bg-image-3.jpg

Addressable String Lights

Addressable String Lights

Custom PCB, Electrical Components

collaboration with Dom Chang

Overview

LEDs are cool but light bulbs are cooler. So I worked with Dom Chang to design a custom PCB for addressable incandescent bulbs that can be strung together. Each 24V bulb can be controlled using standard led libraries and drivers.

BRIEF

For the class ‘Prototyping Electronic Devices,’ we were asked to create any kind of circuit and iterate on it. I’ve been inspired by the lighting at Mister Sunday and wanted to see if one could make addressable string lights. So I roped in my classmate Dom Chang and we got to work.

Process

I found a similar project online. It was a great help as I immediately decided to avoid using any AC bulbs, dropping down to a lower voltage and avoiding lots of pain. I reached out to the creator and he provided his eagle file which gave us a great start. However that design was expensive and large. But we stayed faithful in recreating it on the breadboard at first.

After some struggles with bad soldering, we got it working. The bulb that was spec’d had a downside - it required more amperage than a WS2811 (or any led driver chip) could sink on its own. So to get the size and cost down, we tried out some higher voltage, lower current light bulbs and settled on a 24V ??A bulb and an off brand chip that could operate at 24V. Dom finished up the breadboarded version with the simplified components, we went back and forth with our eagle files until we settled on a design. A week later we had our pcbs from jlpcb.

We set up a little factory, applying solder paste, placing components, baking in the oven, hand soldering on the sockets and jsts, then hot gluing the connections. and boom they worked right out the gate. We tested them with a dmxking and I wrote a touch designer patch to pixel map them. We manufactured ~60 of them 

In the end, we got the cost down to about $2 per board, including all components (excluding our labor of course). Now I’m experimenting with enclosures and we have a clear path to improving the design; wider traces(?sp?) so theres less of a power drop, better choice of connector than jsts.

FINAL THOUGHTS

People seemed to enjoy the end installation. But the difficulty in using a neural network to classify something as subjective as dancing was that there was no objective way to evaluate its results.