From the abstract:
"...feature resolution below 100 micrometers. ... complex solid parts can be drawn out of the resin at rates of hundreds of millimeters per hour."
Key novel feature is the oxygen-permeable, UV-transmitting membrane at the bottom of the tank that creates a thin (down to 20um) inhibited 'dead zone' where the resin can't polymerize. They project the image for the current layer up through the membrane and the dead zone, so the build layer is actually within the tank. This means they can just draw the part up continuously from the top, with no stepping or processing needed after each layer. The thickness of the effective build layer can be controlled by adding a UV light-absorbing dye to the resin, which allows them to optimize for different print speeds.
I didn't see in the supplementary materials, does anybody know if they are using a pressurized air or oxygen source to increase oxygen flow through the membrane?
edit: to repond to my own Q: they can use both pressurized oxygen and pressurized air. Using pressurized air decreases the dead zone but still very workable.
it's like a laser printer but instead of toner it uses gooey resin. project (in UV light) a shape on the bottom of the goop and it will start to solidify in that shape (once it touches oxygen)... so gently pulling some solid surface goop out gets you to the point where you can "print" stuff! Pretty cool! It's not clear to me if the resin solidifies only on contact with UV light and the oxygen level is controlled from below or above, but at any rate the balance of UV light (in a particular projected design) and oxygen is what makes this possible.
I think the oxygen prevents solidification which is what stops the object from sticking to the screen they project the image on to. By the looks of it the oxygen just diffuses from the air through the screen.
First carefully consider what you want to print. For example, this resin type printer is not a great technology for creating strong replacement parts around the house.
Joe DeSimone is something of a legend around this area (RTP). The guy is a prolific inventor at UNC Chapel Hill [1] and is a genius when it comes to chemical engineering. He has also won the MIT Lemelson prize [2]. His partner in this, Ed Samulski, is no slouch either [3].
These guys don't mess around. Their inventions do spin off into real companies with real products, and they have a track record to prove their success. Joe has been involved with projects as diverse as an environmentally-friendly dry cleaning technology [4] to nanocarriers for vaccines [5].
I was ready to write this off as another promising idea that would either never make it to market or experience serious issues with the quality and reliability of the polymers. Knowing now that DeSimone is behind this is all the reassurance I need. This will be a real product, it will make it to market, and it will produce quality 3D prints.
Depending on how they're handling the "tech transfer", they might not actually have much to do with the company. A lot of academics prefer to keep being academics and just collect their royalty checks.
Not at all. Carbon is a company; and building a company is just as difficult as building a technology (though a totally different skill set).
It's not that engineers aren't important, but they shouldn't be running the company. It's not the skill they have cultivated. They do still probably have a significant equity stake, so all is not lost.
I don't disagree that engineers bring a different skill set to the mix. The last thing you might want is engineers or scientists running a young company - though Microsoft, Google, Facebook, HP, Honda, Fairchild, Porsche, etc. turned out OK. ;-)
I just find it odd that Dr. Alex Ermoshkin the co-founder is listed somewhere after "Head of Customer Engagement" for example.
I just noticed that they are listed under the heading "North Carolina Office" and not "Leadership Team" so perhaps it is a simple as academics wanting to stay in academia.
Yeah, that's what I would guess. Many academics simply have no interest in running a business (often because they've tried and found that engineering is more fun).
The site is so cool that it just screams "fake". Cool pictures, not much copy, no details.
The big difference here is that they're doing photopolymerization at the bottom of the tank, rather than at the top like everybody else. This requires a transparent material that passes oxygen on the bottom of the tank, so the action takes place on the surface of that membrane.
They're vague about the details. How long does the membrane last? Is it an expensive consumable? Is the process gas air, or pure oxygen? Why do all the videos show the object being built slightly out of focus?
It seems to generate smooth surfaces nicely, but none of the examples have fine detail or sharp corners.
They've raised 40m from Sequoia, so it's reasonable to conclude there's been some degree of vetting. Given that they show video of a working prototype it's very likely this vetting included independent observation of their printer in action. Their product might be useless for some other reason, but if so it's non-obvious enough for them to have raised a non-trivial amount of money.
I don't know much about 3D printing, but that's the most futuristic video I've seen this decade. It looks like it could just pull a coffee mug or whatever out of nothing and serve it.
One thing to clarify is that this is appears to be an insightful tweak on a popular form of 3D printing, rather than a new technology all together. This uses a standard vat photopolymerization process with a DLP projector. The key difference is the oxygen permeable window that removes the detachment step between layers. This step is a common point of failure and slowness with current inverted photopolymerization printers like the Form1 and B9 Creator.
