5D Library › Architectural Acoustics, Reverberation Time and the Environmental Psychology of Built Space · 中文
Chapter 1 Sabine and Eyring: Reverberation Time and Speech Clarity
The “tail” of a room
Clap once in an empty glass meeting room and the sound does not vanish at once; it trails off in a fading “tail”. Clap in a study lined with bookshelves and carpeted, and the tail is so short that it is hard to hear at all. That tail is reverberation. Acoustics describes it by the reverberation time (often written below): the time it takes, after a source stops, for the sound energy in the room to fall by 60 dB, to one millionth.
In the 1890s Wallace Sabine, a young Harvard physicist, was asked to rescue a lecture room in which speech was almost unintelligible. Using an organ pipe as a source and his own ears and a chronograph as the detector, and carrying different numbers of seat cushions in and out, he measured the length of the tail and arrived at a simple empirical rule: reverberation time is proportional to the volume of the room and inversely proportional to the total absorption in it. The work was published in instalments in 1900 under the title “Reverberation” and later collected in his Collected Papers on Acoustics (Sabine 1922). It is still the first lesson of architectural acoustics.
Derivation: Sabine’s formula
Calculation
Energy balance in a diffuse field. Suppose the sound energy in the room is uniformly distributed and all directions of travel are equally likely (a “diffuse” field), with energy density . It can be shown that the energy falling on unit area of wall per unit time is , with the speed of sound. If the room’s surfaces have areas and absorption coefficients , the equivalent absorption area is (in square metres, or “sabins”). After the source stops,
A fall of 60 dB means , so
This is Sabine’s formula. The constant 0.161 is not a magic number but , and it shifts slightly as the speed of sound changes with temperature.
Calculation
Eyring’s correction. Sabine’s formula has an obvious flaw: if every surface absorbed completely (), sound should disappear at once, yet the formula still gives . In 1930 Carl Eyring took a different view: a sound ray meets a wall once in every mean free path , and at each encounter only a fraction of its energy survives, where is the mean absorption coefficient (Eyring 1930). With reflections per second,
For small , and Eyring reduces to Sabine; as , . Both formulas assume a diffuse field and roughly uniform absorption; when the absorption is concentrated on one surface (an absorbent ceiling alone, for example) both can depart from measured values (Kuttruff 2009).
The program below computes three finishes of the same 8 × 6 × 3 m meeting room. Room A has a hard tiled floor, a plastered ceiling, one glass curtain wall and painted concrete elsewhere; room B is room A with 60 % of the ceiling replaced by an absorbent suspended ceiling; room C adds a carpet, a heavy curtain over half of the glass and a full bookshelf along one end wall. The absorption coefficients are typical textbook-style values, and the furniture and six seated people are given a separate equivalent absorption area; all are illustrative.
Calculation
# Sabine T = 0.161 V / A and Eyring T = 0.161 V / (-S ln(1 - abar)) for one 8 x 6 x 3 m
# room, finished two ways. Absorption coefficients are typical textbook-style values
# (illustrative).
import math
c = 343.0 # speed of sound at about 20 C, m/s
print(f"Sabine constant 24 ln(10) / c = {24 * math.log(10) / c:.4f} s/m")
BANDS = (125, 250, 500, 1000, 2000, 4000)
MAT = { # absorption coefficient per octave band
"tile floor": (0.01, 0.01, 0.015, 0.02, 0.02, 0.02),
"plaster ceiling": (0.01, 0.02, 0.02, 0.03, 0.04, 0.05),
"glass wall": (0.18, 0.06, 0.04, 0.03, 0.02, 0.02),
"painted concrete": (0.01, 0.01, 0.02, 0.02, 0.02, 0.02),
"carpet": (0.02, 0.06, 0.14, 0.37, 0.60, 0.65),
"acoustic ceiling": (0.50, 0.70, 0.60, 0.70, 0.70, 0.50),
"heavy curtain": (0.14, 0.35, 0.55, 0.72, 0.70, 0.65),
"book-filled shelf": (0.30, 0.40, 0.40, 0.30, 0.30, 0.20)}
EXTRA = { # equivalent absorption area in m2 per band: furniture plus 6 seated people
