UNIVERSITY OF CALIFORNIA

RIVERSIDE

Compositional Strategies in Spectral Spatialization

A Dissertation submitted in partial satisfaction

of the requirements for the degree of

Doctor of Philosophy

in

Music

by

Martin Jaroszewicz

March 2015

Dissertation Committee:

Dr. Paulo C. Chagas, Chairperson

Dr. Miller Puckette
Dr. Ian Dicke
 Copyright by
Martin Jaroszewicz
2015
The Dissertation of Martin Jaroszewicz is approved:

Committee Chairperson

University of California, Riverside

 Acknowledgments

I am deeply grateful to my supervisor, Dr. Paulo C. Chagas, for his teaching and the

guidance over the past four years. Without his incredible knowledge, this thesis would

not have been possible.

I would also like to thank Wouter Snoei from the Game of Life Foundation and

Andreas Piesiewicz from TU Berlin for their time and support.

Immense gratitude to Hans-Ulrich Werner for giving me the opportunity to

be in residence at Hochschule Offenburg and connecting me with all the people and

resources pertaining to this thesis.

Finally, I am very grateful to my wife for all the sacrifices and infinite support.

iv
To my beloved wife.

v
 ABSTRACT OF THE DISSERTATION

Compositional Strategies in Spectral Spatialization

by

Martin Jaroszewicz

Doctor of Philosophy, Graduate Program in Music

University of California, Riverside, March 2015
Dr. Paulo C. Chagas, Chairperson

The use of space as a structural part of a composition is a complex issue that

involves rethinking the idea of the sound object – as described by Pierre Schaeffer in his

Treatise on Musical Objects (1966) – and involves considering movement as a means to

manipulate spectrum in the frequency domain. In addition, the contemporary composer

– especially when writing acousmatic music – needs to consider the materiality of the

studio where music is created. These interrelated factors and others, such as the means

of reproduction, the acoustics of the venue, the choice of loudspeakers and eventually

the software that executes the algorithms for complex calculations of movement and

other parameters of sound spatialization, are part of the techné involved in the creative

process. 1 Thus, the choice of sound material and the aesthetics of movement should be

considered to be elements intrinsically related to the technical means of the performance

as spatialization produces an impact on the timbral footprint of sound.

One of the motivations behind the research conducted at EARS was to answer

Agostino Di Scipio’s question: “How can I design the tools that are necessary to realize

my own idea of composition ?” (Scipio 1995a). The other one, was to realize some ideas

inspired by Borges’ story The Library of Babel, which became the starting point behind
1
For a discussion about the significance of techné see (Manning 2006).

vi
the Laberintos series of electroacoustic études, each one related to one aspect of working

with space and using the electronic music techniques described in this thesis.

There is a discussion involving the meaning of working with space from the

point of view of a composer of contemporary electroacoustic music. This dissertation

deals with strategies for working with sounds objects whose trajectories are predeter-

mined by the composer during the composition process, as opposed to using space as a

resonant body to “enhance” music. A new definition of the sound object that considers

its “materiality” and its relationship with space is discussed. It views sound–objects

from a different perspective than Pierre Schaeffer, considering them as volumetric ob-

jects that occupy a space like any object or “thing” – a view in tandem with speculative

realism (Harman 2011). These sound objects are real; they can travel in space and

acquire a tangible property.

I describe a unique approach for dealing with sound trajectories in spatialized

music using 3D modeling software. Although, none of these applications were designed

for music or sound and lack any synthesis capabilities – besides the creation of basic

waveforms – it is possible to extract the data containing the coordinates of a 3D virtual

space in the form of a simple text file. In addition, Blender 2 features a game engine

that can be used to send Open Sound Control (OSC) messages in real time using the

Python scripting language. As an alternative, I have created several externals written

in C for the Pure Data language for the creation of sound trajectories using ideas taken

from parametric design. These objects offer the possibility to work using algorithmic

and stochastic approaches to spatialization simplifying the compositional process. The

technical part of this dissertation deals with the tools I have developed at EARS and

how can be implemented using the techniques exemplified with my own work.
2
Open source 3D graphics and animation software.

vii
 Another aspect of this thesis is the description of the systems available for

spatialization in music and how they differ from commercial systems. At EARS, I

built – and worked with – tools that were specifically designed for composition and

spatialization thus contributing to the techné and the aesthetics that influenced my

musical ideas. The result of the research conducted at EARS and the experienced gained

working with Wave Field Synthesis (WFS) systems – the Game of Life in Netherlands

and the systems at the Technische Universitt Berlin – not only generated new apparatus

but suggested a new rethinking of composition in space.

From a technical point of view, among the programming languages for music

and sound, I found that Pure Data,3 Supercollider,4 Faust, and Chuck are suitable open

source tools for the composer working with electronic music as they are portable and

available for most platforms including Mac, Linux and Windows. 5 In addition, Pure

Data, with its visual approach, is a great tool for quickly sketching musical ideas and

for demonstrating the theory and technique of electronic music without writing lines of

code.

3
Pure Data PD is an object based graphical environment for sound synthesis developed by Miller
Puckette, professor at the University of California San Diego. In PD, like in Supercollider, it is possible
to create custom synthesizers, effects, musical patterns, and sonic and musical machines by connecting
on-screen patch cords, but most importantly, PD is a great tool for sound research, analysis and re-
synthesis.
4
Supercollider is an object-oriented programming language designed specifically for describing sound
processes in real time. SuperCollider was written by James McCartney and is now an open source
(GPL) project maintained and developed by various people. It is used by musicians, scientists, and
artists working with sound.
5
Software portability is the usability of the software in different platforms. For example, a Pure Data
patch can be run in a Macintosh, Linux or Windows computer without any modification of the code or
need to recompile.

viii
Contents

List of Figures xi

I Theoretical Aspects 1

1 Introduction 3
1.1 Space and Spatialization . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.2 Immersion and Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.3 From Stereo to Wave Field Synthesis . . . . . . . . . . . . . . . . . . . . 17
1.4 Acoustic spatialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.5 Electroacoustic Spatialization . . . . . . . . . . . . . . . . . . . . . . . . 28
1.6 Gesang der Jünglinge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
1.6.1 Circular Sound Space . . . . . . . . . . . . . . . . . . . . . . . . 31
1.6.1.1 Répons by Pierre Boulez . . . . . . . . . . . . . . . . . 33
1.7 The Sound Source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
1.8 The Aesthetics of Circular Motion . . . . . . . . . . . . . . . . . . . . . 42

II Technical Aspects 47

2 Sound Fields 48
2.1 Stereo and Panning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.1.0.2 Simple Linear Panner . . . . . . . . . . . . . . . . . . . 49
2.1.0.3 Surround . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.1.0.4 Ambisonics . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.1.0.5 Wave Field Synthesis . . . . . . . . . . . . . . . . . . . 59
2.2 Characteristics of the WFS . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.2.0.6 Delays, Doppler and Distance Cues . . . . . . . . . . . 68
2.3 The Design of The Portable Game of Life System . . . . . . . . . . . . . 69
2.4 The TU Berlin Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

3 Spectral Spatialization 74
3.1 Spectral Domain. The Fourier Transform and its musical applications . 76
3.1.1 Windowing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.1.2 Window creation with Max/MSP . . . . . . . . . . . . . . . . . . 82
3.2 FFT in Max/MSP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

ix
 3.3 Cross-synthesis and Spectral Shredding . . . . . . . . . . . . . . . . . . 87
3.4 Tools and Software for Spatialization . . . . . . . . . . . . . . . . . . . . 90
3.5 Some Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.6 Envelopes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
3.6.1 The Sound Envelope as a metaphor for composition . . . . . . . 94
3.7 Parametric Tools for Spatialization . . . . . . . . . . . . . . . . . . . . . 97
3.8 Parametric Spatialization . . . . . . . . . . . . . . . . . . . . . . . . . . 97
3.8.1 HOA Library externals . . . . . . . . . . . . . . . . . . . . . . . 98
3.8.1.1 hoa.2d.decoder∼ . . . . . . . . . . . . . . . . . . . . . 98
3.8.1.2 hoa.optim∼ . . . . . . . . . . . . . . . . . . . . . . . . 99
3.9 Parametric Spatialization objects for PD . . . . . . . . . . . . . . . . . . 100
3.10 Parametric Spatialization using 3D Procedural Animation Software . . . 106

III Laberintos 112

4 Laberintos 113
4.1 Laberinto 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
4.2 Laberinto I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
4.3 Laberinto III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
4.4 Laberinto IV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
4.5 9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

5 Conclusion 152
5.1 Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
5.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
.1 Appendix A - mj Library Source Code . . . . . . . . . . . . . . . . . . . 160
.1 Appendix B - Spectrograms . . . . . . . . . . . . . . . . . . . . . . . . . 169

x
List of Figures

0.1 Spatialization with the mj library for Pure Data and HOA . . . . . . . . 2

1.1 Odeion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 The simplest stereo panorama produces a void in the center . . . . . . . 18
1.3 The dome of San Marco basilica in Venice . . . . . . . . . . . . . . . . 21
1.4 Beginning of Mozart’s Notturno for 4 orchestras . . . . . . . . . . . . . 22
1.5 Some of the possible spatialization layouts in Mozart’s Notturno for 4
orchestras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.6 Berlioz’s Rex Tremendae . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.7 Distribution plan for Voyage Four by Henry Brant . . . . . . . . . . . . 27
1.8 Excerpt from the manuscript of Gesang der Jünglinge . . . . . . . . . . 30
1.9 Répons’ sound projection . . . . . . . . . . . . . . . . . . . . . . . . . . 34
1.10 Répons’ spatial switch . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
1.11 “Wall” of Speakers at the Zuiderstrandtheater in Den Haag . . . . . . . 44
1.12 Elliptical and straight paths in a WFS system . . . . . . . . . . . . . . . 46

2.1 Theoretical width of an audio image. . . . . . . . . . . . . . . . . . . . . 48

2.2 Simple Linear Panner in PD with Low Pass Filters for smoothing the
signals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.3 Square root panner (left) and Sine-Cosine panner (right) . . . . . . . . . 50
2.4 Logic Pro X. Panning using automation. From right to left to center in
2 seconds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.5 Stereo panning in Max/MSP . . . . . . . . . . . . . . . . . . . . . . . . 52
2.6 XY Stereo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.8 Ambisonics directivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.9 Directivity of a point source encoding with different Ambisonic Levels
(Pérez-López 2014). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.10 Different ambisonics orders and their directionality in Max. The patch
was developed at EARS for the works Laberintos. . . . . . . . . . . . . . 57
2.11 Speaker array, point source and wave generated by the system. . . . . . 60
2.12 Wave Field Synthesis. Huygens’ Principle (Snoei 2014) . . . . . . . . . . 60
2.13 Kirchhoff-Helmutz. Illustration of the geometry (Sascha 2008). . . . . . 61
2.14 Wave Field 2D plane and listener’s perspective . . . . . . . . . . . . . . 61
2.15 WFS Spherical Harmonics . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.16 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2.17 Block diagram of the process generated by Faust svg compiler . . . . . . 67

xi
2.18 a complex path created with the WFSCollider software . . . . . . . . . . 70
2.19 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.20 WFS studio at TU Berlin controlled by parametric spatialization tools
written by the author for the Pure data software. . . . . . . . . . . . . . 72
2.21 Speaker layouts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

3.1 Spectrogram of an entire performance of Pierre Boulez’s Répons . . . . 80

3.2 Hann Window and resultant spectrum . . . . . . . . . . . . . . . . . . . 81
3.3 Hanning Window in Max . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.4 Dividing a spectrum into 8 different buffers . . . . . . . . . . . . . . . . 84
3.5 Generalized cross–synthesis with gen˜in Max/MSP . . . . . . . . . . . . 88
3.6 Source–filter cross–synthesis in cepstrum domain with Max . . . . . . . 90
3.7 Swarm behavior of eight sources after spectral shredding . . . . . . . . . 91
3.8 Percussion, triangle, linear and sine envelopes created in Supercollider . 93
3.9 A more complex envelop applied to a sawtooth oscillator . . . . . . . . . 94
3.10 The shape of the increasing texture resembles a Gaussian envelope . . . 95
3.11 The PD patch showing the creation of an envelope (ADSR) applied to a
table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
3.12 Speakers configuration at EARS . . . . . . . . . . . . . . . . . . . . . . 99
3.13 Simple and complex paths can be created with mjRose . . . . . . . . . . 102
3.14 3-D model of a point cloud at EARS . . . . . . . . . . . . . . . . . . . . 104
3.15 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
3.16 mjRose with different values for n, d and z . . . . . . . . . . . . . . . . 107
3.17 A 3D mesh (left) with a wave deformation (right) in Blender. . . . . . . 108
3.18 Different views of a point cloud. In addition to the creation of curves
using parameters or manually drawing curves, 3D software allows the
composer to visualize the curve in real time from any angle and modify
any vertex or point as desired. . . . . . . . . . . . . . . . . . . . . . . . 110
3.19 Point Cloud . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
3.20 Different paths created with Houdini and Python. See Appendix A for
source code. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

