See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/11599718

Mechanism of sound absorption by seated audience in halls

Article  in  The Journal of the Acoustical Society of America · November 2001

DOI: 10.1121/1.1409970 · Source: PubMed

CITATIONS READS
31 4,736

3 authors:

Noriko Nishihara Takayuki Hidaka

Takenaka R&D, Japan Takenaka Research & Development Institute
16 PUBLICATIONS   219 CITATIONS    24 PUBLICATIONS   533 CITATIONS   

SEE PROFILE SEE PROFILE

Leo L. Beranek
Institute of Electrical and Electronics Engineers
161 PUBLICATIONS   4,079 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Leo L. Beranek on 21 July 2015.

The user has requested enhancement of the downloaded file.

Mechanism of sound absorption by seated audience in halls
Noriko Nishiharaa) and Takayuki Hidaka
Takenaka Research & Development Institute, 1-5-1, Otsuka, Inzai, Chiba 270-1395, Japan

Leo L. Beranek
975 Memorial Drive, #804, Cambridge, Massachusetts 02138
共Received 16 October 2000; revised 10 August 2001; accepted 15 August 2001兲
Four methods are explored for predicting the reverberation times in fully occupied halls for music
as related to the sound absorption by their audiences. The methods for providing audience
absorptions include two that use reverberation chambers, namely, the ISO 354 method 共and other
similar standards兲 共ISO兲 and Kath and Kuhl’s method 共K & K兲 关Acustica 15, 127–131 共1965兲兴, and
two that use average data from halls, i.e., Beranek’s method 共COH兲 关Concert and Opera Halls: How
They Sound 共Acoustical Society of America, Melville, NY, 1996兲兴, and the average audience
power-per-seat absorption which in practice is multiplied by the number of seats (⌬A). These
methods are applied to the calculation of reverberation times in six existing halls, fully occupied,
and the results were compared with actual measurements. The COH method was best for predictions
over the entire frequency range. The K & K method showed the highest accuracy at
mid-frequencies. Both the ISO and the K & K methods yielded wide differences for the
measurements in the 125- and 250-Hz bands. The ⌬A method was as good as the COH method
when the measurements for the six halls were averaged, but showed a wide spread in the predictions
around the average because it does not consider the degree of upholstering of the seats. It was
hypothecated by the authors that the principal reasons for the ISO and K & K discrepancies at low
frequencies were 共a兲 differences between the degree of sound diffusion in actual halls and that in
reverberation chambers, and 共b兲 lack of information on the mechanisms of absorption of sound by
people seated side-by-side in rows, particularly for near-grazing incidence sound fields. First, this
article explores the sound diffusivity in a reverberation chamber and in the halls using CAD models.
A probability density function of the incident angles of the sound rays that impinge on the audiences
is defined and was measured for each case. Using a unique method, the sound absorption coefficient
of each portion of the body and chair in a seated audience was determined in an anechoic chamber
as a function of the incident angle of a sound wave. With adjustments from these findings, the K &
K method can be made to equal the COH method in accuracy at all frequencies. Its forte is that it
can be used for the determination of the sound absorption of occupied chairs from measurements of
a limited number in a reverberation chamber. © 2001 Acoustical Society of America.
关DOI: 10.1121/1.1409970兴
PACS numbers: 43.55.Br, 43.55.Fw, 43.55.Gx, 43.55.Mc 关JDQ兴

I. INTRODUCTION limitation of each prediction method with an eye toward its

Several studies on sound absorption by seated audiences
have been published. Included are Kath and Kuhl 共1964,
1965兲, Davis et al. 共1994兲, and Bradley 共1992, 1996兲. They
examined several approaches for more precise prediction of
reverberation time 共RT兲 in halls by using the results of audi-
ence absorption measurements in reverberant chambers. In
general, they have concluded that the reverberation times in
an actual occupied hall cannot be predicted satisfactorily
from absorption coefficients measured in a reverberation
chamber. Beranek 共1962, 1996兲 and Beranek and Hidaka
共1998兲 proposed a method for predicting the reverberation
times in occupied halls 共with Sabine’s formula兲 in which the
absorption coefficients of audiences were based on measure-
ments of reverberation times made in many halls.
This study attempts to make clear the accuracy and the
a兲
Author to whom correspondence should be addressed; electronic mail:
nishihara.noriko@takenaka.co.jp
use in new halls of modern shapes and/or audience chairs of
new designs. In other words, it is intended 共a兲 to search for
and establish an appropriate method to predict reverberation
times in occupied halls, with focus on the accuracy of the
audience absorption coefficients used in the above methods,
and 共b兲 to determine the reasons for the differences in the
absorption coefficients measured in reverberation chambers
and those measured in actual halls.
II. COMPARISON OF RT PREDICTION METHODS
The residual absorption, which is defined as that for a
room, separate from the audience, orchestra, and any
strongly absorbing materials 共Beranek and Hidaka, 1998兲,
generally contributes to the total absorption power of a hall
about 20% to 30% at mid- and high frequencies. Therefore,
even admitting ⫾3% error in the determination of residual
surface areas, such inaccuracy would not seriously affect the
accuracy of predictions of RT in occupied halls. On the other
hand, measurements show that of the total sound absorption

2398 J. Acoust. Soc. Am. 110 (5), Pt. 1, Nov. 2001 0001-4966/2001/110(5)/2398/14/$18.00 © 2001 Acoustical Society of America

