Sabine's Equation: The Physics of How a Room Sounds

Reverberation as an energy reservoir that leaks

Strike a note in a room and the wavefront doesn't reach your ear once — it reaches it thousands of times, each arrival a copy that has bounced off walls, ceiling, and floor. In a hall a few tens of metres across, sound (traveling at c ≈ 343 m/s in 20 °C air) crosses the room in tens of milliseconds, so within a second a single handclap has undergone hundreds of reflections. After the source stops, this dense hail of overlapping echoes — the reverberant field — decays smoothly rather than as distinct echoes.

Sabine's insight was to stop tracking individual rays and treat the acoustic energy density E as a single quantity that drains from a leaky reservoir. Each time the field strikes a surface of area S and absorption coefficient α, a fraction α of the incident energy is lost. Averaged over the whole boundary, the energy obeys an exponential decay:

• E(t) = E₀ · e^−t/τ, where τ is the acoustic time constant.
• In the diffuse-field approximation the mean rate of surface encounters is c·A/(4V), giving τ = 4V/(cA).
• Here A = Σ Sᵢαᵢ is the total absorption in sabins (units of m²), summing every surface times its absorption coefficient.

The factor of 4 is not arbitrary: it comes from integrating the flux of an isotropic (diffuse) sound field onto a flat surface, ⟨cos θ⟩ over a hemisphere giving the mean free path ℓ = 4V/S — a result Sabine's contemporaries later derived rigorously from geometry.

Deriving RT₆₀ = 0.161·V/A

Reverberation time RT₆₀ is defined as the interval over which the sound-pressure level falls by 60 dB — a factor of 10⁶ in energy, chosen because it spans from a fortissimo orchestral chord (~100 dB) down to a quiet room's noise floor (~40 dB). We just need the time for E to drop by 10⁶.

• Set the decay ratio: E(RT)/E₀ = 10⁻⁶ = e^−RT/τ.
• Take logs: RT/τ = 6·ln(10) = 13.816.
• Substitute τ = 4V/(cA): RT₆₀ = (6 ln10)·4V/(cA) = 24·ln(10)·V/(cA).
• Plug in c = 343 m/s: the prefactor is 24·(2.3026)/343 = 0.1611 s·m⁻¹.

Hence the celebrated result: RT₆₀ = 0.161·V/A, with V in m³ and A in m² (sabins). In imperial units the constant becomes 0.049 (feet). Everything hard about acoustics is buried in one honest place — the absorption A — while the geometry contributes only the room's volume. That separation is exactly why the law is so useful: double the volume and you double the reverberation; double the carpet and drapery and you halve it.

A quick sanity check on Sabine's own numbers: Boston Symphony Hall holds V ≈ 18,750 m³. To land RT₆₀ ≈ 1.8 s, the formula demands A = 0.161·18750/1.8 ≈ 1,680 sabins — the combined absorption of plaster, wood, and (crucially) a full audience, since each seated listener contributes roughly 0.4–0.5 sabins at mid frequencies.

What controls the decay: volume, area, and α

The absorption coefficient α is the single most important material property in the room, running from 0 (perfect reflector) to 1 (perfect absorber, e.g. an open window that never returns the sound). It is strongly frequency-dependent, which is why a good hall is designed band-by-band across the octaves from 125 Hz to 4 kHz.

• Painted concrete / plaster: α ≈ 0.02 — nearly perfect mirror; a bare concrete stairwell rings for seconds.
• Heavy carpet on felt: α ≈ 0.05 (125 Hz) rising to 0.6 (4 kHz) — absorbs treble, leaves bass.
• Thick fiberglass / rockwool panel: α ≈ 0.9–1.0 across mids and highs — the workhorse of studios.
• One adult listener: ≈ 0.4–0.5 sabins each — an audience is often the biggest absorber in a full hall, which is why empty and full halls sound different.

Because A is a sum of Sᵢαᵢ, the strategy for tuning a room is additive bookkeeping: you tally each surface's area times its coefficient, add the people, and read off RT₆₀. A studio designer chasing a dead RT₆₀ ≈ 0.3 s in a 100 m³ room needs A = 0.161·100/0.3 ≈ 54 sabins — perhaps 60 m² of near-total absorption, i.e. wall-to-wall treatment. A cathedral, by contrast, pairs a colossal V (tens of thousands of m³) with hard stone (tiny α), yielding the famous 6–10 s tails that let Gregorian chant bloom.

Where Sabine breaks and Eyring takes over

Sabine's formula has a glaring flaw at its own limit. If every surface were a perfect absorber (ᾱ = 1, an anechoic chamber), physics says a sound must vanish after essentially one reflection — RT₆₀ should approach zero. But plug ᾱ = 1 into A = Sᾱ and Sabine returns RT₆₀ = 0.161V/S, a stubbornly finite number. The law over-predicts reverberation whenever rooms are highly absorptive.

