Você está na página 1de 12

This article appeared in a journal published by Elsevier.

The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elseviers archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright

Author's personal copy

Applied Acoustics 70 (2009) 378388

Contents lists available at ScienceDirect

Applied Acoustics
journal homepage: www.elsevier.com/locate/apacoust

Technical Note

Guidelines for acoustical measurements in churches


Francesco Martellotta a,*, Ettore Cirillo a, Antonio Carbonari b, Paola Ricciardi c
a
b
c

DAU, Politecnico di Bari, Via Orabona 4, 70125 Bari, Italy


DCA, Universit IUAV di Venezia, Dorsoduro 2206, 30123 Venezia, Italy
DIIA, Universit di Pavia, Via Ferrata 1, 27100 Pavia, Italy

a r t i c l e

i n f o

Article history:
Received 20 December 2007
Received in revised form 1 April 2008
Accepted 16 April 2008
Available online 3 June 2008
Keywords:
Church acoustics
Measurement techniques
Worship buildings

a b s t r a c t
The acoustics of churches is a cultural heritage to be preserved as carefully as the artistic and architectural aspects of this particular category of buildings. The acoustic characteristics of an environment could
be measured according to different techniques varying from the numerical quantication, by means of
acoustical parameters, to the recording of the binaural or ambisonic impulse responses of several combinations of sourcereceiver locations. The complexity of this kind of buildings can lead different
researchers to choose dissimilar sourcereceiver arrangements, that yield incomparable results. To prevent different approaches to the problem and to assist in obtaining comparable data, the results of experimental measurements deriving from previous acoustical surveys are statistically analysed in order to
better understand the spatial variation of acoustical parameters leading to the denition of a set of guidelines to standardize the choice of sources and receivers locations. In addition, suitable hardware combinations depending on the purpose of the measurement are nally suggested.
2008 Elsevier Ltd. All rights reserved.

1. Introduction
The acoustics of worship places and churches in particular have
gained increasing importance in recent years. Several research
groups have been studying church acoustics, each one focussing
on different geographical areas [17]. However, a more detailed
analysis of the studies shows that the measurement techniques
are often signicantly different in terms of measurement equipment and calculated parameters. A further source of differences
is the large variability of church shapes, which may induce different researchers to locate sources and receivers in different ways
with the consequence of having hardly comparable measurement
efforts. The latter problem is one of the most important sources
of uncertainties when comparing both the acoustics of different
churches and the measurements of different teams in the same
place.
The following proposal was developed by a team of three Italian
universities within the framework of the national interest program
of scientic research The acoustics of worship places, funded by
the Italian Ministry of Universities and Research, with the aim of
providing a technical and operative support to measurement sessions inside churches of different traditions and of different countries. This measurement program will be used to collect a detailed
description of the acoustic characteristics of one of the most
important group of cultural heritages to be used in order to

* Corresponding author. Tel.: +39 0805963631; fax: +39 0805963701.


E-mail address: f.martellotta@poliba.it (F. Martellotta).
0003-682X/$ - see front matter 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.apacoust.2008.04.004

improve the knowledge of the sound propagation inside this typology of buildings, to preserve the original characteristics in case of
restoration, and to determine optimal approaches to improve the
acoustic conditions inside existing buildings.
One of the most interesting, and worrying, aspects that arise
when dealing with churches is represented by their dimensions
(which vary from small to huge) and by their spatial complexity
which sometimes derives from the original design but, more frequently, is the result of the additions and modications made over
centuries. In addition, sound sources in a church may be located in
different positions, depending on the evolution and on the current
structure of the liturgy.
This proposal, always keeping in mind the general requests of
the ISO 3382 standard [8] and the similar work carried out for theatres [9], takes into account three basic aspects of the measurement procedure: the choice of the source positions, the choice of
the receiver positions, and the equipment to be used.
2. Source positions
The location of the sound sources in a church (Fig. 1) may be
quite varied because even though the sanctuary (or chancel) is
the centre of the liturgical action, there are (regardless of which
Christian tradition is considered) at least two acoustic poles in
this area: the altar and the ambos (or lectern). The pulpit, generally located outside the sanctuary and close to the assembly,
plays a fundamental role in most Reformed churches and in
Catholic churches may provide an interesting insight into the

Author's personal copy

379

F. Martellotta et al. / Applied Acoustics 70 (2009) 378388

S9
1m

S3

S4
S8

S7

>2m

S1

S6a

1m

S6b

1.5m

>1m

1.5m

S5

S2

>2m

Fig. 1. Typical location of sound sources. S1, altar; S2, main altar; S3, ambos; S4, pulpit; S5, choir; S6, organ; S7, congregation; S8, dome; S9, chapel.

sound reinforcement methods used before the electro-acoustics


era. In addition, choir and organ have a well dened position
in the church and when their role during the celebration is signicant they should also be taken into account. The assembly of
the faithful, being an active part during congregational singing,
should be considered as a further sound source. Finally, it cannot
be excluded that other kinds of musical performance may be
hosted in a church (generally with several musicians in front
of the altar). For all these reasons it clearly appears that a complete and detailed analysis of the sound propagation inside a
church is hardly possible. However, in order to make measurement efforts carried out by different teams comparable, the proposal is to assume at least two source locations. The rst one
should be considered as a reference source location placed at
a xed position. The second one should be chosen among the
others as a function of the specic purpose of the measurement
campaign. In order to encourage the use of further sound source
locations, which is particularly recommended in churches where
speech, organ, and choir are important, no upper limit to the
number of locations is specied, the only limit being the time
availability of the church in the desired conditions.
The reference source location (S1), required for the comparisons
among different teams (and among different churches), should be
placed, if possible, in front of the altar currently used for celebrations on the side of the listening area, at a distance of 1.5 m and
on the symmetry axis provided that signicant focussing effects
due to reecting surfaces may be excluded. In Eastern churches
(both Christian and Orthodox), where an iconostasis often separates the sanctuary from the nave, the sound source should be
located in the nave side at 1.5 m from the iconostasis (in this case
source S2 may be used as the actual altar position). In case focussing effects cannot be excluded the source should be shifted by 1 m
from the axis. In any case, the source height should be 1.5 m above
the oor.
The high altar source location (S2), may be employed in
churches where the original high altar has been preserved
(although not used), and its position is raised or signicantly distant from the altar currently in use. The sound source should be
placed on the symmetry axis (provided that signicant focussing
effects due to reecting surfaces may be excluded) and at a distance of 1.5 m from the vertical surface of the altar. In this case
the source height should be 1.5 m above the base of the high altar.
In churches where the altar used for celebration is movable and
has a provisional character source location S2 may replace S1 as

