How a dekatron counts

A dekatron is a gas-filled glass tube with a ring of ten cathodes around a central anode. One spot of glow, orange in a neon tube like this project's A101, sits on one cathode. Send the tube a pair of pulses and the spot moves one place round the ring; send them in the other order and it moves back. The position of the spot is a decimal digit.

That makes the dekatron unusual among display tubes: it does not just show a number, it holds one. It is a counter, a one-digit memory and a display in one envelope. This page explains how a glow can do that, using the A101, the two-guide dekatron in this project.

This project's A101 stepping on the bench, from a 400 V supply (watch on YouTube, recorded 2026-10-19).

Step through it yourself ↓ in the interactive ring below.

The page answers five questions:

1. Why does only one main cathode glow, when all of them are connected to the circuit?
2. How can pulsing a whole group of guides select only the guide beside the glow?
3. Why does the glow move on to the next main cathode rather than back to the old one?
4. Why does reversing the pulse order reverse the count?
5. What keeps the count between pulses?

Step through it

The ring below is drawn from above. Between each pair of main cathodes (circles, K0-K9) sit two guides: a G1 (triangle) and a G2 (square). All ten G1 guides are wired together, and so are all ten G2 guides: the circuit has only two guide wires. The figure drives them the way the bench test did, with two pushbuttons that ground one set of guides each while held (the circuit is further down).

Use Stage ▶ to follow one handoff at a time, or Count +1 to complete a whole step. With the ring focused, the left and right arrow keys move one stage and the up and down keys one count. Under the controls, an expandable voltage record of the bench circuit builds up like a scope trace, in volts from ground, with the cathodes dashed underneath each trace. Count up a few times, then down, and compare the order in which G1 and G2 fall.

The interactive figure needs JavaScript. Here is the ring at rest, with the glow on K0, and the sequence that moves it one step:

Stage	G1 guides	G2 guides	Glow on
Rest	at rest	at rest	K0: count = 0
1 (S1)	grounded	at rest	G1 beside K0 (transfer in progress)
2 (S1, S2)	grounded	grounded	spreading from that G1 to the next G2
3 (S2)	at rest	grounded	G2 beside K1
Rest	at rest	at rest	K1: count = 1

Starting with G2 instead runs the same stages backwards and counts down.

What the figure is, and is not

A qualitative state model, slowed down enormously, not a plasma simulation. The halo is schematic, and each stage in the real circuit is far shorter than shown. The drawing's electrode order and viewing direction are illustrative until checked against the A101 itself. The drive sequence is the bench test's button sequence, which stepped the tube reliably both ways.

Why only one glow

A glow discharge needs a high voltage to strike (start) across a gap, and a lower one to sustain it once the gas near the cathode is ionised. All the tube's cathodes share one anode, and the anode is fed through one resistor.

Before anything conducts, no current flows through the resistor, so the anode sits at the full supply voltage. On the bench the tube struck as the supply reached 289 V, with the cathodes at about 33 V (calculated): an estimated 256 V across the tube. From then on the current through the anode resistor, R_A, drops the anode until only the maintaining voltage is left across the glow: at a 400 V supply the anode measured 210 V and the cathodes 63 V, so 147 V at a tube current of 288 µA. That is plenty to keep the existing glow alive, and far below the roughly 256 V that started the first one. No second glow can start, and the tube settles into one lit cathode by itself. Almost always: the A101's datasheet warns that just after switch-on two cathodes may glow at once, until a reset pulse clears one.

How the glow moves

On the bench, at rest, the guides sit at G+, about 91 V from ground, above the main cathodes at about 57 V (the circuit is below), so the glow has no reason to visit them. To move it, a button grounds one set of guides: 0 V, well below the cathodes.

All ten G1 guides fall together, but they are not all equal. The one beside the glow sits in a cloud of ions and excited atoms left by the discharge, and a discharge starts there far more easily than at a guide across the ring. So the glow hands over to the adjacent guide only. This is a local handoff: the glow grows on the new electrode as it fades on the old one. The discharge transfers between neighbouring electrodes; it is not a glowing spot circling the ring.

