I biult the amps almost five years ago now and frankly don't remember exactly how I got to the layout I have. 8^)
Blowing the dust off of my memory...
This design when I biult it was much discussed on the Joenet and was then referred to as the monkey amp or full monkey amp(don't ask me why). The design originated with Thorsten Loesch I believe.
I think the voltage across R2 is basically the PS for the driver tube and one wants a low impedance AC path around this resistor as in any PS. As R3 is not in this 'loop', I didn't see any reason to bypass it. The 'ground' of the driver tube is the bottom of R2, not the actual ground ground. If I'm wrong in this assumption, someone please explain.
I remember trying the design with and without this cap and liked it better with it. Personally I like resistors best when they are eliminated with some iron or if that is not possible, bypassing them. Do a search on the web for monkey or full monkey and you will find a few variations. Some don't use bypasses in the resistor 'tree' at all. Some bypass R1 & R2 like I did. Check out http://www.diyparadise.com/monkeydarling.html for starters.
As noted on this site, the lower the Rp of the driver the better. That is why I used the 6C45Pi and I run them low and hot to get the Rp down.
Steve Bench's hum bucking circuit cut the really bothersome (to me) 120 Hz hum without having to resort to current regulated DC heating which would have been my next option. It is such a cheap option to try and is easy to retrofit.
My PS is choke input 10H/47uF/10H/47uF from a full wave 400-0-400V winding. I never felt the need to use more capacitance as I can live with 3.5 mV of ripple at the speaker terminals. You seem to be using lots of filtration. I don't think that the PS is the source of your problems. Try bypassing R2.
Regards, David
PS: I also added UTC A20 input transformers (600 ohm:600 ohm) which cured some nasty ground loop problems I had.
