The short answer: useful estimate, not certified instrument

A decibel meter app turns your phone microphone into a sound level estimate. With the right app, a calibrated offset, and a 30–60 second average, it answers everyday questions well: is this nursery 45 or 60 dB? Is the classroom averaging above 60 dBA? Did the new seal cut traffic noise by 5 dB? For those screening and comparison jobs, a phone is genuinely useful.

What it is not: a certified sound level meter meeting IEC 61672. No phone app passes the full standard, sensitivity varies between models, bass is under-read, very quiet and very loud sounds are unreliable, and hidden gain control can quietly shift readings. So treat phone numbers as screening data with an error band — typically ±2 dB calibrated best case, ±5–10 dB uncalibrated — and never as evidence for courts, regulators, or workplace compliance. Our accuracy guide states this plainly, and this article shows the numbers behind it.

What lab studies actually measured

Forget star ratings — these peer-reviewed tests compared apps against Type 1 laboratory meters on the same sounds:

  • Kardous & Shaw 2014 (JASA, 192 apps screened): only 4 of 10 qualifying iOS apps stayed within ±2 dBA of the reference on pink noise at 65–95 dB. Most Android apps tested at the time failed — not because Android is hopeless, but because hundreds of different capsules made per-model calibration impossible.
  • Kardous & Shaw 2016 (with external mics): the same NIOSH SLM app paired with a calibrated external microphone (MicW i436 / Dayton iMM-6 class) measured within ±1 dB — proof the software math is sound and the capsule is the limit.
  • Crossley et al. 2021 (Laryngoscope, 9 iPhone apps): the free NIOSH Sound Level Meter app was the best, with R² 0.97 against the reference — essentially tracking the true level across the tested range.
  • Celestina et al. 2018 (Applied Acoustics): tested whether phones could partially meet IEC 61672 Type 2 tolerances. Some passed some tests, none passed all — the formal way of saying “close, but not certified.”
  • JAMA Otolaryngology 2023 (Apple Watch vs Brüel & Kjær Type 2250): correlation 0.961+ from 125 Hz to 8 kHz, statistically equivalent from 125–3000 Hz and at 6 kHz, small under-5 dB gaps at 4 and 8 kHz. A well-designed consumer indicator, not an occupational meter.
  • JMIR Formative Research 2023 (iPhone 13 Pro): NIOSH SLM within 0.5 dBA of a conventional meter from 40–85 dBA — the best of the tested apps in everyday range.

Pattern across all of them: 40–90 dBA steady sounds are the sweet spot. Below ~30 dB the phone noise floor reads too high; above ~95–100 dB capsules clip and read too low. Quote mid-range Leq values with confidence, treat extremes as directional. See how we measure and how we validate for our own signal chain and limits.

iPhone vs Android vs browser meter: what differs

iPhones are the most consistent phones because Apple ships few models with similar MEMS capsules — developers can pre-calibrate per model, and two iPhones on the same table often agree within 3–5 dB. Android spans hundreds of capsules, gains, and cases — 5–10 dB disagreement on the same sound is normal uncalibrated, though a single calibrated Android is just as useful for before/after on that same device.

Browser meters like the free online meter use the same capsule through the Web Audio API, entirely on-device with nothing uploaded. They request echoCancellation: false, noiseSuppression: false and autoGainControl: false, but some phones and browsers ignore the request — check the Device & Microphone panel to see what was actually applied. Advantage of browser: no install, same technique and calibration workflow as apps, plus Leq, spectrum and CSV/JSON export. Advantage of the best native app (NIOSH SLM on iOS): deeper metrics, dose calculations and history. For a one-off room check, use the browser; for repeated surveys, use NIOSH SLM on iPhone; for all-day awareness, use Apple Watch Noise — details in our iPhone guide.

The 5 factors that decide your error band

1. Microphone capsule. Phone mics favour voice frequencies, roll off bass, and vary in sensitivity — the reason calibration removes bias but cannot fix frequency response. Expect bass-heavy traffic rumble and neighbour subwoofers to under-read; cross-check with dBC and the frequency analyzer.

2. Hidden processing. Automatic gain control turns sensitivity up in quiet rooms and down in loud ones; noise suppression erases the background you want to measure; echo cancellation removes replayed audio. All three are built for calls, not measurement.

3. Placement and distance. This dominates real-world error. A case over the pinhole costs up to 10 dB; grip noise adds 2–5 dB; halving distance adds ~6 dB outdoors; pointing away or pocketing the phone ruins comparability. Fix placement before blaming the app — mic uncovered, facing the source, at ear height, distance noted every time.

4. Room and wind. Hard walls raise levels through reflections; outdoors, wind on an unshielded mic adds 10+ dB of artefacts. Measure rooms as people use them and shield outdoors with a hand nearby or a foam windscreen.

5. Level and duration. Quoting a flickering instant is the classic amateur mistake. Always quote average or Leq over 30–60 seconds, plus maximum for events and minimum for background. Full routine in how to use an online meter and how to measure decibels.

