Decibel Meter: Measure Noise in dBA, Max and Leq

This decibel meter reads your microphone in the browser and shows the current sound level, max, min and Leq, with A, C or Z weighting and a Fast or Slow response. It starts in dBFS, a level relative to the loudest signal your mic input can pass, because no web page can know your mic's sensitivity. Set one calibration offset against a real meter and it shows estimated dBA. That's enough to compare rooms, times of day and before-and-after fixes, but not for a compliance check or legal evidence.

My rule is simple: trust the differences, not the absolute number. Two readings taken with the same device, browser, spot and settings compare well, so if the Leq drops after you close a window, that drop is real even when the absolute value isn't. The absolute value is only as good as your calibration, and the published phone tests further down show how far an uncalibrated reading can wander.

How to use the decibel meter

1. Start the meter. Press Start meter and allow the microphone. The page asks the browser to switch off echo cancellation, noise suppression and auto gain, then prints what the browser granted under the mic selector.
2. Pick the weighting. Use A (dBA) for everyday noise and hearing questions, C (dBC) for bass-heavy noise and peaks, and Z for a flat, unweighted reading.
3. Pick the response. Fast (125 ms) follows speech, door slams and other quick changes. Slow (1 s) steadies a level that keeps jumping, and it's the response OSHA's limits are written for.
4. Place the device. Point the mic toward the noise, keep it away from your body, walls and table edges, and keep it out of wind and air vents. For comparisons, use the same spot and height every time.
5. Calibrate when you need dB, not just differences. Open Calibrate to estimated dB, put the mic next to a sound level meter set to the same weighting, play a steady sound such as pink noise, type the meter's reading and press Match over 3 s. The page sets the offset so its 3-second Leq equals the reference, then labels readings "dBA est." (or dBC or dBZ est.).
6. Measure a full cycle of the noise. Let the meter run through a representative stretch: one full pass of a dishwasher, a whole song from next door, a typical hour of traffic. Read Leq for the average, Max for the loudest moment and Min for the background. Reset stats starts a new period.
7. Stop and release the mic. Press Stop or Escape. Hiding the tab or leaving the page also stops the meter and releases the microphone.

A steady source makes calibration easier: the pink noise generator plays one through a speaker while both meters listen. I'd also play a 1 kHz tone from the online tone generator once: all three weightings are 0 dB at 1 kHz, so A, C and Z should read within a fraction of a dB of each other. A clearly higher C or Z reading means low-frequency room noise is mixing in, so pick a quieter moment to calibrate.

Reading the numbers

The big number is the current time-weighted level. Its label names the weighting and the response: LAF means A-weighted, Fast, and LCS means C-weighted, Slow. Max and Min are the highest and lowest values of that same time-weighted level since the stats last started: at Start, Reset stats, a change of weighting or response, or a calibration match. A single click won't set Max on Slow the way it does on Fast. The meter waits 0.5 s before it starts counting, which keeps the start-up settle out of the stats.

LAeq is the equivalent continuous level: the steady level that carries the same sound energy as everything measured, which is what sound level meters report as Leq or LAT. It's an energy average, not an average of the displayed numbers, so loud moments count far more than quiet ones. A quiet room with a few loud bangs gets a Leq well above its usual background.

Fast and Slow aren't refresh rates. They're exponential time constants: after a sudden step in level, the reading gets to within 2 dB of the new level after one time constant, and when the sound stops it falls about 34.7 dB per second on Fast and 4.3 dB per second on Slow. That's why Slow looks calm next to Fast on the same noise.

Sample peak is the loudest single sample in dBFS, unweighted. When it touches 0, the input is clipping, the card shows a warning, and loud readings are too low because the waveform's tops are cut off. Move back from the source or turn the system input level down. Before calibration, dBFS here is 20 times the log of the RMS value, so a full-scale sine reads −3.0 and digital silence reads −120.0, the card's floor.

dB, dBA and dBC in plain terms

A decibel is a ratio on a log scale. Every 10 dB is ten times the sound energy, and every 3 dB is double, which is why 88 dB isn't a little louder than 85 dB for exposure purposes: it carries twice the energy. On its own, "dB" says nothing about which frequencies were counted, so meters add a weighting letter.

A-weighting turns low bass down hard and leaves the midrange alone, to account for the ear's lower sensitivity to low frequencies. It was first meant for quiet sounds and is now the standard curve for environmental noise and hearing-damage limits. C-weighting is nearly flat and only rolls off at the extremes, which is why NTi Audio's guide to frequency weightings notes that C is often used for peak levels. Z means zero weighting: every frequency counts equally. These are the IEC 61672 values the card's filters follow:

The gap between C and A is a quick bass check. A neighbor's subwoofer, a truck idling outside or a ventilation rumble reads much higher in dBC than in dBA, because A ignores most of that energy. When the two are close, the noise sits mostly in the midrange. To see exactly which frequencies are behind a reading, run the same sound through the audio spectrum analyzer.

