You are on page 1of 22

Simulation

Article
Simulation: Transactions of the Society for
Modeling and Simulation International
2015, Vol. 91(9) 787–808
Integrating acoustic simulation Ó 2015 The Author(s)
DOI: 10.1177/0037549715603480
in architectural design workflows: sim.sagepub.com

the FabPod meeting room prototype

Brady Peters

Abstract
Sound is an important part of our experience of buildings. However, architects design largely using visually based tech-
niques and largely for visual phenomena. Aiming to address this problem, the research presented in this paper proposes
four digital design workflows that integrate acoustic computer simulation into architectural design. These techniques
enable architects to design for both visual and acoustic criteria. The goal is to develop rapid and accessible workflows
for architects that allow acoustic performance to be tuned as geometry and materials are modified at the scale of the
room, and also at the scale of the surface. The discovery and testing of these techniques takes place within the design of
the FabPod, a semi-enclosed meeting room situated within an open-plan working environment. The project builds on
previous research investigating the design principles, the acoustic performance, and the fabrication methods of hyperbo-
loid surface geometry. Four design workflows were developed: two of these investigate the acoustic performance of the
room and use existing acoustic simulation software, and the other two workflows investigate the acoustic performance
of the surface and use custom-written scripts to calculate and visualize sound scattering. This paper presents the back-
ground to the study, outlines the digital workflows, describes how they integrate acoustic simulation, and shows some
of the data produced by these simulations.

Keywords
acoustic simulation, architectural acoustics, architectural design, design workflows, simulation visualization, computer-
aided design, performance-based design

1. Introduction adoption of existing analysis software or the development


of bespoke analysis computer programs.4
The role of the architect is to conceive of a building prior The FabPod is a prototype meeting room for the new
to, and separate from, the construction of that building. To Design Hub building at RMIT University in Melbourne,
do this architects have traditionally used drawing tech- Australia. The FabPod is a free-standing meeting room
niques to explore, collect, and communicate their concepts situated in an open-plan working environment. The chal-
about how the building should be. In contemporary archi- lenge was not only how to design the meeting room to
tectural practice, digital design processes have replaced the have optimal acoustic performance, but additionally how
use of pen and paper and introduced new methods for the to design the structure so it could positively contribute to
conception and exploration of building form. While draw- the acoustic performance of the open-plan working envi-
ing on paper and the construction of physical models are ronment in which it was situated. Meeting rooms are
inherently visual processes and therefore do not accommo- places where decisions are made, creative ideas are
date designing for sound, the digital design environment
can have an aural component.1 Computer simulation tech-
niques have been implemented in acoustic engineering John H. Daniels Faculty of Architecture, Landscape, and Design,
software that gives feedback on digital models of build- University of Toronto, Canada
ings;2 however, there are no architectural design tools that
Corresponding author:
include sound.3 As architects become more computation- Brady Peters, John H. Daniels Faculty of Architecture, Landscape, and
ally proficient, there is a tendency to introduce perfor- Design, University of Toronto, 230 College Street, Toronto, ON, Canada.
mance feedback into the design process either through the Email: brady.peters@daniels.utoronto.ca
788 Simulation: Transactions of the Society for Modeling and Simulation International 91(9)

constructed, and where listening and communicating are geometry. This research project contributes knowledge to
critically important, and so the acoustic performance of the integration of acoustic simulation in the architectural
meeting rooms is an important design criterion. design process and the acoustic design of meeting spaces
Unfortunately, the literature surrounding the acoustic and of open-plan work environments. While the results of
design of successful meeting rooms is not as extensive as this research are specific to this application, it is hoped that
the literature for the acoustic design of concert halls. In the design workflow and simulation procedures, as sum-
this project, the acoustic criteria for the meeting room marized in this paper, will be applicable to a wide range of
were developed from several sources relating to meeting architectural projects involving acoustic simulation.
rooms, offices, and classrooms. Although open-plan work-
ing environments is an area of on-going study, the litera-
ture is much clearer in terms of criteria that define good
2. Background
acoustic performance.5 The FabPod is a prototype meeting room. It is designed to
The FabPod project has several research aims. One of have excellent acoustic performance as a meeting room
the primary aims is to create a series of flexible digital and to improve the acoustic performance of the open-plan
workflows that would allow for design decisions to be working environment in which it is situated. The back-
deferred until late in the design process. The hypothesis ground research for the project focuses on the design of
was that as the complexities involved in the geometry and meeting rooms, open-plan working environments, the use
fabrication were encoded into the logics of a computa- of architectural design methods for sound, and the simula-
tional design system, there would then be more time to tion of sound. The literature review shows that acoustically
explore design options and develop a better meeting room. well-performing meeting rooms have reduced room reso-
A second aim was to integrate acoustic computer simula- nances, no flutter echoes, appropriate reverberation time,
tion into these design processes. The integration of design, and a uniform sound field, which is achieved through the
performance, and fabrication meant that a certain amount design of the form of the room and its surface geometry
of negotiation between competing parameters was neces- and material specification. The performance of the open-
sary. A third aim was to design and build an acoustically plan working space is driven by an appropriate reverbera-
well-performing meeting room. To do so, the consider- tion time, a reduced distraction distance, and a high rate of
ation of acoustic performance at the level of the room and spatial decay of sound from one area to another. In terms
the geometry and materiality is considered. To predict this of architectural design tools for sound, most traditional
performance, four acoustic design workflows were devel- architectural design methods do not allow the prediction of
oped. These workflows are presented and discussed in appropriate acoustic parameters to design for these condi-
relation to digital design and acoustic criteria and some tions. The use of computer simulation is necessary to
future directions are proposed. A fourth aim was to extend engage with the phenomenon of sound and to predict the
previous research on hyperboloid structures.6 acoustic performance of the meeting room and its sur-
The project develops knowledge in terms of the geo- rounding environment.
metric definition of, the development of design tools for,
and the fabrication of hyperboloid structures. Previous
research by the authors has shown hyperboloid geometry
2.1. The acoustic performance of meeting rooms
has unique acoustic properties and can be designed to scat- Meeting rooms are communication spaces. Bradley7 writes
ter sound in different ways.6 An aggregate structure com- that ‘‘the goal of good acoustical design for classrooms
posed of cells of different materials can further modify the and meeting rooms is to make possible clear, accurate and
acoustic performance. Specifically, this paper will present relaxed speech communication.’’ Bradley further adds that
research into the development of digital design workflows good acoustical design of meeting rooms should optimize
that incorporate acoustic computer simulation in the design room acoustics and minimize unwanted noise so that
of a meeting room with hyperboloid geometry that is situ- effective speech-to-noise ratios are maximized. Bradley
ated in an open-plan working environment. These experi- observes that rooms are still designed in terms of rever-
ments have provided the opportunity to obtain several beration time, which only indirectly relates to critical
types of data. Firstly, they offer findings as to new forms room acoustic details. Doelle8 and Vorlander1 separately
for meeting spaces from the perspective of acoustic perfor- identify several fundamental requirements in designing
mance. Secondly, they give guidance on the benefits of rooms for speech. The sound signal itself should have ade-
having an acoustically considered meeting pod situated quate loudness and there must be a high signal-to-noise
within an open-plan work environment. Thirdly, they offer ratio. The background noise levels must be low enough to
observations on how acoustic simulation can be integrated not interfere with the listening environment. The rever-
into a digital architectural design workflow, with particular beration characteristics must be appropriate. While some
focus on the early design stage. Fourthly, they offer further reflections are detrimental to the acoustic performance,
evidence as to the acoustic properties of hyperboloid such as long delayed reflections, flutter echoes, and
Peters 789

