an issue, which might be a pain edit: there might be a cheaper way actually, if you use a diffraction grating you might be able to get high precision positional accuracy from an off the shelf camera by using interference patterns just by looking at the pattern, and since the position is very limited that may require semi real time slicing (e.g. only slice a few layers in advance, since you can assume calibration won't drastically change in a few layers) to compensate the sideways misalignement of the extruders as each extruder only has 2 axes. though a small pair of driven offset screws on the x axis rail might be enough to compensate drift and keep the same gcode while being easier to implement. imo the main issue with the system is that at best you get a 10, maybe 20x improvement for a much higher cost and complexity (when you could just run 10 printers unless you need very fast turnover). (it still is worth pursuing as improvements to idex would be great) imo for very fast printing if cost isn't an issue high power bottom down dlp resin has a much higher ceiling (at the cost of needing a large base volume of resin.) also if we see affordable larger models of resin printers with a pump like the anycubic m7s (but with a 300 or even 400mm bed), that might be a great solution for moderately high speed printing (though there's always the issue of film peeling which doesn't exist in bottom down printing.) there might be a way to use a light resin floating on top of water or glycerin, that way you don't need high volumes of resin to print large bounding box bottom down models. waves might be an issue but probably not since at very high speed you're printing at a constant flow so there shouldn't be that much turbulence.
Nathan...wow! What is so cool; for me, is that nearly everything I print is circular and large. This is my kind of deal. If you wanted to share, I would def do a build...built tons of deltas of the last nearly 10 years but this would totally be brilliant. I actually have some pretty big deltas that I could convert into a 3 gantry for sure BUT... Overhangs???
Do you think that you could do like a 3D mesh level to compensate for the warping ? I'm not sure exactly how it would work but maybe have 4 bed scanners and have them all scan every 0.5 or 1mm Z height and maybe somehow overlap the meshes to be able to figure out where the tool heads are relative to each other? Maybe have markers in the bed to help with aligning the meshes? And then use interpolation to get even finer adjustment for warping. It sounds like it could work but idk if it's feasible to make. Best of luck
It seems like the maximum speed improvement should be 4x with 4 tool heads, and that's only if printing a completely optimized symmetrical part that can keep all 4 heads fully active all the time. Any speed benefit beyond 4x is tuning or hot end improvements that could also be migrated to another standard platform without multiple heads. There are probably edge cases where you need something fast, but in most commercial applications getting 4 in 4x the time in parallel is equivalent to printing them serially on a fast custom setup. In its favor parallel printing doesn't suffer inefficiency when printing unusual things (no symmetry requirements), and aligns the start/stop times to limit the number of times you need to interact with the machines. Pretty cool you got it to work and it's an impressive technical achievement, I just question the practicality.
Truly amazing work! I think it's an awesome idea. Without giving me a heart attack, can I get a rough estimate on your build cost? Would you be able to make a four-way in the center between all four of the extruder bars to hold them perfectly aligned?
regarding the allignment issue: laser and receiver on each printhead, measure their distance from each other and inject offsets into print live. will need a specialised software though.
Brilliant project. What a mission. love it. Would a solid hub linking the ends of the extruder arms make their relative positions more rigid? And then if an extruder is idle make it sit in a rest position outside of the build area so as not to crash into the build. A solid print head gantry with the multiple print arms might help keep print plane alignment neat between the 4 heads as the gantry rises. Probably also needs well coordinated for 4 z axis rise timing and each print head completes it's layer at the same time so flow rates mapped to layer specific speeds of extrusion. A lot of code for the slicer
That is an interesting thought. Originally I was thinking of spinning the whole top half of the machine, but this makes a lot more sense. Or, moving the bed down with stationary gantry up top, that would also do the trick
Very impressive machine! To make the alignment somewhat easier, but still hard: Reverse the machine and use a fixed gantry assembly with all the heads. Let the bed drop down. The bed only needs 3 axes to control precisely, which is much less than however many extruders will be used. To get the vertical alignment done, use a laser and a mirror and an image sensor to see the drift of the spot as the bed is lowered. Or something like an auto collimator. You should be able to get very precise angular twist detection at one particular point. To get toolhead alignment done, use a bed probe that can be rotated around. Might require the probe to rotate precisely about the origin. There may be better solutions out there :/
I like it! But maybe you can extend 1 arm to cover the center? Or 2 arms, offset, extending past the center. Would make for pretty complicated slicing though!
