57 comments

[ 3.4 ms ] story [ 99.3 ms ] thread
Sounds interesting; any details on how it works, what materials it can be used with, what are the consumables, strength, accuracy/resolution, maximum size, etc..?
Here's a little more detail: https://www.theverge.com/circuitbreaker/2017/4/26/15440744/d...

The important thing to remember is that the 3D printing doesn't generate the final part, the plastic binder in the filament has to be burned out and then the part has to "densify", e.g. the actual sintering where the metal particles fuse together creates the final material properties of the part.

The final part will be a lot smaller than the thing coming out of the 3D printer, controlling uniform shrinkage is probably a lot of the secret sauce.

Does it actually has to go trough forging and tempering? They don't have a fusing element in the head?
It goes through two additional steps: burning out the plastic and then putting the part into a furnace to get the metal particles to "fuse" together more closely, aka sintering.

The printing part just forms a "green" part, it's basically like you would make pottery, the actual strengthening of the part happens in a high-temperature furnace.

Looks like fused jet deposited powdered metal printing.

Strong but not nearly as strong as casting (especially without secondary processing), as for materials powdered metal printing is usually either ferrite alloys or nobel metals the former is much more common.

Resolution is in single digit microns or finer.

This is based on similar products not this specific one, it might be faster which is important but there is only that much you can improve on the process.

The speed also to me looks like a result of very limited market in terms of competitiveness. Existing products are few and they rather sell you more machines than drastically improve the printing speed by making it much smaller and focusing on smaller parts.

I don't know if they use induction or optical fusing based on the size I would go for optical so they might have found a good laser or a absorption facilitating epoxy/flux to go with the powder.

Edit: based on a comment here it looks like they don't do fusing as part of the print.

In this case its metal powder with a chemical binder (glue) which usually contains flux and vaporizes at a fairly low temperature.

After the printing the parts go into an oven (usually encased in sand for temperature control)

The oven brings the part to a high enough temperature for the crystal formation but not close to the melting point.

In this process larger crystals can form but not nearly as large as with cast forging.

I don't see the advantages to this over lost-PLA casting. Am I missing something?
That's what I was wondering. Shapeways makes metal parts by a similar process as a service. See their 2009 video.[1] There's a mini-industry making metal jewelry that way.

I notice that Desktop Metal doesn't show their oven in the video. That's the big, expensive, somewhat hazardous unit. "Easily swappable aluminum gas canisters ... safe to use on the shop floor" says the data sheet.[2]

Wazer, the low-end water jet cutter that's still stuck at "pre-order", has a similar problem. They're trying to make a somewhat messy industrial process less messy. There's some hand-waving involved. You end up with a sludge that's a mixture of water, shattered garnet, and whatever you're cutting. They gloss over what you have to do to dispose of that.

[1] https://www.youtube.com/watch?v=B9VOwqtOglg [2] https://desktopmetal.s3.amazonaws.com/Studio-office_friendly...

I was pricing out metal casting after I watched this video:

https://www.youtube.com/watch?v=a-FHY5FNsWc

12 minutes uncut. They go from printed model to working cast piece.

All new, Kaya Cast vac chamber they use is $1k, the Paragon kiln is $800, and the Electro-melt furnace is $350. Add in safety gear and incidentals and you are looking at 3-4 thousand dollars to go from a 3d print to a cast metal piece. Heck, depending on the complexity, the vac may be optional.

Hot metal casting on a small scale isn't complicated or expensive, just messy. Desktop Metal is trying to do this in an office environment. Because nobody has a shop any more.

(The "maker movement" may have run its course. TechShop seems to be in financial trouble.[1])

[1] https://www.bogleheads.org/forum/viewtopic.php?t=177391

Expertise and materials.

Casting aluminum is easy, but not very useful. Doing this yourself you can't guarantee grain structure/size, inclusions/porosity, or temper. Plus it's usually plenty easy to mill from billet since aluminum is so soft.

