This is a repository copy of A concept-based system for the live diffusion of sound via

multiple loudspeakers.

White Rose Research Online URL for this paper:

http://eprints.whiterose.ac.uk/80585/

Proceedings Paper:
Mooney, JR and Moore, D A concept-based system for the live diffusion of sound via
multiple loudspeakers. In: Proceedings of the Digital Music Research Network. Digital
Music Research Network Conference 2008, 07 - 08 Jul 2008, Leeds, UK. . (Unpublished)

Reuse
Unless indicated otherwise, fulltext items are protected by copyright with all rights reserved. The copyright
exception in section 29 of the Copyright, Designs and Patents Act 1988 allows the making of a single copy
solely for the purpose of non-commercial research or private study within the limits of fair dealing. The
publisher or other rights-holder may allow further reproduction and re-use of this version - refer to the White
Rose Research Online record for this item. Where records identify the publisher as the copyright holder,
users can verify any specific terms of use on the publisher’s website.

Takedown
If you consider content in White Rose Research Online to be in breach of UK law, please notify us by
emailing [email protected] including the URL of the record and the reason for the withdrawal request.

[email protected]
https://eprints.whiterose.ac.uk/
 A CONCEPT-BASED MODEL FOR THE LIVE DIFFUSION OF SOUND
VIA MULTIPLE LOUDSPEAKERS

Dr James Mooney Dr David Moore

Newcastle University University of Sheffield
Culture Lab Department of Music

ABSTRACT cert programme, a dynamic and configurable audio routing ar-

chitecture is required.
This paper presents a conceptual framework for sound diffusion: The authors have previously developed the M2 Sound Diffu-
the process of presenting multiple channels of audio to an au- sion System, whose design and implementation has been thor-
dience in a live performance context, via loudspeakers. Termi- oughly documented elsewhere [1, 2, 3]. The prototype system
nology that allows us to concisely describe the task of sound has regularly facilitated live performances since its inaugural
diffusion is defined. The conceptual model is described using performance in March 2004 and, at the time of writing, continues
this terminology. The model allows audio channels (sources) to serve as the principal sound diffusion system at the University
and loudspeakers (destinations) to be grouped logically, which, of Sheffield Sound Studios. The M2 system demonstrated some
in turn, allows for sophisticated abstract methods of control that significant advantages: it is portable quick to set up; signal rout-
supercede the restrictive ’one-fader-one-loudspeaker’ approach. ings can be re-configured comparatively quickly and easily; the
The Resound project – an open source software initiative con- interface supplements familiar fader-based diffusion paradigms
ceived to implement and further develop the conceptual model – rather than replacing them with unfamiliar methods. Three years
is introduced. The aim is, through further theoretical and practice- of practical experience with the system – as well as highlighting
led research into the conceptual model and software respectively, the benefits of the system – has revealed some unforeseen issues
to address the technical, logistical and aesthetic issues inherent [1, p247-265]. It is clear that significant improvements are still
in the process of sound diffusion. Keywords – Sound Diffusion, to be made in the technology of sound diffusion, and that future
Spatialisation, Software, Resound. systems will benefit from a thorough conceptual reappraisal of
what, exactly, sound diffusion involves.