4.1 A contorted passage and path through the library. . . . . . . . . . . . . 114

4.2 A path applied to an ambisonics field . . . . . . . . . . . . . . . . . . . . 115
4.3 Some trajectories used in Laberintos’ ambisonics field. . . . . . . . . . . 115
4.4 Laberinto V screenshot . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
4.5 Algorithmically created letters in Houdini. . . . . . . . . . . . . . . . . . 118
4.6 Laberinto V Frame. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
4.7 Images only move forward while sounds move in all directions. . . . . . 119
4.8 Two excerpts. The first shows micro-glissandi (long sounds); the second,
short sounds. These gestures are spatialized using rotation and point
cloud respectively. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
4.9 Envelopes used for micro-pitch modulation . . . . . . . . . . . . . . . . 126
4.10 A spiral curve used in Laberinto III . . . . . . . . . . . . . . . . . . . . 127
4.11 First measure of Laberinto Borgiano . . . . . . . . . . . . . . . . . . . . 135
4.12 Main motifs in Laberinto Borgiano . . . . . . . . . . . . . . . . . . . . . 135
4.13 Time envelopes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
4.14 A random or ”pseudo random” distribution in PD . . . . . . . . . . . . 150

xii
4.15 Georges Seurat - ”Circus Sideshow” (1889) - detail showing pointillism
technique. Public domain image. . . . . . . . . . . . . . . . . . . . . . . 151

.1 Different Cross-synthesis techniques. . . . . . . . . . . . . . . . . . . . . 169

.2 Sounds with contrasting spectrum. . . . . . . . . . . . . . . . . . . . . . 170

xiii
 Part I

Theoretical Aspects

1
Figure 0.1: Spatialization with the mj library for Pure Data and HOA

2
Chapter 1

Introduction

El universo (que otros llaman la

biblioteca) se compone de un número

indefinido, y tal vez infinito, de

galerı́as hexagonales...

La biblioteca de Babel

JORGE LUIS BORGES

1.1 Space and Spatialization

The use of technology for creating listening spaces to enhance music goes back

to ancient Greece and Rome where engineers built special theaters for music and speech.

Aristotle – 4th century B.C.– knew that the law of reflections could be applied to sound

as well and refers to the analogy between light and sound in his explanation of echoes

and resonances, demonstrating how timbre changes due to distance and absorption.1
1
See Aristotle’s Problemata. Book XI (Aristotle 2001).

3
 A different use of space concerned with music is found in antiphonal practice

– Medieval Gregorian Chant and Responsorial Psalmody – performed during religious

services in the 4th century.2 The difference between antiphonal and responsorial is

the manner of performance. Antiphonal features alternating choirs while responsorial,

a choir responding to a soloist. There were also secular chants based on non-biblical

texts, prose and poetry.

During the 16th. and 17th. centuries composers took advantage of the spatial

properties of cathedrals, such as the St. Mark’s Basilica of Venice, to “color” their music

by exploring timbre possibilities thus expanding their musical ideas. This practice can

be seen in the works of composers from the Venetian region who often wrote for double

chorus. The style grew with the polychoral music of Giovanni Gabrieli (ca. 1553–1612)

who explored the sonic capabilities of mixing up to seven choruses combining different

timbres: low or high voices mixed with diverse instruments.

The practice in San Marco in Venice raises questions about music and space.

Did the composer selected the performance site and how musicians should be placed?

How these decisions affected the work of the composer? Any orchestral work requires

a specific layout, for example, in a standard orchestra we find the violin section to the

left of the conductor – and the audience – and double basses to the right. If halls did

not have reverberation the listener would have a completely different experience as he

would be perceiving great spatial separation.3 In addition, higher frequencies – from

violins – and lower frequencies from the bass section would definitely be perceived as a

sort of spectral separation.

2
In this manner of performing, a soloist or reader sings the first part of the psalm, typically two
halves of a verse, and a choir or congregation responds singing a second part.
3
Note that the layout of the orchestra has nothing to do with the range of the instruments but instead
with volume and visibility. For an example see: Di Grazia, Donna.19th-Century Music Vol. 22, No. 2
(Autumn, 1998), pp. 190-209.

4
 4
Figure 1.1:

The law of reflections – produced by sound bouncing from a surface – was well

known by the Greeks. The work of the Roman Marcus Vitruvius Pollio De Architectura

libri decem (c. 30 BC.) provides an interesting account of early thought in room acoustics

and the use of different materials for acoustic conditioning applied to music and speech.

Vitruvius’ work explains how spaces with no reverberations – such as the open air –

are more suitable for speech intelligibility and reverberant spaces are adequate for the

performance of music (Pollio 2008, 149). For music performance, there were vessels under

the seats of the theaters that in theory would resonate with the frequencies produced

by the instruments:

The Greeks knew that the clear but dry acoustics of the open-air theatron(in
Greek “a place to see”) were excellent for speaking but inadequate for music.
Therefore, the odeion (Greek “a place to hear”) was developed for song per-
formances during musical competitions. A hall of limited size with excellent
acoustics erected on a rectangular ground plan, the odeion featured steeply
raking rows of seats and a flat horizontal ceiling. The walls and the ceiling
with structured surfaces produced well supported reflexiones consonantes,
the open window absorption (Baumann 2011).

4
Odeion. McGraw-Hill Dictionary of Architecture and Construction. Copyright ©2003 by McGraw-
Hill Companies, Inc.

5
 In many cases, large reverberant spaces such as the domes of basilicas of the

roman tradition can produce reverberations times of eight seconds of duration 5 due to

the angles of incidence and non-absorbent materials of their walls.

Throughout the 17th and 18th centuries many architectural techniques were

used to reinforce or “enhance” sound including the use of water under the orchestra 6

and/or the audience. Some of these ideas had unpredictable or no results at all and

some did work. Only during the mid 19th century, the acoustics of the halls were

greatly improved with efficient architecture. According to Dorothea Baumann in her

book Music and Space: A systematic and historical investigation into the impact of

architectural acoustics on performance practice followed by a study of Handel’s Messiah:

“this period deserves to be called the ”golden age” in the history of opera houses and

concert halls” (Baumann 2011, 41).

In the mid 19th, Berlioz writes about the importance of room reverberations

and the effect on the listener: “And one must vibrate with the instruments and voices

and because of them in order to have a genuine musical sensation” (Berlioz 1994).

There is evidence of a clear separation between acoustic spaces designed for

speech and those designed for music. Moreover, the inclination to use reverberant spaces

for music shows that music was meant to be perceived without being localized. In other

words, music needed to “fill” the space as one amalgamated form. On the other hand,

theater pieces took advantage of dry spaces in order for the audience to be able to

localize the voice of the actor/s on stage, for example, to easily follow a plot.7
5
Reverberation – Reverb – time is calculated by measuring the time a reverberated sound takes to fall
to 1/1000 of its peak amplitude. This is referred as the RT60 value because it represents a 60 decibels
(dB) drop from the intensity of the first reflection.
6
The word orchestra originally meant the semi-circular space in front of a stage in a Greek theater.
7
For a brief history of the development of architectural acoustics see Shankland (Shankland 1972)
and for an example of a scientific analysis of the acoustics in a Greek theater see Declercq (Declercq
2007).

6
 The technology for sound reproduction – including the invention of the gramo-

phone in the late 19th century and the use of filters in the 20th century – made it possible

to recreate acoustic spaces with artificial reverberation. The technique is often used in

pop and rock music as recording studios are “dry”, that is, they are designed to not

produce any reverberations . In “classical” music, reverb may be not necessary as most

performances are recorded in “live” halls as opposed to studios. Moreover, the musician

often needs to adjust his playing to the acoustics of the concert halls and playing in a

dry space requires different techniques.

The more realistic the artificial space is reproduced, the more important is

for the conductor to adjust the balance of the orchestra. Then, it is the task of the

architects, engineers, conductors and performers to adjust those parameters to create

and aesthetics such as in the music of the cathedral of San Marco. In addition, it is

possible to virtually recreate any concert hall or space by using convolution reverb, a type

of reverb that is obtained by “capturing” the space with microphones and convolving

the result with the music.

A different aspect of the use of space is the concept of spatialization which is,

as I see it, part of the compositional process; it can be incorporated on the “score” and

in the material supporting the work. That is, as a parameter that can be “written”

in the music by the composer whereas as part of the score or by the use of techniques

and practices – algorithmic or “composed” –Moreover, in this thesis, I use the term

spatialization to refer to the techniques for moving acousmatic sounds from a virtual

source in a real space using computer algorithms and parametric implementations that

can be written in the score thus becoming a structural part of the work.8 Spatialization
8
As are dynamics, pitch collections, tempo markings, etc.

7
as a parameter of composition has strong links to serialism and Elektronische Musik is

especially suitable for working with timbre composition using parametric techniques:
The methods of timbre composition concentrated on parametric and com-
binatorial thinking link elektronische Musik to the polyphony of the Middle
Ages and the Renaissance and to the experimental music of the beginning of
the 20th century, particularly the aesthetics of atonalism, twelve-tone music,
and serialism (Chagas 2014).

Although spatialization deals with localization and movement and not with the

use of natural or artificial reverb, it is possible to combine these two different ideas as

composer Pierre Boulez did in the 20th Century with his work Répons .9

While spatialization does not concern with the acoustics of the space, the

manner of playing or the conductor’s decisions, the concepts are interconnected. Paulo

C. Chagas, in his book Unsayable Music clearly explains the connection between space,

reverb, technology and spatialization:

The acoustics properties of the performance hall play an important role in
determining how the listener perceives the music. Reverberation for example,
can significantly impact both the performance and the listening experiences.
It affects the clarity and liveliness of the sound and the localization of the
sound source. In the case of electroacoustic music, the listening experience
also depends on the quality of the sound system: the audio equipment,
the characteristics of the speakers, and their distribution throughout the
performance space, Sound spatialization, a crucial aspect of electroacoustic
music performance, is in fact a virtual construction of sound space; the
virtual space can be generated by the composition itself[..](Chagas 2014).

It is in the 20th. century when composers could and began to explore other

aspects of the use of space other than the one imposed by the architecture of the hall.

This was possible by the invention of the tape recorder and the loudspeaker. In the last

fifty years, with the growth of the movie industry, new developments in audio spatializa-

tion and multichannel systems such as “surround”, that offer an immersive experience to
9
Pierre Boulez recreates the massive sonorous climaxes of antiphonal chant with his work Répons for
ensemble, soloists and spatialized electroacoustic sounds. Boulez’s work, in addition to a virtual sound
space created by the circular motion of acousmatic and electronic sounds around the listener, presents
a composer who is aware of the spectral changes imposed by the acoustics of the performance space.
Moreover, a closer look at the orchestration and electronic transformations in Répons, reveals how the
composer methodically works with the timbral possibilities of spatialization (Jaroszewicz 2013).

8
the movie aficionado, had been commercialized and standardized. Unfortunately, com-

mercial surround techniques have little to offer to contemporary music; their surround

panning techniques do not work in tandem with the aesthetics of contemporary music

adding little to nothing to the development of spatial techniques and aesthetics.

A very special case in the history of electronic music and the development

of spatial techniques is Karlheinz Stockhausen’s Gesang der Jünglinge [Song of the

Youths] (1955-56), which was the first piece to use multiple loudspeakers creating the

psychoacoustic illusion of circular motion around the listener: an aesthetics that has

influenced composers of electroacoustic music since the 1950s (Decoupret 1998). But

the most salient aspect of the work is undoubtedly the linkage of movement and spectral

transformations:
The revolutionary accomplishment of Gesang der Jünglinge is to provide the
first example of the integration of the composition of timbres with spatial
articulation (Chagas 2014, 111).

