I knew after finishing the Outputs Assignment I wanted to play more with RGB ligts, ideally creating several programs and patterns which responded to differeing inputs. I thought I would start basic by using levers to pull pins and activate different while loops, with different permutations of flipped levers could all yield unique outputs. I also wanted in integrate a pyrometer, and ideally one of the sonar units, so that eye responses and movements could be triggered by proximity.
Back to Index |
Basic "unit" of the RGB circuit. Connected in series, with jumpers from pads on the outer circuits. | ||
Control board. Has three cross-wired RGBs for pin minitoring, and is run through PWMing mostfets, because 18 RGB LEDs are too great a load for the microcontroller. |
The first PNG for the traces I generated was very messy, full of artifacts which seemed to confuse fab modules, and also drastically increase the wait wait for the Roland Mill. So I redesigned the board from scratch again, being sure eliminate artifacts early, and use Gimp to draw solid, smooth traces on the PNG.
Milling, Mounting
Despite redoing the traces twice, I did not catch on to a major flaw until after I printed the board. When designing the board, I manually drew the left eye and then mirror flipped it on to the other half. This simple transformation ignore the fact that some SMT components are chiral, and can't have their traces and still fit on the board. The RGBs were one such component - it was impossible to correctly align the anode on the right side of the board. I decided to press on - the LEDs were wired in series, so I thought they should still work even with half of the components missing. I wanted to press on, and get the control board built.
We're getting down to the end of our boards, and the only large 6" onces left are double sided. This required some tinkering with fab modules and the end-mill fit to get traces cutting clean - the double sided boards seem to bow more, and are thicker.
3D Printing
While the first board was milling (took over an hour. Sorry classmates!) I quickly designed a 3D printed model for a set of pins to mount on the human wireframe. In this case, they will hold the LED panel, but they are versatile and can be used on every 4-way junction on the model. It was my first time using the Makerbot, and it seemed to be just fine to me, althoug I was surprised by the apparent lack of control options in Makerbot's software.
A bifurcated clothes' pin design. | ||
Attached to the model, and glued to the SMT. |
Control Board, Programming
Finally, I milled the control board. Some of the traces were close, so I set the tool diameter to 0.2 (usually 0.4 for 1/64 endmill), and increased the overlap. The board came out almost perfect, except one route was hacked off on the edge. I mended this with a jumper, and then stuffed the board. I stopped using the rainbow-wire as a source of wire (it's too big, the fibers flare apart and make it hard to place, and the rubber coating melts very fast.) I switched to a smaller guage wire, but a single solid strand, with a more resistant plastic coating that was easier to bend into place.
I thought the board looked good, and testing AVR dude revealed no "rc=-1" error. So I packaged up one of the codes I wrote for the 3-LED RBG for Output devices, to test the equipment. Although swtiches were attached, their pins were not set for any input, and the board should have behaved just like my previous board. But it didn't work - despite successfully loading onto the microcontroller, none of the lights turned on.
Control Panel. Wmop womp. |
I tried hard-wiring the LED Panel into a DC power source (I used a 5V charger. It did not work either. I may be because of the aforementioned issue with the reversed traces, and half of the board being empty. At this point I had amassed a good haul of artifacts and experiences, and despite not getting the board functioning by the deadline, I will continue to work to get this online and operational. In the meantime, I'm interested to see if I can find better conditions for using the Sense, so that I could make a full body model perhaps using the composites (using cardboard would be a great visual effect, but also resource very intensive).