This is the third iteration on this power supply design. As mentioned in the discussion of the Rev1 version, there's a problem that arises through an evil interaction between the Standby switch, added sort-of at the last minute of the design, and the capacitively coupled bias supply, that gets its operating current from the same HV winding as the B+ supply. See that discussion for details.. in short, this arrangement pumps the votlage on the Standby switch terminals (and the HV winding ends) down as low as -1100V, which seems... bad, somehow. Particularly since the little Carling Standby switch's insulation is really only good for 1KV, according to UL.
Anyhow, in this version, I just bit the bullet and put another transformer into the design. Sometimes parsimony isn't economy, I guess. With the bias supply no longer associated with the HV supply, I was free to operate it un-regulated, which is actually a good thing, and completely customary for guitar amps. The reason this is good is: as the AC line votlage increases, the HV supply to the PA plates increases proportionately. By allowing the bias voltage to go more negative as the line voltage increases, the plate dissipation in the PA tubes is held relatively constant. And, of course, the aforementioned very high negative swing at the Standby switch is just the usual 'fairly' high voltage, but well within the switch's dielectric ratings.
Other than that change, I substituted some high-reliability axial-leaded Nichicon electrolytics in the filament DC voltage doubler, replacing the radial leaded Vishay 10mF caps I used previously. These new 6.8mF caps take up less real estate on the board (they're taller), but they also have 5,000H lifetime at 105C, and higher ripple current ratings.
I'm leaving the standby switch and B+ voltage limiting series resistor in the negative side of the supply, since it doesn't really matter where it is at this point, so far as voltage standoff.