1. INTRODUCTION
2. AUDIO STREAMS
Sound diffusion involves the reproduction of Audio Streams, via
loudspeakers, to an audience. A common application is in the For the purposes of this article, an Audio Stream is defined as
performance of electroacoustic music from CD, but the princi- any non-symbolic representation of audio, analogue or digital,
ples are relevant in a diverse range of activities including theatre including that which is recorded, synthesized or transduced via
sound, cinema, virtual reality, and indeed any scenario where microphones or pickups. The continuously varying voltage gen-
sophisticated real time control over the auditory environment is erated by a microphone is an example of an Audio Stream. It
required. Where references are made to the performance of elec- is a directly analogous representation, obtained via a process of
troacoustic music, it is to be understood that the same principles transduction, of the fluctuations in air pressure that constituted
could apply in a variety of sound diffusion scenarios. the original auditory event. It is not symbolic, unlike, for ex-
ample, western classical notation or Morse code: these are ab-
Audio Streams can originate from any combination of a wide
stract codified systems that need to be deciphered according to
variety of sources, including recording and playback devices (e.g.
predetermined schemes. Importantly, the Audio Stream is one,
CD players, audio sequencers), microphones and pickups, syn-
and only one, single channel of audio: a stereophonic record-
thesis technologies and audio processing technologies, any of
ing therefore comprises two Audio Streams. (Clearly the two
which may be implemented in hardware or software [1, p17-20].
Streams are related; we will return to this later.) We identify
The Audio Streams are reproduced via a loudspeaker array com-
the Audio Stream as an irreducible unit in the context of sound
prising a potentially large number of loudspeakers. A mixing
diffusion.
device (often an audio mixing desk but, increasingly, a software
implementation) acts as an intermediary between audio sources
and loudspeakers, along with an interface that allows the diffuser 3. COHERENT AUDIO STREAM SETS
(performer) to control the auditory results.
This research focuses mainly on the issues inherent in the Multiple Audio Streams are very often used concurrently to en-
real-time diffusion of multiple Audio Streams. The ergonomics code spatial information. Two-channel stereophony [4] involves
of the interface used to control the diffusion are therefore of the concurrent use of two Audio Streams to encode spatial infor-
key importance. Furthermore, because the specific technical de- mation with respect to a single (nominally horizontal) axis. In
mands of electroacoustic works – in terms of audio sources, the principle, this concept can be extended to any number of Audio
number of channels used, and so on – can vary considerably from Streams. A Coherent Audio Stream Set (CASS) is defined as
composition to composition within the context of a single con- a group comprising an arbitrary number of Audio Streams that
are co-dependent on account of having been used collectively
Email: [email protected], [email protected] to encode spatial information in this manner. Some examples
are given in Figure 1. Collectively, the constituent Streams of
the CASS represent a single, homogeneous spatio-auditory im-
age. Given an appropriate loudspeaker configuration it is pos-
sible to reproduce the spatio-auditory image embedded within
a CASS by simply mapping the constituent Audio Streams di-
rectly to loudspeakers. This distinguishes the CASS from higher
order encodings such as Ambisonics and Dolby’s AC3 encoding,
which require preliminary decoding in order to obtain the con-
stituent Audio Streams before reproduction over loudspeakers is
possible.

(a) Example loudspeaker array. (b) Three potential stereo CLS

mappings: L1-R1, L2-R2 and
L3-R3.

Figure 1. Coherent Audio Stream Sets (CASS)

4. COHERENT LOUDSPEAKER SETS

A Coherent Loudspeaker Set (CLS) is a group of loudspeakers (c) Two potential 5.1 CLS map- (d) 12 Channel CLS circular
pings. mapping with sequencial ”clock
arranged in such a way that it can be used to correctly broadcast a
face” ordering.
Coherent Audio Stream Set. For example, two loudspeakers ar-
ranged according to the normal stereophonic convention can be
used to correctly reproduce a stereophonic CASS. (What con-
stitutes ”correct reproduction” in this context is, clearly, open
Figure 2. Coherent Loudspeaker Sets (CLS). A selection of
CLS groupings are mapped onto the loudspeaker array.
to interpretation.) Similarly, five loudspeakers and a sub-woofer
deployed according to the accepted convention for 5.1, can be
used to correctly play back a 5.1 channel recording. A CLS
can comprise any combination of loudspeakers from the array (a 6. MATRIX MIXING
loudspeaker ”array” connotes all of the loudspeakers in the sys-
tem) that are deemed appropriate for the reproduction of a given Ultimately, sound diffusion concerns the dynamic distribution of
CASS. Thus, an array can contain many more CLSes than there Audio Streams from input channels to output channels. A flexi-
are loudspeakers, and any given loudspeaker could conceivably ble way of achieving this in practice is with a mix matrix, an ex-
be a member of several different CLSes. Some arbitrary exam- ample of which is illustrated in Figure 3. Input busses are shown
ples are given in Figure 2. as horizontal lines, output busses as vertical lines. The points
at which these lines intersect are referred to as input-to-output
nodes (or cross-points); these represent potential input-to-output
5. REDEFINING SOUND DIFFUSION buss routings. Each input-to-output node has an attenuator, as
opposed to simply being switched. In addition to the input-to-
Sound diffusion can now be defined as a practice that involves output nodes we also have the usual input nodes output nodes
the real time broadcast of one or more Coherent Audio Stream for each channel.
Sets via one or more Coherent Loudspeaker Sets. Nowadays it is
not at all uncommon for electroacoustic works to comprise mul-
tiple CASSes comprising variable numbers of Audio Streams.
Consider Jonty Harrison’s eight-channel work Rock ’n’ Roll, de-
scribed by the composer as follows:

I used a six-channel hexagonal [loudspeaker] array

as a surround,’ and I had two solo speakers – the
mains if you like [...] – for close up stuff. What I
would like to do [...] is be able to diffuse the twos
and the sixes completely independently.[5]

What Harrison describes is an eight channel work compris-

ing two Coherent Audio Stream Sets: one six-channel Set and
one two-channel Set. To be able to diffuse the two Sets indepen-
dently will clearly depend on a flexible mix architecture and an
equally flexible user interface. Figure 3. A simple 8-in-8-out Matrix mixer schematic
 Switched-buss mix architectures (as found in most conven-
tional mixing desks) necessitate fixed routing of inputs to out-
puts, and allow us simple control over buss input and output
levels only. In sound diffusion this can be limiting, especially
where multiple and/or greater-than-stereo CASSes are involved.
The mix matrix architecture, on the other hand, allows us to dy-
namically mix any input signal to any output channel in any pro-
portion. Clearly, this is very useful in the context of sound dif-
fusion. Accordingly, the mix matrix is increasingly frequently
implemented in sound diffusion systems, and forms the underly- Figure 4. Example showing the structure of a Collective
ing architecture of, amongst others, the DM-8 system [6], Rich-
mond’s SoundMan system [7], and the M2 Sound Diffusion Sys-
tem, which we will return to later. Recent developments at the 9. BEHAVIOURS
University of Birmingham have also explored the matrix archi-
A Behaviour determines precisely how the parameters of a Col-
tecture [8]. It has additionally been noted that matrix nodes can
lective will be acted upon. More accurately, a Behaviour has one
be used to accommodate extended signal processing, resulting in
or more of its own input parameters and uses these to manipu-
further increased flexibility [2, p36-38].
late the parameters contained within a designated Collective in a
In summary, the mix matrix provides us with a multitude of specific way. A Behaviour’s own parameters can be assigned to
parameters through which we can diffuse Audio Streams. The Physical Interface Devices, thus allowing interactive control.
complexity involved in the individual control of such large num-
As a very simple example, consider a Behaviour called ’Group.’
bers of parameters, however, becomes a concern.
The ’Group’ Behaviour has one parameter of its own – Level –
direct control of which has been assigned to Fader A. The Be-
haviour also has a Collective assigned to it, which contains a
7. PHYSICAL INTERFACE DEVICES AND
number of mix matrix parameters. When the value of Fader A
ABSTRACTED VERSUS DIRECT CONTROL
changes, its value is mapped on to all of the parameters of the
Collective, and thus the fader behaves like a ’group.’
The expression Physical Interface Device will be used to de-
In a more sophisticated example, a Behaviour may be de-
scribe any hardware means of user input into a system. Single
signed to make use of the ordered nature of the Collective in
axis slide potentiometers – faders – are the most common Phys-
some logical manner. A ’Chase’ or ’Sequence’ Behaviour could
ical Interface Devices in sound diffusion. A fader affords us
be used to iterate through the elements of a Collective in turn,
direct control over the value of a single matrix parameter only.
resulting in a semi-automated movement of sound between loud-
A simple eight-in-eight-out attenuation matrix (as shown in Fig-
speakers. In addition to a ’Level’ parameter it might also have a
ure 3) has eighty nodes – eight for the inputs, eight for the out-
’Frequency’ parameter, to determine the rate at which the chase
puts, and sixty-four for the input-to-output nodes – and would
sequence would iterate. In principle, any number of different
therefore require eighty faders. Clearly, then, a direct one-to-
Behaviours could be defined that interact algorithmically with
one mapping of faders to parameters is problematic. If the ma-
matrix parameters in various ways.
trix architecture is to be useful in the context of sound diffusion
it is necessary to have abstracted control over the matrix param-
eters, whereby the fader manipulates multiple parameters indi- 10. PARAMETER SUMMING PRINCIPLE
rectly via some predefined algorithm. If a fader is to afford us