Carl Eyring (Bell Labs, 1930) fixed this by treating absorption per reflection rather than as a continuous leak. If the field survives each bounce with probability (1 − ᾱ) and reflections occur every ℓ/c = 4V/(cS) seconds, the surviving energy after n bounces is (1 − ᾱ)ⁿ, which is exponential in t with the corrected time constant. The result:

• RT₆₀ = 0.161·V / (−S·ln(1 − ᾱ)), where ᾱ = A/S is the mean absorption coefficient.
• For small ᾱ, −ln(1 − ᾱ) ≈ ᾱ, so −S·ln(1−ᾱ) ≈ Sᾱ = A and Eyring collapses back to Sabine — they agree in live rooms.
• At ᾱ = 1, −ln(1−ᾱ) → ∞, so RT₆₀ → 0, correctly matching the anechoic limit.

The practical rule: use Sabine when ᾱ ≲ 0.2 (concert halls, churches, classrooms), where its error is a few percent, and switch to Eyring for absorptive spaces like studios and treated offices, where Sabine can overshoot by 30% or more. Both assume a perfectly diffuse, homogeneous field — the assumption that ultimately limits them.

Measuring RT₆₀ and the diffuse-field assumption

You rarely wait for a true 60 dB drop — background noise usually swallows the last 20 dB. The ISO 3382 standard instead fits the initial slope of the decay curve and extrapolates: measure the time for a −20 dB or −30 dB fall (the T20 and T30 metrics) and scale up to 60 dB. Schroeder's 1965 backward integration of an impulse response turned a noisy, jagged decay into a smooth, repeatable curve, and remains the standard method — fire a starter pistol or play a swept sine, record the room's impulse response, integrate it backward in time, and read the slope.

Every version of the theory leans on the diffuse-field assumption: sound energy uniform in space and arriving equally from all directions. Real rooms violate it. A shoebox hall with parallel hard walls sets up flutter echoes and standing-wave modes; a room much longer than it is wide has a decay that varies with position. Below the Schroeder frequency — roughly f_S ≈ 2000·√(RT₆₀/V) Hz — the modal density is too low for statistics to apply, and the room behaves as a set of discrete normal modes rather than a diffuse continuum. For a 200 m³ room with RT₆₀ = 1 s, f_S ≈ 140 Hz, below which every note excites individual room resonances and Sabine's smooth exponential no longer describes the bass.

From symphony halls to Zoom calls

Sabine's law is the quiet backbone of every space designed for the ear. The optimum RT₆₀ depends on purpose: speech intelligibility wants short tails (a classroom or courtroom targets 0.6–0.8 s so consonants don't smear), while symphonic music wants a rich 1.8–2.2 s to fuse the orchestra and add warmth. Opera sits in between (~1.4 s) so singers' words stay crisp; a recording studio's live room might be 0.4 s, its vocal booth 0.2 s.

• Boston Symphony Hall (1900): the first hall designed with Sabine's formula in hand, RT₆₀ ≈ 1.8 s full — still ranked among the world's best.
• Reverberation chambers: deliberately hard, non-parallel rooms with RT₆₀ up to 10 s, used to measure α of materials by running Sabine's equation in reverse: install a sample, watch RT₆₀ drop, solve for A.
• Open-plan offices & restaurants: excessive RT₆₀ is the physics of a room that is exhausting to talk in; acoustic ceiling tiles are absorption engineering, adding sabins to shorten the tail.
• Digital reverb & VR audio: convolution reverbs and game engines synthesize a room's decay from an RT₆₀ target and a modeled diffuse tail — Sabine's exponential is the seed of the algorithm.

The same one-line equation that saved a Harvard lecture hall now sizes the absorption in a podcast studio and calibrates the reverb slider in your headphones.

Sabine vs. Eyring absorption models and where each holds

Property	Sabine (1898)	Eyring–Norris (1930)
Formula	RT₆₀ = 0.161V/(Sᾱ)	RT₆₀ = 0.161V/(−S·ln(1−ᾱ))
Absorption regime	Live rooms, ᾱ ≲ 0.2	Dead/absorptive rooms, ᾱ → 1
ᾱ = 1 (anechoic)	Predicts finite RT (wrong)	Predicts RT → 0 (correct)
Physical picture	Continuous absorption	Per-reflection survival (1−ᾱ)
Error at ᾱ = 0.5	~40% too long	Near-exact
Sound speed used	343 m/s (20 °C air)	343 m/s (20 °C air)