the reference source location provided that the altar does not
obstruct the propagation towards the assembly.
The ambos (or lectern) source location (S3), generally located in
the chancel area, should be used when the effects of unamplied
speech are of particular interest. Alternatively, the pulpit source
location (S4), located nearby the audience, should be used when
access is possible. In both cases, the source height should be
1.5 m above the oor or, if the balustrade is too high, the height
should be increased to overcome the top of the balustrade by at
least 0.50 m. On the pulpit the source should be placed as close
as possible to the balustrade to simulate speakers natural position.
Given the peculiar character of the source position it would be
preferable to use a sound source having a directivity comparable
to that of the human voice [1012], aiming at the centre of the congregation. Since no standard sound source is available for emulating human voice (ITU-T recommendation P-51 [13] providing poor
constraints on source directivity), the directivity pattern of the employed directional source should be included in the documentation. However, until reliable human voice simulators will be
available, the use of an omni-directional sound source should be
considered as a viable alternative.
The choir source location (S5), should be used whenever a permanent choir sing in the church and should be located at the centre
of the area occupied by the singers or, in case of ancient wooden
choirs (provided that they are still in use), at the centre of the corresponding area. In Orthodox churches and in all those churches
where the choir is generally split into two groups, two sound
source positions should be used, located at the centre of the
respective areas occupied by the singers. In any case, the source
should be at least 1.0 m far from reecting walls and at 1.5 m
above the oor.
The organ source location (S6), should be located as much as
possible close to the centre of the organ pipes (at a distance of at
least 1 m). In case the span of the organ pipes is larger than 6 m
it is recommended to use two positions (S6a and S6b) shifted by
a distance equal to half the width of the organ. In case the height
corresponding to the centre of the pipes cannot be reached it is recommended to place the source at 2 m above the oor.
The congregation source location (S7), should be used whenever
the assembly of the faithful plays an active role during the liturgy
or when the acoustic coupling between this area and the sanctuary
is thought to be poor. The source should be placed at the centre of
the largest area occupied by the congregation at an height of 1.5 m
above the oor.

Author's personal copy

380

F. Martellotta et al. / Applied Acoustics 70 (2009) 378388

The dome source location (S8), should be used in presence of


large domes when the reference source location (S1) is located outside the area covered by the dome. The source should be preferably
located on the symmetry axis at 2 m from the centre of the dome
and 2 m within the geometrical projection of the dome on the horizontal plane, and at 1.5 m above the oor.
Further source positions (S9, etc.), should be used in addition to
the others already mentioned, when the church is characterized by
parts (such as transepts and side chapels) which may reasonably
behave as coupled volumes [14] or which may be used independently (i.e. to celebrate daily masses, for choir performance or
rehearsals). In all these cases the sound source should be placed
at the centre of the area or, in case of independent use, close to
the position of real sound sources (minister, singers, etc.), provided
that the distance from reecting surfaces is at least 1 m and the
height above the oor is 1.5 m.
Finally, in all those churches where the sound system has a particular historical value, or when the analysis of the speech intelligibility
is particularly important, the whole set of the loudspeakers of the PA
system might be used as a sound source, provided that the test signal
is fed directly to the system. In this case the settings should not be
modied with respect to their normal values, and the characteristics of the system and of the loudspeakers, including model, position
and wiring details, should be carefully documented in the measurement log. Feeding the signal directly to the system allows the exclusion of the microphones (and their controls) which may be more
easily moved and modied by the users. In any case this source
should only be used in addition to the two conventional sound
source positions.

work, but the nal choice is generally the result of a compromise


between two contrasting needs. On one side, the need to have an
accurate picture of the variation of the acoustic parameters inside
the church which leads to maximize the number of source and
receiver positions. On the other side, the limited time granted by
the church management for the measurements which, to assure
minimum background noise, and minimize the risk of interruption,
should be conducted late at night.
A general rule to be followed is that the receivers should be
placed preferably in the main listening area (MLA), even though
for churches such area is much less dened than in performing
spaces. In fact, in many churches (especially the very large ones)
the area occupied by pews or seats may vary as a function of the
period of the year with the largest area observed during the major holy days. In addition, it should also be considered that
standing people are not unusual in churches. Consequently, for
the purpose of the acoustical measurements the MLA should
be dened as the combination of the largest area covered by
pews and the area which is more likely to be occupied by standing faithful (Fig. 2).
ISO 3382 standard [8] suggests that receivers should be placed
at 1.2 m above the oor, at a distance of 1/4 wavelength from any
reecting surfaces (corresponding to about 1 m), and at half wavelength from each other. However, the denition of the optimum
placement of the receivers (capable of ensuring an accurate
description of the variations of the acoustical parameters), cannot
rely only on practical considerations but should take into account
the actual distribution of the sound in the room. In order to scientically dene such rules the results of the acoustic measurements
carried out in several churches were statistically analysed.