focusing, the temporal structure of the early reflections than optimal performance, and causes an increased rever-
can add to speech comprehension. Finally, the resonances beration time.16 If the sound pressure is equal in all parts
of the room should be reduced as the distribution of room of a room and sound waves are traveling in all directions,
resonances and the corresponding effect of sound colora- the sound field is said to be ‘‘homogenous’’ or ‘‘diffuse’’.8
tion can detrimentally affect acoustic performance. Sound diffusion promotes a uniform distribution of sound,
Many meeting rooms are small, square, or rectangular accentuates the natural qualities of music and speech, and
in form, and are often only sparsely furnished. These char- prevents the occurrence of undesirable acoustic defects.
acteristics can lead to poor room acoustic performance. Sound diffusion can be created through the creation of sur-
Small, rectangular rooms with hard surfaces develop reso- face irregularities, alternating between sound-reflective
nances between parallel reflecting surfaces, and when the and sound-absorptive surfaces, and the irregular distribu-
dimensions along the major axes are similar this acts to tion of differing sound-absorptive surfaces. The overall
amplify these resonant effects.9 These resonances cause dimensions of the surface irregularities or sound-
certain frequencies to be amplified by up to 20 dB and absorptive materials must be comparable to the wave-
cause the timbre of voices to be changed due to massive length of sound. Surface irregularities also help to reduce
‘‘coloration’’. These effects are annoying and decrease flutter echoes, which occur when sounds persist locally
speech comprehension.10 Noise in meeting rooms is pre- due to multiple reflections between hard, parallel wall sur-
dominantly self-generated. The room does not make any faces.9 In the last decade there has been an increased inter-
sound itself; however, the room determines how the sound est in characterization of sound scattering.17 It has been
arrives at the sound receiver. The lack of sound absorption shown that the correct modeling of sound scattering is nec-
causes people in these rooms to raise their voices as soon essary for the accuracy of geometric room acoustic mod-
as necessary communication between them takes place. els.18 There has also been increased understanding on how
This is known as the ‘‘Lombard Effect’’, the phenomenon important sound scattering is for the acoustic performance
in which speakers increase the level of their vocal output of a space. It has been suggested that one of the reasons
as well as several other vocal changes in the presence of that traditional concert halls sound better than modern
loud background noise in order to improve the intelligibil- halls is because of the greater degree of surface detail that
ity of the speech signal.11 occurred at a variety of scales to provide sound scattering
In his study of classrooms, Bradley12 maximized intel- across a range of frequencies.19 Therefore, in recent years,
ligibility as a function of reverberation time. He found that there has been the development of a standard measure for
intelligibility was maximized at a reverberation time that sound scattering.20
depends on the signal-to-background-noise ratio in a range The FabPod project seeks to create better room acoustic
of 0.2–0.8 seconds. A low reverberation time and signal- performance for meeting rooms through non-parallel wall
to-noise ratio is important to achieve good speech intellig- surfaces to reduce flutter echoes, a non-rectangular overall
ibility.12 More recently it has been found that it is critical geometry to reduce room resonances, a complex interior
to have strong early reflections of sound as well; simply surface to provide sound scattering, and a balance of mate-
striving for a low reverberation time will decrease the rials that not only helps to tune the reverberation time, but
speech intelligibility.13 Human hearing is more sensitive also aids and controls early reflections.
to reflections arriving from lateral directions (at about 45°)
than those arriving from the front or rear.14 Reflections
from above are masked more effectively by the direct 2.2. The acoustic performance of open-plan
sound than are lateral reflections. Sound reflections that
working environments
reach the listener within 50 ms of the direct sound increase
the intelligibility of speech owing to the ability of the ear Two categories of sound should be distinguished. The first
to integrate those sounds. At low levels reflections mani- is sound that serves communication or well-being. For
fest themselves only by an increase of loudness of the total example, speaking and listening, or listening or playing
sound signal, by a change in timbre, or by an increase of music. These sound signals should be transmitted in such
the apparent size of the sound source.15 The target value a way that a good timbre is achieved and speech is under-
for reverberation time for the meeting room was set at standable; these are the principle tasks of room acoustics.
about 0.6 seconds. This number was developed consider- The second is sound that is a nuisance or that is harmful to
ing the research presented by Bradley7 and Bradley and our health.21 This sound could be unwanted sound from
Yang.13 outside the room, unwanted sound from other sound
Another problem that often affects meeting rooms is sources in the room, or distracting sound caused by the
the uneven distribution of absorbing materials. Acoustic sound source interacting with the room itself. This is
absorbing materials, if there are any, are often placed only noise. It is often the balance of positive sound and nega-
on the ceiling. An uneven distribution of absorbing mate- tive noise that determines the acoustic performance of a
rial causes the absorbing material to work at much less working environment.
790 Simulation: Transactions of the Society for Modeling and Simulation International 91(9)

The trend in workspace design is towards forms that sound from one side of the pod to the other. Thirdly, the
promote and support synergies between individuals, teams, material composition of the outside of the pod will help to
and across group boundaries.22 As new technologies allow tune the reverberation time of the open-plan work space.
work to be decoupled from space and time, these new
working processes demand new forms and spatial concepts
that promote community, entrepreneurship, contact, and
2.3. Architectural design tools for sound
exchange.22 The notion of workplace is shifting from that Sound is an important part of architectural design, both in
of being a passive setting for work to that of an active set- terms of achieving its poetic design aims as well as meet-
ting – a tool for getting work done.23 Open-plan work- ing the needs of its programmatic functions, and good
spaces have been shown to have benefits for employers, architects are aware of this and consider sound a key
including a more efficient use of space, a decrease in design parameter. For most of last 2500 years architects
costs, and flexibility in organization. Other benefits for have used empirical guidance for designing the sound of
employees include increased communication and interac- new buildings.27 Wallace Sabine28 introduced the first
tion. However, drawbacks of open-plan spaces include design tool that is able to accurately predict the acoustic
increased noise, distractions, perceived crowding, and performance of new buildings; Sabine’s formula for the
decreased privacy.24 Many of these drawbacks at least par- calculation of reverberation time gave architects the ability
tially relate to the room acoustic performance of the space. to predict acoustic performance prior to construction.
In open-plan spaces, occupants are affected by the sounds Sabine’s study of sound and the dissemination of his
of the activities around them. Too much noise leads to dis- research sparked the development of the field of architec-
traction and too little a lack of speech privacy. tural acoustics. Modernist architects seeking a functional
Traditionally in modern architectural design, absorption justification for everything were quick to seize on the idea
and reverberation time have guided design decisions in that acoustic principles could be a generator of architec-
relation to the design of open-plan working environments; tural form.29 A drawing technique that was used was ray-
however, there is evidence that other types of measure- tracing, where lines were used to visualize and map sound
ments are needed to understand, predict, and evaluate the paths. The invention of electronic recording and measure-
acoustics of open-plan working environments.25,26 The ment devices led to the creation of further acoustic para-
Speech Transmission Index (STI) is a physical quantity meters.27 The construction of scale models was the first
representing the transmission quality of speech with method by which acoustic performance could be simu-
respect to intelligibility. The STI and the spatial distribu- lated. This technique was first used in the 1930s by both
tion of the STI are important to understand acoustic per- Spandock and Jordan.27,30 The use of scale models and
formance in open-plan offices. A new international scale sound sources and microphones allowed for the first
standard describes three key parameters for open-plan listening tests to be performed,31 although scale models
offices: the rate of spatial decay; the distraction distance often have too limited a dynamic frequency range.32
(described as the distance from a speaker where the Developments from the 1930s to 1970s focused on the
STI falls below 0.50); and sound level.5 Target values for refinement of techniques that allowed for the construction
an open-plan offices with good acoustic conditions could of smaller, and therefore more affordable, models.31 The
be spatial decay rate of speech (D2,S) 5 7 dB; main purpose of a physical scale model is the investigation
A-weighted sound pressure level of speech at a distance of acoustic phenomena related to reflection sequences, and
of 4 m (Lp,A,S,4m) 4 48 dB; and, distraction distance the study of diffusion and diffraction.32 Scale models also
(rd) 4 5m.5 offer the added benefit of offering visualization benefits
The design of the FabPod seeks to modify the room for architects and clients.
acoustic performance of the open-plan working environ- Prior to the invention of acoustic computer simulation,
ment in which it is located. The pod geometry and materi- it can see that there are several design tools that have been
ality were adjusted in response to different simulations used by architects: empirical guidance, the calculation of
with the goal to change measured room acoustic criteria. reverberation time using the Sabine formula, the drawing
The experiments were not focused on changing the room’s of two-dimensional (2D) ray-tracing diagrams, and simu-
internal ambient noise or noise from building services, as lation through the measurement of scale models. There
these were considered to be part of the given environment are, however, limitations to these techniques that inhibit
in which the FabPod was inserted. In terms of the the prediction of the acoustic performance of new designs.
FabPod’s acoustic performance, firstly, it isolates distract- While the use of empirical guidance and the careful study
ing meetings from the main work space, allowing of precedent projects has been shown to be a reliable way
extended conversations to occur in a partially acoustically to achieve predictable acoustic performance,33 by design-
isolated pod within the main space. Secondly, by placing ing in this way architects are not able to achieve radically
the pod in the center of the space it breaks the large space new forms, only repeat designs that have been shown to
up into two spaces, thereby increasing the spatial decay of work in the past. The calculation of reverberation using
Peters 791

