. IRCAM Multimedia LibraryMédiathèque de l'Ircam © Ircam, 1996-2008 - Research Institute in Acoustics and MusicInstitut de Recherche et Coordination Acoustique/Musique - Paris, France
 | Serveur © IRCAM - CENTRE POMPIDOU 1996-2005.
Tous droits réservés pour tous pays. All rights reserved. |
Validation of a Computer Simulation Environment for Room Acoustics Prediction
Olivier Warusfel, Federico Cruz-Barney
ICA 95, Trondheim (Norvège) 1995
Copyright © ICA 1995
Summary
The aim of this paper is to present some results on a validation campaign of a
computer simulation environment developed at IRCAM. A brief description of its
characteristics is made followed by the organization of the data base used for
the validation. The paper points out validation steps concerning the
directivity model of the source, the influence of grazing incidence over seat
rows and the spatial distribution of the late energy.
Introduction
The acoustics laboratory at IRCAM has developed a computer simulation
environment that combines various physical models of sound propagation such as
specular reflections, diffuse reflections and diffraction. In 1991, a
measurement campaign was undertaken by IRCAM in several concert halls and opera
houses throughout Europe. This data base has been used to validate the
performances of this environment on the basis of objective quantities such as
energy-time curves and the temporal and spatial distribution of early
reflections and late reverberation.
Description of the Computer Simulation Environment
The computer simulation environment developed at IRCAM, described in
[Mal86][WJ92], combines different physical models of acoustic propagation to
estimate the energy time curves at different receiver locations. The time
sections of these curves are built via the following steps. The early energy
distribution is estimated by the association of diffraction, plane wave
propagation with specular reflections and diffused propagation. The diffraction
model is used in case of direct sound geometrical masking. The plane wave
propagation is simulated with a source image method up to second order
reflections and a cone method for higher orders. The propagation of diffused
energy between the different room boundaries and its imputation to receivers
are computed according to the space discretisation of Kuttruf's integral
equation [Kutt73]. Assuming an exponential decay of the reverberation, the
reverberation time together with the spatial distribution of the late
reverberant field may be obtained from procedures described in [Gilb81][Mal86].
This combination of methods is derived from preliminary studies showing the
complementarity of these models for the estimation of the different criteria
linked with the acoustical quality of a hall.
The Measurement Campaign
The room acoustics laboratory at IRCAM has undertaken a measurement campaign in
several european concert halls, auditoria and opera houses [WJ92]. The interest
of this campaign is, on the one hand, to gather a data base which permits to
make systematical comparisons with the results estimated by the computer
simulation environment and on the other hand, to verify our knowledge in
auditory perception of room acoustical quality.
In order to follow a rigorous comparison between rooms, an experimental
protocol has been adopted for the ensemble of the campaign. It responds to the
following points:
-
-
- Spatial distribution of sources and receivers: A systematic spatial
disposition of sources and receivers has been applied to all rooms so as to
analyse the dependence of the acoustical quality on source and receiver
positions. Impulse responses have been measured for every source/receiver
couple.
-
- Directivity and orientation of transducers: the calculation of
certain criteria need temporal and/or spatial weighting of reflections. Hence
the measurements were done with different microphone sensitivities:
omnidirectional, cardioide and bidirectional (figure of 8). Moreover, as the
directivity characteristics of musical instruments privilege the excitation of
different walls, it appeared important to get a more detailed information on
the acoustical behaviour of the room than what can be afforded with an
omnidirectional source [Kah94]. For that purpose, several orientations of a
directive source were systematically studied (forward, backward, upward and
towards left or right), meaning that a high number of measurements of
source/receiver couples had to be performed.
For the validation of the simulation environment, the geometrical shape of the
different halls is represented in the form of plane facets characterized by
absorption and diffusion coefficients depending on the type of material. Source
and receivers are set out in the model according to the spatial disposition and
orientation used for the measurement campaign.
| HALL | OBJECTIVES MEASURES | SIMULATION
|
|---|
| Number of Src-Rev couples / Number of measures | Number of facets
|
|---|
| Louvres | 95 couples / 195 measures | 200 facets
|
|---|
| Orsay | 89 couples / 177 measures | 219 facets
|
|---|
| Pleyel | 420 couples / 816 measures | 242 facets
|
|---|
| Musikverein | 300 couples / 600 measures | 248 facets
|
|---|
| ConcertGebouw | 341 couples / 630 measures | 254 facets
|
|---|
| ChampsElysées | 419 couples / 881 measures | 504 facets
|
|---|
Table 1. Number of measurements and simulated halls facets of the data
base
Simulation of the Source Directivity
The validation of the different models of acoustic propagation used in this
simulation environment, needs first to verify the accuracy of the simulated
source radiation because it will influence all the process. Especially, the
total power radiated will affect the reverberant field, while the principal
direction of emission will modify the direct sound and first reflections,
consequently affecting acoustical criteria depending on early energy. This
point is of particular importance since for part of the measurements the
receivers were located outside the loudspeakers main radiation lobe (side,
backward and upward orientations of the source).