Being the first electroacoustic work to integrate space and timbre as part of the

composition process, Gesang der Jünglinge opened the path to new possibilities for the

composer of electronic music. Moreover, it changed the way electroacoustic music should

be performed including its technological requirements. Spatialization needs ”dry” spaces

as opposed to large reverberant halls – such as those of churches – detaching music from

the tradition of the concert hall or the church .10

To further clarify the concept of spatialization, we can infer that there are three

different aspects of music and space according to several parameters; the last being the

topic of this dissertation:

1. Architectural Space: The use of technology and architecture to “enhance” music

by modifying physical spaces.

10
For a discussion on the role of the concert hall see Boulez’ Orientations: collected writings(Boulez
1986, 448).

9
 Table 1.1: Different interpretations of the use of space in music.

Architectural Performance Spatialization

Space Space
Example Greek Theater Concert Hall Wave Field Syn-
thesis
Function To enhance or To enhance music To add movement
suppress rever- to sound
berations
Controlled by Shape and Mate- Conductor / Per- Composer
rials formers
Sound Directionality Omni/Direct Omni Direct
Listener’s attention Non- Non-Focused Focused
Focused/Focused
Structural in Composition No No Yes

2. Performance Space: The use of orchestration as a “special effect” with the

intention to modify timbre.

3. Composed Space or Spatialization: The use of space as a container for creating

movement as a structural part of a composition. 11

In order to incorporate spatialization to my own work, I had to approach the

task of composing music from a different perspective. Working with space adds another

level of complexity and I thought spatialization had to be “composed” with the music.

The analogy is the use of dynamics which some composers incorporate last as if it was a

parameter of less hierarchically value. But spatialization, as any aspect of composition

can be approached from different angles and using different techniques. One of them is

algorithmic composition using stochastic methods for moving sounds in space. Another

one is the use of simple trajectories such as circular and elliptical paths. Lastly, the
11
Table 1.1 shows how correlations between music and space result in different ideas that are clearly
defined by their function and the intention of the composer to alter – or not – timbre. It also takes
into consideration the point of view of the listener who has the ability to focus on different aspects of
music such as timbre, harmony, melody, etc. When spatialization is part of the structure of a work,
the intention of the composer is to draw the attention of the audience to movement and the spectral
changes linked to the movement of the sound object as it travels through predefined – composed or
algorithmically generated – trajectories.

10
“composition” – written in the score or somehow notated – of paths that work in tandem

with the music.

Moreover, there are two aspects of composing spatialization: the creation of

trajectories and the spectral transformations sound undergoes as it moves. I found

that a good composition practice was to work with these two elements at the same

time. For example, as a sound moves from point A to point B, it undergoes a timbral

transformation using cross-synthesis techniques in the spectral domain, which I found

to be the best way to link movement and spectra. At EARS,12 I had the opportunity

to developed different strategies by using its 8 channel movable system.

At EARS, I experimented with different speaker layouts beginning with a circle

– for Ambisonics – and finally placing speakers accordingly to the shape of the room

and adjusting for distance using delay and gain compensation techniques. The software

tools at EARS allowed me to try different cross-synthesis techniques. I was also able to

experiment with different approaches for the creation of trajectories. There were several

steps in the creation of a system that can be used to incorporate spatialization as a

structural part of composition and the following items were considered:

• Speaker layout: circular / square / room-adapted

• Spatialization techniques: SPAT / Ambisonics / WFS

• Cross-synthesis techniques : Mixing, source-filter / convolution / cross-modulation

/ square-root convolution / cross-product

• Trajectories: Algorithmic / Parametric / Stochastic

The following chapters introduce the work I developed at EARS, including

different aspects of spatialization techniques, new technological developments, a reflec-

12
The Experimental Acoustic Research Studio at the University of California Riverside is a satellite
facility founded by Paulo C. Chagas and expanded into a studio for spatialization in collaboration with
the author.

11
tion on the aesthetics of sound trajectories, and a brief historical background for the

purpose of connecting ideas about timbre and space, always from the point of view of

composition. The reader will be able to learn about recent practices in sound spatial-

ization and get acquainted with the tools and technologies available to the composer of

contemporary music.

12
1.1.1 Definitions

Some definitions are provided to clarify and introduce the reader to spatializa-

tion terminology:

1. Monoaural: Listening through one ear.

2. Binaural: Listening through two ears. This is the natural way of listening to

sounds. The term is also used for “binaural” recordings , which are made with

an anatomically correct dummy head that captures sounds binaurally with two

microphones attached to its “ears”. There are also binaural microphones that

are designed as earplugs – like headphones – and are able to record binaurally

using a human head instead. The reproduction of binaural recordings is through

headphones as this type of recording technique produces excellent separation of

the sounds.

3. Stereophonic: The word “stereo” comes from the Greeks meaning “solid” re-

ferring to three-dimensional. Originally stereo recordings were intended to be

reproduced through two loudspeakers creating a psychoacoustic three-dimensional

image which in fact is not “stereo” in the strict meaning of the word. Stereo

systems can have more than two channels.

4. Multichannel: A system than can reproduce more than one channel. At the

present time, commercial systems offer up to 7.2 channels, that is seven surround

channels and two subwoofers. There are 10.2 and 22.2 systems not commercially

available yet. The definition includes all the systems listed below.

5. Surround: I use this term referring to multichannel commercial systems such as

5.1 and 7.1 surround. These include an array of speakers surrounding the listener.

The most common layout is the following 13 : Front Left(L), Front Center(C),Front
13
Clockwise.

13
 Right(R), Surround Right(SR) and Surround Left(SR). the sub-woofers can be

placed anywhere as low frequencies cannot be spatialized.

6. High Order Ambisonics A technique for creating a sound field requiring a full-

sphere array where speakers are distributed in a dome surrounding the listener.

7. Wave Field Synthesis A technique for spatialization of sounds in a 2D field with

an array of speakers completely surrounding the listener.

It is convenient for composers to write acousmatic music for commercial sur-

round formats as systems are widely available and most Digital Audio Workstations,

such as ProTools and Logic, offer plugins or built-in tools for bouncing surround files

and for creating surround automation. Moreover, if writing for multimedia, it is possible

to work with pictures and sounds in the same software environment. Multimedia works

– such as my Laberintos V – can be “performed” in a variety of venues including a movie

theater setting; they can be encapsulated in a single file and stream or store in a DVD.

Although surround is convenient, it is limited to positioning and moving sounds from

speaker to speaker using standard panning techniques. The focus of attention is not the

sound itself and movement usually adds little to nothing to the composition. For this

reason, alternative systems have been developed exclusively for music with emphasis on

contemporary practices and acousmatic music.

1.2 Immersion and Fields

The idea of immersion underlies the development of sound spatialization .14

Multichannel systems position the listener inside a space surrounded by loudspeakers

which in addition act as physical boundaries. Surround techniques, which were originally
14
Immersion – without interactivity – creates a sense of depth integrating the listener and the space.
Sounds can be perceived as moving from point to point and in the case of surround systems, they can
only be perceived as moving from speaker to speaker, and only left to right and the opposite.

14
developed for the movie theater, are limited to the movement of sounds which is the

product of a psychoacoustic effect. Moreover, there is no correlation between movement

and timbre.

The use of multichannel systems by composers of contemporary music such as

Karlheinz Stockhausen and Paulo C. Chagas influenced the development of immersive

sonic environments where the listener can experience timbral changes and movement

interrelated as parameters of composition. This idea is further explored by the use of

devices such as the gmebaphone (1973), which is capable of doing spectral spatialization

by splitting frequencies into multiple channels and using cross-over techniques.15 It was a

combination of instruments specially designed for the diffusion of electroacoustic works.

These instruments were built under the premise of considering diffusion of electroacoustic

music inseparable from the process of composition. Christian Clozier argues that music

is a complex ensemble of parts and not single units that move in time:

Music is the only system of symbolic communication and exchange that un-
folds along the irreversible line of time, it is no mere information, nor is it the
simple flow of sequence of single sound units, but rather a complex ensemble
of parts related to each other not only in the present but as fragments of
processing time and future time (Clozier 2001).

Spatialization should be part of that complex ensemble and not be regarded

as a technological “trick”.

The gmebaphone is an immersive system that creates a sound field surround-

ing the listener. Modern surround systems are designed to be “equal” in all directions

and are used for the reproduction of works where sound trajectories are not relevant

and not suitable for music reproduction.16 On the other hand, “geometric systems” are
15
Later versions of the gmebaphone moved away from the cross-over filtering of signals into a software
based instrument with fader motion capture and scene automation with crossfade. It was renamed
cybernephone in 1997 (Emmerson 2007).
16
In the 1990, the European standards organization ITU (International Telecommunications Union)
conducted research for optimal speaker placement. The recommendations are part of the document

15
considered when space is a structural part of the composition as in, for example, Stock-

hausen’s Oktophonie and Paulo C. Chagas’ Migration, the latter – for twelve channels

– “was the first electroacoustic piece that systematically explored the circular config-

uration of 12 loudspeakers in the WDR studio” (Chagas 2008). Chagas also designed

the spatialization system for Stockhausen’s Oktophonie. The intention to create a more

accurate immersive experience is evident by the increasing number of loudspeakers:

• 4 channels. Kontakte. 1958. (Stockhausen)

• 8 channels. Oktophonie. 1991. (Stockhausen)

• 12 channels. Migration. 1995. (Chagas)

• 12 channels. Projektion. 2000. (Chagas)

Systems such as WFS and ambisonics allow for a more flexible approach to

spatialization without the need to work or compose with the loudspeaker in mind. One

of the reasons behind the creation of the mj library was to detach the loudspeaker from

the sound–object allowing the composer to focus on the composition and the spectral

changes instead.

ITU-R BS.775-1. The research was conducted using classical music material and left the surround
channels for “effects” or “ambience”. The document was created well before the development of new
surround methods which give equal importance to all five main speakers which may not be ideal for
music applications. See: The Recording Academy’s Producers and Engineers Wing. Recommendations
for Surround Sound Production. The National Academy of Recording Arts & Sciences, Inc. 2004.

16
1.3 From Stereo to Wave Field Synthesis

First attempts to reproduce sound were made on a single monophonic channel

with the invention of the phonograph in 1877 by Thomas Edison. The ability to record

sounds – and music – paved the way to the development of musique concrète which

began with recordings of natural sounds. Although, it “engaged the persistent myth

of listening to the sound of the world as a source of music creativity” (Chagas 2014,

107), the first phonographs had already inspired musicians to think of it as a musical

instrument.17

The earliest known use of a stereo system was Clement Ader’s experiment

with distant telephone listening in 1881. He demonstrated his invention at the Paris

Electrical Exhibition where the attendees where able to listen to music broadcast live

via telephone from the Grand Opera at Paris. The system consisted of two microphones

on the stage at the opera and stereo pairs of headphones connected in series at the

exhibition. Using the same principle as the stereoscope, which enables a person, by

means of the superposed visual impressions of the two eyes, to see the stereoscopic

images with their natural relief, Adler adds “relief” to sounds.

Suppose two microphonic transmitters, placed on the stage at T and T. Let

one of them be connected by wires with the receiver R, and the other with
the receiver R, these receivers being applied to the ears of the auditor, and
suppose an actor to stand first at A and then at A. In the former position,
as he is nearer to the transmitter T than to T, his singing will be heard
loudest with the left ear; but when he is at A he is nearer to T than to T,
and the right ear will receive the strongest impression. Thus as he goes from
A to A the definite sensation which the auditor will receive will be that of
a diminution of loudness in the left and an increase of loudness in the right
ear, which is the same as he sensation which we experience when a person
who is speaking walks from our left to our right (Nature, 1881).
17
Philip G. Hubert, Jr. ”What the Phonograph will do for Music and Music-Lovers,”
Scribner’s Monthly 46 (1893:May/Oct.), pp.152-4.”Open Letters” section. Can be accessed at:
http://www.phonozoic.net/n0128.htm.