Another great video and hope you continue to develop this project. I completely agree with your assertion that parallel multi head printing is the way forward. My thoughts on some of the issues you raise in this video are: 1) Limitations on the number of axis you can control - I think keeping the control boards separate on each print head is a solution to this - i.e. Each board has 3 AXIS and Extruder control - for its own print head, You just then need to ensure that the control boards synchronise with each other at appropriate points in the print - such as layer changes or exclusion area changes. I think you could then break in print up into a number of 'exclusion zones' per layer - probably the number of print heads x 2 + 1 - so that they can all be printed independently without risk of overlap/head crashes. 2) Alignment - I think software is the answer - I think each print head needs to automatically align itself in 3 dimensions to the same 2 or 3 fixed points in 3d space - then it should be possible for all print heads to then be fairly perfectly aligned with each other. 3) Independence - I still think only having one (IDEX) or two (Your polar machine) axis fully independent is going to limit the shapes you can produce - even producing an oval is going to be difficult as some heads are going to need to be moving at non optimal speeds to allow the other ones to keep up with the different distances they need to travel on the fixed motion of the bed, Therefore I would suggest you keep the polar design - but add another axis of movement at right angles to the in out movement. My 'Revolutionary concept' short purposely included a 'house' or 'Triangle on top of Square shape' to demonstrate the sort of shapes that will only be possible with fully independent movement in the XY axis. I think that the new right angle axis could then be used for a lot of the motion - with the polar movement then only being used to move the print around maybe twice to per layer - to give each head access to a new 'exclusion zone'. 4) Multi colour - I still think having some kind of colour multiplexing on some of the heads is going to be desirable - both the allow larger number of colours than print heads, plus to allow parallel printing of some colours in certain parts of the print. 5). Pausing - it was a shame that you had to cancel your first big print due to time constraints. Personally for really long prints I almost always pause them overnight - I use a non heated build plate and bring each print head down to 35 degrees so that it doesn't need any fans and then just pause the print overnight or when I go out. I usually only pause during a colour change. I have even managed a couple of times to completely power off the printer and 'power interruption restart it' and still get a successful print. Will probably produce some sort of demo myself - probably of an IDEX with a couple of extra independent cantilever heads added in front, plus maybe the bed changed to Polar - but am currently waiting to see what BambuLab come out with in the next few weeks if the rumours are correct - would be very interested to see one of your 'prediction' videos on this as you have been pretty spot on in the past.
Amazing machine sir, excellent work! To me, a pellet fed design feels like the right direction given your comments regarding filament swaps. Looking at the comments regarding the ability to print the entire platform, missing the middle comments, given all heads are printing the same layer, is it possible to move the relative motion to the heads themselves mounted on continuous X cross rails compensating relative head heights with some sort of mount design? Just blue-sky thoughts here about an amazing machine. Please, want to see more.
Insane, thanks man. Cool to see some innovation outside of oem. You should stop printing trash cans and start printing side tables. Also look into rhino 3d grasshopper and the opportunity to make custom gcode for your printers if you aren't using something like this already. Anyway cheers, cool stuff
You can try To make it print upside down and with a small printing Tower for centre to be not hollow. Even though most things that we use are Hollow centre
The “does he even build robots” people are real quiet right now.
Nathan's playing the long game... building machines that will eventually be capable of building robots
yeah. definitely love this 'sleeper' redemption to those peoples
it's not a robot. but that doesn't make it any less technically impressive, this is fucking awesome -- a roboticist.
i love whats been happening here! amazing... while many other channels are clueless about whats next..
@@digitaltoaster that is literally a robot. 3D printers are robots.
i think a big concern is filament cost. The pellets used in injection molding are like 6€ a kg, but one spool of filament cost like 20€+
Yes, material cost is a concern.
This is a sort of roto-moulding alternative that is tool-less/on-demand, so it's not going to be the first choice for low-cost items
20€? More like 10-12€. I pay £7 per 1kg of Sunlu (petg, pla)
Okay PLA/PETG is more like 10 euro if you don't just grab a single off some shop.
I thought the pellets were cheaper too?
filament doesn't need to be expensive though... they just make it that way so you can't compete in the market using only a 3d printer.
Only the ones with money get to make more in a pyramid society.
@@arturdobrzynski6531 Where do you buy filament from?
Respect for the build, I realized that 3d printing isn't cost effective after you said 3 kg for a trash can 😅😂
3D printing is still ideal for 1-off or custom projects.
But maybe I could put 4 pellet extruders on there
@@NathanBuildsRobots Yeah you're absolutely right about 1-off or custom projects. Keep the good work 😊
Jup, that’s the way to go.
@@alengusic4259 that`s because the inaccuracy of the print requires thick walls, otherwise it could be cheaper
@@alengusic4259 it's a proof of concept. The real trashcan is the giant replica of Kelsey Grammer's head. There is where the value lies.
Nathan Builds Robots is the definitive source for quadrupolar high speed printing
Whoa, this is awesome! Nice build!
Thank you Hacksmith! 🫶
No way hacksmith commented 😂
when i first saw your channel i knew you were a gifted engineer who was gonna bring something meaningful to the space (im sure my earlier comments reflect this assumption). this printer is in a league of it's own. i keep rewinding the video because watching it move is magic. thank you for sharing this with the world.
Thanks Mr Baskins, my biggest fan :)
A closed loop computer vision or LIDAR alignment system is the only real solution I can think of. Maybe you could print a special test print and use a 3D scanner to calibrate it.
That would be interesting, like the Lidar calibration 3D Bambu Lab, Creality, and FLSun keeps promising, but something that actually works!
