More service bureaus

In a couple of recent postings I have talked about the Ponoko laser-cutting service. Another very interesting online service bureau is Emachineshop.com which is a machine shop to which you send design files created with their downloadable CAD software. They have some examples of the things people have made. Just poking around their website really makes me drool a little bit. I can't believe I'm not doing something with this.

Big Blue Saw is a service bureau that does waterjet cutting of metal and plastic. They cut very thick pieces of metal, which surprises me, I didn't know you could do that.

A couple more: Fabjectory.com specializes primarily in making physical copies of avatars from games like SecondLife. FluidForms makes pretty flowing shapes for things like vases and pitchers. I haven't read about these yet, and as of this writing I don't know what technology they use, or what design software.

Broadening the definition of "fabber"

I want to broaden the scope of this blog a bit. The word "fabber" is generally accepted as synonymous with "3D printer" but a 3D printer has a lot in common with both CNC machines (routers for wood or milling machines for metal) and laser cutters. There are hobbyists building all of these. All of them make a 2D or 3D shape under computer control with relatively little human intervention, and minimal need for human skill.

How many of these gadgets could be self-replicative in the RepRap sense? For example, could one use a laser cutter (or a laser-cutting service like Ponoko) to cut out pieces and use those pieces to build another laser cutter, thereby driving down the cost of laser cutters? As with RepRap there will inevitably be complicated pieces that can't be made that way. CO₂ lasers are dangerous and expensive, so I don't think this could make the kind of impact in the developing world that RepRap hopes to make. A replicating CNC machine might be a better bet, as Dremel tools are much cheaper and safer than lasers.

That self-replicative idea does fascinate me a good deal. It will, over time, drive down the price of the self-replicating thing. That doesn't mean we'll enter a microeconomic paradise, but it promises at least to be interesting and possibly to raise the quality of life noticeably.

I've haven't blogged too much about commercial machines. I want to do more of that. I admire the hobbyists and their perseverance in the face of difficulties, but the technology appearing in commercial machines will gradually trickle down into the hobbyist arena as patents expire.

RepRap parts available via Ponoko

Ponoko is a very cool on-line laser cutter fabrication service with a wide range of available materials. The idea is that you create a EPS file for the laser cutter to follow, specify the material, and they cut out the pieces and ship them to you. The laser can also engrave lines on the material. EPS files can be generated with Adobe Illustrator or various other similar 2D artwork programs. If you want to make a 3D project (like this table), you make it out of 2D pieces that fit together with slots and grooves. When you upload your EPS file and choice of materials, they figure out how much the laser cutting fee will cost.

Toby Borland (of SMARTlab in the U.K.) has designed a set of laser-cut plywood RepRap parts and made the EPS files available on the Ponoko website. There is a Flickr photo set showing laser-cut RepRap parts and the process of assembling them; I am not sure that's the same Ponoko files and process, or another laser-cutting effort, but it gives you a sense of what's involved, and the level of complexity.

Too-brief overview of DNA nanotechnology

A lot of interesting work has been done with DNA nanotechnology, much of it starting with Nadrian Seeman's work on DNA polyhedra in the mid-90s (1, 2).

Around 2000, Andrew Turberfield (Oxford University's Department of Physics) used DNA to make tweezers, with arms 7 nanometers long.

"Of course it's all very speculative," said Dr Turberfield, "but you can imagine, for instance, little factories on chips doing chemistry or simple assembly. You can think of production lines made up of little motors with different reactants being passed from one place to the next."

Things got really interesting in March 2006 with Paul Rothemund's DNA origami technique. Here is the publication. I was working at Nanorex at that time, and we were all quite excited about it.

In 2005 Turberfield and colleagues described a family of DNA tetrahedra consisting of triangles of DNA helices covalently joined at the vertices to form a mechanically rigid 3D structure. This image of a reduced model of one structure, which is less than 10 nanometers on a side, was created using NanoEngineer-1 Alpha 9. The bowing of the DNA helices is pronounced in this rendering and is the result of electrostatic potential terms included in the customized molecular-mechanics-like force field developed by Dr. K. Eric Drexler specifically for DNA structures. Regarding Turberfield's work, New Scientist wrote:

Now Andrew Turberfield [et al] have shown how carefully crafted DNA structures can be made to self assemble and change shape when sent specific DNA signals. The researchers built tetrahedrons ... using four short DNA "struts" that join at each end. The process exploits the way DNA is held together by complementary bases that form the rungs of a ladder-like structure ... the researchers made cages with two extendible struts that could be independently controlled using different DNA sequences. In theory, it should be possible to create cages in which every strut can be controlled independently, Tuberfield says.

