02Sound and decibels in plain words

The short version: Sound is a small, fast wobble in air pressure, and noise is sound somebody did not ask for. Loudness is measured in decibels, a logarithmic scale where two equal sources add up to only 3 dB more and it takes about 10 dB more for something to seem twice as loud. Every decibel number needs a label - dBA, dBC, Leq, Lmax, DNL, Lden - because each one measures something different, and comparing across them is how most noise arguments go wrong. Distance, walls and pitch then decide what actually reaches you: bass slips through walls that stop voices, and a highway carries much farther than a single horn.

What is sound, exactly?

What you hear: everything, from a bus pulling away to a fridge humming in the next room.

The FAA's plain definition is a good start: sound is "energy transferred through the air that our ears detect as small changes in air pressure."⁴ Noise is "sound that is unwanted," which gives it two sides: a physical one you can measure and a subjective one you cannot.⁴

The pressure changes involved are tiny. The quietest sound a healthy young ear can detect, the threshold of hearing, is a pressure change of about 20 micropascals, which is defined as 0 dB.⁵ The threshold of pain is put at about 140 dB, or 200 million micropascals, by the Federal Highway Administration,⁵ while the FAA puts the loudest sounds we hear without pain at about 120 dB and says most everyday sounds fall between 30 and 100 dB.⁶ The agencies differ on the top end, so "about 120-140 dB" is the honest range. That ten-million-to-one span in pressure is why decibels use a logarithmic scale.

How high and how low can people hear?

What you hear: pitch - the low rumble of a truck engine, the high whine of a scooter motor.

Frequency is how many pressure cycles pass per second, measured in hertz (Hz), and we hear it as pitch. Human hearing runs from about 20 Hz at the bottom to somewhere between 10,000 and 20,000 Hz at the top, according to the FAA.⁶ The top end falls with age and with hearing damage, which is why a teenager can hear a high-pitched hiss an adult cannot.

The ear is not equally sensitive across that range. The FAA describes peak sensitivity between about 500 and 2,000 Hz,⁶ and an acoustics vendor glossary gives a wider band of about 500 Hz to 6 kHz.⁷

Each frequency has a wavelength, the physical length of one cycle in the air. Wavelength is the speed of sound (about 343 m/s) divided by frequency, so a 20 Hz bass note is about 17 m (56 ft) long, 100 Hz about 3.4 m, 1 kHz about 34 cm and 4 kHz about 8.6 cm.⁸ A wave the length of a city bus behaves very differently from one the width of your hand, and that difference explains a lot of what follows.

Why don't decibels add up?

What you hear: a second leaf blower starts across the street, and the noise gets worse, but not twice as bad.

The decibel (dB) is "a logarithmic unit which expresses the ratio of the sound pressure level being measured to a standard reference level."⁹ Because it is a ratio on a log scale, decibels do not add like ordinary numbers.

Two identical sources together are 3 dB louder than one, because doubling the energy adds 3 dB. The FHWA's example: two 90 dB trucks make 93 dB, not 180 dB.⁹ For uneven sources, the FHWA offers a shortcut: if two levels are 0-1 dB apart, add 3 dB to the higher; 2-3 dB apart, add 2; 4-9 dB apart, add 1; 10 or more apart, add nothing.⁵ A loud source swamps a quiet one: the FHWA's 60, 60, 65 and 75 dB sources total about 75.7 dB.⁵

Perception follows a different rule from energy. A 10 dB increase is heard, on average, as a doubling of loudness, so 70 dB sounds about twice as loud as 60 dB.⁹ A 3 dB change is a "barely perceptible change" and a 5 dB change a "readily perceptible change."⁹ The FHWA calls a 20 dB(A) cut "one-fourth as loud."⁵ The "10 dB is twice as loud" rule is an average of how people judge loudness, not physics; in energy terms, 10 dB is ten times the energy.

What is the difference between dB, dBA, dBC and dBZ?

What you hear: a meter showing a much higher number in dBC than in dBA for the same thumping car stereo.

Meters filter the signal before reporting a number, a step called frequency weighting; the letter after "dB" names the filter.

A-weighting (dBA) mimics human hearing at moderate levels. It turns down frequencies below about 500 Hz and above about 10,000 Hz and is nearly flat in between, and it is the basic environmental noise measure used by EPA and most agencies.⁶

C-weighting (dBC) is much flatter and keeps more of the bass. It is "commonly used for the measurement of Peak Sound Pressure level."⁷ A bass-heavy source reads higher in dBC than in dBA, so a big gap between the two is a sign that most of the energy is low.