The presses overlap. While S1 is still held, S2 grounds G2 too, and the glow spreads to the G2 beside it. That G2 is on the far side of the G1 from the old cathode: the G2 on the old cathode's other side is two electrodes away and out of reach. Then G1 is released, leaving the glow on G2, and finally G2 is released. The glow falls to the nearest main cathode, which is now the next one. The two guides between each pair of cathodes act like a ratchet: the order of the pulses decides which way the glow can go.

Reverse the order (S2 first, then S1) and the same reasoning runs the other way round the ring: the count goes down. The same two buttons drive the count in both directions.

Unrolled: the voltages that decide it

Cut the ring open and lay a few electrodes out in a row under the anode. The chart below them is in volts from ground, from the bench circuit (below). Each electrode's bar runs from its own voltage (the dot) up to the anode's (the solid line), so the bar's length is the voltage across its gap. Two marks on each bar stand for the two thresholds of a simplified model: the tick, 147 V above the electrode, is what keeps a glow going; the dashed outline, about 256 V above it, is what fresh gas needs to start one. The 147 V is the maintaining voltage measured at the bench's operating point, and the 256 V the striking voltage estimated from the first strike; real thresholds shift with current and differ from gap to gap, but the model holds them fixed. A gap next to the glow is primed by its ions and starts at a lower voltage (the arrow); by how much was not measured, so the figure gives no number for it.

Because the anode sits 147 V above whichever electrode glows, it follows the glow. When the glow moves onto a grounded guide, the anode drops to 147 V, and the cathode the glow left behind no longer has enough voltage across it to stay lit.

Each change is shown twice: the instant the buttons change, before the glow has moved, and after the handoff.

The interactive figure needs JavaScript. One step from K0 to K1, in volts from ground (gap = anode minus electrode):

Moment	Anode	Cathodes	What the gaps see	Result
At rest on K0	204 V	57 V	Other cathodes 147 V (sustains, cannot strike); guides at 91 V, 113 V	Glow stays on K0
S1 pressed, instant	198 V	51 V	Every G1, now at 0 V, has 198 V: below 256 V, but the G1 beside K0 is primed	Only that G1 strikes
After	147 V	39 V	K0 is left with 108 V, below 147 V	K0 goes out
S2 pressed too	147 V	34 V	G2 at 0 V has 147 V; the one beside the glow is primed	Glow spreads onto it
S1 released	147 V	39 V	G1 rises towards 70 V: 77 V across it	G1 goes out; glow on G2
S2 released, G2 passing 57 V	204 V	57 V	K0 and K1 both have 147 V; only K1 is next to the glow	Glow lands on K1
After	204 V	57 V	G2, back at 91 V, has 113 V	Count = 1

A first-order picture

The two thresholds come from the bench: 147 V measured at one operating point, 256 V estimated from the first strike, both held fixed here. The node voltages are calculated from the bench circuit with the simple anode rule, and come out about 6 V below the two points that were measured. While a released guide carries the glow, it rises only as fast as its 470 kΩ pull-up allows; the figure shows it at the moment it passes the cathodes. Real gaps also differ in geometry, and the ions' influence fades with distance rather than stopping at the neighbouring electrode. The figure shows why the handoff is selective, not how fast it happens.

The circuit on the bench

The voltages on this page come from the circuit the project's A101 was first tested on: a bench supply, a few resistors and two pushbuttons. All voltages are measured from ground, the supply's negative terminal, so a number always means the same thing.

Not a casual experiment

This circuit has 400 V on open wiring. If you build anything like it, it is at your own risk: read the safety notes first, and keep one hand away while it is powered.

The bench circuit (open the drawing full size), with the parts as built: the record of an experiment, not a recommendation. R1 was two 1/4 W 680k resistors in parallel, and each saw about 320 V, above the usual rating for that size. To build it, make R1 from two resistors in series (2 × 160k–180k), as in the bench test's parts table.

