#firefly 2017

Last time, I had just done a full-scale test of the new Witch Lights harness, only to find that the voltage on the far end of the harness was 3.6 volts, too little for the cheap PIR sensors I was using to work properly. I had my hopes pinned on a new set of PIR sensors from SeeedStudio, which are rated to operate with 3-5 volts.

Last week, the new sensors arrived. The Witch Lights are still snaking down my house's stairwell and into the living room, so I took the time to hook up the new LED strips I had just soldered. And on turning the lights on… they browned out.

Great.

Slowly disconnecting and reconnecting each LED strip informed me that two of the four LED strips were bad somehow, and will need to be re-soldered. Somehow, both of them passed my QC testing, but still fail when they're hooked up to the actual harness. Fantastic. Spare LED strips were dragged out, and the offending strips set aside for further punishment investigation.

With LEDs hooked up and running, I was able to confirm the results from the earlier test: the 5 volt PIR motion sensor on the far end starts to go berserk within a few seconds of boot-up, and the lights animate constantly. OK, fair enough.

Time to do some surgery: I extracted the circuit board from the far-end motion sensor housing, and wired the new motion sensor into it. Moment of truth time. Turn the lights on, and…

It worked!

The new 3 volt motion sensors operate just fine in the 3.6 volts provided by the wiring harness, as advertised. The lights no longer trigger randomly, they trigger when a heat source (like my cats) walk by either motion sensor.

Hooray!

Of course, the new PIR sensors are a funny shape, and don't fit into the housings I designed around the original sensors. The next day, I spent a few hours in Solidworks with a pair of calipers, and designed a new housing. Some hours later, I had a new housing sitting on the bed of my original Makerbot Replicator. The new sensor press-fits into the housing. I installed the new housing with no problems, and now I have a fully-assembled, functional set of Witch Lights in my living room.

Which only left the bad LED strips. Grr. Fine, off we go to the workbench. Now, what worries me is, I have an Arduino Uno with a cable hookup that connects to the LED strips once the wireless cables are soldered on. That Uno runs a test animation on the strip. I've been operating under the assumption that if there's a short circuit, the LED strip will brown out when connected to the Uno, and the animation won't play. I used this to test all my strips after soldering.

Turns out, nope. The bad LED strips pass that test just fine. That does not give me the warm fuzzies, let me tell you. So the only real test of the LED strips is to connect them to a fully-deployed wiring harness. Which means I don't get the living room back any time soon.

Update Saturday; May 13, 2017

OK so since I started this entry, I have re-tested and re-worked the LED strips. I ran them continuously for about 16 hours (two battery charges worth), with no problems. That's as good a test as I can devise, so I have to sign off on them, but I'm still wary. This means the second, new set of Witch Lights is fully functional, with its own set of LEDs, ready to deploy. I put away the LED strips and coiled up the wiring harness. It will stay that way until I start reprogramming the Arduino with new code, hopefully in the next month or so.

That left the original Witch Lights, which had the same wiring flaw as the second strand; at critical points, ribbon cable was used as a shortcut to pulling wire. That led to them malfunctioning and constantly animating at Firefly 2016. My goal is to fix this issue before Firefly 2017, on July 4th.

I pulled the wiring harness out from under the Makerbot table, and slowly, laboriously took it apart into segments. It wasn't until I had the conduit laid out on the floor, ready for the wire pull, that I found out that I didn't have enough wire left. Damn. So that was a delay of a few days waiting for wire to come from Jameco.

Once the wire came, I wasted no time in pulling the power lines through the conduit, and started to reassemble the harness. That's a slow process because every time you connect conduit to one of the 3D-printed housings, you have to continuity test the wires to make sure that there's an unbroken chain from the Arduino and power source all the way through the wiring harness. Well, after connecting a long, complicated segment of the harness, the sensor wire failed the continuity test. Somewhere in the harness, a wire had come loose from a screw terminal.

Well, fuck.

I resigned myself to a long search. Then I found the loose wire in literally the first place I looked. OK, I was owed a break on this project I suppose. I accept the good luck. After reconnecting the wire, the harness passed continuity tests. I was able to reassemble the rest of the harness without issue. When it was fully assembled, I plugged in a battery and turned it on, then measured the voltage at the far end of the harness: 5.17 volts.

Huh.

And the debug LEDs on the Arduino tell me that the Witch Lights aren't animating constantly, but are only triggered when I activate the infrared motion sensors.

That's without LEDs hooked up. I don't know what their power drain will do to the voltage at the sensor on the far end. So next up is to do a full-scale deploy of the original Witch Lights, down the stairs and into the living room. My long-suffering wife will love that, but with any luck it will only be for a day, until I can confirm that the lights are working as designed.

After that, I can focus on the software improvements. Right now, if the lights are animating, the Arduino processor is locked up, and it can't respond to any further motion until the lights are done playing their animation. My hope is that, with the help of my friend Jim (a much more experienced programmer who knows C++ and can handle complicated object-oriented code), we can have animation code this summer that can display multiple sprites at once, and respond to motion while animating the lights at the same time.

Both the current and in-development code for the Witch Lights are on github, at https://github.com/jdimauro/witch-lights along with documentation of the build process.