John Sokol's Blog: Building Neon Logic Gates: A Practical Maker’s Guide

Building Neon Logic Gates: A Practical Maker’s Guide

Companion to the pneumatic computer essay. This one is the how-to. If the pneumatic post asked “could a kid watch a computer think?”, this one is the answer for adults willing to handle a few hundred volts. The result: a clear-glass-and-acrylic stack with orange glowing cells that act as 1-bit memory, wired through a PCB underneath into anything from a single visible bit to a four-bit counter. Aesthetic of 1950s computer; build cost of a long weekend.

⚠️ This project uses 150–180 V DC at low current. It’s not as dangerous as mains AC at the same voltage, but it can hurt you and in unusual circumstances kill you. Section 11 covers what you need to do to keep it safe. Read it before you build anything. If you are not already comfortable building HV circuits with current limiting and bleeder resistors, build the all-acrylic pneumatic version instead — it’s the same pedagogy and you can’t electrocute yourself with air.

1. What you’re building

A single cell is a small glass-and-acrylic sandwich about the size of a postage stamp, holding 5-15 torr of neon, with two or three metal electrodes passing through the glass into the chamber. Apply ~90 V across the main electrodes, pulse a third “trigger” electrode, and the gas ignites into a visible orange glow. The cell stays lit until you drop the supply voltage below the sustaining threshold of ~65 V. One cell is a 1-bit memory you can see.

A row of eight cells, mounted on a PCB that handles the supply routing and the trigger pulses, is a visible 8-bit register. You load a number into it by sequentially pulsing the trigger inputs. You read the number off the front by which cells are glowing.

A few rows of cells wired with appropriate cross-couplings is a visible counter or, with more care, a visible adder. The final device is a small wooden-and-glass box about the size of a hardcover book, sitting on a table at the front of a classroom or maker space, glowing softly orange while it counts.

The whole thing runs on a wall transformer, draws under 10 watts, and sits silently. Compared to the pneumatic version: no moving parts, no compressor noise, faster operation (milliseconds per bit, not half-seconds), and the aesthetic is straight out of 2001.

2. How the cell works, in 500 words

Low-pressure neon has a useful property called the Townsend breakdown that makes it behave as a natural bistable element.

Below a critical voltage (the ignition voltage, typically 80-100 V for a 5-10 mm electrode gap at 5-15 torr neon), the gas in the chamber is electrically neutral and effectively an insulator. No current flows. The chamber is dark.

Above the ignition voltage, an avalanche of ionisation starts: a stray free electron is accelerated, collides with a neon atom, liberates more electrons and ions, and the cascade runs away until the chamber is full of plasma. Current flows freely. The chamber glows orange (specifically at 585.2, 614.3, and 640.2 nm — the strong visible neon lines).

Critically, once ignited, the voltage required to sustain the discharge is much lower than the ignition voltage — typically 60-70 V. The cell exhibits hysteresis: it stays lit until the supply voltage drops below the sustaining threshold, at which point the plasma deionises and the cell goes dark.

This hysteresis is what makes the cell a 1-bit memory. Sit the supply voltage at ~75 V — between ignition and sustain — and the cell is stable in either state. If you push the supply momentarily above 90 V (or, equivalently, momentarily lower the effective ignition threshold by injecting electrons via a third electrode), the cell ignites and stays lit at 75 V. If you push the supply momentarily below 60 V, the cell goes dark and stays dark at 75 V. The cell remembers.

The third electrode — the trigger — is the practical knob you control. A small voltage pulse on the trigger (typically 20-30 V on top of the cathode reference) creates a small auxiliary discharge that seeds the main chamber with enough ions to lower the ignition threshold below the running supply voltage. The main discharge fires. After the trigger pulse ends, the main discharge sustains itself.

The whole cycle is fast: ignition takes microseconds, sustained operation is steady, and extinction (when you drop the supply) takes hundreds of microseconds while the plasma recombines.

