Proposal for a new way to combine low cost 3d printers together. Along similar lines, to this idea from the Proper Printing RUclips channel • I built an experimenta...
What makes cheap printers "cheap" is because of the simpler research and development process of producing a basic machine - not because of the parts. The complexity of the hardware and software symbiosis to make this work would cause the printer to be as expensive as existing multi-head printers like the Prusa XL
Almost anything is possible with a 'need' for something new, different or both! I can see whoever programs the software will be 'very good' to do that! 😎
I was thinking of something like this, but i was thinking of having simply the bed move. Imagine having the 4 printers set up in a circle, with some additional hardware you could place all the beds on a single disk and attach a motor to it, each printer would work on one bed, then when a material change is needed the disk would rotate and put bed #1 in the printer #2, with the additional benefit that now printer #1 is free to start a new piece until the original piece it was working on finishes being worked on by the other printers. You could have the benefit of 4 printers working in parallel with 4 possible material changes AND virtually no down time. The software to make this work would certainly be a challenge because all 4 printers have to be syncronised, but the benefits are interesting.
Yes - I am also thinking that a circular bed could help. I've made a short to show one way of doing it where the heads all rotate around a circular bed independently - but I think it would be more practical to have 4xXY printers also collaborating on a circular bed - a bit like the machine that Nathan Builds Robots has built - but with full XY movement, rather than just in/out like he has done. There will as you said be big benefits in terms of colour changes - but I think the real game changer will be the ability to print the same part in parallel across multiple heads.
@@dwuk99 yeah a singular axis with the head on that pivots around a center point is cool but much more complicated to make. The 4 printers on 4 beds that move from one to the next is cool af tho. One problem though is certainly layer time per each printer. If printing in parallel each layer is still gonna be as long as the longest layer currently being printed among the 4 printers, so the actual efficiency of the parallel printing would vary wildly depending on the piece, but I believe there would still be interesting use cases, especially with pieces made of different materials. Think of a part that uses pla for the structure, some tpu for some corners and maybe water soluble supports. A machine like this I believe would smoke any mmu currently on the market, and it does have some additional complexity but if it works it's worth it.
cool idea but you would have 4 motors per print head, x y z and extruder. You would need a I2C or something else to run all those printers because I dont think we have a 16 stepper drive board. All the Y axis's are on the same rail and you would need a way to fit 4 motion systems on it. Rack pinion or a long screw and driving the nut could work. I think 4 belts would take too much room.
Thanks for the comment. I envisage each head having its own donor controller board and psu - with something like a raspberry pi coordinating the 4 printers. Haven't worked out the details of how the Y axis would work - rack and pinion would I guess be easiest - especially if you wanted the printers to be able to be separated.
Agreed - To me the most important benefits are print speed and larger bed size. Filament waste can either be reduced by using Flush-into objects, or re-melted into useful flat sheets.
Is it added complexity though? Maybe marginally more complex than an IDEX setup, but not more complex than a tool changer solution for sure, I have a 3 color machine that does everything through the one head and I literally never use it. You forgot the other major problem, time. This solution has the potential to do multi color/material prints with basically zero additional time, possibly faster in some cases. It's plausible to have one head printing in one area while the other is printing a different color elsewhere. But at the very least the software can have the next head coming over ready to print while the first head is finishing a motion. Reducing changeover time to practically nothing.
Thanks. RUclips shows only "1 Month ago" now. However, I for myself thought about combining two printers for some time now. If you f.e. have two Voron 2.4-style printers, their beds are fixed. Fix the printers side by side. Add a rail system with one movable bed. So the print can be moved between the two printers. If you make the movable bed system easy to remove, you still can use the printers seperatly. Nice idea? Calibration would be hell. Both printers would have to be calibrated precisely. However, not impossible.
I like the out of the box thinking but the filament waste argument is kinda tired and misleading. One a bambu printer a filament swap is about .2g on avg. So it's about 5 swaps per gram or up to 1000-5000 swaps a 1kg roll. Prime towers can be reduced or outright eliminated if you make a sacrificial, but functional, purge to infill part. There is a "long retract before cut" option which has been shown to reduce filament purge by up to 70% too with purge amount tweaking. So if you took a bambu p1s or x1c you would need hundreds of thousands to well over a million filament swaps to get to the break even price of an XL 5 head. and you would then have access to 16 colors instead of 4 in your demo or 5 in the XL case. so for the cost of the 4 prusa mini you can have 4 A1 with 4 color ams lite or 2 p1s with ams with the option of expansion. so 256(cubed) volume X 4 with access to 16 colors(4 per machine) or 256(cubed) volume x 2 with access to 8 colors initially or 32 colors maxed out. (16 per machine)
@@AwwwSnapperz all very good points - which if you look at some of my other videos or my making better used of purge waste thread you will see I am pretty familiar with. To me the biggest issue and hopefully the next big step forward in multi colour printing is the massive time savings that reduced numbers of colour swaps and parallel printing could bring, The ratrig vcore4 ams idex will be a small step in this direction - and I have a feeling that the next BL larger format printer may also be AMS But i believe fully independent multi colour printing is where the real time savings start to come, 4x A1minis with AMS lites would potentially give you 16 colour printing and more than 4x faster print times due to the reduced number of colour swaps.