3. Receiver positions
The volumetric complexity and the plan layout of many
churches make the choice of the receivers locations even more difcult (and subjective) than the choice of the source locations. Several parameters should be taken into account in order to ease the

3.1. Preliminary analysis of the spatial distribution of acoustic


parameters
A sample of 37 churches was analysed in order to investigate
the spatial variation of the acoustic parameters. All the measure-

Ltot

08

07

Dav

06

W
05

04

03

02

09

01

10

L/6

Fig. 2. Arrangement of the minimum number of receivers in a typical church. Six receivers are placed on one side the main listening area together with three mirrored
control receivers. One receiver is located in the chancel area to analyse sound perceived by ministers and one in the side aisle to analyse acoustic conditions in acoustically
unfavorable portions of the space. Area enclosed in dashed line represents the main listening area.
Table 1
Octave bands used to calculate multi-octave bands average of acoustic parameters and corresponding JNDs used

Octave bands
JND

T30

T20

EDT

G10

D50

C80

Ts

LF

IACC

5001k
5%

5001k
5%

5001k
5%

5001k
1 dB

5002k
5% (abs)

5002k
1 dB

5001k
20 ms

5002k
0.05

5002k
0.075

Author's personal copy

F. Martellotta et al. / Applied Acoustics 70 (2009) 378388


Table 2
Mean, minimum, and maximum values of standard deviations of the acoustic
parameters measured in the 37 churches surveyed and expressed in terms of JND

Min
Max
Avg

T30

T20

EDT

G10

D50

Ts

C80

LF

IACC

0.1
0.6
0.2

0.1
0.9
0.3

0.3
2.7
1.1

0.3
2.6
1.2

0.5
3.3
1.6

1.1
6.5
2.9

0.7
4.3
2.6

0.9
2.5
1.4

0.7
2.4
1.4

Table 3
Frequency distribution of standard deviations calculated in surveyed churches
JND

T30

T20

EDT

G10

D50

Ts

C80

LF

IACC

<0.5
0.51
11.5
1.52
22.5
2.53
33.5
3.54
>4

36
1
0
0
0
0
0
0
0

33
4
0
0
0
0
0
0
0

3
18
11
2
2
1
0
0
0

2
10
17
7
0
1
0
0
0

0
6
13
8
6
3
1
0
0

0
0
3
10
6
3
6
1
8

0
1
1
6
10
8
9
1
1

0
3
21
7
3
1
0
0
0

0
4
18
13
2
0
0
0
0

ments were carried out in strict agreement with ISO 3382 standard
[8] by using an omni-directional sound source made of twelve
loudspeakers mounted on a dodecahedron combined with a subwoofer to obtain a frequency response from 50 Hz to 16 kHz. The
signal used was an envelope equalized sine sweep [15]. The microphones were a random incidence GRAS 40AR and a Soundeld
MKV.
The standard deviation of the multi-octave-band averages of
the acoustic parameters measured in each church were considered
and the corresponding standard deviations were then calculated in
terms of just noticeable difference (JND). The latter values for each
acoustic parameter (Table 1), were taken from ISO 3382 draft [16]
circulated in 2004, with the exception of Ts. In fact, for this parameter JND is normally assumed equal to 10 ms, but a preliminary
investigation carried out by the authors suggests that when the
reverberation time is longer than 4 s JND should be much longer,
at least 30 ms. For this reason, in the present analysis JND for Ts
was assumed to be 20 ms. The results of the analysis of standard
deviations are reported in Tables 2 and 3.
It can be observed that reverberation times (T30 and T20) show
the lowest variations, with a standard deviation which is well below the JND limit in all the cases, indicating that both these parameters are quite stable independent of the source and receiver used.
EDT shows higher variations with an average standard deviation of about 1 JND, but with one-third of the values above this value and a maximum of 2.7 JNDs. In this case, as shown in Ref. [17],
a dependence on sourcereceiver distance appears in most cases,
with a trend to increase as a function of the distance (with an average slope of 0.03 s/m). The mean slope tends to decrease as the
dimensions of the church grow, and if the slope is expressed in
terms of distance beyond which a JND may be perceived (later
on called JND distance) the mean value corresponds to 8.7 m.
The analysis of the correlations between JND distance and some
architectural parameters (volume, total length, nave length, oor
surface, mean width, average altar-listener distance), was performed with the hope of nding some sort of relationship with
church specic parameters. As reported in Fig. 2 the nave length
(L) was conventionally assumed as the distance from the altar to
the entrance, the average altar-listener distance (Dav) was dened
as the distance from the altar to the centre of the main listening
area. In all the cases the regression equations were forced to the
origin. The best correlation was found with the mean altar-listener