This is still a layer-by-layer process: the DLP takes a 3D object and uses a 2D projection (in both the mathematical and physical sense) per layer. Due to pixel constraints, this process will produce objects with similar resolution, although may have more organic edges instead of harder ones. I’d bet the software stack being used still slices the object into layers, so the projector still operates in a layer-by-layer fashion, and likely well below the theoretical 60 or 120 layers / second max dictated by frame-rate. The key advantage here, and it’s a big one, is speed.
It's tremendously exciting to see companies tackling the speed problem in 3DP. In the next two years, we will see a 25x improvement on print speeds from companies like HP, Carbon3D, ...
Yes, it does still slice into layers. Those layers were demonstrated in their paper as small as 1µm in thickness.
From the paper:
"Because CLIP is continuous, the refresh rate of projected images can be increased without altering print speed, ultimately allowing for smooth 3D objects with no model slicing artifacts."
I am just a bit sad that this is going to be a commercial venture protected by Intellectual Property. IP is the reason why filament extrusion is so widespread and cheap for DIYers compared to SLA.
I agree, and it is sad. Well, in 20 years when the patent expires we can finally get some progress on this tech, just like when the patents on filament extrusion expired the costs went down two orders of magnitude and print speed and quality up one order of magnitude in the space of a couple years. Until then it's going to stagnate.
20 fucking years. I'm sorry but this is so hopelessly obsolete, it is outrageous. Is there any hope for this to be universally reduced to say 5 years in the foreseeable future?
Just a point here that since two of the primary founders / inventors are on the faculty of UNC Chapel Hill, this technology likely arose from work in their labs on campus. That means it could fall under the UNC Patent Policy [1].
It would then be up to UNC to make the determination on how to best protect the IP and commercialize it. Universities have large patent budgets and do their best to protect their IP until suitable commercial partners can found to take over the patent costs and bring the technology to market.
If it weren't for patents, they likely wouldn't have spent $40million developing it. Someone would reverse engineer it and then launch a Kickstarter and sell it without having to spend the R&D cost.
you can certainly do it as DIYer but to ameliorate it and commercialize it you are going to deal with patenting. SLA patents just recently started to expire and this has already created an explosion of new developments. see here for instance http://qz.com/106483/3d-printing-will-explode-in-2014-thanks...
you may also want to check to formlabs last year, and their litigation on the patenting.
Don't worry. IP protection is only for commercial ventures. Once the details are well understood by hobbyists, you will be able to build one yourself with no worries of infringement.
That said, I'll hold my breath until somebody replicates the optically transparent - oxygen permeable membrane (I guess it is some kind of "holed" plastic or glass, with a goretex-like coating on one side).
You forgot the resin, which is probably isn't just a regular UV-activated resin but could be their own patented mixture to achieve the accurate curing time they need.
That's the harder part to DIY but I guess it's not impossible and there will be similar products soon.
The precise composition of the resin isn't specified, but the online material and methods state:
The ramp test patterns in Fig. 1C were printed with trimethylolpropane triacrylate (TMPTA) using the photoinitiator, diphenyl(2,4,6-trimethyl-benzoyl)phosphine oxide. Other objects were printed with a combination of monomers from Sartomer (CN2920 & CN981), TMPTA, and reactive diluents such as n-vinylpyrrolidone, isobornyl acrylate, and cyclohexane dimethanol di-vinyl ether. We also utilized the photoinitiators, phenylbis(2,4,6-trimethyl-benzoyl)phosphine oxide, 1-hydroxycyclohexyl phenyl ketone, and 2-benzyl-2-(dimethylamino)4morpholinobutyrophenone along with an assortment of dyes from Wikoff and Mayzo.
Well, a 10"x10" bed might be unreasonably expensive for now, but a 3" by 3" SLA printer like this has lots of potential applications, and would only cost around $75 for the Teflon AF 2400 (using your numbers).
You are. Using UV to cure photo-sensitive resin isn't the innovation, it's using the oxygen permeable membrane.
If you look at the Form 1 printer or the B9Creator, there's a mechanical step between every layer where it needs to actuate the platform in order to loosen the resin from the projection window, so it can build the next layer. In the Form 1, it peels the print off, and the B9Creator slides a window. (look for videos on youtube.)