• The anode is fed from the 400 V supply through one resistor, R_A.
• All ten main cathodes return to RTN, which the divider R1-R2-R3 holds at about 57 V rather than at ground. That is the trick that lets a plain switch to ground pull a guide below the cathodes.
• All ten G1 guides are joined, and so are all ten G2 guides. Each set is pulled up through 470 kΩ to G+, about 91 V, which keeps the guides above the cathodes at rest. Pressing S1 grounds every G1 at once; S2 grounds every G2.

The node voltages in each state, calculated from the resistor values:

Buttons	Glow on	Cathodes (RTN)	G+	Pressed guide	Anode
None	a main cathode	57 V (63 V measured)	91 V	—	204 V (210 V measured)
S1	a G1	39 V	70 V	0 V	147 V
S1 and S2	a G1 and a G2	34 V	62 V	0 V	147 V
S2	a G2	39 V	70 V	0 V	147 V

The calculation treats the tube simply: the anode sits 147 V, the tube's measured maintaining voltage, above whichever electrode glows. Where the bench was measured, it comes out about 6 V low. Notice that the cathodes move too. The divider is not stiff, and while the glow sits on a grounded guide the tube current no longer flows into RTN, so RTN falls. The controller board uses the same scheme with transistors in place of the buttons, but regulates RTN, so there the cathodes stay put.

How fast it can count

Slowly. The A101 is rated at 1,000 steps per second, at the slow end of what dekatrons could do. That is no handicap here: the display's plan steps it at up to 200 per second.

What sets the limit is the same thing that makes the handoff work. Every transfer leaves ions and excited atoms around the electrode the glow has just left. They are what primes the neighbouring gap, and they take time to clear: step again too soon and the old position is still primed, so the glow can take the wrong neighbour. Ericsson's 1954 catalogue, describing one of its tubes, names the limit as the de-ionisation time of the gas.

So the gas fill sets the speed. The table collects the tubes whose figures could be traced to a datasheet, a manufacturer's bulletin or Acton's account; a dash means the source does not say.

Tube	Fill	Glow	Maintaining voltage	Rated speed	Source
A101 (Soviet; this project)	neon	orange (this project's bench)	130–170 V	1 kHz (0.01–1000 Hz)	Original datasheet, 19891
Ericsson GC10B	neon and helium	—	—	4 kHz	Acton2
Raytheon CK7978	neon	—	190 V (nominal tube drop)	5 kHz	Raytheon datasheet, 19613
A106 (Soviet)	"a neon mixture"	blue	255–295 V	10 kHz (10⁴ Hz)	Soviet handbook page4
Ericsson GC10D	helium and hydrogen	—	—	10–20 kHz, by drive waveform; up to 40 kHz by Acton's account	Acton2; collectors' survey8
A107 (Soviet)	—	light blue	270–310 V	1 MHz (0.01 Hz–10⁶ Hz)	Soviet handbook page5
Elesta EZ10B (Switzerland, 1961)	— (collectors: hydrogen7)	"weaker, bluish"	about 290 V (half its 580 V supply)	"a little above 1 MHz"	Elesta technical bulletin No. 17, 19616

Two patterns stand out. The slow tubes are the neon ones with the familiar orange glow; the fast ones glow blue. And the fast ones need about twice the A101's maintaining voltage. Their makers' own sheets mostly do not name the gas: the A107's is silent, and Elesta's bulletin describes the EZ10B's speed and glow but not its fill. Collectors' sources say hydrogen for the 1 MHz class, and Elesta concedes the price in light: the EZ10B's glow is weaker. Reaching 1 MHz took more than the tube, too: the bulletin gives a dedicated input stage for it, more critical to build and set up than its 500 kHz trigger.

Ten times too fast: one tube, three sources

The A106's handbook page gives its speed as 10⁴ Hz: 10 kHz. The collectors' site Lampes et tubes transcribes this as "104 Hz", the exponent lost. A third source, Industrial Alchemy, rates the tube at 100 kHz, ten times the handbook. The fill is no clearer: the handbook says "a neon mixture", which cannot by itself explain a blue glow (neon glows orange-red), while Lampes et tubes says argon, whose glow is violet-blue. The table uses the handbook. Collectors' surveys also list a few 100 kHz types, the Sylvania 6909 and 8035 and the Raytheon 8262; without a datasheet to check them against, they are left out of the table.

