This is really well-written but honestly the whole thing screams "why didn't I just use an SDR and Python." Building 12 cascaded filters in discrete components to hit specific cutoff frequencies is... a lot. Like, I get the appeal. You learn filter design, impedance matching, all that. But then he's spending paragraphs tuning Q7's source resistor and measuring bias points with a multimeter to get half the supply voltage. A RTL-SDR dongle and scipy.signal would handle all that in like 50 lines of code and you could adjust everything in real-time without breaking out the soldering iron. I suspect part of the charm is the analog zen thing (which fair) but the practical argument for this approach in 2025 is pretty thin. The recording quality argument also doesn't hold since laptops now have decent audio ADCs. Good on him for the deep dive though, that level of circuit kung fu isn't common anymore.

My Uni project was using ELF waves to monitor solar terrestrial interactions at the magnetopause.

ELF also used to communicate with submarines too.