In both instances, the amount of time spent actuating the mechanical part adds up, and results in a significant amount of time spent in the print actuating the plate. What the oxygen membrane allows us to do is to skip that step between every layer, and simply keep shining a continuous changing image slice of the object as we're pulling the object out of the resin.
Not only does this have the advantage of speeding up prints by orders of magnitude, in materials, the grain of the object influences the type of thing you can built. If you print a stress holding object with the grain orientated in the wrong direction, the part will fail very readily. This way, we have greater design freedom, without worrying about grain direction.
Is this process really better than the one used in the Peachy printer? Peachy printer uses resin tank that is filled with salt water from the bottom (which raises the level continuously, resin floats on saltwater) and the UV curing is done from the top. One huge thing for the Peachy is that you can use any tank, and the size of the print isn't limited. Also, Peachy costs just $100.
Well here you only need a small tank relative to the finished print, also, you don't move the laser head, but only the parts of the item that has already been printed. Everything else stays still, meaning reduced vibration which seems to mean a really steady quality of print.
In Peachy, both the laser head and the print are fixed in place. So it doesn't have the usual moving parts at all (although it uses modulated laser mirrors instead of a projector for cost reasons).
>> What the oxygen membrane allows us to do is to skip that step between every layer, and simply keep shining a continuous changing image slice of the object as we're pulling the object out of the resin.
What keeps the oxygen at the bottom of the tank? Why doesn't it diffuse upward and prevent curing? Why doesn't it get sucked upward and prevent curing? Are there limitations on the geometry needed to prevent oxygen from moving up?
The oxygen inhibits the curing, so it only begins to cure ~20um above the membrane, which means that it never sticks, since it never touches the window, and that can just continually pull it upwards.
SLA is layered, each layer a knife sweeps over it and refills and smooths the uncured resin layer. This is submerged, so drawing the parts upwards pulls in more fluid automatically so everything can be run continuously. What I don't get is that, hydraulically most of the liquid pulled in is going to come from the oxygenated layer. But I'm guessing that reacts with the resin and neutralizes quickly when it is too far from the diffusion membrane that provides the source of new oxygen.
They're doing the "growing" on a membrane at the bottom of the tank, instead of layer by layer at the top of the tank.
That should let them print really quickly and precisely, even compared to the SLA printers you link.
The reason it hasn't been possible in the past is that the polymerization process requires oxygen, which has only previously been available from the air over the liquid resin. Their system has some "secret sauce" for getting it to the bottom of the vat.
You've got it backwards about the role of oxygen--it is being used to inhibit polymerization in this case. They are allowing oxygen to dissolve into a thin layer at the bottom of the vat so that the build doesn't stick to the membrane, and they can have a continuous draw.
The true build layer is actually within the liquid, near the bottom, but not on it.
You're totally right. Oxygen inhibits the polymerization process. Interestingly, oxygen inhibition is also what allows things to be printed in layers (or, in this case continuously) as it allows the newly cured material to bond to the partially-cured inhibited material below it.
Ok, that is a new spin on the resin printers. I have wondered about ways to 'scan print' an entire layer rather than draw it out with a laser. Given the translucency of the uncured material the ability to project on the membrane is really cool.
And I am waaaay relieved they raised 40M in a series A rather than have this be a link to a kickstarter (which I feared) since bringing this to market isn't a kickstarter level kind of thing. Now if they can stay disciplined and not waste the $40M it could be very interesting.
The website is pretty and the copy is tight, but it's nothing pathbreaking in terms of fundamental principles.
This is just bottom-up UV-DLP SLA with a new twist: taking fuller advantage of oxygen permeable materials for the vat to eliminate the recoating step and get to continuous printing. A similar idea was tried, the ill-fated Solidator used a pressurized vat bottom with a permeable membrane. Though from what I know Solidator was not counting on the oxygen inhibition in the same manner as Carbon3d. Solidator's problems were not related to the inhibition technology, just the standard hardware issues faced by many on KS. Other companies have similar technologies as Carbon3D in production or soon to reach the market. But a very talented team and quite the splash of a product launch.
Agreed. I have a more or less exact idea of what they're doing just from the concept description as I design resin-type 3D printers at my day job. I know this concept has been tried, I think they're just the first ones to have gotten it to work.
The other thing to note is that this only works with a certain subset of polymerization reactions--not all polymerizations are oxygen inhibited so if they want to move into truly water-clear and UV stable materials this design won't work.
I really love the approach take by the Peachy Printer [1] guys. I wonder why more people aren't exploring that route (dripping system, or any water-assisted system)? It shares many of the strengths of this one but got far less attention.
