Proc. of the EAA Joint Symposium on Auralization and Ambisonics, Berlin, Germany, 3-5 April 2014

ABSTRACT

Auralization provides a valuable tool that allows architects,

building owners, and other decision-makers to directly experi-

ence the aural implications of design decisions and allows them

to make more informed choices. Standard numerical metrics are

difficult to relate to aural phenomena without significant practice

and frequently fail to capture acoustical issues that are essential

actions between PA systems and room acoustics, HVAC noise,

wall and window transmission, and the subjective effects of

sound masking. In general, clients find the experience of listen-

ing to their as-yet unbuilt spaces to be exciting and useful.

Though most are not trained listeners, they typically move quick-

discussion of the different acoustical treatments presented and of

the overall sound of the space. This helps architects and project

owners to feel connected to the acoustical aspect of the design,

and it helps the team to agree on design decisions that may have

platform.

Over the last ten years, consultants at Acentech have worked to

develop and refine techniques for using auralization as part of the

design process for a variety of buildings and spaces. Much of this

work has focused on the auralization of activity sounds in large

public spaces such as atrium-type lobby spaces in universities

and cultural institutions. The relevant features for auralization of

this type of space are:

in an acoustically appropriate meeting room, it is possible for a

group of listeners, such as a project design team and owner, to

experience the auralization together and to accurately judge

as loudness, reverberance, and speech intelligibility.

To create a perceptually realistic soundfield, it is necessary to

include multiple source locations and to convolve each source

with time-incoherent anechoic sound material. Our early tech-

niques required completing separate convolutions for each source

time-consuming revision and calibration process and leaving no

possibility of level adjustment among sources during playback.

The most recent innovation to this process uses software created

with the MaxMSP development environment to perform all of

the required convolution in real time during the presentation.

This allows for easy toggling on and off of individual sources,

level and timbre manipulation of individual sources, and switch-

ing among various architectural design conditions, all during the

presentation.

AURALIZATION DEVELOPMENT

We conceptualize auralization development as four main pro-

cesses:

 modeling,

 impulse generation,

 source material selection and convolution, and

 playback calibration.

Typically, our modeling begins in SketchUp and is then exported

to CATT Acoustic where material properties are added. The

Proc. of the EAA Joint Symposium on Auralization and Ambisonics, Berlin, Germany, 3-5 April 2014

material with first-order ambisonic (B-Format) impulses re-

sponses for each sound source in the model, and then summed

and rendered the resulting audio as either a four channel (quadra-

phonic) or five channel (5.1 surround) WAV file using simple

decoding filters generated by CATT Multivolver. This process

was effective, but revisions and calibration were time-consuming

since any changes involving source material required repeating

major new wing at the Museum of Fine Arts Boston, which

opened to the public in 2010. This work was presented previous-

ly at the EAA Symposium on Auralization in Espoo, Finland in

2009 [1] and at Internoise 2009 in Ottawa, Canada in 2009 [2].

The auralization presented a familiar architectural acoustics

problem and solution: excessive loudness in a large, reverberant

public space, addressed with the inclusion of large areas of

acoustically absorptive material. The design team and museum

trustees heard a predictive auralization of their future space

during a fundraising banquet, complete with 500 talkative diners,

director. Participants experienced the difficulty of speaking with

design conditions were presented: no acoustical treatment, a

small amount of acoustical treatment, and the recommended

amount of acoustical treatment. All audio was pre-rendered and

played back as four-channel wav files. Anechoic sounds were

gathered from various sources, including original recordings of

conversation and clearing of dishes made in our nearly-anechoic

presentation room, speech from museum audio tour recordings,

and studio recordings of a swing band.

processing significantly reduces total processing time, processing

Proc. of the EAA Joint Symposium on Auralization and Ambisonics, Berlin, Germany, 3-5 April 2014

platform that allows independent development of additional

function objects (called “externals”); this extensibility is critical

when using Max as an auralization tool since many of required

functions, such as convolution and ambisonic encoding and

decoding, are not part of the base MaxMSP object library.

The most important tools for auralization in Max are the convo-

vide some left-right spatial spread. Stereo anechoic material was

used where available, and in other cases monaural material was

disc “ Anechoic Orchestral Music Recording” (1995) and from

Wenger Corporation’s “Anechoic Ch oral Record ings” (2004).

The auralization helped the university music faculty to gain

confidence in the design, particularly its acoustical variability.

Figure 2: CATT Acoustic model of the University of

Massachusetts recital hall, Boston, MA, USA

Margery Milne Battin Hall at the Cary Memorial Building in

Lexington, Mass. is a multifunction auditorium built in 1928.

The hall’s users were unsatisfied with its sound reinforcement

system, and the building’s HVAC system was loud and ineffi-

cient. Acentech was asked to design an upgraded PA system and

to assist the mechanical engineer in quieting the ventilation

system.