In the card, the A and C filters stay within 0.1 dB of the IEC formula from 10 Hz to 10 kHz at 44.1 and 48 kHz, which its test suite checks. When the browser's audio runs below 40 kHz, which some phones and headsets use, the top octave is less exact and the card says so. The settings line under the Microphone menu also shows the rate the mic itself reports, when the browser gives one; a mic that captures below 40 kHz can't carry sound above half its rate, even if the browser runs at 48 kHz.

How loud is it? Common sound levels

The levels below come from the WHO's safe listening questions and answers, which also list how long per week each level can be heard safely. Treat them as typical values, not as targets for your meter: a real hair dryer or car horn can sit well above or below its row.

NIDCD gives ranges rather than single numbers in its page on noise-induced hearing loss: normal conversation at 60–70 dBA, a movie theater at 74–104 dBA, sporting events, concerts and headphones at maximum volume at 94–110 dBA, and sirens at 110–129 dBA. The spread inside each range is a reminder that "a concert" or "a siren" isn't one level.

Safe exposure levels: NIOSH, OSHA and WHO

Exposure limits pair a level with a time. NIOSH sets its recommended exposure limit at 85 dBA over an eight-hour shift and halves the allowed time for every 3 dBA above that. OSHA's permissible exposure limit is 90 dBA for eight hours, measured on the A scale at slow response, and its Table G-16 in 29 CFR 1910.95 halves the time for every 5 dBA. OSHA also calls an 8-hour average of 85 dBA the action level: at or above it, the employer has to run a hearing conservation program.

The NIOSH column is the arithmetic of its own rule, 8 hours halved for every 3 dBA above 85. OSHA's values are read straight from Table G-16. The two differ in their starting point and exchange rate: NIOSH halves the time for every 3 dBA above 85, OSHA for every 5 dBA above 90. NIOSH's limit is also a recommendation, while OSHA's is the enforceable regulation.

Outside work, the WHO frames safe listening per week: 80 dB for up to 40 hours a week, falling to four hours a week at 90 dB, and it calls levels below 80 dB unlikely to cause hearing damage. NIDCD puts it this way: sounds at or below 70 dBA are unlikely to cause hearing loss even after long exposure, while long or repeated exposure at or above 85 dBA can. These are exposure guidelines, not a diagnosis, and an uncalibrated browser reading can't tell you whether you crossed 85 dBA.

Test results: how accurate are phone and browser meters

Across ten published studies of smartphone sound level apps from 2014 to 2021, calibration and an external mic narrow the spread far more than they improve the best case. The best app in each study lands within about 1 dB (median of three studies, range 0.5 to 5 dB); the typical uncalibrated app doesn't. All ten measured phone apps, not browser pages. A browser adds its own audio path and can't guarantee raw input, so treat these numbers as the best case for a meter like this one.

The first row comes from Kardous and Shaw's 2014 test of iOS sound level apps, which ran every app with its original settings; the median is of the absolute A-weighted mean differences in its table. The best-app row takes that study's best A-weighted app (−0.52 dBA), the best of the apps on 100 smartphones tested in a reverberation room (within ±1 dB across four conditions) and the best of five apps on an iPhone 4S, which stayed within 5 dB of a calibrated meter. It mixes a mean difference with "within" bounds, so read it as a ballpark.

Three studies fill the calibrated rows: Kardous and Shaw's 2016 follow-up with external microphones on six iPhones (within ±1 dB of the reference), a 2017 iPod study with an external mic (lab TWA within ±0.5 dBA except one outlier, then 1.7 dBA above paired dosimeters in field use) and a 2019 study in which the NIOSH SLM app with external mics stayed within ±1 dBA in the lab and in a mine. A 2016 study adds that external mics paired with certain apps didn't differ significantly from a Type 1 system.

Weak spots repeat across studies. Android apps varied more between devices in 2014, and a 2019 comparison found Android apps under-reading at 90 dBA and fit only for screening. A 2021 study found apps over-reading sounds below 40 dBA even with calibrated mics, which matters for bedroom and night-noise checks. Older phones measured less accurately in the 100-phone study. On the other side, an app with an external mic met most Class 2 requirements of IEC 61672 in 2018.

What follows for this meter: uncalibrated, expect the absolute number to be off by a few dB and sometimes by much more; calibrated against a real meter in the same spot, the same mic, browser and settings can track levels to within a dB or two in the range the studies covered. I'd calibrate before trusting any number near 80–85 dBA, where a few dB changes the safe time.