These cages are a combination of support material and linear motor, and with the many other DNA tricks being done, they should allow people to build large, complicated, reasonably rigid 3D structures that have controllable moving parts. So this is a very promising development.

A very recent announcement of work by Chad Mirkin and colleagues. They have found a way to use DNA to glue together arbitrary arrangements of teeny gold spheres. People have known for some time now how to make DNA stick to gold spheres, and by careful selection of DNA sequences, Mirkin et al can position groups of spheres in almost any 3D configuration they want.

In light of these developments, Nanorex has narrowed its focus from "general" nanotechnology (anything one might build from common small molecules) to structural DNA nanotechnology. This is likely to be where much progress will occur in the next five years or so. I hope Nanorex will still be around after that, and will be in a good position to shift gears as we move beyond DNA to more general chemistry.

Those commercial 3D printers sure are gittin' purty

Some commercial 3D printers are very pretty. This one prints in colored plastic and is intended to create prototypes in a few hours that can be shown to managers or customers. The claimed resolution of this thing (presumably in all three dimensions) is 450 dots per inch. Drool.

In twenty years, all the patents for this printer will have expired, and it will be possible for hobbyists to make such pretty stuff at such high resolution. Hmm, thinking more about that inclines me to start an economics blog, since I blog about economics a lot elsewhere.

An XYZ platform for fabbing or CNC

I was watching an auction for a CNC XYZ table on eBay that went for $300, item number 200198037915. I would have bid on it if the Z travel hadn't been only 2 inches. It was built from plans from hobbycnc.com and didn't have stepper motors or the machining tool but was otherwise complete. I felt lust in my heart, but that itty bitty Z travel bugged me, so I thought about what could be done to increase it. Here's my general idea.



My hope is that the blue-hatched stage can be made to take either a Dremel tool for CNC milling, or an extruder for fabbing. The result might or might not be self-replicative in a RepRap sense but it would be a cool toy.

RepRap is now half-way to replication

Vik Olliver has made good progress (1, 2, 3) on the goal of self-replication for the RepRap, having now been able to use a RepRap to fabricate half the RepRap's parts.

It's interesting that you can see the size of the volume pixels Vik is working with. These pieces were printed with polylactic acid, I believe.

Unrelated but cool: Kovio is a non-hobbyist company working on a process to inexpensively print working transistors. Early applications will include smart cards, later you'll see wall-sized displays.


Also unrelated but also cool: Fernando Muñiz has been working with UV-cured resins. This will work a bit like the CandyFab, except the uncured resin is still a liquid so under-support structures are still required. Interesting, I'm not sure if it's better or worse than the FDM approach used by RepRap, Fab@Home, and Tommelise. Also, I don't have any idea how environmentally benign these resins are; it's hard to imagine they're as green as polylactic acid.

Big fabbers: houses, boats, factories

How big could a fabber get? Could a fabber build a boat or a house? Here are two big CNC machines, one of which is claimed to do work on boat hulls.




A fabber placing individual drops of building material would be awfully slow for a very large project. One work-around would be to trade away spatial resolution, and let the fabber lay down big handfuls of wet concrete.

Maybe you'd want many fabbers feeding small pieces to an assembler that assembles them into bigger pieces. The assembler must be able to make the small pieces stick together, by gluing them or melting the sides or by using mechanical fasteners such as screws or nuts and bolts. It's possible that the big pieces might then be assembled into very big pieces, and again an assembling machine must be fed from many sources. The assembler would need to be very smart to recognize and correct assembly errors, and would probably need machine vision. This would work well for products from a factory, but might be unsuitable for a house.

A google search for "robot bricklayer" turns up a few modest research efforts. I would have imagined something like the big XYZ stage above with a brick-lifting robot arm, wheeled into position over the site of the future house, but the most advanced effort visible on the web is a standard industrial robot arm picking up bricks instead of doing whatever else robot arms normally do. The arm can't move around the entire volume of the future house, it's not on any sort of XYZ stage, it's just bolted to the floor like any other industrial robot arm. So robotic house construction is still quite a ways off.