FIG. 1. Average ratio of the absorption power of the audience to the total
absorption power in each of the four halls, Hamarikyu-Asahi Concert Hall,
Kanagawa Auditorium, Mitaka Concert Hall, and Tokyo Opera City Concert
Hall. The audience absorption power was derived from reverberation-time
measurements made before and after installation of chairs 共occupied兲. The
halls were completely finished before the seats were installed and all debris
was removed.
in a hall, 60% to 80% is due to the audience, so that its
contribution to RT predominates 共Fig. 1兲. Hence, the remain-
ing contribution is caused by air absorption, which is only
important at high frequencies.
Audience absorption coefficients are usually determined
from either 共a兲 measurements made in a reverberation cham-
ber involving a few occupied seats or 共b兲 measurements
made in an actual occupied hall. In this study, four methods
for determining audience absorption coefficients for use in
halls for music are compared. For either 共a兲 or 共b兲 the sound
absorption power for the audience as a whole 共where air
absorption can be neglected or subtracted兲 has been deter-
mined by
a occup⫽ 共 0.161V/T occup⫺兲 ⫺a other , 共1兲
where a other is determined from the same equation as that
within the parentheses but with T being that measured before
the seats are installed. Each ‘‘a’’ is the product of an area and
an average sound absorption coefficient. Obviously, the ab-
sorption power must be determined at each frequency of in-
terest.
A. Four methods for determining sound absorption
coefficients for seated audiences
1. The ISO method
The standards relating to the measurement of absorption
coefficients 共or powers兲 of the seats, occupied or unoccupied,
in reverberation chambers 关ISO 354 共1983兲, ASTM C423-
90a 共1990兲, DIN 52212 共1961兲, BS 3638 共1987兲, and JIS A
1409 共1998兲兴 stipulate that a group of them be placed near
the center of the floor, 1 m or more away from each wall, and
that room diffusers or microphones be placed in somewhat
the same way as they are in an actual hall. A barrier around
the seating group is recommended in some cases.
J. Acoust. Soc. Am., Vol. 110, No. 5, Pt. 1, Nov. 2001
FIG. 2. Schematic seating array used in the Kath and Kuhl method. When
the seating block is expressed as L⫻W⫻H, the equivalent areas of edges,
S ⬘p , S s⬘ , and S ⬘f , exposed to the sound field in the reverberant chamber are
defined as Eq. 共2兲 in order to compensate for the Waterhouse effect.
The reverberant chamber of Takenaka R&D, which was
used for this research, has seven walls with no parallel sur-
faces and has a volume of 332.8 m3 and a surface area of
289.2 m2, in which 20 curved diffusers, each with an area of
1.8 m2, are randomly suspended. Based on recommendations
in the standards, 20 seats 共seating arrays of 5 seats⫻4 rows
with row-to-row spacing of 950 mm兲 chosen from those ac-
tually used in the six halls1 were installed at the center of the
chamber. During measurements, all five surfaces of these
seating blocks were exposed to the sound field, and Eq. 共1兲 is
used to calculate the absorption coefficients, using as the area
that area over which the seat block is located, S p . When the
absorption coefficients determined by this method are used
to predict reverberation times in an actual hall, the hall area
chosen is also that which lay directly beneath the seating
blocks—no edge correction is applied.
2. K & K method (Kath and Kuhl, 1964, 1965)
This method assumes the seated audience to be a
rectangular-parallelepiped, solid absorber, covering an area
S p . The seats were installed in a 90-degree corner of the
reverberation chamber, in order to minimize the edge effect,
and the top, side, and front part of the solid block were given
the sound absorption coefficients ␣ p , ␣ s , and ␣ f , respec-
tively 共Fig. 2兲. As shown, ␭/8-wide strips are added to the
edge surfaces at the walls to compensate for the additional
absorption by the Waterhouse effect, where ␭ is wavelength
共Waterhouse, 1955兲. Then, the equivalent area of each edge
for the calculation is expressed as follows:
S ⬘p ⫽S p ⫹ 共 L⫹W 兲 ⫻␭/8⫹ 共 ␭/8兲 2 共 top兲
S s⬘ ⫽S s ⫹␭/8⫻1.2 m 共 side兲 共2兲
S ⬘f ⫽S f ⫹␭/8⫻1.2 m 共 front兲 .
During the measurements, the edges of the seating array
on the two sides that were exposed to the sound field were
surrounded by stiff wooden barriers 共1.2 m high, equal to the
height of the head of a seated person兲. The absorption coef-
ficients, ␣ p , ␣ s , and ␣ f , were determined by using simul-
taneous equations and removing the barrier from S s or S f as
appropriate, for five arrangements of seating arrays 共5 chairs
Nishihara et al.: Seated audience in halls 2399
 TABLE I. Absorption powers of audience per person obtained from survey of 20 acoustical textbooks. Their
median value was used for the ⌬A method.
125
Lord and Templeton 共1996兲 0.33
Furrer 共1964兲 0.20
Kimura 共1986兲 0.20
Nagata 共1991兲 0.25
Median 0.23
⫻4 rows, 4 chairs⫻5 rows, 3 chairs⫻6 rows, 6 chairs⫻3
rows, and staggered 5 chairs⫻4 rows兲. The wooden barrier
utilized in this study is made of two layers of 12-mm-thick
plywood nailed to a wooden frame and reinforced with
wooden battens forming squares 40⫻40 cm2. From the pre-
liminary measurement in the reverberant chamber, it was
concluded that the absorption power of this barrier is less
than 1 m2, which has little influence on the results discussed
in the next section.
When an absorption coefficient at a particular frequency
determined by the K & K method is used to calculate the
absorption power of the audience in a real hall, the areas of
the exposed edges are determined accurately from the archi-
tectural drawings. The following principles for obtaining that
total area are applied:
共i兲 For edges adjacent to balcony fronts or walls, sound
absorption by them was disregarded in the calcula-
tion.
共ii兲 The side edges 共of two seating blocks兲 that face each
other, e.g., at an aisle, are treated as one edge as
though they are combined.
共iii兲 The absorption by the other edges, S s and S f , is de-
termined by first taking their length and multiplying it
by the height 共1.2 m兲, then multiplying that product
by the appropriate absorption coefficients, ␣ s and ␣ f .
3. COH method
According to Beranek 共1996兲, the absorption power of
an audience, including that of the orchestra if present, is
defined by the first term in the denominator of Sabine’s for-
mula,
0.161•V
RT⫽ , 共3兲
S T • ␣ T ⫹S R • ␣ R ⫹S N • ␣ N ⫹4mV
where V denotes room volume; m is air absorption coeffi-
cient; and S is a surface area. The subscript T specifies the
audience area, including an edge effect equal to the area of a
strip of 0.5 m wide around the seating area, except where a
side area is adjacent to a balcony front or a wall. The sub-
script R specifies the residual area as discussed earlier. The
subscript N applies to any area of special sound-absorbing
material installed for sound control—there may be several N
terms if there are several different types of absorbing mate-
rials. The vertical parts of steps on floors of the seating area
are not included in the area calculations.
2400 J. Acoust. Soc. Am., Vol. 110, No. 5, Pt. 1, Nov. 2001
Frequency 共Hz兲
250 500 1k 2k 4k
0.40 0.44 0.45 0.45 0.45
0.40 0.55 0.60 0.60 0.50
0.32 0.40 0.43 0.43 0.40
0.34 0.41 0.43 0.42 0.41
0.37 0.43 0.44 0.44 0.43
The absorption coefficients of the audience, ␣ T , were
determined by this method from RT measurements made in
21 halls with full occupancy 共Beranek and Hidaka, 1998兲.
The occupied seats in those halls were classified into three
categories depending on the degree of seat upholstering:
heavily, medium, or lightly upholstered, and their average in
each category was taken as the sound absorption coefficient
for that category. Then, the values of the residual absorption
coefficients ␣ R were estimated from data in Table III of Be-
ranek and Hidaka 共1998兲, unless measurements were made
before the seats were installed.
4. ⌬ A method
For this simple method the absorption power of each
seated person in a hall is determined by dividing a occup 关Eq.
共1兲兴 by the number of seats. In practice, the absorption power
per person is measured in a number of halls and the average
used for the predictions of RTs in other halls. The results of
this method have been reported in the literature 共Furrer,
1964; Kimura, 1986; Nagata, 1991; Lord and Templeton,
1996兲. It is often used because the more tedious determina-
tion of the audience area from architectural drawings is un-
necessary. From a literature survey of more than 20 text-
books, it was found that the values shown in Table I are
usually used. Their median value was employed in this study.
B. Comparison of absorption powers of seated
audiences
The total sound absorptions by seated audiences were
obtained from RT measurements before and after installing
the chairs 共with audience兲 in six halls 共named in Tables II
and III兲, where the area actually covered by the seats was
defined as the audience area S p 共see Fig. 2兲. The upholstering
of the seats in those halls is light, except for Hamarikyu-
Asahi, where the level of upholstering is medium.
Figure 3 compares the average values of the ratios of the
measured to the calculated total absorption power of the
seated audiences in the six halls. The COH method has the
highest accuracy for all frequency ranges, because it is based
on the average value of the sound absorption coefficients
measured in 21 existing halls and the level of upholstering of
the chairs is considered. The ⌬A method looks good on av-
erage, but it has the largest variation range in the RT values
for the different halls which is indicated by the vertical bars.
This is to be expected because ⌬A does not take into account
the upholstering and the arrangement of the seats. The K &
K method gives the best results at mid-frequencies, and is
Nishihara et al.: Seated audience in halls
TABLE II. Six halls in this study for which the occupied data were measured accurately. V is room volume; N is the number of seats; S is the total surface
area; S ta is the total area under and around the seating, i.e., aisles, rear and front and side areas; S a is the area over which the audience sits, with 0.5-m edge
effect included; S p , S s , and S f are the area of the upper, side, and front part based on the principles of K & K method, respectively.

V S S ta Sa Sp Ss Sf
共m3兲 N 共m2兲 共m2兲 共m2兲 共m2兲 共m2兲 共m2兲

Mitaka Concert Hall 5500 625 2315 451 413 312 88 32

Hamarikyu Asahi Hall 5800 552 2570 412 395 277 76 26
New National Theater Medium Theater 7200 1038 2786 834 614 534 116 50
New National Theater Opera House 14 500 1810 5666 1411 1153 891 240 40
Vienna, Musikvereinssaal 15 000 1573 4102 1059 955 690 198 78
Tokyo Opera City Concert Hall 15 300 1632 5843 1312 1052 793 261 39

fairly good at frequencies above 500 Hz. However, a large
deviation, about a factor of 1.5, is observed for the 125- and
250-Hz frequency bands. The K & K method is superior to
the ISO method at medium and high frequencies.
From these results it is seen that of the four, the COH
method is the best for predicting the reverberation times in
halls for music if the degree of upholstering on chairs of
conventional design is known. But what about halls in which
it is hoped to use new designs of chairs? Because the number
of sample chairs is usually small, their absorption properties
will have to be measured in reverberation chambers. The K
& K method is obviously more effective than the ISO
method for frequencies of 500 Hz and above. To be fully
effective, the discrepancies for the low-frequency bands must
be solved. In the remainder of this article, we shall attempt to
diagnose the reason for this low-frequency discrepancy and
to recommend means for eliminating it.
found that such panels are often not effective at low frequen-
cies. It is generally believed that the diffusivity of the sound
field in a hall for music is imperfect at all frequencies, al-
though there is little published evidence to support this con-
clusion. At least, it seems reasonable to assume that ideal
diffusion does not occur in the low frequency range, particu-
larly when the hall is occupied and most of the sound ab-
sorption takes place on horizontal surfaces. It must be re-
marked that this discussion applies to halls with
reverberation times greater than 1.5 s, at mid-frequencies
when fully occupied. The absorption coefficient of a seated
audience measured in a reverberation chamber, ␣ s , will
equal the statistical 共theoretical兲 absorption coefficient, ␣ stat ,
if the diffusion of the sound field is perfect and if excess
absorption by the edge effect and other disturbing factors
共Makita and Hidaka, 1987兲 can be excluded.