the Sabine formula is not only slow, but it does not work introduced that detects sound rays and constructs an
for rooms that have odd shapes or unevenly distributed energy decay signature. This allows the calculation of dif-
material,34 and it only calculates a single parameter, rever- ferent acoustic parameters beyond reverberation time. The
beration time, and does not give a spatial response. The ray-tracing technique combined with the Head Related
ray-tracing diagrams so popular in modern architecture Transfer Function (HRTF) to give the Binaural Room
only work in two dimensions and ignore the fact that Impulse Response (BRIR), which can be convolved with
sound is a wave phenomenon. The use of scale models anechoic sound recordings to create auralizations.1
requires very large, very expensive models to be con- Auralizations allow for virtual spaces to be heard before
structed and requires the use of specialized equipment and they are constructed. Ray-tracing is a primary technique in
techniques not part of the architect’s toolkit. None of these a general class of computer acoustic simulation methods
techniques successfully integrated sound into the everyday called ‘‘geometric methods’’. Other methods include the
design toolkit of the architect. image source method,40 which is based on the principle
that a specular (geometrical) reflection can be constructed
by mirroring the sound source in the plane of the reflecting
2.4. Acoustic computer simulation surface. While this method works well for rectangular
Since 1995,35 computer acoustic simulations for architec- rooms, in rooms of arbitrary geometries this number of
tural projects have not only been sufficiently accurate, but image sources increases exponentially with the reflection
efficient in terms of time and cost.34 Today the computer order, therefore making this unsuitable for studying rever-
simulation of room acoustic performance can be grouped berant sound decay.31 Most commonly available computer
into two approaches: wave-based and ray-based.36 Wave- simulation software programs are based on a hybrid
based techniques attempt to solve the wave equation approach where ray-tracing and image source methods are
numerically. Wave-based techniques are more computa- combined to maximize the accuracy of calculations while
tionally intensive, as they involve the discretization of minimizing the time needed to carry them out. However,
either the space of the room or its bounding surfaces and there are some limitations to the acoustic phenomena that
model the interactions between all of the resulting ele- geometric methods can model; a source of uncertainty is
ments. Ray-based techniques assume that sound travels in how to model the effects of surface diffusion and
a straight line and its reflection from surfaces is computed diffraction.17,18
using geometrical methods. These two techniques can be The other type of acoustic simulations includes those
used to study different aspects of architectural acoustic methods that use numerical techniques. These wave-based
performance. Recent tests show that the majority of acous- methods do not solve the wave equation directly, but they
tic simulation software packages provide valid results.37 try to numerically approximate its solution. These numeri-
While most of these simulation software packages have cal methods are the finite element method (FEM), the
limited abilities to actually create geometry, it is usually boundary element method (BEM), and the finite-difference
possible to export geometry from architectural design soft- time-domain method (FDTD).17 A fundamental difference
ware to the simulation software.38 The trend in architec- between these methods and the geometric acoustic simula-
ture is from the creation of 2D drawings to the creation of tion methods is that they divide the space of the digital
highly accurate three-dimensional (3D) models, digital model into small elements or nodes. These elements then
models that could be acoustically simulated; however, interact with each other according to laws of the wave
there are currently no architectural design tools that incor- movement phenomena. Another difference is that this ele-
porate acoustic simulation.3,4 ment interaction must be calculated for each element in a
To understand the acoustic performance of an interior time step sequence. This leads to large calculation times
space, geometrical acoustic simulation is the currently for large numbers of elements and long durations. As the
accepted method.1 The technique of ray-tracing is com- size of these elements has to be much smaller than the size
monly used in the computer simulation of visual scenes of the wavelength, there is potentially a very large number
and forms one of the algorithmic foundations of the digital of elements. Vorlander1 suggests that there are at least six
design environment. A similar technique can be used to nodes per wavelength. For practical use, wave models are
simulate the acoustic scene of a virtual environment. The typically restricted to low frequencies and small spaces.
first computer simulation of sound, and the first to use However, there are several benefits to using numerical
ray-tracing, was proposed in the late 1960s39 and matured methods: firstly, these methods are characterized by their
in the 1990s.31 Using this technique, a large number of calculation of very accurate results, especially at single
rays are traced from a source point – the sound source. frequencies; and, secondly, these methods will compute
These rays reflect off surfaces within the room following the wave effects of diffraction and interference. These
the geometrical law of reflection. These rays are traced simulation techniques much more accurately model sound
until a cut-off point based on a time factor, or decrease in waves and are therefore ultimately preferably due to their
energy level, is reached. A sound receiver, or receivers, is accuracy.
792 Simulation: Transactions of the Society for Modeling and Simulation International 91(9)

The BEM is the most accurate way of studying acoustic separately.45,46 The description of the project in this paper
diffusion.17 The disadvantages of this approach are that it focuses specifically on the development of digital tools
is computationally intensive and slow. Therefore, acoustic and workflows that allow for acoustic performance to be
diffusing geometries are generally studied in isolation of integrated into the architectural design process.
other objects and boundaries. When predicting acoustic
performance in rooms it is more common to use geometric
3. Methodology
models. The models used in BEM simulation are prone to
human error in meshing. FDTD is a widely used simula- 3.1. Research by design
tion technique in many different disciplines and is becom- Architectural research that uses design work itself as part
ing increasingly popular in acoustics.17 Although these of the inquiry process can be called ‘‘research through
simulations have long computation times, the algorithm is practitioner action’’.47 Archer47 writes that sometimes
easy to understand and implement, and it gives a wide fre- ‘‘the best or only way to shed light on a proposition, or
quency range with a single prediction. FDTD simulation principle, a material, a process or a function is to attempt
uses a volumetric rather that surface meshing and can give to construct something, or to enact something, calculated
accurate prediction of scattering.41,42 FDTD was devel- to explore, embody or test it’’. This is a ‘‘systematic
oped in the 1960s for use in studying the electromagnetic enquiry conducted through the medium of practical action;
field but was first applied to acoustics in the 1990s.43 This calculated to devise or test new, or newly imported, infor-
technique can be used for low-frequency studies,43 sound mation, ideas, forms, or procedures and generate commu-
visualizations,44 and for studying the scattering properties nicable knowledge’’.47 In this category, both quantitative
of different geometrical configurations.41 and qualitative research methods may be appropriate.48 In
practice-based research, methodologies are emergent –
that is, the research strategy grows and unfolds from the
2.5. Background summary
practitioner’s interaction with the research question and
From the background study, it can be seen that there are context.49 This project draws on a methodology of natura-
several critical performance considerations for meeting listic inquiry that places the researcher firmly within the
rooms: the reduction of room resonances; the removal of research process. This research project, similar to much
acoustic defects such as flutter echoes; the tuning of the architectural research, uses a hybrid methodology that is
room to an appropriate reverberation time; the increase in the synthesis of methods and techniques. As the modeling
speech intelligibility through positive early reflections and and simulation is done in a design process, the author is
the reduction in background noise; and the creation of an therefore a participant in the activities that are under
even or diffuse sound field. Most of these acoustic perfor- study.
mance considerations can be dealt with through the manip- It should be noted that this study of acoustic simulation
ulation of the overall geometry of the room, the detailed workflows were not carried out in a controlled manner, but
design of the room’s surface geometry, and the specifica- done within the design process. While performance objec-
tion of material. To simulate both the overall room perfor- tives were known before the project began, their method of
mance as well as the performance of the surface, both evaluation was not defined prior to the project commen-
geometric and numerical simulation methods must be used. cing. The workflows were developed within the design
The acoustic performance of the open-plan working space process and each software found its role naturally accord-
is largely driven by parameters that can be simulated using ing to how useful it was. It was not the intent to objectively
geometric acoustic methods. As this research project aimed compare software tools, but to develop a spectrum of simu-
to develop digital design workflows that incorporated lation approaches that could evaluate different acoustic
acoustic computer simulation in the design of a meeting parameters.
room incorporating hyperboloid geometry located in an
open-plan office environment, the digital design work-
flows needed to incorporate different types of simulation 3.2. Digital design and performance
to predict the acoustic performance of different aspects of All architecture is now digital architecture in the sense that
the design. As no architectural software exists that includes the vast majority of drawing, documentation, and commu-
this functionality, new software and new workflows using nication is done through the use of digital tools. Digital
multiple software packages needed to be developed. The design processes support the modeling, parametric trans-
inclusion of simulation into the architectural design work- formation, and computational generation of geometric
flow allowed for design, performance, and fabrication con- models. In addition, digital design processes support the
siderations to be explored simultaneously. The work done analytical evaluation of environmental performance based
in this project on the development of design tools for gen- on the computer simulation of physical reality.50 Departing
erating hyperboloid geometry and the description of that from previous methods of architectural design where
geometry for fabrication and assembly has been published sketching, drawing, and the construction of physical
Peters 793

