Decibel Addition (Adding Sound Levels)
Two machines of 85 dB side by side do not make 170 dB, but 88 dB: the decibel is a logarithmic scale, so you add energies, not levels. Enter the levels of your sources — measured or taken from data sheets — to get the resulting sound level.
The levels you add must be consistent: all in dB, or all in dB(A), measured at the same point and over the same period. The result is then expressed in the same unit as the values you entered.
Source levels
Why decibels cannot be added arithmetically
Two machines at 40 dB(A) do not make 80 dB(A). That howler is everywhere, and it comes from treating a logarithm as if it were a metre. A decibel is a ratio, written as ten times the log of a power (or twenty times the log of a pressure) against a reference. You cannot add the logs and pretend you added the powers.
The ear is roughly logarithmic (Weber–Fechner is the old name for that). The arithmetic is not. To combine levels you go back to energy, add, then take the log again. Acousticians call it energetic summation. It is the only addition this page will do.
Ltot = 10 · log₁₀( 10L₁/10 + 10L₂/10 + … + 10Lₙ/10 )
That is why two identical, uncorrelated sources raise the level by 3 dB, not by a factor of two on the display.
The +3 dB rule: doubling the number of identical sources
Same level L, two independent sources (not locked in phase):
Ltot = L + 10 · log₁₀(2) ≈ L + 3.01 dB
Double the count, or double the acoustic power, and you pick up 3 dB. Four identical machines: +6 dB. Ten: +10 dB. Halve the power: −3 dB. The rule dies the moment the levels differ — then you need the full sum, or the difference table below.
In fact “identical” here means the same level at the microphone, not the same nameplate. Two equal pumps at different distances are not a +3 dB pair.
Mental arithmetic: the table of level differences (0, 1, 2, 3… dB)
On site you often add two numbers in your head. Let L₁ ≥ L₂, Δ = L₁ − L₂. Add a small correction C(Δ) to the louder one:
| Δ (dB) | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | ≥10 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| C (dB) | 3.0 | 2.5 | 2.1 | 1.8 | 1.5 | 1.2 | 1.0 | 0.8 | 0.6 | 0.5 | 0.0 |
78 dB + 75 dB: Δ = 3, C = 1.8, total 79.8 dB. The exact formula agrees. Once Δ hits 10 dB the quieter source has dropped out of the sum for any field purpose. You can forget it.
dB, dB(A), dB(C): which weighting for which purpose
A bare “dB” is just a ratio. Frequency weightings try to follow the ear.
- dB(A) — A-weighting. Cuts lows and the very top, roughly the 40-phon contour. Default for environmental noise and occupational exposure (ISO 1996, the EU noise directive, HSE in the UK). If a regulation says “dB” and does not specify, it almost always means A.
- dB(C) — flatter, closer to 100 phon. Peaks, bangs, gunfire, and anything where the bass is the story (a slow diesel, a wind turbine at night).
- dB(Z) / linear — no filter. Spectra, power sums, engineering. Not a dose.
For the same noise, A is less than or equal to C, because A throws energy away. Never add an A number to a C number and call it a total.
Pressure level, power level, intensity level: telling them apart
Three different quantities wear the same “dB” jacket. Mixing them is a classic report error.
- Sound pressure level Lp (SPL, often Leq) — what a sound level meter and an eardrum see. Reference 20 µPa.
- Sound power level Lw (SWL) — total acoustic watts leaving the machine, independent of distance. Reference 1 pW (10−12 W). The manufacturer's figure.
- Sound intensity level LI — watts per square metre. Reference 10−12 W/m². In a free field, LI and Lp track each other.
In free field, Lp drops 6 dB per doubling of distance (inverse square). Lw does not move. You can convert one to the other only if you know the distance, the radiation pattern and the environment. This calculator adds pressure levels, not nameplate watts.
Subtracting background noise from a measurement
The meter reads source plus residual. Energetic subtraction peels the residual off:
Lsource = 10 · log₁₀( 10Lmeas/10 − 10Lbg/10 )
Field table, from the difference measured − background:
| Meas. − background (dB) | ≥10 | 6–9 | 4–5 | 3 | ≤2 |
|---|---|---|---|---|---|
| Subtract (dB) | 0 | 1 | 2 | 3 | unreliable |
Under 3 dB of margin the correction is already >3 dB and the reading is junk. Kill the background, or move the microphone in, or stop pretending you measured the source.