17
 Aders experiment is a demonstration of stereo sound spatialization as a psy-

choacoustic phenomenon or “sensation” and of course, he was aware of the changes

in volume on each headphone: “diminution of loudness in the left and an increase of

loudness in the right ear”. Today we use the terms constant linear panning or linear

crossfade. Usually we can change the position of a mono-source signal by feeding each

channel with the same signal and adjusting the relative amplitudes of the channels. Al-

though an interesting way to create the illusion of moving sounds from left to right and

the opposite, linear crossfading does not preserve the loudness and creates a void in the

middle of the stereo front.

Figure 1.2: The simplest stereo panorama produces a void in the center

In the United States Harvey Fletcher, a physicist most famous for his contri-

butions to the creation of the equal-loudness contours, invented a “Wall of Sound” or

“Acoustic Curtain” at the Bell Labs in 1932. For him, the ideal stereophonic system

could work with up to 80 microphones and speakers in a hall. This is perhaps the clos-

est idea to the Wave Field Synthesis systems we have today. His research included the

creation of a system capable of recording and reproducing the entire frequency range

and to be able to synchronize sound with motion pictures because silent film was the

prevalent format for movies. Sound was introduced by using Bell Lab systems but there

was no interest from the film industry to incorporate spatialization to movies. Fletcher

18
wanted to add a spatial dimension to movies by using binaural techniques which he

initially called “auditory perspective” and later became known as stereophonic sound

or stereo (Fletcher 1992, 182).

Fletcher demonstrated a binaural system in 1932 at the Bell Labs World Fair in

Chicago using a dummy head with microphones attached to each ear. As someone walk-

ing around the head would speak, the public listening through headphones connected

to the microphones would get the impression of being surrounded by someone talking

around them. Fletcher later contacted different orchestras and conductors to see if they

would be interested in a spatialization system for music. No one seemed interested ex-

cept Leopold Stokowski, at the moment conductor of the Philadelphia Orchestra and

who later collaborated with Disney in several projects including the first stereophonic

film Fantasia.

It was about 1931 when I first met Stokowski and we made tests of stereo-
phonic sound down at the Academy of Music in Philadelphia where the
Philadelphia Orchestra held its concerts. There was a spare room in the
Academy building which was large enough to house the orchestra so that we
could have them play on the stage and listen to it up in this large room.
In this way we tried experiments until we felt we had developed a stereo-
phonic system. Originally, the theory of this system was that it should have
an infinite number of loud speakers at one end and the same number of mi-
crophones at the other end. However, we found that in stage productions,
three microphones, three transmitting lines and three loud speakers were
sufficient. I’ll not go into the details of the development work that was nec-
essary to produce this. There are several printed papers on it. However,
we did make nine loud speakers expecting that we might have to use three
across and three up and down. However, we found that most of the action
was horizontal and consequently, three loud speakers were sufficient . . . so re-
alistic was the effect that to the audience the act seemed to be taking place
on the stage before them. Not only were the sounds of sawing, hammering,
and talking faith- fully reproduced correctly, but the auditory perspective
enabled the listeners to place each sound in its proper position, and to follow
the movements of the actors by their footsteps and voices [...](Fletcher 1992).

Fletcher understood the importance of spatialization for commercial purposes

such as in film, telematic music and theater. He focused on two aspects: fidelity and

19
localization. During the 1980’s, fifty years after Fletcher’s experiments, A.J. Berkhout

and Diemer de Vries from the Delft University of Technology in the Netherlands devel-

oped the first Wave Field Synthesis system that was based on seismic data analysis –

elastic wave field extrapolation – and research in acoustics (Berkhout 1993).The WFS

technique changes the psychoacoustic stereo paradigm to a model where sound sources

are physically created in an acoustic field. The speaker is not the source of the sound but

part of a system that creates the acoustic field. WFS is costly to implement and there

are few places where it can be experienced. As of today, the most important centers

for research in WFS and music are at the Technische Universität Berlin, Germany and

Institute of Sonology in Den Haag, Netherlands.

1.4 Acoustic spatialization

Although spatial music is usually associated with acousmatic or electroacoustic

music of the 20th Century, the practice of placing a sound in space with the purpose

of using the space as a musical parameter was explored by composers during the early

Christian tradition in the form of antiphonal choral music of cori spezzati .18

The interest in spatial writing was gradually lost 19 but by the end of the 18th

Century composers occasionally placed group of musicians away from the orchestra only

to create a “dramatic” effect. Wolfgang A. Mozart (1756–1791) wrote for four identical

chamber ensembles of strings and horns in his Notturno in D Major, K 286 (1776).20 In
18
The principle is also used in large polychoral compositions – for two or more choirs–. The term
cori spezzati or split choirs was used to describe polychoral singing in Venice in the later 16th century.
qPolychoral singing. Encyclopedia Britannica, 2013.
19
Many factors affected the way music was performed including performance practices, performances
in secular and private venues, the invention of the pianoforte, etc.
20
A nocturne is a piece intended for an evening party.

20
 Figure 1.3: The dome of San Marco basilica in Venice

the score, Mozart notated the four entrances of the orchestras using the word “echo” 21

– Erstes Echo, Zweites Echo and Drittes Echo – corresponding to the entrances of the

second, third and fourth orchestras respectively.

In addition to expressively indicate the “echoes” in the score, the music does

evoke the acoustical phenomenon of echo; during the first ten measures of the piece, the

repeated entrances of the opening phrase become shorter, spanning from sixteen eight

notes to six eight notes at the last entrance. There is no polyphonic writing between

the ensembles until the last bars of the third movement – menuetto – before the “Trio”

where the four orchestras, at one point, sound simultaneously. Another interesting

aspect of the piece is that the trio is played by only one of the ensembles. In Mozart’s

Notturno the ensembles always play sequentially following the same order throughout

the piece. Mozart did not specify where the orchestras should be placed with respect to
21
From a scientific point of view, echoes are delays between approximately 30 milliseconds and about
a second in duration that usually do not change the “shape” of melodies or phrases (Roads 2004). For
more on time shifts and delays, see the corresponding chapter in Miller Puckette’s Theory and Techniques
of Electronic Music.

21
Figure 1.4: Beginning of Mozart’s Notturno for 4 orchestras

22
the audience. Many different combinations are possible and, by analyzing the way the

piece was written, some are more relevant than others.

Although the details of the first performance of Notturno in D are not known,

and there is no indications of a layout in the score, four orchestras allow for the following

combinations, including clockwise and counterclockwise rotation:

Figure 1.5: Some of the possible spatialization layouts in Mozart’s Notturno for 4 orchestras

Hector Berlioz’s Requiem (1837) is another example of the use of space in mu-

sic.22 The piece was written for four antiphonal brass groups placed on each corner

of the stage and a massive orchestra of singers, woodwinds, horns, strings and percus-

sion. Composed for Saint-Louis des Invalides in Paris, Berlioz exploited the acoustic

characteristics of its gigantic dome.

The four brass choirs enter at unison at the Tuba mirum section of the mass

playing an Eb chord then splitting into grandiose antiphons, polyphonic passages and

unisons; the same orchestration idea is repeated throughout the work.

Although Berlioz was aware of the use of space in music and its implications,

the composer creates an impact on the listener as the majestic choir of brass first appears

in the Dies Irae – Tuba mirum – section of the mass 23 (Berlioz 1994).
22
Hector Berlioz. French. (1803–1869).
23
“Day of Wrath”. The Dies Irae was used in the Roman liturgy as the sequence for the Requiem
Mass for centuries being the most famous Christian doomsday chant, painting an apocalyptic vision of
the world’s dissolution into ashes.(Slonimsky 1998, 39).

23
Figure 1.6: Berlioz’s Rex Tremendae

24
 Another example is the work of American composer Charles Ives (1874–1954),

who experimented with polyrhythms and polytonality incorporating a spatial element in

many of his works. His Symphony No. 4 often requires two conductors as the composer

divides the gigantic orchestra into smaller groups and incorporates additional off-stage

ensembles. Ives was strongly influenced by the sounds of simultaneous marching bands

in the town of Danbury, Connecticut, where he was born. This influence can be clearly

seen in his innovate way of writing for overlapping keys and meters and the combination

of European, American and church music.

Ives is an early postmodernist; as he juxtaposes multiple elements and exper-

iments with microtonality, polytonality, polyrhythm an aleatoric elements such as in

his work The Unanswered Question, his intention is to separate different layers of inde-

pendent material. The work features three different layers of disparate material that is

tempo and key independent. In addition, these layers are further separated by the spa-

tial placement of the ensembles: the string orchestra off-stage, the woodwind ensemble

on stage, and the solo trumpet off-stage at a distant position.

The string quarter or the string orchestra (con sordini), if possible, should be
“off -stage”, or away from the trumpet and flutes. The trumpet should use
a mute unless playing in very large room, or with a larger string orchestra
(Ives 1908).

Ives uses a spatial layout to create a sense of atmospheric perspective and

distance, this is written in the foreword of The Unanswered Question where he asks the

trumpet to use a mute if too close to the string ensemble or playing in a small room.

The composer continued exploring the use of space with his Symphony No. 4 where he

takes the technique to a an unprecedented level applying the same ideas to a very large

orchestra.

25
 From the point of view of Auditory Scene Analysis, the spatialization in Ives’

music aids the listener in the segregation of audio streams preventing the computing of

dissonances. In other words, dissonant musical lines played by instruments placed in

different parts of the hall are not perceived as one dissonant passage but as different

melodies. There is no intention by Ives to create new timbres. The juxtaposed materials

remain unrelated throughout the work whose layers are to be discretely perceived by

the listener.24

Henry Brant (1913–2008) continued and expanded the experiments initiated

by Ives. A pioneer in the development of spatial music, he was greatly overshadowed

by European contemporary music composers, especially Karlheinz Stockhausen. Most

of Brant’s works are spatial. His first one being Antiphony (1953) where the orchestra

is divided into five groups and each section play in different keys. The result is a level

of complexity created by dissonant relationships. In 1967, Brant summarizes his ideas

on space and music in relation to timbre (Brant 1967):

• Spatial separation changes harmonic content of a texture – if the music contains

several layers within the same range of an octave, by-product unisons sound con-

fusing if they come from the same location, for example: performers that are close

together sharing a stage. If the same layers occur from widely separate locations

in the hall, the unisons are no longer perceived as unisons but as distinct tone

qualities created by diffusion and distance. In other words, same pitches on same
24
To break a complex sound apart our brain uses simultaneous strategies, the first one being segrega-
tion, itself composed of different strategies that attempt to identify individuals objects or events with
a composite stream of information. Another primitive feature is the regulatory of harmonic structure
or harmonicity. We understand the harmonics series as one event corresponding to one timbre and not
as separates sounds. For example, a note from a violin is one identifiable sound and not a series of
simultaneous streams with different pitches. For more information on the topic see the chapter on sound
cognition in Andy Farnell’s book Designing Sound (Farnell 2010).

26
 [h]

Figure 1.7: Distribution plan for Voyage Four by Henry Brant

instruments that are widely separated in a hall will have different spectral content

that is added by the room’s acoustics.

• Distance hinders coordination – simultaneities in the music, e.g. unisons –, are

difficult to coordinate between two spatial separated groups

• Spatial separation allows for expanded complexity – if writing the music within

a close pitch range e.g octave –,creating no collisions or crossing of textures it is

possible to have clarity within a group that is close together. On the other hand, if

groups are widely separated, the writing can be freed of restrictions hence adding

textural complexity.

• Placement of performers cannot be optional – the layout of groups of performers

in the hall must be carefully planned as part of the work in order to create a

controlled musical result. Brant originally wrote his ideas about the use of space

in the process of composition in 1953 publishing them a year after.

27
 For many years composers focused on three elements of music: melody, har-

mony and rhythm. By considering spatialization a structural element of a work, Brant

suggests that composers need to be aware of texture, spectral content and dynamics. All

these elements add complexity to the music and should not be arbitrarily considered.

1.5 Electroacoustic Spatialization

It is during the early 20th century when composers started to think differently

about music and space. With the invention of the telephone and the gramophone, com-

posers – and the general public – began to develop different ideas about time and space

including timbre. For them, the earliest telephone systems were capable of transporting

a distorted but intelligible human voice – timbre – miles away from one space to another.