abstracted control over multiple parameters in this way, then we Consider the possibility of individual matrix Parameters being
need to be able to specify – firstly – which parameters we want under the simultaneous control of multiple Behaviours: which
to control and – secondly – precisely what kind of control we Behaviour has final control over the Parameter? Of course we
want to have. could decide that multiplicitous control is not allowed, but this
is limiting. Another solution would be to allow the last access-
ing Behaviour to set absolute value, however, this is problem-
8. COLLECTIVES atic when Behaviours all act at the same time. What if we want
one Interface Device to act on the full range of a Parameter and
A Collective is used to specify which mix matrix parameters another to provide fine control over a small range? This exam-
(nodes) a Physical Interface Device will control. A Collective ple requires the Parameter to have two Behaviours operating on
is, essentially, a group of parameters in a particular order. More it: ’coarse’ and ’fine,’ if you like. This kind of action can be
specifically, it is a set of elements in a specific order, where each achieved by adopting a scheme whereby the influences of all of
element can contain one or more matrix parameters. This is il- the Behaviours affecting a Parameter are summed together to
lustrated in Figure 4. In programming terminology a Collective produce the final absolute value. A similar parameter summing
is simply a two-dimensional array of parameter addresses. A concept was employed in the M2 Diffusion System and is ex-
Coherent Audio Stream Set can, therefore, be expressed as a plained in greater detail elsewhere [2].
Collective whose elements can only be matrix input nodes. A
Coherent Loudspeaker Set – at least as far as the routing of in- 11. CLASSIFICATION OF CASS AND CLS ELEMENTS
puts to outputs is concerned – can be expressed as a Collective
whose elements consist only of matrix output nodes. In terms An advantage of software-based systems is that input-to-output
of how Parameters are addressed, a useful system – whereby ad- routing configurations can be saved to disk and re-loaded at a
dresses are given in URL form – is utilised in the OSC protocol later time. In theory, this should save time when setting up per-
[9]. We might, for example, describe an input-to-output node’s formances. However, loudspeaker arrays can vary considerably
attenuation Parameter address as follows: /matrix/in1/out1/att. from performance to performance in terms of the number and
relative positioning of loudspeakers. The routings for one setup A Collective-Behaviour combination can then be bound to a Phys-
may not map directly on to those of another, meaning that the ical Interface Device (e.g. a fader) for user control, and the pro-
system will have to be configured from scratch. cess of mapping Collective-Behaviours can be repeated for as
Consider the following scenario. As part of our diffusion many faders as there are available. Conflict between multiple
configuration we have a fader to control the level of a stereo Behaviours simultaneously acting on the same Parameters is re-
CASS sent equally to two stereo CLSes: ’rear’ and ’distant rear.’ solved via parameter summing within a fixed range. The model
When we raise the fader fully the stereo source emanates from is further refined by way of a hierarchical structure of abstract
both loudspeaker pairs. At a later date, we perform the same metadata identifiers for parameters, which can be used for semi-
work in a different venue, with a slightly different loudspeaker automated CASS-to-CLS routings and to increase interoperabil-
array, which only has one pair of loudspeakers at the rear of the ity between differing loudspeaker arrays.
hall. We now want the fader to diffuse to one pair of ’rear’ loud-
speakers rather than two. We are clearly dealing with closely 14. REFERENCES
related scenarios yet, frustratingly, there is no way to map be-
tween the old configuration and the new, as the former contains [1] J. Mooney. Sound Diffusion Systems for the Live
references to audio outputs that are either absent, or routed to Performance of Electroacoustic Music: An Inclusive
different loudspeakers, in the new configuration. Approach led by Technological and Aesthetical Con-
In order to make the mapping automatically, two things are sideration of the Electroacoustic Idiom and an Eval-
