Question about Pentode reverb drive circuit

[SoulFetish]

You are deriving the grid potential (12V) of the recovery triode from the cathode network of the power tubes. I don't think this is a good idea, as the cathode voltage is not constant at large-signal operation. Also the ripple filtering at this point may not be good enough. May even contain some 60Hz heater hum, especially as the cathode bypass cap is rather low capacitance at 10µ.

Good Point. I did not test the stability under large signal conditions! I actually remember mulling this over when I first saw the circuit and amplifier, and wanted to double check that the Heaters weren't wired that way by a mistake. Initially, I added a 1k series resistor and large bypass capacitor of 220µF from grid to ground to try and stabilize any modulation. But I suppose the cap would just charge up to the peak voltage anyway, right? I removed it as you can see from the schematic.
The cathode bypass cap is rather low for my taste, but it's one of the few areas of the original amp left
I really don't think this stage generated any significant noise though. At least not the nasty noise coupling in at idle I'm talking about. I pulled the tube completely out of the amp with no change in the loud hum. I also went through added a parallel 22µF capacitor to each preamp filter node one at a time to rule out a failing or undersized capacitor. It did quiet down some noise in the 12AX7 PI stage a little, but did not solve the main issue. I'm really suspicious of ground loops and I'm about to go back to my bench and start trouble shooting after this. What makes me nervous about if this is the case is the amount of work I would have to do to lift all the chassis connections in here

Why don't you use a voltage divider from node B and if necessary an additional filter cap across the low-side resistor?
The grid supply must be particularly clean, as the grid is still an input and any tiny ripple content gets amplified by the tube.

I appreciate the suggestion and It's good advice. It's easy enough to make that change accordingly. Here is the change with the option to bypass the screen.

So, I chose the values of !M/51k, because (obviously) I achieve my design screen voltage, minimizes any loading on the stage, and I know I have those values on hand. I'm going to double check that the grid leak value of the parallel combination of these values is okay. Does this seem appropriate? The .33µF was rounded up and should give me a rolloff under 20Hz.
I do remember coming across some literature somewhere advocating leaving the screen unbypassed with evidence indicating an increase in the PSRR of the stage compared to measuring the bypassed stage. I may leave it unbypassed and see where the degenerative (or perhaps regenerative?) feedback takes us.

[Helmholtz]

I appreciate the suggestion and It's good advice. It's easy enough to make that change accordingly. Here is the change with the option to bypass the screen.
[ATTACH=CONFIG]51445[/ATTACH]
So, I chose the values of !M/51k, because (obviously) I achieve my design screen voltage, minimizes any loading on the stage, and I know I have those values on hand. I'm going to double check that the grid leak value of the parallel combination of these values is okay. Does this seem appropriate? The .33µF was rounded up and should give me a rolloff under 20Hz

Looks good to me.

I'm going to double check that the grid leak value of the parallel combination of these values is okay.

This I don't understand. In most cases even 1M is considered low enough for triodes. How would you check?

The .33µF was rounded up and should give me a rolloff under 20Hz

In combination with 51k the corner frequency is 9.5Hz. Higher cap values give better ripple/hum rejection. Easy to test by paralleling caps.

I do remember coming across some literature somewhere advocating leaving the screen unbypassed with evidence indicating an increase in the PSRR of the stage compared to measuring the bypassed stage. I may leave it unbypassed and see where the degenerative (or perhaps regenerative?) feedback takes us.

What screen?
A control grid does not draw current, so an unbypassed grid "leak" resistor cannot produce a feedback effect like with a screen.

[SoulFetish]

Im calling g2 the “screen”, but understand the differences between a true pentode and cascode.
I wasn’t talking about g2/screen current. Any change in power supply voltage, is a change in an unbypaased g2 voltage. Wouldn’t the voltage change at the screen have an effect on the voltage at the anode? And if the power supply modulation was caused by dynamic signal conditions, wouldn’t that be a form of feedback?
Maybe im incorrect or overstating it

[Helmholtz]

Im calling g2 the “screen”, but understand the differences between a true pentode and cascode.
I wasn’t talking about g2/screen current. Any change in power supply voltage, is a change in an unbypaased g2 voltage. Wouldn’t the voltage change at the screen have an effect on the voltage at the anode? And if the power supply modulation was caused by dynamic signal conditions, wouldn’t that be a form of feedback?
Maybe im incorrect or overstating it

Which tube are we talking about? What is g2?
There is no g2/screen in your recovery triode. The control grid of the upper triode in a cascode does not really behave like the screen of a pentode. The typical screen (negative) feedback is caused by the signal voltage across the screen resistor(s) produced by screen current. A triode grid does not have this.
There should not be any signal voltage on the supply voltage. Something else would be the time lagged compression caused by supply sag.

[SoulFetish]

Which tube are we talking about? What is g2?
There is no g2/screen in your recovery triode.

I apologize if I'm butchering common nomenclature. I've seen examples where people have used the term "screen" to refer to the upper triode grid. Perhaps they were using it informally. In any case, it was not my intention to introduce any confusion. As far as the triode grid goes, I suppose that was a bit of inexperience and ignorance on my part. In an all tube cascode, I guess I'm in the habit of referencing the upper tube's grid as grid #2 when using the bottom tube as the input. I feel a bit foolish now, it seems a bit obvious that because the bottom device input terminal is now a gate... I suppose I don't get to assign something "2nd" if there's only f*cking one of them, right? That dont make no sense.
Do I sound like Nigel Tufnel in spinal tap, explaining to Rob Reiner how important it was to be able to turn this amp up to 11 when you need that little extra output.?

There should not be any signal voltage on the supply voltage. Something else would be the time lagged compression caused by supply sag.

I was thinking about this as a component having an effect on the grid voltage. But your right, that's something esle all together. I guess I was just trying to reason why the article advocated leaving the grid unbypassed to improve the PSRR over bypassing the voltage divider. I wouldn't have guessed that was the case, so I wondered if there was a negative feedback component contributing to this. But, I shouldn't have even brought it up without having the text to share and a better understanding of this circuit. I just wanted to reply and tell say my bad, Helm.
I want to get this back on the rails, because you brought something up a couple of posts back, which may be significant and I need to investigate it tomorrow. It reminded me of something, but I'm gonna' get some sleep and work on this a bit tomorrow. To be cont....