Pictures

Video

Schematic

Circuit notes

Construction/mechanical notes

Testing and results

Operation and service manual

Introduction

When it comes to hobby projects, clocks have always been one of my favorites. I've made them with LEDs, planar neon displays, monodigichrons, some with microprocessors, FPGAs, TTL logic, etc, but this time I wanted to make a dekatron and nixie tube clock with a minimum of solid state components, except for maybe a few germanium diodes. As I thought through the design, the cold war theme developed along with some specific goals I had for the project:

• All-tube clock, no semiconductor devices
• Research every part, and use only components and technologies available in 1959
• Authentic military/commercial reproduction, not a unique or artsy style
• Reasonable size, less than approximately 1 cu ft
• Use a mix of Soviet and US components in key parts of the circuit
• Audible alarm
• Low power consumption for tube equipment, less than 60W
• Selectable 60 cps or crystal reference with decent accuracy
• Use original old parts wherever possible
• Learn about how products of that era were designed and manufactured

I chose the design year to be 1959 mainly because I wanted to use Sprague orange drop capacitors, and they were not introduced until that year. I like orange drops because they're more compact and reliable than wax and paper caps, can still be found in old stock, and I like how they look. Also, 1959 was good because the cold war was going strong then, and I had some parts donor gear from that era.

Design, build, initial testing, and documentation took about a year and a half. It was a balancing act of compromises to meet all the goals, but I'm satisfied with the results.

Components

You don't see them much anymore, but in this clock dekatron tubes are probably the most important components because they are the register elements that store and count the current time. Dekatron tubes were introduced the early 1950s and were often seen in particle counters for atomic research. They were also used in some lower speed base-10 computers (Harwell Witch for example), but faded out in around 1960 when solid state binary-based computing took over.

Here's list of the electronic parts and how many of each are used. The tube types are listed after the tube count. Vacuum tubes have heaters/filaments in them, and the cold cathode tubes are filled with a gas like neon and don't have heaters. The use of cold cathode tubes wherever possible is a big part of why the power draw in this clock is low.

I thought it would be fun to make a list of manufacturers and country of origin of the components I used (S-Soviet, B-British). Many of these names bring back fond memories of the electronics I worked on when I was a kid:

Circuit notes and operation

Here are some notes on the circuit and components used. I won't repeat it here, but for a description of how the circuit works, look at section 3 in the manual. The same goes for what all the switches do and how to use the clock. It's more fun to read it in its original 1950s format anyway.

The transformer, which came from a TS-382 signal generator, is oversized but fit the application better than anything else I had or could easily find. The secondary voltages were a bit high, so I bucked the primary with an unused 6.3V filament winding.

The filaments are run close to 5.5V instead of 6.3V because the emission is still more than high enough even at low line, and I think this will dramatically improve the lifetime of the tubes. It will be interesting to see.....

I decided to run the crystal oscillator at the low frequency of 36 kc because it's a multiple of 60 cps, which allows power line operation without extra or redundant frequency dividers. I went to a fair amount of effort to keep power dissipation in the crystal as low as possible for better stability and aging characteristics, and make the oscillator still run reliably. A higher frequency crystal, like a 1 mc AT-cut would have better stability than the one I'm using, but the additional frequency dividers would have exceeded my budget for size and power consumption.

A lot of these Soviet low frequency crystals in 7 pin glass tube packages like the one I'm using, have 3 connections to the crystal, but I don't know why. If you know the story behind this, or what they were originally used for, I'd be interested in hearing about it.

The oven I put the Soviet crystal in, I was told by the seller, is from a Nike Hercules IFC (Integrated Fire Control) system! Nike Hercules was a pre-ICBM anti-aircraft missle system with nuclear capability. The oven originally housed a zener diode string and pot, evidently used as a precision voltage reference.

Multivibrators were the typical circuits used to do division at these frequencies, but impressed by what I read in this article, I decided to try the cathode-coupled injection locked oscillator approach. Initially I had high hopes of dividing by 60 in just one stage. It worked, but due to the tuning, input drive level, and power supply sensitivities I saw, I felt it wasn't robust enough to use in "production".

Instead of going with all neon dekatrons, I used an argon OG3 at V11 instead of an A101 because it seemed more reliable running at 600 cps. It works well, and the octal socket is smaller and more convenient than the big 13 pin socket for an A101.

I mounted V15, the 10 cps dekatron, on the front panel because I ran out of room on the chassis, but I like it there because the "spinner effect" looks really cool. It turns out there is a low-tech way of using this 10 cps counter: I hold a piece of paper with a hole punched in it over the dekatron so I can see only one cathode through the hole at a time. At the same time I listen to the tick from WWV or look at the seconds on another clock in my peripheral vision. By moving the hole in the paper from cathode to cathode to find the closest match of dekatron flash to seconds tick, I can estimate the clock time to within 50ms.

The subminiature tubes used were developed by Raytheon during WW2 for an above-top-secret project: the proximity fuse. They're made to be fired in an artillery shell and can withstand 20,000G of acceleration

Most (52 of the 94) tubes, and about 75% of the resistors in this clock are there only to provide the nixie tube display. A dekatron-only clock would be much less complex, but harder to read.

