Really subjective question

[Yvesm]

Use no cap !
Neither in grid nor in cathode.

Yves.

 

[regal]

You have to look at the time constants- that big ol' cap has to discharge itself through the tube's cathode to return to proper bias after overload. That's assuming that clipping hasn't caused a problem with the current sink.

Making an amp overload-proof is not a trivial proposition and is too rarely done, IMO. Moving the cap doesn't help that, it just allows the substitution of a MUCH more ideal cap (sorry, there's not a 50u cap in the world that has the performance of a one dollar 100n polyprop and foil), and saves cost and parts count.

The bias needed to get the low Rp on this tube is in the 20V's, so LEDs aren't an option. I doubt I could build a PS as quiet for negative grid bias as a 50uf Cap, so I'm back to the ultrapath. If you are saying this is a bad idea and interstage decoupling is best, it probably makes since to turn this into a parafeed amp with a CCS load on both tube sections. Still would need a big cap on the cathode. Oh the choices, I feel as this will never get off the ground 🙂

 

[SY]

I doubt I could build a PS as quiet for negative grid bias

Let's see, if you've got a 25V grid bias supply, a noise of 0.25mV is 100dB below that (good enough, wouldn't you say?). And getting noise that low at low current demand is pretty trivial. Three LM329s stacked in series would cost about $5, give you 21V of bias, and have a noise less than 12uV (-125dB!). And that doesn't include the noise rolloff from the combination of the large grid leak resistor and the tube's Miller capacitance.

 

[Yvesm]

. . . And that doesn't include the noise rolloff from the combination of the large grid leak resistor and the tube's Miller capacitance.

Not to tell about the (necesary low) internal impedance of the driver 😉 . . .
I bet that most of the noise will come from there.

 

[SY]

Excellent point that I had overlooked!