@@NathanBuildsRobots do you imply that lidar on x1s does not work?
Maybe something simpler, All we care about is the distance of the toolhead from the center the machine, you could use a lot of systems to achieve that (a 4 way tensioned rope that is hooked up to a potentiometer is the jankest of solutions but you could make it more fancy to get accuracy)
@@NathanBuildsRobots probably something more like TAMV type of setup setup. cheers, XGZ
@@arturdobrzynski6531 well, its not really LIDAR. that's the issue. literally a cheap interferometer.
But because they use a camera, lasers and an algorithm, they call it "lidar". Thing is the camera they use is a camera, not a LIDAR sensor. True LIDAR doesn't use cameras. the MicroLIDAR they use reads underpowered laser lines VISUALLY, then makes a guess at how long the line is by LOOKING at it like a human would. Looking at angles, ect.
True LIDAR reads reflections off of an object from a high powered laser and the sensors cant see the world like a camera can. LIDAR was also a thing long before camera systems were small and good like they are today and long before algorithms were a thing.
Also, true LIDAR systems cost A LOT. Like 10s of thousands for highly accurate systems. Most of those highly accurate systems are only accurate to about 0.5mm. BAMBU is out here claiming 7 micros. which is impossible. The machine itself is no where near 7 microns accurate let alone that 1080p camera the company is using.
So does it work? sure but not like real LIDAR would and nowhere near the accuracy the company claims. Thats what makes it a gimmick. Might as well bed level with a real micron-level feedback sensor and a traditional bed mesh.
really putting "builds robots" into the name
WOW, this is a whole other level. I recently jumped on the RatRig IDEX wagon and while they have got great control over precision alignment with their visual assisted alignment tool (VAOC) this has so many more degrees of freedom! Exciting to see how far you got this already.
This isn't at the level where it would be plug and play like a Ratrig, but in a few revisions it should be! Just need to fine tune the mechanical build, then overhaul the electronics and firmware.
This project is really cool honestly. It is a very unique rig and an out of the box thinking for solving printing speeds. I have too many questions about your design choices but one of them is.. to get a precise frame and gantry alignment and a stiffer frame, why not make the frame square instead of rectangle? This way, the gantries form a perfect 90° angle X shape and can be connected in the middle.
I’ve been considering making the gantry a solid piece of aluminum plate with all 4 linear rails attached to it. That would make them all more precise relative to eachother, but it also means I the z heights will be coupled together.
I went with the current design to keep things mechanically simple, and be easier to program in firmware, but there are many other design alternatives
@@NathanBuildsRobots Oh - I just suggested this. you're waaay ahead. sooo awesome
Genius build, man! I hope you have grants/investors for your amazing ideas.
Relative positional accuracy is much easier than absolute positional accuracy. A normal 3D printer has great relative positional accuracy, but can get away with pretty poor absolute positional accuracy.
Exactly.
The challenge with multiple printing is uniting the relative coordinate frames to +- 0.01mm.
@brandonhicks7549
Hmm, I think you've got a slight misconception about CNC machines.
Saying "absolute" I think you are referring to a closed-loop system, where there's some coupling of the motion axis to a measuring scale, be it a rotary encoder on the drive shaft, or better, a linear scale on the rail.
(most) 3D printers have no feedback circuit. They run simple open-loop motor commands, because unlike metal-cutting CNCs , there's no forces deflecting the toolhead.
So you can generally assume that G90 absolute positioning command will be accurate, given the accuracy of your machine's kinematic construction, rigidity, so on...
@@NathanBuildsRobots👈THIS
You explain it very well in your video. Trying to align 4 semi-independent coordinate systems and kinematics is a NIGHTMARE .
Massive respect to you for undertaking this epic engineering task ! And you really did pull it off! Those are some functional and efficient prints!
@@Metapharsical I was referring to construction more generally not really 3D printing specifically.
I guess what I meant was that the frame of a normal 3D printer doesn’t have to be all that square to get reasonably accurate prints. The printer shown has to have a high degree of parallelism and perpendicularity between the axes of the different segments.
@@Metapharsical That's very interesting, I didn't think about how the term is used in CNC machining.
In the context of my comment, I am referring to the fact that all 3D printers have error when compared to an ideal, or absolute, coordinate frame.
The best way to unite the printers onto the same coordinate frame is either to
1) Synchronize them all to an absolute/ideal coordinate frame, as in, reducing all positioning error to zero
or
2) Couple all the printheads together in such a way that all printheads will have the same error, so relative to eachother they will all be off the same amount and still just work in the way that single printheads just work.
This is a great project. There are plenty of people building and refurbing cars with 3D printed parts, this could be a game changer for them. First video I have seen of this channel, will be back!!
Man I'm in awe of your design, not only being able to build it but to get the motion working/coded would be mind boggling for me.
