> Bipolar PSU for tube seems like a great idea
Build it and see.
Cost is a real issue. Parts; also wiring. And with vintage gear, reliability.
I've done a +/-300V +50mA/-2mA power supply for a headphone amp with a lonnngtail driver. In retrospect, it was silly to do so.
I have higher respect for John than most latter-day tube writers. But this one sounds like he has never been crushed by a large tube mixer project.
"....fails to provide an unity gain output."
And so? First: tube gain is precious, we don't waste any. Second: the conventional plan on page 1 is unity gain for each input, the plan on page 2 is unity gain for all input strapped together. He knows better; his brain ran off in two directions.
We can easily adjust values for the conventional unity gain per input, IF that's what we want.
He purports to be avoiding feedback but there's no other way to look at a triode, especially one with a whopping big cathode resistor. True, this does avoid mixing all sources within the cathode NFB loop, which may well affect the sound of a busy mix. But heck, why stop here? Bring all sources up to +30dBm and then use pure precision resistive mixing to sum to a +18dBm output. It uses more tubes, but he's already proposed using more tubes than any old-time mixer designer would contemplate.
Now what's this with the negative supply? He could just as good strapped the grids up to +100V and returned cathodes to common. He does lever this into his philosophy of dumping power noise into audio amplifiers in a way to null the output. Whether power noise reduction is a responsibility of each amp or of the power supply is a point to ponder. However in many-stage systems and for "reasonable" crap specs, one good power supply is probably more-buildable than judo-ing many individual stages.
There's other differences. His feedback mixer and his grounded-grid mixer have 10K inputs, the many-tube mixer has hi-Z inputs. We need to feed each input, and I assume with a tube; high level in 10K is hard work for tubes.
Further on.... the "White Cathode Follower inspired" mixer would baffle White. Given two tubes, you could get the same performance in parallel, without awkward voltages. And of course it is just another Feedback mixer, but with a long NFB path.
And he's not going to get negligible input offset voltage drift. In fact it will be pretty horrid for the first couple hours. Best to use the dang capacitor and be stable by the time the talent is ready to go and for years to come.
And where the heck is the physical reference for the virtual ground? On the simple ground-grid mixer the reference is plainly grounded, input current is trivial, the reference input could be connected anywhere (and in a large mixer, it will matter). His WCF-inspired idea has the reference input hanging halfway across a bipolar supply in a 1mA string and showing 100K to the grid. Just the high grid impedance will trash the GG's supersonic response (Miller never sleeps).
Food for thought. Not a fully thought out plan for a complete mixer.