Good idea but only for some solutions, it adds complexity, the one that might work is for the long, one color bed, but I am not sure if warp can be avoided.
defo interesting, i think the big problem you'll find over corexy toolchangers is the cost scale due to each new print head needs 3 motors (XYE) rather than 1 (E). but for 1-2 toolheads i suspect this could be competitive price wise with the right machine. The mini is £382.80 as a kit in the UK, just a non starter at that price. A V2.4 350mm stealthchanger I'm currently working on looks to be on track to cost me around £1k with 6 toolheads (i got a very good deal on a 350mm kit a while back, but still each extra hotend is costing ~£55). (printed part cost is not included as i printed it all myself) parallel single part printing is IMO the most interesting concept and might be worth the extra cost, the strategy to partition work however is going to be a complex problem, especially if the toolheads have potions that can intersect another printers, print path. you'd need to know when these collisions could occur and avoid generating gcode that causes this, some moves will have to be in essence atomic. there's also the question of how work is partitioned, the naive method would be each head gets some cube of area they work in. but this will result in poor utilisation if the dimensions of the object are not the same as the widest point. a per layer work distribution model could probs solve that, but my gut tells me the complexity of this becomes a problem is non trivial.
Thanks for commenting, i agree that the prusa mini is quite expensive for this type of machine, I have done similar videos for other much cheaper machines llike kingroons, bit for some reason the prusa mini videos are getting a lot more views so may go with at least a 2 headed prusa version as the fhe first prototype. A 6 head stealthchanger sounds very interresting, especially for only £50 per head.
@@dwuk99 yeah, still in the build phase, but i managed to hunt out some pretty good deals for all the parts. hopefully when I'm done and have tuned ill put the results up on the voron discord because the cost I've achieved seems to be unheard of
Not sure if that would be really less expensive than an voron IDEX for example but it surely is less expensive than a prusa XL. I like to see those kind of concept anyways. Do you see the hash kinematics concept video? Not as modular as your concept but still interesting. My dream printer would be a IDEX Dual grantry with a convoyer bed. But as other have mentioned it, the harder part is on the software side.
No I hadn't seen that Hash Kinematics - looks like an interesting idea - but unfortunately not independent on the Z and only 2 of the heads are independent on the XY I think, with the other two being dragged around and oozing. IDEX dual gantry is an interesting idea - I have two old IDEX printers and had considered joining them together - but again they will still not be independent on the Y axis if the bed is still moving. I agree that some sort of ejection system would be useful in this design - not sure that a conveyor belt is going to be reliable enough - will probably just implement knocking off the prints with one of the print heads (or maybe an extra head just for that purpose)
While hardware will be exactly the same price as 4 separate printers, the cost of the software will be incredible high, multithreading printers as shown is an incredibly complex computation task which will offset the cost of the printer more than just buying 4 separate printers and loosing some money on material purge blocks and print time. It would be a very cool project showcase though if done and I look forward to it.
Thanks - I don't think the software will be that complex for the simple multi colour printing option. Also parallel printing.of different objects shouldn't be that hard. I agree though that collaborative parallel printing of the same object will be more complex - For prototyping I would probably be planning to pre- split the object into 4 in fusion 360, then tell the slicer that each part is a different colour - so that it splits the object out into separate tools. Then post process the gcode to split it out into 4 separate files, Maybe with some extra processing to change the order of instructions and introduce delays to ensure that the heads stay away from each other.
Doing something like this is easy with RepRap Firmware and a Duet Board, you just need to connect all motors to one board and make a matrix of all the tools and motors, the rest will be handled by the firmware you just need one gcode file with toolchanges which most modern slicers can make and good pre and post scripts for toolchanges
@@martinhuber6702 Thanks yes - I think that is the sort of thing proper printing is doing. My ideal approach would to continue if possible to use the electronics (drivers, cpu, PSU etc) from the donor printers - with a RPI (Klipper or Octoprint like) central controller to coordinate the individual printers. But will look at RepRap with Duet too.
This is a nice idea but has many drawbaks and only limited benefits. One of the biggest problems will be the Software for path planing. There is software out there that is capable of the path planing for multiple toolpath without collisions, but this CAM-Software is very, very expansive. Several thousand dollars expansive. For the "one Objekt, multiple color"-Mode or Copymode, would the software fairly easy, with very little benefits. TheIndipendent toolpath, with the highest benefits would be way to complex.
Thanks for viewing the demo. I see two big benefits with these multi printer ideas 1) much greater print speed and reduced waste for single threaded multi colour prints - I think you could reduce print times for some models by over 75 percent by either reducing the need for colour changes or doing them in parallel. I am a fair way through some software to analyse the benefits of various different approaches and configurations, 2) as you said parallel printing is where even bigger benefits come - for both multi colour or single colour objects. I don't think the software to do this will be too expensive to prototype at least - because I am basing it on the free 3d software blender. Building on the demo's I have already created, I am planning on creating accurate models of the proposed 3d printers in blender - including things I haven't yet included like moving cables etc, plus the actual printed models as they build up. It should then be fairly easy to detect potential collisions by checking for overlapping meshes. Tool paths can then be proposed and then automatically optimised until non colliding routes are found,
As an owner of two idex 3d printers I have long been an enthusiast of 3d printers with multiple hotends. However multiple hotends create multiple potential points of failure, so these days I think someone should try non-stick (in the inside) nozzles coupled with cold pulls. Isn't the cold pull the easiest way to clean a nozzle? So you could cold pull, cut the blob off, eject it (it would be a lot less filament waste) and route in the new filament. This would require a rapid cooling solution for the hotend and active filament route diversion, but nothing super complex.
Interesting idea - I see BambuLab have just announced a more efficient retraction/cut method - but it still leaves a lot of waste int he nozzle that needs to be purged out - so pulling even more out (at a lower temp as you suggest) and then ejecting it somehow near the hot end (rather than trying to manage it like they do in the MMU) has some potential. I guess my thinking on multiple gantries is that if have 2-4 completely separate 3d printers then yes that does introduce more points of failure - but if these printers are reliable then manageable. Then what I am suggesting is when required allow those 4 printers to work together on a single print - to a) Reduce waste and b) increase colour change speed and c) If possible if they can print parts of the same object in parallel, increase overall print speed.
