Short answer: Every tool is tested before release in a local WordPress install, with a scripted Microsoft Edge browser and a fake microphone that plays generated test signals of known level. Across the 15 current tools, 393 checks passed and 0 failed. Another 78 checks are marked “not tested”, mostly real phones, Safari and Firefox, and they are listed on this page instead of being hidden.
Last updated: September 27, 2026.
What a Pass Means, and What It Doesn’t
A check passes when a tool gives the expected result for a known input, within a tolerance written down before the test. For example, a 1 kHz tone at −23 dBFS must read 77.0 ± 1.0 dB(A) on the sound decibel meter. Passing proves that the maths, filters, controls and error messages behave as designed. It does not prove that your phone’s microphone is accurate. No browser tool is a Class 1 or Class 2 sound level meter, and these tests don’t claim otherwise; the limits are set out in online decibel meter accuracy.
The Test Setup
- Site: WordPress (latest) on PHP 7.4 in WordPress Playground, a local test install, with and without Yoast SEO active.
- Browser: Microsoft Edge (Chromium), driven by a script, at desktop size and in 390 × 844 phone emulation.
- Microphone: a fake microphone playing generated WAV files: 1 kHz and 100 Hz sine tones at −23 dBFS RMS, silence, a full-scale square wave, very quiet noise and a 60 + 120 Hz hum.
- Maths: calculation engines run separately in Node.js against hand-worked values and published reference signals.
- On every tool: keyboard access, the blocked-permission message, no sideways scrolling on a phone, zero console errors, and structured data with 0 errors and 0 warnings in the Schema.org validator.
Reference Signals and Standards
Where a standard defines the right answer, the test uses it:
| Check | Reference | Result |
|---|---|---|
| A- and C-weighting filters, 31.5 Hz to 16 kHz, at 44.1, 48 and 96 kHz | Frequency-weighting table in IEC 61672-1 | Within ±0.3 dB up to 4 kHz (inside the Class 2 tolerance); +0.65 to +0.8 dB at 8 kHz; −3 to −4 dB at 16 kHz at 44.1 and 48 kHz |
| 100 Hz tone, dB(Z) minus dB(A) and dB(C) | IEC 61672-1 table: 19.1 dB (A) and 0.3 dB (C) | 19.0 dB and 0.4 dB |
| Integrated loudness, 4 test cases | EBU Tech 3341 (EBU R 128 metering) | −22.99, −32.99, −23.01 and −23.01 LUFS against −23, −33, −23 and −23 (±0.1) |
| Loudness range, 3 test cases | EBU Tech 3342 | 10.00, 5.00 and 20.00 LU against 10, 5 and 20 (±1) |
| K-weighting coefficients at 48 kHz | ITU-R BS.1770 published values | Largest difference 8.88 × 10⁻¹⁶ |
| True peak of a −6 dBFS, 997 Hz sine | ITU-R BS.1770 true-peak method | −6.00 dBTP |
| Noise dose: 90 dB(A) for 2 h plus 85 dB(A) for 4 h | NIOSH REL and OSHA 29 CFR 1910.95 | NIOSH 129.4%, OSHA 25%, action-level dose 50%, LEX,8h 86.1 dB(A) |
| Safe listening times in the chart | NIOSH 85 dB(A) with a 3 dB exchange rate | 85 dB = 8 h, 100 dB = 15 min |
| Adding and subtracting levels | Energy summation | 60 + 60 dB = 63.0 dB; 70 dB total with 65 dB background = 68.3 dB |
| Pressure conversion | 20 µPa reference for dB SPL | 94 dB SPL = 1.002 Pa; 20 µPa = 0 dB SPL |
Results for Every Tool
Each tool has its own written QA report. The table gives the headline of each one. “The common 4” are the gaps every report shares, explained in the next section.