Fabbers as tissue engineering tools

In late 2006, Gabor Forgacs and other researchers pioneered tissue engineering techniques using 3d printers. This obviously is not a hobbyist application but it's very interesting, and could save thousands of lives. The technique involves alternate layers of "biopaper" and "bioink", the former being a temporary scaffolding gel and the latter being a suspension of adult stem cells.

• Gizmodo
  
• New Scientist (Dec 2006)
  
• New Scientist (Nov 2006)

Tommelise project

Tommelise is Forrest Higg's attempt to build his own RepRap-like gadget before RepRap itself is ready for wide distribution. He has spoken at an O'Reilly conference about the RepRap project and he has some fascinating ideas about architecture and how fabbers might relate to it. The recent (early Feb 2008) postings in his Tommelise blog describe his success in connecting stepper motor axles to threaded rods, something I had been wondering about myself.

In the Tommelise FAQ, Higgs mentions Linux and Java (which have been adopted by the RepRap project) as presenting a steep learning curve to people without a software background, citing Microsoft Windows and Visual Basic as more user-friendly alternatives. My own early experiences with Linux required enormous patience. Higgs writes Tommelise has been created for people who aren't particularly clever and may be living in modest circumstances. Any open-source "fabber revolution" (1, 2, 3) will be an empty exercise if it fails to serve such people. Then again, if a genuinely open-source revolution is to occur, we'll eventually need to wean ourselves from Microsoft and make our own tools equally user-friendly.

Interesting Russian project

Here's the website in Russian, and a Google translation to English (click on the "Constructor Kulibin" link). I found this referenced from MAKE Magazine. This is a very interesting project.
They have a great-looking XYZ stage built from a CNC kit. They lower a heating element onto powdered raw material, sintering the raw material as the CandyFab does, except their heating element is a length of nichrome wire instead of a jet of hot air. It gets hot enough to glow, and on the web page they mention that they can work with any powdered material with a melting temperature from 100 to 300 Celsius, including sugar, wax, "Plexi" (plexiglass?), and mixtures such as plastic and sand, plastic and metal powder, powdered paint and sugar powder. Like the CandyFab, each layer of fresh material is laid down on top of the previously worked layer (and I hope that process is automatic as it sounds tedious otherwise) and then you scribble a cross-section on the new layer with the heating element, and then it's time to put down another layer.

The nice thing to this kind of approach is that the unmelted/unfused material provides mechanical support for the built structure. You can build shapes that RepRap and Fab@Home can't make, such as bridges or inverted cones, because any bridge-like part that will go over empty space is built with stuff under it to support it.

This made me curious to start looking around at CNC kits, which could nicely jump-start any fabber project. The XYZ machinery for a fabber is called a "gantry" in CNC language, and there is a very active hobbyist CNC community. Here is a video for a CNC gantry kit that somebody was selling for $195 on eBay. The video itself is for sale ($20) so this is just a teaser.

Here's a few interesting CNC links.

• HobbyCNC.com: CNC router plans for $25
  
• ShopBotTools.com: a very nice gadget but too expensive
  
• MadVacCNC: a 4-foot-by-8-foot CNC table
  
• SolSylvia.com: more CNC router plans, I want this one
  

Make Controller Kit

The Make Controller Kit is a pair of boards that snap together, with a AT91SAM7X256 as the microcontroller. For communication it has USB, Ethernet, CAN, and JTAG, and supports the Open Sound Control protocol, which has interfaces for numerous programming and scripting languages. It offers eight analog inputs with 10 bits of precision over a 0-to-3.3-volt range. There are eight high-current digital outputs that can drive relays or two stepper motors or possibly solenoids. There are four servo controllers. It has an 8-position dipswitch. It costs $109 at the on-line store for Make Magazine.

CandyFab: The Revolution will be Caramelized

Here's a PDF presentation about the CandyFab, an unusual approach to a hobbyist fabber. These guys decided they could relax their spatial resolution in favor of large volume and cheap fabbing materials. Their fabbing material of choice is sugar, much cheaper than the plastics used by RepRap. Some of the things they make are considerably larger than can be made with RepRap. Here's a Flickr photo set.

In September the CandyFabbers came up with a better hot air nozzle that gives them considerably improved spatial resolution, with volume pixels of about 1/16 inch instead of 1/5 inch.

Some MIT folks have built a fabber that makes stuff out of pasta dough.

How does one get started?