models guided designers to final designs, computational models to predict the architectural qualities of a building
methods are now used to generate buildings where the final prior to construction, and in this way the use of simulation
design emerges as the design is evaluated and the computa- is a natural extension of the role of the architect. While a
tional model is modified.51 Aish52 uses the term ‘‘design model is an abstract description of a system, a simulation
computation’’ to define a design method where there is a is the description of the behavior of that system. Unlike a
distinction between the generative description of a building model, a simulation is not wholly descriptive; when a
and the resulting generated model. Performance-based simulation is begun, the outcome or result is not necessa-
design has been defined as an approach in which building rily known. There are two general classes of simulations:
performance becomes the guiding factor in design and this continuum methods and particle methods.56 Particle meth-
approach has long been recognized as an important issue ods make use of ‘‘Monte Carlo’’ algorithms. Such methods
in the use of computers in architecture.53 Using this design use random sampling algorithms where the randomness of
approach, there is an integration of evaluative simulation the algorithm need not correspond to an underlying inde-
processes with digital form generation processes. While it terminism in the system. The acoustic simulation algo-
has been recognized that all architecture results from a rithms of geometric acoustics, such as ray-tracing and its
complex combination of different influences, constraints, variants, can be considered to be particle methods.
and negotiated decisions during the design process, a Continuum methods begin by describing their object of
performance-based design paradigm is one of ‘‘form find- study as a medium described by fields and distributions.
ing’’ instead of ‘‘form making’’. The goal is then to give differential equations that relate to
The digital environment provides great potential for the rates of change of the values for these fields and distri-
new design tools and for the incorporation of acoustic per- butions, and to use ‘‘discretization’’ techniques to trans-
formance; however, working digitally also creates a poten- form these continuous differential equations into algebraic
tial for interfacing with digital fabrication machinery. This equations that can be calculated step-by-step by a com-
changes the relationship of the architect to the drawing, to puter. The acoustic simulation algorithms of numerical
construction, and to the building. The architect can now methods such as FDTD techniques can be considered to be
send fabrication data directly from the computer without continuum methods.
needing to go through the intermediary agent of the crafts-
man. With the wide availability of computer-controlled
fabrication machinery, such as 3D printers, laser cutters, 4. Implementing the algorithms:
milling machines, etc., the architect is now able to easily four digital design workflows
test and evaluate design propositions through prototyping. integrating acoustic simulation
Prototyping and the construction of full-scale demonstra-
tors was used to test acoustic design concepts in a ‘‘real- Four workflows were developed within the architectural
world’’ setting. The demonstrator in this experiment is design environment that integrate different simulation
used not only as way to validate the design objectives and methods and simulate different aspects of acoustic perfor-
test the acoustic performance of the designed object, but mance. These workflows are described in this section. The
also as a design tool itself. Beyond simulation and aurali- first two workflows use existing acoustic simulation soft-
zation, the demonstrator provides a second way in which ware to predict the acoustic performance of the room.
architects can experience and explore sound in space. These acoustic simulation software packages use geo-
metric acoustic algorithms. The last two workflows are
used to predict the acoustic performance of the surface.
3.3. Simulation The simulations in these workflows use the numerical
Simulation is a method related to the question of how modeling algorithms of BEM and FDTD.
‘‘reality’’ is actually constituted and how we can come to
know it. Simulation and modeling research has the theore-
tical presupposition that knowledge of ‘‘a reality’’ can be
4.1. Parallel modeling and simulation
obtained by reproducing that reality in some substitute Workflow 1, parallel modeling and simulation, is based on
medium. In architectural research, computer simulations the accepted workflow that integrates acoustic simulation
are generally used to find that certain aspects of design can into architectural and engineering practice today, see
be insignificant, or significant in terms of building perfor- Figure 1. When compared to architectural design software,
mance.54 Simulation is now a fundamental aspect of con- most acoustic simulation software packages have limited
temporary scientific knowledge and discovery.55 abilities to actually create geometry. In practice, the build-
Simulations allow for the imitation of a real-world pro- ing or room geometry is created in an architectural design
cesses over time, and can be used to show the eventual real software, either by the architect or acoustic engineer, then
effects of alternative conditions and courses of action. exported, and imported into the simulation software.
Architects are familiar with the use of drawings and While a digital model is often created by the architect, this
794 Simulation: Transactions of the Society for Modeling and Simulation International 91(9)

Figure 1. Workflow 1: parallel modeling and simulation.

digital model is often modified or re-made by the acoustic once the specific requirements of this representation are
engineer to reduce data, layer separate according to mate- understood, its production can be automated. Once this
rial, or fix meshing problems. This translation from one process is set up, the generation and analysis of design
representation – the architectural representation, to another options is simple and relatively quick. The time it takes to
representation – the acoustic representation, is necessary run a simulation is much longer than the time it takes to
because of the nature of the acoustic simulation itself. If export the data from the design software and import it into
architects are to include simulation as part of their work- the simulation software. The simulation and its visualiza-
flow, this will impact what architects draw. Odeon, the tion are handled by the simulation software. The novelty
simulation software used in this workflow, requires simpli- and customizable aspects of this workflow lie in the cre-
fied, watertight geometry, layer-separated by material.57 A ation and export of the correct data to the simulation soft-
simple custom parametric script was used to triangulate ware. The simulation algorithms used by the Odeon
geometric entities, including complex double-curved sur- software were used for the simulation of the room acoustic
faces. Vorlander1 recommends that the geometry not have performance.
extensive surface detail. As the materiality of the individ-
ual surfaces is critical, it is important to put the geometric
entities on different levels (or layers) according to their 4.2. Integrated modeling and simulation
material properties. This makes it easier to assign the cor- Workflow 2, integrated modeling and simulation, used an
rect material in acoustic simulation software. A surface’s acoustic simulation program that runs internally within the
material determines its absorption characteristics and a digital architectural design environment, see Figure 2. The
surface’s geometry determines is sound scattering proper- software used was the Pachyderm software, which runs as
ties. Both properties are important to determine room a plug-in to the Rhinoceros CAD software.58 In this work-
acoustic performance. While the determination of material flow, both the geometry and its analysis happen simultane-
type is a design decision, the determination of a surface’s ously within the same software without the need to
scattering coefficient is not so straightforward. There is translate the data between systems. This decreases the time
currently no way to determine the scattering coefficient in needed to make sure that data is compatible and reduces
an architectural software and export this data to an acous- errors introduced in the translation process, for example
tic simulation software. Scattering coefficients must be problems introduced when simplifying complex surfaces
assigned to individual surfaces in the simulation software. to meshes or the fact that some types of geometric entities
The Odeon manual offers a rough guide as to what these are not supported by analysis software.57 It was not the
scattering values should be.57 aim of this research project to validate or compare these
It should be made clear that the geometric representa- particular analysis software tools, but to develop new
tion required for simulation is a different representation workflows to integrate sound into the architectural design
than the one used for architectural exploration, but that process; however, a quick check indicated that the analysis
Peters 795

Figure 2. Workflow 2: integrated modeling and simulation.

