> someday we'll meet and I'll have to buy the beer.
Sure, except I'm not going to Minnesota this week (brrrr!). You drink it for me.
> While the schematic you posted isn't in line with what I'm working on
I hope not. It should work, but is clearly not optimized. The output swing (not even 15V peak) is rather lame for +/-15V supplies. The base resistors in the 2nd stage are a heavy load on the first stage and might be reconsidered. And no, Henk, I have not figured the complete specs, except I figure it will bring a mike up to "line" without any great flaw. (Hmmmm... the bandwidth might be MHz, which actually could be a "flaw" in a high-RF location.)
Also, everybody should note that Rod's design predates common use of Phantom Power, and has no provision or protection from Phantom. Use the usual scheme: a couple of 47uFd caps, + side to the XLR, with a couple 6K8 resistors to a 48V supply. Protection diodes are wise but perhaps not really needed with resistor bias and these sturdy switch-transistors (and no critical DC offset worries).
> it caused a flash to happen
That's what wacky un-optimized ideas are for. Back around 1970 some clever guys drew transistors every which way, in 1980 they added IC opamps, and ever since we have been basically using the SAME few topologies. Because they are simple and they WORK. But the drop in part prices and the rise in labor costs make other topologies interesting. Ten transistors was a lot in 1970, but today the 10 transistors are about a buck while in DIY your value-of-labor is going to be $100+: layout, PCB etching, stuffing, debugging. And in commercial work, parts like these have little effect on selling price. So what can we do with a dozen 10-cent transistors? And how many of our clients need to see "totally discrete amplification"?
(Not that there is anything wrong with ICs. The selection of parts is limited but the parts are super-cheap so we can mask-up schemes too messy to solder. Baked-in matching makes some nice things possible. The main drawback seems to be that ICs only make sense if you can sell 100K of the exact same thing, and the audio market is not that big, not at the high-end.)
So it is time to draw things every which way, throw them on top of each other, and see if anything sticks. Here I just threw two of the same thing down, and saw that they butted-together with a usable result. If this landed next to something fluttering in your brain and went FLASH, groovy.
The idea of using a dual-linear pot to control the gain of both stages is interesting. Dual-linear may be easier to find than reverse-audio. (Though either way, the value needed is low for a pot.) Instead of gain varying linearly as with a single-linear pot, it changes as square of gain. With a small value of R9 in each stage, gain increases rapidly in the high-gain end of rotation, but not as bad as simple linear control. However if you make each stage's R9 about 250Ω and each VR about 1K, you get about 22dB at one end, 36dB in the middle, and about 52dB at the other end. That's only 30dB of range, but the dBs are nicely spread around the rotation. An even better scheme would be to make first stage R9 very low (250Ω is really too high for great noise performance) and the second stage R9 maybe 250Ω or somewhat higher. Then at the high end, the second stage (and overall) gain doesn't soar as you near the end. My pencil isn't sharp enough to work out this peicewise-linear approximation to a linear-in-dB curve. I suspect that for a good 40dB or 50dB range without any sharp kinks you need a triple linear pot, not a common part. (There is much to be said for switches, if you don't need to trim gain on-air.)
> possibly solve a problem I've been wrestling with for a while.
Do what they do in the WWF. Dance around the problem, snarl, do an armlock and a headbutt, then if not pinned yet just whack it with a chair and throw it out of the ring.