Honestly just having a rotating build plate is something I've never seen before. Me coming from a machinist's background, you can make that build plate very stable and very flat like on a vertical lathe. Just have to have the bearings supporting the right places. What you were saying about machine rigidity is absolutely true. No machine is truly square. It's as close as it can be and then you put in the right offsets to counteract its imperfections.
it was cool to see and feel these prints IRL at open sauce, it's just so foreign compared to any other FDM prints i'd seen before
Wow! I knew you were working on this, but I had no idea you designed and built it yourself. I really like the idea of being able to use one quadrant of the bed and treating it like a 4 head tool changer. It would be just like having 4 printheads on a Prusa XL, but you rotate the bed under each head to get the different material or color. If you can just get independent or switching extruder stepper control, you can already accomplish this. To fix a lot of your alignment issues, you can still use bed probing. Provided your rotating bed is stable enough, you can actually probe the surface with each printhead for compensation values. Anyways, super cool! Look forward to more.
I haven’t watched your channel in a while, I’m glad to see you are doing more than just printer reviews, this was good!
I like the polar axis and I knew that was coming back at some point badass
The Duet 3 control board allows for parallel processes. With the X5 expansion board you can control about 10 axes. Firmware also allows for parallel printing. I don't think it's still enough for your application though. How it works in practice is that two printheads will both execute their particular task until they receive a command to sync them up. At that point I think the fastest printhead will wait for the slowest printhead to complete its gcode.
Hmm, I have a half dozen boxes that say "Duet" on them that agree with you
I suggest an epoxy granite frame. Not because it's a practical solution, but because it would be entertaining to watch.
yeah this printer deserves a solid base
Mmm Blanchard ground epoxy granite might be a cheaper ultraflat base and top for this
14:30 'quite easy' lol. Chapeau!
Would be cool to explore the complicated topic of using these multiple heads to print _different_ models. Here we see them either making the same model per head or one big symmetrical model divided into 4 quadrants.
This could be game changing for prosthetics companies. Seems like a perfect machine for it.
You're onto something very special here! Fantastic work.... Side note - Never seen you smile so much. Can tell you're truly locked in with this concept.
3D printer reviews are sooo boring 😑
@@NathanBuildsRobots Least you ain't bitching about money all the time like 'some' reviewers do but actually being proactive 👍
That’s impressive work you did.
I see one flaw in the polar system if printing non-symmetrical parts.
Different heads will have to deal with different tangential speeds.
I’m sure it’s negligible in most cases and could also park the suboptimal head and let the other heads complete the last troubling extrusion.
I can’t imagine the slicer headache
You’re absolutely right, there will be an inefficiency when the Printheads are at different radii where it will have to spin as slow as the Printhead that is furthest away from the center.
With 4 SCARA type arms, the arms could advance or delay their phase so maximum speed would always be achieved
@@NathanBuildsRobots Have you think about a slotted nozzle, let’s say 0.4mm x 1.6mm, that you can dynamically orient.
Have seen it in those huge concrete 3D printer.
Or event even a nozzle with a row of 4x 0.4mm holes all independently feed by 4 filaments and dynamically orientable.
That way, you can still print small features with one hole and then print thick wall by feeding all holes.
An Arachne on steroids
Cantilevered gantrys.. scary lol, amazing concept and build, so this thing should be able to print perimeters and infill at the same time once software in scaled
This may be a lot of the rigidity issue. linking the 4 cantilevered gantries into hub to effectively have a single gantry plate - requires a change in operating procedures but is way more rigid. might be tricky to calibrate a 4 z axis print gantry warp to match any build plate plane deformations
Yes, it could do a dual spiral infill pattern while simultaneously printing 2 perimeter walls.
Or quads symmetric infill and 4 perimeters simultaneously
Incredible work man, stuff like this & ppl like you will make this whole home/ office production thing actually viable for everyday people.
This is exactly the approach we need for faster printing. Unless we find some materials science miracle, we can only melt and extrude plastic so fast with one printhead. More printheads is the answer!
Dude, you're a legend. This is actually, seriously (a part of) the future of 3D printing. I really hope others follow in these footsteps, or you can just release it as a product and make millions, then the posers will follow ;) btw I love you 😘💕
This looks like true innovation. Industrial 3D printing could truly benefit from this technology.
I'm glad someone is finally doing this!!
Incredible. This is the type of printer i would add lasers to the side to cut out patterns in the material. great job !!
Sharks with friggen lasers.
The beauty of this arrangement of toolheads is you can basically add unlimited tools around the perimeter. Specialized nut installation tools, milling heads, lasers, ironing heads, whatever you want, there's plenty of room.
@@NathanBuildsRobots That is what i was thinking right away, the potential for secondary operations after the print is over or ever while it is printing is insane.
Earned a subscribe! My big question. How are you able to maintain motivation when the project gets overwhelming as they often do?
Very impressed.
I’m reminded of The DED (Directed energy deposition) Metal 3D printer in Avatar: The Way Of Water. That thing was just RIPPING metal and the huge structure it was making seemingly grew in real time.
I imagine having independent heads would create a lot of complexity to avoid heads crashing together. Great work anyway 🎉
That's the great thing about this arrangement, almost 0 chance of toolhead collisions, with the exception of the very center of the bed.
