simpleRTA

AI-assisted project. This codebase was created with Claude Code (Anthropic), directed and reviewed by a human author. The analysis chain is verified numerically — 29 tests pin the power normalisation at every transform size and through every window, the band-integration invariants, and the agreement between the band levels and the spectrum they come from — and the running app is checked against its own pink-noise source, which a constant-percentage-bandwidth display must draw flat. It has not been verified against a calibrated measurement microphone or a reference analyser, and no absolute SPL claim is made: the dB offset field is an unverified user-supplied number, and there is no microphone correction.

A real-time audio analyser that runs entirely in a browser tab.

simple-rta.stoatworks-labs.com

Split view on the built-in pink noise. Pink noise is flat on a constant-percentage-bandwidth display, so a correct analyser draws a level line — which is what makes this the picture worth checking a change against.

• RTA at 1/3, 1/6, 1/12, 1/24 or 1/48 octave
• FFT size 2048 to 65536, with a selectable analysis window and overlap
• Peak hold per band, and exponential or infinite averaging
• Spectrograph sharing the RTA's frequency axis
• Full-height bargraph meter — peak, RMS, peak hold and a latching clip indicator, plus broadband Z / A / C readouts

No backend, no accounts, no telemetry. The audio is analysed on the page and never leaves the browser.

Sources

An audio input — a microphone, a measurement mic on an interface, or a return from a console. The browser's echo cancellation, noise suppression and automatic gain are all switched off, so what you see is the signal rather than the voice pipeline the browser would otherwise hand you.

Tab audio — captures what another browser tab is playing. Pick a tab in the picker and tick "Share tab audio". Whole-window and full-screen shares carry no audio on most platforms. Chrome and Edge only.

Pink noise — generated in the page and analysed without going anywhere near the speakers. Pink noise is flat on a constant-percentage-bandwidth display, so this is the check that the analyser reads level before you trust what it says about a room.

Reading the display

Levels are dBFS, on the convention that a full-scale sine reads 0 dB. The offset field adds a constant to everything shown, so if you have a reference of known level you can dial the scale into SPL — nothing verifies that number, and no microphone correction is applied.

The shaded region

At the fine resolutions the low bands are narrower than the transform can actually resolve. Below the frequency stated under the controls, neighbouring bands report shares of a single measurement rather than independent ones, and the RTA shades that part of the graph. It is not a fault: it is the arithmetic of asking a 341 ms transform for 1/48-octave detail at 30 Hz. Lengthen the transform, or use a coarser resolution, and the shading recedes.

Which window

Hann for anything ordinary. Blackman-Harris to see a quiet tone next to a loud one — it buys −92 dB sidelobes at the cost of a wider main lobe. Flat-top when the amplitude of a tone matters more than its frequency; it is accurate to about 0.01 dB wherever the tone falls between bins, and has poor frequency resolution in exchange. Rectangular only for signals that are periodic within the frame.

Development

npm install
npm run dev
npm test
npm run build

The screenshots in this file are taken from the running application, never composited:

node scripts/shoot.mjs --url https://simple-rta.stoatworks-labs.com

Headless Chrome starts the pink-noise source with a real mouse gesture — a synthetic click() confers no user activation, and an AudioContext will not start without one — then lets the analyser settle before capturing.

See AGENTS.md for the measurement model and the traps, and CLAUDE.md for the command reference.

This project is built on other people's work — see ATTRIBUTIONS.md.

Licence

MIT. See LICENSE.