The RepRap build-a-thon in review.

28 January 2009

The weekend of the RepRap build-a-thon at HacDC started off simply: Lyssa and I went to dinner at Konami. We haven't been out for sushi for a number of months, due to my getting sick there in 2007. However, the food is still good and we enjoyed ourselves. I was unusually popular that night; my cellphone kept ringing every few minutes for various and sundry reasons. After dinner I dropped Lyssa off at home, loaded my gear into the trunk of my car, and headed to Hasufin's to pick him up because we were off to HacDC to help set up for the RepRap build-a-thon. We were a bit late to the party, so to speak, but there didn't appear to be much that could or had to be done at the time: the tables were moved and stacks of chairs were in place.

When we arrived we found Adam and Brian diligently hacking away on Adam's RepRap which had been set up a short time before then. They'd gotten it assembled but ran into problems with the extruder head; the drive screw ran as expected and was supposed to force a filament of ABS plastic through the heating element and out of the nozzle but but the softened plastic wasn't moving the way it was supposed to. Once in a while we could coax it to lay a thread but not consistently or for very long. Adam had mentioned that he'd had it working earlier but went back to work on a few things (he'd replaced some of the structural components with parts he'd fabbed) and everything went pear shaped. One thing that was pointed out was that he was using the drive screw extruder, an older design which wasn't considered very reliable.

After Nick got back from picking up Zach all of us hiked a couple of blocks to the Heights, a fairly hip and upscale restaurant on the fringe of the college campus. Most everyone had a later dinner that night; Hasufin had dessert and I had a cup of coffee to relax at the end of the day. We discussed the possibilities of rapid fabrication until well after midnight, at which time we parted ways because all of us had to get an early start the next day. Our trip back home was fairly rapid due to the late hour and lack of traffic and I crashed in plenty of time to get up at 0730 EST5EDT.

By 0900 EST5EDT the next morning, Lyssa and I had gotten up, gotten dressed, and stowed our gear in the trunk once more. We picked up Hasufin and then stopped off at the local Panera Bread to get breakfast and a box of bagels and cream cheese to take with us to the build-a-thon. Neither Hasufin nor Lyssa were particularly pleased with their breakfasts. Maybe I don't have much in the way of standards but my bagel was fairly tasty and suited me just fine. It says something when Lyssa, who ordinarily avoids coffee, drinks a few slugs of mine to get the taste out of her mouth. Following that side trip, we jumped onto route 66 east and set course for the hackerspace.

We arrived around 1000 EST5EDT on Saturday morning, offloaded our gear into the little alcove just off the church's auditorium (where Adam and Brian had set up the night before) and made our rounds. I introduced Lyssa to Nick and Mark before running off to set up the soldering stations at the roundtable and checking on the tool spreads. At one point I had to move my car because I'd accidentally (and creatively) parked in the church bus' parking space, which was to be used that morning. Luckily, I found a parking space on the street running next to Saint Stephen's (and HacDC, by extensions), so it didn't take very long.

The day started off with a presentation by Zach Hoeken, who in heavily involved in the the RepRap community as well as the hacker space NYC Resistor, who has been hacking on this technology for going on three years now. You'd be hard pressed to find anyone in this country who knows more about it than he does simply due to the amount of work he's put into the project as a whole. He's best known for designing the Sanguino board, which is the microcontroller underdlying the latest generation of RepRap hardware. His presentation discussed what RepRap is, what it does, how it was designed, and what's been done with it.

While Zach was front and center speaking, I was running around behind the scenes taking care of last minute fiddly stuff, like running an Ethernet cable from the HacDC space upstairs down the back of the building to the ground floor and into the auditorium so that I could set up a wireless access point. All of the documentation, from the assembly instructions to frequently asked questions to where to download the software can only be found in electronic form, so net.access is a necessity. Unfortunately, as most of the early generation wireless access points I own tend to be flaky, we ran into problems with the AP within minutes of Zach's closing statement: the one I'd set up runs the 802.11b wireless protocols (at a speed of 11 megabits per second) even though the firmware claims to be outfitted for 802.11g (running at 54 megabits per second). When I got the AP in question, I basically looked at it, saw that I couldn't use it to experiment with the open source Tomato firmware, and put it in the closet for later. All day, it ran at 11 megabits per second which meant that everyone else's access bogged down often.