For logic, the trigger structure gives you natural building blocks:

Bit storage — one cell with a single trigger is a 1-bit memory.
OR gate — two triggers on the same cell, either one fires it.
Set-Reset latch — one cell for the set, one for the reset, cross-coupled supply lines.
Sequential logic — multiple cells wired to fire in sequence via shared trigger lines and capacitive coupling.

Combinational AND/XOR are awkward in pure neon logic and were historically done with diodes in front of the trigger inputs. You’ll do the same.

3. Bill of materials

For a build that produces one working cell plus the supply infrastructure to drive it, expect to spend about $400-500 in materials and $200-400 in tools (most of which you can borrow or already have).

Materials for one cell

Item	Spec	Source	Cost
Cast acrylic sheet	3 mm clear	Local plastics supplier, TAP Plastics, Home Depot	$5/sheet (12×12”)
Float glass	3 mm clear, ≥4×4”	Glass shop scraps, eBay, hobby store	$3/piece
Stainless steel pin stock	1 mm dia × 50 mm	McMaster 90145A115	$0.50/pin (need 3)
Buna-N O-ring sheet	1.5 mm thick	McMaster 8634K12	$25/sheet (enough for 50 cells)
Torr Seal epoxy	Two-part	Edwards Vacuum, Kurt J. Lesker	$45/kit
Apiezon Q wax	Vacuum sealing	Kurt J. Lesker, eBay	$25/stick

Gas-handling and electrical infrastructure

Item	Spec	Source	Cost
Single-stage rotary vane pump	≥10 L/min, can hit 10⁻² torr	eBay used (Edwards E2M5, Welch 1400)	$150-300
Neon lecture bottle	99.99%, 1.7 L at 1700 psi	Praxair, Airgas, AGSI	$80-120
Pressure regulator for neon	Dual-stage, 0-30 psi outlet	Praxair, eBay	$80 used
Pirani gauge + readout	Range 10⁻³ to 10² torr	eBay used (Granville-Phillips 275)	$80
Vacuum tubing	6 mm OD nylon or 1/4” copper	Hardware store	$30
Compression fittings	1/4” Swagelok-compatible	McMaster, eBay	$40
Needle valve	1/4” Swagelok	McMaster 4901K23	$35
HV DC supply	0-250 V adjustable, 50 mA, with current limit	XP Power AHV28-P50 or build from voltage doubler	$80 module, $30 DIY
HV trigger pulse driver	One MOSFET HV pulse generator per cell	DIY from CD4093 + IRF830	$5/cell
HV current-limit resistors	100 kΩ, 5 W wirewound	Mouser, Digi-Key	$3 each
Bleeder resistor	470 kΩ, 5 W	Mouser	$2
FR-4 PCB	Custom, two-layer	JLCPCB, OSHPark	$5 for 10 boards

Tools

Tool	What for	Cost if buying
CO₂ laser cutter, ≥40 W	Cutting acrylic and gaskets	Hacker space access
Drill press + 1 mm + 3 mm bits	Through-holes in glass for electrode pass-throughs	Hacker space access
Glass-drilling diamond bits	If drilling glass yourself	$20 set
Toaster oven	Bakeout at 60-70 °C	$40
Small vacuum desiccator or bell jar	Bakeout chamber	$50 used
Multimeter that handles 250 V DC	Voltage testing	Already have
100 MΩ HV probe (or 10:1 divider)	Measuring supply voltage on a meter	$40
Soldering iron, basic electronics tools	Wiring	Already have

The biggest single-item cost is the rotary vane pump. You can avoid the pump entirely by going to a vintage TIG-welder distributor and asking for a used pump from a refrigeration shop — they’re $50-100 and work fine for this pressure range. Or borrow from a university physics lab.

4. Designing the cell

The cell geometry is constrained by three things:

Paschen’s law sets the relationship between pressure, electrode gap, and ignition voltage. At ~10 torr neon, the ignition voltage minimum is around 80 V at a gap of ~5 mm. Smaller or larger gaps raise the ignition voltage.
Visibility wants the glow region as large as possible, so a bigger chamber is better aesthetically.
Power dissipation wants the chamber small. At 75 V sustained and 2 mA, each cell dissipates 150 mW continuously. A four-bit register with all bits lit dissipates 0.6 W, plus another ~10 W in the HV supply and current-limit resistors.

