I had a pretty good idea of what I wanted to do: an absurdly large Nixie watch, completely unpractical but with a lot of character. I spent some time in Blender to make a 3D model.
I couldn’t really get the filament glow right, but the overall feel of the watch turned out okay.
The watch is somewhat “taller” than the digits, due to the battery height but also due to the PCB.
The case was developed using FreeCAD, and the PCB was developed using KiCad. They were developed in tandem, requiring a few iterations to get both right. The trickiest part was getting the hole for the USB port right.
The four Nixie tubes are soldered to the back side of the PCB, the big holes are for the tube “nipples”. The back side is otherwise free of components.
I sent the PCB design to JLC PCB for fabrication.
All components, except for the Nixie tubes of course, are surface mount. I used a stencil to apply an even layer of solder paste, put the components in place with tweezers and soldered everything with an 858D hot air rework station I got off ebay for 30 euros.
I had ordered the PCB with a HASL surface finish. HASL (hot air solder leveling) is the cheapest PCB surface finish that PCB manufacturers usually provide, but it leaves an uneven surface for soldering. I had to re-solder a few of my chips due to bad connections, especially for the QFN packages. What’s worse, poorly soldered QFN pins don’t leave visual clues, and you can’t even probe the connections with a multimeter to be sure..!
The leads of the Nixie tubes were quite crusty and I had a hard time soldering some of them. I could have cleaned them before soldering, but I couldn’t figure out how to clean them quickly and without damaging the tubes.
Time to test the PCB!
Turning on the power to the PCB was quite disappointing; there was no high voltage at the output of my voltage converter.
I tried changing my hand-made toroidal transformer a few times, to see if the number of windings could have an impact on things. I couldn’t really get my head around what could be wrong. My multimeter showed excessively high voltages around the transformer and even buzzed in pain from overvoltage. Something’s really wrong here.
I figured that my career as a transformer maker had to wait. After browsing around for a while, I found the ATB322524-0110, a miniature 1:10 transformer, corresponding perfectly to what I needed. Ordered, installed and it works! Yesss!
After a few additional re-solderings of the Nixie tube driver ICs, everything seemed to be working just fine!
I had made a really simple firmware for the watch, mostly juggling SMBus calls to the various chips on the board while trying to keep things as simple as possible.
I used Simplicity Studio from Silicon Labs for the task. Takes a while to get used to, but the tool is quite indispensable for this family of microcontrollers as the special function pins move around the chip as you activate and disactivate modules..
The battery charging IC has two “watchdog” timeouts with the shortest one at 15 seconds. If the microcontroller does not reset this timer regularly, the charging IC stops charging and reverts to a low-current setting, making the step-up converter malfunction. The second timeout is longer, but cuts all power, turning off the microcontroller. Debugging my code just turned a lot more interesting.
I quickly noticed an issue with my time keeping. When the step-up converter was active, the clock lagged.
Not having the means to fully debug the issue, I figured that it must be a noise issue. The PCB is a two-layer one, with noisy powers and grounds running just past the real-time clock IC. I must be having a signal of some frequency couping into the sensitive oscillator circuit.
In all, I’m very happy as to how this project turned out.
Going further, I might have a go at a four-layer PCB.