Decibel meter
Live level, average, peak and how long it is safe to stay. With an honest warning about accuracy.
Your browser will ask for microphone access. The audio is analysed on your machine and never sent or stored anywhere.
Move this if you have a reference sound meter beside you: adjust until both numbers agree and the rest of the session will be far closer to true.
Following the NIOSH criterion: 85 dB for eight hours, halving the time for every 3 dB above that.
A browser microphone is not calibrated, and the system may apply automatic gain of its own. These decibels are an estimate: good for comparing two places, watching something rise or fall, and spotting a clearly loud environment — not for a legal or medical measurement.
Generated right here, in your browser. What you type never leaves your device.
This tool listens through your microphone and turns what arrives into a decibel level, with its history, average and peak. It answers everyday questions: whether this office is louder than the other one, whether the new fan helped, whether the neighbour has gone past reasonable.
And it is worth saying up front what tools like this rarely admit: a browser microphone is not calibrated. A real sound meter has a capsule with a known response and a certificate; yours has a generic capsule and an operating system that may be applying automatic gain without telling you.
That does not make it useless, it makes it useful for something else. Comparisons and trends are reliable: if one place reads fifteen decibels above another, that difference is real. The absolute number is an estimate, which is why there is a calibration control for when you have a reference device beside you.
How to use it
- 1
Grant microphone permission
The audio is analysed on your machine and never sent or stored. While measuring you will see your browser’s recording indicator, which is normal.
- 2
Keep the microphone still and pointed at the noise
Moving it lets air and handling noise into the reading. Resting the device on a table near the source gives far steadier numbers than holding it.
- 3
Wait a few seconds before reading
Some systems adjust gain during the first second. Watch the average rather than the instant: the top number jumps with any knock.
- 4
Calibrate if you can
If you have a sound meter or a phone app you already trust, put them side by side and move the offset until both agree. From then on the whole session is far closer to true.
- 5
Compare rather than judging one number
Measure the same room with the window open and shut, or before and after switching something off. The difference between two of your own readings is worth far more than the absolute value.
Things worth keeping in mind
What a decibel is, and why the scale misleads
A decibel is not a unit like a metre: it is a logarithmic ratio between two pressures. That is why they do not add like ordinary numbers, and why intuition fails so badly with them.
Adding 3 dB means doubling the sound energy. Adding 10 dB is perceived as roughly twice as loud, even though the energy went up tenfold. And two identical sources together give 3 dB more, not double: two 70 dB vacuum cleaners make 73, not 140.
- +3 dB: twice the energy · barely noticeable
- +10 dB: perceived as twice as loud
- Two identical sources add 3 dB, they do not double
A-weighting: why meters do not measure what is there
A professional sound meter does not report actual pressure: it filters it to imitate how a person hears. Human hearing is fairly deaf to deep bass and very sensitive between 2 and 5 kHz, so a filter attenuates the former and lifts the latter. That is A-weighting, and it is why measurements are written dBA.
This tool applies no such filter, and saying so matters: faced with bass-heavy noise — an engine, a large air conditioner — your reading will come out somewhat higher than an A-weighted meter’s. For voices or broadband noise the difference is small.
How much noise is too much
The American NIOSH criterion sets 85 dBA over eight hours as the daily exposure limit, halving the time for every 3 dB above. At 88 dB that is four hours, at 91 two, at 100 barely fifteen minutes.
Below 85 dB no hearing-damage limit is set, but that does not mean harmless: around 55 dB already interferes with sleep, and continuous noise is associated with stress even where it does no ear damage. The WHO recommends staying below 30 dB in a bedroom at night.
- 85 dB · 8 hours a day at most
- 91 dB · 2 hours
- 100 dB · about 15 minutes
- 110 dB · under 2 minutes
What can falsify your reading
Automatic gain control is the main enemy: it raises the volume when things are quiet and lowers it when they are loud, which is exactly the opposite of what a meter needs. This tool asks for it off, but not every browser obeys.
Then there is the physical. Wind across the microphone produces enormously high readings that correspond to nothing audible. Holding it in your hand adds handling noise. And on a laptop the microphone usually sits beside the fan, so the computer contaminates the measurement once it warms up.
- Automatic gain control still active
- Wind or air conditioning blowing across the microphone
- Holding it rather than resting it
- The machine’s own fan
From the telephone exchange to an exposure limit
The decibel was born at Bell Labs in the 1920s, and not to measure noise but signal loss on telephone lines. The original unit was the bel, after Alexander Graham Bell, but it proved too large for daily use and the tenth part won out.
The first commercial sound level meter arrived in 1932, from General Radio. It was heavy and used valves, but it already carried the key idea: filter the signal so the measurement resembles what a person hears rather than what is there.
The curves underpinning that weighting were published by Fletcher and Munson in 1933, measuring at what intensity different frequencies sound equally loud. Their equal-loudness curves remain the basis of A-weighting, though they have been refined since.
Regulation came much later. The United States set workplace noise limits in 1970 with OSHA, and the stricter NIOSH recommendations — the ones this page uses — arrived in 1998. The idea that workplace noise is a measurable risk rather than merely an annoyance is surprisingly recent.
- 1920s
- Decibel born at Bell Labs
- 1932
- First commercial sound meter
- 85 dBA
- NIOSH limit for 8 hours
- 30 dB
- WHO bedroom maximum
From air pressure to a number
The microphone converts pressure variation into a voltage, and the converter digitises it into samples between -1 and 1. Over that list the root mean square is computed — the square root of the mean of the squares. That value represents the signal’s average energy far better than the peak does, because a single knock does not throw it.
That RMS becomes decibels via twenty times the base-ten logarithm. The result is dBFS: decibels relative to digital full scale, and always negative, because zero is the loudest the system can represent without clipping.
