Anhert - Modern Tools In Acoustic Design Of Concert Halls And Theatres.pdf - Akustyka(1) - dzidzia2603

XIII Session of the Russian Acoustical Society Moscow, August 25-29, 2003

Wolfgang Ahnert, Stefan Feistel and Oliver Schmitz

MODERN TOOLS IN ACOUSTIC DESIGN OF CONCERT HALLS AND THEATRES -USE

AND LIMITATIONS OF COMPUTER SIMULATION AND AURALISATION-

ADA Acoustic Design Ahnert, Berlin, Germany

Computer simulation for precalculation of acoustic properties of concert halls and theatres has become more

and more common in the course of the last 10 years. While in the past scale models (1:10 to 1:20) were used to

evaluate the acoustic properties of a hall under construction, now computer models supply results faster and

more conveniently. But scale models are still being used in parallel because of the following limitations of com-

puter models:

• Certain incorrectness with regards to scattering effects

• Unsolved solution for diffraction behaviour

• Limitation in computer-model resolution

The advantages and disadvantages of computer simulation in comparison to scale-model measurements will be

shown. Examples will demonstrate the growing superiority of computer simulation during the acoustic design

phase. Limits of computer auralization will be explained.

Introduction

Knowledge about the acoustical design of halls and open-air spaces has been around for 2000 years

/1/. It has been scientifically underpinned especially in the last 200 years. For enabling the acoustical

properties of such rooms to be more accurately determined beforehand, room-acoustical investigations

in physical models have been carried out since the beginning of the 20th century /2,3/. The signifi-

cance of these investigations has until now remained unquestioned. With the availability of mainframe

computers such predeterminations are now carried out by way of computer simulations and no longer

by dint of physical models. A trailblazing paper in which Krokstad /4/ first described the use of main-

frame computers for the precalculation of the acoustical properties of a hall was published in 1968.

With the introduction of personal computers in the middle of the 80ies of last Century the use of com-

puter programs for precalculating acoustical properties has increased immensely, so that now there are

available, in addition to the programs developed by universities, a series of highly efficient commer-

cially available programs:

• CATT-Acoustic , by Dalenbeck/Sweden, 1988 Version 1, now version 8.1

• ODEON, by Naylor&Rindel/Denmark, 1991 Version 1, now version 4.2

• RAMSETE , by Faria/Italy, 1994 Version 1, now version 2.xx

• CAESAR , by Vorländer/Schmitz/Aachen/Germany, 1998/2001 Version 0.12, 2001 vers. 0.20

• EASE , by Ahnert/Feistel/Berlin/Germany, 1990 Version 1 , 2002 Version 4.0

• Predominantly for electroacoustical purposes software packages have been developed in par-

allel to these programs. These include commercially available programs, such as:

• CADP1, graphic CAD for MSDos-based program, 1983, Version 1.0

• CADP1, full-graphic CAD for Windows 3.1-based program, 1984, Version 2.0

• BOSE-Modeler : first full-graphic CAD MacIntosh-based program, 1986, Version 1, by K.

Jacob, T. Birkle/Bose/USA

• Acousta-CADD , first full-graphic CAD MS-DOS-based program, 1987, Version 1 by A.

Muchimaru Altec Lansing/USA

• Nexo-CADD , full-graphic CAD MS-DOS-based program, 1988, Version 1, 1988, by

Nexo/France, not available anymore

• EASE , full-graphic CAD MS-DOS-based program with pop-up menus, 1990, Version 1,

1990, by ADA, Germany

• CADP2, full-graphic CAD Windows 3.1 based-program, also running under Windows 95, by

863

XIII Session of the Russian Acoustical Society Moscow, August 25-29, 2003

JBL/USA, 1991, Version 1, latest version 1.25

• Acousta-CADD, Version 2, 1992,

• BOSE-Modeler , 1993, Version 4 and updates

• EASE V ersion 2 and updates, 1993,

• Acousta-CADD, 1996, Production terminated

• ULYSSES, by IFB/Germany (P. Hallstein), 1996, Version 1, now windows version 2.41

• CADP2, 1997, Further development stopped

• EASE for Windows, by W.Ahnert, R.Feistel, S.Feistel, 1999, Version 3.0

• EASE for Windows, by W.Ahnert, R.Feistel, S.Feistel, 2002, Version 4.0 with room acoustic

module AURA, based on CAESAR

In the case of all these programs, computer software for electro-acoustical and room-acoustical simu-

lation has been available in form of separate products with only a small overlap in functionality and

objective.