The cell I recommend for a first build:

Chamber size: 15 mm × 15 mm × 2 mm deep
Electrode gap: main electrodes 8 mm apart, both 1 mm pins
Trigger electrode: 1 mm pin, located 3 mm from the cathode and 3 mm offset from the main discharge axis
Glass plates: 25 mm × 25 mm × 3 mm
Acrylic spacer: 25 mm × 25 mm × 2 mm, with the chamber cut out and a 1 mm wide × 1 mm deep O-ring groove cut around the chamber perimeter

At ~10 torr neon, this geometry gives an ignition voltage around 90 V and sustains at 65 V. The glow fills the chamber visibly — bright enough to read by in a darkened room, soft enough that you don’t need to dim the room lights to see it.

For the laser cutter, draw the acrylic spacer as a single 25 mm square with:

A 15 mm × 15 mm cutout for the chamber
A continuous 1 mm × 1 mm groove around the chamber for the O-ring
Three 1.5 mm holes for the electrode pins (slightly oversized to allow alignment slop)

Cut at standard acrylic settings (50% power, 30 mm/s on a 60 W laser). The O-ring groove is cut as a separate engrave step at 20% power, single pass — this gives you a slot the O-ring can sit in.

5. Bakeout

This is the step you do before assembly. The purpose is to drive adsorbed water and volatiles out of the acrylic so the cell holds its gas composition for weeks instead of hours.

Procedure:

Cut all your acrylic pieces fresh — don’t bake them, cut them, then leave them in air. The fresh-cut surface re-adsorbs water within minutes. Cut just before bakeout.
Place the pieces in a small vacuum desiccator inside a toaster oven.
Connect the desiccator to the rotary pump via a vacuum hose.
Turn on the pump first. Let it pump down to <0.1 torr. Then turn on the toaster oven and set it to 60-65 °C. Do not exceed 70 °C — cast acrylic softens above this temperature and your O-ring grooves will distort.
Let it bake at temperature under continuous pumping for 48 hours minimum, 72 hours preferred. The outgassing rate drops by about a factor of ten every 24 hours during this period.
Cool to room temperature under vacuum. Vent the desiccator with dry nitrogen or filtered argon slowly — fast venting can re-deposit moisture from any condensation on cold surfaces.
Use the parts within a few hours of bakeout. Don’t leave them in open air overnight before assembly.

A small humidity-indicator card inside the desiccator is a useful sanity check: it should remain blue throughout the bakeout.

The same bakeout applies to the O-ring sheet (if it’s Buna-N — Viton is fine without bakeout) and to anything else organic that will be inside the sealed volume.

6. Assembly

You’ll need: baked acrylic spacer, two pieces of glass, three stainless pins, O-ring stock, Torr Seal, and a small flat surface clean enough to work on.

Drill the glass. The bottom glass piece needs three 1.5 mm holes for the electrode pins, located to match the spacer’s pin holes. Use a diamond bit at low RPM with water cooling. Take your time; cracked glass is the most common assembly failure.
Cut the O-ring. Cut a length of Buna-N stock to fit the groove in the acrylic spacer. Join the ends with a single drop of cyanoacrylate (super glue) to form a closed loop. The O-ring should sit slightly proud of the groove surface when placed in.
Set the electrodes. Push the three 1 mm stainless pins through the holes in the bottom glass, leaving about 3 mm protruding into what will be the chamber and about 10 mm protruding below for external connection. Seal each pin to the glass with a small dab of Torr Seal mixed per the package instructions. Cure 24 hours. This is your single hardest seal — take care.
Build the stack. - Bottom glass (with pins) face up - Acrylic spacer (with O-ring in groove) - Top glass - Sandwich the whole thing between two metal plates (aluminum is fine) with nylon bolts at the four corners. - Tighten the bolts evenly until the O-ring is compressed to about 50% of its uncompressed thickness. Do not over-tighten — you’ll crack the glass.
Mount the gas connection. The top glass needs one more small hole, off to the side of the chamber, where the gas-fill tube enters. Pre-glue a short length of 3 mm OD glass or metal tubing into this hole with Torr Seal before final assembly. The tube extends ~10 mm above the top glass for connection to the gas manifold.
Check the seal. Connect the cell’s fill tube via flexible tubing to the rotary pump. Pump down. Watch the Pirani gauge — it should read down to ~10⁻² torr within a minute or two. If it plateaus at higher pressure, you have a leak. The leak is almost always at one of the electrode pins or at the O-ring (which means it wasn’t compressed evenly).
Fill with neon. Once the cell pumps down cleanly, close the pump valve, open the neon supply via the needle valve, and let the cell pressure rise to 10 torr. Close the needle valve. Disconnect from the gas manifold by pinching off the fill tube with a small metal clamp (the tube can be permanently sealed later by heating and crimping, or by gluing the clamp on).