Limits: what moves the number

The mic sets the ceiling. Phone MEMS microphones handle roughly 30 to 120–130 dB SPL, according to Kardous and Shaw's 2014 paper, and the system's input gain can clip well before that. Watch the clip warning near loud machinery, drums or sirens; once it shows, the reading is too low, not too high.

Browser processing is the second variable. Asking for echo cancellation, noise suppression and auto gain to be off is only a request: the MDN reference for autoGainControl says the browser will attempt it but won't fail if it can't. The card reads the result back with getSettings() and warns when a setting stays on or isn't reported. With auto gain on, the meter can't be calibrated in any useful way, because the gain moves with the noise.

Placement changes the reading on its own. Level falls as you move away from a source, so a reading at the desk and one next to the machine aren't comparable. Wind across the mic adds low-frequency rumble that reads as noise, which is what the foam windscreen on a handheld meter is for. Your body and nearby walls reflect sound too; hold the device out and away, or set it down in the same place each time.

Very quiet rooms are the hardest case: the mic's self-noise and the over-reading below 40 dBA from the studies above set a floor. Treat any reading in the 20s or 30s as "quiet", not as a precise value. A calibration also only holds for the chain it was made on. Change the mic, browser, operating system update, input level or device, and match it again; the card warns when the active mic's name differs from the one you calibrated with.

The microphone test is a quick way to confirm which mic the browser is using and that it records at all before you measure. This meter measures sound in the room; for the loudness of a recording or a stream, the LUFS meter is the right tool.

Noise complaints and workplace checks

For a neighbor or landlord dispute, a browser meter is a good log, not proof. Record the date, start time, duration, LAeq and Max for each episode, plus what you heard and where you stood, and keep the settings identical between nights. That pattern is useful to a mediator or an enforcement officer, who will measure with a calibrated meter under the method your local noise rule specifies. Most phone apps aren't accurate enough for legally required measurements, and an uncalibrated browser page is a step below them.

At work, the question is usually an 8-hour average against 85 or 90 dBA, and that's a job for a noise dosimeter or a Class 1 or Class 2 sound level meter. The browser meter can still tell you which area or machine is loudest and whether a fix lowered the level, which is where to point the proper measurement. The WHO itself names the NIOSH Sound Level Meter app for people who want to monitor their exposure with a phone.

Verdict

My verdict: use this decibel meter for relative checks and rough levels, and step up before anything rides on the number. It's a fine way to compare two rooms, see whether a door seal or a rug made a difference, find the loudest corner of a workshop or log when the noise next door starts. Calibrated once against a real meter, it gives a believable estimate in dBA for the same mic and browser. I wouldn't use it for workplace compliance, a legal complaint or a hearing safety decision near the limits. For those, I'd step up to the NIOSH SLM app with a calibrated external mic, or to a Class 2 meter, and to a Class 1 meter when the result has to stand up formally.

Decibel meter FAQ

Are online decibel meters accurate?

Not in absolute terms without calibration. In a 2014 test of 10 uncalibrated iPhone and iPad apps, the median error was 2.5 dBA and the worst about 13 dBA; with a calibrated external mic, three later studies found about 1 dB. A browser adds its own processing, so treat an online meter as a relative tool until you set an offset against a real meter.

How loud is 60 decibels?

About the level of normal conversation. The WHO lists normal conversation at 60 dB and NIDCD at 60–70 dBA. It's well below the 80 dB the WHO gives as the start of weekly listening limits.

How loud is 85 decibels?

About heavy traffic heard from inside a car, in the WHO's table. It's the level NIOSH treats as hazardous and its 8-hour recommended limit, and OSHA's action level as an 8-hour average. Under the NIOSH 3 dB rule, the safe time halves with every 3 dBA above it.

Can I use a decibel meter app for a noise complaint?

Use it to document a pattern, not as proof. Log date, time, duration, LAeq and Max with the same settings each time, then let the enforcing office measure with a calibrated meter, since phone and browser readings aren't accepted as legally required measurements.

Why does this meter read differently from my phone app?

Each app applies its own calibration offset, weighting and averaging, and the browser may process the mic differently from a native app. Match both to the same weighting and response, then calibrate this meter against the app or, better, a real sound level meter.

What is the difference between dB and dBA?

dB is the level with no frequency weighting stated. dBA applies A-weighting, which turns low bass down hard, by 19.1 dB at 100 Hz, and leaves 1 kHz unchanged, to account for the ear's lower sensitivity to low frequencies. Hearing limits from NIOSH and OSHA are written in dBA.