Thanks, yes In do intend building one. I am leaning towards developing mode 4 further (the 2x2 option) - with an optional rotational bed added in the middle to simplify head movement, and allow things like parallel vase mode. The area outside outside of the circular bed could still be used for other parallel single colour prints or prime towers etc. Will probably allow one or two of the individual heads to be colour changing too (for the lessor used colours in say a 7 colour print). Using something like a Palette2/MMU3 or 3d Chamelion.
You might simply use two or more printers with normaly fixed bed, like voron 2.4-350mm-style (the new sovol sv08 might be a budget choice here). Put them in a row. Construct a long movable bed tray allowing to move the bed trough multiple printers. This movable bed construction should be mounted above the existing beds and it should be removable. As the bed only moves while changing colour by moving to another printer, you do not need expensive linear rails. Still, every printer would be a seperate printer with it's own controller. The bed moving addon would have it's own controller. Software would be quite easy. There is a gcode command to set a pin state and a command to wait for a pin state. So all you need is a simple gcode postprocessing: You slice for a multi tool printer, as example a 3-tool. This gcode is seperated for each printer. After a layer is completed, a pin is set for a second to tell the bed moving system to go to the next printer. First printer now waits for a signal on another pin to continue to print (you might reduce also printhead temperature for waiting time and move the printhead to a comfortable position to avoid oozing etc). Bed moves to next printer, sends signal if final position is reached, next printer starts. If you place the rails for the bedchanger outside the original bed area, you still can position the moving bed only on one printer and the others use their original beds, so all three printers print independend. Or you only use 2 for dual material and the third prints something else. The biggest problem would be all the offset, belt-tension and skrew calibration, as all printers must print exactly the same. With a toolchanger the xyz-gantry is the same, you only need to compensate the xyz-offset of the tool. With multiple printers you have to exactly calibrate every printer to exactly the same. Much much harder. I thought about a "bed-mover" a few years back, before coreXY got common and before Klipper was popular. I thought two Deltas with moving bed might be a way. However, checking the differences between two identical Deltas, I actually considered it nearly impossible to get the result to be exactly the same on both printers. To say it simple: It does not matter if the print result of one printer is contantly of by 0.05mm in x-width on a 5cm cube. It would be a very precise hobby printer. However, if one printer has +0.05, the other -0.05 due to different belt tension, the difference would matter as the different materials would not be at the matching positions. This is the same for y of course. And for any slight deviation from perfectly perpendicular axis (skrew compensation). Ambient temperature changes screw up all your calibration anyway. So it wouldnt be a one time calibration, but you have to calibrate before any print ;-( Also, any multiple printer setup is hard to enclose. For PLA and PET-G an open frame printer might work. But ABS or Nylon - not recommended. So while I realy like the thought, imho the practical problems are still a problem. With CAN-bus or USB to connect toolheads, a toolchanger seems much more easy. Actually you easy could disconnect the toolhead due to this. A nice toolhead is not cheap. However, if you have multiple printers, you could make a toolchanging bay for only one, but still make the other toolheads compatible (not automatic toolchange but manually). So you still need only one tool per printer, but if you want multimaterial, you simply deactivate 1 or two of the other printers to lend their tools...
Thanks for your very comprehensive reply. I think what you are describing is very similar to this alternative proposal ruclips.net/video/uXqAenyT-fY/видео.htmlsi=OZiaiirHaGc6lwL2 . I agree that for any type of multi head parallel printer getting them to exactly align would be a challenge, and would need to be fully automated via some sort of probing of pre-determined fixed points. Shared head toolchangers is an interesting idea.
Thinking about your suggestion further- I think a voron 2.4 type printer is a good choice as it is fixed bed (and therefore the heads are fully independent) - then it means that you could in theory have both printers printing part of the model at the same time on a joint add on sliding bed. Will attempt to simulate a parallel multi colour print with 3x 2.4 printers - with one also having a stealth changer.
Nice animation. Autodesk did something very similar in the past, called "project Escher". I imagine the slicing process should be something very complicated - but if you want one printer platform printing at a time, Im sure RRF can handle that. Im saying that because I created a concept where 2 different parts could be printed simultaneously and David Crocker showed up in the comments and said that could be implemented on RRF if needed. Sadly I never had the money and the balls to actually go forward on the project.
Thanks for that info - Project Escher looks like a very similar concept to what I am suggesting - found a video about it - which is 8 years old - so shame it didn't go anywhere. Also thanks for the RRF info too, I was thinking along the lines of clipper - with each 'component' printer having its own cpu - with then some sort of overall RPI like controller coordinating things. Next steps for me is probably to try creating a couple of flat bed cantilevers out of old printers I have. (I came across a design called 'flatpack' which I might try and use first this), Then adding some modded Kingroon KP3S's into the 'array'.
thanks for the comment, agreed - I don't see the max number being fixed - it could be anything from 2 upwards, and 8 or more certainly a possibility once the algorithms for print segmentation and head avoidance worked out. Might get quite expensive though unless cheaper base printer (link Kingroon for example) used.
Yes - don't see why it wouldn't work - if you add some extra fully independent cantilever print heads along the side- so that they can move along the belted bed independently of the belt movement.
