User-visible changes to the meter and the reference content, newest first — including the measurement bugs we got wrong and had to fix.
Last updated
A measurement tool asks you to trust a number. The most useful evidence we can offer is a record of what has changed and what has been wrong, so the entries below include our own errors — most notably a bug that capped every reading at 60 dBA and made the OSHA panel read "OK" in every room on earth. If you find something else, write to [email protected] and it will end up here.
The homepage used to be a landing page describing the meter, with the tool itself one click away at /app. The working meter is now on the homepage directly, and /app permanently redirects there. The homepage also gained a decibel reference table, NIOSH and OSHA exposure limits side by side, guidance on getting an accurate reading, and a full FAQ.
Added this changelog along with pages covering how a reading is computed, how accurate it actually is by device type, privacy, terms, and contact. The accuracy page replaces a single flattering number with ranges by situation, and states plainly that we have not yet run our own laboratory comparison.
The FAQ said DecibelPro had no premium tiers. That stopped being true when Pro shipped, and it had not been updated. It now describes accurately what is free and what Pro adds.
Cut the delay between tapping a control on the meter and seeing it respond, and did the same for the shared header and footer.
This was the most serious measurement bug we have found. The reading was capped at 60 dBA by construction: the level was derived from the browser's byte frequency data as though those bytes were linear amplitudes, when they are already a decibel scale mapped onto 0–255. Full-scale white noise — the loudest signal that can exist digitally — measured 56.7 dBA. The 85 dBA action level and the 90 dBA permissible exposure limit were therefore unreachable, and the OSHA panel read "OK" in every room, including a dangerous one. The level is now taken from the time-domain waveform, where RMS is exact, with A-weighting applied as a spectral correction. The same white noise now reads 86 dBA and correctly trips the action level. Clipping is detected and surfaced rather than reported as a confident number.
Expanded the noise ordinance guides to cover more US metros, each read from the municipality's own published code rather than from secondary summaries.
Added analytics coverage of the signup and Stripe checkout path, including failure cases, so broken purchases surface instead of being invisible.
The "how loud is…" articles now include a live reading that compares the room you are in against the level the article is about.
The pricing page still advertised sound-type identification after it was withdrawn, and offered subscription-only capabilities to one-off buyers who did not have them. Both were corrected, and the report generator was cut back to the single path that actually ships.
An experimental feature that tried to name the source of a sound was withdrawn. Measured accuracy did not justify presenting its output to users as a claim about what they were hearing.
Added a calculator for 8-hour time-weighted average and dose percentage under both the OSHA 5 dB and NIOSH 3 dB exchange rates.
The methodology page documents how readings are computed today, and the accuracy page sets out how far they can be trusted.