I recently asked a question about inductor flyback in relay computers on Retrocomputing.SE. The preferred modern way to snub the inductive kick of a relay's coil is to add a flyback diode across the coil. But considering the history of diode technology, I doubt that today's solid-state diodes were the solution used in the era of relay computers.

At that time (roughly pre-1950), vacuum tube diodes were considered more reliable than solid-state diodes. Which made me wonder: Would a vacuum-tube diode even work as a flyback diode? I mean actually working its purpose as a rectifier, not as an expensive spark gap.

(There is something satisfying about using an outdated technology to solve the problems of another outdated technology, even if they had been competing technologies: vacuum tubes versus relays.)

  • \$\begingroup\$ The forward voltage drop would be 'large' by today's standards. Which actually may be a good thing if you can otherwise tolerate it. On the order of 50 V, memory serving, at currents in the 1/4 to 1/2 amp. This means that the coil would discharge the field energy much faster than it would through a forward biased semiconductor diode. You will need something to operate the filament to heat up the cathode, though. And you may need to keep an "engine hour" meter on it so that you know when to consider replacing it. ;) Maybe look up the dual, 5U4GB? \$\endgroup\$ Jan 24 at 21:58
  • \$\begingroup\$ Are you asking if it could be used, or was it ever used for that purpose? Or used even if it might not make any sense to use it for that purpose? I mean, you could simply use nothing, or just a resistor, or RC snubber, for example. \$\endgroup\$
    – Justme
    Jan 24 at 22:16
  • \$\begingroup\$ @Justme: I am asking if it could be used. In other words, is there anything about them that would make them ineffective for this purpose. Sure, there are better ways, but that's not the question here. \$\endgroup\$
    – DrSheldon
    Jan 24 at 22:18

2 Answers 2


Yes but no.

First, about diodes.

Vacuum diodes are surprisingly competitive, actually, in certain V/I ranges -- if you ignore heater power, and noise if applicable.

For example, 6AL5 has a knee around 0.4V; granted this is at some microamperes, but it's somewhat impressive for just some hot metal in a vacuum. Not that the on/off ratio is very much, either: in fact, it's still passing forward current at -0.4V -- yes, forward current; it is, in fact, a heat engine, producing a minuscule amount of power (~nW), just as a dimly lit solar panel does (and for similar physical reasons, actually). This is roughly comparable to contemporary schottky or germanium diodes, but at much higher voltage (330V peak rated). And the capacitance isn't bad (~3pF). Of course, the thermal noise is roughly the incremental plate resistance, at the temperature of the cathode, so it won't win you a whole lot if you need say an RF mixer of especially low noise figure.

Or take the 6CL3, an average "damper" diode rated 350mA 5.5kV, sporting a 16V drop at 350mA. Compare with modern HV rectifiers which typically use stacks of 4 or more to reach such voltages, it's... well no, it's not comparable, not nearly, but it's interesting that it's in the same order of magnitude at least. It gets closer if you need high voltage at say MHz -- the vacuum diode has no reverse recovery, it's a schottky diode (emphasis on "hot carrier"!). And with capacitance in the 6-9pF range, it's better in some respects than SiC schottky (which have quite high capacitance especially at low voltages).

But of these properties, neither is relevant for relay snubber use. And voltage drop or power dissipation is essentially irrelevant (perhaps outside of highly repetitive applications i.e. vibrator supply). So it's down to overall capacity, and cost.

I don't have catalog pricing handy from the time, but I'm guessing the tube was more expensive than the diode, including the cost of socket, wiring, extra heater power, space taken up, etc.

But even more importantly, a simple capacitor or R+C will do the same, probably at a fraction of the price of either. The peak voltage isn't important for snubbing contacts, just the initial rise. As long as the voltage rises slower than the contacts open (the breakdown voltage increasing as they separate), you're done, arcing avoided. Just a solid capacitor has the downside of increasing turn-on current, making not just wear but contact welding also a problem; the addition of a series resistor limits charging current.

A further option is just dumb old ballast: a load resistor in parallel with the coil. The peak voltage is no more than maximum coil current times load resistance, and the expense is increased power dissipation while on, which the coil already dominates so it's not that big a deal. Nothing was especially efficient back then, and if you're talking relay logic, how do you really express the efficiency of something that could have asymptotically zero power consumption (from an information theoretic point of view)? You don't, you fall back on basic engineering principles: does the product meet specifications, user expectations, etc.

Even then, I think a lot of applications just didn't care, and used bare contacts and a bit of luck.


It could have been used. For example, the coil current for Post Office 3000 series relays used in some early computers was in the low tens of mA.

A vacuum tube dual diode such as the 5AR4 would have a voltage drop in the few volt range (< 10V). There is almost 10 watts consumed by the heater, however.

Of course that would slow the the operation, and may not have been desirable. A similar effect could be achieve at little cost and much higher reliability by paralleling the coils with a resistor (saving heater power, but increasing power for energized coils).

Various solid-state diodes were also available in the time period where relay computers were used, such as the 1N34 (1946), copper-oxide (used in some multimeters) and selenium. What, if any, coil suppression was actually used in relay computers (and telephone exchanges, which were close cousins) is another question entirely.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.