> I have background in electrical engineering
Sorry to hear that. Schools and training sessions don't give you the whole real-world picture.
> 600 ohm on R7 resistor before output transformer
Yes, but.....
The optimum Large-Signal impedance is quite different.
> ...but eddie Cilleti used 11k primary output tranny.
What can the 12AX7 actually DRIVE? We see it has 105VDC across 47K; round-off to 100V and 2mA.
If we used 1:1 ratio so the 12AX7 cathodes saw 600 ohms, the maximum voltage is 600*2mA or 1.2 Volts. And the distortion will exceed 5% by 1 Volt. And the cathode-follower gain would be about 0.5.
Using 11K load we can swing less than 11K*2mA or 22V peak at primary. An 11K:600 impedance ratio is 4.28:1 voltage ratio. We can get almost 22V/4.28 or 5V at secondary. And the gain is like 11K/11.6K or 0.94, the distortion is roughly 18 times lower or well under 0.5%.
That's for true 600 ohm load. When this thing was new, these were still around. We should ask what happens on a hi-Z load. Ignoring transformer inductance, the 600:600 iron which gave gain near 0.5 would un-sag to gain of 0.9, a significant rise. The max peak output of 1.2V would seem to rise to 30V, which could be bad. OTOH with 11K:600 iron loaded in hi-Z, the gain rises from 0.94 to maybe 0.96, "no change", the max peak output seems to rise from 5V to 7V, not a big deal.
> don't we have to have (output resistance) = (resistance on primary of output transformer).
Almost never. We usually want the driving small-signal impedance to be << load. When gain (and NFB) is cheap, we use opamps with 1 ohm small-signal impedance to drive 600 ohm transformers. In power engineering, the impedance of the dam and generator is much lower than the impedance reflected through the distribution transformers.
"Matching" gets the maximum power possible from a given source. But it also drops the voltage to half of the unloaded voltage. It would be awkward to get 120V in the middle of the night but 60V in the evening. That would also cause waste heat in the generator equal to the total load. Instead we scale the generator for 10%-20% sag at full load, then add negative feedback to regulate to 2% error from light load to heavy load.
Same in audio: we notice 10% sag as we connect different loads. When telephones were passive, the more phones connected to a conversation the lower the level. Now that we have amplifiers, we like to have sources <<600 and loads >>600. Typically 50-100 ohms output and 10K-22K input.
Low source Z to a transformer also reduces bass distortion. (Or allows a smaller core for tolerable distortion.)
So the impedance seen looking into the 1566's "lo-Z" output is about 11 ohms due to 12AX7. There will be another 30-60 ohms impedance due to winding copper loss. The output Z is under 100 ohms, which was low enough for the smaller systems where a 1566 would be used.
