Decibel Meter

See how loud it is right now, and how many dB a change makes

The decibel meter turns the microphone signal into a sound level every 0.1 seconds and shows the current value, the average (equivalent level) over the whole run, the maximum and the background level (L90). The chart recorder draws the last 60 seconds like a pen on paper, and the distribution strip under the level bar shows how much time was spent at each loudness.

The default unit is dBFS, the size of the digital signal coming from the microphone. Phone and laptop microphones differ in sensitivity and input volume, so adding a fixed number to dBFS to get sound level meter dB gives very different results from device to device. With a sound level meter or a calibrated measurement app at the same spot, the reading can be matched once and turned into a dB(A) estimate. Without a reference, use it to compare before and after on the same device.

When a number for loudness helps

A sound can feel loud in the moment, but it is hard to compare with one heard yesterday. A level measured at the same spot with the same microphone makes "how much quieter" a difference in dB instead of an impression.

• Appliances and windows: compare fan or air purifier speed settings, or how much outside noise a closed window removes. Uncalibrated dBFS still shows differences within the same device. Because the value is unweighted, though, a comparison dominated by low sounds, such as traffic noise outside a window, can come out different from what the ear perceives.
• Night noise logs: keep a record of when noise happened and how loud it got, using the full chart, the maximum and L10. It is not a compliance measurement, so use it only as supporting notes for a landlord, building manager or local authority.
• Study, work and sleep spaces: find the times and spots with the lowest background level (L90). The average alone can make a quiet room look noisy because of a few loud events.
• Hearing health: at a concert or workshop, a calibrated dB(A) estimate gives a rough idea of how long the NIOSH guideline allows a person to stay.

Steps for a first measurement

1. Place the device on a desk or stand. Held in the hand or moved around, it picks up handling noise as well.
2. Press Start and choose Allow when the browser asks for microphone permission. Merely opening the page never turns the microphone on.
3. Stay quiet for about 10 seconds so the background level (L90) and the chart settle.
4. To compare conditions, type a measurement name, press Save and start new, change one condition and measure again. From the second measurement on, the difference from the first is shown.
5. With a sound level meter or calibrated app available, enter its value in the Calibration field of the settings table and press Match current level. The unit changes to dB(A) estimate and the everyday sound scale appears.
6. When finished, press Stop to turn the microphone off. Resume adds to the same measurement, while New run clears the statistics.

What is the difference between dBFS and a dB(A) estimate?

dBFS is a unit where the largest level a digital signal can hold is 0. The tool squares the samples of each 0.1 second block, averages them (mean square) and converts the result with 10 × log10(mean square). A sine wave of amplitude 0.1 reads −23.0 dBFS, not −20 dBFS, because the mean square of a sine wave is half the square of its amplitude. Meters that follow the convention of setting a full-scale sine wave to 0 dBFS (AES17) show the same sine wave as −20 dBFS, so allow for this difference when matching numbers with other software. If every sample is zero the calculation gives minus infinity, so the display says "Silent".

dB(A) weights each frequency according to how the human ear hears it, with very low and very high sounds counting for less. The tool builds the A-weighting formula of the international standard IEC 61672-1 as a digital filter, applies it to the microphone signal and adds the calibration offset to produce the dB(A) estimate. As the table shows, a low rumble can look large in dBFS and still shrink a lot in dB(A).

How to calibrate, and what accuracy to expect

Calibration is a single measurement of how many dB must be added to the A-weighted dBFS of this microphone to get the real dB(A). After starting, play a steady sound (a fan, an extractor or steady music), type the dB(A) value shown by a sound level meter or calibrated app at the same spot into the calibration field and press Match current level. The tool stores the difference from the A-weighted average of the last 3 seconds as the offset.

1. Hold the reference device and this microphone close together and facing the same way. Any difference in distance ends up inside the offset.
2. Calibrate on a sound that stays steady for more than 3 seconds, not on speech or door slams that start and stop.
3. The offset is saved in this browser together with a microphone ID (a random string used only on this site) and applies only to the same microphone. Switching to a different microphone, such as an earphone microphone, returns the display to dBFS.
4. Calibrate again after a browser or operating system update or a change in input volume.