How does one start to build one's own fabber? RepRap and Fab@Home both offer instructions. There is of course the caveat that the technology is new and experimental and bleeding-edge, so it's not a shrink-wrapped thing where you simply tear open the packaging and start using it. My goal in this posting is mostly to decide whether it makes sense for me to start work on a fabber. My early conclusion is that I'd like the field to mature a little bit more, but it might be fun to tinker with just the 3-axis motion part (check out the video), probably using this microcontroller board.

The RepRap folks have a page about constructing their version 1.0 fabber, called "Darwin". They recommend that you join the RepRap Research Foundation, which supports new fabber builders, and you can purchase parts from their on-line store.

Fab@Home has a Getting Started page with links to their catalog and the list of materials that you can fab with. A pre-assembled Fab@Home fabber will set you back about $3600 plus shipping, currently with a 6-to-8 week lead time, so I guess people are buying them. The Fab@Home is an impressive thing, and good looking.

Hobbyist fabbers today look the way Linux did in 1993. In five or ten years fabbers will be much more common and much more polished, but the people tinkering today will have 99% of the fun. Linux in 1993 was not at all user friendly, everything needed to be hand-tweaked, and you needed to understand a lot of it to use any of it, and the same was true with cars in 1910, and with fabbers now.

Brilliant RepRap video (thanks to Emeka Okafor)

I am deeply indebted to Emeka Okafor, author of the Timbuktu Chronicles blog and director for the TEDGlobal 2007 conference in Tanzania, for stumbling across this brilliant Poptech video of Professor Adrian Bowyer, the inventor of the RepRap fabber. I would also like to thank Mr. Okafor for giving attribution to my nanotechnology blog, and call attention to his postings on technologies that can help Africa and other developing regions. There is an Emeka Okafor who plays basketball, I'm not sure if it's the same guy.



Bowyer talks about the economics behind the project, particularly its ability to empower communities that are now economically depressed. There is some yummy game-theory stuff in the paper linked here that does not get mentioned in the video, check it out. He also talks about using polylactic acid (wikipedia) as a printing material for the RepRap. This is significant because you can make PLA from starchy vegetables like potatoes and corn, and when you're finished using your PLA object, you can compost it to help grow next year's crop of starchy vegetables. You can have a closed-loop local manufacturing economy that doesn't require trucks or trains or ships to move products around. In fact there are several materials under consideration, and thought has been given to printing a single product from multiple materials. The Fab@Home folks also have an impressive list of materials that can be fabbed, including chocolate.

I got curious about PLA and did a little googling. In a RepRap forum there is a discussion of just how easy it is to turn starchy vegetables into PLA. From the sound of it, it is non-trivial and demands that the person attempting it be quite knowledgeable. One person compares "home PLA production today to home biodesiel production 20 years ago, when it was arcane, a little dangerous, and rare, but theoretically possible" and notes that for many people it will simply not be economical compared to mail-ordering some PLA. I found a place that sells utensils, plates, and cups made from PLA. NatureWorks appears to be a source for PLA in ready-to-work form.

3D printer in a knick-knack store

Make Magazine has a note about an Umbra concept store in Toronto which now has a 3D printer (some people also call them "fabbers"). The store can use it to fabricate novel items, and the store chain designers visit the store to create and fabricate designs while chatting with customers about the process. The little white widgets to the left of the printer are some of its products.



I'm interested in 3D printers, but I haven't dedicated the time to build my own, as some people have started to do. It's intriguing to imagine what 3D printers might accomplish in combination with automated design techniques such as genetic algorithms (here are some more GA links).

At the present time, 3D printers are the closest things to real nanofactories, and they present limited versions of many of the same challenges that nanofactories will bring, such as copyright issues and the bumpy ride toward a post-scarcity economy.

Starting a fabber blog

Lately I've been thinking and posting a bit about fabbers (also called 3D printers), primarily on my nanotechnology blog. I think the topic (and my growing interest in it) is rich enough to deserve its own blog. I am particularly interested in affordable hobbyist fabber projects, something I might be able to fool around with myself.

The fabber idea is pretty simple. Take a hot glue gun and three stepper motors. Use the stepper motors under computer control (with appropriate mechanics) to position the hot glue gun at a specific XYZ point, and deposit a drop of hot glue. The glue cools and you move to the next XYZ point. Use this arrangement to draw a glue pattern on a horizontal surface, then move up a little bit and draw the next layer, and then the next. Soon you've got a 3D object of almost any shape you wish. A few of the details can vary -- it's not really glue, it's typically a polymer like polylactic acid -- but that's the basic idea.