Around the auditorium we'd stationed tables for various purposes. There was a how to solder roundtable at the back that attracted not a few people during the course of the day. There was one table near the front headed by Zach for building the drive-screw extruder head. Near the back of the auditorium, almost inside of the alcove where Adam's RepRap was set up, one table was set aside for punching out laser-cut parts from large sheets of acrylic backed with what amounts to masking tape. In the HacDC workshop upstairs, a couple of people were soldering together the control circuitry for the extruder head, the driver boards for the motors, the temperature and optical positioning sensors for the toolhead, and the Sanguino mainboard that ties the whole shebang together.

The thing about buying a RepRap (technically a RepStrap, because the parts are pre-manufactured rather than made yourself) is that it consists of many, many, many small parts that have to be identified, punched out without damaging them, and fitted, slid, or bolted togther. As with many kits, some of these parts are barely the size of a thumbnail while others make absolutely no sense if you've never worked with them before. I recall seeing only three acrylic parts that had any identifying marks on them, and they were all flags for the optical positioning sensors.

As excited as this project makes me, I have to be honest: this is definitely not some for beginners. To successfully assemble a RepStrap you will need the patience of Job, the wisdom of Solomon, and the luck of Inspector Gadget to put it together correctly, calibrate it, and get it running. Having studied engineering under the tutilage of Tony Stark is strictly optional. Making this system work requires knowledge spanning many different skillsets, which I am not ashamed to say that I lack - the ability to visualize how oddly shaped parts fit together (sometimes a dozen simultaneously) out of context is essential. Knowing something about cutting and shaping metal, even something as simple as metal rods is also highly important given that the superstructure is comprised of them. The ability to fit together and solder electronic components in the form of a kit (you don't have to buy the parts singly) is also necessary (though it might be possible to purchase the boards already assembled).

Around 1400 EST5EDT, Hasufin, Lyssa, and I gathered up Serge and Brian to go to lunch. Brian was supposed to do a presentation on 3D modeling using Art of Illusion, an open source 3D package coded in Java, but general consensus was that we were running on empty and needed to get something to eat if anything else was going to happen. I sent Nick a couple of text messages and we struck out for the Heights, where we'd gone the night before. When confronted with a 45 minute wait for a party of six, we decided to backtrack to the Cuban restaurant one block back. It was at this time that we ran into Nick trying to catch us up. I don't recall the name of the Cuban place we had lunch at but the horribly slow service made the fairly tasty food not worth the wait. By the time we got back Lyssa, Hasufin, and I were feeling the effects of sleep deprivation from the night before, and opted to call it a day and head for home.

Around 1930 EST5EDT, Lyssa and I were both up and around and wishing that we'd gone grocery shopping because we didn't actually have anything around the apartment to make for dinner. On a lark, I rang up Orthaevelve, who already had plans to go to dinner with Nicole, Hasufin, and Mika. She didn't mind the extra company, however, so a short jaunt brought us to her place, and from there we invaded Tequila Grande shortly before it closed for tex-mex. Not long after we got home, we wound up passing out for the night because I had to get up at 0800 the next morning to pick up Hasufin and get back out to HacDC for the final day of the build-a-thon.

As it turned out, an indeterminant number of people were there late into the night assembling the very tiny acrylic parts of the motor mounts, the Cartesian bot, and the extrusion head. Meanwhile, Ash built the power supply into a decorative housing and mounted the Arduino mainboard to the top. Zach had constructed his pinch wheel extruder, which works better than the drive screw model due to the fact that it exerts more force against the filament of feedstock, which makes it more reliable. It's also a much simpler design when you compare it to the rest of the device. As with many things, simpler systems tend to work better and are less prone to malfunction. All in all quite a bit of progress had been made the night before toward getting the RepRap assembled.

As it turned out, one of the major problems of the day was that there were too many people packed into the HacDC space all talking at once. All of the overlapping voices made it very difficult to ask any meaningful questions because the important, lengthy questions simply couldn't be heard. I had a very difficult time finding out what was going on and what needed to be done, and I'm still not sure how I managed to convey to Hasufin that we should pick up lunch before the teleconference with Adrian Bowyer (the original inventor of the RepRap) who happens to be in the United Kingdom. His presentation went on for about an hour, during which time he described his motivation for inventing a rapid prototyper that (hypothetically speaking) anyone could set up in the garage or basement. At the end, I asked him a question that I've been pondering for a while: how much time is really required to maintain a RepRap once you've built one and got it running? Right now they're kits and at this stage in their development they require a good deal of care and tweaking to keep them operational, as evidenced by Adam and Brian's last minute debugging and tweaking on Friday night. Adrian replied that he spends about 15% of his time performing basic maintenance on his RepRap, with the rest of the time spent either designing upgrades or using it to fabricate objects. I don't know how that compares with most industrial machinery, but that seems a bit high to me.