Z-weighting (dBZ) is "zero" weighting: "a flat and unadjusted frequency response between 10 Hz and 20kHz."⁷

Meters also apply a time weighting that controls how fast the display reacts: Fast is 125 milliseconds, Slow is 1 second, and Impulse (rarely used) is 35 milliseconds. A peak reading has no time constant at all - it catches the single highest instantaneous pressure.⁷

The rule for reading any noise number: 70 dBA and 70 dBC are different measurements of the same sound, and neither can be compared with the other. The free NIOSH phone app lets you switch among A, C and Z weighting.¹⁰ For how this plays out with concert systems, see The Show File's guide to concert sound.

Which number is the "real" level? Leq, Lmax, L10, L90 and SEL

What you hear: a quiet street, then a motorcycle, then quiet again. What was "the" level?

Leq (equivalent continuous level) is the energy average: "the level of continuous sound over a given time period that would deliver the same amount of energy" as the fluctuating sound.⁴ An A-weighted Leq is written LAeq. Many US state highway agencies use Leq(h), the hourly A-weighted Leq.⁹

Lmax is the loudest moment in the period. It ignores duration entirely: the FAA notes that a gunshot and a nearby freight train can share the same Lmax even though the train lasts far longer.⁴

L10 and L90 are statistical levels: the level exceeded 10% of the time and 90% of the time. L10 is common in traffic work; L90 is a standard measure of ambient or background sound.⁷ If a complaint is about a constant hum (see hums and drones), L90 tends to capture it. If it is about passing trucks, L10 and Lmax do.

SEL (sound exposure level) takes one event's total energy and expresses it "as if that event had occurred within a one-second period."⁴ For most aircraft overflights, SEL runs about 7-12 dB higher than the Lmax of the same event.⁶

What are DNL and Lden?

What you hear: an airport or rail report says your street is "DNL 62" or "Lden 58."

These are long-term, 24-hour averages with a penalty added for noise at night, when it bothers people more.

DNL (day-night average sound level, also written Ldn) is the US standard. It counts A-weighted sound between 10 p.m. and 7 a.m. as 10 dB louder than it really is.⁶ DNL is the standard metric for all FAA aviation noise studies.⁴ Because it is an energy average, many quiet events and a few loud ones can produce the same number: the FAA notes that DNL 65 results from 1 event a day at SEL 114.4 dBA, 10 events at SEL 104.4, or 100 events at SEL 94.4.⁴ Residents and airports can look at the same DNL and describe very different experiences; the aircraft section covers how that plays out around airports.

Lden (day-evening-night level) is the European version. By default the day runs 07:00-19:00 with no penalty, the evening 19:00-23:00 with +5 dB, and the night 23:00-07:00 with +10 dB; member states may shift the hours.¹¹ The EU's Environmental Noise Directive uses four indicators: Lday, Levening, Lnight (the night average on its own) and Lden.¹²

DNL and Lden are cousins, not twins: different hours and the evening penalty give the same sound different values, so do not read one as the other.

How much quieter is it twice as far away?

What you hear: a jackhammer fades fast as you walk away; the highway hum follows you for blocks.

It depends on the shape of the source. A point source - one horn, one jackhammer - spreads sound in a sphere, and the level drops about 6 dB(A) each time you double your distance: the FHWA's example goes from 90 dB(A) at 15 m (50 ft) to 84 dB(A) at 30 m (100 ft).⁵

A busy road behaves like a line source, a long stream of point sources that spreads as a cylinder. It drops only about 3 dB(A) per doubling: 90 dB(A) at 15 m becomes 87 dB(A) at 30 m.⁵ Over soft ground such as grass, the usual traffic analysis rate is 4.5 dBA per doubling,⁹ because soft ground absorbs about an extra 1.5 dB(A) per doubling; over hard ground, the FHWA reports levels 1-2 dBA higher for the first rows of homes near highways.⁵

So moving from 15 m to 60 m (four times the distance) cuts a single horn by about 12 dB but a highway by only about 6 dB over hard ground. The FHWA says traffic noise is usually not serious beyond about 150 m (500 ft) from heavy freeways or 30-60 m (100-200 ft) from lightly traveled roads.⁹

Weather bends sound too, an effect the FHWA calls atmospheric refraction. Downwind, sound is bent toward the ground and levels rise; upwind, it is bent away and levels fall, and studies cited by the FHWA measured wind effects of up to 7 dB(A) within 100 m of a road.⁵ On a sunny day, warm air near the ground bends sound upward; at night, a temperature inversion - cool air under warmer air - bends it back down and can raise levels at a listener, though the FHWA says temperature effects are not substantial within about 61 m (200 ft) of a road.⁵

Do noise barriers work?

