LTP PI cg2 question

[Chuck H]

I'm working on a design and I'd like to try coupling the reverb at the "other" grid of the LTP. All the amps I've built use the cg2 .1uf bypass cap as seen in all Marshall and Fender amps. But vintage VOX amps seem immune to the charms of this capacitor?!? I think I can see what it does by being there. Effectively reducing grid impedance so that electrons can freely pass from the cathode to the plate (corrections accepted). But what happens when it isn't there. As it seems it doesn't need to be. Both the Aiken and Valve Wizard articles on the LTP pretty much just say 'Make sure this cap is big enough to pass all frequencies to ground for best results.' But I haven't found a comprehensive discussion about omitting it or using the non inverting LTP grid as an input.

TIA

EDIT: cg2 would also remove local NFB, yes? So there would be a gain difference. I'm not good with the formulas so a simpler explaination would be appreciated.

[Steve Conner]

Hi Chuck

The capacitor provides a signal path from the second grid to ground, which is needed for balance of the PI. Without it the grid wouldn't move relative to the cathode and there would be no output from the second PI section.

It doesn't matter whether you connect it straight to ground, or couple a signal in through it, as long as the reference is there.

When global NFB is used, the capacitor is one of the time constants in the loop. If you make it the same size as the PI input coupling capacitor, you can get motorboating. So, designs with NFB oversize it to 0.1uF. The AC30 didn't have NFB.

[Chuck H]

Thanks Steve. I guess I don't have my head wrapped around this though. How can the cap being grounded have the same result as if it's coupling a channel if that channel isn't being used? Well, whether I get it or not, there it is.

[Steve Conner]

The signal you're coupling in through the capacitor will have an output impedance. For proper PI functioning, that has to be relatively low, even when the channel "isn't being used".

So for example, if you turned off the reverb by opening a switch in series with that 0.1uF cap, the PI would stop working. If you turned off the reverb by shorting the signal to ground, all would be well. But on a system level, both of those options would be a form of "not being used".

In the AC30, both of the channel outputs have an impedance of 50-100k or so at most, and of course when a channel volume is set to zero, the wiper is a short to ground: the good kind of "not being used" as above.

This is one of these linear-superposition-mathy things that's hard to grasp intuitively. The old engineers were good at it, though, as they didn't have PSpice and had to get maximum functionality out of minimum component count. They couldn't afford the luxury of neat (and bland-sounding ) op-amp blocks with infinite input impedance and zero output impedance, as we do nowadays.

[pdf64]

Yes, the intuitive thing is that the grid impedances should be low to enable effective grid decoupling, but low compared to what?
ie the WEM Dominator / Marshall 18 Watt feed the grid caps with a 500k vol control pots, so the impedances there could rise to 250k, dependant on the pot setting. That seems high but on reflection, the grid input impedances will be >1M, due to the grid leak resistors being bootstrapped.
It would be interesting to investigate whether the vol pot setting of the 'other' channel affected the LTP outputs balance on this arrangement.
Pete.

[Steve Conner]

The bootstrapping works for DC in the common mode, but does it work for signals in the differential mode?

I'm not sure, but I think it doesn't, because differential mode signals are supposed to cancel at the cathodes, where the bootstrapping voltage comes from. Then the AC input impedance is just 1M. Or maybe 2M if you measure one input with the other one shorted...

[Chuck H]

He he... I knew this would happen. A discussion of the undiscussed PI design parameters. This forum is the best for that sort of thing bar none.

I actually did get as much as to think the unused channel would need to be low(ish) impedance. Figuring the high input impedance of the LTP, and in light of the info given, I figured that for a grounded C2 to be similar to an ungrounded C2 that it must be an impedance thing and the whole point was an AC 0V reference. And this is exactly the sort of info I wanted since I now know to design the rest of the circuit to offer the lowest possible impedance at either input of the PI.

And... Since the LTP input impedance kind of tanks when clipping, and the VOX amps have an ungrounded (higher impedance) cg2, I wonder if this has anything specific and profound to do with the famous AC30 overdrive sound.

[Gaz]

Just had to mention that I've left this cap out on three builds by accident! Drove me nuts. The amps still sounded pretty good, just put out 1/4 of the power.

[Steve Conner]

Well there you go, a new kind of power reduction switch that doesn't hurt the tone.

Patent it quick before Mesa does!

[EFK]

I'm not Randall Smith, but I did stay at a Holiday Inn Express last night......

(Thanks Steve, now I have to clean all the coffee off my screen)