I built my first digital clock back in the mid 70s. It was a Heathkit of course (one of many I built as a teen) and it had a cool looking Nixie-like display. Fast forward to today when I recently came across a YouTube video featuring a clock with real Nixie tubes. Most of the tubes available today were manufactured back in the 70s and 80s for instrumentation, test equipment, etc. but apparently there has been a revival of these devices for digital clocks in particular so I thought why not go for it. The first challenge was locating the tubes but that actually proved easier than I thought thanks to eBay. I found four tubes from a Russian seller (type IN-8) which were labeled as NOS (new old stock) for $75.00. These are high voltage devices (upwards of 170VDC or more needed for activating the internals). To avoid having to scratch build a supply, I decided to purchase one (again from eBay). A detailed description of the boost converter kit I bought can be found at Threeneuron's Pile o'Poo. Below is the finished board:
Pretty nifty little device and it only takes up a bit more than one square inch of space.
Next came the components needed to control the clock functions and drive the displays. I chose an ESP8266 based WiFi module for the controller (ESP8266DevKitC-02D). The controller periodically retrieves the time from an NTP server, then converts it into four bytes (32 bits) and outputs it via an SPI interface to a serial to parallel logic converter (HV5530). This particular converter has 32 channels, each capable of handling up to 300V, so it worked perfectly for my application.
To the right is the controller and voltage regulator along with the various wiring harnesses needed to connect with the display circuit board. The blue section to the right of the controller is the WiFi antenna.
One minor issue cropped up when using this particular controller and driver and that involved their respective supply voltages. The ESP8266 is a 3.3v device while the HV5530 uses 12v. Therefore, to connect the two a level shifter is required. This was easily solved by using an N-channel MOSFET and two pullup resistors for each control line (shown below):
 |
Level shifter
|
One issue that all Nixie tubes have in common is something called "cathode poisoning". As the individual digits (cathodes) are turned on, they emit microscopic particles which can coat the unlit elements. Over time, enough material can coat the other cathodes to the point that they start to go dim. To reduce this effect, an anti-poisoning routine is needed in software. By turning on each digit in rapid succession over a given period of time, the poisoning process can be minimized. So, prior to turning the tubes off for good each night, I run my routine for sixty seconds, then shut down (I decided to turn off the tubes at 11:00p to help lengthen their useful life which is stated to be in the neighborhood of 10,000 hours. I have read, however, that the posted numbers for these tubes is very much on the conservative side but why take any chances?).
A circuit with this many components required a custom printed circuit board so instead of rolling my own, I used EasyEDA software to create the schematic and then layout the components. The gerber/drill files generated by the software were then emailed to JLCPCB, a PCB fab house in HongKong. I received my board in less than 2 weeks (actually, the minimum number of boards that can be ordered is five but total cost including shipping was only $15.00 so no complaints from me). Here is an image of an early board created by the PCB software after I made an initial component placement:
 |
| Early PC board prototype |
For comparison, here is the actual board I received from the fab house along with a few components I soldered on:
 |
| Bottom view - partially populated PC board |
A couple of minor changes noted between the blue and green board images are due to updates I made to the schematic later on. I decided to add a neon tube between the hour and minute Nixie tubes to act as a colon and I also realized I didn't need one of the voltage level shifters after all. For what its worth, the 44-pin chip was a real bear to solder in. Not my best soldering job but it works. The finished board is seen below (I left off one tube to show the backlight LED found under each tube:
 |
| Top view - showing a backlight LED |
To wrap things up I decided to make my own enclosure and chose maple and walnut because I thought the contrasting colors would look nice. All pieces came from 1/4" stock.
I used double sided carpet tape to attach the top piece and screws for the bottom piece. A clear finish of linseed oil was then applied to protect and beautify the wood.
 |
| Time (designated by colon off) |
 |
| Date (designated by colon on) |
 |
| Display elements highlighting differences in depth |
 |
| Rear view |
Update (Sep 2023)
Currently, the display is only on between the hours of 8:00a and 11:00p to prolong tube life. This is true regardless of whether the room its located in is occupied or not. I decided to go one step further in my quest to improve tube longevity by adding a light sensor. This was easily achieved by adding a photo detector and resistor combination connected to the controller's ADC input. A low light condition results in a low voltage reading (ie. counts) on the analog input while a high light condition results in a high voltage reading. Now, whenever the room light is turned off (indicating the absence of a person) the display will also turn off. Conversely, when someone enters the room and turns on the light the display will turn on.