What you hear: the tall concrete walls along a highway, and the quieter yards behind them.

Yes, within limits. The FHWA's handbook says a barrier that barely breaks the line of sight between road and listener gives about 5 dB(A) of reduction, a good barrier approaches 10 dB(A), and the practical maximum is about 20 dB(A) for walls and 23 dB(A) for earth berms.⁵ A row of buildings gives about 3 dB(A), plus about 1.5 dB(A) for each extra row, and a dense belt of vegetation 5 m tall and 30 m wide can give up to 5 dB(A).⁵ A barrier material of 20 kg/m2 or more provides at least 20 dB(A) of transmission loss, so the wall itself is rarely the weak point - sound bending over the top is.⁵

Why can I hear bass through my wall but not voices?

What you hear: the neighbor's subwoofer thumps clearly; the words of the song are gone.

Walls stop sound mainly by mass, and the mass law says transmission loss rises about 6 dB for each doubling of frequency and about 6 dB for each doubling of the wall's surface mass.⁸ Run that backward and the same wall blocks about 6 dB less every time the pitch halves. An acoustics firm's example: a single layer of 5/8-inch drywall that blocks 40 dB at 1 kHz would block about 34 dB at 500 Hz, 28 dB at 250 Hz and 22 dB at 100 Hz.⁸ Voices and treble get knocked down hard; bass does not.

There is a second problem. Typical two-layer drywall walls have a mass-air-mass resonance, a frequency where the wall flexes like a drum, usually around 50-100 Hz.⁸ The common US wall rating, STC (Sound Transmission Class), is averaged over 125-4,000 Hz, so that weak spot can sit below the range the rating checks.⁸ A wall with a respectable STC can still pass a kick drum, which is why bass is the usual complaint between neighbors and venues (see nightlife).

Finally, long bass waves travel through structure. They couple into floor slabs and stud cavities and go around the wall, a path called flanking, and door seals that help with speech do little for bass.⁸ These figures come from an acoustics vendor, not a standards body.

What do hard walls and tall buildings do?

What you hear: a siren that seems to come from everywhere at once on a narrow downtown street.

Hard, flat surfaces reflect sound rather than absorbing it. Tall facades keep throwing sound back into the street instead of letting it spread. The verified evidence is indirect: the FHWA reports that two parallel reflective highway barriers can degrade each other's performance by 2-6 dB(A), that reflections from an opposing barrier typically add 1-2 dB(A), and it recommends a width-to-height ratio of 10:1 or more to keep the effect small.⁵ A narrow street between tall buildings is a much tighter ratio than that.

One peer-reviewed estimate: a 2010 paper by Okada, Yoshihisa and Kuno in Acoustical Science and Technology modeled street canyons with an elevated road, checked against a 1:40 scale model, and put the mean increase from multiple reflections between buildings at "up to about 8 dB" in built-up urban areas.¹³ That figure is for one street layout and a modeled average, not a rule for every downtown block.

What is a sound level meter, and what is Class 1?

What you hear: nothing - but this is the instrument a city inspector or acoustic consultant brings.

Professional meters are graded under the international standard IEC 61672-1 as Class 1 (precision) or Class 2 (general purpose). The older US terms under ANSI S1.4 are Type 1 and Type 2, and both are still in use.¹⁴ At 1 kHz the tolerance is +/-1.1 dB for Class 1 and +/-1.4 dB for Class 2; at 20 Hz it is +/-2.5 dB versus +/-3.5 dB, and the gap widens at high frequencies.¹⁵ OSHA treats a Type 2 meter as accurate to +/-2 dBA.¹⁴

Which class is required depends on the job. In the UK, the Control of Noise at Work Regulations 2005 require at least Class 2, while ISO 20906 for unattended aircraft noise monitoring requires Class 1.¹⁵ The tolerance figures above come from a meter manufacturer rather than the standard itself, which is behind a paywall.

Is a phone decibel app accurate?