A calibrated phone app can come close to a reference meter. When NIOSH researchers (Kardous & Shaw, 2014) compared sound measurement apps with a reference meter, two of the apps that met the conditions had mean differences of 0.07 dB (unweighted) and −0.52 dB(A), and two others were within ±2 dB. Those results come from specific iOS devices and apps, though, and the NIOSH app itself states an accuracy of ±2 dBA while noting that Android devices cannot yet be validated. Read the dB(A) estimate of this web tool with a wider margin than that.

Reading the average, last minute, L90 and L10

Decibels are logarithmic, so values cannot simply be added and divided. The average (Leq) here is the mean of the mean squares of all 0.1 second blocks, converted back to dB. This is the equivalent continuous sound level used in noise measurement. Loud sounds carry far more energy, so a short loud event pulls the average up a lot.

• L90 (background): the level exceeded for 90% of the measuring time. It comes close to the sound that is always there once occasional noises are set aside.
• L10 (loud events): the level exceeded for 10% of the time, which describes how loud the frequent interruptions are. A large gap between L10 and L90 means a quiet place with occasional noise.
• Last minute: the average (Leq) of only the last 60 seconds. The overall average keeps remembering an early loud sound, while the last minute follows the current situation quickly.
• Max and min: the loudest and quietest 0.1 second blocks. For a very brief sound such as a hand clap, they read lower than the instantaneous peak.

Comparing conditions

The measurement comparison works on one rule: same spot, same microphone, same duration, one condition changed. As in the sample on first load, measure about 30 seconds with the fan off and save it as "Fan off", then switch the fan on, measure the same length and save as "Fan on". The second row then shows the difference from the first. The bar is the typical range from L90 to L10, the dot is the average and the vertical line is the maximum.

Two equally loud sounds together carry twice the energy, which is about 3 dB more. Even without calibration, differences within the same device read correctly, because the offset adds the same number to every value and leaves the gap between two measurements unchanged. Comparing a phone measurement with a laptop measurement, on the other hand, mixes in the microphone difference and means little. That is why changing the microphone starts a new measurement, and a measurement taken with another microphone shows "Other microphone" instead of a difference.

Everyday sound levels and hearing guidelines

With a calibrated dB(A) estimate, the examples below help to judge the size of a sound. They are everyday noise examples published by the US CDC, and the ranges are wide because distance and equipment vary.

For hearing health, the recommendation of the US NIOSH is the most widely used: no more than 85 dBA averaged over 8 hours, with the allowed time cut in half for every 3 dB increase. When a calibrated average (Leq) rises above 82 dB(A), the readout shows the daily time allowed under this guideline.

The WHO and ITU safe listening standard sets 80 dB for adults and 75 dB for children, for 40 hours a week, for personal audio devices such as earphones and headphones. This tool cannot measure the sound inside an earphone, so treat that standard only as a reason to turn the volume down.

OSHA, NIOSH and neighbor noise: can this tool be compared with US limits?

Workplace noise in the United States is covered by two different numbers. OSHA's permissible exposure limit in 29 CFR 1910.95 is 90 dBA for 8 hours, and Table G-16 halves the allowed time for every 5 dB increase. NIOSH recommends a stricter 85 dBA with a 3 dB exchange rate. OSHA also requires a hearing conservation program when exposure reaches an 8-hour time-weighted average of 85 decibels, which the standard calls the action level.

• The OSHA values come from Table G-16, which is read at slow response on the A scale. The standard also says that impulsive or impact noise should not exceed 140 dB peak sound pressure level.
• The OSHA action level of 85 dB is measured as an 8-hour TWA on the A scale (slow response). Noise from 80 to 130 decibels counts toward that measurement.
• The NIOSH column is calculated here from 85 dBA for 8 hours with the allowed time halved every 3 dB, and rounded.

Neighbor noise works differently. There is no single federal limit for noise between neighbors or apartments in the United States, so the rule depends on the state, county or city ordinance, and many of those set limits by zone and time of day or only describe noise that is "unreasonable". The number a local ordinance sets, and the way it must be measured, are specific to that place, so check the ordinance of the city in question.