Speed had a price beyond the voltage. Hydrogen seeps out through the glass, even on the shelf, and helium-filled GC10Ds have almost always lost their gas by now⁸. Neon was the stable choice, and most dekatrons ran far below the hydrogen tubes' speeds. For faster inputs, the answer at the time was to put vacuum-tube counting stages in front: for one of its tubes, Ericsson's catalogue advises vacuum-tube decades ahead of the dekatron above about 20,000 counts per second. Transistor counters later made the whole arrangement obsolete.

A tube that remembers

Between steps, nothing in the bench circuit records the count. Both guide wires sit at rest, identical for every digit. The count exists only as the position of the glow, and the discharge keeps itself there: the anode-resistor balance above holds it on whichever cathode it occupies. The pulses carry no number, only "one more" or "one fewer".

In the finished display the point is literal. On the controller board, nine of the ten main cathodes are joined together at the socket and return on one wire; only K1, the index cathode, has a wire of its own (turn on Show connections in the figure). The controller cannot ask the tube which digit it holds. It can only see the index light once per revolution and count the pulses it has sent since, so what the controller holds is an estimate of the count; the count itself is the glow's position in the tube. Normally the two agree. After a missed step they differ, and the controller can only notice at an index visit, if then: one index cathode cannot catch every missed step. The controller's firmware is planned to re-home the tube whenever the index disagrees, and after any power interruption.

The memory is volatile. It lasts exactly as long as the discharge: switch off and the count is gone, and at the next power-up the glow strikes wherever it strikes. That is why a dekatron circuit needs the index cathode, or a reset, to find a known starting point.

Engineers in the 1950s used exactly this as computer memory. The Harwell Dekatron computer kept its numbers in 720 dekatrons, each one a decimal digit you could watch.

The memory has one more quirk: it does not like to sit still. A glow left on one cathode for a long time changes the tube: Ericsson's 1954 catalogue warns that the GC10B's maximum speed drops if the glow rests on one cathode for more than 2,000 hours. The A101's datasheet allows at most an hour of running slower than one step per 100 seconds, and after slow running (a step per second or less) recommends twenty minutes of exercise at 100 steps per second or more. The Harwell restorers met the same problem: dekatrons that had spent a long time with the glow on one cathode became "sticky", and they planned a way to spin a whole store of them. A dekatron that shows a fixed number for hours is, in the datasheet's terms, being misused. This display's controller will therefore never hold the glow still for long: it exercises the tube now and then, and switches it off when there is nothing to show.

A short history

• 1945–1947. John Reginald Acton, then at the valve maker Cossor, finds that a pulse can move a glow discharge from one cathode to another, and that it stays put when the pulse ends. A colleague wishes aloud for "a valve that has ten states built into it"; Acton imagines moving the glow along a line of ten cathodes.
• 1948. At Ericsson Telephones in Beeston, Nottingham, Acton demonstrates a first prototype in November. By the end of the year the ten-way GC10A, filled with helium and counting up to about 400 pulses per second, is ready for production.
• 1949–1950. A provisional patent in Acton's name is granted on 18 January 1949. A colleague, Ray Bacon, suggests the name Dekatron, which appears in print in May 1950.
• 1950s. The faster GC10B, filled with neon and helium (about 4,000 pulses per second), becomes the standard tube, used by the thousand in Geiger-counter scalers, timers, early calculators and cash registers. The helium-hydrogen GC10D goes faster still (see How fast it can count). The design is copied in Europe, the USA and the USSR.
• 1951. The Harwell Dekatron computer, designed at Harwell from 1949, first runs. Its store is 720 dekatrons. It works until 1973; restored at The National Museum of Computing, it is rebooted on 20 November 2012 and recognised in 2013 as the world's oldest original working digital computer.
• Mid-1960s. Transistor counters make the dekatron obsolete, and Ericsson stops making them. Equipment already installed lives on: dekatron timers still controlled spot welders at the end of the century.

