How accurate is this meter?

Most online meters answer this question with a single figure, and the figure is almost always "±2 dB". That number is defensible in some circumstances and misleading in others, so this page gives it a scope instead of a headline.

The distinction that matters is between absolute accuracy — how close a reading is to what a calibrated meter would show — and repeatability, how consistent readings are with each other on the same device. They are not the same, they differ by a lot, and repeatability is what most real uses depend on.

The meter converts a full-scale-relative level into a sound pressure level by assuming that a full-scale signal equals 94 dB SPL, the standard 1-pascal calibrator reference. That assumption is fixed for every visitor, because there is no per-device calibration control. The full derivation is on the methodology page.

Whether that assumption holds for your hardware depends on your microphone's sensitivity and on how your operating system stages gain before the browser sees the signal. MEMS microphones are typically specified with a part-to-part tolerance of around ±3 dB before any of that, and the gain staging differences between manufacturers are larger still. This is why the same room can read differently on a phone and a laptop sitting beside each other, and neither is malfunctioning.

One thing works in our favour. Browsers enable automatic gain control, noise suppression, and echo cancellation by default, and all three actively change signal level — the exact quantity being measured. This meter disables all three. Evaluations of smartphone measurement apps attribute much of their disagreement with reference instruments to leaving that processing on, so switching it off removes a large and well-documented error source. It does not remove the sensitivity variation above.

Because device error is largely a fixed offset, it mostly cancels when you compare readings from the same device. If your phone reads 4 dB high, it reads 4 dB high in the machine room and 4 dB high in the office, and the 12 dB difference between them is real.

That makes the meter genuinely reliable for the questions people usually bring to it: which room is worse, whether a noise is getting louder over weeks, whether hearing protection changed anything, whether the neighbour's Saturday is worse than their Tuesday. Screening against a threshold — is this over 85 dBA? — is the weakest use, because that one depends on absolute accuracy.

If you have access to a calibrated meter even once, measure a steady source with both at the same position and note the difference. That single number is your device's offset, and applying it by hand corrects most of the absolute error for that device in that browser. It will not correct frequency response, so an offset derived from a steady tone transfers imperfectly to a very different noise.

The practical technique — device position, microphone port, hard surfaces, measurement duration — is on the homepage, and matters more than most people expect. Holding the phone wrong costs more accuracy than the calibration assumption does.

Use a calibrated Type 1 or Type 2 sound level meter, or engage an occupational hygienist, for any of the following: OSHA compliance documentation, enforcement or regulatory submissions, litigation and insurance claims, formal noise ordinance complaints where the authority specifies an instrument class, and hearing conservation program monitoring. This meter is not certified to IEC 61672 and has no calibration traceability, which is usually the first question asked of a measurement in any of those settings.

A browser reading is the evidence that tells you whether paying for a real measurement is justified. That is a genuinely useful thing to know, and it is what this tool is for.

We have not published our own laboratory comparison against a reference instrument. The ranges above are bounded by published research on smartphone sound measurement and by the known properties of the signal chain, not by our own testing of this meter across a device matrix. We would rather say that plainly than present a protocol as though it were a result.

If you have a calibrated meter and want to help, the comparison that would be most useful is: a steady broadband source at several levels between 60 and 95 dBA, measured simultaneously by the reference and by this meter at the same position, repeated across device and browser combinations, with the model, OS version, and browser recorded for each. Send results to [email protected]. Anything published from that will name the devices and show the spread, including the results that make us look bad.

• Kardous, C. A. & Shaw, P. B. (2014). Evaluation of smartphone sound measurement applications. J. Acoust. Soc. Am. 135(4), EL186–EL192.
• Kardous, C. A. & Shaw, P. B. (2016). Evaluation of smartphone sound measurement applications using external microphones — a follow-up study. J. Acoust. Soc. Am. 140(4), EL327–EL333.
• NIOSH Science Blog. So how accurate are these smartphone sound measurement apps?