Looked a bit more into actually modding a CR30 - but not sure it will be that practical to fit more heads onto it - have just proposed this belt type add on for Bambu Lab Core XY printers. Not as good as this fully independent proposal - as wouldn't allow parallel printing - but might be fairly practical and will definitely bring massive speed and waste improvements to big multi colour prints. ruclips.net/video/uXqAenyT-fY/видео.html
Would be easier, unfortunately if you replace the independent Y movement with the bed moving you can't do much parallel printing. I may well however have a moving bed too - that maybe only moves once or twice per layer - just to move the prints to a more convenient place for the heads to reach. T
Instead of using toolchanger we will build a printerchanger. 3D printing with robotic arms does something similar. But in this design, even on the video demonstrating the idea, author could not comply the collision of printing heads and the problems and difficulties that arise from this.
Demo movements were created by hand - so you are correct that there are one or two slight overlaps. I am working on some better examples auto created from gcode - plus will include full collision detection. I agree though that for some models the need to avoid collisions will slow things down to almost current 'single threaded' speeds - but for others I think it will be possible to get close to 4x speed for single colour prints, and potentially even more than 4x speeds for multi colour.
@@dwuk99 Titan Robotics created a similar printer 7-8 years ago, but now they make bed slinger with a toolchanger. So perhaps the complexity of the development is not worth the benefits it provides. However, it is always interesting to see good engineering work, so I hope that you will succeed 🙂
Any idling print heads still ends up needing a priming block possibly in a region it can't reach so multiple priming blocks. no difference from a tool changer aside from more moving points of failure. Its one of those ideas that sounds cool in your head, but practically, it is useless.
No more points of failure than 4 printers I think - and there are plenty of print farms with a lot more than 4 printers in them that work 24 hrs a day. I think the biggest challenge is probably keeping the heads in alignment - so that would need to be fully automated somehow - perhaps like CNC tool changers do it, perhaps several times during each print. Probably easiest to have one prime tower in the middle where all heads can reach, but it is possible that having more than one prime tower could increase overall print speed - because for example head2 could be priming in parallel to head1 finishing off its current print.
@dwuk99 you have more points of failure and higher chance of failure because you are trying to move the entire x gantry + z axis instead of just y so you definitely have way more failures than if you just ran 4 printers individually. If one of them is down and the issue is the y axis now all 4 are down. It offers no mechanical advantage and will suffer at higher speeds in which that case you might as well go for a tool changer. If it's a IDEX like system, all that mass changing directions is just going to cause interference and artifacts on the others.
@@am_stephanos Agreed that changing the printers from moving bed to moving gantry will present rigidity challenges - which might need to be addressed by some sort of top rail. The mechanical advantage over IDEX or tool changers though is parallel printing - without fully independent print heads you are limited to serial printing.
Thanks for the comment. I think I have included more space than you actually need - so I don't think the whole thing will need to be quite as long as I have shown. Re moving bed, for single head at a time multi colour printing for example I agree that a moving bed would be simpler. However the real driver for this sort of design is parallel printing. With fixed heads and a single moving bed every head would be limited to printing more or less the same thing at the same time, which would work for some models, but would severely limit what can be done in parallel. Nathan Builds Robots current rotary design is I think more or less as you suggest - will be interesting to see how he gets on with it and whether he ends up adding another degree of movement to his print heads.
Well, there are some printers that do that, like this one ruclips.net/video/XBJMz457pEo/видео.html&ab_channel=Proper Printing the firmware and software is the biggest hassle. Different kinematic systems and configuration and implementation in slicers.
Thanks yes - I have been thinking about this for a while (mainly in relating to thinking about what the perfect next Bambu Lab printer would be - see my A1 Mini version of this video), and the Proper Printing video you link to did inspire me to release these videos. I agree that one of the biggest challenges will be the software to coordinate the various printers, align them with each other and coordinate the axis movements to make sure that they don't hit each other. Then the slicer would also be a challenge - I see this challenge being addressed a bit like parallel processing happens in multi core CPUs - by perhaps looking at the generated GCODE, analysing it, and working out how to split the movements out so that they can be done in parallel.
Thanks yes I have been shown the adobe video - shame it never got developed - I would still like to progress something along these lines. Before purchasing lots of extra printers I will probably start with an IDEX bed slinger, with a couple of extra fully independent cantilevered heads added (making 4 heads in total) to see how far I can get with it.
@@dwuk99 the problem is the slicer not the printer you need to divide up the print and have overlapping layers so you're not trying to but join two sections
@@TheShorterboy Agreed - how I see it working is that you divide up each layer into sections - probably double the number of print heads (so 8) - so that the heads to not clash with each other while they are printing their allocated section, plus have the sections overlap between layers to help with strength. For initial prototyping I would expect to divide the model into the 8 sections prior to slicing - probably in Fusion 360 or using some sort of python script to split the STLs. Then slice the model as 8 colours, then post process the generated GCODE to split out the 8 colours into separate files for each print head.
@@dwuk99 the simplest form I can think of is as the slicer process's each layer offset the X zero reference by + or - some value then slice that layer then repeat then maybe sync the gcode (no idea how) this will give give the overlap
![OsamaSon - ik what you did last summer [Official Music Video]](http://i.ytimg.com/vi/4DIQDnu3mr4/mqdefault.jpg)
What makes cheap printers "cheap" is because of the simpler research and development process of producing a basic machine - not because of the parts. The complexity of the hardware and software symbiosis to make this work would cause the printer to be as expensive as existing multi-head printers like the Prusa XL
Agreed - the software will be a challenge - but if it can be demonstrated to work then I think parallel printing would be a good way to boost speed.
Almost anything is possible with a 'need' for something new, different or both! I can see whoever programs the software will be 'very good' to do that! 😎
I was thinking of something like this, but i was thinking of having simply the bed move. Imagine having the 4 printers set up in a circle, with some additional hardware you could place all the beds on a single disk and attach a motor to it, each printer would work on one bed, then when a material change is needed the disk would rotate and put bed #1 in the printer #2, with the additional benefit that now printer #1 is free to start a new piece until the original piece it was working on finishes being worked on by the other printers.
