Cyber Brows - The Hardware

I've come to love surface mount components. They're just so gosh darn small and cute. This allows really compact designs. Of course this comes with its problems as well. They're just so gosh darn small and hard to see!

So far I've managed well. Passive 1206 are massive, 0804 are easy and 0603 are doable. On the chip side, I mostly buy SOIC packages which are easy to solder with only a sharp tip and thin solder. But more and more, very interesting chips are only available in QFN packages. These often have leads only underneath the package. These require hot air or a reflow setup.

3 axis accelerometerThe PIC12F1840 is available in a convenient 8 lead SOIC but the MMA8453QT triple axis accelerometer is only available in a 16 lead QFN package and man is it small. Here's a picture of it beside a 0.1" spacing header.

I wasn't sure whether my idea of sharing an I/O line between the LEDs and the I2C would really work so I knew I'd have to make a prototype board before committing to having PCBs made. I usually make my prototype boards via toner transfer but I wasn't sure whether I could make traces thin enough to support the QFN.

16 lead QFN Landing Pattern
QFN Footprint printed on laser printer

When I placed the appropriate QFN footprint into my PCB design program dipTrace, I found I needed 8 mil traces. Printing this on my cheap Samsung ML-2240 showed that this was just doable. At this size, undercutting the copper during etching is a real concern. In fact, when I made the prototype board and an additional breakout board (seen below) I could see that with 1 oz copper, traces much thinner than 8 mil would likely disappear. Surprisingly though, the two boards I made etched fine on my first attempt. A bit of good luck there.

16QFN Breakout Board
Toner transfer breakout board created to test the function of
the accelerometer. Note the thin 8 mil traces to the QFN part .

Soldering the 16 QFN package onto the prototype board proved pretty tricky. I used a sticky ChipQuik solder flux (SMD291) to hold the chip in place. Because there are no visual guides, lining the chip up with the footprint involved minute moves followed by close visual inspection from the sides with a loupe. Once things were lined up, the actual soldering with my Aoyue 968 hot air rework station was relatively easy. Once the solder melts, the chip snugs into place due to the solder's surface tension. A light tap on top was all that was required to make sure of good contacts.

On the PCB boards from Gold Phoenix, soldering was MUCH easier. The silk screen box arround the footprint made lining up the chips a snap.

The prototype board worked really well although I rerouted a few things before committing to a batch of boards from Gold Phoenix. I ran into one issue while programming the boards. Apparantly when some manufacturers tell you that a maximum voltage is 3.6V they really mean it. I was a bit cavalier while programming the PIC12F1840. It's tolerant up to 5.5V but the accelerometer is only tolerant to 3.6V. I was powering the PIC from my PICKIT 2 programmer. I had set the PICKIT voltage to 3V but occassionally, for no reason apparant to me, it would jump to 5V. I lost two accelerometers this way. In the end I attached the 3V coin battery holder and powered the circuit from that. Since then I've had no problems. The PICKIT 2 recognizes that the circuit is powered and works with that.

This is the toner transfer prototype. In the picture the QFN accelerometer has
been removed due to it being damaged during programming. The ribbon cable goes to
the programmer..

The LEDs required a bit of experimentation to find a good part. The prototype was made using red chip leds (Digikey 475-1416-1-ND) in a 1206 package because I had them on hand. These put out about 26 mcd of light at 20MA. Although they looked good in the dark, they were dim in room light. Combing through Digikey, I found three cheap alternatives that I hoped would work better. These were Digikey 754-1133-1-ND that output 220 mcd, 754-1165-1-ND that output 550 mcd, and finally 475-2820-1-ND that output 925 mcd, all at 20mA. In the end I used the brightest as they are quite dazzling even in room light and not much more expensive.

When I did a battery longevity test I used the original 26 mcd LEDs. In the configuration with a Cylon type display, the setup ran continously for about 3 days on a dollar-store 3V CR2032 coin cell.

Toner TransferCopper
Patterns for Toner Transfer

Gerber Files Here

Assembly Instructions
Assembly Guide


Next Section: Cyber Brows - The Software