Nixie Watch, Part 1: Development Board

This is a multipart instructable. The development board presented here is not a final project, but draft work presented in hopes of helping others (and possibly some collaboration towards a better final project).

Two ultra fabulous nixie watches currently grace the web. The first was created by Jeff Thomas ( http://www.amug.org/~jthomas/watch.html ). This watch has four small end-view tubes for time display. The nixies are powered using a camera flash circuit and an AA battery. Two button cells power the clock logic. This watch is logic based, and does not contain a microcontroller. David Forbes at the Cathode Corner has the only nixie watch currently available for sale ( http://www.cathodecorner.com/nixiewatch/ ). This watch uses two small side view nixies to display the time. Time keeping is done on a PIC microcontroller. High voltage for the nixie tubes is provided by a boost converter IC and transformer. This watch is powered by a single 'CR2' lithium camera battery.

Both of these designs employ much surface mount mojo. Consider the picture of Jeff's nixie watch: seven surface mount ICs and a dozen or more surface mount transistors. The goal of this instructable is to create a prototype 'poor man's' nixie watch using as few surface mount components as possible. In Part 1:Development Board, we will consider various design problems while creating an all through-hole version of the nixie watch. Part 2: Surface Mount Mojo, will present the final design incorporating a few surface mount components to save space.

The watch display is a single IN-2 Russian nixie tube. These are not the smallest nixie tubes available, but I have a big bag of them on-hand:
These tubes are cheap, I think I paid ~0.50 cents each.
These tubes are readily available from tube vendors and Ebay.
These tubes are socketed, rather than soldered, for easy replacement.
These tubes will work at lower voltages (~140 volts @ 0.9mA).

This is a tall tube (its longest dimension being height), so the design will use height wherever possible to save space. The Eagle socket footprint was created using Nick de Smith's value socket calculator:

( http://www.desmith.net/NMdS/Electronics/soculator.html )

Drivers connect a cathode (a number inside the tube) to ground, causing the connected digit to glow. Driver choice is always interesting. Three options are prevalent in nixie tube design. First, ten NPN transistors (one for each cathode/digit) can be connected to the tube pins. A microcontroller pin connected to the base of each transistor is used to light the appropriate digit. This driver is seen in Jeff's nixie watch – notice all the surface mount mojo in the picture of the PCB. To learn more about using discrete transistors with nixie tubes, check out this excellent tutorial by Jason Harper:

( http://home.earthlink.net/~jasonrandharper/NixieTransistors.pdf )

David Forbes uses a Darlington driver IC in his nixie watch. This is a modern IC that contains 8 transistors, and is similar in function and use to discrete transistor design discussed above. As I understand, this IC can't cope with the full voltage from the nixie cathodes. To get around this problem zener diodes are placed on each input pin to clamp the voltage at an acceptable level. I believe David used this approach, but I can't seem to find a citation. Assuming he did use these diodes, it added 16 tiny tiny surface mount zeners to the design.

The third way to drive nixie tubes is to use an antique BCD (binary coded decimal) chip. These chips are designed to drive nixie tubes and have no problem coping with the voltages involved. Four pins on the BCD chip receive input from the microcontroller while 10 additional pins connect to the 10 nixie cathodes. The input pins are interpreted as a binary number and cause the corresponding cathode pin to be grounded. For example, if all four pins input pins on the BCD are low (representing 0/0/0/0 in binary, 0 in decimal) then the BCD pin corresponding to '0' will be grounded, causing the '0' cathode in the nixie tube to light. Common BCD chips include the 74141 and KD155 Soviet equivalent.

The Soviet KD155 was chosen for this design. These chips are cheap (~0.50 from tube vendors and Ebay) and require no additional components. The down side is that these old chips only come in a DIP-16 package. I think thats OK – I really like Russian lettering on the chip and consider it as much a curiosity as the nixie tube itself. It takes a little extra space, but adds additional novelty and eliminates oh-so much surface mount mojo.

A PIC 16F684 micontroller is at the heart of the watch. This uC keeps time, runs the KD155 BCD/nixie display (previous step), and controls the power supply (next step). This is a 14 pin part available in DIP and SOIC (.150) packages from Microchip.

For development purposes we use the calibrated internal oscillator running at 125Khz to count seconds. The final design will use a precision watch crystal to save power (see final step & part 2).

The power supply has been the most difficult and evolving part of the design. Nixie tubes require high voltage, easily generated with an inductor coil as in my nixie switch mode power supply:
( https://www.instructables.com/ex/i/B59D3AD4E2CE10288F99001143E7E506/ )

Things get harder when we want to run from batteries at a low voltage. There are several dedicated boost converter chips that can be used to generate the required voltage, David Forbes uses the LT1308B. I avoided using such a chip because it adds an additional (surface mount) component. It also promotes a vendor specific design that doesn't contribute to understanding what actually goes on in our SMPS. Any uC can drive my SMPS without major modifications.

