[kragen]

Gas-discharge tubes are kind of niche today, and vacuum tubes have been more important since at least the 01920s, but in the 01960s, gas tubes weren't that niche; people used them for signal switching, voltage regulation, light detection, breakover elements for relaxation oscillators, digital displays https://en.wikipedia.org/wiki/Nixie_tube, counters https://en.wikipedia.org/wiki/Dekatron, and other memory devices. https://en.wikipedia.org/wiki/Neon_lamp#Applications has a bit of a catalog of uses, which references this actual catalog of devices from 01966 https://archive.org/details/ge-glow-lamp-manual-1966/page/n1... and Miller's 01969 book https://www.tiffe.de/roehren/neon.pdf. Geiger tubes are still used for detecting ionizing radiation, although we have other alternatives today. And of course most forms of electrical lighting have been gas discharge lamps since Davy invented the first commercial electric light around 01809: the open-air arc lamp, the neon lamp or tube, the fluorescent tube, the mercury light, low-pressure and high-pressure sodium-vapor lamps, strobe lights, and continuous high-intensity discharge xenon lamps.

The 01951 book I learned digital logic from, by Dennis Ritchie's father and two of his Bell Labs colleagues, has a chapter on switching with "electron tubes, both vacuum and gas-filled," and "semi-conductors": https://archive.org/details/TheDesignOfSwitchingCircuits/pag....

(Fluorescent light bulbs, by the way, do have a heated filament, and do work by thermionic emission, though cold-cathode fluorescents like those used in old LCDs don't.)

The respective niches of vacuum tubes and gas switching tubes could be very crudely summarized as high speed and high reliability. Even primitive vacuum tubes had switching times in the microseconds, and by WWII it was below a nanosecond, like transistors, but they relied on hot filaments that eventually burned out. Cold-cathode gas tubes, by contrast, essentially never break, but they take close to a millisecond for the gas to deionize so they can stop conducting. They can switch higher-frequency signals, but they can't switch on and off faster than that. Keister, Ritchie, and Washburn say of hot-cathode gas tubes:

> The speed of response of the tube is contingent primarily on the ionization and de-ionization times of the tube. Depending upon the gas, the ionization time ranges from a fraction of a microsecond to several microseconds; the de-ionization time is ordinarily of the order of a hundred to a thousand microseconds, though lower values have been achieved. The tube, then, can respond very rapidly to input signals applied to operate the tube, but considerably more time must be allowed for extinguishing the tube.

When I first read this when I was eight, "a hundred to a thousand microseconds" presumably sounded incredibly fast, but of course it's painfully slow for computation. Of cold-cathode tubes, they say:

> Moreover, since the cold-cathode tube has no filament, no standby current is consumed. The speed of response, though somewhat less than that of the hot-cathode gas tube, is sufficient for most applications. The ionization time depends upon the time necessary to transfer the discharge from the starter gap to the main gap, and it is generally less than a hundred microseconds. Main gap de-ionization times are of the order of one to ten milliseconds.

You might hope that this would have improved since 01951, but, as far as I can tell, it never did.

They continue:

> Because of its suitability to switching circuits, the electron tube circuit examples contained in the remainder of the chapter are, in the majority of cases, based on the cold cathode-tube.

(They do, however, include a few vacuum-tube circuits.)

The rest of the book is about relays. Vacuum tubes and semi-conductors were, from their point of view, niche.

[kragen]

One of the most surprising applications for gas tubes in Miller's 01969 book, which I hadn't read before, is capacitive touch control of a 100mA solenoid. His Fig. 6–15 on p. 74 consists of four neon tubes (two T2–27–WR500, two 5AB-B), four resistors, two .001μF capacitors, a 4μF capacitor, a freewheeling diode for the solenoid, and the solenoid itself, and is powered by a 160VDC supply. When you touch the "on" or "off" touchplate, which is grounded through a 5.6MΩ resistor, your body capacitance momentarily provides a path for the 160V to ground before the touchplate capacitor charges up and blocks it. This kicks on the respective T2–27–WR500, which has a series 5AB-B to ground through a 6.8kΩ resistor shared between them. The two series pairs of tubes, connected on the high side through the other cap, form a flip-flop; the "off" pair has a 10kΩ resistor feeding it from the positive supply, while the "on" pair instead is supplied through the solenoid being controlled. When one pair turns on, the big cap couples a negative–going pulse to the other pair to turn it off.

12 low-precision components is pretty good for providing a flip-flop, high-voltage power switching†, capacitive touch sensing, and indicator lights. Miller seems to imply that such circuits were commonplace at the time.

______

† I think the circuit is switching 3mA with 50 volts across the solenoid, so a respectable 150mW, even if the 5AB-B is only rated for 0.3mA. The T2-27-1WR500 is rated for 3mA. The 5AB-B has a maintaining voltage of 50–60V, the T2-27-1WR500 of 60–70V, so the voltage left across the series combination of the 10k high-side resistor and the 6.8k low-side resistor is something like 50V when the "off" side of the flip-flop is conducting, and 50V/16.8kΩ is just under 3mA. I assume the solenoid must have comparable resistance.

[Animats]

Once upon a time, there were three branches in US electronics - Bell System, IBM, and everybody else. You're reading the Bell System viewpoint.

In the Bell System, most electronic components came in rectangular metal cans, often hermetically sealed, usually labelled "Western Electric NNNN Network". The Bell System loved inductors. Inductors don't wear out. They often used unusual inductors, such as saturable reactors, or inductors with a copper slug. For the same reason, they liked gas-discharge tubes, although they're not suitable for amplifying audio.

IBM liked plug in cards. Some cards in tabulating machines had moving parts connected to drive shafts. Tube computers had plug-in subassemblies.[1] This allowed maintenance of large machines in the field. Thyatrons were used in some early printers, as the drivers for the printer magnets. But not for logic - too slow.[2]

Everybody else had metal chassis with tubes on top and everything else underneath. Military gear would have extra hold-down arrangement for tubes, and often metal tubes, but usually stayed with the metal chassis form factor.

[1] https://www.righto.com/2018/01/examining-1954-ibm-mainframes...

[2] https://bitsavers.trailing-edge.com/pdf/ibm/logic/223-6746-1...

[kragen]

Yes, agreed. But neon lamps weren't limited to the Bell System; everybody who worked in electronics at the time used them for something, though not as commonly as things like capacitors and resistors.

Microwave gear of course had to use vacuum tubes until transistors got fast enough; as you say, although you can switch a voice signal with a neon lamp, you can't amplify it that way, and microwaves are a million times faster than voice. (Amplifying voice with a saturable inductor, a so-called "magamp", had its day too, though magamps are rarely seen today outside of ATX power supplies.)

But a lot of electronics didn't have to run at microwave frequencies or even voice frequencies; motor frequencies or powerline frequencies were enough.

Which page are you referring to in [2]?