That's a cute little article.

The key diagram is the one that shows the signal path through the amplifier. Input feeds grid, plate feeds next grid, final output is from plate. Everything else is supporting circuitry.

Note that between each stage there's a capacitor in the signal path. That's to block DC. If you want an amp that amplifies DC, each stage has to run at a higher voltage than the previous stage. The plate must be above the grid in voltage. This was a huge headache in tube computers, both analog and digital.

Transistor circuits don't have the increasing voltage problem. Outputs and inputs are in the same voltage range. That's because transistors are current gain devices, not voltage gain devices.

▲

monocasa 10 hours ago | parent [-]

> Note that between each stage there's a capacitor in the signal path. That's to block DC. If you want an amp that amplifies DC, each stage has to run at a higher voltage than the previous stage. The plate must be above the grid in voltage. This was a huge headache in tube computers, both analog and digital.

You can also stick a voltage divider (and probably some diode clamping) in there to pull the signal off of the plate down to a grid compatible voltage for the next stage if you're just doing digital computing. That was the most common setup I've seen in tube based computing. They tended to play pretty nice with the resistors needed for the plate current anyway so it wasn't that much extra RC constant delay.

▲

HPsquared 10 hours ago | parent [-]

That won't help with the power consumption though, I guess. (Or is that a rounding error compared to everything else?)

	▲	monocasa 10 hours ago \| parent [-]
		It's not exactly what I'd call a rounding error, but it's manageable. But yeah, tube computing in general is an exercise in building a really fancy space heater. I'm trying to keep my tube computer I'm building down to ~3KW, and that's probably the biggest actual constraint on design complexity.