I used the nixie power supply from the previous instructable as a starting point. One input (100uF/16V) and one output filtering capacitors (1uF, 250V, high-temp) are used. At such a low load (1 tube at ~ 1mA) nothing more is required. The supply voltage sense resistors are eliminated – the watch will not measure the supply voltage, it will be hard-coded with ideal values. The rest of the design is exactly as presented in the SMPS instructable: IRF740 FET, BYV26C diode, 100uH inductor, and TC4427A FET driver. The 16F684 uC drives the SMPS.

The rub -
My first draft design used a small 12 volt alkaline battery (V-23). I though -WOW- I'll use this small cell with a zener to provide 5.1 volts for the PIC and feed the SMPS directly from the battery @ 12 volts. Nope. The V-23 is probably a stack of 10 x 1.2 volt button cells, but I haven't opened it to be sure. These cells cannot provide enough current for the circuit – even without a nixie tube installed. Voltage from the battery fell to under four volts at any significant amount of current. The IRF740 FET needs a bare minimum of 5 volts to switch at all. Efficient switching is more in the order of 10 volts. At four volts the SMPS FET is never actually switched, thus no high voltage is generated.

The fix -
I looked over David's design and noticed the CR2 3 volt lithium battery he uses. A quick googling showed that this cell is intended to power a camera flash – that means plenty of current is available. This cell is much bigger, but should have a longer life than the V-23.

The rub (part II) -
The CR2 is only a 3 volt battery – we have the same problem as before, nowhere near enough voltage to switch the FET. I was feeling awfully confident that I could come up with a solution. I tried to 'boot strap' the FET – feed a little of the SMPS output voltage to a simple 15 volt resistor/zener regulated supply, then use that voltage to switch the FET through the TC4427A FET driver. No go. Adequate voltage didn't accumulate, the FET didn't switch.

The fix -
Where do we go from here?
1. Redesign the power supply using the guts of a disposable camera flash. This could work, but I would like the option of buying new, standard components.
2. Give up and use two batteries. The supply works fine when using the CR2 for supply and the V-23 (12 volt) battery just to switch the FET. 2 x 3 volt coin cells work well, and would fit nicely under the larger CR2 battery.
3. Use a DC/DC charge pump to double the CR2 output to 6 volts. This is an appealing option. Only a single battery is needed, but an additional (surface mount) IC, 2 capacitors, and 2 diodes are required.

This development board was made to evaluate the last two options. It can test various button cell batteries to determine their usefulness as a FET driver supply. It can also be used to test the suitability of a DC/DC charge pump as the FET drive supply. The charge pump has its own caveats that might rule out its usefulness (dubbed: the rub part III). From the Microchip TC7660 datasheet:

( http://ww1.microchip.com/downloads/en/DeviceDoc/21465b.pdf )

The voltage thus created...becomes (2 V+) – (2 VF), or twice the supply voltage minus the combined forward voltage drops of diodes D1 and D2.

So the CR2 @ 3 volts and 2 good diodes with an (ideal) 0.6 drop:
4.8 volts=(2*3)-(2*0.6)

Thats might just barely do the job. Perhaps different diodes might help, or cascading two charge pumps. This all makes dual batteries seem like a small inconvenience for the space and complexity saved.

1. Evaluate FET switch supply: A charge pump will be tested. There is a certain je nous se quoi about the single battery design that almost out weighs the complexity. On the other hand, the simplicity of dual button cells (and space savings) is elegant in itself.

2. Design Interface: The watch needs to be activated by a button if the battery is to last more than an hour or two. A touch sensor based on a Darlington transistor would be cool. The interrupt pin of the uC is brought out to a pin on the development board so that various interfaces can be tested.

3. Move to surface mount: There is a lot of space on the PCB under the battery. A .150 SOIC PIC 16F684 and TC4427A fit nicely. The SOIC package seems like a reasonable size, its not a tiny TQFN or anything. All other components will still be through hole. Most parts will be readily available, the others would likely have to be ordered anyways. Moving these two ICs under the battery saves enough space that we can really start to call this a watch (perhaps a pocket watch?).

4. Add watch crystal: An external crystal (specifically the 32.whatever kHz watch crystal) will be more accurate than the internal 32kHz crystal, and use less power than the calibrated internal 8MHz crystal. It might also be used on a counter pin to increment the PIC timer while it is in sleep mode. The timer interrupt could wake the PIC to increment the seconds, but otherwise be in a VERY low power state.

Project files:
These files are very draft-y, the archive includes:
Development board in Cadsoft Eagle format.
A ton of footprint libraries that go with the PCB (socket, inductor)
Bare-bones SMPS firmware. Enough to run the power and display a digit on the nixie.