That's a classic article. Given the number of McIntosh amps sold over the years, and the enduring popularity of the QUAD II, I wouldn't totally write off cathode feedback - but for someone building at home with off the shelf parts, it's not really a realistic option. I'm sure that Sowter could do a one-off, but get our your wallet!

Back to the original topic, I think the 6AS7G was used most widely by the military, in all sorts of regulated power supplies. In my years of collecting I have come across multiples of military surplus 6AS7G and 6080, but not even one in a consumer targeted box or package. I even built a regulated power supply using one, and it worked really well for what it was.

One strike against the 6AS7 for audio is it's not that linear - sure, all the second harmonic is cancelled by push-pull, so it still does decent with distortion measurements this way, but have a look at the characteristic curves, they are not evenly spaced at all, and lots of "knee" near cutoff. In comparison to a linear triode such as a 6SN7 or a 2A3, they are very compromised.

The 6AS7 is not as linear, no, but the distortion that is does generate is mostly second and fourth harmonics, which get cancelled away as you said. In simulations using LTSpice, I'm able to get lower distortion from push pull 6AS7Gs than 2A3s, and for the simple reason that the 2A3, while producing less distortion overall, does produce more 3rd and 5th order.

You also have to keep in mind that the 2A3 is a filamentary type tube. There is much more 60 cycle hum to contend with than the 6AS7G, and that will mix with the actual audio signal to produce more IM distortion. The 6B4G is even worse, the 6A5G and 2A3H types are the only 2A3 variants, aside from the 6AS7G, that I know of that are relatively free of this pitfall.

Attached are some plate curves, the 2A3, 6AS7G, and 300B in that order, generated from Ayumi's extremely accurate tube models; they agree with the plate curves published by RCA and Western Electric. Sorry, they're kind of dark.

Wow a full $3 more (net price) than the 6W4 damper diode , that was a good deal more expensive .

In the context of modern designs, it's possible to use DC on the filaments of tubes like the 2A3, and get rid of that issue. I know some people almost religiously stick to AC heated filaments, but in my 2A3 amp I used DC and am very happy with the result.

The bias is then off somewhat along the length of the filament. If the reference to ground (or the self bias resistor) is halfway between the ends of the filament, then the bias is only shifted by 1.25 volts, which is negligible. If one end of the filament is grounded (or tied to the self bias resistor) now our bias is off at the other end by 2.5 volts, which, with a high perveance type valve, ought to be avoided. Still, for the 2A3 the DC filament situation is manageable.

The situation for the 6A3 and 6B4G is worse. 3.15 volts if the filament is tied to ground or the self bias resistor at the midpoint, a whole 6.3 volts if one end is grounded. This is non-negligable just from looking at the plate curves...

Still, a 2A3 on DC heaters, while more likely to fail prematurely does sound nice.

If we are considering modern advantages, then there is one area where the 6AS7G is far and away superior to most other tubes. It is ironic, considering that it was probably the biggest stumbling block to widespread adoption back in the day: cost!

I purchased the OPT for my 6AS7G amplifier project from Sowter in the UK today. The total cost was £250 or so with shipping.

I also managed to buy NOS JAN 6AS7G, in the ST envelope, for $10 a piece today. I bought 20. Assuming I have to replace the 6AS7G once every year or two, I'm looking at a 20 to 40 year supply of output tubes.

In order to secure a 20 to 40 year supply 'monoplate' RCA 2A3s, or Sylvania 2A3Ws, or Raytheon 2A3Hs, I would have to sell a kidney...