done using Pachyderm was consistent with the results from hyperboloid structures.6 While these experiments demon-
Odeon. In this workflow, the simulation is handled by the strated that this workflow works, its disadvantages were
acoustic simulation plug-in and virtually no data prepara- that it takes a considerable time for each iteration (days),
tion is necessary, as the simulation software and the design and there were still outstanding issues with the process in
software run in the program. This workflow is conceptually terms of achieving accurate and reliable results.59,60
the most integrated of the four workflows in terms of mov- Therefore, for the FabPod, we looked to develop a purely
ing between design and acoustic performance evaluation. digital design workflow to obtain the scattering coefficient
The simulation algorithms used by the Pachyderm software from surface geometries. Unlike Workflows 1 and 2,
are based on geometric acoustics. which used geometric acoustics and abstracted sound as a
vector, Workflows 3 and 4 model sound waves themselves
and therefore use a numerical modeling algorithm.
4.3. Simulating the sound of the complex surface Workflow 3, simulating the sound of the complex surface,
While the previous two workflows simulate acoustic per- used the OpenBEM library,61 an open-source library of
formance of the room, the next two workflows simulate BEM algorithms. A diagram of Workflow 3 is shown in
the acoustic performance of the acoustic surface. While Figure 3. The workflow involved the creation of the com-
many acoustic simulation software packages include the plex geometry in the computer-aided design (CAD) soft-
effects of sound scattering, few actually enable the mea- ware, taking a section through this geometry, translating
surement of the sound scattering properties of complex the geometry into a text format using a custom computer
geometric surfaces. So, given that the FabPod has a com- script, and finally measuring the scattering coefficient
plex geometric surface, and the sound scattering perfor- using a custom script running in MATLAB.
mance of this surface is a crucial part of the design, the
sound scattering properties of this surface needed to be
simulated. Previous research has developed a digital/physi- 4.4. Visualizing sound
cal design workflow integrating architectural parametric Workflow 4, visualizing sound, used numerical methods to
modeling with rapid prototyping and laboratory measure- simulate sound waves (see Figure 4). This workflow was
ments to measure sound scattering coefficients.59 In a fur- not used to directly calculate the scattering coefficient, but
ther experiment, this workflow was then successfully used was used to create visualizations of sound waves and their
to measure the acoustic performance of different interaction with the complex geometry of the room and
796 Simulation: Transactions of the Society for Modeling and Simulation International 91(9)

Figure 3. Workflow 3: calculation of sound scattering coefficient using the boundary element method.

Figure 4. Workflow 4: visualization of sound scattering using the finite-difference time-domain method.

surface. This workflow used an implementation of the using a custom computer script. This produces a discre-
FDTD scheme. The algorithm was implemented in the tized grid of points that maps the space and geometry of
Processing computer programming language. The algo- the surfaces. It does not, however, yet include a mapping
rithm as implemented assumes that all surfaces are reflect- of material properties, but simply what is solid and what is
ing and so does not include absorption properties. The void. This data file is then used by the Processing software
inclusion of different absorption characteristics applied to to produce both a visualization of the sound waves as well
different surfaces in the model would alter the simulation as still images and data of the sound pressure level over
results, and indeed could increase sound scattering as alter- time measured at specific microphone locations. Although
nating surface absorption characteristics can cause sound the method has demonstrated that it can calculate the scat-
scattering.17 In this design workflow, the geometry is first tering coefficient,41 the current implementation does not
created in the CAD software, then exported as a data file yet do so.
Peters 797

of the FabPod. A visualization of the FabPod is shown in


Figure 5.

5.1. Digital design workflows


The digital design workflows of the FabPod project
required temporal flexibility so that acoustic design goals
and desired fabrication qualities could be considered early
in the design process. The project aimed to create digital
workflows that allowed for design decisions to be deferred
in one area of the project, without halting the work in
other areas. The workflows (see Figure 6) used both stan-
dard software and custom software. The research did not
seek to engage or propose a holistic software environment
in which all aspects of the design exercise could be inte-
grated; instead, the project aimed to create workflows that
connected separate software packages and specialized
tools. The output of each model could be used for acoustic
simulations. A set of parametric models were developed
Figure 5. The FabPod three-dimensional visualization. that allowed the designers to independently address the
overall form, and the surface patterning and material dis-
tribution, for a proposed design. Within these workflows,
5. The experiment: designing, it was hoped that the acoustic performance simulation
fabricating, constructing would inform the design of both the global form of the
meeting pod and the form and distribution of cells for opti-
The FabPod project builds on previous research by the mal scattering, reflection, absorption of sound; give feed-
authors in the creation of digital workflows that includes back on the sound propagation in and around the pod; and
performance analysis, design visualization, and the cre- provide data against which the acoustic testing of the com-
ation of fabrication data. While acoustic simulation is the pleted pod in use can be compared.
focus of this paper, the digital workflows that generated
the hyperboloid geometry were an important part of this
process, as these provided the architectural configurations 5.2. Performance goals
that were evaluated by the acoustic simulation routines. The performance goals for the FabPod can be described as
These workflows needed to consider the data needs of the to reduce sound transmission to and from the interior of
different acoustic simulations, and how the calculated per- the pod to a level that would cut disturbance and distrac-
formance could then re-inform the geometry and material tion to meeting participants and people working outside

Figure 6. The scope of design tools and their interactions with acoustic simulation software.
798 Simulation: Transactions of the Society for Modeling and Simulation International 91(9)

the pod; to accomplish the right level of sound scattering


within the pod for good speech intelligibility without hav-
ing to resort to a level of absorption that will deaden the
space, thereby creating a lively space for meeting and
debate; to provide a barrier to the propagation of sound in
the large open-plan space separating ‘‘loud’’ and ‘‘quiet’’
working areas; and to meet these performance goals with
the design of a meeting pod that is not entirely enclosed,
has an open doorway, is open to the ceiling, and has good
air circulation; in other words, a leaky acoustic space. The
FabPod meeting room needed to fit a large meeting table
and seating for eight people; the approximate interior
dimension of the square room is about 4–5 m by
about 3–3.5 m high, giving a volume of approximately
50–90 m3 (see Figure 7). The FabPod seeks to create an
acoustically better meeting room through three specific
features: a non-rectangular overall geometry and non-
parallel wall surfaces to reduce resonances; a highly
articulated interior surface to provide significant sound
scattering and reduce flutter echoes; and a balance of
materials adjusted to control acoustic absorption, therefore
producing a space without echoes or resonances but that
also does not sound ‘‘dead’’. This set of design criteria led
the team to a scheme for an aggregate structure composed
of different ‘‘cells’’. These cells could then be assigned
different types of material, they would have a complex
geometric profile to create sound scattering, and they Figure 7. The FabPod section and plan.
would be brought together in non-rectangular formation.
be varied, allowing the acoustic performance of the inte-
5.3. Geometry and fabrication constraints rior or exterior of the space to be tuned. The hyperboloid
faces are made from either Plexiglas, spun metal, or stiff,
Several key decisions drove the definition and constraint
acoustic-absorbing felt. The planar faces are made from
of the design space of the project. This investigation was
either Plexiglas or stiff, acoustic-absorbing felt. The acous-
inspired by anecdotal evidence of the sound-diffusing
tic absorption and scattering coefficients of the materials
properties of the surfaces at Antoni Gaudı́’s Sagrada
were necessary for the digital simulations.
Famı́lia.6 These geometries are based on the hyperboloid
of one-sheet, which became a fundamental research driver.
The circular hyperboloid has the known geometric prop-
5.4. The form of the meeting room
erty that the intersection of any two common hyperboloids
results in a planar curve when these hyperboloids are The overall form of the meeting room was determined
oriented equidistant from, and aligned to, a common point; using acoustic simulations primarily using Workflow 2,
in other words, when both hyperboloids are oriented nor- but also Workflow 1. As the form of the room was devel-
mal to and lie on a sphere (see Figure 8). Using these oped using a parametric modeling software, many different
spheres as set out geometry, different overall meeting geometric results could be developed very quickly. Rapid
room geometries could be explored (see Figure 9). This simulation and analysis was important for this part of the
creates overall room shapes that deviate from the regular, design process, and the integrated nature of Workflow 2
square format of meeting rooms, thus reducing the chance worked well. The results showed that, when considering
of resonances and flutter echoes. only room shape, an overall convex form had the lowest
The structure is composed of ‘‘cells’’ where each cell reverberation time (see Figure 11). It was discovered that
contains one hyperboloid that is trimmed by an irregular the calculation times for Workflow 2 took longer than
edge shape defined by a Voronoi pattern. The Voronoi Workflow 1; unfortunately, this meant that speed gained
cells were defined by a timber frame with a hyperboloid through integrating the analysis in the design software was
face on one side and a planar face on the opposite side lost due to an extended calculation time. It is speculated
(see Figure 10). The acoustic importance of this strategy is that one reason for this is that Odeon was developed in
that the materials of both the inside and outside face could such a way that all octaves are handled using a single
Peters 799

Figure 8. Hyperboloid intersections. The geometric rules ensure planar intersection curves between hyperboloids: (a) intersection
curves of spheres on a common sphere; (b) mirroring planes at the transition from one sphere to another.

