Mic Capsule Preamp DIY

The major takeaway from these designs is that electrets (pre-biased condensers) are always inferior to externally biased ones due to the self-noise. This is an important factor when considering which capsule we choose. Making a 48-80V inverter is next on my list of things to do and since I have both electret and unbiased capsules of similar design it's going to be interesting to see how they perform. 🙂

Nice to see the OPA Alice getting more coverage - I see that JLI offer a pre-made one for people who aren't too handy with a soldering iron. Through hole is tough enough, but when you get into SMD (and double-sided SMD in particular) the difficultly level starts to rise exponentially.

The 1642 has been superseded by the 1656 which features lower input noise (4.3 vs 5.1nV per root Hz at 1K) but this is still quite a bit louder than a pure FET solution (0.9nV or better for some designs). The 1642 is comparable to the bipolar 5532, but is more power efficient and preferred.

What I have learned from studying this design though, is that the noise floor cannot deliver performance better than 10 nV since it's chaining two amps in series and since that's random it won't cancel out. 

I'll admit that I figured, what's a few nV among friends right? Well as Matt has shown with his premium USB 3 mic, a mic amplifier might have to boost the signal by 40-60 dB - which is 100 to 1000x so in order to reach the required levels for subsequent stages. So 10 nV is added to the self-noise from the capsule (worse in an electret) and all of that is amplified by 1000x at which point it may become signifiant depending on the program material. Rock music isn't going to be an issue and most speech is OK but more subtle use cases (particularly acoustic guitars) will likely have noticeable artefacts. Albeit not necessarily intrusive. I've found the NE5532, as power hungry as it is, to be an excellent amp despite some authors getting a bit sniffy over it.

Worth repeating though, that a lot of early solid state mixing desks from the 1970s and onward used the classic uA741 - so your Beatles, Pink Floyd, Zeppelin (etc.) that you play through a 1972 Linn Sondeck [ https://www.linn.co.uk/uk/turntables ] and £10,000 amp and speakers was probably mixed on a humble 741. 😉

Out of curiosity (you know, because ya have to), I nipped over to Omicalculator to get the equivalent noise figures (Johnson-Nyquist) noise for a typical resistor. https://www.omnicalculator.com/physics/resistor-noise

This is going to spoil your day (made me sit up a bit too).

The 1G resistor on the input stage, required so as not to load the capsule at low frequencies - generates (I hope you're sitting down) generates a typical total noise figure of 569 uV (over 1/2 a mV) of noise across the audio band. As bad as that looks we also have to contend with the fact that manufactures quote a noise figure for Op Amps at a specific frequency, typically 1Khz, so the total noise for any given circuit is going to be higher.

I think I need a little lie down now.