Field Guide · v1.0
A real-time audio spectrum analyzer with a live RTA, scrolling waterfall, and 1/3-octave bars. Everything is measured on-device from the microphone — nothing is recorded, nothing leaves your phone.
Open the app and allow microphone access when asked. That's it — you're live. The trace responds instantly to any sound in the room: speak, clap, or play a tone and watch it land at its frequency.
The top bar is your instrument readout; the plot below it is the spectrum itself.
dBFS(A).The full-height RTA. Best for tone-by-tone work: tuning, finding hums, watching harmonics.
Spectrum on top, waterfall below — the instrument view. See the now and the last minute at once.
Full-screen spectrogram. Time scrolls downward, newest at the top; color is level, from deep blue-black (quiet) through violet and ember to pale amber (loud). One waterfall row is written per analysis frame, so the time axis is honest.
31 bars, one per standard 1/3-octave band from 20 Hz to 20 kHz. Band edges follow the exact base-10 IEC 61260 midbands. Bars running hot (above −20 dBFS) tint amber at the top.
FFT size trades frequency detail against reaction speed. All settings persist between launches.
| FFT size | Frequency resolution* | Feel |
|---|---|---|
| 2,048 | 23.4 Hz | Fastest response, coarse low end |
| 4,096 | 11.7 Hz | Quick, general-purpose |
| 8,192 | 5.9 Hz | Default — resolves bass notes cleanly |
| 16,384 | 2.9 Hz | Finest detail, slowest to react |
*at a 48 kHz input. You can also set the display's dB floor (−60 to −140) and frequency range.
Why a tone's octave bar can differ from its spectrum peak: the octave bars sum all the energy inside each band, while the spectrum shows energy per frequency. For a pure tone they agree; for noise or dense material the bands read higher — both are correct, they answer different questions.
Why the noise floor moves with FFT size: bigger FFTs split the same noise across more bins, so each bin reads lower. Tones stay put — that's the giveaway for telling a real tone from noise.
The gear menu holds the three tools that complete the workflow: make a test signal, compare against a goal, and document the result.
Signal generator. Pink noise, white noise, a sine, a 20 Hz–20 kHz log sweep, and a warble tone, with an adjustable level in dBFS. For measurements, feed the system under test by cable or USB — the phone's own speaker is not a flat source. The mic keeps analyzing while the generator plays, so one phone can drive and measure at once.
Target curves. Settings → Target curve overlays a dashed reference shape — Flat, a −3 dB/oct tilt, or a bass-lifted Car curve — on the spectrum. It auto-fits its vertical position to the live measurement, so it is always a relative shape, never a calibrated-level claim. Tune until the bars hug the dashes.
Snapshots. The camera button freezes the current spectrum with all its settings. Gear menu → Snapshots lets you recall any saved measurement as a ghost overlay (the before/after line), or share it as a publication-ready image or a CSV. Every export carries a metadata header — FFT size, window, weighting, ENBW correction — so the numbers stay interpretable long after you forget the setup.
| Symptom | What it means |
|---|---|
| MIC OFF | Microphone permission is off — enable it in system settings (paths in §01) and return to the app. |
| ERROR badge | The audio input was lost (a call took over, or a device switched). Tap ▶ to restart — the app also recovers automatically from calls and headset changes on iOS. |
| Everything reads low | Check the weighting: A-weighting pulls bass readings down by design (−19 dB at 100 Hz). Switch to Z for raw levels. |
| Waterfall stopped | You're paused (HOLD) — that's by design; the whole picture freezes together. |
| Trace feels sluggish | Large FFT sizes react slower — drop to 4,096 or 2,048 in Settings (iOS) for fast transient work. |
Audio is analyzed on-device in real time and immediately discarded. The app never records, stores, or transmits audio — it has no network features at all. When paused or backgrounded, the microphone is released entirely.