In order to be compatible with most standard characterization data of
loudspeakers, a simple radiation model has been introduced. The user may
indicate either a directivity index DI, the vertical and horizontal dispersion
angles at -3dB or -6dB, or different combinations of these elementary models of
the source. The DI and source power of the loudspeaker used for measurements
were obtained from a standard 20 points measurement method, described by the
norm AFNOR NF S 31-026. The aperture angles at -3 and -6 dB were directly
determined from the polar diagrams of the source measured for each 1/3 octave
band.
The validation of the source directivity model is carried out by comparing the
measured and the estimated energy of the direct sound FOD (using a 5ms window)
for all source/receiver couples over three octave bands: 250Hz, 1KHz and 4KHz.
Means and standard deviations of all source/receiver couples in the 3 octave
bands were calculated.
At 250Hz, for receivers located at the floor seating area, an over estimation
of the direct sound was found. Moreover, it showed a positive correlation with
distance, leading to a mean error of 7dB and a standard deviation of 4dB. This
observation is linked to the excess attenuation of sound propagating accross
the seats under grazing incidence. From the correlation analysis a correction
factor could be derived for each hall (table 2). This correction factor depends
on the mean elevation angle of incidence of the source, and reaches its maximum
for a mean elevation of approximately -2.5deg.. These values seem coherent with
the curves reported in [DL94].
| HALL | grazing incidence
correction (dB/m) | Mean Elevation Angle
of Incidence | STD(dB)
without correction | STD(dB)
with correction
|
|---|
| Louvres | 0,30 | +15° | 1,60 | 0,97
|
|---|
| Pleyel | 0,50 | -2,6° | 3,78 | 1,00
|
|---|
| Musikverein | 0,36 | -5,0° | 3,00 | 1,50
|
|---|
| ConcertGebouw | 0,30 | -6,3° | 2,66 | 1,36
|
|---|
Table 2. Influence of grazing incidence at 250Hz octave band on the
error estimation of direct sound. Correction factor, mean elevation angle and
standard deviations of the error estimation (with and without the corrections)
are indicated for each hall.
At 1KHz and 4KHz octave bands, the mean difference on FOD is less than 1dB
showing that the radiated power of the simulation model is correct (Table 3).
Yet, the dispersion of the differences noticed for the 4kHz octave band shows
that the model is not well adapted to describe the field radiated by the source
because of the presence of three pronounced lobes in the vertical plane. Better
results are obtained when considering only measurements for which the receiver
faces the main directivity lobe of the source (|Azimut| < 60deg. and
|Elevation| < 5deg.). Still, in that case, it is difficult to separate the
respective influences of the source model accuracy, the room model
approximations (degree of geometrical fidelity) and the measurement
uncertainties (source and receiver location and orientation, repeatability of
the measurement system, ...).
| All Src/Rev couples | |Azim| < 60° |Elev| < 5°
|
|---|
| 1KHz | 4KHz | 1KHz | 4KHz
|
|---|
| MEAN (FODsim - FODmeas)[dB] | -0,5 | 0,8 | 0,7 | 1
|
|---|
| STD (FODsim - FODmeas)[dB] | 2,6 | 3,8 | 2,1 | 2,2
|
|---|
Table3. Means and standard deviations of the error estimation (FODsim -
FODmeas), computed for all Source/Receiver couples and for specific couples
where the receiver faces the main directivity lobe of the source.
Simulation of the Spatial Distribution of the Late Reverberation
This part is dedicated to the estimation of the late energy distribution within
the hall. In the process, after a transition time, the cone tracing method and
the direct diffusion processes are stopped. The residual intensities present on
the boundaries are then driven by a transition matrix that describes the
behaviour of the room under Lambert's law diffusion hypothesis. The
amplification of this residue during the exponential decay is calculated for
each boundary and the corresponding energy contributions are checked on each
receiver location. Hence the late energy distribution will depend on the
following parameters:
-
-
- - distribution of the residual intensity when the cone tracing method is
stopped,
-
- - amplification of this residue during the exponential decay,
-
- - geometrical coupling of the receiver with each surface.
