eeeek---I've unleashed a monster (apologies to Hiraga). I started writing the below and there are a zillion posts since. But this one was more or less after the second-to-last of Svart's:
As far as gain, it would be nice to explore Paul Stamler's suggestion and have the first stage gain fixed---this also eliminates the need for reverse log. So the proposed structure is first stage fixed gain, attenuator, second stage fixed gain.
With that in mind though, I would now solicit input for what, within reason, should be our maximum input level. Figure we run up to nearly the rails of the suggested +/- 28V system. Let's give a little room and say +/-25V peak out; maybe internally single-ended, maybe internally balanced. Should we suppose nothing hotter than 100mV rms? That would work out to a numerical gain of about 177 for the first stage.
[added after some look over of the recent posts: maybe it needs an output attentuator if feeding boards that will be easily overloaded]
If the front end of this is doing overall about 1nV/rt Hz at the input, and the second stage is comparable, this would mean that you would start losing significant S/N ratio when the second stage noise gets comparable referred to the first stage. As well you have to consider the noise associated with the attenuator. There are dissipation considerations here as well---if we limit attenuator dissipation to 1/4W the impedance is 1.25k; the worst-case attenuator thermal will be at -6dB and be equivalent to a 313 ohm R or ~2.3nV/rt Hz, so that will dominate the second stage noise but still be totally negligible compared to first stage noise.
But our overall 70dB requires that the second stage have 25dB gain. Assuming a comparable output swing and balanced, this would allow as much as 2V rms swing at the second stage input, which would correspond to 11mV rms at the first-stage input. For the 100mV input the attenuator would have to be set to about -19dB to avoid second-stage output overload. The attenuator noise and second-stage input noise sums rms to 1.75nV/rt Hz while the attenuated first-stage output noise is about 20nV/rt Hz, so we are not degrading the first stage noise significantly.
I?ll have to run these numbers again and check, but it seems as if the 2 stage with intervening attenuator approach would work.
The predicted S/N is decent; neglecting noise in the source: at maximum gain, 1nV/rt Hz at the input (which could get lower if we really work at it---still a bit high compared to the 1.58nV/rt Hz of a 150 ohm resistor, but getting there) becomes 447uV in a 20kHz passband at the output, below 35.6V rms, or about 98dB. At lower gains like the one explored for hot sources the number improves to about 50uV rms out, or about 117dB.
If the mic had thermal noise of 150 ohms the max gain S/N drops to about 93dB. Note that if the preamp were noiseless, with this gain and output swing the best S/N possible would be still only 94dB. And of course real mics and rooms are going to be noisier.
I?m supposing FETs in the front end hence essentially no current noise, and no need for input d.c. blocks unless the mic is phantom-powered. If even lower noise is demanded the easiest way to get there would be with low rbb bipolars. Unfortunately both the FETs and the bipolars considered violate the easy-to-source requirement, although there are some easier-to-get varieties of both that could come within hooting distance.
