Adventures in Neon Discharge Bulbs

—— 001 ———-

When I was young, my dad took me to a magazine store. There were all kinds of magazines – gardening, hunting, world politics, race cars, doll collecting, you name it. Being the kind of kid who was often looking through the ventilation slots in the back of the TV to watch glowing tube filaments, I was quick to pick up an electronics magazine. This one happened to be the January 1969 issue of Elementary Electronics.

Cover of the magazine I grabbed

Cover of the magazine I grabbed

The first build-it-yourself project I opened to had several NE2 neon bulbs, some resistors and capacitors, a couple transistors and a battery. Looked simple without being too simple. The author, Herb Friedman (W2ZLF), described how the lights would blink in interesting ways, making great example of what some authors would have called an “idiot display”, pure blinkenlight fun, for little cost or effort. With bleepy chirpy sounds too.

“And soon the listener finds that exactly at the moment he thinks he has discovered the rhythm of the sound pattern, the light pattern makes a slight change and the sounds follow with a similar change in sequence. At one moment you may hear a clock inside
the box (accompanied by a glide tone), while the next moment your clock shifts into a sweep that mystifies.”

Great! I will build one.

While I already understood how one NE2 with one resistor and one capacitor worked to make a blinker, I couldn’t comprehend the messier network of this magazine project. Just what will it do?

Schematic that caught my interest when I was a kid

Schematic that caught my interest when I was a kid

I wondered what would happen if the capacitors were connected in other ways. Why six bulbs? What if we added a seventh – what would be good ways to add extra capacitors? I had only two maybe three NE2 bulbs taken out of old radios and whatever. I aimed to get a few more and build this gadget.

A PDF of the whole Jan/Feb ’69 issue is available at

I never did build that thing – other projects became more important. Now that we live in an age of advanced computing and sophisticated FOSS software which anyone can have, who needs to build anything? We can simulate it on a computer.

(See my recently made Neon Blinker animation on Youtube)

A blinking light illustrates one of the simplest feature of a wave or any type of oscillation: repetition in time. The number of blinks per second, or per minute, is the frequency. The time between blinks is the period. Use whatever time unit is convenient.

frequency: how many blinks per second, or per minute.
period: how long it waits between blinks.

This circuit is the simplest oscillator one can build. Only three parts! (Besides the power source.)

Simple NE2 relaxation oscillator

Simple NE2 relaxation oscillator

How it works: Initially, there’s no charge on the capacitor. Charge from the power source flows through the resistor, accumulating in the capacitor. As the capacitor’s voltage rises, it eventually reaches the “breakdown voltage” (or “strike voltage”) of the neon bulb, typically around 80V to 90V but could be higher. This much voltage makes a strong enough electric field between the electrodes that electrons are peeled off their atoms, making a plasma. As electrons recombine with positive Ne+ ions, light is given off. Once started, the plasma does not require as much voltage to be maintained. Typically about 35V to 45V or so will be found across the bulb in typical use.

This glowing plasma conducts electricity, discharging the capacitor until its voltage drops to the extinction voltage of the neon bulb, {{V_e}}. At this point, the plasma dissipates, losing conductivity and glow, in only a millisecond becoming insulator. Then the cycle repeats, the resistor charging the capacitor as before.

So, what’s the simplest way to model an NE2 neon discharge lamp in SPICE?
I’m on Linux, using ngspice. Models can be found out there in the www, but I think it’d be more fun and educational to create something from scratch, starting simple and building it up.

The most important feature of the neon lamp is that is acts as a switch, normally off (infinite resistance) but can be on, with some nonzero resistance. The switch is turned on by a certain threshold voltage, then turns off at some lower extinction voltage. Basic hysteresis. Luckily, it happens that (afik) all flavors of SPICE have a standard switch component, “S”, which does everything we want. Although, instead of using S’s built-in Ron resistance parameter, we’ll use an explicit resistor in series, so we can diddle around with it later. NE2s aren’t exactly linear when on, but we’ll deal with that another day. For now, SPICE’s S will do fine for a basic first attempt at simulating an NE2 bulb.

