Nixie Tube Clock

Introduction

It seems nixie tubes have had somewhat of a revival recently with DIY clock kits becoming available on the usual marketplaces (Ebay, Aliexpress, Etsy, etc). While the kits commonly use surplus Soviet-made nixie tubes, the driving circuits employ modern electronics; Microcontrollers, shift register ICs and a boost converter to generate the supply rail for the tubes.

I thought it would be fun to build a nixie clock that used period-correct technology for the rest of the electronics - ideally only with cold-cathode gas tubes like the nixies themselves. Turns out I'm not the first to have this idea. one example being Sgitheach's first trigger tube based nixie clock. My design is heavily inspired by theirs, with timing derived from the 50Hz AC mains frequency and a series of trigger tube ring counters driving the nixie tubes directly. I sort to reduce the number of trigger tubes needed since they are becoming harder to obtain and ended up independantly coming up with a circuit design similar to Sgitheach's later nixie clock designs

The Trigger Tube

A trigger tube is type of cold-cathode gas tube (like a neon bulb) with a third electrode placed between anode and cathode. This third electrode, called the trigger, is used to ionise the gas between it and either anode or cathode, reducing the breakdown voltage between anode and cathode. A trigger tube can therefore use a low voltage, low current signal to switch a higher voltage, higher current source. Once triggered, the only way to deionise the gas and cease conduction is for the voltages between all terminals to be reduced below the holding voltages.

I selected MTX-90 trigger tubes as these are readily available and inexpensive due to being ex-soviet military surplus.

With the trigger electrode floating, anode-cathode breakdown voltage is >200V and with the trigger connected to cathode, >150V (thus the anode-trigger breakdown being circa 150-200V). The trigger-cathode breakdown voltage is around 80-90V however since the gas takes time to ionise the peak voltage of a short trigger pulse may need to be higher than this. During trigger-cathode ionisation, the anode-cathode breakdown voltage can be reduced to circa 100V. The holding voltage for both anode-cathode and trigger-cathode is around 60V, thus to deionise the tube either both anode and trigger need to be reduced to below 60V, or with the trigger floating/high-impedance only the anode needs to be reduced below 60V.

Trigger Tube ring counting circuit

The core logic of the clock is implemented with a series of ring counting circuits. These use a chain of n trigger tubes which are sequentially ionised, thus each ring can count from 1 to n (or 0 to (n-1)).

All trigger tubes in a ring are common-anode connected via a resistive dropper, Ra, to the positive supply while each cathode has its own capacitor decoupled cathode dropper network, Rk1, Rk2, Ck. When power is initially applied to the circuit, the supply voltage is sufficient to cause anode-cathode breakdown of the tube with the lowest breakdown voltage, without application of any trigger signal. This in turn pulls down the common-anode voltage low enough that all other tubes are below their untriggered anode-cathode breakdown voltage, preventing more than 1 tube from being ionised concurrently.

The cathode decoupling capacitor, Ck, of the ionised tube then charges up, limited to a voltage determined by cathode dropper resistors, Rk1, Rk2. The voltage dropped across the cathode resistors (~100V) in series with the holding voltage of the tube (~60V) brings the common-anode voltage high enough to allow a triggered anode-cathode breakdown of the other tubes but not high enough for untriggered breakdown. Trigger pulses are generated by AC-coupling a square-wave clock applied at the IN port via capacitors Ct. The amplitude of the clock signal is scaled such that the AC-coupled pulses by themselves are insufficient in amplitude to trigger the tubes.