Amazing idea. I love the whole concept. Definitely has some limitations like you mentioned, but I can see a lot of positive development on these designs in the future. Excellent work!
This is one of those really cool builds that 99% of people would have no use for, I like it :D
Increased the print speed but also the amount of moving parts and possible errors from clogs on the nozzles to issues with offsets and belts, increased the misprints by 4x as well
Super cool! This definitely seems like the next big thing once its refined for large scale printing.
Man my brain is fuc*!$ haha😅 when i scrolled the thumbnail seemed to me like an advanced Minecraft mob farm. Im ruined 😂😂
Compact and elegant design for such massive capacity!
As you said, there's certainly more ridgid kinematic configurations, but you've got firmware limitations.
Combining Cartesian and Polar motion 🤩 It's absolutely fantastic what you achieved here!!
You have to build some drum shells with that! With that size you could even make a bass drum!
Would 600mm x 800mm be large enough
@@NathanBuildsRobots sure, there are many different sizes. From the Wikipedia article "Sizes range from 16 to 28 inches (41 to 71 centimetres) in diameter while depths range from 12 to 22 inches (30 to 56 centimetres), with 14 to 18 inches (36 to 46 centimetres) being normal." en.wikipedia.org/wiki/Bass_drum
It would just be a matter of printing in one of the standard sizes and then adding the appropriate hardware. Typical drum shell materials are wood, metal or plexiglass, so 3D printing should work
Insane that you were able to code all that up to make it work.
I gotta say I had my doubts when you first showed us the parts of the machine and said you're gonna build something out of them, but I was indeed surprised how quickly you got it built. Or to be honest, that you even got a working machine out of them :D Well done!
What's next? More development on the firmware side? Will we get more videos about the progress? How are the shoes holding up?
Shoes are falling apart but still functional.
I'm split between spending more time working more on the firmware or hardware.
The firmware change will require new electronics, so either way it will require a partial rebuild.
This looks amazing, wish you the best of luck!
Very cool. My brain is on fire with ideas about alternative ways to arrange those X gantries and produce a kinematic where all the print heads could reach across the middle, or at least 100% of the platform when it's rotated. As for the frame alignment, you could definitely add a frame warp compensation to the kinematics to account for the irregularities in the frame. The tricky part would just be taking those measurements, so maybe lasers (passing through guide holes) could be involved. Keep up the good work!
That thumbnail looks like atmospheric reentry
In terms of axis control, you're better off with a "bedslinger" in the same style as milling machines. Then you can use a circular bed with a rotational axis to use any number of print heads you like with any number of subdivisions/symmetries
Good point, I think XY position control on the bed, and the quad gantries on a solid plate is the sturdiest configuration with the least complexity
AWESOME MACHINE!
Ok I think I've got some ideas you might not have considered for axis alignment.
REAL IDEA:
1- put load cells in each of the print heads.
2- Use one print head as a "master" print head, and have it print a large circle.
3- use the load cells in all of the print heads to locate the edge of the circle
4- use this at multiple locations to figure out where each head is in space and adjust the hardware positioning in software accordingly.
The main benefit is that it doesn't use any external hardware to register alignment but instead the actual position of one of the other heads to calibrate them all.
At the end of the alignment protocol, have each print head print a circle on top of the others to check and see how close they are to each other.
Idea other - do the opposite axes need to be independently cantilevered? Or could you effectively make an "X" (not exactly, but with 2 overlapping cross bars). That would help greatly with the parallelism/perpendicularity alignment of the heads across from each other. Also you could have a single print head extend further than the others in order to reach the middle of the bed.
COMBO BONUS IDEA:
1 - use "X" idea from idea "other" above, and the load cells to perform regular bl touch style bed leveling, but then use the lead screws to actively "torque" the frame into the correct shape. for the first few layers at least. But that's another software solution.
The perpendicularity and parallelism of this machine is crazy important but it should be achievable mechanically with enough thinking about it... And Blanchard ground 1 inch thick pieces of tool steel LMAO.
BUY A DUET MAIN BOARD + EXPANSION CARD TO CONTROL THIS MACHINE AND THANK ME LATER
This is proper impressive. Well done.
Your Channel is about to blow up! Great information!
I think I'd start with one IDEX arrangement that crosses the center of rotation, so you can print in the center of the bed, then add more radial axes. For alignment, best option would be a large surface plate to get the frame planar a straight edge to get one rail true to itself, and then a test indicator to let you adjust the other rails parallel to that one.
First time watching your channel, and already amazed with you did.
Really impressive!
Great work Nathan! It has been a lot of fun following your project. What about having different size nozzles in the same print? E.g. 1mm for large areas with less precession, but more speed and 0.4 when you need those details and precision. Do you think the extra print heads will add much to the slicing time? I been printing some lattice structures lately and the slicing takes forever. I hope GPU slicing will be a thing soon.