necessary. Firstly, we need a system whereby loudspeakers can uation of Existing Systems. PhD thesis, Univer-
be identified according to their role rather than by which audio sity of Sheffield, 2005. Online at http://www.james-
outputs they are connected to. This way loudspeaker pairs could mooney.co.uk/research/sound diffusion.
be labelled ’rear,’ ’distant rear,’ and so on. Secondly, we need to
[2] D. Moore. Real-time Sound Spatialization, Soft-
be able to identify ’rear’ and ’distant rear’ as functionally related.
ware Design and Implementation. PhD the-
More specifically, ’distant rear’ could be defined as a hierarchical
sis, University of Sheffield, 2004. Online at
subdivision of ’rear.’ This process has been described more fully
http://music.dept.shef.ac.uk/phd thesis/rtss.pdf.
elsewhere [1, p278-80].
Such labelling could also be applied to CASS elements. (Re- [3] A. Moore, D. Moore, and J. Mooney. M2 Diffusion The
call that in our conceptual model CASSes and CLSes are both Live Diffusion of Sound in Space. In Procedings of the
simply special instances of parameter Collectives.) This means 2004 ICMC International Computer Music Conference,
that mappings between CASSes and CLSes could take place pages 317–320., Miami FL., November 2004.
semi- or fully automatically if required. The left channel of a [4] A. Blumlein. Improvements in and relating to Sound-
stereophonic CASS – let’s call this Audio Stream A – might be transmission, Sound-recording and Sound-reproducing
associated with the labels ’Mono’ and ’Left.’ Two loudspeak- Systems. International Pat. No. 394325, 1931. Online
ers within a CLS – Loudspeakers x and y – might be classified at http://www.doramusic.com/patents/394325.htm (17th
as ’Mono,’ ’Left’ and ’Mono,’ ’Right,’ respectively. If asked to June 2007).
identify a suitable automatic mapping for Audio Stream A, we
[5] J. Mooney, editor. Sound Diffusion Systems for the Live
would choose Loudspeaker x, because its identifiers match more
Performance of Electroacoustic Music: An Inclusive Ap-
closely the labels of the Audio Stream A than do those of Loud-
proach led by Technological and Aesthetical Consideration
speaker y.
of the Electroacoustic Idiom and an Evaluation of Existing
In effect, what we are describing is a hierarchical structure of
Systems, chapter Interview with Jonty Harrison. 2005.
abstract metadata identifiers.
[6] C. Rolfe. A practical guide to diffusion. 1999. Online
at http://cec.concordia.ca/econtact/Diffusion/pracdiff.htm
12. RESOUND (20th June 2007).
In order to realise the conceptual model and develop it further, [7] Richmond Sound Design. Soundman server theatre sound
the authors have founded the Resound project [10]. This is an design / audio sound control software. 2007. Online
open-source software project licensed under the GNU Public Li- at http://www.richmondsounddesign.com/sm-sspecs.html
cense (GPL). The software comprises a server component, which (20th June 2007).
amongst other things implements the mix architecture and pa- [8] S. Wilson and J. Harrison. Introduction to beast’s
rameter addressing paradigms described, and a graphical client, diffusion research project. 2007. Presenta-
which deals with user input as well as facilitating the manipula- tion given at CBSO Centre, Birmingham, 27th
tion of Collectives and Behaviours. May 2007. Reference online at http://www.ea-
studios.bham.ac.uk/BEAST/concerts/2007.html.
13. CONCLUSION [9] M. Wright and A. Freed. Open sound control: A new
protocol for communicating with sound synthesizers. In
In summary, the conceptual model of sound diffusion is as fol- Procedings of the 1997 ICMC International Computer Mu-
lows. The ’raw materials’ are Audio Streams; these are grouped sic Conference, pages 317–320., Thessaloniki, November
in to Coherent Audio Stream Sets (CASSes). The CASSes are 1997.
then diffused, via a process of abstracted control over a mix
[10] J. Mooney and D. Moore. Resound project. Online at
architecture comprising a matrix of Parameters, to an array of
http://resound.sourceforge.net (19th June 2007).
loudspeakers comprising multiple Coherent Loudspeaker Sets
(CLSes). Abstracted control is achieved by defining two things:
a two-dimensional array of target matrix parameters (a Collec-
tive), and the way in which these parameters behave (a Behaviour).