Figure 9. Overall meeting room form: the two design families in which the geometry of alternate union and difference of sphere
are shown.

Figure 10. Individual hyperboloid cell: the form of the cell is defined by a Voronoi pattern, the overall build-up designed to reduce
sound transmission, and the surface materials defined by results from acoustic simulation.
800 Simulation: Transactions of the Society for Modeling and Simulation International 91(9)

Figure 11. Comparison of five different room shapes using Pachyderm and their calculated reverberation times.

Figure 12. Spatial distribution of Speech Transmission Index (STI) for the open-plan working environment with and without
meeting room pod.

round of calculation, whereas in Pachyderm, each octave space. While Workflow 2 was useful for quick iterative
band is handled by an independent round of ray-tracing studies, and therefore useful for the determination of a
calculation. While this did slow down the overall calcula- wide variety of room shapes, the detailed analysis of room
tion, this approach allows the user to specify frequency- acoustic performance was done using Workflow 1, as this
dependent scattering coefficients. A suggestion would be allowed for the calculation of additional acoustic para-
to build in an option where only one round of ray tracing meters such as the STI and the calculation and visualiza-
is performed to provide a way to calculate fast and effi- tion of spatial distributions of acoustic parameters. As an
cient sketch solutions. It is further speculated that, as all example, the STI, or degree of intelligibility, was calcu-
surfaces in the 3D CAD models in Rhinoceros are com- lated. This measure needed to be maximized inside of the
posed of non-uniform rational B-spline (NURBS) geome- meeting room, but minimized in the open-plan working
try, the necessary calculation of ray intersections with this space. While inside the meeting room speech should be as
type of geometry may take longer than the calculation of clear as possible, outside the meeting room, in the open-
intersections of rays with triangulated geometry. plan working space, speech can often disturb others work-
ing nearby. The distraction distance (rD) is the distance
where the STI falls below 0.50. As can be seen in
5.5. The performance of the room Figure 12, the addition of the pod structure creates a
Workflow 1 was used to do the detailed acoustic analyses ‘‘shadow’’ of low STI values, an area where noise from
of the meeting room and its surrounding open-plan office the other side of the pod will not be found to be disturbing.
Peters 801

Table 1. Study of critical parameters for acoustic performance of the open-plan working environment.

Parameter No pod With pod

rD – distraction distance (meters) 7.65 5.98


rP – privacy distance (meters) 28.62 24.62
D2,S – rate of spatial decay of SPL(A) 3.75 4.54
Lp,A,S,4m – nominal SPL(A) at 4 meters 52.24 52.06
Lp,A,B – background noise level 37.63 37.63
DL2 (A) – slope of spatial sound distribution (decibels) 5.56 6.85

SPL(A): A-weighted Sound Pressure Level.

Table 1 summarizes findings from the simulations. simplification of the complex geometry of the hyperboloid
These calculated values were averaged from several simu- cells to a triangulated geometry (see Figure 14). As the
lations. When compared to the previously discussed target simulation algorithms used do not calculate the scattering
values, the spatial decay of speech (D2,S) was too low with coefficient, yet require the scattering coefficient for accu-
or without the pod; the A-weighted sound pressure level racy,18 the acoustic performance of the complex surface
of speech at a distance of 4 m (Lp,A,S,4m) 4 was too high must be calculated using another means.
in both cases; and distraction distance (rD) was also too
high in both cases. However, it can be seen that the case
with the pod is better than without, and that the values are
5.6. The performance of the geometry
approaching what the ISO5 suggests for good performing Workflows 3 and 4 use numerical simulation algorithms to
spaces. simulate the wave nature of sound and calculate the per-
Workflow 1, to create geometry in one software and formance of the complex acoustic surface. As mentioned
analyze in another, allows for specialist software to be in the introduction, there are two main approaches to room
used and each simulation gives potentially a great deal of acoustic modeling: geometrical acoustics, in which sound
information about a particular design option. Because of is abstracted as rays; and, numerical methods, in which
the accuracy of the simulation process, and the variety of sound is modeled as waves.36 Workflows 1 and 2 use
acoustic parameters that the software is able to calculate, methods of geometrical acoustics. Siltanen et al.36 write
this was the primary workflow used. Figure 13 illustrates that the use of these methods ‘‘means that all the wave-
three different spatial distributions of acoustic performance based phenomena such as diffraction and interference are
parameters. The reverberation time was calculated both missing’’. However, the analysis software, Odeon, which
with sound sources inside the meeting pod and outside. is used in Workflow 1, has developed a mathematical
The calculated reverberation time when the sound source work-around that approximates the some of the effects of
was placed outside of the meeting room was consistent diffraction.18 Workflow 3 calculates the sound scattering
across the space; however, when the sound source was coefficient using the BEM. The hyperboloid geometry was
placed inside the meeting room the reverberation time generated using a computational design script and differ-
showed significant spatial variation due to volume and ent options of hyperboloid geometry were virtually con-
material differences between the inside and outside of the structed into several test pieces. The variables that were
pod. Sound propagation was approximated using the geo- tested were the size of the hyperboloids, the depth of the
metric algorithms of Workflow 1. This spatial distribution hyperboloids, and variation in depth of the hyperboloid
showed how the sound would ‘‘leak’’ out of the open door- wells. Multiple 2D sections were taken through each
way of the meeting room into the open-plan office. option and the scattering performance of these was simu-
Despite the advantages of Workflow 1 using an estab- lated using Workflow 3. It was found that the denser
lished simulation software, this workflow is limited in hyperboloid geometry produced more sound scattering
terms of feedback into the design process. The results over the applicable frequency range (see Figure 15). The
from the acoustic simulation software cannot be fed into a density and depth of the hyperboloids were also limited by
computational optimization loop (at least using Odeon). fabrication constraints. Although it was hypothesized that
Through the use of parametric or computational scripting, a variation in the depth would positively contribute to the
geometry modification can be linked to evaluation pro- sound scattering performance, it was found that the varia-
cesses to allow for design performance optimization.62,63 tion in depth of the hyperboloids did not appear to have an
The workflow, as it is not fully integrated, is also difficult effect on the amount of scattering. The measurements pro-
to use as a sketch tool; it does not allow for a quick analy- duced by Workflow 3 have only been validated against
sis of many different options. Workflow 1 required a previous results by the author and not against established
802 Simulation: Transactions of the Society for Modeling and Simulation International 91(9)

Figure 13. Three spatial distributions of acoustic parameters calculated using Workflow 1: (1) reverberation time (T30) with
sound source in the workspace; (2) reverberation time (T30) with sound source in the meeting pod; (3) sound propagation outside
of the meeting pod visualized using sound pressure level.
Peters 803

models. The results from Workflow 3 were useful in two


distinct ways in the design process: firstly, to compare
between options for surface geometries, and secondly, to
use the calculated scattering coefficient in acoustic simula-
tion software to get more accurate results.
Workflow 4 was used to visualize sound waves and
evaluate different surface geometries, and the relation of
surface geometry to overall form. Workflow 4, using the
FDTD, produces sound waves that emanate from a sound
source and reflect around a space affected by the geometry
of the surface (see Figure 16). These animations demon-
strate the effects of sound scattering and of sound diffrac-
tion. This workflow was useful in that it mapped the sound
pressure level over time and recorded levels at particular
microphone positions. The recorded levels were used to
see if flutter echoes were present. Figure 17 illustrates two
conditions, one where flutter echoes are present and the
corresponding sound pressure levels, and one where the
sound is diffused by a complex acoustic surface.