Gold/silver is also relatively easy. You need better equipment since the temperatures are 2x higher. Strength isn't important, so casting stuff in gold/silver is quite common. Outside of jewelry there is basically no need for it.

Casting something more complex like good steels, or platinum, or nickel alloys- casting will be very difficult. Molds aren't so trivial, you need preheats, and a mistake leads to liquid metal on the floor or an explosion.

Sintering can be used with challenging alloys and requires almost no expertise. Everything comes certified and gets heat treated in the oven. If your time is valuable enough to warrant a metal 3d printer, it's valuable enough to make learning an entire discipline too expensive.

For $360k you could just hire a materials scientist for three years, and give them $20k/year to play with.
With product development times of 6-12 months, it doesn't make sense to pay 360k to get results in 3 years if you can pay 360k and get results in few months.
They claim a resolution of 50 microns which is far removed from 'single digit' microns, and generally too much for precision mechanics.
Waiting for the day that printer will be able to print copies of itself.
Even if you can print all the structural portions, there's a big problem printing electrical components e.g. motors, and then an even bigger problem printing control electronics...

Although, I think if they end up able to print robust, ready-to-use metal components with reasonable strength (say strong enough for a bicycle frame), that would be a pretty good leap forward. The base tech of what you can do on a desktop doesn't seem like it would quite get there today though.

I would think the next logical step (after metal printing comes down in price) will be to start printing mixtures of metals and plastics, which would eventually lead to being able to print more complicated things such as motors.
Bearings and round cylinder things need to be really precise.
I don't understand why this hope is still prevalent.

3D printers are just another tool in the manufacturing toolbox. The sets of materials and structures printers are made from barely overlap with the sets of materials and structures that the same printers make.

If 3D printers could make themselves, I think they would look a lot more like biological organisms and a lot less like laser copiers.

That said, the Desktop Metal parts I've seen are beautiful.

People think that the only thing standing between them and designing cool new useful things is the final "model of thing -> thing itself" transformation, and that if they just had a magic machine to do this automatically, the rest would be easy.

Much like people think that the only thing standing between them and designing cool new useful software is all that pesky syntax, and that if they had some magic application to turn pictures into software, the rest would be easy.

Maybe it is already here... we just call it "life"! Hmmm... maybe that is life is - someone's printer?
Anyone know who the primary customers for these will be?
Any engineering shop that's prototyping and possible creating production products. I got to see the prints first hand, pretty solid stuff. Sadly they hadn't printed a T-800 terminator yet, but give it time. Skynet is alive!
(comment deleted)
For a nice CGI movie on their high volume model:

https://www.desktopmetal.com/products/production/

Very impressive video. Questions that page completely doesn't answer, though: How much does the system cost? ($5k to "reserve", but what's the final cost? Their "downloads" page, of all places, has a price sheet for the "studio" system, but not the "production" system.) What does the software workflow involve, from CAD software to printing?

Also disappointing that, despite having an innovative piece of hardware, they still treat the software as heavily proprietary as well.

> How much does the system cost?

Follow the reserve page through two more clicks.

Studio system is $120,000

Production system is $360,000

The first is printer, debinder and furnace combined. So close to a total cost. The latter is printer only, so think of that figure as a secondary deposit.

> despite having an innovative piece of hardware, they still treat the software as heavily proprietary as well.

My client created an innovative piece of hardware without investing in proprietary software. In a couple of months, he found cheaper clones offered on aliexpress and taobao. While the design was good, the product failed commercially. Then the company found me, I’ve created heavily proprietary software for the next version of their hardware. The product sells OK so far.

In addition, creating software from scratch specifically for the given hardware delivers better overall value for the end users.

If I may, what was the product and what is within the proprietary software which differs it from the clones? Because given China manufacturing scale it is given that every other thing will have a clone on aliexpress or taobao.
A 3D printer.

On the very high level, I’ve implemented a small subset of what Autodesk Netfabb premium does, the subset relevant to the hardware my client is offering. That Autodesk software costs $500/year, I guess that’s what differs our product from the clones: a clone would cost $500/year extra.