381

distance. The regression equation (Fig. 3a) is statistically signicant


(R2 = 0.53, p < 0.001) and shows that the JND distance is 0.45 times
the mean sourcereceiver distance, suggesting that the spacing between receivers may grow together with the church dimensions.
G shows variations of the same magnitude of EDT, with an average standard deviation of 1.2 JNDs and a maximum of 2.6 JNDs. The
variations are expected because several researches showed that G
decreases as the sourcereceiver distance grows [6,7]. The slope
varies between 0.27 dB/m and 0.04 dB/m, with a mean value
of 0.127 dB/m (corresponding to a JND distance of 7.9 m). Again,
correlations between JND distance for each church and corresponding architectural parameters were investigated. The best results were obtained by taking into account the mean altar-listener
distance (Fig. 3b). The regression equation is highly signicant
(p < 0.0001) and shows that JND distance is about 2/5 of the mean
altar-listener distance.
Energy ratios show the highest variations, with average values
of the standard deviation varying from 1.6 JNDs for D50 to 2.9 JNDs
for Ts. C80 stays in between with a standard deviation of 2.6 JNDs.
As shown in Ref. 6 energy ratios are strongly dependent on source
receiver distance with D50 and C80 decreasing and Ts increasing as
the distance grows. For C80 the slope varies from 0.12 dB/m to
0.53 dB/m with a mean value of 0.27 dB/m and a JND distance
of 3.7 m. For Ts the slope varies from 2.2 ms/m to 11 ms/m with
a mean value of 5.5 ms/m and a JND distance of 3.6 m. The analysis
of the correlations with architectural parameters shows that for
C80 the JND distance is best related to the nave length. The regression is highly signicant (R2 = 0.661, p < 0.0001) and shows that
JND distance is about 1/10 of this measure (Fig. 3c). Ts also shows
the best correlation with the nave length and even though the statistical signicance is lower (R2 = 0.515, p < 0.001) the regression
equation shows that JND distance is 1/11 of that measure (Fig. 3d).
LF and IACC show a similar behaviour, with an average standard
deviation of 1.4 JNDs and a maximum of 2.5 JNDs. Unlike the energetic parameters, whose variations are mostly dependent on
sourcereceiver distance, the variations of LF and IACC are more
sensitive to relative sourcereceiver position and to wall proximity
which may cause strong lateral reections [18]. In fact, the above
calculations are based on the whole set of receivers of each church,
including side aisles, transepts and alike. If only the receivers in the
main listening area are taken into account, the average standard
deviation lowers, and a further reduction may be obtained by considering only receivers far from reecting walls. In this case the
standard deviation becomes 0.9 JNDs for LF and 1.0 JNDs for IACC,
with maximums of 1.3 and 1.6 JNDs, respectively, suggesting a
more uniform distribution with differences hard to be detected
subjectively (about 2/3 of the churches show a standard deviation
below the JND). A subset of churches was selected to analyse the
variations of receivers located close to walls, showing higher mean
values (due to the stronger lateral reections) and standard deviations generally below the JND limit (Table 4), indicating again substantial uniformity.
In terms of variation as a function of sourcereceiver distance
two opposite trends were observed for LF. In elongated churches
LF mildly decreases as the distance grows, with a JND distance
varying between 6.5 m and 34.5 m with an average of 16.9 m.
The JND distances show poor correlation with architectural parameters but are systematically larger than the corresponding values
calculated for monaural parameters. In squared and central-plan
churches LF increases as the distance grows, with a JND distance
varying between 3.8 m and 19 m and an average value of 7.8 m.
Even in this case no signicant correlation with architectural measures was observed but JND distances were always bigger than the
corresponding values calculated for monaural parameters.
These observations were conrmed by simulations carried out
with the software CATT-Acoustic. Three simple shoebox rooms of

Author's personal copy

382

F. Martellotta et al. / Applied Acoustics 70 (2009) 378388

b
JNDdist = 0.45Dav
R2 = 0.531

20

20
18

JND distance for G (m)

JND distance for EDT (m)

25

15

10

JNDdist = 0.383Dav
R2 = 0.6062

16
14
12
10
8
6
4
2
0

0
0

10

20

30

40

50

Average Source-receiver distance (m)

10
JNDdist = 0.103 L
R2 = 0.649

20

30

40

50

9
JNDdist = 0.092 L
R2 = 0.515

8
JND distance for TS(m)

JND distance for C80 (m)

10

Average Source-receiver distance (m)

7
6
5
4
3
2

7
6
5
4
3
2
1

0
0

20

40

60

80

100

20

40

60

80

100

Nave length (m)

Nave length (m)

Fig. 3. Plot of JND distance calculated for EDT (a) and G (b) vs. mean altar-listener distance, and for C80 (c) and Ts (d) vs. nave length.

Table 4
Mean values and standard deviations (in brackets) of average values from 500 Hz to 2 kHz of LF and 1-IACC measured in a selection of four churches as a function of the distance
from walls
C1

Walls
Centre
Difference

C2

C3

C4

LF

1-IACC

LF

1-IACC

LF

1-IACC

LF

1-IACC

0.31 (0.048)
0.23 (0.073)
0.078

0.68 (0.045)
0.49 (0.058)
0.192

0.27 (0.046)
0.19 (0.024)
0.076

0.56 (0.12)
0.49 (0.037)
0.119

0.36 (0.072)
0.26 (0.036)
0.099

0.79 (0.043)
0.76 (0.072)
0.029

0.35 (0.032)
0.20 (0.042)
0.155

75.8 (0.09)
0.66 (0.012)
0.095

xed height (20 m) and different plan (30  30 m, 20  40 m, and


20  60 m) were analysed. Simulations were carried out at 1 kHz
only, all surfaces were assumed to have an absorption coefcient
of 0.10 and a scattering coefcient of 0.30. The results show that
the spatial distribution of LF is strongly dependent on room geometry. A signicant dependence on sourcereceiver distance can be
observed when the room is elongated (Figs. 4b and c and 5a), with
a mild decrease of LF as the distance grows (and consequently the
angle of the lateral reection becomes smaller). Conversely, in the
square room (Fig. 4a) the values increase monotonically. A transversal variation, with decreasing LF values moving towards the
centre of the room, can be observed in all the cases, in particular
at positions within 1020 m from the source, where the standard
deviation appears larger than JND (Fig. 5b). It is interesting to
observe that, as a consequence of the later arriving lateral reections, in the wider room the longitudinal variation is slower than
in the narrower rooms.

In order to take into account the effect of columns and pillars,


which frequently divide churches into naves, the simulations were
repeated using the same dimensional ratios but adding a series of
simple rectangular pillars (Fig. 4df). For such surfaces the scattering coefcient was assumed to be 0.7 to account for diffraction effects. In this case the general trend to increase up to a distance of
1020 m from the source and then decrease is still visible both in
the nave and in the aisles (Fig. 5c). However, LF values in the aisles
are markedly higher (by about 0.1) as a consequence of the shadowing effect of the columns (masking frontal sound) and of the
strong lateral reections. This also determines signicant transversal variations in the aisles as a function of source visibility, as illustrated by the high standard deviation (Fig. 5d). Conversely,
transversal variations in the nave are generally negligible, with
standard deviation well below JND, even at points close to the
source. Transversal variations in the main nave are signicant only
when its width is greater than 15 m.

