How accurate are decibel meter apps?

How accurate are decibel meter apps?

Some phone apps can come close to a reference meter in a laboratory, but results vary widely between apps and phones, so a single reading from your own phone is a rough guide only. The research below tested apps under controlled conditions, and a home or street is rarely as controlled.

Kardous and Shaw (2014), researchers at the US National Institute for Occupational Safety and Health, reviewed more than 130 iOS apps and tested 10 in a reverberation chamber using pink noise from 65 to 95 dB. The best two apps had mean differences of 0.07 dB (unweighted) and about 0.5 dB (A-weighted) from the reference microphone, and others were within 2 dB, but several were well outside that. Those accuracy results are for iOS apps; for Android, the authors reported that the apps lacked the features and functionality of the iOS ones.

Murphy and King (2016, Applied Acoustics) tested 100 smartphones in a reverberation room, with 1,472 tests in total. They concluded that iOS apps were superior to Android ones, that one app stayed within 1 dB of the true level across four reference conditions, and that there was a significant relationship between a phone's age and its accuracy.

A newer study by Sasindran and colleagues (Acta Acustica, 2026) tested only three apps on two phones, in a small sound-proofed box, with white noise and tones of 3 to 10 kHz at roughly 60 to 100 dB, against a Class I meter. The most accurate app for white noise had an average departure of 2.9 dB. In the paper's tonal results, the best app averaged a departure of 13.2 dB, falling to 6.5 dB after a calibration step. One app gave spurious values such as minus 1370 dB. An abstract from Poojary and Nayak (2026, an ear, nose and throat journal) reports that budget Android phones showed errors of up to 9.5 dBA, and says it included real-world measurements as well as laboratory ones.

Why do the readings differ?

Microphones are the first limit. Murphy and King found that the apps they tested were less efficient at measuring at background and high noise levels, and that the quality of the small microphones varied between devices. That means very quiet and very loud places are where a phone is most likely to mislead.

Calibration is the second. Kardous and Shaw ran the apps with their original calibration settings, and Murphy and King did not calibrate the phones either, to mimic real use. Kardous and Shaw (2016) later found measurements within 1 dB of the reference when the same four apps were paired with external calibrated microphones, which is extra equipment most people will not have.

Weighting matters too. Defra's noise maps use A-weighted levels (our decibels guide explains weighting), so an app set to an unweighted or different mode is not comparable. Handling is a further source of error. Kardous and Shaw say they did not assess field effects such as temperature, humidity or objects near the microphone, and, as a general acoustic point rather than a study result, a phone case over the microphone opening, a hand held close or wind across the phone can all change a reading, though none of the studies we read tested that.

Why is a few minutes not the same as a day or night level?

The levels on Defra's noise maps are long-term averages. Defra describes Lden as an annual average over 24 hours, with a 5 dB penalty for evening and 10 dB for night, and Lnight as the annual average for 11pm to 7am. Defra adds that the modelling averages out variation from weather, time of day and time of year.

A phone app, even a good one, shows the level at one place over the minutes you ran it. It is a spot measurement. Comparing it with a modelled Lden is not like for like, because a quiet ten minutes and a busy ten minutes at the same spot could differ a good deal. Our guide to what Lden and Lnight mean explains how the long-term figures are built.

What are phone apps sensible for?

They are most useful for rough comparisons. Using the same phone, held the same way, you can compare two rooms, two sides of a house, a window open against closed, or the same spot at different times. Relative differences are more trustworthy than the absolute number, though we have not found a study that tested that claim for home use, so treat it as a sensible habit rather than a proven one.

Do not lean on an app for absolute numbers. Murphy and King found apps less accurate at background and high noise levels, so very quiet or very loud readings deserve extra caution. Our 10-minute sound walk guide shows a routine where an app supports what you hear and does not replace it.

What should I not use an app for?

Do not rely on an app reading as legal evidence, as proof of compliance with a planning condition or standard, or for checking a property against a WHO guideline. Where a measured level has to stand up in a dispute or a condition, a proper measured survey by a qualified person is the right route, and for legal or planning questions you should take advice from a solicitor or your council.

If you want a view of the long-term picture before you visit, the AcoustaPLOT check gives a modelled road and rail noise estimate for an address. It is modelled rather than measured, so it complements a visit and your own ears; it does not replace them.

Limits of this evidence

Several of these studies used reverberation rooms and played steady artificial noise, which is easier than a real street with traffic, wind and echoes. The samples of phones were limited in some studies, the Android study tested only three apps on two phones and high-pitched tones, and the Poojary and Nayak result is from an abstract only, and apps and phones change quickly, so the named results may not match the version on your phone.

We did not find a study comparing app readings with annual modelled levels at homes, or one run in UK streets. We also did not test any app ourselves. The sensible conclusion is cautious: apps can help you compare, and they cannot tell you the day or night level of an address.