You could have the benefit of 4 printers working in parallel with 4 possible material changes AND virtually no down time.
The software to make this work would certainly be a challenge because all 4 printers have to be syncronised, but the benefits are interesting.
Yes - I am also thinking that a circular bed could help. I've made a short to show one way of doing it where the heads all rotate around a circular bed independently - but I think it would be more practical to have 4xXY printers also collaborating on a circular bed - a bit like the machine that Nathan Builds Robots has built - but with full XY movement, rather than just in/out like he has done. There will as you said be big benefits in terms of colour changes - but I think the real game changer will be the ability to print the same part in parallel across multiple heads.
@@dwuk99 yeah a singular axis with the head on that pivots around a center point is cool but much more complicated to make. The 4 printers on 4 beds that move from one to the next is cool af tho. One problem though is certainly layer time per each printer. If printing in parallel each layer is still gonna be as long as the longest layer currently being printed among the 4 printers, so the actual efficiency of the parallel printing would vary wildly depending on the piece, but I believe there would still be interesting use cases, especially with pieces made of different materials. Think of a part that uses pla for the structure, some tpu for some corners and maybe water soluble supports. A machine like this I believe would smoke any mmu currently on the market, and it does have some additional complexity but if it works it's worth it.
Interesting idea! Definitely out side the box. Anxious to see your results
This is seriously cool as hell.
cool idea but you would have 4 motors per print head, x y z and extruder.
You would need a I2C or something else to run all those printers because I dont think we have a 16 stepper drive board.
All the Y axis's are on the same rail and you would need a way to fit 4 motion systems on it. Rack pinion or a long screw and driving the nut could work. I think 4 belts would take too much room.
Thanks for the comment. I envisage each head having its own donor controller board and psu - with something like a raspberry pi coordinating the 4 printers. Haven't worked out the details of how the Y axis would work - rack and pinion would I guess be easiest - especially if you wanted the printers to be able to be separated.
Lots of added complexity to battle purge blocks or filament poop.
Agreed - To me the most important benefits are print speed and larger bed size. Filament waste can either be reduced by using Flush-into objects, or re-melted into useful flat sheets.
Is it added complexity though? Maybe marginally more complex than an IDEX setup, but not more complex than a tool changer solution for sure, I have a 3 color machine that does everything through the one head and I literally never use it.
You forgot the other major problem, time. This solution has the potential to do multi color/material prints with basically zero additional time, possibly faster in some cases. It's plausible to have one head printing in one area while the other is printing a different color elsewhere. But at the very least the software can have the next head coming over ready to print while the first head is finishing a motion. Reducing changeover time to practically nothing.
if there were the necessity though, having a machine like this would be insanely better than the down times mmu systems have
sick idea, i hope to see it a reality
Seriously? Nobody noticed this was posted on April Fools day?
Thanks. RUclips shows only "1 Month ago" now. However, I for myself thought about combining two printers for some time now. If you f.e. have two Voron 2.4-style printers, their beds are fixed. Fix the printers side by side. Add a rail system with one movable bed. So the print can be moved between the two printers. If you make the movable bed system easy to remove, you still can use the printers seperatly.
Nice idea? Calibration would be hell. Both printers would have to be calibrated precisely. However, not impossible.
I like the out of the box thinking but the filament waste argument is kinda tired and misleading.
One a bambu printer a filament swap is about .2g on avg. So it's about 5 swaps per gram or up to 1000-5000 swaps a 1kg roll.
Prime towers can be reduced or outright eliminated if you make a sacrificial, but functional, purge to infill part.
There is a "long retract before cut" option which has been shown to reduce filament purge by up to 70% too with purge amount tweaking.
So if you took a bambu p1s or x1c you would need hundreds of thousands to well over a million filament swaps to get to the break even price of an XL 5 head. and you would then have access to 16 colors instead of 4 in your demo or 5 in the XL case.
so for the cost of the 4 prusa mini you can have 4 A1 with 4 color ams lite or 2 p1s with ams with the option of expansion.
so 256(cubed) volume X 4 with access to 16 colors(4 per machine)
or 256(cubed) volume x 2 with access to 8 colors initially or 32 colors maxed out. (16 per machine)
@@AwwwSnapperz all very good points - which if you look at some of my other videos or my making better used of purge waste thread you will see I am pretty familiar with. To me the biggest issue and hopefully the next big step forward in multi colour printing is the massive time savings that reduced numbers of colour swaps and parallel printing could bring, The ratrig vcore4 ams idex will be a small step in this direction - and I have a feeling that the next BL larger format printer may also be AMS But i believe fully independent multi colour printing is where the real time savings start to come, 4x A1minis with AMS lites would potentially give you 16 colour printing and more than 4x faster print times due to the reduced number of colour swaps.
Good idea but only for some solutions, it adds complexity, the one that might work is for the long, one color bed, but I am not sure if warp can be avoided.
defo interesting, i think the big problem you'll find over corexy toolchangers is the cost scale due to each new print head needs 3 motors (XYE) rather than 1 (E). but for 1-2 toolheads i suspect this could be competitive price wise with the right machine. The mini is £382.80 as a kit in the UK, just a non starter at that price.