Adding levels octave band by octave band
A spectrum is a row of band levels — 63, 125, 250, 500 Hz, 1, 2, 4, 8 kHz, and so on. The overall linear level is the same energetic sum this page already does, one term per band. To get an A-weighted overall, apply the A correction to each band first, then sum. Doing it the other way (sum linear, then “A” by eye) is not the same number.
Lglobal = 10 · log₁₀( Σ 10Li/10 )
Example: 70 dB at 125 Hz, 75 at 500 Hz, 72 at 2 kHz → about 77.6 dB linear. That is also how you rebuild LEX,8h from a logged spectrum, after A-weighting.
What the calculation does not tell you: perceived annoyance and the +10 dB doubling of loudness
Energetic sum is physics. Loudness is a different curve. A rule of thumb that has survived a century of psychoacoustics: +10 dB is heard as “about twice as loud” (Stevens, roughly). +3 dB — a doubling of energy — is a small step in a workshop, often just noticeable in steady noise.
Annoyance then piles on extras the formula never sees: tonality, impulsiveness, time of day, whether the noise “means” a neighbour or a job. Lden and Lnight try to fold some of that into a single number. They are still not a verdict on how angry someone will be at 2 a.m.
Regulatory limits and noise exposure values at work
In Great Britain the Control of Noise at Work Regulations 2005 (and in the EU, Directive 2003/10/EC) use the same three rungs. OSHA in the US is a different ladder — 90 dB(A) PEL, 5 dB exchange rate — so do not paste a European action value into an American plant.
- Lower exposure action value — LEX,8h = 80 dB(A) and/or peak 135 dB(C). Inform, offer hearing protection.
- Upper exposure action value — 85 dB(A) / 137 dB(C). Programme of control, hearing protection compulsory in the zone.
- Exposure limit — 87 dB(A) / 140 dB(C) at the ear, after the protector's attenuation. Must not be exceeded.
Peaks (impacts, shots) live on the C-weighted peak, not on A. LEX,8h is an eight-hour energy average: four hours at 88 dB(A) is already an 85 dB(A) day.
Worked example: a workshop with four machines
Four A-weighted levels, all at the same workstation: 84, 82, 79 and 76 dB(A).
- 84 + 82 → Δ = 2 → C ≈ 2.1 → 86.1 dB.
- 86.1 + 79 → Δ ≈ 7 → C ≈ 0.8 → 86.9 dB.
- 86.9 + 76 → Δ ≈ 11 → C = 0 → still 86.9 dB(A).
Ltot = 10 · log₁₀(108.4 + 108.2 + 107.9 + 107.6) ≈ 87.2 dB(A)
Call it 87 dB(A). That is above the upper action value. Hearing protection in the zone is not a debate; reducing the two loudest machines is the actual work.
Common mistakes: inconsistent levels, different measuring points, unequal durations
- Adding 80 + 80 = 160 — physically empty. The energetic answer is 83.
- Mixing A and linear, or A and C — different filters, not addable.
- Different distances — Lp is a point. Normalise, or do not add.
- Ignoring duration — Leq,T is an energy average over T. A one-minute slam and an eight-hour hiss with the same “level” are not the same dose.
- Skipping the background strip — you then publish the room, not the machine.
- Adding Lw as if it were Lp — different references, different geometry.
Coherent sources: when two loudspeakers make +6 dB
The +3 dB rule assumed incoherent sources — independent machines, random phase. Two radiators playing the same sine, in phase, at the same point, add as pressures, not as powers: +6 dB. Out of phase they can cancel. A stereo pair on a mono bass note sits somewhere between those extremes, which is why a path difference of λ/2 is a notch, not a +3.
Workshop machines are incoherent. PA stacks at one frequency are not. Know which sum you are doing.
Leq, dose and the 3 dB exchange rate
Leq,T is the constant level that would carry the same energy over time T. The European (and UK) exchange rate is 3 dB: +3 dB halves the allowed time. 88 dB(A) for 4 h equals 85 dB(A) for 8 h. OSHA's 5 dB exchange is more lenient on the clock and must not be mixed into the same spreadsheet.
This calculator does not know duration. Feed it simultaneous levels, then fold time in afterwards with Leq.
Before you add: the inverse-square step
A manufacturer's Lw or a catalogue “85 dB at 1 m” is not the level at the ear. In a free field, 6 dB per doubling of distance; in a real room, reverberation fills in the far field and the slope flattens. Convert every source to the same point before you type the row of numbers, or the sum is a mash of geometries.
Never add A-weighted and C-weighted values, and never add levels taken at different distances as if they were the same quantity. The logarithm will still produce a number. The number will not be a measurement.