That is why a reading from this tool cannot show that an OSHA or NIOSH limit, or a local ordinance, has been exceeded. Official measurements require a calibrated sound level meter, a defined measuring position and, for workplaces, a full-shift dosimeter or time-weighted average. A phone microphone with automatic processing does not meet that. The full chart is useful as a dated note of when a sound happened, to bring to an employer, safety officer, landlord or local noise authority.

When the numbers look strange

• The value stays at the same level: automatic gain control or noise suppression raises quiet sounds and removes steady ones. Check the "Voice processing" line in the settings table for what the browser reports. Even when asked to turn these off, the device may not comply.
• Input hit the maximum warning: a loud sound right in front of the microphone clips at the digital maximum (0 dBFS) and reads lower than it really is. Move the microphone away from the sound or lower the operating system input volume.
• Muting: if the microphone is muted by the operating system or the device, the status line says so, and that time is left out of the log until the microphone is unmuted.
• Wind and hand noise: a fan blowing directly onto the microphone, or holding the device in the hand, adds loud airflow and friction sounds.
• Permission errors: the microphone must be allowed for this browser in both the site settings next to the address bar and the operating system privacy settings. A meeting app that is using the microphone can prevent it from opening, and in-app browsers inside messengers often block it, so open the https address in a regular browser.
• Screen turns off: while measuring, the tool asks supported browsers to keep the screen on, but locking a phone can stop the microphone. For a long session, plug in the charger and keep the screen on.

Limits

• Before calibration the value is the digital microphone input (dBFS), not sound pressure (dB SPL) or dB(A). It cannot be used to compare different devices.
• The dB(A) estimate is an approximation fitted to a single reference value entered by the user. The microphone's frequency response, the device's voice processing and the error of the reference device all remain in it.
• At 44.1 and 48 kHz the A-weighting digital filter matches the standard formula to within about 0.1 dB from 20 Hz to 5 kHz, but high sounds are counted low: about 1.2 to 1.5 dB low at 10 kHz and about 6 to 9 dB low at 16 kHz. When the sample rate drops to 16 kHz, as with a Bluetooth headset, even 5 kHz reads about 1.5 dB low.
• The values are mean squares over 0.1 second blocks, which is not the same as the Fast (0.125 s) or Slow (1 s) time weighting of a sound level meter, and the maximum is lower than the instantaneous peak.
• This is not a measurement for legal noise limits, occupational exposure assessment or evidence in a dispute. Those need an official method and a certified sound level meter.
• Sound is neither recorded nor uploaded, and measurements disappear when the page is closed. Only the calibration offset and the microphone ID are stored, and the microphone name is not.

Reference date The sources and regulations were checked on 2026-10-20.

Sources

• 01OSHA: 29 CFR 1910.95 Occupational noise exposureTable G-16 (90 dBA for 8 hours, time halves every 5 dB), action level of 85 dB 8-hour TWA, 140 dB peak impulsive limit
• 02CDC NIOSH: Noise and Occupational Hearing LossRecommended exposure limit of 85 dBA (8 hours), allowed time halved for every 3 dB
• 03CDC: Listen Up! Protect Your HearingEveryday noise examples from a 30 dB whisper to 110–129 dB sirens
• 04WHO and ITU: Safe listening devices and systemsPersonal audio device limits of 80 dB (adults) and 75 dB (children), 40 hours a week
• 05Kardous & Shaw (2014) Evaluation of smartphone sound measurement applications, J Acoust Soc Am 135(4)Mean differences of 0.07 dB (unweighted) and −0.52 dB(A) for two apps
• 06CDC NIOSH: Sound Level Meter AppCalibration for built-in and external microphones, accuracy ±2 dBA, Android not validated
• 07A-weighting (IEC 61672-1 weighting function R_A(f))A-weighting curve formula, normalized at 1 kHz
• 08MDN: MediaDevices.getUserMedia()Secure context required; errors for denied permission, no device and device in use
• 09MDN: MediaTrackConstraintsRequests for autoGainControl, noiseSuppression and echoCancellation
• 10W3C: Web Audio APIProcessing microphone samples directly with an AudioWorklet

Frequently asked questions

Tools to use with it