A V2.4 350mm stealthchanger I'm currently working on looks to be on track to cost me around £1k with 6 toolheads (i got a very good deal on a 350mm kit a while back, but still each extra hotend is costing ~£55). (printed part cost is not included as i printed it all myself)
parallel single part printing is IMO the most interesting concept and might be worth the extra cost, the strategy to partition work however is going to be a complex problem, especially if the toolheads have potions that can intersect another printers, print path. you'd need to know when these collisions could occur and avoid generating gcode that causes this, some moves will have to be in essence atomic. there's also the question of how work is partitioned, the naive method would be each head gets some cube of area they work in. but this will result in poor utilisation if the dimensions of the object are not the same as the widest point. a per layer work distribution model could probs solve that, but my gut tells me the complexity of this becomes a problem is non trivial.
Thanks for commenting, i agree that the prusa mini is quite expensive for this type of machine, I have done similar videos for other much cheaper machines llike kingroons, bit for some reason the prusa mini videos are getting a lot more views so may go with at least a 2 headed prusa version as the fhe first prototype. A 6 head stealthchanger sounds very interresting, especially for only £50 per head.
@@dwuk99 yeah, still in the build phase, but i managed to hunt out some pretty good deals for all the parts. hopefully when I'm done and have tuned ill put the results up on the voron discord because the cost I've achieved seems to be unheard of
im bad at math i miss calulated the cost its actually £60 per hotend
Not sure if that would be really less expensive than an voron IDEX for example but it surely is less expensive than a prusa XL.
I like to see those kind of concept anyways.
Do you see the hash kinematics concept video? Not as modular as your concept but still interesting.
My dream printer would be a IDEX Dual grantry with a convoyer bed. But as other have mentioned it, the harder part is on the software side.
No I hadn't seen that Hash Kinematics - looks like an interesting idea - but unfortunately not independent on the Z and only 2 of the heads are independent on the XY I think, with the other two being dragged around and oozing.
IDEX dual gantry is an interesting idea - I have two old IDEX printers and had considered joining them together - but again they will still not be independent on the Y axis if the bed is still moving.
I agree that some sort of ejection system would be useful in this design - not sure that a conveyor belt is going to be reliable enough - will probably just implement knocking off the prints with one of the print heads (or maybe an extra head just for that purpose)
While hardware will be exactly the same price as 4 separate printers, the cost of the software will be incredible high, multithreading printers as shown is an incredibly complex computation task which will offset the cost of the printer more than just buying 4 separate printers and loosing some money on material purge blocks and print time.
It would be a very cool project showcase though if done and I look forward to it.
Thanks - I don't think the software will be that complex for the simple multi colour printing option. Also parallel printing.of different objects shouldn't be that hard. I agree though that collaborative parallel printing of the same object will be more complex -
For prototyping I would probably be planning to pre- split the object into 4 in fusion 360, then tell the slicer that each part is a different colour - so that it splits the object out into separate tools. Then post process the gcode to split it out into 4 separate files, Maybe with some extra processing to change the order of instructions and introduce delays to ensure that the heads stay away from each other.
Doing something like this is easy with RepRap Firmware and a Duet Board, you just need to connect all motors to one board and make a matrix of all the tools and motors, the rest will be handled by the firmware you just need one gcode file with toolchanges which most modern slicers can make and good pre and post scripts for toolchanges
@@martinhuber6702 Thanks yes - I think that is the sort of thing proper printing is doing. My ideal approach would to continue if possible to use the electronics (drivers, cpu, PSU etc) from the donor printers - with a RPI (Klipper or Octoprint like) central controller to coordinate the individual printers. But will look at RepRap with Duet too.
Youre literally just making up some bullshit
This is a nice idea but has many drawbaks and only limited benefits.
One of the biggest problems will be the Software for path planing. There is software out there that is capable of the path planing for multiple toolpath without collisions, but this CAM-Software is very, very expansive. Several thousand dollars expansive. For the "one Objekt, multiple color"-Mode or Copymode, would the software fairly easy, with very little benefits.
TheIndipendent toolpath, with the highest benefits would be way to complex.
Thanks for viewing the demo. I see two big benefits with these multi printer ideas 1) much greater print speed and reduced waste for single threaded multi colour prints - I think you could reduce print times for some models by over 75 percent by either reducing the need for colour changes or doing them in parallel. I am a fair way through some software to analyse the benefits of various different approaches and configurations, 2) as you said parallel printing is where even bigger benefits come - for both multi colour or single colour objects. I don't think the software to do this will be too expensive to prototype at least - because I am basing it on the free 3d software blender. Building on the demo's I have already created, I am planning on creating accurate models of the proposed 3d printers in blender - including things I haven't yet included like moving cables etc, plus the actual printed models as they build up. It should then be fairly easy to detect potential collisions by checking for overlapping meshes. Tool paths can then be proposed and then automatically optimised until non colliding routes are found,
As an owner of two idex 3d printers I have long been an enthusiast of 3d printers with multiple hotends. However multiple hotends create multiple potential points of failure, so these days I think someone should try non-stick (in the inside) nozzles coupled with cold pulls. Isn't the cold pull the easiest way to clean a nozzle? So you could cold pull, cut the blob off, eject it (it would be a lot less filament waste) and route in the new filament. This would require a rapid cooling solution for the hotend and active filament route diversion, but nothing super complex.
Interesting idea - I see BambuLab have just announced a more efficient retraction/cut method - but it still leaves a lot of waste int he nozzle that needs to be purged out - so pulling even more out (at a lower temp as you suggest) and then ejecting it somehow near the hot end (rather than trying to manage it like they do in the MMU) has some potential. I guess my thinking on multiple gantries is that if have 2-4 completely separate 3d printers then yes that does introduce more points of failure - but if these printers are reliable then manageable. Then what I am suggesting is when required allow those 4 printers to work together on a single print - to a) Reduce waste and b) increase colour change speed and c) If possible if they can print parts of the same object in parallel, increase overall print speed.