Author's personal copy

383

F. Martellotta et al. / Applied Acoustics 70 (2009) 378388

Fig. 4. Plot of calculated LF distribution at 1 kHz. Top panel: simple rectangular rooms: (a) 30  30 m; (b) 20  40 m; (c) 20  60 m. Bottom panel: rectangular rooms
subdivided into naves: (d) 30  30 m; (e) 20  40 m; (f) 20  60 m.

30

30x30
Standard Deviation (%)

25
Mean LF (%)

15

20
15
30x30

10

20x40
5

20x40
20x60

10

20x60

0
0

10

20

30

40

50

60

50
Nave 30x30
Nave 20x40
Nave 20x60
Aisles 30x30
Aisles 20x40
Aisles 20x60

45

Mean LF (%)

40
35
30
25
20
15
10

Standard Deviation (%)

10

20

30

40

50

60

Longitudinal distance from source (m)

Longitudinal distance from source (m)

15
Nave 30x30
Nave 20x40
Nave 20x60
Aisles 30x30
Aisles 20x40
Aisles 20x60

10

5
0

0
0

10

20

30

40

50

Longitudinal distance from source (m)

10

20

30

40

50

Longitudinal distance from source (m)

Fig. 5. Plot of averages (a) and standard deviations (b) of LF values calculated in the rooms reported in Fig. 8, along transversal axis and plotted as a function of longitudinal
distance; (c) and (d) are the same as (a) and (b) but calculated in the rooms with pillars.

3.2. Effect of geometrical symmetry on acoustic measurements


Six different churches, having different plan and typology, but
all being substantially symmetrical were analysed. The results of
the comparisons between the measurements carried out in 22
pairs of receivers placed in symmetric locations when the sound
source was located in front of the altar on the symmetry axis are
discussed. As well as in the previous discussion the differences be-

tween pairs of corresponding receivers were expressed in terms of


JND (Table 5). It appears that for all the acoustical parameters,
apart from LF, the mean differences are within the JND limit.
Reverberation and strength show the highest accuracy, while
early decay time and energy ratios show relatively higher differences due to their much higher sensitivity to small differences in
the sound eld. Similarly, lateral fraction is sensitive to small differences in time and direction of arrival of early reections and

Author's personal copy

384

F. Martellotta et al. / Applied Acoustics 70 (2009) 378388

Table 5
Mean differences, standard deviations, and condence limits of acoustical parameters measured in symmetrical positions in nearly symmetrical churches (all values expressed in
terms of JND)

Mean
Standard deviation
Condence interval

T30

T20

EDT

G10

D50

C50

C80

Ts

LF

0.19
0.15
0.06

0.28
0.18
0.08

0.69
0.51
0.22

0.22
0.14
0.06

0.34
0.21
0.09

0.67
0.43
0.19

0.50
0.34
0.14

0.62
0.45
0.19

1.10
0.57
0.24

consequently shows an average value (and an upper condence


limit) slightly exceeding the JND limit. When only perfectly symmetrical churches are taken into account the mean difference for
LF lowers to 0.86 JNDs with an upper condence limit of 1.04,
showing a considerable improvement of performance, indicating
that good matching between symmetrical positions may only be
obtained if the room is symmetrical and the receivers are accurately placed.
3.3. Practical considerations
Taking into account the results of the above mentioned analysis a few practical rules may be proposed for the receivers placement. In particular, it appeared that the maximum distance
between receivers that allows an acceptable description of the
spatial variability of monaural parameters may be assumed proportional to the church dimensions and, taking into account the
strictest conditions deriving from Ts, for which JND distance is
L/11, it may be concluded that each receiver may cover a circular
area with a radius equal to the JND distance. Consequently, allowing some superposition among circular areas, the resulting distance between receivers should be about 1/6 of the length of
the main listening volume, approximated by the entrance-to-altar
distance (Fig. 2).
In churches where the nave has an exceptional width (>15 m)
the receivers should be preferably arranged according to a quincunx pattern in order to have some receivers within a distance of
5 m from the walls and take into account the variations in LF and
IACC as a function of the distance from the reecting surfaces. In
narrow churches with pillars or columns dividing the nave from
the aisles, the receivers should be arranged along a single line at
varying distances from the columns in order to account for their effect on lateral reections and inter-aural correlation.
If the church is symmetrical the receivers may be located only
on one half of the church, provided that at least three control
receivers are located symmetrically on the other half of the main
nave. The receivers should be distributed throughout the whole
area if more than 30% of the multi-octave-band averages of the
acoustical parameters (as suggested by ISO 3382 draft [16]) calculated in control points show differences higher than the corresponding JND when the source is in position S1. However, if this
analysis cannot be carried out on-site and the discrepancy appears
after the survey is completed, the lack of acoustical symmetry
should be clearly stated in the measurement report. For non-symmetrical sound sources the control receivers should also be included in order to determine the differences in sound
propagation. In any case, the main line of receivers should coincide
with the centreline of the main pew area, with the rst receiver located at the centre of the third row from the altar.
In addition to the above mentioned rules, in order to analyse
the acoustical behaviour of all the parts of a church, it is recommended to place one receiver close to the sound source (within a
distance of 1 m, in order to analyse what speakers and musicians
perceive), and at least one receiver in each secondary volume.
This group should include the sanctuary, transepts, choirs, large
chapels and all those accessible volumes which, because of a par-