"hard": (3.5, 5.0, 6.5, 7.5, 8.0, 8.0),
"soft": (4.5, 6.5, 8.5, 9.5, 10.0, 10.0)}
L, W, H = 8.0, 6.0, 3.0
V, S = L * W * H, 2 * (L * W + L * H + W * H)
ROOMS = {
"A glass + concrete": ([("tile floor", L * W), ("plaster ceiling", L * W),
("glass wall", L * H),
("painted concrete", L * H + 2 * W * H)], "hard"),
"B A + panel ceiling": ([("tile floor", L * W), ("plaster ceiling", 0.4 * L * W),
("acoustic ceiling", 0.6 * L * W), ("glass wall", L * H),
("painted concrete", L * H + 2 * W * H)], "hard"),
"C wood, books, cloth": ([("carpet", L * W), ("plaster ceiling", 0.4 * L * W),
("acoustic ceiling", 0.6 * L * W),
("glass wall", L * H / 2),
("heavy curtain", L * H / 2), ("book-filled shelf", W * H),
("painted concrete", L * H + W * H)], "soft")}
def rt(room):
surfaces, extra = ROOMS[room]
out = []
for k in range(len(BANDS)):
A = sum(MAT[m][k] * a for m, a in surfaces) + EXTRA[extra][k]
abar = A / S
out.append((0.161 * V / A, 0.161 * V / (-S * math.log(1 - abar)), abar))
return out
print(f"room {L:.0f} x {W:.0f} x {H:.0f} m: V = {V:.0f} m3, S = {S:.0f} m2,"
f" mean free path"
f" 4V/S = {4 * V / S:.2f} m")
print("band (Hz) " + "".join(f"{b:>7d}" for b in BANDS) + " mid (500-1k)")
for room in ROOMS:
r = rt(room)
mid = lambda i: (r[2][i] + r[3][i]) / 2
print(f"{room:20s} abar " + "".join(f"{x[2]:7.2f}" for x in r))
print(f"{'':20s} Sab. " + "".join(f"{x[0]:7.2f}" for x in r) + f" {mid(0):.2f} s")
print(f"{'':20s} Eyr. " + "".join(f"{x[1]:7.2f}" for x in r) + f" {mid(1):.2f} s")
print("same room, uniform absorption: Eyring / Sabine ratio")
print(" " + " ".join(f"a={a:.2f}: {-a / math.log(1 - a):.3f}"
for a in (0.05, 0.1, 0.2, 0.4, 0.6)))
# Sabine constant 24 ln(10) / c = 0.1611 s/m
# room 8 x 6 x 3 m: V = 144 m3, S = 180 m2, mean free path 4V/S = 3.20 m
# band (Hz) 125 250 500 1000 2000 4000 mid (500-1k)
# A glass + concrete abar 0.05 0.05 0.06 0.07 0.07 0.07
# Sab. 2.47 2.73 2.24 1.96 1.85 1.78 2.10 s
# Eyr. 2.41 2.67 2.18 1.90 1.78 1.71 2.04 s
# B A + panel ceiling abar 0.13 0.16 0.15 0.17 0.18 0.14
# Sab. 0.99 0.83 0.86 0.75 0.73 0.89 0.80 s
# Eyr. 0.92 0.76 0.79 0.68 0.67 0.83 0.73 s
# C wood, books, cloth abar 0.17 0.24 0.27 0.35 0.41 0.38
# Sab. 0.78 0.55 0.48 0.37 0.31 0.34 0.42 s
# Eyr. 0.71 0.48 0.42 0.30 0.24 0.27 0.36 s
# same room, uniform absorption: Eyring / Sabine ratio
# a=0.05: 0.975 a=0.10: 0.949 a=0.20: 0.896 a=0.40: 0.783 a=0.60: 0.655The hard room A has a mid-frequency (500–1,000 Hz) reverberation time of about 2.1 s, the typical order of magnitude for the glass meeting room that “feels noisy the moment you walk in”; with the absorbent ceiling, room B falls to 0.80 s, and room C to 0.42 s (0.36 s by Eyring). The last line shows that the two formulas diverge as absorption grows: at a mean coefficient of 0.1 the Eyring value is only about 5 % shorter than Sabine’s, at 0.4 about 22 % and at 0.6 about 35 %. In heavily absorbent rooms (studios, a heavily damped study) Sabine’s formula therefore overestimates the reverberation time.
At 125 Hz room C still has 0.78 s, much longer than at mid frequencies: carpet, curtains and books absorb mainly middle and high frequencies, and low frequencies need dedicated low-frequency absorbers. Chapter 2 shows that the low end of a small room needs different physics altogether.
The same numbers carry an often overlooked consequence. In a diffuse field, the reverberant level produced by a given source is (with the sound power level), so it changes only with . Since , the mid-frequency absorption area is about 11.0 m² in room A and about 55.2 m² in room C, so their reverberant levels differ by dB. The same voices, footsteps and scraped chairs “hang in the air” about 7 dB less in room C.
Reverberation and speech: STI and
Reverberation harms speech by dragging the tail of one syllable into the next and flattening the fluctuations of the speech envelope. Houtgast and Steeneken turned this into a physical index, the speech transmission index (STI) (Steeneken and Houtgast 1980). The intensity envelope of speech fluctuates at rates between about 0.5 and 16 Hz, and both reverberation and background noise reduce the “modulation depth” of these fluctuations.
