>> the biggest difference seems to be that the #66 uses a two-transistor input stage (like the B Unltragain, Midas and some others) whereas the Green uses one (like Mackie).
> Don't have the Green-sch here, but (instead of considering the number of discrete devices) wouldn't the biggest difference be that it's a 'different kind' of pre? (current feedback, iiric)
The gross difference is: some BJT+opamp designs take feedback around the whole thing, some just in each stage.
Rod's #66 has the BJTs working with only local feedback. That's hard to do well at high gain and/or level unless the transistors are compounded (Rod #66 PNP/NPN compound).
Many others bring the opamp outputs back to the input emitters. This gives enough feedback to use simple transistor inputs and get vanishingly low THD numbers. There is a subtlety here: if the opamp is compensated for unity gain, and the BJTs add gain inside the loop, it is sure to oscillate. And we don't want to over-compensate for less than unity gain, because most opamp slew-rate is barely good enough at unity-gain compensation. So if you'll note: the overall feedback resistors are (almost always) bigger than the transistor collector resistors. That keeps the opamp working at a gain somewhat over unity, with the BJTs supplying most of the bulk gain, and the opamp reserve gain just fighting BJT and opamp nonlinearity.
On paper, overall feedback may be a little better. In real life, simple things are often best, and the local feedback variant like Rod #66 has simple feedback.
There isn't any compelling reason for these opamps to be "good" opamps. They need great audio performance, but the DC performance can be awful. They can have high input current, and would work fine with DC offset of many-many-many millivolts (even 600mV, meaning we don't need diff-pair inputs). That's what I was hinting at: we do NOT need full opamp specifications, so why build something "universal" for a very specialized job?
And if you want to avoid opamp-think completely: take the BJT input of Rod #66 and do it twice, first feeding the second via caps. To duplicate his gain structure, the emitter resistor in the second stage should be fixed around 75 ohms. However you might instead vary both emitter resistors with a dual pot. And for reasonable range of gain, a dual-linear pot would give a square-law curve not that different from a "proper" reverse-audio taper pot. The output impedance is quite high, like 4K4, so you want a couple emitter followers. This BJT topology isn't optimized for output swing, but you could still reach +18dBm which is a usable value. (Or use hi-volt transistors and scale-up to say +/-30V.) The CMRR isn't "infinite" and varies with gain, but could be 20dB at low gains and 40dB at high gains, more than enough for any place you should be using recording gear.
Oh: all these plans were using "Current Feedback" 20 years before it became the darling buzz-word of the chipmakers.