Total assembly time per cell: ~3 hours including epoxy cure (which you do overnight). With practice you can build five cells in parallel in roughly the same time.

7. The HV supply

You need a DC supply that delivers 0-180 V at up to 50 mA, with current limit and a bleeder resistor.

The cheap path: a 0-30 V variable DC bench supply driving a 6:1 voltage doubler made from off-the-shelf parts. Two 1 µF/400 V electrolytic capacitors, two 1N4007 diodes, configured as a Cockcroft- Walton ladder. Input 24 V AC (from a small line transformer), output ~150 V DC. Add a 100 kΩ series resistor for current limit and a 470 kΩ bleeder across the output.

The cleaner path: buy an XP Power AHV series module or equivalent. They cost ~$80 and give you a clean 0-250 V variable output with built-in current limit.

Either way, always include the bleeder resistor. When you turn the supply off, the capacitors retain charge for many minutes otherwise. The bleeder drains them within 10 seconds.

For the trigger circuit, you need a 30 V pulse, ~10 ms duration, floating reference, into the trigger electrode. Easiest implementation: a CD4093 Schmitt-trigger NAND gate driving an IRF830 MOSFET that switches a small 30 V supply onto the trigger pin via a 10 kΩ resistor. The MOSFET is rated for 500 V, so even if the cell shorts to its trigger you don’t kill the driver. One driver per trigger input.

Lay all of this out on a single board underneath the cell array. Through-hole construction, generous spacing between high-voltage traces (≥3 mm), and a clear silkscreen showing which traces are at which voltage. Mount the supply on the underside of the wooden case, behind a barrier so nobody’s fingers can reach the live nodes.

8. First light

You’ve built one cell. You’ve built the supply. Time to make it glow.

Confirm the cell has the right gas pressure with the Pirani gauge.
Connect the cathode pin to the supply ground via a 1 kΩ resistor (this provides a small reference and limits any transient).
Connect the anode pin to the supply positive via a 100 kΩ series resistor.
Bring the supply up slowly from 0 V. Watch the cell.
At around 85-95 V, the cell will spontaneously ignite — a sudden orange glow filling the chamber. The supply current will jump to 1-2 mA.
Lower the supply slowly. The cell will continue glowing down to about 60-65 V, then suddenly extinguish.
Note both voltages. These are your cell’s actual ignition and sustaining voltages. They’ll vary a few volts cell-to-cell.

If it doesn’t ignite by 100 V: the gas pressure is too low or too high (Paschen curve has a minimum), or the gap is too small. If the pressure is way off, vent the cell and refill more carefully.

If it ignites way below 80 V: the gas is contaminated with moisture or air (so you have a higher proportion of nitrogen, which ionises more easily). Pump and refill.

If the glow is white/pink instead of orange: definitely contamination, probably nitrogen from a leak. Pump down, refill.

If it ignites but flickers: could be a leak (slow contamination changing breakdown voltage), or could be supply ripple. Add a 1 µF cap across the supply output.

Once you have stable orange glow with reproducible ignition/sustain voltages, set the supply at the midpoint (typically ~75 V) and verify the cell sits stable in either lit or dark state for at least 60 seconds. That’s your working bistable.