A DC bias voltage tapped from the ionised stage is applied to the trigger of the following stage via Rb such that the DC bias voltage together with a positive AC-coupled pulse is sufficient to trigger only the tube in the following stage. Rt provides current limiting to the trigger electrode and prevents current sourced or sunk though the trigger electrode back-driving the previous stage. When following stage's tube is triggered it temporarily pulls down the common-anode voltage (until its Ck charges), extinguishing the previous tube. With each successive clock, the tubes are ionised sequentially. Output clock signals can be tapped from any of the cathode terminals, providing a division-by-n of the input clock signal with 100%/n duty cycle. It is necessary to reduce the value of Ck when tapping off the cathode to drive a capacitive load (input of the next ring counter), such that the overall cathode capacitive load (Ck + Cload) remains matched between stages, otherwise the previous stage may not extinguish and the counter will enter a deadlocked state with multiple tubes ionised concurrently. The purpose of using two cathode resistors is that is allows a higher voltage output signal, while providing a lower voltage signal to bias the following stage.

Driving a nixie indicator tube with a ring counting circuit

A nixie indicator tube can be combined with a ring counter, providing both clock division and displaying the state of the counter on the nixie tube. Sgitheach's design used an extra trigger tube to buffer the signal from the ring counter and light the nixie tube. I wanted to eliminate the requirement for extra trigger tubes and also the flickering caused by running the nixie tubes from an AC power supply.

The nixie tube, having a common anode and seperate cathode for each digit can be placed in series with the trigger tubes of a ring counter circuit, however, this introduces a few additional nuances which must be optimised for the circuit to work.

Firstly and most obviously, the supply voltage must be increased in order to breakdown both the nixie and trigger tubes in series - in practice we can get away with a little less than the addition of the nixie breakdown (~180V) and trigger tube breakdown (~250V), since one tube will breakdown before the other we only require one breakdown voltage in addition to the holding voltage of the other, therefore around 350V is sufficient.

Secondly, due to the physical arrangement of the electrodes inside the nixie tube, the unlit cathode electrodes float at a voltage somewhere between that of the lit cathode and the anode, since the unlit cathodes are physically positioned between them. This has the consequence that the voltage of one or more of the unlit cathodes may float high enough to cause an undesired breakdown of the connected trigger tube. To mitigate this, balancing resistors (Rbal) are connected to a common rail between all tubes to bring the voltages of the lit and unlit digits closer together therefore allowing them to be treated as a quasi-common cathode. The value of balancing resistors must be chosen to both limit the anode voltages of trigger tubes on 'unlit' digits as well as prevent too much leakage current being conducted through 'unlit' digits in the nixie tube such that they noticably glow giving a ghosting effect to the display. Simultaneously, we can also reduce the voltage dropped across the cathode resistors (Rk) to further reduce the trigger anode voltage and prevent unwanted breakdowns, however, this also reduces the amplitude of trigger tube cathode voltages used as an output signal, thus a buffer circuit with higher sensitivity is required to drive the input of an equivalent ring counter circuit.

I experimented with connecting the balancing common-rail to a DC supply, in the hope this could reduce the 'ghosting' effect on the display and/or better limit the floating cathode voltages, however, the effect was insignificant versus just letting the common-rail float.

'Buffer' circuit

The following 'buffer' circuit is designed to take a 0-50V square wave or pulsed input (such as from a nixie ring counter stage) and produce a 0-100V output sufficient to drive another stage. The circuit acts as a set/reset flip-flop by the use of a 2-tube ring counter with seperate trigger inputs. The set (ClockIn) input is increased in sensitivity by changing the values of cathode resistors (Rk1, Rk2) in the other tube such that the DC bias is increased. The reset (ClockReset) input can be provided by a regular clock - 5Hz is used in practice. A positive output trigger is produced on the reset, with the output otherwise staying high. A momentary switch (SW1) which pulls down the output allows the user to manually produce an output pulse, for instance, to manually set the following counter. Output filtering provides bandwidth-limiting which in turn limits the amplitude of AC-coupled output pulses as well as providing some rudimentary switch debouncing.

Nixie ring counters with less than 10 counts (e.g. 6 counts for tens of minutes or seconds) can be created by omitting the trigger tube stages that are not required. It is, however, advisable that balancing resistors be populated for all of the cathodes in the decimal nixie tube, including unused digits.

Power supply

For prototyping I built a simple supply with 360Vac driving a silicon bridge rectifier and several regulated DC outputs provided by gas voltage stabiliser/regulator tubes in series.