| Tool and version | What was tested | Passed | Not tested |
|---|---|---|---|
| Sound Decibel Meter 2.0.0 | 1 kHz and 100 Hz tones in dB(A), dB(C) and dB(Z); filters against the IEC table; reference matching (+3.0 dB offset); CSV and report files; silence, clipping and blocked-permission messages | 33 | 6: the common 4, a real room with a real microphone, an hour-long session |
| Tone Generator 1.0.0 | Output frequency (440.00 Hz, 1,234.50 Hz), level matching between waveforms, sweeps, beeps, noise colours, left and right channels, WAV files parsed and measured | 38 | 5: the common 4, listening on real speakers |
| Background Noise Test 1.0.0 | Leq and L90 of a steady 77 dB(A) tone, 60 Hz mains-hum detection, quiet-room and too-short tests, shared calibration | 23 | 5: the common 4, real rooms |
| Frequency Analyzer 1.0.0 | Peak of a 1 kHz tone within ±2 Hz, a two-tone file (440 Hz and 3 kHz), FFT sizes, image export, non-audio file error | 21 | 5: the common 4, real instruments and long files |
| Decibel Chart 1.0.0 | 23-row chart, computed safe times, search and filters, level checker at 88 and 120 dB, readable without JavaScript | 19 | 4: the common 4 |
| Volume Level Comparator 1.0.0 | Ratios for 70 vs 80 dB (2× loudness, 10× energy, 3.16× pressure), noticeability wording, 5-second microphone measurement | 21 | 5: the common 4, real before-and-after measurements |
| Decibel Calculator 1.0.0 | Every tab against hand-worked values: adding, subtracting, energy average, distance, ratios, dBm, sones and phons | 30 | 4: the common 4 |
| Microphone Test 1.0.0 | Verdicts for good, very quiet, silent and clipping input; 5-second recording and download; processing on and off; blocked permission | 21 | 6: the common 4, real headsets and USB or Bluetooth mics, Safari’s recording format |
| Classroom Noise Monitor 1.0.0 | Green, amber and red states, presets, too-loud counter, chime timing, limits saved on the device | 22 | 6: the common 4, a classroom projector, a real class |
| Noise Monitor 1.0.0 | Logging in a background tab, events over the limit, L10, L90 and Lmax, CSV and report files | 17 | 6: the common 4, a phone with the screen locked, a 24-hour run |
| Audio Loudness Meter 1.0.0 | EBU Tech 3341 and 3342 cases, K-weighting coefficients, true peak, mono files, streaming-target verdicts | 32 | 6: the common 4, 30-minute files, 5.1 surround files |
| Decibel Level Widget 1.0.0 (on Is 85 dB dangerous? and the 14 other level pages) | All 15 pages, noise-dose checker (85 dB for 8 h = 100%), listening steps exactly 10.00 dB apart, accessibility labels | 23 | 4: Safari, Android phone, Firefox, listening on real speakers |
| SPL Converter 1.0.0 | Pascals, intensity, peak, water reference, microphone sensitivity (12.6 mV/Pa = −37.99 dBV/Pa), number formats | 35 | 5: the common 4, very old browsers |
| Noise Exposure Calculator 1.0.0 | NIOSH, OSHA and EU dose for single and mixed exposures, hearing-protection derating, share links, input errors | 28 | 5: the common 4, a physical noise dosimeter |
| Online Hearing Test 1.0.0 | Threshold search with scripted listeners of known hearing (found exactly), left and right check, result notes, hearing-age test, downloads | 30 | 6: the common 4, real ears and headphones, Bluetooth delay |
| All 15 tools | 393 | 78 (0 failed) |
What We Could Not Test
Every report lists the same four gaps, because the test setup had no Apple device, no physical Android phone and only a Chromium browser:
- Safari on Mac and iPhone (a fallback for Safari’s older audio interface is written into each tool, but it wasn’t run).
- Chrome on a physical Android phone.
- Firefox on desktop.
- Saving each plugin’s settings screen in WordPress admin: admin saves hang in the test install on Windows. It is the same settings code in every tool.
The tool-specific gaps are in the table above. The biggest are real rooms and real microphones for the meters (the tests used synthetic tones through a fake microphone), an hour-long meter session and a 24-hour noise log (memory is capped by design, but the full runs weren’t done), a physical dosimeter to compare the exposure calculator with, and real ears with real headphones for the hearing test.
If you use a tool on one of these setups and something looks wrong, a report that names your device and browser helps most; see the contact page.
Bugs the Tests Caught Before Release
The tests exist to find problems. Some that were fixed before the tools went live:
- The first version of the sound meter kept every frame in memory, so long sessions slowed down and could crash. Version 2.0.0 uses running totals and was tested again from scratch.
- The background noise test could report 60 Hz mains hum as 50 Hz, because nearby frequency bins overlap. It now checks within ±2.5 Hz of each hum frequency.
- The tone generator’s “both ears” setting was 3 dB quieter than one ear, and its WAV files peaked at −9 dBFS instead of −6.
- The decibel chart rounded short safe times to whole minutes, so 110 dB showed 1 minute instead of 1.5.
- The loudness meter’s true-peak search skipped one of its four interpolation phases.
- The exposure calculator showed “no limit” for OSHA between 80 and 89 dB(A), which contradicted its own safe-time table.
Content and Numbers
The same approach applies to the guides. Safety figures such as NIOSH’s 85 dB(A) limit, OSHA’s 90 dB(A) limit with a 5 dB exchange rate and the everyday levels in the decibel chart come from the official sources on our references page, and the tools compute them with the same formulas the articles quote. How sources are chosen and how corrections are handled is described in the editorial policy.
When Tools Are Retested
- A new version of a tool gets a new QA report with the full test list, not only the changed feature.
- A reported bug is first reproduced with a known signal or value, then fixed, and the check that would have caught it joins the list.
- When a standard or official guideline behind a tool changes, the tool and the pages that quote it are checked again.
The date at the top of this page changes whenever the results table changes.
Check Your Own Device
The tests can’t cover your exact phone, so the best check is your own: compare the meter with a reference meter or a known sound and set the offset, as described in calibrating your microphone for a decibel meter. Each step from microphone to reading is explained in how the sound decibel meter works, and the limits of any browser reading are set out in the disclaimer. Every tool is listed on the sound tools page.