ticular combination of acoustical materials and shape, independent of their degree of acoustic coupling with the main volume,
may rise place to a different acoustic behaviour. In these cases
the receiver should be placed at the centre of the oor area corresponding to the secondary volume or in proximity of xed seating positions (such as the ministers seat or behind the altar in
the chancel or at the centre of the stalls in the choir, if present).
In case of stringent time limitations, priority should be given to
the receiver close to the source and to those areas which are
more likely to be occupied by a signicant number of persons
(aisles, transepts, choir).
In case of large domes, especially if the focal point is close to the
oor (or better to peoples heads), at least six receivers should be
used in order to detect possible irregularities in the sound eld
[19]. The receivers should be distributed according to a quincunx
pattern, as reported in Fig. 6. In case of small domes only the rst
three positions could be used. When the symmetry of the church is
not perfect or when a more detailed analysis is required, additional
positions mirrored along the main axis may be used.
Combining all these things together, the actual minimum
number of receivers may be dened and a convenient layout organized (Fig. 2). Even though the use of an optimum number of
receivers, obtained by doubling the minimum number of receivers
in the MLA is strongly encouraged, in most cases the best solution
could be an adapted conguration, with a number of receivers
higher than the minimum but tted to the geometry of the church
under analysis and to the selected sound sources, ensuring the best
compromise between accuracy, time restrictions, and ease of placing the receivers. A simple rule of thumb to better t the layout of
the receivers to the geometry of the church could be to align
them to the architectural grid (i.e. of the spans), trying to locate
the rst receiver as close as possible to the centre of the third
row of pews. Possible adapted arrangements of receivers are
shown in Figs. 79 for three typical church plans.

01
03

02
04

05

06

Fig. 6. Suggested arrangement of the receivers under large domes.

Author's personal copy

385

F. Martellotta et al. / Applied Acoustics 70 (2009) 378388

4. Source and receiver combinations


The acoustic characteristics of a church are strongly dependent
on the combination of sourcereceiver positions taken into account. A complete description of the acoustic behaviour of a place
should involve the use of the largest set of source and receiver
locations. However, time restrictions often prevent such degree
of detail because moving two or three microphones with their
cables in many points is a very time consuming activity. It was stated in the previous sections that a minimum measurement setup should involve at least two source locations (S1 and another
one depending on the main purpose of the measurement) and
the minimum number of receivers. However, in order to encourage
testing the church with more than two source positions it is suggested that, apart from source location S1 which should be always
used in combination with all the receivers, when more than 2
source locations are used, a reduced number of receivers may be
used in agreement with the scheme reported in Table 6.
The choice of the receivers to be used in the simplied set-ups
should be made by taking into account the principle of the maximum coverage of the MLA including, preferably, the extreme positions (the closer and the farther receiver location). In symmetrical
churches where the receivers are located only in one half of the
MLA the control receivers should be used to ensure a better coverage of the MLA when non-symmetrical sound sources are used.
Examples of combinations of sources and receivers in typical
church plans are reported in Figs. 79.
Given the signicant inuence that occupancy may have on the
acoustics of churches, measurements in occupied conditions are
strongly encouraged in addition to the unoccupied set of measurements. In this case, a single source position (preferably S1) may be
accepted, in combination with at least three receivers located
respectively on the front and on the back of the assembly, and
one located in the chancel area. In order to make comparison between occupiedunoccupied conditions receiver locations should
coincide with those used in the empty conguration. Care should
be taken in order to prevent unwanted noise entering the microphones. The occupants should be distributed throughout the MLA
and their number (together with the area they actually cover)
should be reported in the measurement log.
In any case, a measurement of the background noise in onethird octave bands should be made, possibly during daytime in order to provide a complete description of the acoustic environment

12

11

11
03
08

07

08

05

06

04

04

S1

10
01

09
S6

Fig. 8. Arrangement of an adapted number of receivers in combination with three


source positions in a typical central-plan church with large dome. Source S1 (altar)
should be used in combination with all the receivers. Source S8 (dome) should be
used in combination with receivers from 01 to 06, 11, and 12. Source S6 (organ)
should be used in combination with receivers 01, 04, 06, 08, 09, 11, and 13.

of the church during the celebration and to allow STI calculations


by means of impulse responses. Measurements should be preferably made at each receiver position, but the minimum conguration
should include at least three receivers located respectively on the
front and on the back of the assembly, and one located in the chancel area to take into account different parts of the church in their
relation with the envelope and its opening. Each measurement
should last for at least ve minutes.
5. Measurement equipment
Four different equipment set-ups are proposed as a function of
the purpose of the measurement, of the instruments available and
of the time given for the measurement. The basic set-up is mostly
intended for a simple and quick characterization of the place by
means of the measurement of the reverberation time and monaural acoustic parameters, consequently it is also the less demanding
in terms of equipment and also non-repetitive sound sources may
be used. The intermediate set-up is intended for a more detailed

10

S4

03

S8

02

05

07
06

S5

S1

09

12

13

01

14

15

02
13

Fig. 7. Arrangement of an adapted number of receivers in combination with three source positions in a typical basilican-plan church. Source S1 (altar) should be used in
combination with all the receivers. Source S4 (pulpit) should be used in combination with odd receivers from 01 to 09 and with receivers from 10 to 12. Source S5 (choir)
should be used in combination with receivers 01, 02, 05, 08, 09, 10, 12, and 15.

Author's personal copy

386

F. Martellotta et al. / Applied Acoustics 70 (2009) 378388

11

12

10
S6

S1

07

09
08

05
06

03
04

01

14

13

S5

02

S10

15
Fig. 9. Arrangement of an adapted number of receivers in combination with four source positions in a typical single nave church. Source S1 (altar) should be used in
combination with all the receivers. Sources S5 (choir) and S6 (organ) should be used in combination with receivers 01, 02, 05, 08, 09, 10, 12, and 14. Source S10 (chapel)
should be used in combination with receivers 01, 05, 13, and 15.

Table 6
Recommended sourcereceiver combinations

5.1. Basic set-up

Source

Receivers in the
main volume

Receivers in secondary volumes

S1, altar

All + 3 control rec.

All

50% + 2 control rec.