Very cool video, keep up the good work, and maybe try to build one!
Thanks, yes In do intend building one. I am leaning towards developing mode 4 further (the 2x2 option) - with an optional rotational bed added in the middle to simplify head movement, and allow things like parallel vase mode. The area outside outside of the circular bed could still be used for other parallel single colour prints or prime towers etc. Will probably allow one or two of the individual heads to be colour changing too (for the lessor used colours in say a 7 colour print). Using something like a Palette2/MMU3 or 3d Chamelion.
You might simply use two or more printers with normaly fixed bed, like voron 2.4-350mm-style (the new sovol sv08 might be a budget choice here). Put them in a row. Construct a long movable bed tray allowing to move the bed trough multiple printers. This movable bed construction should be mounted above the existing beds and it should be removable. As the bed only moves while changing colour by moving to another printer, you do not need expensive linear rails.
Still, every printer would be a seperate printer with it's own controller. The bed moving addon would have it's own controller.
Software would be quite easy. There is a gcode command to set a pin state and a command to wait for a pin state. So all you need is a simple gcode postprocessing: You slice for a multi tool printer, as example a 3-tool. This gcode is seperated for each printer. After a layer is completed, a pin is set for a second to tell the bed moving system to go to the next printer. First printer now waits for a signal on another pin to continue to print (you might reduce also printhead temperature for waiting time and move the printhead to a comfortable position to avoid oozing etc). Bed moves to next printer, sends signal if final position is reached, next printer starts.
If you place the rails for the bedchanger outside the original bed area, you still can position the moving bed only on one printer and the others use their original beds, so all three printers print independend. Or you only use 2 for dual material and the third prints something else.
The biggest problem would be all the offset, belt-tension and skrew calibration, as all printers must print exactly the same. With a toolchanger the xyz-gantry is the same, you only need to compensate the xyz-offset of the tool. With multiple printers you have to exactly calibrate every printer to exactly the same. Much much harder.
I thought about a "bed-mover" a few years back, before coreXY got common and before Klipper was popular. I thought two Deltas with moving bed might be a way. However, checking the differences between two identical Deltas, I actually considered it nearly impossible to get the result to be exactly the same on both printers. To say it simple: It does not matter if the print result of one printer is contantly of by 0.05mm in x-width on a 5cm cube. It would be a very precise hobby printer. However, if one printer has +0.05, the other -0.05 due to different belt tension, the difference would matter as the different materials would not be at the matching positions. This is the same for y of course. And for any slight deviation from perfectly perpendicular axis (skrew compensation).
Ambient temperature changes screw up all your calibration anyway. So it wouldnt be a one time calibration, but you have to calibrate before any print ;-(
Also, any multiple printer setup is hard to enclose. For PLA and PET-G an open frame printer might work. But ABS or Nylon - not recommended.
So while I realy like the thought, imho the practical problems are still a problem.
With CAN-bus or USB to connect toolheads, a toolchanger seems much more easy. Actually you easy could disconnect the toolhead due to this.
A nice toolhead is not cheap. However, if you have multiple printers, you could make a toolchanging bay for only one, but still make the other toolheads compatible (not automatic toolchange but manually). So you still need only one tool per printer, but if you want multimaterial, you simply deactivate 1 or two of the other printers to lend their tools...
Thanks for your very comprehensive reply. I think what you are describing is very similar to this alternative proposal ruclips.net/video/uXqAenyT-fY/видео.htmlsi=OZiaiirHaGc6lwL2 . I agree that for any type of multi head parallel printer getting them to exactly align would be a challenge, and would need to be fully automated via some sort of probing of pre-determined fixed points. Shared head toolchangers is an interesting idea.
Thinking about your suggestion further- I think a voron 2.4 type printer is a good choice as it is fixed bed (and therefore the heads are fully independent) - then it means that you could in theory have both printers printing part of the model at the same time on a joint add on sliding bed. Will attempt to simulate a parallel multi colour print with 3x 2.4 printers - with one also having a stealth changer.
Nice animation. Autodesk did something very similar in the past, called "project Escher". I imagine the slicing process should be something very complicated - but if you want one printer platform printing at a time, Im sure RRF can handle that.
Im saying that because I created a concept where 2 different parts could be printed simultaneously and David Crocker showed up in the comments and said that could be implemented on RRF if needed. Sadly I never had the money and the balls to actually go forward on the project.
Thanks for that info - Project Escher looks like a very similar concept to what I am suggesting - found a video about it - which is 8 years old - so shame it didn't go anywhere. Also thanks for the RRF info too, I was thinking along the lines of clipper - with each 'component' printer having its own cpu - with then some sort of overall RPI like controller coordinating things. Next steps for me is probably to try creating a couple of flat bed cantilevers out of old printers I have. (I came across a design called 'flatpack' which I might try and use first this), Then adding some modded Kingroon KP3S's into the 'array'.
why not do the wide bed and long bed types both at the same time for up to 8x speed with only 4 long
thanks for the comment, agreed - I don't see the max number being fixed - it could be anything from 2 upwards, and 8 or more certainly a possibility once the algorithms for print segmentation and head avoidance worked out. Might get quite expensive though unless cheaper base printer (link Kingroon for example) used.
interesting concept, but I dont think this will be cost effective tbh... its too complex...
Did you consider a belted bed like the CR30?
Yes - don't see why it wouldn't work - if you add some extra fully independent cantilever print heads along the side- so that they can move along the belted bed independently of the belt movement.