To demonstrate the predicted effects of our various design rec-

ommendations, and to help the client prioritize their use of lim-

presented based on the system noise level and spectrum meas-

ured in the auditorium, and two proposed upgrades to the HVAC

system were presented based on levels and spectra calculated

according to ASHRAE guidelines. Since the auralization is being

processed in real time, the background noise, PA system, and

curtains can be toggled during playback with a continuously

running source. This allowed for seamless comparisons between

the different options and made the differences much easier for

the clients to understand.

Proc. of the EAA Joint Symposium on Auralization and Ambisonics, Berlin, Germany, 3-5 April 2014

Because the convolution takes place with first order ambisonic

impulse responses, each source requires four channels of real-

time convolution. While we have successfully had over 50 chan-

nels of convolution running concurrently on a quad-core CPU,

the 140 channels of convolution that would be required to gener-

ate the responses of all 35 sources in real time is not practical. To

reduce the number of concurrent convolution channels, we

Shark and Ray Touch Tank, where guests can pet small sharks

and stingrays. The space has a naturally high background noise

use after washing their hands. When combined with the added

noise of as many as 100 excited elementary school students, the

hall containing the Touch Tank can be extremely loud. The New

England Aquarium was interested in ascertaining the noise im-

pact on the animals in the tank, as well as exploring potential

room treatments to reduce the overall noise level for the comfort

of their staff and guests.

that appropriate audio source material was used. Since the audi-

ence for this auralization was intimately familiar with the mod-

elled environment the sounds presented had to match their expec-

many sounds within the space, including the background sound

of the unoccupied space, the hand dryers, and the sound of ex-

cited children. The speech of one of the educators who directs

patrons in safely touching the animals was transcribed and re-

recorded anechoically. Since we were also concerned about the

noise impact on the animals in the water, we also recorded the

background sound underwater using a hydrophone, and measured

the transfer function impulse response between a microphone just

above the water and a hydrophone in the tank.

The auralization was very successful, but in an unusual way: in

this case we demonstrated that additional acoustical treatment

would not make a substantial difference in the loudness of the

space, and that the most effective option for quieting the exhibit

a five storey glass, wood, and stone atrium featuring a café with

seating all around the opening. Despite our presentation of calcu-

lated reverberation times and expected background sound and

speech intelligibility levels, the client remained unsure that the

extensive acoustical treatments we recommended were truly

needed. After further discussion, it was decided that an auraliza-

tion would be the best way for the client to make an informed

decision regarding room treatments, and to develop appropriate

space. Of particular interest to the client was the level of café

with light footfall noise to boisterous conversation and the

sounds of eating and tables being cleared. The level of ambient

activity can be varied in real time, and for each activity level

chosen by the operator appropriate activity samples are dynami-

cally selected and fed into the convolution engines for the vari-

ous modelled activity sources. A close-mic’d recording of a

speech by an Olin School professor was used as the sound source

in the Forum, which was both context-appropriate and helped to

anchor the audience’s sense of place in the modelled environ-

ment. This recording was made by the school for public relations

Proc. of the EAA Joint Symposium on Auralization and Ambisonics, Berlin, Germany, 3-5 April 2014

Five varying levels of acoustical treatments were modelled,

ranging from the initial design to fully treating every available

surface. Five listener locations are available, including one inside

a faculty office overlooking the atrium which incorporates the

transmission loss of the window glazing and allows for three

different glazing types to be auditioned. Of course, all of these

parameters are freely variable at run time without a need to stop

and restart the source recording. See Figure 5 for an illustration

of the graphical user interface (GUI) used to select the various

options available in this auralization.

in a way that they had not been able to from a written report.

final decision was that fully treating the Forum and atrium was a

requirement for the success of the project. They were also alerted

to practical considerations regarding scheduling of Forum pres-

entations and the need to control Café activity during these pres-

entations.

While not an architectural design problem per se, The Virtual

Paul’s Cross Project [7] represents the current cutting-edge of

Acentech’s auralization program. This project was a joint inves-

London in 1666.

The Paul’s Cross auralization includes the modelled acoustics of

the open churchyard and its surrounding buildings, a 2 hour and

15 minute anechoically-recorded sermon, and an “artificially

intelligent” crowd which is variable in real time from between

selects appropriate reactions from a custom-produced sample

library. Birds fly overhead, horses trot past at the edges of the

Proc. of the EAA Joint Symposium on Auralization and Ambisonics, Berlin, Germany, 3-5 April 2014

were able to specify model parameters and then automatically

generate all of the ambient sounds, with that added benefit that

no two listens through the sermon are exactly the same. Also, not

needing to pre-render the audio substantially reduced the size of

the data assets. As an example, the audio assets for the auraliza-

tion total 2.6 GB of data. If all configurations of the auralization

were to be pre-rendered, it would result in 240 GB of data, al-

Music and Sound Technology, Zurich University of the

Proc. of the EAA Joint Symposium on Auralization and Ambisonics, Berlin, Germany, 3-5 April 2014