German composer Karlheinz Stockhausen may have been aware of Brant’s

writings since they predate Stockhausen’s “spatial music manifesto” Musik im Raum

(1959) written after his acousmatic piece Gesang der Jünglinge (1955–56) and Gruppen

(1955–57) for 3 orchestras.

In my Gesang der Jünglinge I attempted to form the direction and movement

of sound in space, and to make them accessible as a new dimension in musical
experience. The work was composed for 5 groups of loudspeakers, which
should be placed around the listeners in the hall. From which side, by
how many loudspeakers at a time, whether rotation to left or right, whether
motionless or moving, how sounds and sound groups are projected into space:
all this is decisive for a comprehension of the work (Stockhausen 1958).

Stockhausen opposes Brant’s ideas about the use of space. In his “manifesto”,

the German composer dismisses Gabrieli’s use of partial techniques. He argues that

Berlioz and Mahler used spatialization on their works only to add drama accusing them

of being “too theatrical”. He argues that direction is the only spatial feature of sound

worthy of compositional attentions because it could be serialized. The composer claims

28
that perception of distance and timbral changes due the acoustics of a hall are param-

eters that cannot be serialized hence cannot be a structural part of a composition. In

contrast, when a sound is placed in a circle around the audience, parameters such as lo-

calization and speed of movement can be operated with exact proportions. It is possible

to create “the scale of localities corresponding to the scales of pitch, duration, timbre

and loudness” (Miller 2009).

Stockhausen’s music and ideas were novel at the time. Today, a sound orbiting

a space in circular motion can be parameterized as a function of time obtaining phase

for a given radius in a two dimensional polar coordinate system, allowing the sound to

be placed at a precise location, for example, using Wave Field Synthesis or High Order

Ambisonics. The perception of distance and timbral changes are a consequence of the

room’s acoustics and cannot be “composed” but highly controlled using artificial reverb

algorithms and software tools.

1.6 Gesang der Jünglinge

Gesang der Jünglinge is Stockhausen’s fourth acousmatic piece, being, Etude

(1952, musique concrète), Studie I (1953, musik elektronische) and Studie II (1954,

musik elektronische) his previous works for tape. His first three pieces are monophonic

works; Studie I and II are completely serialized compositions. In contrast, Gesang der

Jünglinge was written for five channels and incorporates the idea of fusion between

sounds, in this case: the sound of a human voice and electronic generated sounds. This

is the first time a composer applies the idea of morphing acoustic sounds with artificially

generated spectra to create music. According to Paulo C. Chagas, Gesang der Jünglinge:
[...] develops an aesthetics of hybridism that integrates different kinds of
sound material. Inspired by Meyer-Eppler’s research in phonetics, Stock-
hausen explores the living quality of voice to create electronics imitations

29
 Figure 1.8: Excerpt from the manuscript of Gesang der Jünglinge

of phonemes and other elements of spoken speech and language such as for-
mants, articulation and intonation[...] Stockhausen applied serial principles
of composition to the generalization of this material as well to other sound
dimensions and to the organization of the formal levels of the composition
(Chagas 2014, 110).

Gesang der Jünglinge is serialized work that employs the idea of functions of functions –

transposition of the series onto its own elements – developed by Boulez (Decroupet 1998,

100). The text from the prerecorded material is transformed by permutations that ob-

scure its meaning. Although continuing the German tradition of serialism, Stockhausen

gives a historically important step forward from early experimental stages of electronic

music and musique concrète showing a more open minded approach to composition

departing from the aesthetical constrains imposed by the studios in Cologne and Paris.

Gesang is historically important for another reason. It includes five channels

of audio: four speakers surrounding the audience and one hanging from the ceiling at

30
the center of the hall, and it is the first electronic piece to serialize the projection of the

music in space. For its first performance at its 1956 premiere, the fifth loudspeaker was

set up on stage. The first four channels where played by a four-track tape machine. the

fifth by a separate machine. After the performance, Stockhausen mixed the fifth track

onto the first track (Smalley 2000, 11).

Gesang’s spatialization seems to aid to clarify the serial texture not as a dis-

crete parameter of the composition. This aspect of Gesang shows strong evidence that

the composer was well aware of Brant’s ideas with regards to spectral content of textures

in spatial music. Although the logical step would have been to go from experimenting

with monophonic playback to stereo, the idea from working with multiple discrete sound

sources can be attributed to Henry Brant who wrote for five orchestras before Stock-

hausen’s Gruppen (1955). 25

1.6.1 Circular Sound Space

The aesthetics of circular motion, emerged after the creation of a device Stock-

hausen designed himself: the Rotationslautsprecher or Rotationstisch which allowed

sound images to be rotated 360 degrees within a two dimensional plane. The device

consisted of a table capable of rotating a loudspeaker that sit on top. Rotation was first

controlled manually, later motorized. There were four microphones in a square array

capturing the sounds emanating by the rotating speaker. The sounds were recorded into

a newly acquired four-track tape recorder, all sounds to be mixed later with mono tape

recorders (Manning 2006). This approach to the process of composition was not simple

or straightforward requiring different stages of recording and mixing. One important

aspect of the process that is usually not considered by many scholars is the fact that
25
Antiphony I (1953) was written for an orchestra divided into 5 sections situated in different parts of
the hall to create an effect the composer called spatial-polyphony (Brant 1977).

31
the Rotationslautsprecher made noise and generated wind as the table rotated. On top

of air and noise there was the addition of the Doppler effect as the sound from the

loudspeaker move from one speaker to the next (Decroupet 1998).

Although Stockhausen aimed for great precision and control over the serialized

material, the various monophonic tracks could never be synchronized perfectly, the

recorders would work at different speeds, the oscillator would never play back the exact

frequency of a sine tone the composer meticulously calculated (Manning 2006). All these

elements added to the aesthetics of the composition. The result of the parameterization

of electronic sounds in early electronic music could never be as “precise” as the ones

created by today’s digital computers but they incorporate to the aesthetics of the works.

The rotating loudspeaker mechanism was used for Kontakte (1959–1960), for

four-channel tape, piano and percussion. By using this device, Stockhausen created

a new aesthetics of sound spatialization that will be explored and expanded by other

composers such as Pierre Boulez notably in his work Répons. The aesthetics of circular

motion is still today the most popular approach to sound spatialization.

The idea of a moving sounds in space created an aesthetics that cannot be

dissociated with the importance of the materiality on the resultant sound space. The

circular motion was not randomly chosen by Stockhausen. It is the consequence of a

real physical room where microphones could be placed in a quadrant with a rotating

table in the middle. The recorded sounds were to be reproduced in a concert hall of

rectangular shape with the speakers oriented to the audience at the same angles. The

microphones are replaced by speakers, the Rotationstisch by an audience.

For Stockhausen, moving sounds in a circular fashion was not enough. Con-

sidering that sounds are “non-stationary objects”: movement is related to space as it is

related to time. The spatial perception of a sound moving in space is hard to achieve

32
with only two speakers. As the perception of a sound moving left to right on a stereo

setup is a psychoacoustic phenomena, it only works when the listener is at an equidistant

distance from the two speakers. We localize sounds by their variations on their timbres

and the different times of arrival of sound waves to our left and right ears. In order to

make the circular sound space a structural part of a work, sounds need to change as they

move. Stockhausen applied morphing techniques 26 from one sound object to another as

sounds rotate. This technique originates the concept of spectral spatialization. As the

sounds morph while moving in circular motion, they make transitions clear. No matter

where the listener is, sounds belong to a space which in return shapes the spatiality of

the physical space.

1.6.1.1 Répons by Pierre Boulez

Since the beginnings of the microcomputer revolution during the mid 1970s

and through the 1980s, processing of large amount of data became faster and cheaper.

Research institutions such as Ircam in France applied the technology to computer music,

with special emphasis on real-time signal processing. Composer and Ircam’s director

Pierre Boulez, and a group of collaborators designed and developed a machine that was

capable of synthesis, spatialization and transformation of sounds in real time: the 4x.

Boulez used the computer for his composition Répons which he wrote as a showcase for

Ircam. The machine was capable of doing the following real-time transformations:

• The modulation of a solo instrument by a synthetic sound.

• Time shifts systems, mostly delays and echoes.

26
Mixing is not morphing. Mixing is a change in volume – fade-in and fade-out –. One can transition
from one sound to another by fading them accordingly but morphing requires combining two spectra.
There are several techniques, for example one can take the amplitude (from its frequency bins in the
spectral domain) of one sound and apply the gains of another sound.

33
 Figure 1.9: Répons’ sound projection

• The spatialization of sound at different speeds controlled by dynamics (of the

soloists).

The spatialization was implemented using an envelope follower as follows:

• A microphone captures the sound of the performance

• The sound is stored in a buffer that contains an envelope

• The envelope (buffer) is divided into four sections

• Each section is transferred to the loudspeakers using a flip-flop unit. 27

Spatialization in Répons is performed in the time-domain. The portion of the

sound that is first captured by the microphone is the first one to be transferred to a

loudspeaker as in a FIFO system.28 It was not possible with the 4x to perform a FFT,

that is to say, to apply transformations in the spectral-domain.

Boulez wanted to get rid of the “tyranny of the tape” as in Mario Davidovsky’s

Synchronisms where performers need to precisely synchronize with a fixed media leav-

ing no room for personal interpretation and the freedom of internal time as opposed to

metronomic time (Jaroszewicz 2013). Boulez positioned six instrumental soloists and
27
In electronics, a flip-flop unit is a circuit that has two states (0 or 1) and can be used for storing
state information. These circuits are the building blocks of digital computers.
28
FIFO: First In, First Out. An abstraction related to ways of organizing and manipulation of data
relative to time and prioritization.

34
 Figure 1.10: Répons’ spatial switch

six loudspeakers surrounding the audience, leaving a small orchestra at the center of the

performance hall. The “surround” instruments are highly resonant: piano, glockenspiel,

harp, vibraphone and cimbalon. The work was inspired by medieval responsorial chants

(Répons) and the spatialization is used to provide multiple “responses” to a soloist’s

question. After the opening movement at the entrance of the electronics and the intro-

duction of spatialization the 4x captures a series of arpeggios and creates a new sound

field by opening a circular space.

The attention of the audience is thereby suddenly turned away from the
center of the hall to the perimeter, where the soloists and the speakers are.
The audience hears the soloist’s sounds traveling around the hall without
being able to distinguish the paths followed by the individual sounds. The
overall effect highlights the antiphonal relation between the central group
and the soloists by making the audience aware of the spatial dimensions
that separate the ensemble from the soloists and that separate the individual
soloists as well (Boulez 1988).

In this case, the trajectories are not intended to be perceived as such. They are a timbral

extension of the soloists. Velocities and volume envelopes are map to speed of rotation

with the implementation of a switching mechanism. An arpeggio is stored and divided

35
into four pieces which are later spatialized by the matrix. Although spatialization in

Répons seems to be executed in the time-domain, the four different “dissected” sections

of each arpeggio correspond to different parts of the spectrum as sounds change their

spectral content over time. A different approach – and technology – than Stockhausen,

in Répons, there is no morphing or cross-synthesis between sounds.

1.7 The Sound Source

In his article “Composition in Circular Sound Space”, Paulo C. Chagas argues

that the materiality of the studio where music is created influences the aesthetics of

sound spatialization. According to Chagas, space as an embodiment of composition

has a special role in the history of electroacoustic music – elektronische Musik – and

how it relates to serial music. Most importantly, Chagas describes how sound can

acquire a tangible property thus becoming material:“When space becomes a parameter

of composition, sound develops a “tactile” dimension. Similar to a body, it occupies a

unique position in the space from which it can exclude other spaces.”

At EARS, I had the opportunity to experiment with systems for spatialization

using different layouts and software including Ircam’s Spat, ICST’s Ambisonics Tools

and the HOA Library. I worked with different speaker layouts and different sound

sources – ranging from acoustic instruments to computer generated noise – in order

to test sounds that had contrasting spectra and repetition rate. The results of the

research were applied to a series of spatialization etudes. The apparatus at EARS and

the physical space influenced the compositions of the etudes resulting in an aesthetic

imposed by the materials. Moreover, upon listening and working with different Wave

Field Synthesis systems in which sound sources become “physical”, a different way of

36
thinking was needed in order to understand what it meant to compose with sounds that

occupy a real space and exclude other spaces.

