Mike's Spacetime

Luxdrive LED RGB units

Having made the decision to go forward with the super-bright luxdrive LEDS meant that we needed to build LED drivers. These are basically current regulators which keep the current below a maximum per LED and also provide the correct voltage drop from the supply voltage of 5V. The kind of LED we decided to use is new enough that few cheap off-the-shelf solutions exist although some do. Mark was able to find some for 15$ per channel, but we needed 56 * 3 plus spares which would come to 2500$, too much for our budget. Mark thus set out to design our own board, custom fabricated they would come to about 600$.
Mark designed the board, made some proto boards using his inkjet printer, etched them and with Richard’s help started assembling it. Most of the components were SMD type components, which are considerably more tricky to solder, so we started training people to solder these. Olli, Maja and others were soldering test boards.

Richard discusses the printed circuit design

The central component of the board is a small integrated circuit, sadlyit was only available in XYZ packaging, which means that the 7 pins forthe package are not only surface mount style, they don’t even extendoutwards from the package, but instead consist of little pads on theunderside of the package.

This is the tiny central integrated circuit at the center of the original design. Note the lack of extended pins.

This turned out to be the bain of Richard and Mark’s existence for about a month. By the end of June they had attempted to solder 6 boards, each with 3 channels – but only 1 or two boards ended up being functional, even after several soldering attempts on that central chip. Slowly but inevitably we concluded that while in theory manufacturable, the tiny component was not made for human manipulation.

We were stuck. We realized there was no way we could manufacture 54functional boards at a yield of 20% or so in less then a month.The LED heads themselves also turned out to be trickier to solder then imagined. Each head consisted of a 4×4″ aluminium plate with two luxdrive LEDs screwed to it (into predrilled and tapped holes), with wires soldered to the LED pads. Ash and Richard spent an hour soldering a single LED head. Thus that would mean another 120 man hours of work just for the LEDs.

Soldering cables to the light head

Had we taken up a task to momentousness for us to achieve ? We had already overrun our planned go-no-go date by two weeks and due to long lead times on ordering the boards and some of the components which had to come from China (several weeks) our time was running out.

We called the critical meeting at ATC. The mood was mixed. On one hand we had made great progress, we has a design that was likely to work, in principle. Structure, fabric ,reflectors and UI stations had been prototyped successfully and in principle the electronics and lighting components had been too.On the other hand it was unclear we would be able to manufacture the electronic parts needed in less then 4 weeks leading up to August. We discussed the possibilities. We could simply continue, build as many light boards as we could and downsize the project as necessary – It would make a pretty light wall or a small chill space or something nice, even with as few as 10 working light panels. We had managed to create 2.5 working panels, surely we could make 10. This way would be rather unsatisfying but we could always return to the playa the year after with the full cube. In the mean time mark was working on another, potentially simpler to manufacture board, but we had no way of knowing if it would work. After all it had taken weeks to get the previous design working.Should we wait and delay the go decision another week giving him time to complete and test the design ? This would put the other areas of the project at risk, since parts couldn’t be ordered in time. But the small possibility of being able top manufacture all the boards due to an improved design was highly appealing.
Furthermore, our funding situation had potentially improved by this point – it was looking likely that we could get fund matching from Microsoft and Google since the project was run through a non-profit organization (Shunpike). So with more money at hand, could we use the off-the-shelf solution ? It would put an extra $2000 onto our budget but maybe we could deal with that now.
Or do we stop entirely, go to the burn with no art, have a good time,and build the cube for the year after ?

The mood in the room was somber. We voted on the options. Shakingly we decided to delay the decision one more week, wait for the results from Mark on the new board, risk the entire project but leave the chance of pulling the entire thing through and achieve the original vision.

My mood was incredibly stressed, i had trouble sleeping, couldn’t function at work, couldn’t think of anything useful and felt generally paralyzed.

Barry, Richard and Mathew discuss the first set boards fresh out of the oven.

Mark was working on designing a different board with a different central IC, which was not as suitable as the previously used component, but itcame in a package with actual pins and was thus much much easier to solder. However the new design was entirely untested and required extra components on the outside. There would not be enough time to get a proto-board etched, shipped, tested and the resubmit the final tweaked design. A simple, home-etched prototype worked, which gave us confidence but really we only had one shot at this.

Using the experience gathered from the previous boards Mark tweaked and tweaked the printed circuit design, trying to avoid areas which would form easy solder bridges, or bring parts in clashing distance. All of this had to be done theoretically, on the computer screen. I remember staring at the design on tuesday night at 2am, and ask naive questions and make naiive suggestions. Mark would think about it and either give me a solid answer or it would make him think about it and he’d adjust the design. Richard & Marked looked over the board some more the next day and then went for it, it was ordered.

Old board (left) new board (right)

Mark etched a protoboard and Richard was able to assemble it much easier then the previous design. All the channels works with less then half an hour of debugging!
We had a working and manufacturable design. The one remaining question was thus whether or not we had to hand solder every board of if could mass produce them using a mask and oven procedure. The latter involves applying solder paste onto every solder pad, using a cut stencil, placing each component and then warming the entire board in an oven to a precise temperature, enough to melt that solder but not enough to damage the components. This is essentially the way commercial electronics is produced, except that the placing of the components is done by robots rather then humans. Still, this way of manufacture hugely reduces manual labor and more importantly reduced the chance of damaging components by touching them with the soldering iron for too long. While Mark has been trying o convert a toaster oven into a solder-melt oven using some custom thermostat electronics, Barry got a friend from Google involved in the project who had a lot of experience in producing boards in this way. Together they managed to get this process working for our board. Matt used his home-built laser cutter to produce a series of stencils for the solder paste. It took some time to get the thickness of the solder mask just right. Initially they were too thick and the components would flow away and twist as the solder melted underneath them.

The printed circuit board with solder paste applied.

As part of the enormous electronics manufacturing effort led by Olli (Many other components such as the power boards, LED heads and cables still had to be hand made) Matt, Richard, Becky and many others were able to manufacture all 65 boards (54 + spares) in less then two days ! Two days ! Matt spent many hours under the microscope checking the boards and repairing any incorrectly deposited components.
The biggest electronics obstacle was overcome, with the quality of the final product exceeding our wildest dreams. In fact, only a single board broke on the playa, everything else worked solidly.