As the crowd thinned out through the course of the afternoon I was able to roll my sleeves up and work on the RepRap itself in a meaningful manner for a change. Hasufin spent most of the afternoon building and rebuilding the Cartesian bot assembly while I worked on aligning the sides, building and attaching the diagonal supports, and jigging the edges so that they were as close to perfectly straight as they could be made.

The diagonals. Oh, gods, the diagonals.

The diagonals are the pairs of cross-bars on each side and on the bottom of the device which add structural support by bracing opposite corners together. The thing about them is that every time you add a pair of diagonals to the framework the whole thing racks slightly, thus knocking structural members out of alignment. You then have to go back and re-jigger the rest of the diagonals to return the shape of the framework to a perfectly cubical state. This takes an insane amount of effort because every one of the preceeding sides has to be re-done whenever you add another side. Also, the acrylic nature of the attachment points makes them much more delicate than reinforcing structures really should be. You can't really tighten them by hand without risking blowing them out, but you can't really use your fingers either because the head of the bolt and nut that are supposed to clamp them down are fairly tiny.

By 2000 EST5EDT, Hasufin, Ash, and I called it a night.

I appreciate being able to buy a RepStrap from Ponoko, I really do. Finding a company that'll laser cut something out of acrylic for you using a design that you send them without having to order a few thousand at a time is rare due to the cost of running an industrial laser cutter. Acrylic is a light material which makes certain aspects of assembly easier for the beginner. The down side is that acrylic is fragile: tighten a screw a little too much and it'll spiderweb if you're lucky and shatter if you're not. Anyone who's ever built a PC chassis with a window in the side knows about this if they've overtightened a screw. The acrylic components are also prone to stress fractures due to the purposes they serve. Hasufin estimates that those parts will probably have a 3-6 month lifespan under normal use before they crack or splinter. In short, acrylic is definitely not the right material for robotics in general (books have been saying that for years) or the RepRap in particular. An ideal ownership strategy would appear to involve buying a couple of dozen feet of ABS plastic filament from your local welding supply store and printing out one or two sets of spare parts to squirrel away for later. When the acrylic parts finally give out you'll have spares handy, and the spares will actually be stronger and more durable.

Something else that we noticed about the Ponoko kit was that a few of the parts were incorrectly cut, though someone at the build-a-thon who was handy with a Dremel was kind enough to carve them to fit on short notice. The assembly instructions are obtuse to the point of being unuseful in critical places, which meant that finding and downloading a .pdf file that contained rotatable and manipulatable 3D schematics was essential, though the UI took some getting used to. I must confess, I was skeptical of Adobe adding 3D graphics to the PDF file format, but now I understand the usefulness of this feature. It is my considered opinion that the design needs to be better documented if it's going to be useful. For a while last weekend we referred back to Adam's RepRap often because looking at it made more sense than the blueprints.

Another sticking point is that some of the parts, like the 5mm long grub screws that are supposed to fit inside the corner braces are damnably difficult to get in place. There are other ways to provide structural support that don't like to wiggle loose and fall out when you need them the most. I'm all for clever engineering, you understand, but there's an important distinction between 'clever' and 'too hard for anyone to use'. Frankly, I've fallen in love with this particular project and I hope to work on it some more on my own time, but before I go singing its praises I have to be a realist. It's a kit. More accurately, it's a research project geared toward hardware hackers first and foremost. It's not ready for prime time, when everyone will have a fabber set up in their basement or garage to crank out a new lock for the front door when the old one breaks. The RepRap also has some engineering deficiencies that will have to be worked out, but this is the perfect community to do that. Some more exotic toolheads (such as grinders, routers, and extruders of different materials) will have to be perfected; this will also require adding functionality to the 3D modelling software to take into account multiple materials, its native file formats will have to be updated for same, and the C&C software will have to be reworked. This would be a difficult but certainly not impossible task.

But will the RepRap be more versatile and user friendly inside of two or three years? I certainly hope so, and at the rate the design is evolving, I think it'll get there.