When is phone accuracy good enough? A practical table

Match the tool to the stakes:

  • Good for phones (calibrated, Leq 30–60 s): nursery checks (white-noise machine under 50 dB at the crib), classroom averages above 55–60 dBA hurting focus, rehearsal and commute checks, appliance comparison at fixed 1 m, before/after home fixes logged in the Noise Journal.
  • Useful as a briefing log, not proof: neighbour noise diary — “living room 63 dBA Leq 15 min, peaks 78 dBA, dBC 81, window closed, 22:10–22:25” tells a surveyor exactly where to deploy certified kit and saves survey cost.
  • Never a phone job: workplace exposure under OSHA or the EU directive, planning and nuisance disputes, product specs, commissioning, anything below ~30 dB or above ~100 dB, or any number that will be challenged. That needs an IEC 61672 meter, lab calibration, and a competent surveyor reporting Leq plus L90/L50/L10 with uncertainty.

Feed any measured Leq plus duration into the noise exposure calculator, and judge it against safe exposure limits and the full decibel chart.

Get to ±2 dB in 5 minutes: calibration + 60-second routine

Step 1 — Set up. Remove bulky cases or align cutouts, uncover the bottom-edge mic pinhole, hold at ear height where people are, silence yourself (your voice adds 10–20 dB), keep the tab foreground and screen awake. Note device, place, distance, weighting and duration.

Step 2 — Calibrate once. Play a known 94 dB reference tone or compare against a trusted second meter at the same position for 30–60 seconds, enter reference minus reading as the stored offset in the Calibrate dialog, and retest — the same tone should now read correctly. Full method in the calibration guide.

Step 3 — Measure Leq, not instants. Record average or Leq over at least 30 seconds (60 for rooms), note minimum for background and maximum/peak for events, repeat from a second spot. Agreement within a few dB means you can trust the picture; wild swings mean something pulses nearby. Log repeats with exports so outliers stay visible.

Sanity checks: a quiet bedroom at night should sit roughly 25–40 dB, conversation ~60 dB at 1 m, city traffic ~70–85 dB at the kerb — see 10 everyday sounds. If a library reads 70 dB, suspect a blocked mic, suppression, or a nearby fan — not a loud library.

Red flags: when not to trust the number

Distrust and retest when two phones on the same table differ by more than ~5 dB after calibration, when bass dominates but dBC barely exceeds dBA, when the reading never moves (gain control frozen), when wind or handling noise is audible, or when the number sits under 30 dB or over 100 dB. All are capsule or processing limits, not real silence or real danger — change position, remove the case, disable processing where possible, recalibrate, and average longer.

Bottom line: apps do not fail because the math is wrong — they fail when the capsule, the processing, or the technique is wrong. Fix those three and a free browser meter or the NIOSH app becomes a genuinely honest screening instrument. For everything else — curiosity, comfort, sleep, study, rehearsal, travel — measure a minute, decide with data, and read what a decibel meter does next.

Reference tool only — not for legal, medical, occupational, or regulatory decisions.

Frequently asked questions

Are decibel meter apps accurate?
The best apps are within about ±2 dB of professional gear on steady 40–90 dBA sounds when calibrated, and typical uncalibrated phones disagree by 5–10 dB. They are good screening tools for everyday decisions, not certified instruments for legal or workplace proof.

Why do two phones show different dB on the same sound?
Different microphone sensitivity and frequency response, plus per-device automatic gain, noise suppression and echo cancellation. Cases covering the pinhole alone can cost up to 10 dB. Calibrating each phone against the same reference removes most of the gap.

Can I use a decibel app for legal or workplace proof?
No. Courts, regulators and occupational schemes require IEC 61672 certified meters with traceable calibration and defined positions, durations and reporting. Use phone logs only as a dated briefing for the professional surveyor.

How do I calibrate a decibel meter app?
Measure a known 94 dB reference tone or a trusted second meter at the same position for 30–60 seconds, compute reference minus your reading, and store it as the offset. Retest the same tone to confirm, and recalibrate after phone or OS changes.

Which decibel meter app is most accurate?
In peer-reviewed tests the free NIOSH Sound Level Meter app on iOS is consistently the most accurate phone option, within about 2 dB of a Type 1 reference and within 1 dB with a calibrated external microphone. Avoid apps that cannot calibrate or export.

Sources and further reading

  • Kardous & Shaw 2014, JASA — Evaluation of smartphone sound measurement applications (192 apps screened, 4 iOS apps within ±2 dB on pink noise 65–95 dB).
  • Kardous & Shaw 2016, NIOSH follow-up — NIOSH SLM within ±1 dB with calibrated external microphone.
  • Celestina et al. 2018, Applied Acoustics — partial IEC 61672 Type 2 compliance testing of smartphones.
  • Crossley et al. 2021, Laryngoscope — accuracy of 9 iPhone apps (NIOSH SLM best, R² 0.97).
  • JAMA Otolaryngology 2023 — Apple Watch Noise app vs Brüel & Kjær Type 2250 Class 1 (r ≥ 0.961).
  • JMIR Formative Research 2023 — NIOSH SLM vs Decibel X vs SoundMeter X on iPhone 13 Pro (NIOSH within 0.5 dBA, 40–85 dBA).
  • WHO — Make Listening Safe / Deafness and hearing loss Q&A (80 dB 40 h/week adult weekly allowance; 3 dB exchange).