David had a specific printer in mind, an Ultimaker, and, out of a desire to drink as deeply as possible from the well of experience, we decided to order it in kit form. We placed the order and waited eagerly for its arrival. Or at least I waited. David went out and got a commission from the Autodesk Gallery to build a redesigned version of his "Big Ball Maze," a sort of giant toy maze better seen than explained. Here's one view of it:
That's David in the background, and Gideon in the fore, "playing" the machine. As you can see, as goes the disc in Gideon's little hands so goes the table and the balls (should have been a single ball, but that's Gideon for you) upon it. But I'm getting ahead of myself.
Having invented the thing and built more than one of them David already knew how to make a Big Ball Maze, but the agreement he had struck with the Gallery called for a redesign with the goal of introducing 3D-printed elements wherever reasonable (again, in line with Autodesk's strategic interest in this technology). This meant, at the very least, designing and printing in plastic a number of parts that were constructed from wood in earlier models, and for this task, as well as for other miscellaneous labor, David invited me in as half partner on the project. To be clear, David did not need me for any part of this work: in addition to being the original designer of the piece he was already a competent 3D designer, whereas I had no almost no experience with these tools and had spent far less time than he researching the printers themselves. But he was facing a deadline and I had a spare garage. In any event, I agreed and immediately made my first, and arguably most important contribution, by insisting that we not wait for our printer kit to arrive (it was being shipped from the Netherlands and I know from long experience not to rely on speedy delivery across that particular route) but that we instead find a way to get access to Autodesk's fleet of Objet 3D printers.
Printing on the Objets had a number of implications, the most important of which were that a) each print run meant asking for favors (someone else had to load the file and oversee the print operation) and though these were very willingly given we naturally felt inclined to limit the number of iterations; and b) we could print without a thought given to "overhangs," i.e., the points in our design that would have to be printed with nothing underneath them, because the Objets can print with two materials simultaneously, one being the "final" plastic, the other a different kind of plastic that is easily washed away and which can be used to provide removable supports in those areas of a design where overhangs exist. Don't try to reread that last sentence, just look at this, a picture of our first print:
See the hole through the middle? It's easy enough to print the bottom of that cylinder but what happens as you get half way up and have to start closing the circle? There you encounter an overhang. The beauty of the Objet is that it fills in the cylinder as it goes with a separate plastic that you later wash out. So in this picture you see one big hole where we've removed the support plastic and six small ones around it with the support plastic still in them. Here's what it looked like once we cleaned it all up:
There's the design on the screen and there, lower left, is the piece as printed. Let's think about that for a second: I made this on my computer, and then magic happened, and then there it was, on my desk. Amazing.
That was our first print on Autodesk's printers, a draft print in every sense: we hadn't perfected the design, we didn't print it in an especially sturdy plastic, but it was certainly good enough to test the precision of the printers (fantastic) and to see back home whether the piece made sense as part of the maze table we were building in my garage. It did, and so we went ahead and designed all the parts we thought we'd need for this project. Then, again on Autodesk's machines, we printed them, and here are most of the results:
We were confident these pieces would hold up under the small strains they would face in operation. The ball and socket pieces, however, were facing a much more serious mechanical test: the entire weight of the maze table rests on the socket, which in turn must glide over the ball as the table moves. Would the plastic as printed be able to handle the load over time? Would it heat up and deform, or wear away? Unfortunately, by this point the deadline was upon us: the maze was due to be delivered to the Gallery so that it could be packed and shipped in time for the SXSW Create event. Product testing would have to happen in the field.
Things did not go well at SXSW, but not because the plastic wasn't strong enough. The ball and joint performed perfectly, but designing the motor and spindle supports as separate pieces proved to be a mistake. In the picture below you see (moving from screen right to screen left) the motor mounted on its support and attached to the spindle via a metal coupling (the double cylinder of metal in the middle of the photo). The spindle then extends through one tower via a skateboard bearing and on through the next tower at far left.
The part holding the motor and the part holding the spindle are two separate pieces and were designed as such because this made it easy to get at the six screws attaching the motor to its support (the small bolts near the coupling), but we didn't anticipate how difficult it would be to get precise alignment between the spindle support and the motor support. As it turns out if the spindle is even slightly misaligned with the motor's shaft the load on the motor is greatly increased. In initial (and all too brief) tests the problem didn't show itself, but after some hours of continuous use at the event one of the motors burned out. We arranged for its replacement, but it was only a matter of time before it happened again, so after the maze got back to San Francisco we decided to redesign this and some of the other parts before final delivery to the Gallery.
By the time the maze returned to us we had received and built our Ultimaker. This meant that we could now do our own production at home and that, in turn, meant we could iterate rapidly and repeatedly, designing, printing, evaluating, and then designing again until the desired part was in hand. Some 16 versions in, of which five were printed at least in part, we settled on this design for the combined motor-and-spindle support:
Three key improvements here: first, the spindle and motor supports are now a single piece and thus hold the motor in perfectly straight alignment to the spindle; second, the spindle supports are smaller, which reduces the amount of plastic used (we eliminated the exterior flanges for the same reason) and lets you get at the motor screws; and third, we added an hexagonal hole at the top of each spindle tower, allowing us to pin the bearings in place so that they cannot migrate out of their housings as they spin. Pretty slick, at least onscreen, but what about the many overhangs this presents?
Somewhat to our surprise, the Ultimaker, at the right temperature and speed, didn't have as much trouble with overhangs as we'd anticipated. Here's a movie of it printing another piece, our redesigned foot peg holder, one which also features embedded hex bolt seatings:
Mesmerizing, isn't it? The overhangs don't print perfectly, small loops of plastic will sometimes be misaligned or will droop a bit...
...but since all of the holes in our design are there to be filled by bolts or bearings or some other bit of hardware these flaws don't show.
We spent a good deal more time refining our designs, figuring out how solid to make our fills (i.e., how hollow to make the interior of the printed parts; we found 20% fill to be enough in most cases), experimenting with print orientation to reduce the risk of parts shearing under load, and so on. Having a printer on our communal desk was simply an incredible luxury, and we've found its busy little sounds to be great company on a late night.
And many a late night there was. For me at least doing 3D design takes a lot of time. It's terrifically enjoyable and appeals strongly to my puzzle-solving instincts, but I'm still not very good at it and tend to make a lot of false starts. (Amateur-level 3D design programs--I used our own 123D Design almost exclusively; Dave used the professional-grade packages Fusion, Inventor, and Adobe's Illustrator--are not very good at versioning so a false start often means starting over from scratch.) That having been said the majority of our time--and this took a LOT of time, many nights each week for a couple of months in a row, so thanks here and everywhere to our patient wives; for my non-wife readership please take this as explanation for the paucity of postings since the beginning of the year--was spent on non-printing activities, like carpentry, finishing, and, for Dave, programming. In an ideal world we'd be able to print the entire thing. That world isn't far off.
Cool hobby!
ReplyDelete