What you hear: your own phone telling you the bar is at 94 dB.

Some are good; many are not. In 2014, NIOSH researchers examined 130 iOS apps and tested 10, and examined 62 Android apps, of which only 4 "partially met" their criteria.¹⁶ Three iOS apps came within +/-2 dBA of a reference meter.¹⁶ The best, SoundMeter, averaged -0.51 dBA off. A single Android app showed "wide variance" from phone to phone, and the authors concluded that Android and Windows apps did not meet occupational needs.¹⁴ A 2015 follow-up found that with a calibrated external microphone, apps agreed within +/-1 dB of professional instruments, and NIOSH announced plans for its own free app.¹⁷

That app, the NIOSH Sound Level Meter, was announced in a NIOSH blog post dated January 2017. It reports LAeq, Lmax, Lpeak, time-weighted average and noise dose.¹⁰ In a 2019 evaluation, it read about 2 dBA off with an iPhone's built-in microphone in lab pink-noise tests and within 1 dBA with a calibrated external microphone; in an underground mine it was within about 2 dBA.¹⁰

As of its last review (February 16, 2024), NIOSH's app page says the app "is only available on iOS devices," and that "verifying the accuracy of an Android-based app is not currently possible" in its lab.¹⁸

A good iPhone app is fine for awareness - knowing whether a room is in the 70s or the 90s. For anything formal, a calibrated Class 1 or Class 2 meter is what counts.

How loud is loud? A rough decibel ruler

What you hear: everything on one scale.

The American Academy of Audiology's 2023 poster lists rustling leaves at 20 dB, a whisper at 30, a quiet library at 40, moderate rainfall at 50, normal conversation at 60, traffic at 70, an alarm clock at 80, a lawnmower at 90, earbuds at full volume at 100, concerts and car horns at 110, a jet at takeoff at 120, a jackhammer at 130 and fireworks at 140 dB.¹⁹ The poster gives no weighting and no distance, so treat it as a sense of scale, not as measurements to compare with a meter reading.

NIOSH offers a more practical yardstick: if you have to raise your voice to be heard by someone about 3 ft (1 m) away, the level is likely 85 dBA or more; if you have to shout, it is likely 95 dBA or more.²⁰ NIOSH's recommended workplace limit is 85 dBA averaged over 8 hours,²⁰ with allowed time halving for every 3 dB above that.²¹ At the other end, the NIDCD says sounds at or below 70 dBA are unlikely to cause hearing loss even after long exposure.²² The health section covers what those limits mean.

What is a noise map?

What you hear: nothing - a noise map is a model, not a recording.

Noise maps are computer models of long-term average levels from roads, rail and aircraft. In the US, the Bureau of Transportation Statistics released its first National Transportation Noise Map on March 21, 2017, modeling aviation and Interstate highway noise as a 24-hour LAeq from 2014 data.²³ It found more than 97% of the US population potentially exposed to transportation noise below 50 dBA.²³ BTS describes the modeling as worst case, with no shielding from buildings or terrain, and meant for tracking trends rather than assessing impact at any one home.²³

In Boston

The national BTS map models aviation and Interstate noise across the whole country,²³ so Boston's Interstates and Logan Airport fall inside it. Because DNL is the standard metric for all FAA aviation noise studies,⁴ any FAA noise study around Logan reports in DNL, not in the Lden figures used in European cities. The physics in this section is the same on a Boston side street as anywhere else: a single horn fades about 6 dB(A) per doubling of distance, a highway only about 3 dB(A).

What Boston does have is modeled Logan contours. According to the Massachusetts Department of Public Health's Logan Airport Health Study (May 2014), Massport produces annual DNL contours around Logan ranging from 60 to 75 dBA in 5 dB steps, using the FAA's Integrated Noise Model.²⁴

Elsewhere in the world

The physics is universal; the paperwork is not. The US uses DNL for aviation and Leq(h) or L10 on highways,⁹ while the EU uses Lden and Lnight.¹² EU law requires strategic noise maps every five years for large cities and major roads, railways and airports, where the US federal map is voluntary.¹² EU maps estimate how many people are annoyed (from Lden) and sleep-disturbed (from Lnight), and the Directive itself "does not set any limit value."²⁵ Meter standards split along the same line: US practice often cites ANSI Type 1 and Type 2, and the UK and Europe cite IEC 61672 Class 1 and Class 2.¹⁵

Report an error in this section