The choice of sound sources – synthesized or sampled – and their transfor-

mations in space change the way sounds are perceived: they occupy a space therefore

acquire a “tactile” dimension. In order to understand this transformation, the topic

necessitates a different thinking style, one in which sound is not seen as what it appears

– perceived – but as what it really is. Thus, a real world perspective is required that

defines sound objects as things or “concrete” objects.

A thing has the quality of being something that occupies a space in time. In

other words, a space-temporal concrete object that in global 21st century culture we

associate with material things, things we can observe and are perceptible by touch,

which for most people, is the ultimate proof of existence. We can perceive a flower by

its fragrance but it becomes a flower when we are able to see it and define its shape. We

reaffirm its quality as object by using a combination of senses; a rock becomes a rock

when we are able to see it, touch it,29 move it, measure its weight and feel its shape.

A rock and a flower have different space-temporal qualities, as sound does. A

rock occupies a space and might not change its shape in a human perceivable time lapse.

We know it is changing by looking at the erosion produced by the elements. For us, the

rock is the same rock. On the other hand, a flower grows from a seed and decays in a

time span that is shorter than our own. Rocks and flowers are both concrete objects and

can be physically displaced while retaining its identity. Men, like the flower, perceivable

change over time but we still retain our identity even if those transformations are not

part of the natural aging process. This identity is conscious and experiential. Sounds
29
It is possible to physically touch sounds as demonstrated by Miha Ciglar’s tactile con-
troller which uses 97 ultrasonic transducers to focus acoustic radiation. “The result is a rel-
atively strong and spatio-temporally precise tactile reproduction of the projected audio signal”.
http://www.ciglar.mur.at/sonicbeams.html. Last accessed 01/20/2015.

37
morph as they travel in space following a trajectory and their identities become part of

the musical form.

For the physicalist and the materialist, the world is comprised of small particles

or atoms that create things that posses unity. Iannis Xenakis (1922-2001) – composer

and architect – applied these ideas as a metaphor for composing with microsounds. In

his book Formalized Music: Thought and Mathematics in Composition, Xenakis uses

the concepts of “granular sounds” and “sound-points” defined by an audible area called

“screen”. This screen defines the boundaries of the thing, in this case, a sound-thing or

a sound object (Xenakis 1992, 51).

A “thing”, to be considered as the same thing as it moves in space and time,

should preserve those boundaries that define it. For example, given that we cannot per-

ceive ultrasounds with our human ears,30 a sound object is limited by its frequency com-

ponents but the combinations of those components are what defines and gives a sound

its identity, which we call timbre. For example, the sound of a locomotive changes over

time depending on parameters such as the speed of the train, but it preservers its iden-

tity; the frequency components undergo small variations but preserve their relationship.

Then timbre is analogous to shape and the contour of the shape is the screen. This idea

corresponds with our modern conception of matter. We think every object is composed

of smaller particles, from elements to atoms to point-like particles. These conceptions

thus remain local belonging to some spatial scale. When does a sound object begin and

when does it end when we only look at its spatial dimension? If the frequency compo-

nents keep their relationship over time then the object preserves its unity but then what

is the space the sound occupies or is there an occupied space?

30
Sounds above 20,000 Hz, varying from person to person.

38
 It is difficult to define boundaries for some objects and especially human created

things. Considering a sound–object that preservers its unity as time advances, it may

occupy a space in our consciousness and does occupy a physical space as sound waves

traverse an elastic medium such as the air. A perturbation of the pressure of the medium

can be measured with a sound level meter and a sound field can be established. The

boundaries of the sound field cannot be established by the listener and the sound object

is in reality a hyperobject. It goes beyond the limits of human perception. An attribute

of a hyperobject is non–locality; these objects are massively distributed in time and

space.

Examples of hyperobjects are icebergs, the universe, radiation and music (Mor-

ton 2013). Imagine a helicopter flying in a city like N.Y. with a dense distribution of

skyscrapers. For the listener on the street, the sound from the helicopter only means “he-

licopter”. It does not carry other information as the reverberations makes it impossible

for the human ear to know where the sound is coming from, and for the listener to point

to its source. Then, another attribute of a sound object is to be able to be localized.

Then the sound–hyperobject becomes a sound–object establishing its spatio–temporal

unity.

For everyday material objects, space unity takes precedence over temporal

unity (Garcia 2014, 35); an object that can be moved retaining its shape has more

permanence than and object that changes as it moves, thus creating a hierarchy. Spatial

consistency is an attribute of a thing. In music, a motif, has spatial consistency as it

reappears unchanged or slightly varied as the music progresses. This idea was explored

by Wagner who used shorts musical ideas – the leitmotiv – associated with physical

objects or characters in his cycle Der Ring des Nibelungen. In Wagner’s music, these

objects do not undergo a temporal transformation by themselves or are affected by

39
external forces. Leitmotivs are sound objects that are associate to a greater object and

from a different perspective, sound objects in Wagner’s music acquire another dimension

as they carry a musical idea attached to them. For example, as with the fragrance of a

flower signifying the flower, the “ring” leitmotiv signifies the ring.

If an object changes and loses its unity as it is being spatially displaced, it be-

comes “unthingly”, bounded with the environment and of an inferior hierarchy. Spatial

consistency is an attribute of objects that allow us to re-identify them as time progresses.

In music, this consistency creates form and coherence. If a sound–object lacks the at-

tribute of localization remaining a hyperobject, there is an impossibility to reduce it to

a spatio–temporal object. But as material things change and our perception of them

also changes over time, their spatial limits are different at different scales, for example, a

sound particle or microsund can be part of a stream of sound particles, the latter having

its own timbral qualities. As things are constantly changing at different temporal scales

an insect might live one day and a rock could take millions of years to dissolve. Accord-

ing to Tristan Garcia in his book Form and Object, for the formalist or metaphysical

abstract things are objects that can exists regardless their spatio-temporal attributes:

abstract things exist outside those boundaries. “Many things exist, and we cannot do

without the concept of “thing”. In the absence of things, the world becomes undiffer-

entiated: the world is a self-saturated whole in itself which knows no differentiation”

(Garcia 2014, 37).

In acousmatic music, the sound world of a composition could be a mass of

non-localized sound-objects. For example, in an ambience or soundscape composition,

where does the sound object begin and where does it end?

According to Pierre Shaeffer, the sound object is not the instrument that has

been played (Schaeffer 2003, 58). For example, if we listen to a recording of a band

40
whose instruments are unknown to us, what is what we listen to? For him, that is

the sound object. The sound object is neither the instrument nor the tape which is a

support for the sound. This tape can contain many different sound objects and when

the composer manipulates sound by splicing and pasting its pieces, he is not modifying

the sound but creating new sound objects. For Schaeffer, the sound object is not an

emotional state because the object does not change from one listener to another or

between our different moods or attention spans. The object does exceed our individual

experiences: the visual, auditory, tactile impressions and the way we interact with it

through perception, memory, imagination, etc.

Shaeffer defines the object as an independent thing that is perceived from a

level that is deeper than the acousmatic reduction (Schaeffer 2003, 50). There is no

need to interact or find its significance. When we listen to a speech we focus on the

concepts that are transmitted by the speaker, the sound is a significant. On the other

hand, the sound objects appears when we do this different level of listening that is

more rigorous which Schaeffer calls l’écoute réduite. It is the sound itself that takes our

attention without any meaning attached to; we disregard its significance. A deeper level

of listening is achieved by detaching the information about the sign. If we listen to the

sound of a helicopter in a recording, it will be hard not to think about a helicopter at the

first listening, which, according to Schaeffer, is listening by reference. At the moment

when we do not think of the helicopter anymore and instead our attention focuses on

the sound object, we are doing a deeper level of listening and the sound object reveals

itself.

With Schaeffer’s ideas, it is very difficult to see how a sound object can become

a thing or how the definition of object applies to an acoustic sound. For us, it is missing

some elemental features in our definition of things. As mentioned above, an attribute

41
of a hyperobject is that it is non-local and it is massively distributed in time and space.

Music is a hyperobject, but in a micro scale we could consider Pierre Schaeffer’s sound–

object to be a hyperobject as it is a non-localized object. We could also put in this

category Ambient music and ambisonics fields for background support. These non-

localized sounds occupy the whole space and they are meant to have no clear beginning

or end, they usually give the listener a background layer for other sounds. Brian Eno,

describes immersion being the point of Ambient music; “music to swim in, to float in,

to get lost inside” (Eno 2004, 95).

Soundscape composition is different “as it tends to involve material offering a

connection with a listener’s life” (Landy 2007, 45). In order to find the sound object in

a soundscape it is necessary to apply a deeper level of listening; but l’écoute réduite is

not what the composer of a soundscape wants as he intends to communicate something

and tries to offer and interpretation of the material that was transformed. On the one

hand, Ambience music offers the listener a space to explore without having to search

for meanings, on the other hand, spatialization, gives the listener the opportunity to

attach to, follow and hear the transformations of the sound object as it morphs into

something else, all in a clearly defined sound space by composed or algorithmically

generated trajectories.

1.8 The Aesthetics of Circular Motion

We shall focus now on one of the aspects that define the sound object: the

property of localization. According to Chagas, the aesthetics of circular motion works

very well for the ears. Extended experimentation with circular panning techniques at

EARS and at the Wave Field Synthesis system in TU Berlin resulted in excellent per-

42
ception of circular motion regardless of the system (Chagas 2008). This perception is

enhanced by the the visual layout of the speakers, which are usually positioned sur-

rounding the listener. Complex trajectories are more difficult to follow with the WFS

system rendering the best results.

But an object that moves elliptically – loops – around the listener creates more

than movement, it becomes part of the structure of the composition by adding to its

form. The use of concentric circles in a WFS plane can create form as objects move

towards the listener at different time intervals. If sound objects preserve their amplitudes

throughout their orbits, then a special event occurs when it reaches the closest point to

the listener. This varies of course, depending if the listener is static or moving inside the

space. For static listeners, there is something that comes back musically. For example,

if an orbit takes one minute to complete, and the object completes an orbit during the

piece, this could be perceive as A (object’s position closest to listener X) and A again

when object arrives at the starting location. As a part of a piece, this could be perceived

as A - B - A if at some point the object disappears from the listener’s ears. We have

a similar experience when listening a theme that returns during the recapitulation of

a sonata. The sound object becomes a building element of the musical form like a

recurring motive.

Another approach is to use orbits that reach listeners at different time intervals.

This creates a new challenge for the composer as he needs to think about form from a

different perspective. In some cases, the composer could be dealing with as multiple –

and complex – forms as there are listeners in the space, even if he wrote the piece with

a particular form in mind.

This also raises the question about form and the perception of the work by

different listeners at different points in the listening space. A moving audience is very

43
 Figure 1.11: “Wall” of Speakers at the Zuiderstrandtheater in Den Haag

likely when concerts are held in large warehouses or works are part of sound installations.

To give an example, the Zuiderstrandtheater in Den Haag held in 2014 a series of concerts

where sound was projected from a gigantic array of speakers and the audience was free

to move, walk or lay around the listening space.

How many objects can be continuously followed? It varies. It is possible to

follow the orbital paths of four 31 objects sounding at the same time. I experimented with

sounds including repetitive patterns (drum loops, clicks, noise, short samples) and those

provided no problems for the listener but is it possible to follow the location of many

objects that transform in time? Usually when there is a great timbral transformation, for

example by means of cross-synthesis, where sound A gradually acquires some properties

of sound B eventually morphing into B, and the paths are not elliptical, there is chance

to lose the connection with the sound. But the question is difficult to answer because

if the listener is not static then, what does it mean to follow a sound in space? With a
31
Stockhausen talks about spatial depth and the ability to hear up to 6 layers in Four Criteria of
Electronic Music. Lecture IV, Kontakte given at the Oxford Union on May 6Th 1972.