Calculation
The modulation transfer function. For an exponentially decaying diffuse field, the modulation transfer ratio at modulation frequency is
where the first factor is the loss from reverberation () and the second the loss from the signal-to-noise ratio SNR. Each is converted to an “apparent SNR” , clipped to ±15 dB and mapped linearly onto 0–1; these are averaged over 14 modulation frequencies (0.63–12.5 Hz) and combined over 7 octave bands (125 Hz–8 kHz) with weights to give the STI. The program uses the male octave weights and redundancy corrections of IEC 60268-16 from the 2011 edition onwards (International Electrotechnical Commission 2011), but omits the auditory masking and threshold corrections and ignores direct sound; it is a simplified, textbook version.
Clarity is the ratio (in dB) of the energy arriving in the first 50 ms to the energy arriving later. For a pure exponential decay, .
Calculation
# Speech Transmission Index from the modulation transfer function of a diffuse room
# (exponential decay) plus steady noise:
# m(F) = 1 / sqrt(1 + (2 pi F T / 13.8)^2) / (1 + 10^(-SNR/10)).
# Octave weights alpha/beta as in IEC 60268-16 (2011+, male); no auditory masking or
# threshold corrections, no direct sound: a simplified textbook STI. Clarity C50 for an
# exponential decay.
import math
FM = [10 ** (k / 10) for k in range(-2, 12)] # 14 modulation freqs, 0.63-12.5
ALPHA = (0.085, 0.127, 0.230, 0.233, 0.309, 0.224, 0.173) # 125 Hz ... 8 kHz
BETA = (0.085, 0.078, 0.065, 0.011, 0.047, 0.095)
def mti(T, snr):
tis = []
for F in FM:
m = 1 / math.sqrt(1 + (2 * math.pi * F * T / 13.8) ** 2) / (1 + 10 ** (-snr / 10))
snr_app = max(-15.0, min(15.0, 10 * math.log10(m / (1 - m)))) if m < 1 else 15.0
tis.append((snr_app + 15) / 30)
return sum(tis) / len(tis)
def sti(Ts, snrs):
M = [mti(T, s) for T, s in zip(Ts, snrs)]
return (sum(a * m for a, m in zip(ALPHA, M))
- sum(b * math.sqrt(M[k] * M[k + 1]) for k, b in enumerate(BETA)))
def c50(T): # early (0-50 ms) / late energy, dB
return round(10 * math.log10(math.exp(13.8 * 0.05 / T) - 1), 1) + 0.0
LABEL = [(0.30, "bad"), (0.45, "poor"), (0.60, "fair"), (0.75, "good"),
(1.01, "excellent")]
word = lambda x: next(w for t, w in LABEL if x < t)
print(" T (s) | C50 (dB) | STI at SNR +30 | +15 dB | +10 dB | +5 dB | 0 dB")
for T in (0.3, 0.4, 0.5, 0.6, 0.8, 1.0, 1.5, 2.0, 3.0):
row = [sti([T] * 7, [s] * 7) for s in (30, 15, 10, 5, 0)]
print(f" {T:3.1f} | {c50(T):+5.1f} | {row[0]:.2f} | "
+ " | ".join(f"{x:.2f}" for x in row[1:]))
ROOMS = { # Sabine T per band of the three rooms of p01; 8 kHz taken equal to 4 kHz
"A glass + concrete": (2.47, 2.73, 2.24, 1.96, 1.85, 1.78, 1.78),
"B A + panel ceiling": (0.99, 0.83, 0.86, 0.75, 0.73, 0.89, 0.89),
"C wood, books, cloth": (0.78, 0.55, 0.48, 0.37, 0.31, 0.34, 0.34)}
for name, Ts in ROOMS.items():
s15 = sti(Ts, [15] * 7)
print(f"{name:20s}: STI {s15:.2f} ({word(s15)}) at SNR 15 dB,"
f" {sti(Ts, [5] * 7):.2f} at 5 dB; C50(1k) {c50(Ts[3]):+.1f} dB")
# T (s) | C50 (dB) | STI at SNR +30 | +15 dB | +10 dB | +5 dB | 0 dB
# 0.3 | +9.5 | 0.83 | 0.79 | 0.71 | 0.60 | 0.46
# 0.4 | +6.6 | 0.78 | 0.74 | 0.67 | 0.57 | 0.44
# 0.5 | +4.7 | 0.74 | 0.70 | 0.64 | 0.55 | 0.42
# 0.6 | +3.3 | 0.70 | 0.67 | 0.61 | 0.53 | 0.41
# 0.8 | +1.4 | 0.64 | 0.61 | 0.57 | 0.49 | 0.38
# 1.0 | +0.0 | 0.59 | 0.57 | 0.53 | 0.46 | 0.36
# 1.5 | -2.3 | 0.50 | 0.48 | 0.46 | 0.40 | 0.31
# 2.0 | -3.9 | 0.44 | 0.43 | 0.41 | 0.36 | 0.27
# 3.0 | -5.9 | 0.36 | 0.35 | 0.33 | 0.30 | 0.22
# A glass + concrete : STI 0.43 (poor) at SNR 15 dB, 0.36 at 5 dB; C50(1k) -3.7 dB
# B A + panel ceiling : STI 0.61 (good) at SNR 15 dB, 0.49 at 5 dB; C50(1k) +1.8 dB
# C wood, books, cloth: STI 0.75 (excellent) at SNR 15 dB, 0.57 at 5 dB; C50(1k) +7.4 dBAt an SNR of 15 dB, a reverberation time of 0.5 s gives an STI of 0.70, 1.0 s gives 0.57 and 2.0 s 0.43, while falls from +4.7 dB to about 0 dB and then to −3.9 dB. The three meeting rooms reach STIs of 0.43 (“poor”), 0.61 (“good”) and 0.75 (“excellent”) at 15 dB SNR; with louder background noise and an SNR of only 5 dB they fall to 0.36, 0.49 and 0.57. Noise and reverberation are two independent losses, and reducing one cannot fully make up for the other.