Today, higher demands of the users and the increasing power of available personal computers (PCs)

make it a straightforward approach to combine these two components into an integrated system solv-

ing most of the currently requested simulation tasks. Now EASE4.0 has become a powerful tool of-

fering more than the sum of its parts. It merges the extensive loudspeaker database and the powerful

3D-CAD interface with a fully featured room acoustical simulation tool calculating room acoustical

parameters and binaural impulse responses for auralization.

Room-acoustical simulation software is often limited to loudspeakers of simple and magnitude-only

directivity data. The assumption of sound sources with an ideally spherical directivity is on the one

hand a result of the necessary reference to theoretical approaches for obtaining “clean” room charac-

teristics without source characteristics. On the other hand it is a requirement for the calculation of

room acoustical parameters according to ISO3382. In contrast to that it is doubtless a fact that loud-

speaker manufacturers and engineers planning electro-acoustical installations are mainly interested in

direct sound investigation tools but not in scientific room acoustics. To overcome the limitations of

both approaches in sound field computation, a MS Windows DLL (dynamic link library CAE-

SAR2EASE, called EASE AURA module) was developed. It includes a combined image source / ray

tracing algorithm together with a scattering model.

In this article the pros and contras of a complex computer simulations will be explained in comparison

with results obtained by physical scale-model measurements.

II.

General overview on room acoustical design techniques

A) Scale model measurements

The scale model measurements go back to the 30thies of the last century:

impulse response measurement in a

1:5 scale model in 1934 /2/

The model room is excited by means of a pulse generator and the signal picked up by a microphone or

a model-size dummy head is transferred to an evaluation unit (formerly an oscilloscope, nowadays a

computer). To achieve similarities with the real room, the model wall materials within the frequency

range used in the model have to be adapted in their (absorptive and scattering) behaviour to that of the

real materials in the real frequency range (f model – f real * model scale). Entire databases have meanwhile

864

XIII Session of the Russian Acoustical Society Moscow, August 25-29, 2003

been established to this effect.

The advantages of this scale-model measuring method are:

• High real-time similarities

• Short acquisition times (measuring times)

• Simple comparative measurements in case of changes in the secondary structure

• Scattering and, in part, diffraction effects are detectable

The disadvantages of this scale-model measuring method are, on the other hand:

• The heavy dissipation in air requires costly compensating measures (therefore measurements

have sometimes been carried out even in a nitrogen environment)

• High costs involved in case of modifications in the primary structure of the model (height,

length, width)

• The transducers (see fig: 1) have a finite bandwidth (mostly only 100 kHz; thus there are no

measurements beyond 5 kHz possible in 1:20-scale models)

fig: 1

B) Computersimulation

Geometrical acoustics

The models for all algorithms based on geometrical acoustics theoretically require plane waves and

plane, infinitely sized, locally reacting surfaces with uniform impedance distribution. This allows to

model sound propagation by using sound particles moving along straight lines generating sound rays.

No wave phenomena like diffraction are considered. Although the previously stated conditions are

never met, a model based on geometrical acoustics can be used with limited errors as long as all sur-

faces are large by comparison with the wavelength. This defines a lower frequency limit.

Based on this model two different basic algorithms were derived, namely the ray-tracing and mirror-

image methods. All common room and electro-acoustical programs are based on these models, as for

instance: cone tracing, triangular beam tracing etc. But as prior investigations revealed, it was neces-

sary to extend the model of geometrical acoustics by surface scattering to account for rough or struc-

tured surfaces. Generic mirror-image methods do not support scattering, with ray-tracing it can be

easily implemented.

