The objective of the No-Si-Clock is a digital clock with nixie tube readout in Valve/tube technology. I have used absolutely no silicon chips or transistors, let alone a microcontroller. I have used a few germanium diodes to keep the tube count acceptably low. The 1Hz frequency is derived from the mains frequency (50/60Hz). Frequency division is based on vacuum and gas-filled tubes. I am using 31 tubes in this project. 

The block diagram contains: 

This is now a finished, working project. The schematic diagram is available in the Files section (Kicad). Also the physical design of the aluminium base plate is avaible there (FreeCad). Check out the video below. 

I have done numerous clock projects in the past, some with nixie readout and some with other technology. I wanted to challenge myself and decided to go No-Si.

The design encompasses seven dekatrons, which are the counting elements. A dekatron is a neon-filled tube, with one anode and a great number of cathodes (typically 30), arranged in a circle. At any moment in time, a discharge is present at one of the cathodes and as long as nothing happens, the discharge continues on that cathode. This makes the dekatron a memory element. Ten of the cathodes represent a stable state, while the other 20 are transient. These 20 are connected together in two groups of 10 and by applying negative impulses to these groups, the neon discharge can be transferred from one to the next stable cathode. This makes the dekatron a counting element. The dekatron also has one of more outputs, each of which generates an impulse once per revolution.

One dekatron is used to divide the mains frequency by 5 and then by 10. This results in 1Hz.

There is another type of dekatron, that can count to 12 and can also be used to divide by 6.

Starting with 50 Hz (or 60 Hz) mains frequency, the dekatrons will do the following.

Dekatron type 6802 (or GS12C) divides by 5 (or 6) -> 10Hz

Dekatron type 6802 divides by 10 -> 1Hz... Hurraah!

Dekatron type 6802 divides by 10 -> one impulse per 10 seconds

Dekatron type GS12C divides by 6 -> one impulse per minute

Dekatron type GS10C divides by 10 -> drives the minute units and outputs an impulse per 10 minutes

Dekatron type GS12C divides by 6 -> drives the 10 minutes and outputs an impulse per hour

Dekatron type GS12C divides by 12 -> drives the hours

Fig. 1. The "one beat per minute generator". Mains frequency 50Hz is divided by 5, 10, 10 and 6.

The first stage divides the mains frequency (50 or 60Hz) by 5 (or 6). In principle, this is just a 'dekatron spinner' circuit, like the ones that can be found on the internet. But the difference is that the output impuls from 2 of the dekatron's cathodes is taken, amplified and inverted by one of the triodes in a E90CC double triode. For each rotation of the spinner, two such impulses are generated (10Hz) and fed to the second stage, which divides the frequency by 10, which gives a one Hertz (1 Hz) impulse.

The stages that drive the nixies, pose an extra problem. The outputs of the dekatron are not suitable to drive the nixies just like that. The signals need to be amplified and inverted. This will be handled by the E90CC's. These double triodes were specifically designed back in the day for computer and other digital purposes. They are unsuitable for audio applications, among others because the suffer from microphony, and that's very nice because that keeps the price down. I was able to lay my hands on the required number of E90CC's for €2 each.

Fig. 2. Driving a nixie tube from a dekatron requires 10 triodes.

If you study figures 1 and 2 carefully, you will notice that the coupling between the stages is done using a 2.2nF capacitor. This is to decouple the high anode voltage from the rest of the circuit. This leads to short impulses. This is typical of digital tube technology and different from what we are used to in transistor and IC technology. Nowadays dc-coupling, without a a capacitor is the norm. An added advantage of ac-coupling in this circuit is that the power consumption (in the triode and its anode resistor) is lower, because the triode only opens briefly.

However, the circuit that drives the nixie tubes in fig. 2 cannot rely on ac-coupling, because the nixies have to stay on longer than the duration of an impulse. therefore, the coupling of the dekatron to the triodes and the triodes to the nixies is relized without capacitors.

There are four nixie tubes, but I will have only three dekatron stages driving the nixies. I have chosen for a twelve-hour clock and I happen to have a few dekatrons with 12 steps instead of the usual 10. So for the hours I use one of these and it drives both the hour's units and tens. For the '11' and '12' I use some diode logic with germanium diodes in order to drive both the '1' (in the leftmost digit) and the '1'/'2' (in the unit hours digit).

Fig. 3. The two nixies for the hours are driven by just one dekatron that counts to 12. Some clever germanium diode logic takes care of the rest. Later I decided that D4 and D6 are not really needed (can be shorted), so they have gone out of the window.

Quite late in project, I realized it would be handy to set the time after starting the clock. Therefore I added a double pole, three position (DP3T) rotary switch. In the middle position, the clock works normally. In the left position, the input to the minutes section is rerouted to the 1Hz signal. In this way, the minutes can be adjusted. In the right position, the input to the hours section is rerouted to the 1Hz signal, so the hours can be set. In the junk drawer labeled knobs, I found a knob with the inscription Time adjust. I thought: bingo!

A note about the power supply for the filaments. The circuit consumes 80W. Most of this is used for heating the filaments. This is why there is a 100VA transformer for the filament voltage. I am using mostly E-series valves, like the E90CC. These all require 6.3V. The 100VA transformer that I found, has secondary voltage of 25V. By stringing 4 E90CC's in series, this gives 6.25V per tube, which should be fine. So four of these strings and I can supply the filaments of all 16 E90CC's. But in my circuit were also 3 EY88's (rectifiers), also requiring 6.3V each. Initially, I just put 4 EY88's in the circuits (so one dummy) and put them together in an extra string. But it felt like a waste of energy (10W per EY88, against 2.5W for a E90CC). And besides, I needed the room that the dummy rectifier took, to give way to the switch for setting the time.

Then came the ide to replace the EY88's by PY88's. Same specs, only difference being the required filament voltage of the PY88 is 26V. I guessed they would work on 25V too, so I could wire these in parallel, directly to the transformer. Then I could get rid of the dummy rectifier and make room for a switch. Two problems solved in one go!

In hindsight, maybe the xY88 was not the right choice as a rectifier anyway, because of its huge power consumption. E.g. the EY91's maximum anode current is lower, but would have also done the job, and consumes only 2.6W. And it has the same socket as the E90CC, and no top cap, so maybe esthetically better. But then again, there is no 25V equivalent for this tube. Thought UY92 for a second, but it's limited to 145V. Would have been good for the -12V supply, though. Other alternatives: 25W4, 25X4, 25Z3, 25Z4, 25AX4