I have some simple use cases that I can develop first, working on that now
Probably the easiest way to ensure each rail is parallel is mounting a dial indicator to one and measuring the distance to the other as you move the rail up and down, this is very common when making or preparing machine tools. Just need to be careful with how rigid the mounting for the indicator is and what surfaces you are actually measuring. For example doing this will not guarantee squareness to the base, and you'll need to measure the front and side surface of each rail for both axis of the rail that can be out of parallel. Use the same rail to measure the adjacent rails to avoid compounding errors, and adjust the opposite corner rail with the dial mounted to it while doing your best to make it agree with both newly adjusted rails if you don't have a dial stand long enough to reach it from your reference rail.
I’m pretty sure something like this could work on duet firmware and boards. They build much more sophisticated motion systems.
Duet/Reprap firmware is critically underrated!
That’s an amazing project! I think, the bed adhesion could be solved by using some pressure measuring sensors like the prusa mk4.
I also think, that the machine is more suitable for pellet printing.
Also, once it is possible to control each extruder by itself (so 6 axis), it would make sense to build in a gyro on each printhead to compensate for any misalignment of a particularly axis
It would be interesting to translate this to a linear configuration. Multiple print heads along a single belt or having the printheads arranged in a grid above a larger bed could accomplish something similar with potentially less mobing parts and less concern for the resolution changing towards the center of the rotating bed
I think a belt style 3d printer would be a good candidate for true multi head printing. You could put the heads in a pipeline, each responsible for e.g. a third of the total height by mass
I forsee two major issues:
1.) you can never reach the center of the buildplate.
2.)when you want to print asymmterical shapes, the accelerations of your "X" axis' will be very geometry dependant ans eratic. This will cause a lot of artifacts and oozing defects.
With the current mechanical design, it can reach center with all 4 printheads, just not simultaneously.
12:00 - The multitude of out of tram conditions in a 3D printer of this design could largely be calibrated in software, once you have all four print heads completely independent of each other, similar to Z axis offset after creating a print surface map.
13:00 - Never mind. 😀
Just a thought but to get coverage in the center, maybe putting a pillar down from above between the gantries. Maybe an upside down T gantry that can rotate and allow a 5th extruded to move around and compensate for the blank spot.
I think the heads are designed at an offer to allow them to go beyond the end of the arms and therefore be able to reach the centre (one at a time).
Pretty cool idea and design. Come on now, the 3d printing community is interested in knowing can you change out the massive 1.8mm nozzles to a 0.4 or 0.5mm nozzle and print out a Benchy!! Kidding - very interesting and well done.
To solve problem with many many axes, I think you need take 4 controllers - one to every print head(or more if you need more print heads), and of course will need rebuild Klipper firmware on this case of printing.
You just gained my admiration mate. No kidding. Hatts off.
I think for your project reprap firmware is better suited. Proper printing has built a dual toolhead machine that has all axis independent, so a separate xyz for each of the extruders.
Btw klipper had for a long time support for idex printers (dual carriage), copy/mirror has been added this year in mainline klipper. So no changing pin assignments, you just park one tool, activate the other and continue.
just as an Idea: with independent z axis control it might be possible to do vase mode with 4 layers done in one rotation of the build plate (not exactly 4 layers in one rotation, but it should cut the print time of vase mode prints down to 1/4 of the original time) each print head does one rotation on one level and when the newly printed layers from the other extruders arrive you could move up the extruder whilst printing up 3 layers. If you stagger the end of the ends of the circles of the different extrusions you also don't need to move up 3 layers immediately but a bit more diagonally up. not sure if it makes complete sense but this could be a fun application of this design
edit: for g-code generation you could essentially run the same g-code on each extruder, but starting out on different layers and also staggered in phase of the rotation of the buildplate, this could ease up on the software implementation of this.
Mechanics do wheel alignment using string before lasers, might be worth a look at. Probably does help all the issues
Very interesting project! I’d look into pellet printing. I work with some very big companies doing large scale high flow pellet and filament using gantry and nonplanar printing. But pellet is definitely the future of 3D printing…
Lovely machine! Hope you figure out some sort of software for it!
This is a cool printer concept! We're Tesseract Technologies, and you may have seen some of our previous comments here on YT. We're currently developing a high-speed 3D printer using custom-made BLDC linear motors. However, due to budget constraints, we've decided to start by offering an affordable, high-speed linear motor that is plug-and-play compatible with existing control boards using step and direction signals.
We're also planning to release a video soon, where we'll explain a unique approach using an H-bot configuration. In this setup, multiple X-axes can share a Y-axis, driven by several linear motors, allowing for overlapping print heads. This system is simpler and more efficient than complex belt-driven mechanisms, enabling extremely high-speed 3D printing with multiple print heads.
You can check out an image of this concept on our channel. We're excited to see how your design evolves and to share more about our own journey!
Thats good.
I wish someone would combine all innovations into one machine.