Turning dBFS into decibels of sound pressure requires knowing how much real pressure corresponds to that particular microphone’s maximum signal. That figure appears on the datasheet of a calibrated capsule, and does not exist for a laptop microphone. All that can be done is add a fixed offset — which is exactly what the calibration control does.
So the control is not decoration. Without it, the number you see depends on which microphone you have and what gain the system is applying. With a reference beside you, that offset becomes a one-off calibration good for every measurement on that machine.
The history keeps the last hundred and eighty values, one per frame, covering about three seconds at sixty a second. That is enough to see the shape of a noise — whether it is constant or arrives in knocks — which a single number never conveys.
Calibrating without a meter, using what you have
With no reference device you can still get reasonably close using sources of known level. It is not exact, but it beats leaving the default.
- 1
A closed room at night, no appliances, no traffic. That sits around 30 dB. Adjust the offset until the reading approaches it.
- 2
At a metre, a calm conversation runs about 60 dB. If after the previous step it reads 45 or 75, the offset is wrong.
- 3
A vacuum cleaner at a metre is 70 to 75 dB. A hairdryer, 80 to 90. Crude references, but useful as a control.
- 4
Do not chase a perfect match on one: find the offset that leaves all three reasonably close. Write it down, because you will need to set it again.
- 5
The calibration belongs to that microphone. Another laptop, another phone, or a headset microphone each need their own.
If something is not working
Symptom:The reading is far higher than seems reasonable
- Likely cause:
- Nearly always wind or moving air across the microphone, or the machine’s own fan.
- What to do:
- Move the microphone away from any draught and away from the computer. On a laptop, rest it on another surface and wait for the fan to slow. Check the calibration offset too, which may have been left high.
Symptom:The number will not drop even in silence
- Likely cause:
- Automatic gain control is raising the volume because it finds no sound, which is exactly the opposite of what is needed.
- What to do:
- Turn off microphone enhancement in your system sound settings. On Windows it is in the device properties; on mobile, try another browser, since not all allow it to be disabled.
Symptom:It reads very differently from my phone app
- Likely cause:
- Neither is calibrated, and each uses a different default offset.
- What to do:
- Use the calibration control to make them agree with both sources side by side. And bear in mind that if the app applies A-weighting and this page does not, a difference will remain on bass-heavy noise.
Symptom:The peak jumps enormously with any knock
- Likely cause:
- That is correct: peak records the loudest instant, and an impact triggers it even lasting milliseconds.
- What to do:
- Judge an environment by the average, not the peak. Peak is useful for the opposite job: catching isolated impacts that an average smooths away.
Symptom:The reading changes a lot when I move the microphone
- Likely cause:
- Sound pressure falls with distance, and quite fast: about 6 dB less at twice the distance.
- What to do:
- Fix a distance and note it. Without that, two of your own readings are not comparable. Resting the device somewhere fixed beats holding it.
Symptom:No microphone prompt appears on my phone
- Likely cause:
- You are inside an app’s built-in browser, which blocks access without asking.
- What to do:
- Use the app menu and choose to open in Chrome or Safari. From a normal browser the prompt appears as expected.
Things worth knowing
A decibel measures a ratio, not a sound
It always compares two values. In acoustics the second is a reference pressure fixed by convention at 20 micropascals, roughly the quietest sound a young ear can detect.
Two vacuum cleaners do not make double the decibels
Two identical 70 dB sources sum to 73, not 140. The scale is logarithmic, so doubling the energy adds three decibels and little more.
The quietest room in the world measures negative
Record-holding anechoic chambers drop below -20 dBA. Past a certain point people begin hearing their own blood circulating, and staying inside more than a few minutes becomes unpleasant.
Hearing damage depends on time, not just level
Every 3 dB above 85 halves the tolerable duration. A concert at 105 dB exhausts the daily dose in about four minutes.
A-weighting ignores bass on purpose
Because the ear largely ignores it too. Which is why a deep engine can read modestly in dBA and still shake an entire house.
Frequently asked questions
Is it accurate?
For comparisons yes; for an absolute value no. A browser microphone is not calibrated and the system may apply gain of its own. The difference between two of your readings is reliable; the exact number is an estimate, which is why there is a calibration control.
Can I use it for a noise complaint?
No. A legally meaningful measurement requires a certified, calibrated sound meter, usually class 1 or 2, and often an accredited technician. This tool tells you whether that step is worth taking.
Why does my reading differ from a real sound meter?
Two reasons compound: the lack of calibration, and that this page applies no A-weighting. A meter filters bass to imitate the ear; we do not, so against bass-heavy noise our reading comes out higher.
What level is dangerous?
The NIOSH criterion sets 85 dB over eight hours, halving the time for every 3 dB above: 91 dB is two hours and 100 dB barely fifteen minutes. For sleeping, the WHO recommends staying below 30 dB.
Is the audio stored?
No. Sound is analysed in your browser to compute the level and discarded immediately. Nothing is recorded, nothing is sent, and nothing remains when you close the tab.
Does it work with an external microphone?
Yes, and it usually gives steadier readings than a built-in one, because you can place it away from the machine’s fan. Remember to recalibrate: every microphone needs its own offset.
Related tools
Spectrum analyzer
Live spectrum and spectrogram of any source, with the window function in your hands and the real resolution printed on screen.
Microphone test
Watch the waveform and level as you speak, then record five seconds to hear yourself the way everyone else does.
Speaker test
Left, right, a 20 Hz to 20 kHz sweep and a polarity check. Every tone is generated in your browser.
Fix my speaker
A sustained low tone moves the diaphragm far enough to push water out. Turn the volume all the way up and point the speaker at the floor.
Updated