III. PHYSICAL DIFFUSION IN A REVERBERANT

1
␣ s ⬵ ␣ stat⫽ 1
2
冕0
␲ /2
␣ 共 ␪ 兲 cos ␪ sin ␪ d ␪ . 共4兲

CHAMBER AND IN A HALL

In regard to the diffusivity of the sound field in a rever-
beration chamber, Balachandran et al. 共1967/1968兲 reported
that it is almost perfect for mid- and high frequencies if
sound diffusing panels are installed. Dämming 共1991兲 has
Since the term sin ␪ in Eq. 共4兲 corresponds to the theo-
retical probability density function of the sound incident
angle distribution in the perfectly diffuse sound field, the
‘‘effective’’statistical absorption coefficient in a room can be
defined by modifying Eq. 共4兲 as

TABLE III. Key as in Table II but ␣ p , ␣ s , and ␣ f are the absorption coefficients associated with S p , S s , and
S f areas, respectively.

Frequency 共Hz兲

125 250 500 1k 2k 4k

Mitaka Concert Hall ␣p 0.33 0.44 0.81 0.93 1.06 1.12

␣s 0.04 0.11 0.30 0.22 0.18 0.11
␣f 0.12 0.20 0.27 0.17 0.13 0.07
Hamarikyu Asahi Hall ␣p 0.46 0.59 0.87 0.98 1.12 1.08
␣s 0.16 0.30 0.32 0.21 0.11 0.12
␣f 0.09 0.20 0.28 0.24 0.10 0.05
New National Theater Medium Theater ␣p 0.37 0.55 0.80 0.89 0.96 1.00
␣s 0.12 0.10 0.36 0.30 0.24 0.21
␣f 0.18 0.24 0.29 0.19 0.16 0.15
New National Theater Opera House ␣p 0.42 0.54 0.77 0.89 0.99 1.07
␣s 0.00 0.03 0.38 0.33 0.25 0.23
␣f 0.17 0.14 0.31 0.26 0.16 0.11
Vienna, Musikvereinssaal ␣p 0.30 0.38 0.79 0.97 1.13 1.12
␣s 0.00 0.10 0.24 0.21 0.13 0.13
␣f 0.05 0.23 0.28 0.16 0.06 0.11
Tokyo Opera City Concert Hall ␣p 0.37 0.43 0.75 0.87 0.99 1.06
␣s 0.01 0.04 0.31 0.30 0.18 0.20
␣f 0.08 0.15 0.26 0.25 0.09 0.14

J. Acoust. Soc. Am., Vol. 110, No. 5, Pt. 1, Nov. 2001 Nishihara et al.: Seated audience in halls 2401
FIG. 3. Average values of the ratios of measured to calculated audience
absorption power. The former is derived from measured reverberation times
and the latter is predicted by four methods, ISO, K&K, COH 共Beranek,
1996; Beranek and Hidaka, 1998兲 and ⌬A 共total absorption per person兲, for
the six halls shown in Table II. The vertical bars indicate the variation range
of the ratio by ⌬A method, for which the spread is substantially more than
that for the other three methods.
兰 ␲0 /2␣ 共 ␪ 兲 cos ␪ P 共 ␪ 兲 d ␪
␣ field⫽ , 共5兲
兰 ␲0 /2cos ␪ P 共 ␪ 兲 d ␪
where P( ␪ ) is the probability density function for the actual
hall and ␪ means the incident angle of each sound wave
arriving at a specified receiving point.
The shape of the reverberation chamber in Takenaka
R&D was precisely reconstructed in the computer as a CAD
model, including the 20 diffusion panels which were ran-
domly suspended in the space. An audience area equivalent
to 2.8⫻3.8 m2 was marked off on the floor. A ‘‘receiver’’ was
located in the center of this area. In the calculation, the au-
dience area was assumed to be perfectly absorptive and the
height of the receiving point was 1.2 m. The function P( ␪ )
was calculated using the ray tracing method, where the num-
ber of rays traced was 100 000 and the maximum reflection
order was up to 40 or the maximum time up to 1000 ms.
Figure 4 shows the numerical results for the Takenaka
reverberation chamber, where the effective sound energy dis-
tributions per unit area, cos ␪P(␪), in both the vertical and
horizontal planes are plotted against the incident angle ␪. The
solid line gives cos ␪P(␪) for every degree and the solid
points are for cos ␪兵⌺P(␪)其, calculated every 15 degrees. In
the vertical plane, it is seen that the values for the CAD
model behave similar to the theoretical value for perfect dif-
fusion 共sin 2␪, dotted line兲.
These same measurements were made in a CAD model
of Boston Symphony Hall (V⫽18 750 m3), a typical
shoebox-shaped hall 共see Fig. 5兲. Perfect sound absorption
for the audience areas was assumed at all frequencies. By
comparison with Fig. 4, cos ␪P(␪) vs. ␪ reveals an uneven
distribution, particularly approaching grazing incidence
共above 50 degrees兲 in the vertical plane and at lateral inci-
dence 共particularly ⫾45 to ⫾75 degrees兲 in the horizontal
plane. In this experiment, the receiving position was near the
2402 J. Acoust. Soc. Am., Vol. 110, No. 5, Pt. 1, Nov. 2001
FIG. 4. Typical example of the sound energy distribution per degree and per
unit area as a function of 共upper兲 the incident angle in the median plane, and
共lower兲 averaged distribution at intervals of 15 degrees in horizontal plane.
The data were obtained in a CAD model of the reverberation chamber of
Takenaka R&D. A sound absorbing layer (2.8⫻3.8 m2) was placed at the
center of the floor of the ‘‘chamber.’’ The dotted line is the theoretical value,
sin 2␪, that assumes perfect diffusion and the black dots mean the average
distribution at intervals of 15 degrees in the vertical plane. The sound source
(h⫽0.5 m) is placed at the corner of the chamber and the receiving point
(h⫽1.2 m) is in the center of sound absorbing object.
fore-aft center of the main floor, but was shifted to the side of
the center line by a few meters. This difference was found to
be more apparent under a balcony or at seats on the upper
balcony, indicating an abundance of deviated reflections
from specific directions, as one would expect. This result
shows that the 共effective兲 statistical absorption coefficients in
actual occupied halls are different from those in a reverbera-
tion chamber with ideal diffusion.
As a supplementary discussion, these results may de-
pend on the applicability of the ray tracing method in the
frequency range of 125 to 250 Hz. The validity of the results
must be associated with the existence of a sufficient number
of modes in each frequency range as long as each surface
area is properly subdivided in relation to the wavelength. In
a room, the average density of eigenfrequencies is approxi-
mately equal to 4 ␲ V f 2 /c 3 , where V and c are room volume
and sound velocity, respectively. Taking into account the val-
ues of V above, i.e., 18 750 m3 for the hall and 332.8 m3 for
the reverberation, the number of eigenfrequencies at 63 Hz in
the hall is 23, which is similar in magnitude to that at 500 Hz
Nishihara et al.: Seated audience in halls