Three transformers' (TR1-TR3) 120Vac secondaries are wired in series to obtain 360Vac which is rectified into a 430V DC bus. The 330ohm resistors limit inrush current during startup and improve power factor of the supply. Gas voltage stabiliser tubes (T1-T4) of type StR75/60 and 0B2 (regulating voltages ~78V and ~108V respectively) produce four regulated outputs at 108V, 185V, 293V and 370V respectively. R3 sets the bias current to around 27mA, which is sufficient to power the clock circuits (~20mA) plus the minimum current requirement for the stabiliser tubes (5mA). Gas stabiliser tubes are used similarly to Zener diodes whereby the voltage remains near constant though a range of bias currents. Since the 185V supply is not used in the final clock design, the power supply could be further optimised to need only three stabiliser tubes (e.g. 150V+150V+75V arrangement) and adjust the clock's circuits to suit the slightly different supply voltages accordingly.

I investigated continuing the theme of using all cold-cathode gas tubes and no silicon for the powersupply, however, adapting the design for a gas rectifier presents some hurdles. 'Cold cathode' rectifier tubes (e.g. a 0Z4G tube) do not achieve their full reverse-breakdown voltage capability until the cathode becomes heated by ionic bombardment. Accordingly, ionically heated rectifiers are specified with a minimum load current to ensure reliable starting and running. While our load of cold cathode gas tubes in the powersupply voltage regulation and clock circuitry do draw the minimum current at operating voltage, they do not draw any appreciable current until the supply voltages reach near-operational voltages. In practice, when gas rectifier tube is used with such a load it can conduct potentially damaging reverse currents, due to the high inverse voltages applied to the tube while the cathode remains cold. A circuit to draw a minimum load at lower voltages could be implemented, however not easily with only gas tubes. Additionally, in order to obtain the needed voltages (>=430V DC), two rectifier tubes would be needed, each providing half of the supply voltage from their own high voltage (circa 700Vac center-tapped) secondary winding and with a choke to handle the negative-resistance characteristic. I decided to stick with the silicon rectifed supply design as I didn't want to go down the rabbit hole of 'tube-specific' / custom wound magnetics or filament supplies which would probably make that the most expensive (and bulky) part of the design. 240V/120V PCB mount transformers are widely available at the power ratings required.

Mains clock recovery

Deriving a clock from the mains frequency was historically an accurate way of time keeping, with mainland Australia maintaining less than 5 seconds of accumulated error. Although the guarantee of accumulated error has recently been removed. The accumulated error in the current mains system remains to be seen - this may be a topic I investigate in the future. To obtain a 50Hz square wave, the mains tranformer secondary voltage is clipped by a neon bulb. This output is in turn used to drive 2X and 5X ring counters to divide the 50Hz clock down to 25Hz, 5Hz and 1Hz respectively. The 1Hz output is used to drive the seconds nixie digit while 5Hz is utilised to provide a reset signal with a rising edge soon after the 1Hz edges.

Hours counter

The final piece of the puzzle is to create a two-digit nixie counter that counts to 24 (0-23), for the hours display. We can accomplish this by cascading a decimal (0-9) nixie counter with a 0-2 nixie counter, plus additional circuitry to reset the combined counter back to 0 after it reaches a count of 23 (as opposed to 29). To accomplish this, a dedicated buffer circuit is added for triggering a '4' on the least significant digit (LSD) with biasing provided by the most significant digit (MSD) nixie counter such that the buffer is only allowed to be set when the MSD is at a count of '0' or '1'. Additionally, biasing for a '0' count is such that triggering is enabled when either the LSD is '9', or, when the LSD is '3' and MSD is '2'. Thus, both nixie counters increment decimally until the count is '23', at which point triggering is disabled for the '4' on the LSD while '0' is enabled to trigger instead. Upon the LSD changing to '0', the MSD is triggered as per normal when the LSD rolls over to '0', changing the MSD from '2' to '0' without noticeable delay.

