Just for completeness on this, I finally finished work on the 3rd prototype last week. I discovered a few things that wouldn't have helped during my disassembly. The 220nF capacitor I thought I'd bought was actually a 22nF capacitor and had been since day one. There were one or two dodgy solder joints and crimps, mainly on power connections. The most recent discovery is the new tung sol EF806 I had bought runs out of spec as it causes untold chaos to voltages across V1 & V2which results in a lot of audible hiss. I decided to do what I could to build the "MK3 prototype" using aluminium. I had some old 19" rack blanking panels so opted to build it on that.
What with all the prior issues I'd had, I decided to just go to town on this one and make as many of the supposed improvements that the design "needs". (according to some) This meant taking the 12AX7/ECC83 out and use a medium mu high GM valve for the phase splitter. I opted to go for 6SN7/6CG7/6FQ7 in preference to 12BH7 or 12AU7, mainly down to the simpler heater connections. The power transformer had the capacity to cover the increased heater current. I also took Andy's advice and built a CCS on a piece of veroboard. A more controversial change was the suggestion of using a "combination bias" using resistors and zeners to help drop the quiescent without sacrificing much if any power. It seems to have cured the red plating issues and I can easily get the 20W that the amp is rated for. The disadvantage is there is it's not as tolerant to valve mismatch as plain cathode bias. Hopefully there is enough give in it that the ageing of the valves doesn't prove troublesome.
This is the schematic I ended up at after a few feedback network tweaks. I think there may be room for improvement with the feedback as I've simply gone for getting it stable no matter what. I think the step network may be stronger than it needs to be. The target was 17dB of feedback and I think I got about 16.3dB with 18K.
https://***********************/65535/...1309f8f6_h.jpg
CCS built on veroboard using what I could find in my component stash.
https://***********************/65535/...fa4982a5_h.jpg
100Hz at the onset of clipping. A big win for me as there wan't a hint of ringing.
https://***********************/65535/...2ae35dda_h.jpg
20Hz at the edge, again no ringing.
https://***********************/65535/...61e37273_h.jpg
1KHz sq looks quite good.
https://***********************/65535/...e351bd69_h.jpg
10KHz is showing some HF bandwidth limitation but some of that appears to be input capacitance which I think relates to a piece of foil backed card I used behind the veroboard to act as a shield. (there is slight rounding over of the input signal after the veroboard)
https://***********************/65535/...31d5f16d_h.jpg
The input veroboard was created to help fit the infrasonic filter.
https://***********************/65535/...2a001b18_h.jpg
The complete interior. It's a bit big as it stands but it made building easier. I will be working on the final construction being a bit smaller, closer to 400x250mm.
https://***********************/65535/...a1fdc5be_h.jpg
I've been using it now for a couple of days over the last week and think it sounds great. It's not easy to differentiate between it and the solid state power amp I normally use. I think since this is no longer a proper 5-20, merely based on one, a mod could move this thread into the DIY section. It's closest relatives would be the Dynaco A-420 modified mullard design or the Eico HF-50/HF-60. I'll try to keep this updated as I build the pair in their final chassis.