9. Triggering for memory operation

With the supply at 75 V (between ignition and sustain), you want to flip the cell between lit and dark with pulses.

To set (light up): pulse the trigger electrode to +30 V relative to the cathode for ~10 ms. The trigger creates a small auxiliary discharge in the chamber that locally lowers the ionisation threshold for the main gap. The main discharge fires. After the trigger pulse ends, the main discharge sustains itself.

To reset (extinguish): pulse the supply briefly below the sustaining voltage. The simplest way is to put a small MOSFET in series with the supply line that briefly drops it (a “kill switch”) for ~1 ms. The plasma deionises and the cell goes dark.

With the set and reset capabilities wired to two pushbuttons on the front panel, you have a manually-operated 1-bit memory you can flick on and off and watch. It’s already a satisfying object at this point. Put it on a shelf for a day and look at it from time to time. Verify it holds state for hours without drifting (it should).

10. Building out to a register

For an 8-bit register, you build 8 cells side-by-side on one PCB and wire them with shared supply, shared reset bus (kills all bits simultaneously when you want to clear), and individual trigger inputs (one per bit).

The PCB layout puts the cells in a 1×8 row, 30 mm apart, with the HV supply rail running along the top edge and the cathode-ground rail along the bottom. Each cell’s anode connects to the HV supply via a shared current-limit resistor (or, more conservatively, individual resistors so a failed cell doesn’t take out the supply).

Mount the trigger pulse drivers on the underside of the same PCB, one per cell. Bring the trigger inputs out to a header connector on the back. Drive them from a microcontroller — an Arduino Nano works fine — that sequences the trigger pulses according to whatever pattern you want to load.

An 8-bit binary counter is the simplest demonstration: the Arduino counts internally and on each tick re-loads the register with the new value, displayed as glowing bits. Counts up from 0 to 255 over four minutes if you tick once per second. Watching the bits propagate is the entire point.

A visible adder is more elaborate. You build two 4-bit input registers, the adder logic (NAND gates in solid-state, not in the neon — the neon is for the visible state), and a 4-bit output register. Set the inputs by flipping switches that pulse the trigger inputs. Press a “compute” button. The adder logic computes A+B and triggers the appropriate output bits. The output register shows the sum.

The neon cells are the display + state memory. The combinational logic in between is solid-state. This is exactly how the Harwell WITCH (1949) worked: dekatrons for storage and visible display, relays for the actual logic. You’re recreating that architecture at maker scale.

11. Safety

The non-negotiable rules.

Bleeder resistors on every capacitor in the supply. A 470 kΩ / 5 W resistor across the main output capacitor drains it to <30 V within 10 seconds of power-off. Without it, the supply caps can hold a hazardous charge for 30+ minutes.
Current limiting on every cell. The supply itself should have a hard current limit at ~50 mA. Each cell should also have a series resistor (100 kΩ or higher) so that even if a cell shorts, the steady-state current is limited to <2 mA.
Live conductors behind glass or behind a barrier. Nothing conductive at >40 V should be touchable from the outside. The front of the device is glass over the cells. The back of the device is a wooden panel or a perforated metal cover with the high-voltage routing inside.
One-handed operation rule. When the device is powered on, keep one hand in your pocket. If you must touch anything, only do so with one hand. This prevents current paths through your chest.
Verify dead before touching. Before working on the powered-off device, measure the supply rail with a multimeter to confirm <10 V. The bleeder resistor should make this automatic but verify.
Mains isolation. The HV supply should be galvanically isolated from the mains via a line-frequency transformer or an off-the-shelf isolated module. Don’t use a mains-rectified supply without isolation.
Plug-in warning label. A small label on the back of the device saying “180 V DC inside. Service by qualified personnel only. Allow 30 seconds after disconnect before opening.” Yes, the bleeders drain it in 10 seconds; the label says 30 to be conservative.
No kids inside. Children can interact with the front of the device — toggles, pushbuttons, watching the glow. They should never have access to the back. This means either screwed-shut panels (not snap-on covers a curious 8-year-old can pop off) or a key-locked back panel.

