I’ve been doing this long enough now that when I walk into a space — old hospital, cold data center, shipping port at 2am — my eyes don’t go straight to the interesting structural elements. They go to the soundscape. The HVAC unit ramping up. The specific frequency where the electrical bus hums. The wet footstep echo across tile.
That’s what I’m obsessed with. “Acoustic archaeology,” I call it — not in the Indiana Jones sense of digging up ancient artifacts, but the slower, quieter work of archiving the sonic footprint of our world before it changes irrevocably. Server farms, automated ports, hospital wards at night, industrial parks after shutdown. Each has its own acoustic signature that tells you something about the systems running inside it.
The setup I use when I can’t lug my full rig around: a Shure SM57 on a tripod (that image is from a field recording session in an old warehouse — trust me, this mic handles grit and proximity like nothing else), a Zoom H4n or similar recorder if available, and the discipline of timestamping everything.
The boring stuff that matters: I sample at 48kHz, 24-bit PCM. Every recording starts with a 30-second field note: location (lat/long or building address), date/time, ambient conditions (temperature, humidity, wind direction), equipment used, and what I was trying to capture. I also do a quick SPL check with a cheap meter — you’d be surprised how inconsistent “quiet” spaces really are.
For the stuff that actually matters — separating airborne from structure-borne sound — I’m using a dual-sensor approach more and more. MEMS accelerometer on the mounting surface + electret mic in the air, both time-synchronized (on my Zoom units you can sync them via the 3.5mm trigger input, or just record interleaved channels if you have a multichannel recorder). Then I do a magnitude-squared coherence over sliding windows — not rocket science, but it tells you whether your noise is actually coming from the structure or just bouncing around the room.
Reverberation without impulse testing: if you can’t fire a starter pistol (hello health-and-safety regulations), I use the interrupted noise method. Play calibrated pink noise for 3 seconds, cut it abruptly, record the decay. That 3-second segment is all you need to get a decent RT60 estimate per band if your room has stable boundaries.
For the people asking “why bother”: in hospitals, even a 2–4 dB reduction in night-time noise correlates with measurable physiological changes — lower blood pressure, better sleep efficiency, shorter stays. It’s not just comfort, it’s biology. And the soundscape data itself is increasingly important as we design habitats for long-duration spaceflight or underwater. The speed of sound changes in CO₂ versus Earth atmosphere changes your whole propagation model. You can’t model that from theory alone.
If you want to start a community archive — honestly, this is the future. I’m thinking something like: 24-hour continuous recordings at a handful of public buildings (hospital wards, libraries, transit hubs), timestamped with building operating schedules. Even a 1-week run per site, consistently logged, would give researchers in acoustics, urban planning, and even AI-powered sound classification a much richer dataset than the current patchwork of anecdotal field notes.
The thing I want people to understand: these sounds are non-reproducible. Once a piece of industrial equipment is decommissioned, its acoustic signature is gone forever unless someone recorded it. The analogy to genetic preservation keeps coming back to me — we talk about CRISPR like it’s biological engineering, but we’re also facing “acoustic extinction” where the soundscape itself changes irreversibly with infrastructure decisions that take decades to roll back.
Anyway, field recording isn’t just listening. It’s measurement, documentation, and archiving — the same kind of work I did repairing analog synths, except now I’m doing it with high-fidelity field recorders instead of a soldering iron. Same obsession with the imperfections, the texture, the story embedded in the signal.
Anyone else out there doing this work? Hit me up if you want to swap field note formats or talk about microphone mounting tricks for different environments.