FIG. 5. Key as in Fig. 4, but for the model of Boston Symphony Hall. The
sound source position is 3 m from the stage edge on the center line at a
height of 1.5 m, and the receiving position is near the center of the main
floor but shifted to the side a few meters.
in the reverberation chamber, i.e., 26. This implies that the
sound field in the hall at 63 Hz is equivalent to that in the
reverberation chamber at 500 Hz. Next, note the ‘‘Schroeder
frequency’’ 共Schroeder, 1954兲 that defines the lower limiting
frequency at which a statistical treatment is permissible.
With it, letting T equal the reverberation time, f s
⬵2000冑T/V. For the hall (T⫽2 s), f s ⫽21 Hz, and for the
reverberation chamber (T⫽12 s), f s ⫽380 Hz.
IV. OBLIQUE ABSORPTION COEFFICIENT OF AN
AUDIENCE
Even if P( ␪ ) for a particular hall were known, another
variable that is needed to evaluate Eq. 共5兲 is the oblique
absorption coefficient of the audience, ␣ ( ␪ ). If the prepon-
derance of distribution of sound rays in a hall approaches
grazing incidence 共see Fig. 5, upper, above 50 degrees兲, ␣ field
in a hall will certainly be different from ␣ stat in a reverbera-
tion chamber.
Figure 6 is a numerical example of ␣ ( ␪ ) for a homoge-
neous plane porous material, 25 mm in thickness, with rigid
backing 关see also Fig. 6 in Hunt 共1939兲兴. The absorption
coefficient is greater than 0.6 at frequencies of 1 kHz and
above and the variation is gradual as a function of the inci-
dent angle, at least up to an angle of 70 degrees. At frequen-
J. Acoust. Soc. Am., Vol. 110, No. 5, Pt. 1, Nov. 2001
FIG. 6. Example of the oblique sound absorption coefficients for a homo-
geneous plane porous layer with rigid backing at 1/1 octave center fre-
quency of 125 to 4 kHz 共flow resistance per cm: 65 g•s⫺1•cm⫺3; thickness:
25 mm兲.
cies below 500 Hz the absorption coefficient is smallest at
normal incidence and increases rapidly as the angle gets
closer to 80 degrees. We believe that one can expect a similar
tendency for the audience absorption. If so, it follows that
␣ field at mid- and high frequencies is less affected by the
degree of sound diffusion in the sound field, because there
␣ ( ␪ ) is less dependent on the incident angle. At low frequen-
cies the value of ␣ field are highly dependent on the incident
angle distribution for individual rooms. This hypothesis is
consistent with the measurements of Fig. 3 which show that
the greatest differences between the reverberation-chamber
absorptive powers of occupied chairs by the K & K and ISO
methods and those measured in actual halls are significantly
greater at low frequencies.
A. New measuring method of oblique sound
absorption coefficients
Oblique sound absorption coefficients are difficult to
measure, especially for an audience, and no established pro-
cedure is available, although such absorption coefficients
would appear to be one of the essential factors in room
FIG. 7. Block diagram of the new measuring method of oblique absorption
coefficient by digital signal processing. In step 1, the direct sound, E 0 (t), is
measured in an anechoic chamber with no sample, which is used as the
coefficient of a FIR filter. Next, the reflected signal, E r (t), is extracted from
the observed signal, E 0 (t)⫹E r (t), measured with samples by adding the
electrically generated signal with the inverse phase, ⫺E 0 (t), using the FIR
filter in step 2. Note that the microphone must be at the surface of the
sample and that the distance between the loudspeaker and the microphone
must be the exactly same for both steps.
Nishihara et al.: Seated audience in halls 2403
acoustics. Several methods of measurement have been re-
ported 共Hunt, 1939; Ingard and Bolt, 1951; Hirata, 1972兲, but
the accuracy is known to deteriorate as the incident angle
approaches that of grazing.
The seated audience in a hall for music is a collection of
separated bodies a meter and a half in depth 共floor to top of
head兲. Because the audience absorption coefficients for
sound fields approaching grazing incidence are very impor-
tant, any existing method of obtaining them is inappropriate.
We are reporting here a new laboratory measurement proce-
dure, using digital processing, for determining the sound ab-
sorptions of an inhomogeneous-in-depth sample, such as a
seated audience. The primary characteristic of this method is
that for any sample it eliminates the direct sound and retains
the reflected sound. In principle, this method is equal in ac-
curacy for any incident angle, and it can be applied to objects
with irregular surfaces or shapes if, for them, a hypothetical
plane surface is defined over which the absorptions of the
parts of the array can be collected. This definition of oblique
absorption coefficient is not as exact as that for acoustic ma-
terials with flat surfaces; however, it should be noted that, in
practice, an audience area is assumed to be a flat plane with
a certain height. This assumption is the same for all four of
the RT calculation procedures.
FIG. 8. Photograph of the configuration for the measurement of the oblique
The procedure is shown in Fig. 7. As the first step, the absorption coefficients for an audience seated in chairs on a plywood board
impulsive direct sound, E 0 (t), generated by a loudspeaker is in the anechoic chamber of Takenaka R&D. Eighteen mannequins dressed in
measured in an anechoic chamber by a microphone at dis- proper clothes simulated a live audience. Note that the microphone can be
tance L 共with no object present兲 共see step 1兲. The waveform, placed on any part of a mannequin’s body. The loudspeaker is rigidly held at
2-m distance from the microphone by a pole 共shown covered with sound-
E 0 (t), is stored as a coefficient of the finite impulse re- absorbing material兲.
sponse, FIR, filter in step 2. In the upper part of step 2, the
same sound signal is generated with a sample installed. The
共Fig. 8兲. The total sound absorption of each mannequin with
microphone must be on the surface of the sample and the
normal clothes was chosen to equal that of an average per-
distance L between it and the loudspeaker must be exactly
son, ␣ s , as determined by prior measurements in the rever-
the same as it was in step 1 for any angle. The observed
beration chamber.
waveform is E 0 (t)⫹E r (t), where the latter means reflected
Each mannequin and seat was subdivided into nine parts
waveform. Finally, the electrical signal ⫺E 0 (t) with a delay
共head, face, shoulder, breast, abdomen, thigh, shank, feet,
time corresponding to the spacing, L, between the loud-
and seat back shown by 15 numbers兲 and the remaining ar-
speaker and microphone is generated from the FIR filter, and
eas, that is, the bare floor and the uncovered seatback
both signals are summed to yield the reflected term, E r (t). In
共shaded兲 共see Fig. 9兲. Fifteen measuring points plus the two
this experiment, the pressure reflection coefficient, R( ␪ ), as
shaded areas were defined as the representative ones for
a function of frequency, is given by the following equation, if
␣ i ( ␪ j ) (i⫽1,17). The assumption here is that this simulated
the microphone is placed at the surface of the material:
seated audience is equivalent to a rectangular-parallelepiped
FT兵 E r 共 t 兲 其 block just as had been assumed in our earlier determinations
R共 ␪ 兲⫽ , 共6兲
FT兵 E 0 共 t 兲 其 of RTs in a hall or a reverberation chamber. The total absorp-
tion coefficient for the top surface of this block, as shown by
where FT represents Fourier transformation. In our experi-
ment, L was 2 m and the sampling frequency was 22.05 kHz.
The application of this measuring method for any porous
material is presented in the Appendix.

B. Measurement of oblique absorption coefficient of

audience
This new type of measurement was executed for a seated
audience in the anechoic chamber of Takenaka R&D
关8.4 m共l兲⫻7.8 m共w兲⫻9.8 m共h兲兴. Eighteen chairs (6 chairs
FIG. 9. Occupied area per seated person and the definition of the represen-
⫻3 rows, distance between each row: 950 mm兲 held manne-
tative measuring points 共䊉兲 in each part, i.e., head, face, shoulder, breast,
quins, and were supported by a wooden board 共12 mm thick- abdomen, thigh, shank, feet, and seat back, and the other two areas shown
ness兲 laid on the otherwise acoustically transparent floor by the hatching.