I wonder how it would perform with a controllable dipping system and a DLP.
It looks like the object is drawn from the resin through adhesion to the build platform, so I'm curious about the maximum weight this method is capable of lifting. Any rough estimates?
This is going to take off like a rocket. It's genius.
Maybe some chemistry/materials engineer can figure out ahead of time how to take a polymerized object and convert it back to a usable resin. Really, we have to start analyzing the end product for recycle-ability before the tech explodes.
Agreed. This is more like what I want from 3D printing. The smelly ABS 3D printers were never really suitable for indoor use, but this I can see myself using!
Smooth curved surfaces have been a limiting factor for 3d printing in regards to optics. It sounds like this could be used for either directly printing some lenses or printing molds that could be used to create lenses depending on their level of precision. This would be awesome.
This looks similar to what FormLabs does, granted the approach is different but it reminded me of them.
I would be curious is the projection technique is limiting right now in the build size? Their prototype (or what was shown in the video) seemed relatively small.
Continuous is kind of a misleading marketing term isn't it? If it projected, it is in frames. Just like that the upward movement is in frames / steps. In essence the resolution is still as high as the resolution of the step motors moving the build plate up and it's layer by layer.
Not arguing that it is not a cool technique, just saying that they marketing it again with bogus wording.
I've seen many FDM printers do continuous, albeit vase-like, prints. Shall we market them as Continuous Filament Fuser?
Projectors and motors are analog devices. Even step motors do not move from one position to the other instantly.
In theory you could tell the build plate to go up 100 microns and tell the projector to change the image while taking into account what happens between the two stages. While the motor moves and the lights change color. So you can create different kinds of continuous transitions between the stages.
This applies to every discussion about digital/analog of course.
A "movie" depends on the tech presenting it. For typical projectors, yes. It's discrete frames.
Other displays could vary! A display capable of incrementally updating the picture could show motion as a series of changes without there actually being a frame, just deltas... Done quickly, this would approximate what the human eye does.
In the context of this process, "frame" would refer to the changes to the projected image. Each of those changes would be a "frame"
But, if the object being rendered is actually in motion during the cure, there will be interpolation between those "frames", resulting in a very analog like product.
The motion would be "frames" too, as each micro-step would presumably be a controlled atomic thing, but those movements would not necessarily need to be keyed to the changes in the projected image.
Take both of those up in terms of rates and precision, and it's all going to blend together, particularly as both push the material to it's change rate limits.
Think "motion blur" when motion exceeds capture rates, or in the case of analog film, where reality "smears" onto the film while a shutter is open.
This looks amazing though I'm not really sure it is the speed that has been the thing holding back 3D printers, although that is certainly an aspect of it.
I still think the biggest thing stopping adoption is most people not having any idea what to print on a 3D printer, even if buying, operating, calibrating and maintaining one was cheap and easy.
Speed is a very big market impediment for 3D printing.
You can't mass deploy printing kiosks at consumer stores if it takes four hours to make a mug or a trinket (you can, but it's absurd). If it takes five minutes you can and people will buy all sorts of custom products that way.
Someone on Reddit mentioned it taking 20 hours to print half a skull. That's ridiculous. This will do it in probably 20 or 30 minutes, and it'll likely get faster with improvements in the next couple of years.
20 hours to print a plastic skull you have just designed on a kit printer is still quick and cheap enough to feel amazing. Most small objects I do take about an hour or so and it is still cool to have drawn something and then go off and watch something on tv and come back to a working part.
I agree that having a massive increase in speed is cool, but the materials range for UV cured stuff isn't that high. If we are talking wishlists, I'd rather have plastics and metals in one print and be taking a few hours, than a UV cured object in minutes, but that is because of the applications I am interested in. I do think this tech is very very cool though, especially given the detail level you can get at that speed.
I know this is probably the wrong comment to hang this on, but I was just thinking about it reading your mixed mode comment.
I can see shipping container type units fitted internally with multiple 3D printer types that are basically little mobile factories. Any part you need (withing certain volume and material limitations) can be produced on demand. These could easily be transported to remote locations to fully support all kinds of activities.
By the way, they're hiring. And they're looking mostly for software people, who are steeped in web tech. They have some very interesting and challenging ideas for what they want to build on that front!
I got to meet the software side of the Carbon3D team (and printer) not too long ago, and came away extremely impressed.