Ray tracing

A sound source emits sound particles at t=0s . The directivity of the source can be modelled either by

giving each particle a frequency and direction-dependent amount of energy and using a uniform direc-

tion distribution, or by particles having all the same energy, but different density distribution. The

latter way would require a simulation run for each frequency, while the former allows parallel proc-

essing for all frequencies. Frequency-dependent scattering needs some special treatment to allow par

865

XIII Session of the Russian Acoustical Society Moscow, August 25-29, 2003

allel computation.

The sound particles fly along straight lines (rays) only changing their direction when hitting a bound-

ary surface. Here the absorption properties of the wall affect the energy of the particle. When calcu-

lating the new direction of flight for the particle, the frequency-dependent scattering δ is taken into

account. δ specifies the fraction of the reflected energy which is not geometrically reflected but scat-

tered. Different implementations can be found to simulate scattering effectively. Lambert's law which

defines the scattering for an ideally diffuse surface is usually used as directional characteristic.

Finally each time a sound particle hits a receiver, the current energy of the particle is recorded in an

Echogram taking air loss, receiver directivity, and the distance-dependent attenuation for a spherical

wave into account. For receivers modelled by constant-sized volumes like spheres or cubes, the latter

is automatically considered by the room angle decreasing with the distance under which the receiver is

seen by the source, see figure 2.

fig. 2

Usually ray tracing does not take phase into account, because the time resolution of the Echogram is

low and only energy is considered. Additionally the phase shift of the various reflecting materials is

quite unknown. Therefore interference effects cannot be modelled seriously.

Mirror image method

Like in optics a sound source produces a virtual source, when it is positioned in front of an infinite

reflecting surface. If the surface is not reflecting ideally, the mirror image has less power than the

original one. In the generic method each source is mirrored at each surface, see fig. 3. This generates

image sources of first order. These sources recursively mirrored again at all other surfaces generate

sources of higher order. By this procedure a large number of possible mirror images (MI) is calculated.

But not all of these MI are visible for every receiver. To check the visibility the path to each mirror

image has to be traced through the room. This algorithm is limited to low orders of reflection. For

higher orders the calculation time is rising tremendously because of the exponential increase of possi-

ble sources.

Instead of the generic mirror-image method it is possible to use ray tracing to scan the room for valid

paths between source and receiver. Each of these paths represents a valid mirror image. Some care has

to be taken to ensure that no mirror image is detected twice. Different implementations can be found

on the market. One idea is to exclude this case by the same algorithm as in triangular beam tracing.

The room is scanned with triangles which do not overlap and thus any path to a mirror image will only

be found once. Otherwise each new mirror image must be compared with the previously found

sources.

866

XIII Session of the Russian Acoustical Society Moscow, August 25-29, 2003

Using the hybrid method allows to calculate high-order mirror images. Even for rooms with many

walls orders of 20 and above are possible. The calculation time and therefore the accuracy is scalable

in some extent depending on the demands by setting the maximum angle between two rays.

fig. 3

Wave acoustics

In some cases wave-related calculation models are applied to solve room-acoustical problems. For

small rooms or low frequencies FEM/BEM solvers are available. But the use of these models is no

common practice. Although the theory of FEM/BEM is not limited to low frequencies, the number of

mesh nodes and elements increases with the frequency to the power of three. Computer memory re-

quirements and calculation times reach impractical values quite easily.

III.

Room acoustical design techniques in EASE

Room acoustical Simulation with AURA module

In EASE 4.0 these computer-algorithms, based on geometrical acoustics, have been implemented now

in cooperation with the RWTH University Aachen in a module named A nalysis U tility for R oom

A coustics (AURA). This chapter presents the actual implementation of the previously described algo-

rithms.

Ray tracing

In this implementation the source directivity is modelled by a directivity-dependent amount of energy

and an evenly distributed density over all directions. To allow parallel calculation even with different

scattering coefficients, a split technique separating reflected and scattered particles is implemented. To

867

Anhert - Modern Tools In Acoustic Design Of Concert Halls And Theatres.pdf

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika: