The three bands below come from two thresholds that recur in every official
document on the subject. Sound at or below 70 dBA is not expected to damage
hearing however long you sit in
itNIDCD.
From 85 dBA upward, long or repeated exposure
doesNIDCD.
Between the two, the level alone does not decide it — duration does.
-
< 70
Safe for as long as you like. Prolonged exposure is not expected to damage hearing.
-
70–85
Duration starts to matter. Below the limit, but not a level to spend a whole day in.
-
85 +
At or above the NIOSH limit. Protection is recommended regardless of how long it lasts.
Here is a short reference table for placing a reading. Every figure is the
range its source publishes, not something read off this page, and the bar
tracks the lower end of each range — so a row that straddles a boundary is
shown in the gentler band, not the harsher one. The full
decibel chart carries the long version.
Everyday sounds, as published by NIOSH and NIDCD
| Sound |
dBA |
Level against a 130 dBA scale |
Source |
| Normal conversation |
60–70 |
|
NIDCD |
| Ceiling for unlimited exposure |
70 |
|
NIDCD |
| Movie theatre |
74–104 |
|
NIDCD |
| Motorcycle or dirt bike |
80–110 |
|
NIDCD |
| Vacuum cleaner |
85–90 |
|
NIOSH |
| Lawn mower |
85–90 |
|
NIOSH |
| Power tools |
85–90 |
|
NIOSH |
| Headphones at maximum volume |
94–110 |
|
NIDCD |
| Concert |
94–110 |
|
NIDCD |
| Bar or nightclub |
95 + |
|
NIOSH |
| Large sporting event |
95 + |
|
NIOSH |
| Chain saw, bulldozer, ambulance siren |
95 + |
|
NIOSH |
| Siren, close |
110–129 |
|
NIDCD |
| Fireworks show |
140–160 |
|
NIDCD |
Three of those rows have pages of their own, because the ranges are wide enough
that the average is useless on its own:
how loud is 70 decibels,
how loud is 85 decibels, and
how loud is 100 decibels.
Press start, allow microphone access, and a number appears within a second or
two. That is the sound level around you right now, A-weighted, in dBA.
Hold the device out toward whatever you are checking rather than against your
chest or in a pocket, and give it half a minute. Sound in a real room is not
steady — a fridge cycles, a road has gaps, a machine has a duty cycle — so a
single glance catches whatever the room happened to be doing at that instant.
The session max and the running Leq underneath the number exist for exactly
this reason. The max tells you how bad the worst moment was. The Leq tells you
what the whole period averaged out to, weighted by energy rather than by
arithmetic, which is the way exposure actually accumulates.
The Slow and Fast buttons set how quickly the display follows the sound. Slow
averages over about a second, Fast over an eighth of a second. Slow is the
default here because Slow is what NIOSH recommends for a sound level
meterNIOSH,
alongside A-weighting and a 3 dB exchange rate. Fast is useful when you are
trying to catch a short event — a door, a bark, a passing motorbike — and
useless for judging a room, because it never settles.
If you want the number to line up with a meter you trust, there is a
calibration offset folded away below the trace. It shifts every figure on the
page by a fixed amount you choose. It is off by default and it should stay off
unless you have something real to compare against; setting it by feel makes the
number worse, not better. How to do it properly is its own page —
calibrate a phone decibel meter.
Decibels are a ratio on a logarithmic scale, which has two consequences people
routinely get wrong.
The first is that you cannot add them. Two machines at 80 dBA running together
do not make 160 dBA, or even 90. They make about 83. Doubling the acoustic
energy adds roughly 3 dB, which is why the exposure tables below step in
threes.
The second is that small-looking differences are not small. Ten decibels is
roughly ten times the sound energy, and to most people it sounds about twice as
loud. The gap between a room at 65 dBA and the same room at 75 dBA looks like a
rounding error written down and is enormous in the ear.
The A in dBA is a weighting curve. Human hearing is far less sensitive to
very low and very high frequencies than to the middle of the range where speech
sits, and A-weighting discounts the extremes to match. It is the weighting that
every piece of hearing-conservation guidance is written in, which is the only
reason this page uses it. There are other curves — C-weighting keeps much more
of the bass, which is why it is the right tool for a nightclub or a firing range
— and the difference between them is not academic if you are dealing with
low-frequency noise. That comparison lives on
dBA vs dBC.
This is the part most decibel meters leave out, and it is the part that decides
whether a reading should worry you.
Risk from noise is a product of level and time, not level alone. NIOSH puts it
plainly: a sound level at a given moment can sit above the exposure limit
without creating risk, provided the rest of the day is quiet enough to balance
it
outNIOSH.
The number above is a level. It is not an exposure, and this page cannot
compute your exposure, because it does not know what the other twenty-three
hours sounded like.
What the limit actually says is this. Eight hours at 85 dBA uses up 100% of a
day's allowance. Every 3 dBA on top halves the time you get:
Time to reach 100% of the daily noise dose, per the NIOSH REL
| Level | Time to 100% |
|---|
| 85 dB(A) | 8 hours |
|---|
| 88 dB(A) | 4 hours |
|---|
| 91 dB(A) | 2 hours |
|---|
| 94 dB(A) | 60 minutes |
|---|
| 97 dB(A) | 30 minutes |
|---|
| 100 dB(A) | 15 minutes |
|---|
Read that table next to the reference table above and the shape of the problem
appears. A concert sits at 94 to 110 dBA. At the bottom of that range you have
an hour. At the top you have minutes. Two hours of headphones at maximum volume
is not a marginal decision.
That arithmetic is what a noise dose calculator does, and it is the honest
companion to a level readout: it is the thing that turns a number into an
answer. The one caution is that the NIOSH limit is an occupational limit, built
around a working week with quiet evenings and weekends around
itNIOSH.
It was never designed for concerts, power tools at the weekend, or earbuds on a
commute, and applying it to those understates the total.
For life outside work the relevant figure is lower and much less well known. A
1974 EPA report recommended 70 dB(A) averaged over twenty-four hours as the
environmental limit, chosen to protect 96% of the general population from
hearing loss, with separate thresholds of 55 dBA outdoors and 45 dBA indoors
for speech interference and
annoyanceNIOSH.
Those last two are the numbers that matter for a noisy street or a neighbour,
and they sit forty decibels below the workplace limit — which is why a room
that is nowhere near damaging your hearing can still be unliveable. The EPA
figures were explicitly not issued as a standard or a regulation, and no US
agency enforces them; they are a reference point, not a line anyone can be held
to.
Not very, and the site would rather say so than imply otherwise.
Your microphone was designed to make speech intelligible on a phone call. It
applies its own automatic gain, its frequency response is not published by
anyone, and it runs out of headroom somewhere around 100 to 105 dB — which is
precisely where a number starts to matter. Nothing in a browser can undo any of
that.
The size of the gap has been measured. Chucri Kardous and Peter Shaw, both at
NIOSH, compared ten iOS sound measurement apps against a reference
instrumentNIOSH.
On A-weighted levels the closest app sat 0.52 dBA under the reference and the
two furthest sat about 13 dBA under it; three of the ten came within ±2 dBA and
seven did not. Those were native apps, on one hardware platform, written by
people who could test against the actual phones. A web page reading whichever
microphone a visitor happens to own has less to work with than that.
So the published stance here is a band, not a figure: treat what you see as
within roughly ±10 dB of what a hardware sound level meter would show. CDC says
the same thing in its own words — sound measurement apps can give you a sense
of area noise, but may not satisfy regulatory
requirementsNIOSH.
This page does not conform to IEC 61672 and must not be used for occupational
compliance, a legal filing, or a medical decision. If you need a number that
survives being challenged, you need Type 1 or Type 2 hardware, and the tradeoffs
are laid out under
online vs hardware decibel meter and
how accurate is an online decibel meter.
What survives all of that is comparison on one device. Most of the error is
systematic: your microphone is wrong in a consistent direction, so the error
largely cancels when you carry the same phone from the living room to the kerb,
or check the same machine before and after a repair, or compare Tuesday night
to Wednesday night. The difference between two readings is worth far more than
either reading alone.
Two limits get quoted in this field and they are not interchangeable.
NIOSH recommends 85 dBA averaged over eight hours, on a 3 dB exchange
rateNIOSH.
OSHA enforces 90 dBA over eight hours, on a 5 dB exchange
rateOSHA.
The second is the law in a US workplace; the first is what the research
supports.
The distance between them is not a rounding difference. Over a forty-year
working life, exposure at the NIOSH limit is estimated to leave about 8% of
workers with material hearing impairment; at the OSHA limit, about
25%Sheppard et al..
One in four against one in twelve, for the same forty years. Which number you
are held to depends on where you work, not on your ears.
NIOSH vs OSHA noise limits goes through both in detail.
Note also that NIOSH recommends hearing protection whenever levels exceed
85 dB(A) regardless of how long the exposure
lastsNIOSH,
which is a simpler rule than the dose table and a better one to carry around.
Rough, comparative, on-the-spot checks where being within ten decibels is
enough to act on.
Is this room worse than that one. Carry the same phone between them and
compare. The error follows the phone, so the gap is real even when neither
number is. The tool that does it properly — timing both captures and printing
the difference in decibels and in sound energy — is
compare two places.
Is this machine getting louder. Check it now, note the figure, check it
again after a service. A vacuum or a mower that has drifted upward is telling
you something — both sit at 85 to 90 dBA when
healthyNIOSH,
which is already the level where NIOSH wants protection. There is more under
how loud is a vacuum cleaner and
how loud is a lawn mower.
Should I have brought earplugs. A venue reading in the nineties settles the
question, and the dose table tells you how long you have. See
how loud is a concert.
Is that bike as loud as it sounds. Motorcycles run from 80 to 110 dBA
depending on the machine and how it is being
riddenNIDCD,
which is a thirty-decibel spread — an enormous range in the ear. More at
how loud is a motorcycle.
Building a record of a neighbour problem. With one large caveat: a reading
from this page is not evidence and will not survive a challenge. What it is
good for is the pattern — same phone, same spot, same time of night, over
weeks. A pattern is what a complaint needs, and the absolute numbers are the
part a council or a landlord will replace with their own. How to do that
properly is under document neighbor noise.
A rule of thumb that needs no phone at all, from NIOSH: if you have to raise
your voice to be understood by someone an arm's length away, the level is
probably
hazardousNIOSH.
At 85 dBA that is a raised voice at three feet. At 95 dBA it is a shout.
Three sibling tools sit alongside this one, in the Tools menu at the top of
every page. The sound spectrum analyzer answers what a
sound is made of rather than how loud it is, which is the question a hum, a
whine or a rattle actually poses — a fridge compressor, a neighbour's subwoofer
and a laptop fan can all read the same on a single number. The
online tone generator goes the other way and plays a
frequency you choose, for testing a speaker, hearing what a room does to a
note, or handing the analyser a signal whose right answer you already know. And
the overnight noise monitor is this meter left
running all night: it writes the level down once a second and lists every
stretch that crossed your own warning line, with the time it started and how
long it lasted.
The microphone stream is read in the browser, turned into a number, and thrown
away. No audio is recorded, buffered to disk, or sent anywhere — there is no
server in the path at all, because there is nothing to send. Closing the tab
ends the session and leaves nothing behind. The CSV export is built in the page
from data already on screen and saved by your own browser.
The only thing stored on your device is the calibration offset, if you set one,
and the language and theme you chose. Full detail is in the
privacy policy.
Yes, and there is no account. The meter, the trace and the CSV export all work
without signing up for anything.
Yes, in Safari and in Chrome, and on Android. It needs a browser with the Web
Audio API and microphone permission, which every current mobile browser has.
iOS will ask for permission every time unless you allow it for the site.
Because they have different microphones with different frequency responses and
different automatic gain behaviour, and neither has been calibrated against a
reference. A ten-decibel disagreement between two handsets in the same room is
ordinary. This is the single clearest demonstration of why the absolute number
should be treated as an estimate and the comparison treated as the useful part.
No. It is not a Type 1 or Type 2 instrument, it does not conform to IEC 61672,
and it has no calibration certificate. For anything that has to survive being
challenged you need hardware and, usually, someone qualified to operate it.
In practice a phone microphone runs out of headroom somewhere around 100 to
105 dB, and above that the page reports a lower bound rather than a level — the
clipping warning appears when the input hits the ceiling. A siren or a
fireworks display will read far too low for this reason.
Only to load. Once the page is open the audio processing is entirely local, and
losing your connection does not stop the meter.