Electrical safety was in its infancy in the 1950s it seemed, and grounding of equipment was optional or nonexistent. The power plug I'm using, with its swing-away ground prong, was invented in 1956 by Milton Morse, holder of over 100 patents and founder of APM Hexseal. Evidently it was popular on Air Force equipment, and the NSN number still shows up in the database.

Some of the parts I used have been in my junk stash for years, and have a personal history and memories attached to them. For example, I remember purloining the the pushbutton and toggle switches out of an Electronics Associates, Inc. Pace 261 analog computer back in 1979. I think the BNC connector, fuse holder, and probably some other parts came from that analog computer too.

Construction/mechanical notes

The power supply, oscillator, and most of the frequency divider circuitry is mounted directly on the chassis, everything else is on phenolic terminal boards mounted either above or below the chassis. I bought blank terminal boards and swaged in the Keystone Electronics #1559-2 turret terminals where they were needed using the Keystone TL-2 punch tool in a drill press. I bought 600 terminals and had only 4 left when I was finished. Not only is it hard to believe there are that many terminals in the clock, but also that about 500 feet of wire was used.

I bought a supply of Soviet A101 dekatrons to use for the project, mainly because they were the cheapest ones I could find that are register tubes, meaning that all the cathodes are available externally. I soon discovered why they were cheap; the sockets must be made of nearly pure unobtainium. My solution was fire up the lathe and make my own sockets out of 1.5 inch black delrin rod and Molex connector crimp terminals. I used terminals I already had on hand, and unfortunately I don't know the Molex part number for them. They look similar to the 01189 series, except the overall length is only 0.77", which is good because it makes for a shorter socket. Here's a drawing showing the dimensions of my homemade A-101 sockets.

Choosing a type and size of cabinet for the clock was a tough decision. I wanted something that would show off some of the inner workings, but it had to have a military or commercial look and be authentic from that era, so anything home made with wood or plastic was out of the question. I had done some prototyping to get a pretty good idea of the amount of circuitry needed, and therefore an estimate of the volume required, so when I saw an unused but nicely aged Bud C-975 cabinet with a hinged top and wrinkle finish on Ebay, I jumped on it. As a side note, I found the C-975 in the 1959 Allied catalog for $9.01, or only $7.20 if you bought 50 of them. As you might guess, fitting 94 tubes and all the other parts into a 9x15x11" box got a bit tight in places, but that was one of the challenges that made the project fun.

I wanted an aluminum front panel with engraved and paint-filled markings like the old military and HP gear, so I had it made by Front Panel Express . At the same time I had them make the sheet metal panels that the nixie and dekatron sockets are mounted on. I was very happy with the results and the cost was reasonable, about $240 for all three pieces. I doubt they're of any use to anyone, but just in case you're curious, here are the design files for the front panel and sub panel.

The rest of the aluminum parts I made myself, borrowing the use of sheetmetal brake, shear, and corner notcher when needed. Since I didn't have a 2" punch, I made the holes in the chassis side panels with a flycutter in a drill press. The 3" hole for the power transformer was made the same way. Making holes with a flycutter like this is a bit dangerous, and all the cutting oil flying around made a mess, but the results were great.

I wanted a yellowish/gold passivated finish which is electrically conductive applied to all the aluminum parts because that is typical of what was, and still is, used on aluminum electronic chassis. I found a local plating shop that could do the chromate conversion process (aka Iridite or Alodine) using hexavalent chromium. I'm glad I was able to get the authentic toxic process rather than the lame ROHS version that will probably be the only one available some day.

Military gear of this vintage usually had hardware plated with yellow cadmium, and I used it too wherever I could. At prices of $2 a screw or something like that (if you can find it at all), I didn't go out and buy any "new" cad plated hardware, but instead used what I could scrounge from my junk box.

I've always liked the way high quality vintage electronics had the component reference designators printed on both sides of the chassis, often sealed in with a clear or yellowish coating, applied either through a rectangular mask, or sloppily brushed on. That look was a must-have for the clock. Since the chassis had to be formed and plated before printing, there was no way I could think of to use a silk screen, at least not on the inside of the chassis. I decided to use a rubber pad printing set, the kind where you assemble individual letters in a holder, press on an ink pad, and print on the chassis. I used a StazOn permanent ink pad for this. When all the printing was done I used blue painter's tape to mask off everything except a small window around each reference designator and then sprayed it with urethane. It was a very time consuming process, but it looks authentic and serves the intended purpose of protecting the markings.

Cost notes

I wanted to know how much it would have cost for a hobbyist to build a clock like this in 1959, and what that amount would be in today's dollars. I gathered my information from a variety of internet sources, but my favorite was the old Allied Electronics catalogs at alliedcatalogs.com.

Parts cost in 1959 (from old catalogs and my estimates): $2130
1959 parts cost inflated to 2013 dollars, (inflation factor 1959 to 2013 = 8.0): $17040

The median income of a fairly well paid "technical worker" (engineer) in 1959 was $8600.

If you're interested in the component cost data I used to come up with the total, it's in this text file.

So at a cost of about 25% of a year's salary, it's unlikely that a clock like this would have ever been built by a hobbyist - at least not a hobbyist who was still married by the time it was over! What was my cost in 2013? About $1600, low enough to keep peace in the house.