Wow, that's an impressive presentation. I'm interested in the tech behind that website; can anyone suggest how it was done? I see several elements that are in vogue:
Scroll down slightly to reveal top menu bar
Animation at the top
Scroll down and a left frame of animation appears, while allowing scrollable text on the right.
Scroll down more, left animation disappears, and we find an embedded Youtube video.
Where can one read more about this type of design, such as online tutorials and the like?
Whoa - pretty awesome. I can see CLIP tech being really practical for soft robotics (print the actuators directly rather than casting the elastomer from a 3D-printed mold)
Is there anything on how cheaply parts can be produced? They suggest that it could be used for production parts --presumably ones that can't be produced by other processes -- and that's great, but what about for inexpensive one-offs, whether they be hobbyists/inventors/mechanical hackers making stuff, or for things that are highly customized (say earphones that are shaped exactly to your ear).
Printing from top to down from a plate is a much better approach because it works faster. The whole layer can then cure instead of only a single point. This seems to be the main difference to conventional 3d printers.
The website is inaccurate in some aspects but this new top down approach could change the 3D printing industry. Maybe it could also make 3d printers cheaper because less mechanical parts are needed.
A few years ago I 3D printed a head I sculpted in Maya, and took a video of the print. [1] Seeing the video of the Carbon3D in action makes me want to buy one just so I can make a cooler video :)
This is great! The point the website makes that really intrigues me is the vast amount of materials this technique opens up to 3-D printing. I wonder if this tech will ever make it into the hands of the consumer.
Don't you get the same effect (of a 'dead zone') by printing immersing the object rather than pulling the object from the pool of material? The top layer is exposed to the air directly.
The oxygen enters the chamber from the permeable membrane that makes up the tank base, and creates a liquid-polymer barrier between the solidifying elements and the bottom of the tank. This means that the work doesn't fuse to the tank bottom, and can therefore be built not in a layered(1) fashion, but continuously. Therefore, both faster and (hypothetically) smoother.
(1) I'm sure there will still be some directional effects in the result---the light is only coming from one direction, for example. But you definitely shouldn't expect to see the "sandwich stacks" effect you get from extrusion-filament 3D printing.
at the bottom of the page there is a video -- it looks like they cure the resin to make a thin layer across the whole plunger/panel, at the very least there is the small layered grid that holds it on before they start "printing" the object. seems like cured resin will stick to the panel, so they just have to do some simple layered curing before they engage the actual object
What is the precision and how small can an object be? How small can the printer be? Can we make nanoprinters that are controlled from the cloud to mutate the models they create as necessary?
Nanovaccines that can change in response to mutating threats...
I feel like I have understood the importance of 3D printing for the very first time which makes sense, I have tended to be a little slow.
It's interesting that you ask about nanovaccines in relation to this article. The lead inventor and CEO on this project also founded another company that works in that exact area [1].
As to the precision of Carbon3D, the best I can figure based on their recent publication is printed layers as small as 1µm. That's pretty darn good resolution for a 3D printer.
I found this whole thread very exciting, I as down-voted into oblivion very quickly so I stopped sharing my thoughts but my mind ran on for a while...
Self-healing machines, buildings and devices also sprang to mind.
Basically, if the printer is small enough to be a part of the object and there were a way of determining what it needs to print then the object need never be broken (for very long) in fact: the object doesn't even need to be defined as a cup, a pair of trousers, a bicycle. If the printer can embed itself into anything it prints then it could literally morph according to a given requirement.
It will apparently be the cover story. (Does this count as Sequoia breaking Science's embargo?): https://twitter.com/sequoia/status/577651625545748480
From the abstract: "...feature resolution below 100 micrometers. ... complex solid parts can be drawn out of the resin at rates of hundreds of millimeters per hour."
Key novel feature is the oxygen-permeable, UV-transmitting membrane at the bottom of the tank that creates a thin (down to 20um) inhibited 'dead zone' where the resin can't polymerize. They project the image for the current layer up through the membrane and the dead zone, so the build layer is actually within the tank. This means they can just draw the part up continuously from the top, with no stepping or processing needed after each layer. The thickness of the effective build layer can be controlled by adding a UV light-absorbing dye to the resin, which allows them to optimize for different print speeds.
Here's a relevant patent, issued in 2014, listing the 3 founders of carbon3D (formerly EIPI systems): http://www.freepatentsonline.com/20140361463.pdf
And... founder Joe Desimone also gave a Ted Talk tonight, so soon we'll even get to see a splashy 18-minute long talk about the technology. https://conferences.ted.com/TED2015/program/schedule.php