I think LEDs are rapidly becoming the “put a bird on it” of my blog…
Something like two years ago, I had a friend who wore a pair of Christmas light earrings around the holidays. They were the big bulby kind, and they looked so much like the real thing that she was constantly asked if they actually lit up. That got me thinking: how hard would it be to make a pair of Christmas light earrings light up?
I purchased a few glass Christmas light bulbs and quickly set to work cutting them apart:
I even bought a special diamond wheel for my Dremel tool for the project. The plan was to remove the base from one bulb and the top from another and then reassemble them with a small circuit inside that powered an LED.
I really wanted to match the even glow of frosted holiday lights which ironically meant that I couldn’t use frosted bulbs. Frosted bulbs are typically coated on the inside with a frosting layer that flakes off very easily. The plan was to use non-frosted bulbs and then frost the inside with some glass etchant:
I even had plans to build a base station that looked like a conventional light socket. Screwing the bulb into the socket would start it charging. At the time, this was achieved using a tilt sensor and…to be honest, I never got a chance to work out the details.
See, no matter how bright of an LED I tucked inside my frosted bulb, I was still unable to achieve that brilliant even glow of an incandescent light. This was partly because of how small of a point light source an LED represents but also because high intensity LEDs tend not to be the diffuse variety. Even scuffing up the LED with sandpaper didn’t help much. I benched the project.
My pre-project shopping spree left me with a dozen Christmas lights in various states of broken and a small collection of very very small lithium polymer batteries:
These things are seriously tiny. They were the smallest I could find online, and they just so happened to fit into the base of a Christmas light. Although their 19mAh of charge is tiny compared to just about any other battery I’ve dealt with, it can still provide enough energy to run an LED for a few hours.
I found them on All-Battery.com, and unfortunately, they’re no longer in stock (unless you want to order 30,000 of them). I found a seller on Alibaba who sells them, but they’re out of stock as well.
So a few weeks ago, I realized that I was sitting on some pretty rare parts and decided that I wanted to finalize my earrings.
The goal of this project was to create a pair of tiny light-up earrings that are totally unobtrusive and look and feel more or less like normal earrings. This involved creating the smallest circuit board I’ve ever made and packing it into a tiny case where none of its components would be accessible from the outside.
It’s also the first project I’ve ever made that can recharge its own batteries.
Before I dive into the details of the battery management and other electrical aspects of the design, I think I should cover the aesthetics and some of the non-electrical design decisions I made along the way.
The final circuit and battery combo looks like this:
The six lights on the front are programmed to light up in sequence going around the circle.
Now, you might be wondering how I developed the firmware for such a tiny platform that doesn’t have so much as a programming header. My original thought was to work out the firmware on a larger development platform and then when I was satisfied, program the parts before soldering them down. Of course, you can’t exactly plug a QTFN part into a breadboard for programming, so that would require something like this:
And these usually cost around $80. I wasn’t quite ready to drop that much dough on something I probably wouldn’t get a whole lot of use out of, so I came up with a different solution.
Both of the earrings started their lives like this:
You might already recognize the bottom right portion as the front face. The top right section is the “brain” which houses the microcontroller and just about everything else, and the top left section is the programming header.
The idea was that once I was happy with the firmware, I was able to permanently slice off the programming header and the LED face. The LED section was then reattached with some small wires that wrapped around the battery.
As you can imagine, this process really pushed my fabrication skills to the limit. Many of the traces on this PCB are only 10 thousands of an inch thick. The boards came out much better than I expected. And that was kind of a problem.
When printing a board, you have the option of leaving the dimension layer on or off. The dimension layer shows all the lines that represent the outer edges of the board. Because this layer serves as a method of documentation rather than a representation of anything physical on the board itself like silkscreen or copper, it usually has “zero” thickness which means that the printer makes it as small as possible while still being visible.
Typically, this line is so thin that large portions of it don’t make it to the end of the etching process. This is why I had no qualms with running traces straight through this outline between the three mini-boards.
Surprisingly, this time my boards came out so well that these tiny lines of copper were actually shorting out some of my traces! Seriously, look how small these things are:
That clear looking fiber is a piece of hair from my head; I measured its thickness to be 0.003 inches. That’s a seriously tiny trace. I didn’t even have to do anything special to get this result. I just carefully followed my normal printing methods.
Because of their size, these boards were also the first of mine to require 0402 components. Just about all of my boards up to this point have used 0603 components which are 0.06″x0.03″. These new parts are 0.04″x0.02″. They’re small. Small enough that I had to whip out these bad boys just to see what I was doing. For my earlier prototypes, I actually managed to assemble each component by hand with careful use of a pair of tweezers, but I soon switched to solder paste.
One thing I learned this time around is that solder paste doesn’t need nearly as high of a temperature to melt as regular solder does. For a while, I was cranking my hot air rework gun up to 360-400C figuring that it should be around the same temperature as my soldering iron.
What I later learned is that only 260C or so is required to get solder paste to flow. I bet you can tell which of these two PCBs was made after I figured this out:
Here’s a good closeup of the board:
The open pads on the top and bottom are where the charging pins connect. The battery connections are on the LED board.