2404 J. Acoust. Soc. Am., Vol. 110, No. 5, Pt. 1, Nov. 2001 Nishihara et al.: Seated audience in halls
 the hypothetical plane in Fig. 10共a兲, is given by
n
兺 i⫽1 ␣ i 共 ␪ j 兲 •S i
␣共 ␪ j 兲⫽ n , 共7兲
兺 i⫽1 Si

where S i is the area of each part of the seated audience

FIG. 10. 共a兲 Definition of the incident angle ␪ j . The hypothetical plane, at
which the oblique incident absorption coefficient is defined, is at the head
height of the seated audience. The black dots mean receiving points where
the microphone was located. The equivalent surface area of face, breast, and
seat back are shown as an example. A, B, C, and D mean the measuring
seats that were placed at the front and second rows of a total of 6 chairs
⫻3 rows. 共b兲 The seated mannequins in the 18 chairs. Seats A–B and C–D
were measured separately to observe the influence of seats in front of others.
reached by the incident sound wave at an angle ␪ j , and ␣ i is
the absorption coefficient for that area. In the specific calcu-
lation of each area S i , the human body was subdivided and
replaced by several rounded shapes. Thus, S i is the area of
that surface of any one shape that, approximately, faces the
direction of the incident wave. The areas so determined are
shown in Table IV. When the sound wave impinges on the
bare floor or uncovered 共shaded兲 seatback directly, the cor-
responding projected surface S i was defined as that perpen-
dicular to the wave direction. We note that the microphone is
placed successively at the 17 points, the distance between the
source and the microphone is always L, and the angle ␪ j is
that between the vertical line and that connecting the sound
source and the microphone at the measuring point on the
mannequin.
To consider the influence of the front chairs, the mea-
surement was executed for two mannequins 共designated A
and B兲 in the front row and two 共C and D兲 at the second row
关Fig. 10共b兲兴. Two types of chairs, one lightly upholstered
共type used in Vienna’s Musikvereinssalle, Wisner-Hager KG兲
and the other medium upholstered 共type used in Vienna’s
Burgtheater, Wisner-Hager KG兲, were selected, which are
typical for concert halls and opera houses 共Fig. 11兲. The
measurements were carried out at intervals of 15 degrees
between 0 and 90 degrees and 0 and ⫺90 degrees.
As an example, Fig. 12 compares ␣ i (45°) of the Burgth-
eater model at each point on the mannequins. It is seen that
the oblique incident absorption coefficients at seats A and B
have nearly the same values as those at seats C and D, except
for the abdomen position. Next, Fig. 13 shows ␣ i ( ␪ j ) mea-
sured at the mannequin head as a function of ␪ j . The absorp-
tion coefficient at 250 Hz is higher than that at 125 Hz and at
125 Hz it increases as the incident angle approaches grazing
like that for a plane porous material 共as Fig. 6兲. It is inter-

TABLE IV. Surface area S i in (⫻102 ) cm2 of each part of one mannequin in a seated audience corresponding to the incident angle ␪ j onto the vertical plane
defined in Fig. 10共a兲. The ‘‘other’’ areas are determined for any angle ␪ j by projecting the 共perpendicular and horizontal兲 shaded areas in Fig. 9 onto a surface
at the front of the mannequin perpendicular to the direction of travel of the incident wave. The shaded areas are those that receive energy from the incident
wave that is not blocked by a mannequin. The numbers in parentheses are the percentage ratios of S i ’s to total areas.

Seat
␪ j 共deg兲 Head Face Shoulder Breast Abdomen Thigh Shank Feet back Other Sum

90 4.3 2.6 6.9

共63兲 共37兲 共100兲
75 4.3 7.9 8.6 20.8
共21兲 共38兲 共41兲 共100兲
60 4.3 7.9 6.1 10.9 29.2
共15兲 共27兲 共21兲 共37兲 共100兲
45 4.3 7.9 6.1 19.3 9.0 14.9 12.6 74.1
共6兲 共11兲 共8兲 共26兲 共12兲 共20兲 共17兲 共100兲
30 4.3 7.9 6.1 19.3 9.0 14.4 17.6 12.8 91.4
共5兲 共9兲 共7兲 共21兲 共10兲 共16兲 共19兲 共14兲 共100兲
15 4.3 7.9 6.1 19.3 9.0 14.4 10.4 37.4 20.8 129.6
共3兲 共6兲 共5兲 共15兲 共7兲 共11兲 共8兲 共29兲 共16兲 共100兲
0 4.3 7.9 6.1 19.3 9.0 14.4 10.4 3.9 47.3 20.5 143.1
共3兲 共6兲 共4兲 共13兲 共6兲 共10兲 共7兲 共3兲 共33兲 共14兲 共100兲

J. Acoust. Soc. Am., Vol. 110, No. 5, Pt. 1, Nov. 2001 Nishihara et al.: Seated audience in halls 2405

FIG. 11. Photographs of two types of audience chairs: Musikvereinssalle
model 共left in both photos兲 and Burgtheater model 共right in both photos兲.
esting to see that the sound absorption of all the mannequin
surfaces transferred to the hypothetical plane of Fig. 10共a兲 is,
at 125 Hz, approximately the same 共0.4 average below 60
degrees兲 as that of the porous material 共see Fig. 6兲 with rigid
backing 共also 0.4 average below 60 degrees兲. Figure 14 is an
example of a similar measurement of ␣ i ( ␪ j ) in the horizontal
plane. The variation in this plane is not as extreme as that in
FIG. 12. Comparison of ␣ i ( ␪ j ) 共defined at the hypothetical plane兲 for an
audience in Burgtheater model chairs for ␪ j ⫽45 degrees versus the measur-
ing positions on the seated audience. The numbers on the x axis are those
shown in Fig. 9. Eighteen mannequins (6 chairs⫻3 rows) formed the audi-
ence block. The measurement was executed for four mannequins, i.e., A and
B, and C and D, which were placed, respectively, at the front 共solid line兲 and
the second row 共dotted line兲. The averaged values for the two side-by-side
seats at each row are shown in the figure. The ‘‘thigh,’’ ‘‘shank,’’ and ‘‘feet’’
absorptions are not shown because the sound wave does not strike there.
2406 J. Acoust. Soc. Am., Vol. 110, No. 5, Pt. 1, Nov. 2001
FIG. 13. Averaged ␣ i ( ␪ j ) at the head position of four mannequins at the
front and second rows for Musikvereinssalle 共solid line兲 and Burgtheater
共dotted line兲 model chairs as a function of incident angle ␪ j in the median
plane.
the vertical plane. Obviously, a three-dimensional measure-
ment of ␣ i ( ␪ j ) is a time-consuming procedure, but, from
comparison of Figs. 12 and 14, one can assume that ␣ i ( ␪ j )
determined by moving the loudspeaker around in a circle on
a horizontal plane 共angle ␪ j held constant兲 is nearly the same
at all points on the circle. In Fig. 15 it is seen that the values
of ␣ ( ␪ j ) 关see Eq. 共7兲兴, the absorption coefficients of the hy-
pothetical plane, increase as the incident angle ␪ j approaches
grazing incidence, and this characteristic resembles that of
the behavior of a porous material 共Fig. 6兲. This finding con-
firms our assumption in the previous section, which was
based on Fig. 6.
V. ABSORPTION COEFFICIENT OF AN AUDIENCE IN
A REVERBERATION CHAMBER AND IN CAD
MODELS OF HALLS FOR MUSIC
The effective statistical absorption coefficient of Eq. 共5兲
that reflects the practical sound incidence distribution in the
sound field is rewritten in the discrete form
FIG. 14. Comparison of ␣ i (30°) of Burgtheater model chair in the vertical
plane, but for 45 degrees 共solid line兲 and 0 degrees 共dotted line兲 in horizon-
tal plane. The numbers along the x axis are those shown in Fig. 9. The
‘‘shank’’ and ‘‘feet’’ absorptions are not shown because the sound wave does
not strike there.
Nishihara et al.: Seated audience in halls
 TABLE VI. Five concert halls used for the calculations in this study. V is
room volume; S is the total room surface area including that of the audience
area.

Mean
V S free path No. of
共m3兲 共m2兲 共m兲 Surfaces

Mitaka Concert Hall 5500 2316 9.5 657

Hamarikyu Asahi Hall 5800 2570 9.0 1990
Vienna Musikvereinssaal 15 000 4102 14.6 383
TOC Concert Hall 15 300 5843 10.5 1187
Boston Symphony Hall 18 750 5778 13.0 861

The values of cos ␪ j P(␪ j) were obtained, as for Figs. 4 and 5.