Fascinating, ambitious, innovative. Regarding vertical alignment issues of the four arms. Did you consider moving the print bed vertically rather than all the individual arms? Kind of like Bambu style CoreXY. Another strategy would be to lock the four linear rails together in the center with a strong x connection. Alternately they could all be tied together with a periphery frame. Like how a Voron style CoreXY moves the entire gantry as a single unit tied together.
vase mode is gonna be crazy with this kind of printer
Good luck and keep us posted on your progress
Super cool looking forward to seeing more developments.
Is there a reason for the wall thickness being so high?
I'm assuming 1-2 walls doesnt work well due to alignment issues tku mentioned.
I feel like the squareness of the frame could be fixed/helped, if diagonal steel cables could be put on the sides, back, top and bottom, and tensioning them in order to callibrate the whole structure
I'm recommending steel cables, or threaded metal rods so that it can more easily be callibrated, as opposed to having to presicely machine the frame/diagonal pieces, but then again, another issue that could arrive with that is the frame bowing out in the middle
and to allow for the printer to print in the middlle, you might consider allowing 1 of the print heads to go a bit past the middle so that that spot can be reached, however that would make the print heads be asymetrical, which would require them to have individual control (unless if that is actually already possible, just not technically due to the print heads having the same g-code)
great printer.
impressive that the first print worked
I guess the "easiest" way to fix almost all misalignement issues would be to have a constant calibration routine with something similar to vive trackers (I wonder if just using stock trackers would give a good result), you have basically a barcode scanner (sweeping line laser) and multiple light sensors that have a very high speed clock, and you time a flash then a sweep in 2 axis, since there would be a single reference sweep, the relative position of each arm should be fairly precise.
you can do a few sweeps on each layer with static positions (e.g. min middle max) if you aren't doing fully 3d printing.
that would require painting allmost all of the printer matte black, because reflections are an issue, which might be a pain
edit: there might be a cheaper way actually, if you use a diffraction grating you might be able to get high precision positional accuracy from an off the shelf camera by using interference patterns just by looking at the pattern, and since the position is very limited
that may require semi real time slicing (e.g. only slice a few layers in advance, since you can assume calibration won't drastically change in a few layers) to compensate the sideways misalignement of the extruders as each extruder only has 2 axes.
though a small pair of driven offset screws on the x axis rail might be enough to compensate drift and keep the same gcode while being easier to implement.
imo the main issue with the system is that at best you get a 10, maybe 20x improvement for a much higher cost and complexity (when you could just run 10 printers unless you need very fast turnover).
(it still is worth pursuing as improvements to idex would be great)
imo for very fast printing if cost isn't an issue high power bottom down dlp resin has a much higher ceiling (at the cost of needing a large base volume of resin.)
also if we see affordable larger models of resin printers with a pump like the anycubic m7s (but with a 300 or even 400mm bed), that might be a great solution for moderately high speed printing (though there's always the issue of film peeling which doesn't exist in bottom down printing.)
there might be a way to use a light resin floating on top of water or glycerin, that way you don't need high volumes of resin to print large bounding box bottom down models.
waves might be an issue but probably not since at very high speed you're printing at a constant flow so there shouldn't be that much turbulence.
This is an awesome build. Nice work. 😍
Nathan...wow! What is so cool; for me, is that nearly everything I print is circular and large. This is my kind of deal. If you wanted to share, I would def do a build...built tons of deltas of the last nearly 10 years but this would totally be brilliant.
I actually have some pretty big deltas that I could convert into a 3 gantry for sure
BUT...
Overhangs???
Do you think that you could do like a 3D mesh level to compensate for the warping ? I'm not sure exactly how it would work but maybe have 4 bed scanners and have them all scan every 0.5 or 1mm Z height and maybe somehow overlap the meshes to be able to figure out where the tool heads are relative to each other? Maybe have markers in the bed to help with aligning the meshes? And then use interpolation to get even finer adjustment for warping. It sounds like it could work but idk if it's feasible to make. Best of luck
It seems like the maximum speed improvement should be 4x with 4 tool heads, and that's only if printing a completely optimized symmetrical part that can keep all 4 heads fully active all the time. Any speed benefit beyond 4x is tuning or hot end improvements that could also be migrated to another standard platform without multiple heads.
There are probably edge cases where you need something fast, but in most commercial applications getting 4 in 4x the time in parallel is equivalent to printing them serially on a fast custom setup. In its favor parallel printing doesn't suffer inefficiency when printing unusual things (no symmetry requirements), and aligns the start/stop times to limit the number of times you need to interact with the machines.
Pretty cool you got it to work and it's an impressive technical achievement, I just question the practicality.
Truly amazing work! I think it's an awesome idea. Without giving me a heart attack, can I get a rough estimate on your build cost? Would you be able to make a four-way in the center between all four of the extruder bars to hold them perfectly aligned?
regarding the allignment issue:
laser and receiver on each printhead, measure their distance from each other and inject offsets into print live. will need a specialised software though.
Brilliant project. What a mission. love it. Would a solid hub linking the ends of the extruder arms make their relative positions more rigid? And then if an extruder is idle make it sit in a rest position outside of the build area so as not to crash into the build.
