To avoid running into an XY problem, here is the problem I really have:
I have a RODE NTG2 microphone, and a Zoom H4n Pro recorder. The NTG2 has an XLR output, balanced, and it operates from either a single AA battery, or 48V phantom supply.
The H4n is a portable recorder that records to SD card, up to 24bit/96KHz uncompressed WAV or MP3. It has a combination XLR+1/4 jack input. If you plug a 1/4 connector, it is a line-level, unbalanced input. If you use the XLR connection, it's a mic-level balanced input. The recorder automatically switches between inputs depending of what type of connector you plugged.
The general consensus on the web is that this mic isn't "hot enough" for this recorder, as the recorder expects higher inputs. For example, I have to set maximum gain input in the recorder and speak right into the microphone to get -12dB peaks in my recording. This is not acceptable since this is a "shotgun" microphone and you're not supposed to speak right into it.
Also, when using the microphone with 48V phantom power, you get more gain out of the mic. Apparently +6dB. I haven't accurately measured, but indeed, switching phantom on and off, the mic sounds "much louder" with phantom than with batteries.
The solution proposed by most people online is to use a device called FetHead that will boost the mic +27dB and this makes it manageable for the Zoom. But this device is almost $100 and I don't live in the US so it's a long delay for shipping.
So I googled for several options and, as usual with audio circuit googling, I ended up on ESP's site, particularly in the Project 66.
In this project the first stage is two discrete amplifiers based on regular NPN/PNP general purpose transistors, and a opamp-based balanced to unbalanced stage.
Why does he use a discrete preamp? Is this just an old design?
Then I googled for other options. I found others with discrete preamp stages, others with instrumentation amp stages, another circuit with JFET, etc.
I have all sorts of ICs in my junkbox: NE5532, OP177, OPA2134. JFET I only have BF256. I also have BC847/857 for SMD. I don't have the more exotic THAT1512 which seems like an ideal solution.
Finally the question:
Should I use a balanced input/unbalanced output opamp, and take the mic-level output from my mic to line level? My recorder doesn't have line-level balanced inputs.
Should I just boost the signal from the mic from mic-level balanced to mic-level balanced? I was thinking I could use just the first stage of Project 66 for this. Leave out the opamp section.
Best Answer
A discrete stage (with PNP input transistors) into a medium-noise opamp like the 5534 is a common way of getting low noise, without the expense of low noise opamps. With a bit more care, it's not hard for a discrete stage to even beat the THAT Corp amp you linked.
NB the secret to low noise in this simple configuration is to keep the emitter resistor (R9/VR1) low - and that means high gain, which means overloading and distortion on strong signals. So set the gain for the performance you need.
I prefer to switch between different values of R9 (with VR1 shorted) for 10dB steps in gain, the last thing you need is a noisy volume control, though R14 helps there. If you do this, keep the highest value R (lowest gain) permanently connected, like R14, and switch other R's in to reduce it, or you can expect DC thumps.
Your BC847s will do for Q2,Q4. The 857s may be a little noisier than their recommendation; I'd try, and upgrade them later if you need.
If you can hear or measure the difference between R9=22 and R9=68 ohms there, with 68 ohms being noisier, (after allowing for the gain difference) then there's not much point improving the transistors Q1,Q3 over your BC857s.