FIG. 15. Oblique absorption coefficients at the hypothetical plane of a
seated audience, at the head height for Musikvereinssalle 共solid line兲 and
Burgtheater 共dotted line兲 model chairs, based on measurements of Fig. 8.
兺 ␪␲ /2⫽0 ␣ 共 ␪ j 兲 cos ␪ j P 共 ␪ j 兲
␣ field⫽
j
. 共8兲
兺 ␪␲ /2⫽0 cos ␪ j P 共 ␪ j 兲
j
The probability density function P( ␪ j ) in this equation was
determined by the CAD model of the reverberation chamber
described in Sec. III. As a precaution, ␣ field was obtained for
five receiving points in the reverberation chamber, and the
numerical differences among them were very small.
Table V is a comparison of ␣ s , ␣ stat , and ␣ field deter-
mined in the reverberation chamber for seated audiences on
the two kinds of the chairs of Fig. 11. Here, ␣ s was measured
using 20 chairs 共5 chairs⫻4 rows兲 with stiff barriers 共height:
1.2 m兲 around them. The three absorption coefficients for
each chair are in good agreement with each other as the
difference between ␣ s and ␣ field is within acceptable ranges.
In this case, ␣ field is in good agreement with ␣ stat since the
diffusivity of the sound field in the Takenaka reverberation
chamber is adequately large even at these low frequencies.
This result indicates that both equations, namely, 共7兲 and 共8兲,
The computer CAD models were constructed precisely from
architectural drawings. The surface details of the models
were sized to fit the wavelengths at 125 and 250 Hz. While
the number of surfaces needed to constitute a CAD model
varies from hall to hall, the conditions that were used to
calculate P( ␪ j ) for all five halls are 共1兲 about 0.1 to 0.8
surface per m2 共total number of surfaces divided by the total
surface area兲, 共2兲 perfect absorption at the audience area, 共3兲
height of receiver is 1.2 m from the floor, and 共4兲 sound
source position is 3 m behind the stage edge on the center
line and its height is 1.5 m. In the analysis of ␣ field 关see Eq.
共8兲兴, more than 12 receiving positions were chosen for the
calculation and averaged. The standard deviation of ␣ field
was about 0.02 for 125 and 250 Hz.
As the next step, the edge effect must be taken into
account if one is to compare ␣ field and ␣ hall . Here, ␣ hall
means the absorption coefficient of a seated audience calcu-
lated from RT values measured in the five halls and ␣ field are
the calculated values using the measured P( ␪ j ) quantities for
each hall.
The mean squared sound pressure 具 p r2 典 at a point located
at a vertical distance x from a rigid surface in a perfectly
diffuse sound field of a reverberation chamber 共Waterhouse,
1955兲 is

冕 冕
yield the same measured values ␣ s in the reverberation 1 2␲ ␲ /2
chamber. 具 p r2 典 ⬀ „1⫹cos 共 2kx•cos ␪ 兲 …•sin ␪ d ␪ d ␸
2␲
The same comparison of 共7兲 and 共8兲 was carried out for 0 0

five concert halls 共see Table VI兲 using CAD models: Tokyo, sin 共 2kx 兲
Hamarikyu-Asahi, Boston Symphony Hall, Vienna Musikv- ⫽1⫹ , 共9兲
2kx
ereinssaal, Tokyo Opera City Concert Hall, and Mitaka Con-
cert Hall. Their architectural/acoustical features are de- where k⫽2 ␲ /␭. For xⰇ␭, the incremental value of squared
scribed in Beranek 共1996兲 and Hidaka et al. 共2000, 2001兲. sound pressure from 0 to x is

TABLE V. Comparison of the three kinds of absorption coefficients, ␣ s ,

␣ stat , and ␣ field . The coefficient ␣ s was measured in the Takenaka rever-
⌬p 2 ⬀ 冕
0
x
共 具 p r2 典 ⫺1 兲 dx⬇ 冕0
⬁
共 具 p r2 典 ⫺1 兲 dx⫽
␭
8
. 共10兲
beration chamber by the ISO method 共with stiff barriers around seated man-
nequins兲. ␣ stat and ␣ field were calculated by Eqs. 共4兲 and 共8兲 where ␣ ( ␪ j )
Because perfect diffusion is almost fully established in the
was measured in the anechoic chamber 共see Fig. 8兲 and P( ␪ j ) was measured
in the CAD model 共see Sec. III兲. reverberation chamber, the sound pressure increase is calcu-
lated as ␭/8 times the surface area of the chamber if the
Musikverein type Burgtheater type effects of edge and corner of the wall surfaces of Fig. 2 are
Hz 125 250 125 250 ignored 共Kath and Kuhl, 1961兲.
On the other hand, the sound incidence distribution dif-
␣s 0.26 0.44 0.43 0.64
␣ stat 0.26 0.56 0.35 0.65
fers in a hall for music as discussed before. Equations 共9兲 and
␣ field 0.27 0.56 0.35 0.66 共10兲 must be modified by replacing the sin ␪ term with P( ␪ ).
Then, one obtains

J. Acoust. Soc. Am., Vol. 110, No. 5, Pt. 1, Nov. 2001 Nishihara et al.: Seated audience in halls 2407
TABLE VII. Comparison of measured and calculated sound absorption co-
efficients for five halls, ␣ hall and ␣ field , which were obtained, respectively,
from 共a兲 measured values of RT under occupied conditions and 共b兲 the
effective absorption coefficient by Eq. 共8兲. In the former, the corrections by
modified Waterhouse edge effect, that is, ␭/3 addition to the audience area,
are included.

Hz ␣ hall ␣ field
Hamarikyu Asahi Hall 125 0.48 0.41
250 0.73 0.69
Boston Symphony Hall 125 0.52 0.31
250 0.71 0.61
Vienna Musikvereinssaal 125 0.47 0.38
250 0.64 0.68
TOC Concert Hall 125 0.41 0.31
250 0.61 0.61
Mitaka Concert Hall 125 0.48 0.32
250 0.63 0.63

FIG. 16. Comparison of the average of the ratios of audience absorption

⌬p 2 ⬀ 冕具
0
x
共 p r2 典 ⫺1 兲 dx
power derived from measured reverberation times to those predicted by K &
K method in the six halls shown in Table II and the new method in the five
halls shown in Table VI at low frequencies.

⫽ 冕冕
0
x ␲ /2

0
cos 共 2kx•cos ␪ 兲 • P 共 ␪ 兲 d ␪ dx, 共11兲 the K & K method as accurately as the COH method in Fig.
3. Furthermore, this new method is applicable for the predic-
and, by rewriting in the discrete form for numerical purpose, tion of the RT of a particular hall, where chairs with a new
␲ /2 x type of design are contemplated, and the COH method is not
⌬p ⬀
2
兺
␪ j ⫽0
P共 ␪ j兲 兺
x m ⫽0
cos 共 2kx m •cos ␪ j 兲 •⌬x•⌬ ␪ , applicable.

⌬x⫽0.5␭, ⌬ ␪ ⫽15 degrees. 共12兲 VI. EXAMPLE

The sound pressure increase ⌬p was evaluated in the 2
CAD model by using Eq. 共12兲 substituting P( ␪ j ) for each
hall 共see Fig. 5 for one hall兲, and, as a result, it was found
that the average sound pressure increase in the five concert
halls listed in Table VI rises about 135 times that in the per-
fectly diffuse sound field, that is, about ␭/3 instead of ␭/8 of
Eq. 共10兲 for the reverberation chamber. Now, these values for
125 and 250 Hz are 90 and 50 cm, respectively. If converted
to the audience surface area in order to calculate the absorp-
tion coefficient of the audience in the hall, they correspond to
the total floor area 共using full area of aisles and other areas兲
at 125 Hz for many cases and the area defined by the COH
method at 250 Hz. In other words, the empirical definition of
audience areas by the edge effect by Beranek 共1960兲 共0.5 m兲
has been proved at 250 Hz theoretically.
By using this definition of the audience area, the absorp-
tion coefficient of the audience in each hall was calculated
and compared with actual value ␣ hall in Table VII. The re-
sults show that they are in satisfactorily agreement with each
other, both at 125 and 250 Hz excluding that at 125 Hz in
Boston Symphony Hall.2 That is to say, the discrepancy of
the K & K method at low frequencies as shown by Fig. 3
was solved by the new measuring method and definition of
audience area, simply by the increase in the audience area S p
due to the added 90-cm edge strip at 125 Hz and 50-cm edge
strip at 250 Hz. Figure 16 shows the average values of the
ratios of the measured to the calculated total absorption
power of the seated audiences by K & K method and the new
method in the five halls shown in Table VI. Accordingly, one
can obtain the absorption coefficient of an audience by using
The Tokyo Opera City 共TOC兲 Concert Hall of Tables II,
III, and VI is used as an example. The sequence of calcula-
tions and the numerical results are shown in Table VIII.
From 500 to 4000 Hz, the K & K reverberation-chamber
method was employed for TOC, as discussed in Sec. II A 2,
because the oblique absorption coefficients of a seated audi-
ence in a large room are less dependent on the incident
angles of the traveling waves than at lower frequencies. For
the two lower frequencies the procedure differs from the K &
K method. The logical progression is sketched briefly below.
The lettering corresponds to that for the lines in Table VIII.
In calculations of reverberation times at frequencies above
500 Hz, the sound-absorbing power of the air must also be
included.
共a兲 The sound absorption coefficients for the top, side, and
front of the audience block, e.g., ␣ p , ␣ s , and ␣ f , as
shown in Table III, were obtained in the reverberation
chamber according to the K & K method. The particu-
lar seats chosen for TOC were used.
共b兲 From architectural plan-drawings of TOC, the areas
S p , S s , and S f were determined, as shown in Table II,
and the products of the absorption coefficients in 共a兲
and these areas were calculated and given for each of
the frequencies.
共c兲 The actual sound absorbing power at each frequency as
measured in the completed TOC is listed here. At the
higher frequencies, the air absorption was subtracted
out, before the numbers shown here.
共d兲 It is seen from the ratios of 共c兲 to 共b兲 presented in this

2408 J. Acoust. Soc. Am., Vol. 110, No. 5, Pt. 1, Nov. 2001 Nishihara et al.: Seated audience in halls
TABLE VIII. Example 共Tokyo Opera City Concert Hall, TOC兲 of the procedure for reverberation chamber measurements of sound absorption coefficients by
the K & K method above 500 Hz and new measurements at low frequencies to the prediction of audience absorption in halls for music.