6. Conclusions
This paper defines four workflows that integrate sound in
an architectural design process: as separate design and
acoustic analysis software; as an integrated analysis plug-
in for design software; as a plug-in or separate software to
calculate the acoustic performance of materials; and as a
tool for the creation of visualizations of acoustic perfor-
mance. A broad conclusion of the paper is that there is a
Figure 14. Simplification of hyperboloid surface geometry into need for more software that allows architects to simulate
triangular facets. acoustic performance and design for sound. There is an

Figure 15. Boundary element method calculation of scattering coefficients for sparse and dense configurations of hyperboloids,
where the single scattering value (measured at 707 Hz) required by Odeon is noted.
804 Simulation: Transactions of the Society for Modeling and Simulation International 91(9)

between geometry creation and geometry analysis, is fluid.


It has been observed in practice that computation and its
relation to digital manufacturing are redefining roles in the
building industry.64 The findings of this project suggest
also that simulation allows for the redefinition of roles. As
Yanni Loukissas65 states,

From project to project, architects and engineers are looking


for ways in which they can represent architecture in order to
gain a professional upper hand . it is not enough to simply
talk about these technologies in terms of one person at a
machine, . but understanding the implications of computer
simulations requires a broad view of how they affect systems
of relationships.

The author recognizes that these particular workflows are


still under development and need to be improved; how-
ever, as discussed, in general design workflows are devel-
oped during the design process and must remain flexible.
As typical with architecture projects, the design tools that
were developed had to constantly be updated as new con-
straints were discovered and new design opportunities
were investigated. A classic ‘‘wicked problem’’, these
simulation tools needed to be adapted to new input geome-
tries, and output data. The applicability of the findings
benefitted from being carried out using a research-by-
design methodology. In regards to this point, a conclusion
is that not only do the workflows need to be flexible
between projects, but also within a project, as the design
goals shift and new potentials are found as the design
develops. Specific tools also need to be flexible in terms
of having a variety of inputs and outputs.
In the initial design stages of an architecture project,
sketching rough concepts of form and material are critical,
and it is important that designers are able to gain acoustic
performance feedback on developed options. Workflow 1
has the potential to allow this. In this project, Workflow 1,
the use of separate software for acoustic simulation, pro-
vided the most clear, detailed, and useful results, although
these did not include information on the sound scattering
Figure 16. Visualization of sound waves in the FabPod meeting performance. This was because of the efficient workflow
room. setup to move geometry from the design software to the
analysis software, coupled with the speed of the analysis
software and its multiple ways of analyzing and outputting
increasing trend towards performance-driven architecture the data.
and these workflows contribute to this discussion. As As computational techniques now allow the exploration
simulation and analysis tools are integrated into design complex geometries and include performance evaluation,
software, architects are better able to predict their designs architects are beginning to see the potential for including
before they are built – this is a fundamental definition of acoustic performance as part of the design process. The
what it means to be an architect. However, even though FabPod experiment demonstrated that there is an interest
these tools are becoming more available, and more linked by architects to create spaces and surfaces that can define
to the design environment, the loop of how the analysis sound in a more specific way. However, while there are
results inform the design is yet to be worked out. This now established acoustic simulation software packages for
project suggests that this distinction is hard to define. The acoustic engineering, there is no fully featured acoustic
boundary between architecture and engineering, and design software for architects. Workflow 2, the integration
Peters 805

Figure 17. Using finite-difference time-domain method acoustic simulation to test for flutter echoes.

Figure 18. Exterior view of the constructed FabPod (photo Figure 19. Interior view of the constructed FabPod (photo
credit: John Gollings). credit: John Gollings).

of acoustic performance analysis in architectural software, performance, but there needs to be a way to evaluate
demonstrates a benefit as it allows for the testing of ideas design aspirations, and Workflows 3 and 4 were developed
quickly. However, and perhaps more critically, as this to evaluate the acoustic performance of specific complex
workflow integrates analysis in the CAD design environ- surface geometries. The findings were that current tools
ment, this gives the potential to set up iterative optimiza- and workflows available to architects for the evaluation of
tion processes that could be linked to generative processes the acoustic performance of complex surfaces and differ-
and this has been demonstrated in other research.62,63 As ent material build-ups are lacking. Current approaches do
more designers begin using computational methods and not include the ability to analyze architectural designs at a
realizing their potentials, this type of workflow will scale relating to surface geometry or material. These eva-
become more prevalent. luation routines needed to be constructed almost from first
Another observation from this experiment, and of pre- principles, as there was very little simulation software that
vious research,59 is that architects are interested in the measures the acoustic performance of complex surfaces.
design of architectural surfaces that have a complex geo- The method of calculating sound scattering by using the
metric and material configuration (see Figures 18 and 19). BEM digitally appears to be faster and more inexpensive
These surface configurations can be driven by acoustic than the previously examined process of using rapid
806 Simulation: Transactions of the Society for Modeling and Simulation International 91(9)

prototype models. The FDTD algorithm provided a fast 2. Rindel JH. The use of computer modeling in room acoustics.
and effective way to visualize sound. These visualizations J Vibroeng 2000; 4: 41–72.
provided a deeper understanding on how the complex geo- 3. DeBodt K. Digital resonance: sound parameters in architec-
metry affected the sound waves and complemented the tural design. In: Oosterhuis K and Feireiss L (eds) Game, set,
measurement of the sound scattering coefficient. It would and match: on computer games, advanced geometries, and
digital technologies. Rotterdam: Episode, 2006.
be of great benefit if the FDTD tool was developed to also
4. Peters B. Acoustic performance as a design driver: sound
provide some measure of scattering performance.
simulation and parametric modeling using Smartgeometry.
Architects are used to using abstractions. The discipline Int J Architect Comput 2010; 8: 337–358.
is based on its ability to draw and model and use this to 5. ISO. Building acoustics. In: Acoustics—measurement of room
predict the performance of buildings before they are con- acoustic parameters—part 3: open plan offices. Geneva:
structed. What was found in this experiment was a curious International Organization for Standardization, 2011.
tension between rays and waves/vectors and raster that 6. Burry J, Davis D, Peters B, et al. Modelling hyperboloid
emerged with the investigation of sound and its relation to sound scattering: the challenge of simulating, fabricating and
representation and performance. The two approaches (rays measuring. In: design modelling symposium, Berlin,
and waves) require different analysis techniques, and dif- Germany, 10-12 October 2011.
ferent workflows, but also most critically, different ways 7. Bradley J. Acoustics for speech in classrooms and meeting
of representing and therefore conceptualizing and design- rooms. Ottawa: National Research Council of Canada (NRC-
CNRC), 2003.
ing a project. There is a need for the implementation of
8. Doelle L. Environmental acoustics. New York: McGraw
numerical modeling techniques to understand wave phe- Hill, 1972.
nomena of sound for architectural design. 9. Long M. Architectural acoustics. Burlington, MA: Elsevier,
An aspect of the simulated sound field that was found 2006.
to be of importance to architectural designers was the spa- 10. Fuchs HV, Zha X and Drotleff H. Relevance and treatment
tial distribution of acoustic properties. This was important of the low frequency domain for noise control and acoustic
in some cases, for example reverberation time of the open- comfort in rooms. Acta Acustica Acustica 2005; 91: 920–927.
plan office, for a more uniform value, and in other cases, 11. Junqua J-C. The influence of acoustics on speech production:
for example the STI of the open-plan office, for a field of a noise induced stress phenomenon known as the Lombard
different values. While architectural designers were inter- reflex. Speech Comm 1996; 20: 13–22.
ested in meeting the target values for acoustic criteria, they 12. Bradley J. Speech intellibility studies in classrooms. J Acoust
Soc Am 1986; 80: 846–854.
were more interested in new phenomena that could be
13. Bradley J and Yang W. Effects of room acoustics on the
investigated and new experiences that could be introduced.
intellibility of speech in classrooms. J Acoust Soc Am 2009;
This desire to explore and define new types of acoustic 125: 1–12.
performance was an important factor in developing flex- 14. Burgtorf W and Oehlschlägel HK. Untersuchungen über die
ible workflows and tools. richtungsabhängige wahrnehmbarkeit verzögerter schall-
The FabPod project developed and tested four specific signale. Acta Acustica Acustica 1964; 14: 254–266.
workflows for integrating acoustic performance into archi- 15. Kuttruff H. Room acoustics. 4th ed. London: Spon Press,
tectural design. While these are not proposed to be answers 2000.
to all design situations, they do point to several general 16. Nagata M. The effects of diffusion on room acoustics. In:
approaches that could be developed further to provide Nagata acoustics newsletter. Tokyo: Nagata Acoustics, 2009.
sound design tools for architects. 17. Cox TJ and D’Antonio P. Acoustic absorbers and diffusers: the-
ory, design and application. London: Taylor & Francis, 2009.
18. Christensen CL and Rindel JH. A new scattering method that
Funding combines roughness and diffraction effects. In: forum acusti-
The author’s research was funded by the Danish Ministry of cum 2005, Budapest, 29 August–2 September 2005.
Higher Education and Science’s Industrial PhD program. The 19. Cox TJ and D’Antonio P. Engineering art: the science of
FabPod project received funding from the Property Service concert hall acoustics. Interdiscipl Sci Rev 2003; 28:
Group and The Design Research Institute at RMIT University as 119–129.
well as The Australian Research Council through its support of 20. Vorlander M and Mommertz E. Definition and measurement
the ARC Discovery Grant Challenging the Inflexibility of the of random-incidence scattering coefficients. Appl Acoust
Flexible Digital Model. 2000; 60: 187–199.
21. Mommertz E. Acoustics and sound insulation. Detail prac-
tice. Basel: Birkhauser, 2009.
References 22. Kleibrink M. The world of office work in flux—usage strate-
1. Vorlander M. Auralization: fundamentals of acoustics, model- gies and freedom of choice. In: Schittich C (ed.) Work
ling, simulation, algorithms and acoustic virtual reality. environments: spatial concepts, usage strategies, communi-
Berlin: Springer, 2008. cations. In detail. Basel: Birkhauser, 2011.
Peters 807