At least one in the same volume


where the source is located

S2,
S3,
S4,
S5,
S6,

high altara
ambos
pulpit
choir
organ

S7, congregation

S9. . ., extra

At least 2 receivers at
10 m from the source
At least 6 receivers
under the dome
At least 2

PA system

50%

S8, dome

All
None
At least one in the same volume
where the source is located
None

If S2 is used as reference position in place of S1, then the same receiver


combination suggested for S1 should be applied.

characterization of the place by means of the measurement of


monaural acoustic parameters and lateral fractions. In this case
only electro-acoustic sound sources are accepted but the octaveband analysis is restricted to the six bands from 125 Hz to 4 kHz.
The advanced set-up is intended for the more complete characterization of the place by means of the measurement of the impulse
responses and of all the acoustic parameters. In this case the octave
bands taken into account should include 63 Hz and 8 kHz and multi-channel microphones should be used (including 1st order ambisonic, higher-order ambisonic and microphone arrays). The
rendering set-up is intended to measure high quality impulse responses, allowing both the calculation of all the acoustic parameters and, above all, a realistic laboratory reproduction, by means
of convolution with anechoic material, of the actual listening conditions observed in the church. High signal-to-noise ratio and a at
response of the system (up to 16 kHz) are the additional requirements in this case. The playback systems may include headphone
and transaural presentation for binaural signals, ambisonic decomposition on 2D or 3D loudspeaker arrays for both 1st order or
higher-order ambisonic measurements respectively made with
four-capsules or spherical arrays of microphones [20,21]. Details
for each conguration are reported below and a summary is
reported in Table 7.

Sound source: every sound source complying with the ISO 3382
requirements [8], including impulsive sources such as blank pistolshots, balloons, etc. In exceptional cases, when the use of pistolshots and balloons may lead to low signal-to-noise ratio at the
lowest frequencies the organ (if present) might be used as an additional sound source, provided that it is mentioned in the measurement log. In fact, a reverberation time of several seconds should be
scarcely affected by the decay time of low organ stops which is
rather short, normally less than 250 ms, as the volume of the wind
case is rather small compared with the pipes. Low open ue pipes
should be used, and in order to avoid the excitation of only some
separated frequencies of the room, clusters of at least four semitones should be played.
Sound signal to be used with electro-acoustic sources: pink/white
noise, MLS, sweeps in agreement with ISO 18233 [15,22].
Microphone: omni-directional.
Frequency range: from 125 Hz to 4000 Hz in octave bands, or
from 100 Hz to 5000 Hz in one-third octave bands. A sampling rate
of at least 44.1 kHz at 16 bit is recommended.
Measurable parameters: reverberation times (as a function of
the available signal-to-noise ratio), EDT, and (only for impulsive
measurements and omni-directional sources) energetic parameters and stage support indices. In order to calculate strength
(G) the power level of the electro-acoustic sound source must
be measured in anechoic or reverberant room, in agreement
with ISO 3382. Speech Transmission Index (STI) may also be calculated provided that background noise measurements are also
done.
The signal acquired by the microphone may be stored on a highdelity digital recorder in order to be post-processed.
5.2. Intermediate set-up
Sound source: electro-acoustic omni-directional sound source
complying with ISO 3382 standard.
Sound signal to be used with electro-acoustic sources: MLS, linear,
logarithmic, and equalized sweeps in agreement with ISO 18233
[15,22].
Microphones: omni-directional and gure of eight (or B-format).
Frequency range: same as basic conguration.

Author's personal copy

387

F. Martellotta et al. / Applied Acoustics 70 (2009) 378388


Table 7
Summary of the four measurement congurations
Conguration
Basic

Intermediate

Advanced

Rendering

Sound source

Every source complying with ISO 3382


Noise, impulse, deterministic

Microphones

Omni-directional

Same as advanced, but constant amplitude


equalized sweep is preferred
B-Format + dummy head

Frequency
range (Hz)
Sampling
Measurable
parameters

1254000

1254000

Electro acoustic, omnidirectional + sub-woofer


Deterministic (MLS, sweep)
preferably equalized
Omni + gure of 8 (or Bformat) + dummy head
638000

Same as advanced, Lw > 100 dB if V > 50,000 m3

Signal

Electro-acoustic
omni-directional
Deterministic (MLS,
sweep)
Omni + gure of 8

44.1 kHz, 16 bit


T30, EDT and, with restrictions on the
source, C, D, Ts, G, STIa

44.1 kHz, 16 bit


T30, EDT, C, D, Ts, G,
STIa, LF, LG

44.1 kHz, 16 bit


T30, EDT, C, D, Ts, G, STIa, LF, LG,
IACC

48 kHz, 24 bit
Same as advanced + realistic laboratory
reproduction

6316,000

STI measurements based on impulse responses may be made provided that background noise is measured.

Measurable parameters: all the monaural acoustic parameters


(as specied in the Basic conguration) plus lateral fractions LF
and LG.
5.3. Advanced set-up
Sound source: equalized electro-acoustic omni-directional
sound source, complying with ISO 3382 standard, with an additional sub-woofer to cover the lowest frequencies.
Sound signal to be used with electro-acoustic sources: equalised
MLS or equalized sweeps in agreement with ISO 18233 [15,22].
Microphones: omni-directional, gure of eight (or B-format),
and dummy head simulator with binaural microphones. Experimental microphones, such as anemometric pu probes or microphone arrays may be used provided that their characteristics are
clearly specied in the measurement log.
Frequency range: from 63 Hz to 8000 Hz in octave bands, or
from 50 Hz to 10,000 Hz in one-third octave bands. A sampling rate
of at least 44.1 kHz at 16 bit is recommended.
Measurable parameters: same as intermediate conguration plus
inter-aural cross-correlation coefcients.
5.4. Rendering set-up
Sound source: same as advanced conguration. In order to provide a suitable signal-to-noise ratio, the use of a source with a
power level of more than 100 dB is recommended in very big
churches (with volumes larger than 50,000 m3).
Sound signal to be used with electro-acoustic sources: constant
amplitude equalized sweep [15,22] is recommended. However,
any signal capable of providing a S/N ratio of at least 45 dB over
the whole spectrum may be accepted.
Microphones: B-format, dummy head simulator with binaural
microphones. Higher-order ambisonic microphones and any experimental microphone set-up may also be accepted provided that its
characteristics are clearly specied in the measurement log.
Frequency range: from 63 Hz to 16,000 Hz in octave bands, or
from 50 Hz to 20,000 Hz in one-third octave bands. A sampling rate
of at least 48 kHz at 24 bit is recommended.
Measurable parameters: same as advanced conguration plus
allowing a realistic laboratory reproduction, by means of convolution with anechoic material, of the current listening conditions observed in the church.
6. Conclusions
A set of guidelines was proposed to simplify and normalize the
choice of source and receiver locations and to suggest suitable