BTW, if I didn’t mention already, I’m using ngspice-26 on a 64-bit quadcore Arch Linux machine.

First, the simplest rough NE2 model possible, in a file named neon001.sub

Neon NE2 super-simple version 001
* Lamp turns on at 80V, off at 40V, with on resistance given as model parameter

.model ne2sw  SW (Ron=41ohm VT=60V  VH=20V)

.SUBCKT Neon001  1  2
R1  1  11  45ohm
S  11 2   1 2   ne2sw

Then a circuit to use it, a simple RC flasher:

Testing NE2 models as simple one-bulb flasher
.include neon001.sub

Vpwr   Vplus  0   120V
R1     Vplus  2   220K
C1      2   20      2.2uF
RCmeas 20    0      1ohm     ; to measure current through C1
X1      2   21   Neon001
RNmeas 21    0   1ohm        ; to measure current in neon bulb

Run a transient simulation in ngspice

tran  3msec 5sec  uic
plot v(2)
Voltage on capactor from SPICE simulation

Voltage on capactor from SPICE simulation

Odd, that one lower peak dropping down lower than the others. This is normal when the time step of a simulation is around the same size, or larger, than the time it takes for some fast process to occur. In this case, the discharging of the capacitor through the plasma-filled NE2 happens in less than a millisecond. The integrator using t_delta=3ms flies right over that, computing a system state after the event based on varying conditions at varying amounts of time before the event. With a 3ms step, it can’t accurately compute how much charge the capacitor loses.

Try a slightly different time step:

tran 3.1msec 3 uic
Capacitor voltage, slightly different time step.

Capacitor voltage, slightly different time step.

The pattern of lower peak variation is different. This is purely a simulator effect, nothing to do with real life. Since we are interested in simulating flasher circuits with multiple lights, it would be good to solve this problem. There’s about a 10V difference between the lowest of the low peaks and the highest of them. That amount of simulation error make useless any simulation of more complex circuits.

Two ways to solve this:
1. smaller time steps. This is easy. We get way more data, but our CPUs have not trouble dealing with that. 0.1ms might be good.
2. Modify the NE2 model to involve more capacitance, inductance, physical time delay effects of the plasma, so that the discharge process takes longer. Still, we need smaller time steps since 3ms was way too big.

tran 0.07msec 1.2 uic
plot v(2)
With smaller time step

With smaller time step

Tightly zoomed into a time span of a couple microseconds, we see how fast the discharge occurs. There are time steps within the discharge downslope. Probably not enough for accuracy, but we are glad that the curve bottoms out at 40V

Zoom-in of time axis showing discharge event

Zoom-in of time axis showing discharge event

You may have noticed in the SPICE file that I had put a one-ohm resistor in series with the capacitor. We can plot the current flowing through the capacitor with it.

plot v(20)
Current flow in capacitor

Current flow in capacitor

Current flow in capacitor, magnified to show it's positive, not zero, most of the time

Current flow in capacitor, magnified to show it’s positive, not zero, most of the time

The secone plot is zoomed in tight to a small voltage range, to show that the current is never zero, but positive as the capacitor charges for almost the entire cycle, and then briefly in 1/1000th the time discharges with a large current.

In switching and oscillating circuits like this, performing in a steady cycle, it’s normally a good rule of thumb (and physics) that the average current in each capacitor is zero.

An engineer building a neon discharge flasher like this would be wise to choose a type of capacitor which can handle the sudden reverse current flow, and have small ESR and lead inductance.

Another one-ohm resistor is in series with the neon bulb. By plotting v(21) and zooming in tight, we find zero current during most of the cycle, with

Current flow in the NE2

Current flow in the NE2

NE2 current, magnified. It really is zero most of the time.

NE2 current, magnified. It really is zero most of the time.

The average is not zero.  The NE2 is either off, like a piece of insulator, or it’s on, positive current flow for a brief time.

The important thing we have found with our crude NE2 model is that it works, but we must choose small time steps to avoid false variations of oscillation cycles.

I will write later about the basic NE2 multivibrator.