How our browser audio tools work

1. Permission and capture

Microphone tools call the browser’s MediaDevices.getUserMedia() after a user action. The browser controls permission and supplies an audio stream in a secure context. The meter requests echo cancellation, noise suppression, and automatic gain control (AGC) off when those constraints are supported. The spectrum tool also requests those processing features off. Browsers and hardware may still process the input. The microphone test requests voice-oriented processing on by default, so its digital levels can differ from the meter’s.

The live stream is connected to a Web Audio AnalyserNode. Microphone input is not routed to the speakers for measurement. Stopping the tool releases its stream. Analysis and visualizations run locally; see the Privacy Policy for related storage and service details.

2. Signal strength, RMS, and dBFS

The meter reads 2,048 time-domain samples from the analyser and calculates root mean square (RMS). Zero and very small input use a finite digital display floor of −100 dBFS; the displayed estimate is also capped at 0–130 dB. These floors are not measured acoustic silence: RMS = √(Σ sample² / N). Squaring and averaging makes this a useful summary of signal strength across the window instead of using one momentary peak. It then calculates dBFS = 20 × log10(RMS). This is a digital level relative to full scale; it is not sound pressure in pascals and cannot by itself tell us acoustic SPL.

The meter requests a new reading about every 100 ms. Each reading uses the analyser’s current sample window, so it is a sequence of short snapshots rather than an integrating sound-level-meter time weighting.

3. Displayed dB and alignment

The meter adds a default 80 dB reference offset to the computed dBFS value, then confines the displayed estimate to 0–130 dB. The 80 dB value is a display reference, not an independently validated calibration for your microphone. For a steady, unclipped source, “Align Reading” changes the offset so the arithmetic session average matches an attainable entered reference; the adjustment is saved in this browser on this device and existing session values are recalculated.

One constant offset can help comparisons near the sound level and setup used for alignment. It cannot correct changing AGC, clipping, microphone frequency response, frequency weighting, or different acoustic conditions. The meter does not implement A or C frequency weighting and does not claim dBA or dBC.

4. FFT and frequency views

Web Audio’s AnalyserNode provides frequency-domain data using an FFT. The meter uses an FFT size of 2,048 and analyser smoothing of 0.8 to draw its frequency view. The dedicated Spectrum Analyzer offers FFT sizes from 2,048 to 16,384 and adjustable smoothing. Larger FFT sizes provide narrower frequency bins but use a longer slice of audio; smoothing steadies the display while making fast changes less immediate. These plots describe the captured digital signal, not calibrated sound pressure at each frequency.

5. History, summary, and quality

The meter keeps the most recent 30 seconds of timestamped readings for its history graph. Its minimum and maximum are the lowest and highest displayed estimates seen in the current session. Its average is the arithmetic mean of those displayed sample estimates across the session, not a time-integrated equivalent sound level or a professional meter’s Leq. The summary continues across the session even as older points leave the graph.

The meter labels a signal “Low” at or below −60 dBFS and “Clipping” when a sample peak reaches at least 0.98 of full-scale amplitude. A “Stable” label comes from low variation across recent digital readings. These are interface quality cues, not accuracy grades. Frequency smoothing is applied by the analyser; the meter’s RMS readings themselves are not given an A/C filter or professional fast/slow time weighting.

6. Playback tools

Speaker Test, Headphone Test, Bass/Subwoofer Test, and Tone Generator create or play test signals in the browser. Their frequency and level controls describe the digital source. Actual sound at your ears or in the room also depends on system volume, audio hardware, speakers or headphones, placement, and the room. A playback test is not a calibrated measurement of acoustic output or hearing ability.