Frequency 共Hz兲

125 250 500 1k 2k 4k

共a兲 The absorption coefficients of top, side, and front ␣p 0.37 0.43 0.75 0.87 0.99 1.06
area of audience block measured in the ␣s 0.01 0.04 0.31 0.30 0.18 0.20
reverberation chamber by K & K method 共from
Table III兲. ␣f 0.08 0.15 0.26 0.25 0.09 0.14
共b兲 Total audience absorption, product of 共a兲 and 299 357 686 778 836 898
each area, S p ⫽793, S s ⫽261, and S f ⫽39, m2
共from Table II兲.
共c兲 Total audience absorption measured in TOC using 466 614 650 705 711 744
Eq. 共1兲, m2
共d兲 Ratio of 共c兲 to 共b兲 1.56 1.72 0.95 0.91 0.85 0.83
共e兲 Calculation of absorption coefficient for TOC 0.31 0.61 ¯ ¯ ¯ ¯
using Eq. 共8兲
共f兲 Areas for TOC used according to Sec. V 共from 1312 1052 ¯ ¯ ¯ ¯
Table II兲, m2
共g兲 Product of 共e兲 and 共f兲, m2 407 642 686 778 836 898
共h兲 Ratio of 共c兲 to 共g兲 1.14 0.96 0.95 0.91 0.85 0.83

line that the K & K method underestimates the audi-
ence absorption in the 125- and 250-Hz bands.
共e兲 From the sound absorption coefficients ␣ ( ␪ j ) mea-
sured in an anechoic chamber 共see Fig. 8兲 and the prob-
ability density functions P( ␪ j ) determined in a CAD
model, the ‘‘effective’’ absorption coefficients ␣ field
关see Eq. 共8兲兴 were determined at 125 and 250 Hz. Ob-
viously, this is the expensive part of the new method,
because the procedure of Fig. 8 is required, and an
accurate CAD model must be achieved.
共f兲 From the results of Eq. 共12兲, the ‘‘effective’’ areas at
125 and 250 Hz are calculated from the floor plan. This
turns out to be the equivalent of using the entire hori-
zontal area under and around the seating, S ta , at 125
Hz. For 250 Hz, the equivalent area is that over which
the audience sits, plus 0.5 m around the edges 共except
when adjacent to walls or balcony fronts兲.
共g兲 Finally the ‘‘corrected,’’ predicted, total-audience ab-
sorbing powers for these frequencies are presented
here, along with the higher frequency values by the K
& K method taken from 共b兲.
共h兲 Shown is a set of the ratios for TOC, one of the five
halls whose ratios were averaged to obtain the lower
curve of Fig. 16.
It is possible to predict RT with higher accuracy for new
types of chairs than that predicted by the COH method for
‘‘average’’ chairs.
VII. CONCLUSIONS
To predict the reverberations times of a hall for music in
the occupied condition, one must know the sound absorption
per unit area by the audience. The accuracy of four existing
methods, ISO, K & K, COH, and ⌬A, for determining the
sound absorption coefficients of an audience were compared
with measurements in six halls. The COH method, which
predicts the absorption coefficients of audiences from mea-
surements made in actual halls, was found to be the most
appropriate at all frequencies. The ISO method was least
successful, while the ⌬A method was good on average, al-
though it led to widely varying values of reverberation times
among the halls because it does not take into account the
types of upholstering or chairs. Both the ISO and the K & K
methods overestimated the reverberation times at low fre-
quencies by a factor of 1.5 or more. The K & K method,
which uses a reverberation chamber for the measurements,
was most accurate at mid-frequencies. This article extends
the accuracy of the K & K method to the low frequencies, a
desirable goal because this method can then be used to de-
termine audience absorption coefficients of new types of
chairs or upholstering in a reverberation chamber.
It was surmised that the difference between audience
absorption coefficients at low frequencies in reverberation
chambers and those measured in halls for music was caused
by the differences in the diffusion of their respective sound
fields. The first stage of the investigation was to determine
the actual state of diffusion, first, in a large reverberation
chamber 共with the audience test block in the center兲 and,
second, in six halls fully occupied. It was found that the
audience areas in occupied halls are bombarded by a greater
percentage of sound waves near grazing incidence than in
reverberation chambers 共Figs. 4 and 5兲. Numerically, the
probability density of the incident sound waves was deter-
mined.
In the next stage the absorption coefficients of a seated
audience at 125 and 250 Hz as a function of angle of inci-
dence of a sound wave was determined in an anechoic room
by a novel procedure 共Figs. 8 –15兲. The absorption coeffi-
cients at these low frequencies increase greatly as the angle
of incidence increases from 60 to 90 degrees.
Using the K & K method in the reverberation chamber,
the absorption coefficients of the seated audience were de-
termined by the audience area, which were the sum of the
top, sides, and correction areas shown in Fig. 2. Using
these coefficients, the reverberation times were calculated
for six occupied halls and compared with those measured.