The A101 in this project is a Soviet two-guide dekatron with a ten-cathode ring and an index output. Its product sheet comes from the Reflector plant and is stamped December 1989, late in the dekatron's life; a fuller production history is still to be sourced.

This build

The tube on this page is not hypothetical. On the bench (2026-10-09) the project's A101:

• struck at a 289 V supply, with an estimated 256 V across the tube;
• held its glow at a maintaining voltage of 147 V, at 288 µA;
• stepped reliably in both directions, judged by eye, with the guides grounded: a swing of about 90 V, to about 63 V below the cathodes. The datasheet's control pulse is a 135–200 V swing (as its test circuit applies it), so the bench used roughly half to two-thirds of it.

The bench fixture these results come from: the circuit above on a perf board in the white clamshell, the tube in a printed socket housing, the bench supply behind, at 400 V. Photographed 2026-10-20; more in Build and investigate. The video at the top of the page shows it running.

The controller board uses the same scheme with the buttons replaced: RTN is regulated at about 50 V, and transistor switches driven by an RP2040 ground the guides, about 50 V below the cathodes. It also gives the index cathode K1 a wire of its own, so the controller can sense it. The details are in the engineering documents: bench test and dekatron controller design.

Sources

• John and Robin Acton, Dekatron – The Story of an Invention, Resurrection 61, Computer Conservation Society. The invention, the GC10 tubes and their decline.
• The National Museum of Computing, The Harwell Dekatron / WITCH. The computer, its store and its restoration.
• Ericsson 1954 Dekatron catalogue, transcribed on electricstuff.co.uk. Striking and maintaining voltage, glow transfer, guide sequencing and the de-ionisation limit, for Ericsson's tubes rather than the A101.

---

1. Reflector (Sovtek), A-101 product sheet, accepted December 1989 (scan via tube-tester.com): counting rate 0.01–1000 Hz. The sheet gives neither fill nor maintaining voltage; those are from tube-tester.com's A101 page ("Neon filled", 130–170 V). The glow colour is as seen on this project's bench: the classic orange of neon. ↩

2. Acton, Dekatron – The Story of an Invention (above): GC10B "neon/helium mixture", 4,000 pulses per second; GC10D "Helium and Hydrogen", "up to 40,000 pulses per second". ↩↩

3. Raytheon, CK7978 Technical Information, 15 January 1961: "cold cathode gas-filled 5 KC medium speed bi-directional ring stepping decade counter", "direct visual neon glow readout", input frequency 5000 cps, nominal tube drop 190 V. Copy via Industrial Alchemy. ↩

4. A106 handbook page (Soviet reference book): "Свечение — синее. Наполнение — смесь неона" (glow blue; fill a neon mixture); frequency 10⁴ Hz; maintaining voltage 255–295 V. Transcribed by Lampes et tubes as "104 Hz", argon, violet-blue; rated 100 kHz by Industrial Alchemy. ↩

5. A107 handbook page (Soviet reference book): "Цвет свечения — голубой" (light blue); counting rate 0.01–10⁶ Hz; maintaining voltage 270–310 V; life at least 1,000 hours. No fill given. Scan via Lampes et tubes. ↩

6. Elesta AG, Technische Mitteilungen No. 17, February 1961, "Die Dekadenzählröhre EZ 10 B": "Zählfrequenzen von 1 MHz", a cut-off frequency "wenig über 1 MHz", "das schwächere bläuliche Glimmlicht", a 580 V supply "etwa zweimal so hoch wie die Brennspannung", and a separate input stage for 1 MHz. Scan via Lampes et tubes; transcribed on radiomuseum.org. ↩

7. tube-tester.com, EZ10B: "Hydrogen filled"; Industrial Alchemy also gives hydrogen. ↩

8. Industrial Alchemy, Dekatron glow-transfer counting tubes: a collector's survey of dekatron types, including the GC10D's 20 kHz rating with sine-wave drive and 10 kHz with square pulses, hydrogen's leakage through glass, and the GC10D's lost helium. ↩↩