Why do decibel apps show different readings?

Because each app guesses differently about uncalibrated hardware. Different microphones, hidden OS processing like auto gain, different calibration offsets, A-weighting versus none, and different averaging windows can stack up to 5–10 dB of disagreement between two apps measuring the same sound.

Last updated:

The five sources of disagreement

  1. Microphone hardware. Every phone model’s mic has its own sensitivity and frequency response. The same app on two phones can read several dB apart before software even starts.
  2. OS audio processing. Phones quietly apply auto gain control, noise suppression, and echo cancellation tuned for calls. An app that doesn’t explicitly disable these is measuring a signal the OS already altered. (Our browser meter disables all three.)
  3. Calibration offset. Converting a digital signal level (dBFS) to sound pressure level (dB SPL) requires an offset that depends on the exact microphone. Apps ship different per-model profiles — or one global guess.
  4. Weighting. One app shows raw dB, another shows dB(A). On bass-heavy noise those can differ by 10 dB or more by design, not by error.
  5. Averaging window. “Fast” (125 ms), “slow” (1 s), peak-hold, or a smoothed UI number — each responds differently to fluctuating noise, so simultaneous readouts won’t match.

How to get comparable readings

Use one app, on one device, with the same weighting and averaging every time — relative comparisons (“the bedroom is 12 dB quieter than the street”) survive calibration error even when absolute numbers drift. For absolute accuracy, calibrate against a reference: NIOSH’s research found apps with calibrated external microphones land within about ±1 dB of professional meters.

The honest baseline

No uncalibrated app — ours included — should promise instrument precision. That’s why the online meter labels every reading an estimate (±10 dB), and why apps that ship device corrections and let you fine-tune against a reference — like Decibel Shield — get closer to the truth than a generic browser reading can.

← All questions