“Optimal reverberation”: standards, conventions and an unsupported number
The author’s framework stated that the optimal reverberation is 0.4–0.6 s for lecture rooms and focused work, and 1.5–2.2 s for concert halls and spaces for meditative chant. On checking, these need to be separated.
Evidence
An upper limit for speech rooms. The American standard ANSI/ASA S12.60-2010 requires that furnished, unoccupied classrooms of up to 283 m³ (10,000 cubic feet) have a mid-frequency reverberation time of no more than 0.6 s, those of 283–566 m³ no more than 0.7 s, with background noise no higher than 35 dB(A) (Acoustical Society of America 2010). This is an upper limit, not an “optimum”.
The program above also shows that, in the simplified model, STI keeps rising as reverberation time falls; it does not peak at 0.4 s. Physically, the cost of making a room too “dry” lies elsewhere: reverberant sound contributes to the speech level for distant listeners, so with too much absorption speech becomes weaker at the back, and talkers lose the “support” of the room and have to push harder. None of this is captured by the STI.
Evidence
Rooms for music. Music needs some reverberation to blend voices and give a sense of fullness. Beranek’s measurements and evaluations of concert halls around the world show that occupied symphony halls mostly have mid-frequency reverberation times around 2 s, with opera houses shorter (Beranek 2004). These figures come from design convention and the judgements of listeners and critics, and they vary with the type of music, the size of the hall and occupancy.
Note
Correction to the author’s framework. “0.4–0.6 s” is a reasonable design target for speech rooms, but it should be stated as “no more than about 0.6 s, depending on volume”, not as an experimentally established optimum. No research basis was found for “1.5–2.2 s for spaces of meditative chant”: many traditional chant spaces (stone churches, temple halls) have much longer reverberation, and preferences depend mainly on the chant tradition and taste; this book sets no numerical target for such spaces. Whether a room makes people “calm the moment they enter, or irritable and tired” is a fair question, but no controlled study was found that links reverberation time directly to a sense of calm; it is stated below as a hypothesis.
Hypothesis · not established
Moderately absorbent workspaces reduce perceived “noisiness” and fatigue mainly by lowering the reverberant level rather than by raising STI (not established). A testable consequence: in one room fitted with movable absorbers to give = 0.4, 0.8 and 1.6 s, with the sources of activity noise held constant, ratings of noisiness and of fatigue after half a day should change monotonically with reverberant level (not with STI). If ratings still differ between conditions with the same reverberant level but different STI, or do not differ across the three conditions at all, the hypothesis is weakened.
Chapter summary
Sabine’s formula follows from an energy balance in a diffuse field, the constant 0.161 being ; Eyring’s formula keeps of the energy at each reflection, and the more absorption, the more the two differ (about 22 % shorter for Eyring at a mean coefficient of 0.4). In the illustrative calculation, one meeting room falls from about 2.1 s in glass and concrete to 0.80 s with an absorbent ceiling and 0.42 s with books and fabrics, its STI rising from 0.43 to 0.75 and its reverberant level falling by about 7 dB. “About 0.6 s” for speech rooms is a standard’s upper limit rather than an experimental optimum; about 2 s for concert halls is design convention; “1.5–2.2 s for meditation spaces” has no basis. The next chapter turns to where reverberation theory breaks down: low frequencies in small rooms.
References
Fan, Y. (2026). Architectural Acoustics, Reverberation Time and the Environmental Psychology of Built Space: From Sabine's Equation and Fractal Visual Dimensions to Cognitive Performance, Read Beside Classical Chinese Spatial Design [建筑声学、混响时间与空间环境心理学的物理模型] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23782090 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).