🎤 Public Speaking Acoustics: Reverberation, Critical Distance and Speech Intelligibility

2D interactive simulation · educational model
The science

Reverberation. Sound that hits the walls is partly absorbed, so the diffuse field decays exponentially. The Sabine reverberation time (60 dB decay) is RT60 = 0.161 V / A, with room volume V (m³) and total absorption A = Σ Siαi (m²); Eyring's variant −0.161 V / (S ln(1−ᾱ)) is better for very absorbent rooms and is shown beside it. The decay constant of the energy is τ = RT60/13.8 because 13.8 = ln(106). People absorb about 0.45 m² each; the simulator assumes 1.5 people per m² on 60% of the floor.

Direct plus reverberant sound. A speaker of power level LW and directivity Q (2 on a stage in front of a wall) produces at distance r the direct level Ld = LW + 10 log(Q/4πr²) and the nearly uniform reverberant level Lr = LW + 10 log(4/R), with room constant R = A/(1−ᾱ). They are equal at the critical distance rc = √(QR/16π): beyond it extra distance no longer lowers the level, but intelligibility keeps falling because reverberation smears syllables.

Simplified STI. Speech is amplitude-modulated at 0.6–12.5 Hz. The room reduces the modulation depth to m(F) = |Ed + Er/(1 + j2πFτ)| / (Ed+Er) and background noise N reduces it further by 1/(1 + 10−SNR/10). The apparent signal-to-noise ratio 10 log(m/(1−m)) is clipped to ±15 dB, mapped to 0–1 and averaged over 14 modulation frequencies. Ratings: <0.30 bad, 0.30–0.45 poor, 0.45–0.60 fair, 0.60–0.75 good, >0.75 excellent.

Honest simplifications. One broadband value instead of seven octave bands with masking and gender weighting; perfectly diffuse reverberant field (no echoes, flutter or focusing); no loudspeaker system; Q is constant, voice level is a single number; a real STI measurement or acoustic software is needed for design. The result is a teaching tool for why long reverberation and noise make talks hard to follow.