A solid print head gantry with the multiple print arms might help keep print plane alignment neat between the 4 heads as the gantry rises. Probably also needs well coordinated for 4 z axis rise timing and each print head completes it's layer at the same time so flow rates mapped to layer specific speeds of extrusion. A lot of code for the slicer
You could try putting the z axis in the middle, that removes 3 sources of variability, now you just need to align 1 rod, rather than 4.
That is an interesting thought. Originally I was thinking of spinning the whole top half of the machine, but this makes a lot more sense.
Or, moving the bed down with stationary gantry up top, that would also do the trick
great job nathan, you just made me miss 3D printing
Very impressive machine!
To make the alignment somewhat easier, but still hard:
Reverse the machine and use a fixed gantry assembly with all the heads. Let the bed drop down.
The bed only needs 3 axes to control precisely, which is much less than however many extruders will be used.
To get the vertical alignment done, use a laser and a mirror and an image sensor to see the drift of the spot as the bed is lowered. Or something like an auto collimator.
You should be able to get very precise angular twist detection at one particular point.
To get toolhead alignment done, use a bed probe that can be rotated around. Might require the probe to rotate precisely about the origin. There may be better solutions out there :/
I like it! But maybe you can extend 1 arm to cover the center? Or 2 arms, offset, extending past the center. Would make for pretty complicated slicing though!
Another great video and hope you continue to develop this project.
I completely agree with your assertion that parallel multi head printing is the way forward.
My thoughts on some of the issues you raise in this video are:
1) Limitations on the number of axis you can control - I think keeping the control boards separate on each print head is a solution to this - i.e. Each board has 3 AXIS and Extruder control - for its own print head, You just then need to ensure that the control boards synchronise with each other at appropriate points in the print - such as layer changes or exclusion area changes. I think you could then break in print up into a number of 'exclusion zones' per layer - probably the number of print heads x 2 + 1 - so that they can all be printed independently without risk of overlap/head crashes.
2) Alignment - I think software is the answer - I think each print head needs to automatically align itself in 3 dimensions to the same 2 or 3 fixed points in 3d space - then it should be possible for all print heads to then be fairly perfectly aligned with each other.
3) Independence - I still think only having one (IDEX) or two (Your polar machine) axis fully independent is going to limit the shapes you can produce - even producing an oval is going to be difficult as some heads are going to need to be moving at non optimal speeds to allow the other ones to keep up with the different distances they need to travel on the fixed motion of the bed, Therefore I would suggest you keep the polar design - but add another axis of movement at right angles to the in out movement. My 'Revolutionary concept' short purposely included a 'house' or 'Triangle on top of Square shape' to demonstrate the sort of shapes that will only be possible with fully independent movement in the XY axis. I think that the new right angle axis could then be used for a lot of the motion - with the polar movement then only being used to move the print around maybe twice to per layer - to give each head access to a new 'exclusion zone'.
4) Multi colour - I still think having some kind of colour multiplexing on some of the heads is going to be desirable - both the allow larger number of colours than print heads, plus to allow parallel printing of some colours in certain parts of the print.
5). Pausing - it was a shame that you had to cancel your first big print due to time constraints. Personally for really long prints I almost always pause them overnight - I use a non heated build plate and bring each print head down to 35 degrees so that it doesn't need any fans and then just pause the print overnight or when I go out. I usually only pause during a colour change. I have even managed a couple of times to completely power off the printer and 'power interruption restart it' and still get a successful print.
Will probably produce some sort of demo myself - probably of an IDEX with a couple of extra independent cantilever heads added in front, plus maybe the bed changed to Polar - but am currently waiting to see what BambuLab come out with in the next few weeks if the rumours are correct - would be very interested to see one of your 'prediction' videos on this as you have been pretty spot on in the past.
I have some info on the new Bambu, but I think it will be a few months out.
I was not impressed until I've seen the rotary bed. This is an interesting improvement, wonder why there are not a lot of persons using this design.
This is so cool!
Now this could be the beginning of something new. I am not sure if this idea has been attempted before.
Amazing machine sir, excellent work! To me, a pellet fed design feels like the right direction given your comments regarding filament swaps. Looking at the comments regarding the ability to print the entire platform, missing the middle comments, given all heads are printing the same layer, is it possible to move the relative motion to the heads themselves mounted on continuous X cross rails compensating relative head heights with some sort of mount design? Just blue-sky thoughts here about an amazing machine. Please, want to see more.
Insane, thanks man. Cool to see some innovation outside of oem. You should stop printing trash cans and start printing side tables. Also look into rhino 3d grasshopper and the opportunity to make custom gcode for your printers if you aren't using something like this already. Anyway cheers, cool stuff
For the code you can use full control xyz it’s definitely the easiest way to make multi axis gcode
to calibrate it you could use a metod like VAOC of idex vcore 4, a camera on the build plate to correct the offset of the nozzles.
You can try To make it print upside down and with a small printing Tower for centre to be not hollow. Even though most things that we use are Hollow centre