If a 5687 can be the final driver, it can also function as the input as well. The PSRR of the common cathode amp will allow a single stage B+. All the stages are class A, which makes it even more tolerant of potential PS noise. For those of us with more gain in the linestage than we need( right now ), a gain structure like this should not be an issue.

Now, with a higher Z primary, we'll be able to run higher B+ and remain Class A. Higher B+ is quite useful for the driver stage. Swinging the +/-70V from a 300V supply won't be so difficult with a 5687 anyway. For NOS valves, the plate ratings may be close. I would take the Sovtek 2A3 for this one. It is more tolerant of 2A3+ dissipation.

Fixed bias will keep the cathode circuit free from big and imperfect capacitors. Individual bias will require seperate filament supplies, and a split CT on the PP grid choke.

I have been itching to build just such an amp. I have a few possible OPT's to experiment with. Which ever one sounds better, gets torn down and put on the NC winder at Heyboer. They are all about the same size. Cosidering how much the bigger S-265-Q cost to clone and modify, the little ones could probably get made at Hammond-level price tags. Unfortunately, the bigger multi-grid finals are calling to me...must get that out of my system( or perhaps thoroughly into it ). The idea of an 80W 813 amp has me all worked up.

On the power Iron. I have in mind a rather expensive piece. Multi tap primary and two voltage CT secondary. It would be able to deliver B+ from an L-C filter from ~320V all the way to 750. So AC would be 400 to 850, with about a half dozen steps in the middle. The long-term, multiple build power TX. Buy one now, instead of several later. Probably cost less than double a single-project piece.
cheers,
Douglas