MODELLING SOUNDSCAPES
Presentation by
Rob Bullen
Architectural Acoustics
Is it a
concert
hall?
Y
Detailed modelling of
reverberation characteristics,
clarity, spatial separation, etc.
Audible simulation (?)
N
Is it a lecture
theatre, place
of worship or
similar?
Y
Calculate STI, check
reverberation, maybe
some other features
depending on $
N
Is it an
apartment?
N
Y
Discuss AAAC star rating,
select partitions
accordingly (& meet BCA)
Meet the BCA &
AS 2107
cafe, hospital ward,
call centre, dance
studio, library, gym,
foyer, courtyard, …
Can we do more in the “also-ran” spaces?
Rather than just meet criteria, can we think about and evaluate the
soundscape in the space?
The level and quality of sound from all sources – including
differentiated & undifferentiated, conscious & unconscious
To do this we need:
•
indicators of goodness – preferably numerical
•
ways to evaluate a soundscape that doesn’t exist yet
•
design tools to achieve the desired soundscape – these are just
normal acoustic tools – partitions, finishes, sound masking, etc.
The Physical Soundscape
Roof/Ceil
paircraft
Hroof
hroom
hroom
hroom
Window
ppeople
hroom
hroom
Hwind
hroom
ptraffic
The Mental Soundscape
Conscious
(Conscious
Awareness)
Information
Aircraft
Aesthetics
Baby
Annoyance
Speech
Mood
Unconscious
Automatic behaviour
Performance effects
Stress?
The Task of Acoustic Design:
Engineer the physical soundscape in such a way that it will
produce a desired mental soundscape.
Sound
pressure at
ears
Physical
Mental
The Interface Between Physical and
Mental Soundscapes
Formation of a mental soundscape seems to involve at least 3
separate processes:
•
detection – frequency analysis of the incoming signals in
slices of about 30 ms to produce a form of spectrogram
•
“scene analysis” – breaking the spectrogram into streams
in such a way that they probably represent objects in the
real world
•
informational and attentional masking – picking out the few
streams that will be reach conscious awareness. The rest
will form part of the unconscious.
Some Preliminary Assumptions and Rules
of Thumb for Designing a Soundscape
Unconscious Sound
1. In any soundscape the streams with low SPL (as well as
unstreamed noise) will generally be relatively stable with time.
Call this the “background” and identify its level with either the total
LAeq of these low-SPL streams or the LA90 of the total noise. This
sound is always unconscious (after some time in the space).
2. For most spaces, the only control you need on background sound
is on the level. The range of levels in AS 20107 is a reasonable
guide to acceptable levels of background (unconscious) sound.
Some Preliminary Assumptions and Rules
of Thumb for Designing a Soundscape
Conscious Sound
3. For a time-varying sound source, if LAmax is at least 10dB below
the background, it will be energetically masked (i.e. inaudible).
[For unusual spectra, e.g. very low-frequency sound, you may
require the SPL to be 10 dB below background in each octave
band.] Otherwise the source will form a stream and has the
potential to become conscious.
4. If LAmax of a time-varying source is at least 5dB above the
background, assume that the stream will become conscious.
[Again, to be more conservative you can apply this to each octaveband separately.]
5. A source that is both conscious and unwanted will cause
annoyance.
Some Preliminary Assumptions and Rules
of Thumb for Designing a Soundscape
Sound Quality
6. For non-speech sound, to a first approximation sound quality is
described by reverberation time. For example:
< 0.5 secs = “soft”, “sophisticated”, “dead”. People will
tend to speak more quietly. Required for amplified music.
> 1 sec = “buzzy”, “live”. People will tend to speak more
loudly. Required for classical music.
7. Similarly, to a first approximation speech quality is described by
STI.
In short, I propose the following
general specifications for a soundscape:
•
Background SPL
•
What should be inaudible and what should be not intrusive?
•
RT
•
Where do you want speech to be intelligible / not intelligible?
Evaluating a Soundscape
•
You can calculate the total background level, the LAmax from timevarying sounds, the RT and the STI, then see whether the
soundscape will fit your design intent
Or…
•
You can simulate the soundscape.
[Demo of SoundScape program here]
SoundScape: A few technical details
SOUND
SOURCE
Sound
file
reader
SOUND
PATH
From sound
sources
Amp:
source
level
Splitter
TL
Filter
Mixer
Amp:
distance
To reverberation
Splitter
ROOM
From sound
paths
Mixer
Mixer
Reverberation
Direct
To sound paths
Mixer
To direct
SoundScape: A few technical details
TL Filter: Fits a digital filter to an arbitrary specified set of TL
values, approximately. (Filter is 2 conjugate poles + 2 conjugate
zeros.)
Reverberator: Feedback delay network
In
delay 1
filter 1
delay 2
filter 2
delay 3
filter 3
…
…
Average delay
= mean free path / c
=4V/Sc
Out
Filter reduction, dB
= delay * 60/RT
In fact, the ability to add sources at
different levels is useful for other things …
•
Simulations of environmental noise for potential residents?
Experience indicates this is dangerous and difficult to get
right, but in some limited scenarios could be useful (e.g.
difference between single and double glazing).
•
When does a noise sound tonal?
•
How well can you REALLY measure a sound in the presence
of background?
Conclusions …
•
For acoustic design of “also-ran” spaces, it’s possible to do better
than just meet AS 2107 and BCA. The acoustic properties of these
spaces are important to most people in their normal lives.
•
Because these spaces are so diverse, we need a general-purpose
methodology to set acoustic goals and evaluate a design. The
methodology should allow non-experts (e.g. architects) to
communicate with experts in a common language.
•
The concept of a soundscape can do this, with a few descriptive
parameters:
• background SPL
• audibility and intrusiveness of foreground sources
• RT and STI
Conclusions …
•
From this, an expert can do calculations and produce a design
that is likely to satisfy the requirements.
•
Alternatively or additionally, simulation of the soundscape can
help greatly both in defining the requirements and in achieving
them. Where necessary, in simple cases a non-expert can
produce a design that is, at least, a lot more satisfactory than if
they had ignored acoustics altogether.
•
What’s more simulations are fun. Which makes them good for
students.