Decibel Meter

A decibel meter estimates ambient sound level through the microphone and displays the result in dB. In a browser, this should be treated as a practical reference rather than a calibrated scientific measurement, because built-in microphones, automatic gain control, operating-system filters, browser behavior, and device placement all affect accuracy. The reading can still compare quiet and noisy situations, check fan or street noise, test an audio setup, or show whether a change reduced sound level. For workplace safety, legal evidence, construction acceptance, or professional acoustics, calibrated sound level meters and defined measurement procedures are required. Audio processing stays local in the browser.

Sound level estimation in the browser starts with the MediaDevices.getUserMedia({audio: true}) API, which exposes the selected input device as a MediaStream. The stream is routed into an AudioContext, then through a MediaStreamAudioSourceNode and an AnalyserNode. Calling AnalyserNode.getFloatTimeDomainData(buffer) every animation frame fills a Float32Array of normalized samples in the range [-1, 1]; from these samples the page computes the root-mean-square amplitude rms = sqrt(sum(x_i^2) / N), and converts it to a relative level in decibels full scale as dBFS = 20 * log10(rms). Because dBFS is a digital reference (0 dBFS = full-scale square wave), it is not an absolute physical pressure level. Converting dBFS to dB SPL (the physical pressure scale used by Class 1 and Class 2 IEC 61672 sound level meters) requires a calibration offset measured with a reference source such as a 94 dB SPL pistonphone, which uncalibrated laptop and phone microphones do not provide. A getFloatFrequencyData() spectrum can be derived from the same AnalyserNode for FFT-based octave-band analysis, but A-weighting and slow/fast time integration (125 ms / 1 s) defined in IEC 61672 are not part of the Web Audio default. Readings drift with automatic gain control, OS-level noise suppression in echoCancellation/noiseSuppression constraints, microphone directivity (omnidirectional MEMS vs cardioid USB), and distance attenuation that follows the inverse square law (-6 dB per doubling of distance in free field). Public health thresholds the gauge labels against include the WHO recommendation of 85 dBA for 8 h and the OSHA 1910.95 permissible exposure of 90 dBA for 8 h with a 5 dB exchange rate.

• MediaDevices.getUserMedia({audio: true}) requires a secure context (HTTPS or localhost) and a per-origin user gesture; denial cannot be reversed without a site-permission reset.
• AnalyserNode.fftSize must be a power of two between 32 and 32768; getFloatTimeDomainData() runs in O(fftSize) per frame at the audio context sample rate (commonly 44100 or 48000 Hz).
• RMS-to-dBFS uses 20*log10(rms); guard against log10(0) by clamping rms to a small epsilon (e.g. 1e-10) to avoid -Infinity at silence.
• Browser AGC and noiseSuppression constraints are enabled by default in Chrome and Edge; pass {echoCancellation: false, noiseSuppression: false, autoGainControl: false} to keep the level meter from chasing its own gain.
• dBFS is not dB SPL: a published 94 dB SPL pistonphone or sound calibrator is required to convert browser readings into physical pressure for OSHA, ISO 9612, or NIOSH workplace use.
• WHO recommends 85 dBA / 8 h exposure with a 3 dB exchange rate; OSHA 1910.95 uses 90 dBA / 8 h with a 5 dB exchange rate, so the same dBA reading maps to different allowable durations under each standard.
• Inverse square law gives roughly -6 dB per doubling of distance in free field, so microphone placement repeatability of a few centimeters is usually the dominant measurement error.