44
WFS system it is possible to physically “follow” the sound by walking around the space

especially if the source is created as a plane wave instead of a point source .32

Extensive experimentation and listening with a spatialization system is re-

quired in order to find out what works and what does not for the composer. As spa-

tialization becomes a structural part of a musical work, aesthetic choices are left to the

composer who can “preview” his work using a binaural tool. Nonetheless, one must

spend hours listening and fine tuning paths, dynamics, velocities, etc. The process itself

is like any other aspect of a composition. I had the opportunity to listen to many works

written for the Game of Life system and found unique approaches when working with

moving objects. Some composers see spatialization as a way to create interesting paths

that are more visually and conceptually appealing than their musical outcomes. But

a great example of a very effective use of circular motion is the piece ”Iron Age” by

Robert Henke. The composer uses static concentric circles distributed around the field

creating a form that has different timing for listeners located at different points inside

the field.

Complex patterns with irregular shapes are difficult to follow, they depend on

the speed of movement and the timbral characteristics of the objects. The interpolation

of continuous movement between two points may not be perceived by the listener but

this is hard to achieve even with sounds in the real world.

Using circular or elliptical orbits and straight paths with a WFS system is an

excellent combination to provide the listener with a sense of motion and form, as the

WFS produces a good sense of depth; sounds can be positioned close to the listener’s

head and can gradually disappear moving away from the listener.

32
See the Wave Field Synthesis. Section 3.1

45
Figure 1.12: Elliptical and straight paths in a WFS system

46
 Part II

Technical Aspects

47
Chapter 2

Sound Fields

2.1 Stereo and Panning

A panorama is a wide angle representation to conform to a flat or curved

background, which surrounds or is unrolled before the viewer.1 In audio, panning is the

technique to place a monaural –one channel – sound in the stereo sound field between

the left and right speakers to create a sense of space. For example, a performer can be

virtually positioned in a semicircle (0°–180°) in front of the listener.

Figure 2.1: Theoretical width of an audio image.

1
Encyclopædia Britannica Online, s.v. “panorama”, accessed January 21, 2015.
http://www.britannica.com/EBchecked/topic/441452/panorama.

48
 The field of an audio panorama is called the image. Although it is delimited

by a semicircle of 180°, a width of 90°or 60°is used (Farnell 2010, 221).

There are several techniques for panning including the following:

• Simple Linear Panner

• Square Root Panner

• Cosine Panner

2.1.0.2 Simple Linear Panner

The simplest form of panning by equally decreasing and increasing the power

of both left and right signals. A control signal (x: 0.0, 1.0) is used to calculate (1 − x).

Figure 2.2: Simple Linear Panner in PD with Low Pass Filters for smoothing the signals.

As shown in Figure 2.2, the signal loses power at the center. The square root

law panner and the sine-cosine law panner compensate for that loss. See Figure 2.3. The

sine-cosine panner is smoother at the edges as it approaches them at 45°. The latter

is also good for creating the impression of circular motion in front of the listener, for

example, it can be used for placing the musicians of a large orchestra like in a concert

hall. For small ensembles –rarely panned hard left or right – a square root panning

49
 Figure 2.3: Square root panner (left) and Sine-Cosine panner (right)

technique is more desirable as it does have better response at the center (Farnell 2010,

222).

In a stereophonic setting, the output of the speakers is adjusted to a point

towards the center of the listening space – a concert hall, studio, etc –. This point

between the speakers is known as the sweet spot. Panning, due to several psychoacoustic

mechanisms, gives the listener the impression that sound moves. This is perhaps the

oldest and most common technique that has been implemented for the reproduction of

music. Surround sound such as 5.1 is based on the same technique and also requires

– like stereo – that the speakers are laid out using always the same setup. This is

a channel based technique because the mixing is done working on each channel. For

example, in Apple’s DAW Logic Pro X one can automate how sound moves from left to

right individually on each track (Figure 2.4).

Although it seems fairly easy to use the pencil tool in a DAW to add movement

to sound, it only works with stereo or surround panning. For complex spatialization

techniques it is not possible to use a commercial DAW such as Logic or ProTools,

instead, it is necessary a dedicated programming language for audio such as Pure Data

or Supercollider, especially when working in real time. Alternatively, for the creation of

trajectories, 3D modeling software can be used in combination with Python scripting.

50
 Figure 2.4: Logic Pro X. Panning using automation. From right to left to center in 2 seconds

The example below was written in Supercollider where a grain of sound is

panned using a two channel equal power panner class.

 
1 var env=Env.new([-1, 1, 0], [1, 1]);

2 SynthDef("pan2",{ Out.ar(0, Pan2.ar(Dust.ar(200),

3 EnvGen.kr(env,doneAction: 2),0.3)) }).play;

 

Another and better approach to panning is to use the Equal Distance Crossfade

technique, where the curve of level differences needed for a specific angle, is approximated

by the Blumlein Law:

gainL − gainR
sin φ = sin ω (2.1)
gainL + gainR

where gainL and gainR are the gains for each channel, φ is the angle of the

virtual sound source and ω is the angle formed by the loudspeaker. Blumlein Law works

only for frequencies lower than 600Hz and a listener’s head pointing directly forward.

To correct this we can apply the tangent law 2 :

gainL − gainR
tan φ = tan ω (2.2)
gainL + gainR
2
These equations can be easily implemented using C-like expressions in Max/MSP or Pure Data.

51
 Figure 2.5: Stereo panning in Max/MSP

and compensate for when the source is in central position φ= 45◦

√
2
gainL = cos(φ) + sin(φ) (2.3)
2
√
2
gainR = cos(φ) − sin(φ)) (2.4)
2

Blumlein Law was named after Alan Blumlein, an English electronics engi-

neer who in 1931 developed what he called “binaural sound” when working at EMI in

England. His technique used a two channel stereo system for recording and playback

as opposed to Ader’s stereophonic technique for broadcasting music. It is important to

note that recording techniques begun in the middle 1800s with the invention of recording

devices such as the phonoautograph, the paleophone and the later phonograph which

was capable of both recording and reproduction of sounds.3

Stereophonic sound was a practical approach to spatialization limited by the

number of speakers necessary to reproduce a wave front that could position a sound

source moving from left to right on a theater’s stage or a cinema screen. Stereo as the
3
The phonoautograph, the paleophone and the phonograph were invented by Edouard-Leon Scott de
Martinville in 1857, Charles Chros in 1877 and Thomas Edison in 1857 respectively.

52
 Figure 2.6: XY Stereo microphone placement technique

standard format for the music industry today serves the purpose of virtually reproducing

an image – a phantom image – of the location of instruments in a concert hall or studio.

Most recordings of live events are done using two microphones. Common microphone

placements techniques are x-y, ORTF, NOS, Stereosonic, MS and parallel spaced. Many

of these techniques only vary in their angles of maximum response.

2.1.0.3 Surround

Surround refers to an array of speakers surrounding the listener. Five-point-

one (5.1) and Seven-point-one (7.1) are the most common surround formats adopted by

the movie industry today. In surround, the .1 or 0.1 is the Low-Frequency channel or

Low-Frequency Enhancement (LFE) that is connected to a subwoofer speaker.

Other formats such as quad – four speakers in a square – never became com-

mercially popular due to a reluctancy from the public to add more speakers to their

home-entertainment systems. In addition, there were different competing formats for

LPs and tapes between the late 1960s and through the 1970s (Holman 2008, 7). Today,

the general public has access to inexpensive surround systems for the home which in-

clude a surround decoder or receiver, woofers and a subwoofer. However, quadraphonic

53
systems – sometimes called 4.0 surround – have been adopted by composers willing to

add the element of spatialization to electroacoustic music. It works for placement of

sounds on each channel as discrete direct radiators at 90◦ . Pans from the front to the

surround speakers do not work. As an aesthetic choice it might be desirable to work

with quadraphonic sound when the location of a loudspeaker wants to be perceived as

the source of the sound.

The 0.1 is widely used in music to accurately reproduced the low end of an

acousmatic piece and it should be used for frequencies lower than 120Hz. In addition

to aesthetic reasons, a dedicated subwoofer decreases the intermodulation distortion of

the main loudspeakers when handling large amounts of low bass 4 (Holman 2008, 59).

2.1.0.4 Ambisonics

Ambisonics – 1st order or B-format and Higher Order Ambisonics – is a tech-

nique developed by Michael Gerzon (1945-1996) in the 1970’s as an alternative to quadra-

phonic systems which he considered to be incapable of good results:

The fault lies partly in studio equipment incapable of giving good quadra-
phonic results, and partly in erroneously conceived “quadraphonic” systems
which leave the apparent localization of sounds at the mercy of the listener’s
imagination (Gerzon 1974).

Ambisonics can spatialize sound fields using 2D and 3D multi-speaker systems.

It decomposes the sound field at a point using spherical harmonics up to a certain order

(1st-order ambisonics, 2nd order ambisonics, etc). In most implementations it only

accounts for plane waves, that is, only the direction of the source is considered.
4
Intermodulation distortion is a type of non-linear distortion that comes with a non-linear relationship
between the input signal and the resulting sound. For example, a common distortion is created when a
smaller sound is masked by a larger sound. According to Floyd Toole in “Sound Reproduction” this is
not a problem with loudspeakers unless they are put into a limiting condition.

54
 Figure 2.7

Ambisonics can be used for recording and playing back multichannel audio

in a 2-dimensional or planar space (pantophonic system) or a 3 dimensional space or

“full sphere” (periophonic system) (Figure 2.7). This system encodes the signal in three

channels for pantophonic systems and requires an extra “height” channel for periophonic

ambisonics encoding. The main purpose of using Ambisonics is the creation of a surround

sound field. It is possible to render virtual acoustic spaces with sound sources that can

be positioned and moved around a space in real-time.

The main advantage of using ambisonics for musical composition is isotropy.

5 Ambisonics is isotropic meaning that sounds from all directions are treated equally.

Another advantage of using Ambisonics is that the position of the speakers can vary from

square to circles to regular polygons. All speakers are used to localize sound creating a

sound field that is stable regardless of the listener position. In order to understand how

panning works is best to understand how 2D panning works first. We saw that panning

is a technique for positioning a monophonic sound within a stereophonic image. Panning

uses differences in gain that are fed into the loudspeakers nearest to the virtual sound
5
Uniform in all orientations; it is derived from the Greek isos (”equal”) and tropos (”way”). For
example, a bell is an isotropic radiator of sound. The sun is an isotropic point source of light.

55
source. In a stereo system, using linear panning or the square of the cosine function,

the gains or the squares of the gains add to 1.

p
p
gainLp + gainRp = 1 (2.5)

or

gainLp + gainRp = 1 (2.6)

In Ambisonics, all channels add to 1 at the same time thus creating a sound

field instead. The panning functions are defined for 3 dimensions. In addition to the

recording of the sound source, a spherical array of microphones will also record room

information that can be reproduced (decoded) independently of the speaker setup.

The decoding of Ambisonics can vary from different degrees of accuracy de-

pending on the desired directionality for each speaker that is used to create the virtual

sound field. This directional accuracy is given by the so-called order of Ambisonics. The

zeroth order corresponds to a mono signal of equal loudness for all speakers. An order

of 7 with an octophonic speaker layout gives a good amount of directionality while a

higher order localizes the sound at one speaker as shown in Figure 2.9.

Ambisonics is not speaker-dependent or listener-dependent. Although there

is a sweet-spot and better images are created at the center of the field, an accurate

representation depends on the order. As shown in Figures 2.10a, the ambisonics order

defines the accuracy. As ambisonics does not provide information about distance, when

working with synthesized sources, it is possible to simulate distance encoding with the

use of artificial reverb and filters. In addition it is also possible to simulate sound

sources with other shapes other than points (Pérez-López 2014). Ambisonics – 1st order

56
 Figure 2.8

Figure 2.9: Directivity of a point source encoding with different Ambisonic Levels (Pérez-López

2014).

(a) Order zero (b) Order seven (c) Order 35

Figure 2.10: Different ambisonics orders and their directionality in Max. The patch was devel-

oped at EARS for the works Laberintos.

57
– is a good system for composing a sound field that surrounds the listener without

moving sound objects. For example, it is possible to create a sound field with only four

speakers, but for moving sources, higher orders are needed – and more speakers – as

the order defines the level of localization accuracy. For example, one of the advantages

of using Ambisonics is the ability to record using an array of microphones to capture a

soundscape:

Recorded 1st . order sources, although of a lower spatial resolution, capture

“reality information” from the original source, including proximity, image
size, the real balance between events in the original scene and a unified 3D
image of the fore- mid- and background – or in other words the sense of
environment and perspective (Barrett 2010).

