I wanted to make a desk clock, but didn't want to use the 7 segment displays so common in store-bought ones. Once I came across Numitron displays in a magazine, I knew they were perfect for my clock, and so this project was born. It's currently a work in progress, but I'll update this description once the project is finished.
After my design considerations the other day, I started designing a schematic to reflect them. I created it in KiCad, and it's currently made up of two sheets. One contains the microcontroller, power supplies, and RTC; the other has the Numitron digits and their respective controllers.
While I was creating my schematic, I had to change some things from my initial design. First, I realized that I wanted to be able to vary the brightness of the Numitrons. I did this by creating a separate voltage supply, and making it adjustable by way of a digital potentiometer. Secondly, I wanted battery backup for my RTC; the easiest way to do this was to add an external IC (the DS3231) that supports battery backup from a coin cell.
I'll be testing this circuit on a breadboard within the next couple of days. If you'd like to view the current schematic, refer to the GitHub repository that I created to store the project files.
The Numitrons I'll be using for this project are the Sovtek IV-9s, as they seem to be standard across most Numitron projects like this. I'll wire up 6 of them total, so I can display HH:MM:SS, one Numitron for each digit.
As far as control, I decided to use an ARM microcontroller, because I'm trying to move away from Arduino's level of abstraction. The mBed online IDE was recommended to me as being easy to use, and compatible with many different boards; all I needed to do was choose one. After looking through various mBed boards, the F303K8 board seems to be exactly what I need: It provides a chip with an internal RTC, a compact form factor, and low power usage, all at a $10 price point.
The chip can't control Numitrons directly due to current requirements, so I needed some way to connect them. I was going to use 6 standard 74HC595 shift registers, but then I realized the Numitrons require 20mA of current per filament, which is way more than the HC595s can supply. Luckily, TI makes a power logic version called the TPIC6B595 that can supply up to 150mA per output, so I decided to use 6 of those (There are 8 filaments per Numitron, and conveniently 8 outputs on each shift register). I can also daisy-chain the shift registers, so controlling the 6 Numitrons only requires 3 pins.
I'll have a schematic and more updates on the project soon; I'm really excited to build this!