Room acoustics in practice

A loudspeaker never operates alone; it drives a room. After the direct sound leaves the source, walls, floor, ceiling, stage, furniture and audience produce delayed, filtered copies. The measured and perceived result is therefore a coupled loudspeaker-room-listener system.

1. Direct and reflected sound

The listener receives direct sound first, followed by early reflections and then a progressively denser reverberant field. The time structure can be inspected in an impulse response or ETC (Energy Time Curve). A reflection delayed by Δt has an extra path length approximately:

Symbols: Δd is path-length difference in metres, c sound speed in metres per second and Δt time difference in seconds. For example, 1 ms corresponds to about 0.343 m near 20 °C.

This simple relation is powerful when identifying wall, floor or ceiling reflections.

2. Sabine equation - a first approximation

For a sufficiently diffuse reverberant field with moderate average absorption:

Symbols: RT₆₀ is 60 dB decay time in seconds, V room volume in cubic metres and A equivalent absorption area in sabins. The Sabine approximation assumes a diffuse field and relatively low average absorption.

where V is room volume and A the equivalent absorption area. Sabine becomes inaccurate in strongly damped spaces or when the field is not diffuse. Eyring’s model behaves better at high average absorption:

Symbols: S is total boundary area, ᾱ mean absorption coefficient and ln the natural logarithm. Eyring’s equation is often more suitable at higher absorption, but it remains an approximate diffuse-field model.

S is total boundary area and ᾱ average absorption coefficient.

3. Room modes

In a rectangular room, modal frequencies can be approximated by:

Symbols: f_nmp is a mode frequency of a rectangular room, c sound speed, L_x, L_y, L_z room dimensions and n, m, p non-negative integer mode indices that cannot all be zero.

Axial modes involve one dimension and tend to be strongest; tangential modes involve two dimensions; oblique modes involve all three. A useful simplified energy hierarchy is that tangential modes can be roughly 3 dB lower and oblique roughly 6 dB lower than comparable axial modes, although real rooms depart from ideal theory.

4. Why we do not EQ a deep spatial null

A cancellation caused by two paths can remain a cancellation after EQ because both paths are amplified together. A 20 dB boost into a geometric null wastes headroom and may overload the system while barely changing that position. Move the source/listener, change boundary distance, alter multiple-source geometry or use treatment instead.

5. Early reflections and comb filtering

Strong early reflections create frequency-dependent interference and can blur localisation and speech. The audibility depends on delay, level, direction and spectral content. Reflection control is therefore not simply “make everything dead”; useful diffusion and later energy can support spaciousness while strong early reflections may need attenuation or redirection.

6. Absorption and diffusion

Porous absorbers dissipate particle motion and are most effective where particle velocity is high. Against a rigid boundary, pressure is maximum and particle velocity minimum, so deep LF absorption requires thickness or spacing approaching a useful fraction of wavelength. Resonant absorbers - membrane/panel or Helmholtz types - can exploit pressure maxima at boundaries and target a narrower low-frequency range with less physical depth.

7. Critical distance

Critical distance is where direct and reverberant sound energies are roughly equal. A useful approximation involving source directivity Q and room properties is:

Symbols: r_c is critical distance in metres, Q_d source directivity factor, V volume, RT₆₀ decay time and A equivalent absorption area. Here Q_d is not quality factor; the relation is a diffuse-field approximation.

More directional loudspeakers can push critical distance farther into a reverberant room, improving intelligibility. Delay loudspeakers exploit the precedence/Haas effect: with suitable arrival-time and level relationships, localisation remains attached to the earlier main source while the delayed system increases useful direct level.

8. Measurement time window and frequency resolution

A short time window suppresses later reflections but reduces low-frequency resolution; a long window improves LF frequency resolution but includes more room energy. The uncertainty relation between time and frequency resolution means there is no single “best” window for every question.

9. RT60, EDT, C50/C80 and STI

RT60 describes decay time; EDT emphasises the initial decay and can correlate better with perceived liveliness. C50 and C80 compare early to late energy for speech and music. STI estimates speech-transmission quality. Each metric answers a different question, so none is a universal room-quality number.

10. In event sound, the audience is an acoustic element

People absorb and scatter sound, especially in the mid/high range, and physically block paths. An empty-room tuning can change substantially once an audience enters. Measurement and system design should anticipate the occupied condition rather than optimising an empty venue in isolation.

11. Schroeder frequency

Below the Schroeder region, individual modes dominate; above it, the modal density is high enough for more statistical room-acoustic descriptions to become useful. A common estimate is:

Symbols: f_s is Schroeder frequency in hertz, RT₆₀ is in seconds and V in cubic metres. Below it individual room modes dominate; above it a statistical description becomes more useful, with a gradual rather than absolute transition.

This boundary is approximate but helps explain why bass treatment and placement are mode-specific while higher-frequency decay can be discussed more statistically.

12. Boundary interference - SBIR

A loudspeaker near a wall produces a delayed reflected copy. For a simple rear-wall geometry with source distance d, the first cancellation is approximately:

Symbols: f_null is the approximate first quarter-wave cancellation frequency, c sound speed and d perpendicular distance from source to a reflecting boundary. This is a simplified model of one rigid planar boundary.

For d=0.86 m, this is about 100 Hz. Moving the source closer to or farther from the boundary shifts the cancellation. EQ cannot remove the underlying path difference.

13. Measurement-microphone placement

One microphone position describes one point, not a room. For audience systems, measure representative areas and heights, keeping reference geometry consistent. Spatial averaging is useful when the design goal is robust coverage rather than a perfect trace at one seat.

14. What EQ can and cannot do

EQ is excellent for correcting minimum-phase tonal deviations and shaping a system’s target response. It cannot make a long reverberation time short, remove a geometric reflection, repair a deep spatial cancellation everywhere, or make poor directivity uniform. Treatment, placement, aiming, delay and source geometry solve different classes of problems.

15. Bass traps - why thickness matters

At a rigid wall, particle velocity tends toward zero while pressure is high. Porous absorbers need depth/air gap to reach regions of larger particle velocity at low frequency, which is why thin foam does little for deep bass. Resonant traps respond to pressure and can target modal frequencies more compactly, at the cost of narrower bandwidth and greater tuning sensitivity.

16. Stage, DJ booth and microphones

Stage surfaces and DJ booths create local reflections and cavity effects. Open microphones add gain-before-feedback constraints; their polar patterns interact with monitor placement and room reflections. Moving a microphone or monitor by tens of centimetres can matter more than several dB of EQ.

17. The audience area is not one point

A system should be evaluated spatially: front/back, left/right, near/far and representative height. A beautiful single-position response is not the objective if most listeners receive different tonal balance or timing. Good room/system engineering optimises the distribution, not the screenshot.