The evaluation is carried out comparing estimated and measured cumulated
energies of reverberation tails after a transition time (counted from the time
of emission of the source pulse). The minimum transition time is set according
to the maximum source/receiver distance. Different transition times were
studied to evaluate their influence on the prediction of the spatial
distribution of the late reverberant field. Table 4 presents the results for a
transition time of 280ms for two room examples: the Salle Pleyel and the
ConcertGebouw. The spatial distribution is described by the values collected on
different receivers regularly spaced (5,6 m) in the audience area. The
homogeneity of the distribution is evaluated for both the room and the
simulation model by its ambitus and its standard deviation. The correlation
between estimated and measured distributions is computed.
| Residual energy after 280ms | 250Hz | 1kHz | 4kHz
|
|---|
Pleyel
(20 receivers) | max-min [dB] room\sim | 6.1\6.2 | 4.6\7 | 4.6\6.8
|
|---|
| std [dB] room\sim | 1.54\1.7 | 1.2\1.9 | 1.3\1.8
|
|---|
| Correlation (room\sim) | 0.56 | 0.88 | 0.84
|
|---|
ConcertGebouw
(18 receivers) | max-min [dB] room\sim | 3.4\1.6 | 1.4\1.9 | 1.9\2.7
|
|---|
| std [dB] room\sim | 0.9\0.5 | 0.4\0.5
| 0.5\0.7
|
|---|
| Correlation (room\sim) | 0.55 | 0.45 | 0.76
|
|---|
Table 4. Distribution of the late reverberant energy cumulated in the
time interval [280ms ; [[proportional]][ for two rooms. Comparison of the
ambitus within the room and the simulated model. Correlation between measured
and simulated spatial distribution of the late energy.
These results show that the model is able to predict the heterogeneities of the
distribution of late energy although it tends to over estimate the dispersion.
The correlation reaches high values when the dispersion is significant. These
observations hold for the measurements carried out with figure of 8
microphones. Moreover, computations made with an earlier transition time
(130ms) still allow a good estimation of what the late distribution of sound
within the room will be. This remark is interesting regarding the fact that,
for such an early transition, the boundaries of the room have only been
"illuminated" by low order rays (< 3).
Conclusions
A systematic validation program is currently performed on a software
environment dedicated to room acoustics prediction. The first steps of the
validation have been pointed out and some interesting results have been
obtained regarding the modelling of a directive source and of the spatial
distribution of the reverberant energy in a hall.
The importance to take into account the radiation of a source for a relevant
acoustical diagnosis, has lead to perform measurements with a directive source
oriented in different directions, and to introduce a directivity model in the
prediction system. This model proved to be efficient for the simulation of the
acoustical power radiated by the source, and for the radiation description up
to 2KHz. For high frequency bands, the presence of pronounced lobes needs a
more refined model unless the measurements are restricted to direct sound
incidence inside the main lobe. The results have confirmed the need for the
introduction of a correction for the excess attenuation of sound propagation
under grazing incidence over seat rows. The diffuse propagation model provides
an efficient estimation of the spatial distribution of the late reverberant
field.
Further studies will focus on the validation of the prediction of acoustical
criteria linked to the early energy distribution.
References
[Mal86] C. Malcurt. PhD Thesis. Université de Toulouse,
1986
[WJ92] O. Warusfel & J. P. Jullien. Une campagne de mesures
objectives et perceptives en acoustique des salles. proceedings of the
2nd french conference on acoustics, Arcachon, France, April 1992
[Kutt73] H. Kuttruff. Room Acoustics, Applied Science Publishers,
London, 1973
[Gilb81] E.N. Gilbert, An iterative calculation of reverberation time. J.
acoust. Soc. Am, 69,178, 1981
[Kah94] E.Kahle. Influence of size and composition of the orchestra on the
perception of room acoustical quality. Proceedings of the Wallace Clement Sabine Centenial Symposium, Cambridge, Ma.,U.S.A. J. acoust. Soc. Am., pp. 207-210, June, 1994
[DL94] W. J. Davies & Y. W. Lam. New attributes of seat dip attenuation.
Applied Acoustics, 41, pp. 1-23, 1994
____________________________
Server © IRCAM-CGP, 1996-2008 - file updated on .
____________________________
Serveur © IRCAM-CGP, 1996-2008 - document mis à jour le .