Complete 24-hour clock top-level

With that, here is a top-level schematic of a functional 24-hour clock:

PCB Layout

The PCB is a simple 1.6mm 2-layer FR4 board, to save on production cost. The trace clearance for the required voltages and need for both sides of the board to route signals excludes the ability to implement a ground plane, however this is not required for signal integrity since the signal bandwidth is only a handful of KHz. Since this project is more an art display piece than something I'd run 24/7 or intend to commercialise, EMI was not a concern.

The Nixie tubes are proudly displayed top-centre, with trigger tube logic below accordingly, functionally its own clock as well as just a blinkenlights display. The powersupply occupies the left hand side of the board and clock generation on the right. Along the bottom are push buttons to manually set each nixie counter to the correct time.

All trigger tubes are socketed so they can be exchanged easily. This is somewhat neccesary since mismatches in breakdown and holding voltages can cause problems with the operation of the circuits - some of the tubes are 40 or more years old after all. The nixie counters are the most critical circuits since the tolerance for variation in trigger tubes is the tightest. Tubes which don't function correctly in those positions can therefore be exchanged into less critical positions such as the clock generation and buffer circuits. Nixie tubes are soldered to daughter boards in order to mount them parallel with the main PCB.

Enclosure

I designed a simple enclosure out of laser cut acrylic to show off the circuits on a desk top while keeping curious fingers away from hazardous voltages (Achtung! Alles lookenpeepers). The enclosure also functions to mechanically align the trigger tubes and nixie tube daughter boards via holes and slots respectively, since the slop in the 0.1" header pins/sockets otherwise allows the tubes to be skewed. Aluminium actuators extend the pushbuttons through the front panel to set the time.

Final Product

Errata

There is one minor mistake in the circuit design - I thought I was being clever by removing the buffer before the seconds nixie counter (page: Digit10NoBuf_Sec), in order to eliminate two trigger tubes. Of course, the seconds counter still works without the buffer since the 5x divider produces a larger enough pulse signal to drive it, however, this prevents the set button from working since the output from the 5x divider is normally low and pulses high, whereas the output of a buffer is normally high and pulses low. It's still possible to set seconds by simply pulling one of the trigger tubes in the counter out of its socket, waiting until the counter is blocked by the missing tube and then inserting the tube when the time is correct.

Discussion

Binning trigger tubes

Trigger tubes can be 'binned' before going to the effort of soldering header pins onto tubes which may be too far out of spec to function correctly in the clock. To accomplish this I first populated up to the first 5X divider, then used clip leads from one of the trigger tube sockets in the 5X divider and tested all the tubes I had one by one. I produced three groups of tubes; 'normal sensitivity' tubes that worked correctly in the divider, 'low sensitivity' tubes which would not trigger in the divider, and, 'high sensitivity' tubes which would trigger but then not extinguish when the following tube was triggered. The observed 'sensitivity' can be for a variety of reasons, such as differences in the breakdown or holding voltages. It's likely that you may be able to use the tubes that have 'high' or 'low' sensitivity by populating an entire subcircuit with same sensivity tubes, however, you may need to change the anode or cathode resistances to get the circuit to function. If you are to replicate this design I'd recommend buying at least 50% more trigger tubes than required to ensure you can find a subset that work together.

A note about ambient light and long term storage of cold cathode tubes.

Cold cathode tubes require ambient light (photons) or another source of radiation to lower the voltage required for ionisation. Many tubes contain a radioactive source, often krypton-85 gas, to provide reliable operation in the dark. I don't believe the MTX-90 tubes contain a source, either that or the source has significantly decayed over the past few decades as the tubes I have experience significant dark effect. Many tubes won't trigger if the clock is placed in darkness. I believe a blue or UV light can be used to remedy this. Tubes stored for long periods can also have trouble starting and require some exercise before they ionise easily again.

Downloads

 KiCad Project [.zip 17.9MB]

 Solidworks Assembly [.zip 7.0MB]