Looked a bit more into actually modding a CR30 - but not sure it will be that practical to fit more heads onto it - have just proposed this belt type add on for Bambu Lab Core XY printers. Not as good as this fully independent proposal - as wouldn't allow parallel printing - but might be fairly practical and will definitely bring massive speed and waste improvements to big multi colour prints. ruclips.net/video/uXqAenyT-fY/видео.html
maybe the bed moving would be better
Would be easier, unfortunately if you replace the independent Y movement with the bed moving you can't do much parallel printing. I may well however have a moving bed too - that maybe only moves once or twice per layer - just to move the prints to a more convenient place for the heads to reach. T
Instead of using toolchanger we will build a printerchanger.
3D printing with robotic arms does something similar. But in this design, even on the video demonstrating the idea, author could not comply the collision of printing heads and the problems and difficulties that arise from this.
Demo movements were created by hand - so you are correct that there are one or two slight overlaps. I am working on some better examples auto created from gcode - plus will include full collision detection. I agree though that for some models the need to avoid collisions will slow things down to almost current 'single threaded' speeds - but for others I think it will be possible to get close to 4x speed for single colour prints, and potentially even more than 4x speeds for multi colour.
@@dwuk99 Titan Robotics created a similar printer 7-8 years ago, but now they make bed slinger with a toolchanger. So perhaps the complexity of the development is not worth the benefits it provides.
However, it is always interesting to see good engineering work, so I hope that you will succeed 🙂
The hardware is the easy part. The problem is always the software 😂
Any idling print heads still ends up needing a priming block possibly in a region it can't reach so multiple priming blocks. no difference from a tool changer aside from more moving points of failure. Its one of those ideas that sounds cool in your head, but practically, it is useless.
No more points of failure than 4 printers I think - and there are plenty of print farms with a lot more than 4 printers in them that work 24 hrs a day. I think the biggest challenge is probably keeping the heads in alignment - so that would need to be fully automated somehow - perhaps like CNC tool changers do it, perhaps several times during each print. Probably easiest to have one prime tower in the middle where all heads can reach, but it is possible that having more than one prime tower could increase overall print speed - because for example head2 could be priming in parallel to head1 finishing off its current print.
@dwuk99 you have more points of failure and higher chance of failure because you are trying to move the entire x gantry + z axis instead of just y so you definitely have way more failures than if you just ran 4 printers individually. If one of them is down and the issue is the y axis now all 4 are down. It offers no mechanical advantage and will suffer at higher speeds in which that case you might as well go for a tool changer. If it's a IDEX like system, all that mass changing directions is just going to cause interference and artifacts on the others.
@@am_stephanos Agreed that changing the printers from moving bed to moving gantry will present rigidity challenges - which might need to be addressed by some sort of top rail. The mechanical advantage over IDEX or tool changers though is parallel printing - without fully independent print heads you are limited to serial printing.
Instead moving the toolheads, you should move the bed but as you need tons of space behind and the front... I don't know if it would be a great idea.
Thanks for the comment. I think I have included more space than you actually need - so I don't think the whole thing will need to be quite as long as I have shown. Re moving bed, for single head at a time multi colour printing for example I agree that a moving bed would be simpler. However the real driver for this sort of design is parallel printing. With fixed heads and a single moving bed every head would be limited to printing more or less the same thing at the same time, which would work for some models, but would severely limit what can be done in parallel. Nathan Builds Robots current rotary design is I think more or less as you suggest - will be interesting to see how he gets on with it and whether he ends up adding another degree of movement to his print heads.
Well, there are some printers that do that, like this one ruclips.net/video/XBJMz457pEo/видео.html&ab_channel=Proper Printing the firmware and software is the biggest hassle. Different kinematic systems and configuration and implementation in slicers.
Still, nice job and agree
Also don´t forget project Escher ruclips.net/video/m4A9pbkx5KE/видео.html
Thanks yes - I have been thinking about this for a while (mainly in relating to thinking about what the perfect next Bambu Lab printer would be - see my A1 Mini version of this video), and the Proper Printing video you link to did inspire me to release these videos. I agree that one of the biggest challenges will be the software to coordinate the various printers, align them with each other and coordinate the axis movements to make sure that they don't hit each other. Then the slicer would also be a challenge - I see this challenge being addressed a bit like parallel processing happens in multi core CPUs - by perhaps looking at the generated GCODE, analysing it, and working out how to split the movements out so that they can be done in parallel.
Have created another version of the proposal video with a more direct reference to the ProperPrinting design - plus some other options
Genius brain!
The channel Propper printing is working on this kind of a device in his latest yt video. -> Proper printing ruclips.net/video/XBJMz457pEo/видео.html
adobe had one of these 10 years ago, it's still on youtube
Thanks yes I have been shown the adobe video - shame it never got developed - I would still like to progress something along these lines. Before purchasing lots of extra printers I will probably start with an IDEX bed slinger, with a couple of extra fully independent cantilevered heads added (making 4 heads in total) to see how far I can get with it.
@@dwuk99 the problem is the slicer not the printer you need to divide up the print and have overlapping layers so you're not trying to but join two sections
@@TheShorterboy Agreed - how I see it working is that you divide up each layer into sections - probably double the number of print heads (so 8) - so that the heads to not clash with each other while they are printing their allocated section, plus have the sections overlap between layers to help with strength. For initial prototyping I would expect to divide the model into the 8 sections prior to slicing - probably in Fusion 360 or using some sort of python script to split the STLs. Then slice the model as 8 colours, then post process the generated GCODE to split out the 8 colours into separate files for each print head.
@@dwuk99 the simplest form I can think of is as the slicer process's each layer offset the X zero reference by + or - some value then slice that layer then repeat then maybe sync the gcode (no idea how) this will give give the overlap
Isn't that sort of a toolchanger with extra steps?
Yes, a bit like a tool changer - just cheaper and a lot faster.