If you want to reproduce this project, that's great! But before you begin, beware of a few things.

This projects involves voltages up to 400V DC!

If you are afraid of high voltages or not used to working with it, don't do this project. It is a little bit dangerous. I accept no responsibility for any accidents that occur from attempting to reproduce (parts of) this project or otherwise. Do anything at your own risk.

It is quite a difficult build. If you have no experience with tubes/valves, forget it.

You will need a digital oscilloscope with at least two channels. I advise against an analog scope, because some of the signals are not repetitive.

This clock is a show piece, a conversation starter, something to boast and/or brag about, a novelty or whatever you name it, but it is not something to turn on 24×7! It consumes 80W, heats up your room and the tubes have a life expectancy of a little more than one year.

Do not let children and/or adults get near. There's high voltages everywhere, the thing gets hot, it's heavy and fragile.

The design depends on the mains frequency. Mine is built for 50Hz. If your mains frequency happens to be 60Hz, don't worry. You just replace U1 (6802) by U5 (GS12C) and the first stage will not divide by 5, but divide by 6. Problem solved. This does mean you need one extra GS12C instead of a 6802, though.

I built this project without a pcb, in the old style of tube technology with so-called point to point wiring. I regret this decision. Point to point wiring is soo laborious, tedious, error prone and what not. So if you have the chance, design a pcb!

Dekatrons with 10 steps, like the 6802 are not hard to find, in my opinion. But:

The 6802 has not all of its 10 cathodes wired to separate pins on the base. The cathodes 1, 2, 3, 4, 6 and 7 are tied together. This is OK for in the 1BPM generator (the prescaler that converts 50Hz into 1/60Hz). But this makes the 6802 unsuitable in the unit-minutes section, because we need all cathodes wired to seperate pins in order to drive the nixie. This is why I chose the GS10C for the unit-minutes. I happened to have only one of those and more 6802's, so that is why 6802 are still in the circuit.

I use GS12C's in three places. This is a dekatron that counts to 12 and has all cathodes individually exposed. I guess this thing is hard to get. I was just lucky to find a handful of these on the local marketplace a few months ago. They have no socket, they have lugs that you can solder directly to. A suitabel alternative is the CV1740 or GS12D, but I'm afraid these are also made of unobtainium.

An alternative would be to use normal 10-step dekatrons and then design a reset mechanism, so that it jumps back to 0 after hitting the 6. That is feasible. You will end up with more tubes and I guess more is betterer.

The E90CC double triodes on the other hand are easily obtainable and dirt cheap (I bought a lot at €2 each). Why? The E90CC was designed "for use in computer circuits" and is perfectly unsuitable for audio applications because, as the datasheet clearly indicates: the E90CC is "not intended to be used in circuits critical as to hum, microphony or noise".  One cannot thank the engineers at Philips too much for these built-in features!

About the transformers. I found a suitable 25V 100VA transformer in my junk bin. There are two additional transformers needed: a small 9V transformer for the -12V line and a 2x110V transformer for the 200V and 400V DC lines. If you can't find these, nowadays you can have a transformer custom built on aliexpress for next to nothing.

Think about your mechanical construction first. I mounted all the tube sockets in an aluminium base plate and put all the other components and wiring under the hood. I devised a hinge mechanism using a metal rod from an old ink jet printer to make the hood easily opening, to be able to work on the internals. As already said, next time I would make a pcb, for point to point wiring is terrible.

Build the power supply section first. In the beginning, the load is still low, so as it is an unregulated power supply, all voltages will be a little bit on the high side. I used a variac to compensate for this. It is important that the V-HT is always exactly 400V (+/-5V). Measure this voltage regularly and correct if necessary. I inserted R74 (1k5) on sheet 'power supply' to bring down the V-HT a little bit. You may have to tune R74. The -12V can deviate a little bit, but not too much. The nixie voltage (200V) isn't that critical.

When you have the power supply running, the next thing to do is make the 'spinner' U1/U5 work. This dekatron makes five full circles every second. States 0 and 5 generate a 10Hz positive going impulse train that goes to gate 1 of double triode E90CC V1 . You will need an oscilloscope to verify that the Vpp equals about 10V. Also verify that on the anode 1 of V1, there is a similar pulse train, but negative going and with a much higher Vpp (>100V). See figure below. Yellow is pin 0 of the dekatron, blue is the anode of the triode.

Build the circuit around R10/R11/C18/C3, connect to G1/G2 and check the results. Both should give a negative going pulse, but one shifted in time a few millisconds with respect to the other. This is what makes a dekatron step to the next stage. See here an oscilloscope view of these pulses.

Yellow is G1 and blue is G2. Both are negative-going pulses of 40V or more. First, G1 is the most negative. After 100μs or so, G2 is the most negative. This causes the dekatron to take a step.

These circuits are repeated with slight variations, with each stage of the clock. However, while the first four dekatrons are AC-coupled (through a 4.7nF capacitor) from their cathodes to the grid of the triode, the remaining three are DC-coupled. This is required, because the triodes V5 etc. drive nixie tubes. Without DC-coupling, the nixie would only light up for less than a millisecond.

Ouch, I spot an error in the circuit diagram. R43 should not be connected to HT (400V) but only to UNIXIE (200V). Otherwise, the germanium diode D1 will not survive! Same as R60, which is drawn correctly.