So I threw together a schematic at work (sorry 'bout the MSPaint ;- ] )
This is what I meant by 'opamp like' buffer stage..
Fully Bipolar design, cause thats the way I like it ;- ] I hear BJT's are good for audio. Whatever. If someone suggested FET's would be good, they should be rather drop-in-able.
Buffered difamp input - Sorta darlington'd, from what I've seen in simulations it doesn't really make much difference connecting the collector of the buffer to power rail or to the other transistor's collector. Someone might advise otherwise. Using a pnp input stage in an ACTUAL opamp I designed and built I pulled the input bias current down to around 20nA with 1mA current sink. So, with the worst 2n3904 the input impedance should be plenty low enough for just about any ap... And with something fancy like the BC550C's from the B22 (with beta's in the range of the 500's) or something like the Analog Devices MAT02 integrated matched pair (also high beta around 400-800) this input could rival FET input impedances. The resistor at the bottom of the input stage makes a Widlar current source incase you need to tune that current down from the reference current.
One gain stage means easier to put together. Also, one gain stage avoids any gain instability - no 'internal' compensation required. (with one gain stage of phase shift you only get 180degrees.. not enough for positive feedback)...
And from my experience (with the PNP input stage of my opamp) with 1mA driving the whole first stage (so .5mA per side) I got a gain of 100 driving a cascode second stage.. This active load isn't quite as sophisticated as my other design, but its also driving a much higher impedance... So.. I bet it does ok.. Av of 100 should be plenty.
Active Loaded Complimentary Common Collector Common Collector Cascaded Output Stage... Also referred to as Diamond Buffer ;- ] ... 'Cept this one is more efficient than some designs because of the current source loads and the easy tunability - you can pretty much set your max output current based on either what you want to drive or the current you have to spare...
This design can pretty much have the same quiescent current as a opa2134, meaning that it could pretty easily drop into some amps (I'm thinking specifically the SOHA cause I just built one).. Figure IRef as 1mA, driving 2mA quiecent through the output stage and 1mA through the input stage.. 4mA.. Same. Second amp should also be able to use the same reference current as well (of course, that should be tested to make sure there isn't some crazy feedbackandforth loop :- ] That means ~7mA total.. Then whatever you drive into your load.
So... What do people think? Might this be worth looking into? I figure it might be a useable alternative in some other designs too.. Feedback relieves need for dc servo's and intense bias trimming.. And bias trimming could be added to the input stage as well (just have to figure out the best way).