hardware combinations for acoustic measurements in churches.


The aim was to improve the comparability of measurements made
by different teams and, above all, allow comparisons of the acoustic characteristics of different Christian worship buildings. The proposed criteria, mostly based on the experience of the authors in
measuring acoustic characteristics of a large number of churches,
took into account practical, architectural, and liturgical aspects,
pointing out, where necessary, problems relevant to a particular
religious tradition. Proposals were also supported by quantitative
observations based on the results of previous measurements. The
guidelines were nally improved by the contribution of researchers and specialists of both acoustic measurements and worship
building acoustics. Anyway, further suggestions to improve the
method are welcomed.
Acknowledgements
The authors would like to thank Dr. Davide Bonsi, Dr. David
Lloyd Klepper, Dr. David Lubman, Prof. Jrgen Meyer, Dr. Rendell
Torres, Prof. Michael Vorlnder, and Prof. Teolo Zamarreo for
their invaluable help in reviewing and improving this proposal
with their comments. The authors are also indebted to Most Rev.
Nicola Bux for his precious advice on liturgical problems. This
study has been carried out within the national interest program
of scientic research The acoustics of worship places, funded by
the Italian Ministry of Universities and Research.
References
[1] Carvalho APO. Objective acoustical analysis of room acoustic measurements in
Portuguese Catholic churches. In: Proc Noise-Con 94, Ft. Lauderdale, Florida;
1994. p. 80510.
[2] Carvalho APO. Relationships between subjective and objective acoustical
measures in churches. Build Acoust 1997;4(1):120.
[3] Desarnaulds V. De lacoustique des glises en Suisse. Une approche
pluridisciplinaire. Doctoral Thesis. EPFL; 2002.
[4] Magrini A, Ricciardi P. Churches as auditoria: analysis of acoustical parameters
for a better understanding of sound quality. Build Acoust 2003;10:13558.
[5] Meyer J. Kirchenakustik. Frankfurt am Main2003. Verlag Erwin Bochinsky.
[6] Cirillo E, Martellotta F. Sound propagation and energy relations in churches. J
Acoust Soc Am 2005;118(1):23248.
[7] Zamarreo T, Girn S, Galindo M. Acoustic energy relations in Mudejar-Gothic
churches. J Acoust Soc Am 2007;121:23450.
[8] ISO-3382. Acoustics measurement of the reverberation time of rooms with
reference to other acoustical parameters. Geneva, Switzerland: ISO; 1997.
[9] Pompoli R, Prodi N. Guidelines for acoustical measurements inside
historical opera houses: procedures and validation.
J Sound Vib
2000;232(1):281301.
[10] Chu AWT, Warnock ACC. Detailed directivity of sound elds around human
talkers. Report B3144.6 prepared for Public Works and Government Services
Canada, September; 2001.
[11] Bozzoli F, Viktorovitch M, Farina A. Balloons of directivity of real and articial
mouth used in
determining speech transmission index. In: 118 AES
Convention, 2831 May 2005, Paper 6492.

Author's personal copy

388

F. Martellotta et al. / Applied Acoustics 70 (2009) 378388

[12] Behler GK. Sound source for the measurement of room impulse responses for
auralizations. In: Proc 19 ICA, Madrid, 27 September 2007, Paper RBA-06-009.
[13] ITU-T Recommendation P.51. Articial mouth. Geneva, Switzerland: ITU;
1997.
[14] Anderson JS, Bratos-Anderson M. Acoustic coupling effects in St. Pauls
Cathedral, London. J Sound Vib 2000;236(2):20925.
[15] Mller S, Massarani P. Transfer-function measurement with sweeps. J Audio
Eng Soc 2001;49:44371.
[16] ISO/CD-3382-1. Acoustics measurement of the reverberation time Part 1:
performance spaces. Geneva, Switzerland: ISO; 2004.
[17] Cirillo E, Martellotta F. Acoustics of Apulian Romanesque Churches:
correlations between architectural and acoustic parameters. Build Acoust
2003;10(1):5576.

[18] de Vries D, Hulsebos EM, Baan J. Spatial uctuations in measures for


spaciousness. J Acoust Soc Am 2000;110(2):94754.
[19] Magrini A, Ricciardi P. The acoustic eld under the dome in a central plan
church: measurement and simulation. In: Proc 13th ICSV, Vienna, Austria, 26
July 2006.
[20] Gover BN, Ryan JG, Stinson MR. Measurements of directional properties of
reverberant sound elds in rooms using a spherical microphone array.J Acoust
Soc Am 2004;116(4):213848.
[21] Moreau S, Daniel J, Bertet S. 3D soundeld recording with high order
ambisonics objective measurements and validation of a 4th order
spherical microphone. In: 120th AES Convention, Paris, 2006, Paper 6857.
[22] ISO 18233. Acoustics application of new measurement methods in building
and room acoustics. Geneva, Switzerland: ISO; 2006.

Você também pode gostar