23. Vischer J. Towards an environmental psychology of work- 42. Redondo J, Picó R, Avis MR, et al. Prediction of the
place: how people are affected by environments for work. random-incidence scattering coefficient using a FDTD
Architect Sci Rev 2008; 51: 97–108. scheme. Acta Acustica Acustica 2009; 95: 1040–1047.
24. Maher A and von Hippel C. Individual difference in 43. Botteldooren D. Finite-difference time-domain simulation of
employee reactions to open-plan offices. J Environ Psychol low-frequency room acoustic problems. J Acoust Soc Am
2005; 25: 219–229. 1995; 98: 3302–3308.
25. Petersen CM. Limiting annoying noise in open-plan offices. 44. Yokota T, Sakamoto S and Tachibana H. Visualization of
In: joint Baltic-Nordic acoustics meeting 2008, Reykjavik, sound propagation and scattering in rooms. Acoust Sci
2008. Technol 2002; 23: 7.
26. Nilsson E and Hellstrom B. Acoustic design of open-plan 45. Burry J, Williams N, Cherrey J, et al. Fabpod: universal digi-
offices. NT technical report. Taastrup: Nordtest, 2010. tal workflow, local prototype materialization. In: CAAD
27. Addis B. A brief history of design methods for building futures, Shanghai, China, 3–5 July 2013.
acoustics. In: third international congress on construction 46. Williams N, Cherrey J, Peters B, et al. Hubpod: a prototypi-
history, University of Technology Cottbus, Germany, 20–24 cal design system for acoustically diffused enclosures. In:
May 2009. prototyping architecture, London, 2013.
28. Sabine WC. Collected papers on acoustics. Cambridge, MA: 47. Archer B. The nature of research. Codesign 1995;. January:
Harvard University Press, 1922. 6-13.
29. Jones PB and Kang J. Acoustic form in the modern move- 48. Roth S. The state of design research. Design Issues 1999; 15:
ment. Architect Res Q 2003; 7: 75–85. 18–26.
30. Barron M. Auditorium acoustic modelling now. Appl Acoust 49. Gray C and Malins J. Visualizing research. Aldershot:
1982; 16: 279–290. Ashgate, 2004.
31. Rindel JH.Modelling in auditorium acoustics—from ripple 50. Malkawi AM. Performance simulation: research and tools.
tank and scale model to computer simulations. In: forum In: Kolarevic B and Malkawi AM (eds) Performative archi-
acusticum, Seville, Spain, 2002. tecture: beyond instrumentality. London: Spon Press, 2005.
32. Gade AC. Recent experiences using room acoustic scale 51. Kolarevic B. Back to the future: performative architecture.
models. In: proceedings of the Institute of Acoustics, 2006. Int J Architect Comput 2004; 2: 43–50.
33. Krokstad A. The hundred years cycle in room acoustic 52. Aish R. First build your tools. In: Peters B and Peters T (eds)
research and design. In: Svensson P (ed.) Reflections on Inside smartgeometry: expanding the archtiectural possibli-
sound. Trondheim: Norwegian University of Science and ties of computational design. Chichester: Wiley, 2013.
Technology, 2008. 53. Oxman R. Performance-based design: current practices and
34. Rindel JH. Room acoustic prediction modelling. In: XXIII research issues. Int J Architect Comput 2008; 6: 1–17.
Encontro Da Sociedade Brasileira De Acustica, Salvador, 54. Grout L and Wang D. Architectural research methods. New
Brazil, 18–21 May, 2010. York: John Wiley & Sons, 2002.
35. Vorlander M. International round robin on room acoustical 55. Orrell D. The future of everything: the science of prediction
simulations. In: 15th annual congress on acoustics, from weather to chaos and complexity. New York: Thunder
Trondheim, Norway, 1995. Mouth Press, 2007.
36. Siltanen S, Lokki T and Savioja L. Rays or waves? 56. Winsberg E. Science in the age of computer simulation.
Understanding the strengths and weaknesses of computa- Chicago, IL: University of Chicago Press, 2010.
tional room acoustics modeling techniques. In: proceedings 57. Christensen CL. Odeon room acoustic program: version 10.0
of the international symposium on room acoustics (ISRA users manual. Lyngby: Odeon A/S, 2009.
2010), Melbourne, Australia, 2010. 58. van der Harten A. Pachyderm acoustical simulation: towards
37. Bork I. Report on the 3rd round robin on room acoustical open-source sound analysis. Architect Des 2013; 222: 138–
computer simulation—part II: calculations. Acta Acustica 139.
Acustica 2005; 91: 753–763. 59. Peters B and Olesen T. Integrating sound scattering mea-
38. Peters B. Integrating acoustic analysis in the architectural surements in the design of complex architectural surfaces:
design process using parametric modeling. In: forum acusti- informing a parametric design strategy with acoustic mea-
cum, Aalborg, Denmark, 2011. surements from rapid prototype scale models. In: eCAADe
39. Krokstad A, Strom S and Sorsdal S. Calculating the room 2010, Zurich, 2010.
acoustical response by the use of a ray tracing technique. J 60. Choi Y-J and Jeong D-U. Some issues in measurement of the
Sound Vib 1968; 8: 118–125. random-incidence scattering coefficients in a reverberation
40. Allen J and Berkley D. Image method for efficiently simulat- room. Acta Acustica Acustica 2008; 94: 769–773.
ing small room acoustics. J Acoust Soc Am 1979; 65: 943– 61. Juhl P and Henriquez VC. Openbem: open source matlab
950. codes for the boundary element method, http://www.open-
41. Redondo J, Picó R and Roig B. Time domain simulation of bem.dk/ (accessed 15 December 2013).
sound diffusers using finite-difference schemes. Acta 62. Echenagucia TM, Astolfi A and van der Harten A.
Acustica Acustica 2007; 93: 611–622. Interactive design methods for complex curved reflectors in
808 Simulation: Transactions of the Society for Modeling and Simulation International 91(9)

concert halls. In: international symposium on room acous- Author biography


tics (ISRA), Toronto, 9–11 June 2013.
Brady Peters is an assistant professor at the John H.
63. Palma M, Sarotto M, Echenagucia TM, et al. Sound strength
Daniels Faculty of Architecture, Landscape, and Design
driven parametric design of an acoustic shell in a free field
environment. In: international symposium on room acoustics at the University of Toronto. He is a director of the
(ISRA), Toronto, 2013. Smartgeometry organization, and he was formerly an
64. Knippers J. From model thinking to process design. Architect associate partner at the award-winning architecture firm
Des 2013; 222: 74–81. Foster + Partners.
65. Loukissas Y. Co-designers: cultures of computer simulation
in architecture. New York: Routledge, 2012.

You might also like