J. Acoust. Soc. Am., Vol. 110, No. 5, Pt. 1, Nov. 2001 Nishihara et al.: Seated audience in halls 2409
The comparison is shown by the upper 共dotted兲 curve in
Fig. 16.
Finally, it was shown mathematically that using both 共a兲
the probability density of the incident sound waves as a func-
tion of angle of incidence and 共b兲 the measured sound ab-
sorption coefficients of an audience as a function of angle of
incidence, the sound energy 共pressure squared兲 impinging on
an audience is larger in the halls 共proportional to ␭/3兲 as
compared to that in the reverberation chamber 共proportional
to ␭/8兲, that is to say an amount of about 135 larger. This
energy increase can be accounted for in the calculations of
the reverberation times in actual halls, simply by increasing
the area of the audience in the halls. At 125 and 250 Hz, the
results show that absorption coefficients determined by the K
& K method may be improved 共lower curve of Fig. 16兲 by an
FIG. 17. Anechoic chamber test: 共a兲 Waveform of direct sound E 0 (t) that
is used for the coefficients of the finite impulse response 共FIR兲 filter. 共b兲
Residual value of the waveform by the operation E ⬘0 (t)⫺E 0 (t), where
E 0⬘ (t) is also the waveform measured without sample and E 0 (t) is that given
by 共a兲 above 共c兲 FT兵 E ⬘0 (t)⫺E 0 (t) 其 /FT兵 E 0 (t) 其 , which shows that the can-
cellation of the direct sound by this method function with enough accuracy
for wide frequency range, because the residual error is less than 2% above
200 Hz.
2410 J. Acoust. Soc. Am., Vol. 110, No. 5, Pt. 1, Nov. 2001
FIG. 18. Example of a measurement made, using this method, of the
pressure reflection coefficient R for a glass wool sample 50 mm in thick-
ness with rigid backing. The sound wave traveled perpendicular to the sur-
face. The dashes are the calculated values based on the empirical equa-
tion by Delany and Bazley 共1970兲 with flow resistance per cm, 24
g•s⫺1•cm⫺3, which is the certified value of the test sample. The anomalous
value 共above one兲 at low frequencies is caused by edge diffraction, because
the sample size was limited to 910⫻1820 mm. For measurement of audi-
ence chairs in this article, the sample size was large enough to avoid this
phenomenon.
increase in area equal to a strip around each of the audience
blocks in a concert hall of 90 cm at 125 Hz and 50 cm at 250
Hz 共excluding the edges adjacent to walls or balcony fronts兲.
It is believed that these corrections will apply to every type
of seat. They will differ a little from one hall to another, but
the results are at least as precise as the Beranek 共1996兲 and
Beranek and Hidaka methods 共1998兲, which were found to
be the most accurate of those methods studied before this
procedure evolved.
APPENDIX: AN APPLICATION EXAMPLE OF NEW
MEASURING METHOD OF OBLIQUE ABSORPTION
COEFFICIENT
As a preliminary experiment, the measurement was ex-
ecuted in an anechoic chamber so that the numerical error
that will result from elimination of direct sound was evalu-
ated. Figure 17 shows 共1兲 the direct sound, E 0 (t), measured
at step 1 in Fig. 7, 共2兲 residual value of the waveform by
the operation E 0⬘ (t)⫺E 0 (t), where E 0⬘ (t) is the waveform
measured with no sample at different air condition, and 共3兲
FT兵 E 0⬘ (t)⫺E 0 (t) 其 /FT兵 E 0 (t) 其 . It is impossible to completely
eliminate direct sound signal since it involves effects of
fluctuation of medium or digital rounded errors. Never-
theless, it is judged from the result of Fig. 17共c兲 that re-
sidual error by this cancellation method is less than 2%
above 200 Hz, which is an acceptable value for usual
measurement.
Figure 18 is a measurement example of the normal inci-
dent reflection coefficient for plane porous materials, i.e.,
glass wool of 50 mm in thickness, with rigid backing. The
measurement demonstrated that this method could give
stable results for all frequency ranges. It should be noted that
perfect elimination of direct sound requires adequate control
of temperature and humidity in an anechoic chamber, which
was done in these tests.
Nishihara et al.: Seated audience in halls
1
For each hall, two sets of seats were measured, i.e., the manufacturer’s
samples and the seats actually installed. The sample seats were measured to
decide on the design, but a problem arose with the installed seats. In three
of the six halls, the upholstery cloth on the installed seats was back-
sprayed. The coefficients reported here are for the seats as actually in-
stalled.
2
For Boston Symphony Hall, the absorption of the elevated wooden floor
was included as part of the seated audience absorption.
ASTM C423-90a 共1990兲. ‘‘Test method for sound absorption and sound
absorption coefficients by the reverberation room method’’ 共American So-
ciety for Testing and Materials Standards兲.
Balachandran, C. G., and Robinson, D. W. 共1967,1968兲. ‘‘Diffusion of the
Decaying Sound Field,’’ Acustica 19, 245–257.
Beranek, L. L. 共1960兲. ‘‘Audience and seat absorption in large halls,’’ J.
Acoust. Soc. Am. 32, 661– 670.
Beranek, L. L. 共1962兲. Music, Acoustics and Architecture 共Wiley, New
York兲.
Beranek, L. L. 共1969兲. ‘‘Audience and seat absorption in large halls. II,’’ J.
Acoust. Soc. Am. 45, 13–19.
Beranek, L. L. 共1996兲. Concert and Opera Halls: How They Sound 共Acous-
tical Society of America, Melville, NY兲.
Beranek, L. L., and Hidaka, T. 共1998兲. ‘‘Sound absorption concert halls by
seats, occupied and unoccupied, and by the hall’s interior surfaces,’’ J.
Acoust. Soc. Am. 104, 3169–3177.
Beranek, L. L., Hidaka, T., and Masuda, S. 共2000兲. ‘‘Acoustical design of
the opera house of the New National Theater, Tokyo, Japan,’’ J. Acoust.
Soc. Am. 107, 355–367.
Bradley, J. S. 共1992兲. ‘‘Predicting theater chair absorption from reverbera-
tion room measurements,’’ J. Acoust. Soc. Am. 91, 1514 –1524.
Bradley, J. S. 共1996兲. ‘‘The sound absorption of occupied auditorium seat-
ing’’ J. Acoust. Soc. Am. 99, 990–995.
BS 3638 共1987兲. ‘‘Method for measurement of sound absorption in a rever-
beration room.’’
Dämmig, P. 共1991兲. ‘‘Model Investigation into Sound Fields in Reverbera-
tion Rooms,’’ Acustica 75, 105–120.
Davis, W. J., Orlowski, R. J., and Lam, Y. W. 共1994兲. ‘‘Measuring audito-
rium seat absorption,’’ J. Acoust. Soc. Am. 96, 879– 888.
Delany, M. E., and Bazley, E. N. 共1970兲. ‘‘Acoustical Properties of Fibrous
Absorbent Materials,’’ Appl. Acoust. 3, 105–116.
DIN 52212 共1961兲. ‘‘Testing of architectural acoustics; measurement of
sound absorption coefficient in a reverberation room’’ 共German National
Standards兲.
Furrer, W. 共1964兲. Room and Building Acoustics and Noise Abatement 共But-
terworths, London兲, pp. 70–71.
J. Acoust. Soc. Am., Vol. 110, No. 5, Pt. 1, Nov. 2001
View publication stats
Hidaka, T., Nishihara, N., and Beranek, L. L. 共1996兲. ‘‘Mechanism of sound
absorption by seated audiences in concert halls,’’ Presented at the 3rd Joint
Meeting: Acoustical Societies of America and Japan, J. Acoust. Soc. Am.
100, 2705共A兲.
Hidaka, T., Nishihara, N., and Beranek, L. L. 共2001兲. ‘‘Relation of acoustical
parameters with and without audiences in concert halls and a simple
method for simulating the occupied state,’’ J. Acoust. Soc. Am. 109,
1028 –1042.
Hidaka, T., Beranek, L. L., Masuda, S., Nishihara, N., and Okano, T. 共2000兲.
‘‘Acoustical design of the Tokyo Opera City 共TOC兲 concert hall, Japan,’’ J.
Acoust. Soc. Am. 107, 340–354.
Hirata, Y. 共1972兲. ‘‘Measurement of Absorption coefficient by electrically
canceling method,’’ J. Acoust. Soc. Jpn. 28, 416 共in Japanese兲.
Hunt, F. V. 共1939兲. ‘‘Investigation of room acoustics by steady state trans-
mission measurement, I,’’ J. Acoust. Soc. Am. 10, 216.
Ingard, U., and Bolt, R. H. 共1951兲. ‘‘A free field method of measuring the
absorption coefficient of acoustic materials,’’ J. Acoust. Soc. Am. 23, 509.
ISO 354 共1985兲. ‘‘Acoustics—Measurement of sound absorption in a rever-
beration room’’ 共International Organization for Standardization兲.
JIS A 1409 共1998兲. ‘‘Method for measurement of sound absorption coeffi-
cients in a reverberation room’’ 共Japanese Industrial Standards兲.
Kath, U., and Kuhl, W. 共1961兲. ‘‘Einfluss von Streufläche und Hallraumdi-
mensionen auf den Gemessenen Schallabsorptionsgrad,’’ Acustica 11, 50–
64.
Kath, U., and Kuhl, W. 共1964兲. ‘‘Messungen zur Schallabsorption von Per-
sonen auf Ungepolsterten Stuhlen,’’ Acustica 14, 50–55.
Kath, U., and Kuhl, W. 共1965兲. ‘‘Messungen zur Schallabsorption von Pol-
sterst hlen mit und ohen Personen,’’ Acustica 15, 127–131.
Kimura, S. 共1986兲. Architectural Acoustics and Noise Control 共Shoukoku-
Sya, Tokyo兲, p. 201 共in Japanese兲.
Lord, P., and Templeton, D. 共1996兲. Detailing for Acoustics 共E&FN SPON兲,
p. 195.
Makita, Y., and Hidaka, T. 共1987兲. ‘‘Revision of the cos ␪ Law of Oblique
Incident Sound Energy and Modification of the Fundamental Formulations
in Geometrical Acoustics in Accordance with the Revised Law,’’ Acustica
63, 163–173.
Nagata, M. 共1991兲. Acoustic Design of Architecture 共Ohmu-Sya, Tokyo兲, p.
215 共in Japanese兲.
Schroeder, M. R. 共1954兲. ‘‘Die statistischen Parameter der Frequenzkurven
von großen Räumen,’’ Acustica 4, 594.
Waterhouse, R. V. 共1955兲. ‘‘Interference patterns in reverberant sound
fields,’’ J. Acoust. Soc. Am. 27, 247–258.
Nishihara et al.: Seated audience in halls 2411