dumping a lot of heat

Yeah, about a kilowatt per U sounds about right.

I was just running a Dynatron copper CPU cooler at about 300 watts, and the exit air was painful. I might spec 250 on that one, which would be uncomfortable but not dangerous.

One could put a bunch of kilowatt-class mosfets on a couple of those coolers and make a monster class-D amp.

By standard 'safe' design, tubes have a pretty limited life.

By 1956, the heater was no longer the weakest element in the life equation for these parts - glass electrolysis was.

Electrolytic caps and their use has always been an issue. Cuffing the tubes not only enforces distance to other components, but reduces radiant effects in the viscinity.

Win, win.

RL

Tubes were awful. Still are.

It's astounding that people built computers with tubes. Some had hardware floating point!

I have used some 2 watt resistors that showed about 600 degf by iron/constantan but a red glow within could be seen by eye. Well above stated limits of 2 watts. Can't remember the name they went by but they were'nt wound types. The aluminum housed resisters are handy. 50 watt types have done much duty here. Hul

john lark> I'm thinking about building a biggish rackmount dummy load box. It

I had a case where there was a powerup surge, and it destroyed the big Mil-type alum case resistors. They generally opened and shorted to the case.

The Welwyn thick-film porcelain-on-steel things worked great, bolted to the chassis.

I have thought many times, but never got around to make an electronic load with many parallel circuits. At JLCPCB the mounting cost is low, so no biggie

The idea is to spread the heat out, not relying on one element, but many. Imagine a big PCB, like A3 size, spread out with 100 equal circuits electronic loads. Frequency response would be high, power dumping high also. Use a aluminum PCB like those used for LED lamps to get even lower Rth

Needs a fan too, somewhere.

A CPU cooler is cheap and has a fan. It is a bit klunky to mount on a PCB.

There are some nice heat-sinkable resistors. My TO-220s are cheap; I'll TDR one and see how fast it is.

People make crazy high power aluminum nitride microstrip RF resistors too.

I tested a Caddock DPAK surface-mount thick film resistor and it was pretty good. I'd expect the cheaper flat Ohmites or Riedons to be similar, likely better. I should TDR some.

Fly by wire in the Avro Arrow. . . .

The techniques for designing with them are quite different from transistors and ICs, you have to think a different way; they aren't just poor transistors, they have a different lifestyle altogether. They also have some advantages over semiconductors:

Those properties aren't needed most of the time but when they are, valves are a lot easier to design with than transistors if you are conversant and comfortable with the technology. There are still very few single transistors that can out-perform a humble EF91 from the

Yup. EF91 is known as "Not the valve that won the war."

Pentodes are *noisy*, at least 10 times more than some 3 cent jfet.

There are a couple of high-voltage tubes that were, until recently, worth at least considering. But multi-KV fets are more sensible these days.

I used to use the 1B3 HV rectifier diode as an amp, with the filament voltage modulating the conductivity. The downsides were the bandwidth and the X-ray hazard.

One could make an optocoupled tube half-bridge, or full-bridge, with maybe a 30 KV supply. I guess you'd use batteries for filament power.

That's because it wasn't available until 1947. :-)

Not necessarily. The Cockroft-Walton multipliers made by Philips, for generating high-energy X-rays, superimposed RF on the capacitor chain and derived the heater current from that. (It's somewhere in the Philips Technical Review but I can't remember where.)

Only slow overloads. Fast ones depend on the thermal time constant of the bit being heated by the overload. Some time later the energy reaches the heat sink but but then the damage is done.

I've just accidentally mixed up the anode and grid pins of one of the triodes in an ECC91. It drew about 100 mA for a few seconds with no damage. That's equivalent to mixing up the Base and Collector connections on a transistor and subjecting the Base-Emitter junction to about 10 times the rated maximum Collector current. How many transistors would survive that, even with the biggest heatsink available?

Mosfet data sheets usually have SOAR curves.

IXFH400N075T2 is rated for 1000 amps and 1000 watts (with astericks) and 30 kilowatts for 25 uSec.

No vacuum tube can survive 1000 amps because no vacuum tube can conduct 1000 amps.

A thyratron or an ignitron can conduct 1000 amps.

A krytron is even better.

Into the gate?

So no doubt it's properly mounted. Hard to beat electric stove elements on ruggedness and price.

Joe Gwinn

Don't do that.

It's not hard to protect a mosfet from failure. The only way to protect a tube from failure is to not use it.

A 1B3 makes a nice high-voltage capacitor, which is very reliable.

I have some beautiful tubes. 833 transmit jug. Some Blue Arcturus things. A krytron. Some acorn and prox fuse tubes. A few gorgeous CRTs and PMTs.

The last time I designed a tube into a circuit was around 1990, to drive a sensing grid in an air ionization measurement. It was an 811A, and was able to drive the grid to -500V (with respect to the other grids) in a microsecond or so, and then get completely out of the way of the charge measurement.

Tubes are still unequaled for that sort of thing—high voltage, high impedance, very low volume.

They stink for everything else, even apart from being big, inconvenient, power-hungry, and expensive.

Cheers

Phil Hobbs

...but that is the comparison with what I did by accident. I did it to a device that was designed as an RF amplifier with a rated dissipation of less than 3 watts and you were comparing it with a semiconductor that was massively bigger - and now you say don't do that - and the device needs extra protection components.

That was the point I was making: you can get away with mishaps in a valve circuit that you can't get away with in a comparable transistor circuit.