What Validation Means
Validation is the act of comparing this tool's readings against a trusted reference instrument (ideally a calibrated Class 1 or Class 2 sound level meter, e.g. IEC 61672) under controlled, documented conditions. It is how you move from "the meter shows a number" to "the meter is within X dB of a reference under these conditions."
We publish the method anyone can follow, and a template for recording results. We do not publish fabricated numbers. When real measurements are contributed under this protocol, they will appear in the table below with full conditions.
Repeatable Test Protocol
The goal is a fair, side-by-side comparison — not a single lucky peak match. Follow these steps:
- Choose a stable source. A pink-noise speaker at fixed volume, a running appliance, or steady traffic heard from a single spot. Avoid varying speech or music with large dynamics.
- Place devices together. Reference mic and test device (phone/laptop) side-by-side, same orientation, same distance from the source — ideally 0.5–1 m, on a stable surface, mic ports unobstructed.
- Match settings. Set both to the same frequency weighting (use A) and comparable time weighting (Fast on both if available). Note weighting, mode, and calibration offset.
- Measure 30–60 seconds. Record the average / Leq, not momentary peaks or instants. Leq is the correct aggregate for exposure-type comparisons.
- Repeat 3–5 times. Lift and re-place the test device between runs to capture positioning variance. Keep the source level constant.
- Calculate the offset. For each run:
offset = referenceAvg − meterAvg. Average the offsets across runs. - Apply & re-validate. Enter
offset = reference − readingin the Calibrate dialog, then repeat steps 4–5. Agreement should be within ~1–2 dB when the setup is stable. - Document everything. Device, browser, mic, distance, duration, weighting, calibration, ambient conditions, and raw numbers.
What to Record for Each Run
- Device & OS, browser & version, microphone identity (from Device & Microphone Information).
AudioContext.sampleRate, channel count, and whether echo cancellation / noise suppression / AGC were actually off.- Source, distance, room (or outdoor), orientation, weighting (A/C/Z), time weighting (F/S/I), mode (RMS/Peak).
- Duration, reference average & Leq, meter average & Leq, instantaneous min/max/peak, calibration offset applied.
- Temperature/humidity if available (optional, but helpful).
Results Template (copy into your notes or export)
| Run | Reference Avg (dBA) | Meter Avg (dBA) | Δ (Ref − Meter) | Leq (Ref / Meter) | Conditions |
|---|---|---|---|---|---|
| 1 | — | — | — | — | e.g. Pixel 7 / Chrome 126, 1 m, A/Fast, offset +30 dB, 60 s, pink noise |
| 2 | — | — | — | — | — |
| 3 | — | — | — | — | — |
| Avg | — | — | — | — | mean offset → apply via Calibrate |
CSV/JSON exports from this tool include the raw history you need to compute averages externally. Save the export alongside the reference's log so the comparison is auditable.
Published Results
No first-party lab dataset yet
This site has not yet published a controlled lab validation dataset. When one exists, rows will be added to a table here with: date, device, browser, reference instrument & calibration cert, source, distance, duration, weighting, offset, and per-run and mean differences — all reproducible under the protocol above.
This is intentional transparency. We will not invent numbers to fill a table. The methodology and template above are the standing offer: anyone following the protocol produces results that are comparable.
Have a careful side-by-side you would like to contribute? Contact us with the full conditions and raw exports. Submissions are reviewed and, if reproducible, credited (or anonymous if preferred).
What Validation Can and Cannot Tell You
- Can: quantify the systematic offset of your device under those conditions, and how stable that offset is.
- Can: reveal whether AGC / DSP remained active despite being requested off.
- Cannot: prove universal accuracy across all devices — microphones and DSP vary widely.
- Cannot: turn a phone into a certified instrument — frequency response and dynamic-range limits remain even after offset correction. See Methodology and Accuracy.