At 150-180 V DC with ~50 mA current limit, this device is in the “will hurt, very unlikely to kill” range for a healthy adult under dry conditions. With the safeguards above it’s roughly as dangerous as the inside of an open CRT TV from the 1990s, which is to say: respect it, but don’t be paralysed by it.

If your maker space’s safety policy has a hard cap at “low voltage DC only,” this project is over that line and you need to talk to your space’s leadership before starting.

12. Troubleshooting

Cell ignites at the wrong voltage. Almost always gas pressure or gas purity. Pump down to <10⁻² torr (verify with the Pirani gauge), hold for 10 minutes to let outgassing equilibrate, refill with fresh neon. If pressure is the problem, you’re hitting the wrong part of the Paschen curve — go up or down 5 torr and re-test.

Cell ignites OK initially but extinguishes after a few seconds. Current limit is too aggressive. Drop the series resistor from 100 kΩ to 50 kΩ.

Cell ignites and won’t extinguish. Current limit is too loose, or the supply voltage is too high. The sustaining voltage is fundamental to the cell — if your supply is above it, the cell stays lit. Drop the supply 10 V.

Glow is purple/pink instead of orange. Air leak. Nitrogen contamination glows pink. Find the leak by squirting isopropyl alcohol on suspect seams while pumping — IPA’s vapour pressure changes the Pirani reading at the leak point.

Glow is bright orange but flickers. Supply ripple. Add filtering caps to the supply (1 µF / 400 V across the output, plus a small 100 nF ceramic for high-frequency noise).

Glow is dim and uneven. Either pressure is too high (most likely) or the chamber is too deep relative to the electrode gap. Re-pump and re-fill at lower pressure (try 5 torr).

Trigger doesn’t fire the cell reliably. Trigger pulse amplitude too low or duration too short. Increase to 50 V / 50 ms and retest. If still unreliable, the trigger electrode is too far from the main discharge path — re-make the cell with the trigger closer to the cathode (within 3 mm).

Cells holding state become unreliable after a few weeks. Outgassing. Re-pump and refill. If happening within days, you have a slow leak — find it via the IPA technique.

13. Going further

Things you can build on top of the basic cell:

A multi-decade counter. Use commercial dekatrons (eBay) as the counter elements; use your custom cells as the visible “active” bit indicator showing which decade is at non-zero. A 4-decade counter counts to 9999, fits in a hardcover-book-sized case, and runs indefinitely from a wall transformer.

A self-clocking ring oscillator. Two cross-coupled cells with an RC delay between them will oscillate at a frequency set by the RC time constant. Build a chain of these as a visible clock generator, each cell flashing in turn at a rate the kid can see.

A visible shift register. N cells in a row, with each cell’s trigger connected to the previous cell’s anode via a capacitor. A pulse on the first trigger propagates down the register at a rate set by the capacitor values. Watch a single lit bit walk down the row.

A pneumatic-and-neon hybrid. Connect the output of the pneumatic adder to the trigger inputs of an 8-bit neon register. The mechanical computer feeds its answer into the visible electronic display. Two substrates, one demonstration. Worth the extra weekend.

A teaching kit. Once you have a stable cell design, document it as a kit and put it on Tindie or Crowd Supply. The market is small but loyal — Nixie-tube clock builders are a real subculture and they will find this and love it. Open-source the PCB and laser-cut files on GitHub; build a small business selling the consumables (electrodes, gas cylinders, O-ring stock).

Coda

The neon logic gate is one of the few projects where the way the device looks while operating is also exactly how it operates. A transistor is just a black SOIC package and you have to take the datasheet’s word for it. A neon cell is a chamber where you can literally see the plasma. The pedagogy and the aesthetic are the same thing.

The 1950s knew this. The Harwell WITCH, on display now at the National Museum of Computing at Bletchley Park, has been running since 1951 and is still pulled out for demonstrations. Visitors don’t remember the relays; they remember the dekatrons, glowing as they count.

What you’ll build is in that lineage. Smaller, simpler, less ambitious, but the same idea. Computation you can see.

If you build one, send me a photo.

John Sokol's Blog

Friday, May 29, 2026