Another thing, 3D printers are too different, they use many different technologies (FDM, SLA, DLP, SLS, etc.), and within each there’re different printing materials. These things are very relevant to e.g. slicing and support generation. Compared to the software like Netfabb which has to support all of them, our software only supports a single printer and a carefully selected set of printing materials. That’s why, hopefully, we were able to deliver better overall UX.

> In addition, creating software from scratch specifically for the given hardware delivers better overall value for the end users.

That much makes perfect sense, and I would imagine that something as novel as a new metal 3D printer like this would want custom software to drive it as well. That doesn't prevent that custom software from being Open Source. And when you're talking about a product like this, in this price range, designed to support manufacturing with very tight tolerances, clones (and adoption of such) seem quite unlikely, even less so if they keep innovating with it. I'd rather see the "stickiness" provided by support, rather than software.

Because inevitably, software written by a single vendor for their hardware always feels very particular to their workflow, not integrating with anything they didn't anticipate, and not allowing much flexibility. And CAD workflows are exactly the kind of area where everyone has their own set of customizations they want to be able to do. I'd like an Open Source library and an API, not an opaque kiosk.

> I'd rather see the "stickiness" provided by support, rather than software.

I’d rather see the product just works not requiring much support. Ideally, not requiring any support at all. This works OK for consumer electronics. I don’t see why industrial hardware can’t be designed the same way.

> not integrating with anything they didn't anticipate

Desktop software often solves the interop problem by supporting open file formats on input. For 3D printing, the de-facto one is STL. You can export this format from every CAD or 3D modeling software.

> I'd like an Open Source library and an API, not an opaque kiosk.

I prefer opaque kiosks. Opaque kiosks just do their job. Good ones do it well. Libraries and APIs require substantial amount of time to RTFM, integrate, debug integration, update integration with API updates, and so on.

Also, not everyone is a programmer. I can use APIs and libraries when I need to, but most people just can’t. For an average Joe, these things are useless. Even harmful: creating good libraries and APIs has costs (documentation, examples, developers support); these resources might have been spent on useful features instead. The equation changes if the product attracts large user base and third-party developers start using these APIs to create value. But market for 3D printers is currently very small, i.e. this is unlikely to happen.

That's the kind of HQ video that gets you that sort of funding.
Jamison Go reported that his beetleweight fighting robot, Silent Spring, used a drum weapon made using a Desktop Metal system. The drum was printed using 4140 steel, then hardened the traditional way.

It apparently yielded very good results, winning the beetleweight category at Bot Blast. You can search YouTube for videos of it in action.

I am really surprised to see Jamison being mentioned here as if he were a household name. I remember him from the Invention Studio at Georgia Tech.
He's well known among people who are into that sort of thing.
i <3 jamison hes so dreamy
Yeah, but nothing compared to Charles Guan, or so I've heard.
I heard his cat ears could win battlebots on their own, but he likes to give the competition a chance
What 3D printer or service -- in any material, metal, wood, plastic, glass, anything -- can print something affordably i.e. $5 for something that's say 1" by 5" by 5" ???
So where are they at right now in terms of progress? R&D? Prototype built?

Why didn't somebody already build something like this?

I wonder if they can stop casting porosity. There's a potential to save millions.
This machine is definitely some extremely exciting stuff! Most companies I work for (I'm in the US) unfortunately does not have much of a machine shop any more. They either got closed down during the recession or the machinist had to leave for health/age reasons and the role never got fulfilled again. No one wants to have to setup a machine shop in their building since it takes a lot of money and logistics. According to management, "It's not worth it".

That being said, I am not sure how competitive the production unit will be. A lot of the commercial industry does not necessary care about weight reduction or aesthetics. I can't foresee that the parts made by these machines will be anymore cheaper than traditional sintering or more accurate than milling.

Maybe im just a debbie downer but I don't see how these could possibly compete with metal sintering and traditional milling and CNC work. An inferior metal alloy with a huge price tag that will require machining anyways.