There are professional and industrial fabbers with prices starting at about $50,000. But more interestingly, there are hobbyist projects to build much more affordable fabbers. The two currently prominent hobbyist efforts are the RepRap project (wikipedia entry) started by Adrian Bowyer at the University of Bath in the UK and the Fab@Home project started by Hod Lipson at Cornell. There are others but these two have the highest visibility and, as far as I can tell, the largest numbers of participants.

The Fab@Home fabber looks more polished than the RepRap, but I find the RepRap more interesting. Partly because it's more affordable (a getting-starting price somewhere around $400 versus $2300) but also because Bowyer is more committed to an open-source approach and is more interested in the implications of that approach. He very intentionally designed a machine that could fabricate most of its own parts and could therefore mostly copy itself. If the machine becomes popular, its price will quickly drop (building one today might cost a good deal more than $400 and a very large investment of tinkering time) to roughly the price of the few non-copyable parts and the raw plastic for the rest.

Videos and links, RepRap and Fab@Home

Since I've been writing a lot about fabbers lately, I've decided to start a fabber blog and start migrating my fabber postings over to it, starting with this one. Fabbers are only peripherally related to advanced nanotechnology (the economics look similar) and I'd like the fabber blog to go into a level of detail that's not appropriate here.

As far as economic similarities, a fabber looks a lot like a crude nanofactory, and raises many of the same societal concerns but in a smaller, safer way. One of the popular speculations about mature nanotechnology goes like this: (1) sufficiently advanced nanofactories will be able to make almost any desired product from materials found in nature, so (2) the price of physical goods drops to nearly zero, and then (3) money ceases to exist and we all live in a post-scarcity society free of poverty, disease, and war.

It's an appealing simple notion, probably too simple. Even when the necessities of life are available essentially for free, humans always envy other humans and there will still be a premium to pay for things beyond the survival level. Economic demand will exist as long as we're still human, and money will too. Besides, physical goods aren't the only things we spend money on. I can imagine a robot bus driver at some future time, but a robot doctor seems a long way off, and it's hard to imagine the board of directors that will appoint the first robot CEO.

Brilliant RepRap video (thanks to Emeka Okafor)

I've moved this posting over to my new Fabber blog.

Xilinx Spartan 3 FPGA eval board

I was thinking I wanted to do something with the ARM91SAM7 eval board and a Xilinx FPGA, and I discovered that Xilinx also sells a eval board for their Spartan-3 FPGAs for $150. Here's the user guide (PDF). Big surprise, it's available at Digikey.

The Spartan-3 series has hardware multipliers, blocks of RAM, and "distributed RAM bits" which you can spread around your design as needed. Pretty sweet. As I mentioned in an earlier posting, there are readily available HDL tools for FPGA design on a Linux platform. Hardware design shouldn't necessitate the indignities of Windows.

Here is a ridiculously affordable SAM7 board that I need to investigate. It lacks RS-232 level converters, and you need to kludge the JTAG stuff (probably drive it with a parallel port somehow or other). Also interesting are these two open-source simulators for the ARM7 architecture.

The idea would be to use the SAM7 eval board to program the FPGA on the Spartan-3 eval board. FPGA programs are pretty large, so maybe this should be done by feeding the bits through the USB cable. Then you have some GPIO bits on the SAM7 connect to IO pins on the FPGA, and other IO pins on the FPGA to the outside world.

Some years ago, there was a guy who got his hands on some prototyping hardware of this sort (PowerPC based, I think I recall) and he set up a paid-subscription website where subscribers could submit compiled code and run it on his hardware. The cost of a subscription was small compared to the price of buying one's own hardware. According to EE Times Asia, Hitachi was doing the same thing in 2002.

Hitachi Semiconductor America Inc. is trying to move remote engineering to the next level by letting customers tinker with microcontroller hardware and software tools via a Web browser.

Working with DevelopOnline, Hitachi has set up several remote development stations for its H8 microcontroller family. For a fee, engineers can access these remote engineering laboratories from a PC at any time. Hitachi launched the service with its H8/3664 microcontroller device and plans to expand the program during the next several months to include other members of the H8 line and the company's SuperH devices.

One could make two of these gadgets and connect one to an Apache box to set up a remote development website like this. The second gadget is used to tweak/observe GPIO pins, interact with serial debug, etc. There would need to be some mechanism for allocating time fairly among multiple subscribers. The website would want lots of example code (or pointers to code findable on the web) for people to get started.