Looks like this trick is very promissing... but does anybody really understand what it does ?? it's quite confusing to me.
Jeff detailed the basic action. The excerpt from ToneQuest is talking about how a single tube using an unbypassed cathode resistor has negative feedback due to the current flow through the resistor. The voltage developed across the resistor is directly proportional to the amount of current flowing, and that current is wobbling up and down during a signal. That means your bias voltage is also wobbling up and down, with the end effect of counteracting the current the tube is passing due to an input signal. Hence it is "negative feedback" but at one given tube stage.
Normally, we counter this effect in a preamp tube by bypassing the cathode resistor, which bypasses the alternating current around the resistor and keeps the voltage across the resistor steady. It restores gain lost due to negative feedback.
In a push-pull output stage, you could use a shared cathode resistor, and not need a bypass cap. If you assume somewhat matched tubes, then during a signal swing one tube is conducting more current while the other tube is conducting less current by more or less the same amount. The total current passing through the resistor remains unchanged, and so the voltage across the resistor remains unchanged. Therefore, there is no negative feedback effect, even without a bypass cap.
All I have just written is a restatement of exactly what Jeff already said.So tell me please how this works?
Well, if we don't use a common cathode resistor, then there will be negative feedback at the cathode due to the current flow during a signal. We would have to bypass each resistor to eliminate the negative feedback and have full power. If we did not bypass the resistor, the negative feedback due to the resistor reduces the current swing through the tube due to signal, but that also
reduces the voltage swing at the plate of the output tube, because the voltage swing is caused by a current swing through the primary impedance of the output transformer.
In the end, less current swing times less voltage swing equals less power output.
What about the bypass caps? If we bypass, we get full power. But what's up with the funky effect of connecting the negative ends together??
Look again at Geezer's drawing.
If the negative ends of both bypass caps are grounded, are they not effectively connected together? Yeah, they are connected to true ground too; however, if you disconnect the negatives from ground, but keep them connected to each other, they have a "virtual ground" at their junction and have the same end effect. In fact, 2 caps connected in this manner are essentially in series, and the net value is half of the value of each individual cap. The pair of (maybe 22uF) caps could be replaced with a single 10uF cap. This idea (a pair of bypass caps being replaced by a single cap of half-value connected between the 2 non-grounded points to bypass) has been used before by Quad amps and was explained by Morgan Jones in Valve Amplifiers
If you wanted to replace the pair of bypass caps, you'd want a non-polarized cap, but you'd need ~10uF which might be bulky in a film cap, and a non-polarized electrolytic is basically 2 back-to-back electrolytic caps in one package (and much more expensive). If you did replace with a single cap, I'm not sure how effective a single pot would be from one leg of the cap to a single output tube cathode, but it could be tried out for the sake of experimentation.
Anyway, a single pot between the 2 caps as Geezer shows destroys the bypassing effect due to a virtual ground as the resistance is increased. In the end, this is atypical, fairly unusual, but not voodoo or make-believe. Pretty cool rediscovery/re-application if you ask me.