Ambisonics is a technique that requires two stages: encoding and decoding. A

mono signal enters the system and it is encoded into a number of different channels –

depending on the HOA format – and is sent to the a decoding algorithm of the same

order to be distributed to N speakers. During the encoding process the sound waves

are projected into the spherical harmonics using four channels with omnidirectional w,

x, y and z -directional information. The decoding process reconstruct the spatial scene

for a minimum number of speakers in a circular array specified by the formula:

2 ∗ (order + 1) (2.7)

Mathematically, the directional encoding of a sound source S on three orthog-

onal directions x,y,z of the unit vector u - or rather an incident plane wave u carrying

a signal S results in the following equations:

58
 


W =S







√ √



X = 2~u.~xS = 2 cos θ cos δS


(2.8)

 √ √
Y = 2~u.~xS = 2 sin θ cos δS







√ √



Z = 2~u.~xS = 2 sin δS



where the vector ~u is described by the spherical harmonic (θ, δ) (Daniel 2001).

2.1.0.5 Wave Field Synthesis

Wave field synthesis is also a spatial sound field representation with the purpose

of the reproduction of auditory scenes.6

WFS is among the newest techniques for sound spatialization and requires large

number of speakers to create a virtual auditory scene. The system completely overcomes

the limitations of stereophonic and ambisonics regarding the listener’s position in the

sound field such as the sweet-spot. Rather, there is a sweet-area delimited by the layout

of the speakers. Given the number of speakers required to implement the technique,

attempts have been made using a linear distribution thus creating a planar listening

area as shown in Figure 2.11.

WFS is based on the principle of Huygens which states that when there is

a wave front, it is possible to synthesize the next wave with an infinite number of

small sources or spherical waves. Sound sources can be located outside or inside the

front, though, the sound source should always remain before the listener. In other

word, the listener is positioned inside an enclosing array of speakers (Figure 2.11). In
6
Auditory scenes refer to all audible sounds belonging to a sound field from the perspective of a
listener. Auditory scene analysis aims to study how the human brain reconstructs sound. There are
practical applications in the field of artificial intelligence such as machine listening

59
 Figure 2.11: Speaker array, point source and wave generated by the system.

WFS, loudspeakers are used as secondary sources. This way, virtual sources can be

synthesized in the listening area by making use of the Huygens principle.

Figure 2.12: Wave Field Synthesis. Huygens’ Principle (Snoei 2014)

From a mathematical 7 point of view this principle states that a wave field at

time t+δt can be synthesized by replacing the wave front at time t by an infinite number
7
For an extensive discussion behind the physics and mathematics of WFS see (Sascha 2008).

60
 Figure 2.13: Kirchhoff-Helmutz. Illustration of the geometry (Sascha 2008).

Figure 2.14: Wave Field 2D plane and listener’s perspective

of secondary sources at an infinite small distance δx from each other; this is basically

the Kirchhoff Integral (Van Dorp Schuitman 2005) which states that it is possible to

reconstruct the amplitude of a source at a given point inside of an enclosing surface:

e−jkr e−jkr ∂P
Z    
1 ∂
PA (r, ω) = P − dS (2.9)
4π S ∂ n̂ r r ∂ n̂

A WFS system creates a point source at the desired location. The point source

is where the wave will start to propagate in all directions like an inflating balloon (Snoei

2014).

but WFS is not a 3 dimensional system thus it only uses the plane that inter-

sects with the wave field as shown in Figures 2.14. This simplifies the wave field to a 2

61
 Figure 2.15: WFS Spherical Harmonics

dimensional system of coordinates: (x,y) and from the listener’s point of view the first

waves to arrive provide localization cues. 8

As shown in Figure 2.12 the array of speakers recreates the waves originating

in a virtual source as if they were passing through a “glass” 9 that transforms them in

“real” waves, each speaker reproducing a portion of the source. Thus, for the system to

work, all speakers must be working together and must be synchronized. Given that there

is no sweet spot, all listeners hear the source radiating from the exact same location – the

source has an absolute position – and a physical wave is generated .10 The acoustics

of the room responds accordingly and does not play an important role in localization.

The integral in 2.9 tells us that we need an infinitesimal small distance between

the speakers 11 but this is impossible in practice and one of the physical limitations of

the system, as loudspeakers are used to recreate a wavefront. This approximation of

sound sources by loudspeakers results in a spatially discrete source distribution with

8
Localization cues are ITD (Interaural Time Difference, IID (Interaural Intensity Difference) and
Filtering provided by the shape of the head.) ITD is the difference in time sound arrives at the left and
right ears and IID is the measure of the amplitudes arriving at each ear. Combined they are used in a
model known as the head transfer function.
9
The Huygens’ principle was first applied to optics.
10
As opposed to ambisonics or stereophonic systems that produce a psychoacoustic effect.
11
r is the distance to the source.

62
speakers located next to each other as close as possible. The resulting discretization

effects may be described in terms of spatial sampling . For the human range of hearing

(about 20,000 Hz) loudspeakers have to be placed at a distance less than 1 centimeter

apart, something not possible considering the size of available speakers and the quantity

needed for the construction of the array (Rabenstein 2006). The system will distort

when representing frequencies above the spatial aliasing frequency as frequencies start

to overlap. Distorted 12 frequencies will not contribute to spatial cues and the listener

perceives them as coloration (Van Dorp Schuitman 2005). Speakers that are approx-

imately 17 centimeters apart will have an spatial aliasing at 1000 Hz. so to in order

to avoid distortions in the human hearing range, secondary sources should be less than

one 13 centimeter apart. Fortunately, higher frequencies do not contribute with spatial-

ization cues as much as frequencies in the range 800-1600 Hz where the brain uses a

combination of cues from both ITD and IID.14

For the system to calculate the wave field, each speaker has to process an

algorithm that considers the distance from the speaker to the sound source in meters

d, the speed of sound (344m/s) c, delay d/c and the scaling amplitude factor in a 2

dimensional system of coordinates.

Basically, calculations are based on the same theory used to calculate an ITD

panner 15 but from the perspective of a speaker array in space with a system of coor-

dinates that includes a virtual and a real space. Knowing the distance between each

speaker, the attenuation – loss of intensity due to the air – and the position of the source

accordingly to a system of coordinates we can calculate the distances using trigonometry:

12
In this context, distortion refers to spatial distortion and the ability to properly localize sound
13
8.5mm
14
ITD is most effective at frequencies below 700 Hz because of the wavelength, stopping at around
1500Hz where IID is used instead (Farnell 2010).
15
A simple panner with the addition of a delay line.

63
 q
dL (r, d, α) = (r · sin(α))2 + (r · cos(α) − d2 )2
q
dR (r, d, α) = (r · sin(α))2 + (r · cos(α) + d2 )2

q q
d2 d2
delta(r, d, α) = dL (r, d, α)−dR (r, d, α) = r2 − d · r · cos(α) + 4 −r2 + d · r · cos(α) + 4

Figure 2.16 below shows a system of coordinates with the origin at the center

of the leftmost speaker on the x axis. The y axis is oriented as the depth and the unit is

1 meter. A sound source has positive coordinates and the listener negative coordinates.

Figure 2.16

We can express speakers’ positions as a function of their indexes, going from 0

to n:

Si (d) = (d · i, 0) i = 0..n (2.10)

If the speaker line has a length Sn and (x, y) are the speaker coordinates then:

p
Di (d, x, y) = (x − d · i)2 + y 2 (2.11)

64
 Sound pressure p is a force per unit area, in N/m2 and the peak sound pressure

1
of a wave is inversely proportional to the distance decreasing r for a distance r 16 so we

1
have to use a level scaling factor Li proportional to Di for each speaker (Bole 2008):

1
Li (d, x, y) = (2.12)
Di (d, x, y)

and delay:

Di (d, x, y)
Deli (d, x, y) = (2.13)
343m/s

An implementation only taking into account delay times and amplitudes for

each speaker is shown by using the Faust programming language.17 The hypothetical

scenario has 8 speakers that are 10 centimeters apart – for simplification – with the

sound source located at (0,5) – far left – as shown in Figure 2.16 s:

 
1 import("math.lib");

2 import("music.lib");

3 import("filter.lib");

5 d = 1; // speakers at 1 meter apart

6 x = 0; // in meters

7 y = 5; // in meters

8 nSpeakers = 8;

10 Quad(x) = x * x ; // x squared

11 D(d,i,x,y) = Quad(x - (i - 1) * d) + Quad(y) : sqrt ; // distance calculation for each speaker

12

13 // Amplitudes assignments:

16
Sound pressure is an absolute measurement and applies to a point in space where the measurement
is made without taking into consideration the direction of the wave. Sound Pressure Level (SPL) is a
ration given in decibels (Farnell 2010).
17
Faust is a functional programming language that offers high-performance signal processing at the
sample level thus suitable for low level DSP operations. http://faust.grame.fr/. It can be used for
building VST plugins, PD externals, Max and others.

65
14 Amp(d,i,x,y,sig) = sig / D(d,i,x,y) ;

15 OutA(d,1,x,y,sig) = Amp(d,1,x,y,sig) ;

16 OutA(d,i,x,y,sig) = OutA(d,i-1,x,y,sig), Amp(d,i,x,y,sig) ;

17

18 // Delay amounts assignments:

19 R(d,i,x,y) = fdelay1s(D(d,i,x,y) * SR / 343) ;

20 OutR(d,1,x,y) = R(d,1,x,y) ;

21 OutR(d,i,x,y) = OutR(d,i-1,x,y), R(d,i,x,y) ;

22

23 // sequence composition:

24 Out(d,n,x,y,sig) = OutA(d,n,x,y,sig) : OutR(d,n,x,y) ;

25 process = Out(d,nSpeakers,x,y) ;
 

The program generates output signals yi for i ∈ [1, 8]

1. y1 (t) = 0.2

2. y2 (t) = 0.196116135138184

3. y3 (t) = 0.185695338177052

4. y4 (t) = 0.171498585142509

5. y5 (t) = 0.156173761888606

6. y6 (t) = 0.14142135623731

7. y7 (t) = 0.12803687993289

8. y8 (t) = 0.116247638743819

2.2 Characteristics of the WFS

There are three different types of sources:

1. Point Source

2. Focus Point Source

3. Plane Wave

66
 Figure 2.17: Block diagram of the process generated by Faust svg compiler

The point source gives an absolute position for every listener in the room, when

the listener moves the source is still perceived as being in the same position. If there are

multiple sources, there is a sense of perspective of the sound field. A focus point source

is a source that is located in front of the speaker array.18 The location of these sources

will be perceived accurately only by listeners who are in front of the source. Finally,

plane waves are sources that have the same angle for every listener. They “”follow the

listener when he moves inside the space. The latter is mostly used for reverb/reflections

or for emulating a conventional stereo reproduction 19 (Snoei 2014).

18
Sources located in the real space as opposed to the virtual space have negative x coordinates.
19
It is possible to emulate any stereophonic or surround setup when placing sources –virtual
loudspeakers– accordingly.

67
2.2.0.6 Delays, Doppler and Distance Cues

In the example above delay lines are interpolated using a 1st-order Lagrange

interpolation .20 Given that there must be some sort of interpolation when sources are

moving, there is a cost on CPU if applying higher orders in real time but they give

better sound quality and sound localization. In addition, when sources move from the

“virtual” space behind the speakers to the “real” space in front of them, delay times are

inverted causing a “click”. The Game of Life uses a crossfading technique to solve this

issue, other systems may use interpolation techniques but according to Wouter Snoei

these are not as effective. 21

Doppler shifts are generated as a natural side effect on WFS systems and

happen because of the distance between the source and the listener: normal Doppler

shift, and the distance between the source and each speaker: Speaker Doppler shift. The

former can be canceled out but the latter, caused by the discontinuity of the array,

cannot be canceled and could be exacerbated by the cancellation of the normal Doppler

shift, producing a chorus effect (Snoei 2014).

It is possible to improve distance cues by using filters and adding a global

amplitude roll-off relative to a reference point or line.22 As with any system, one can

apply a low-pass filter to mimic distance cues as higher frequencies decay faster than

lower ones.
20
”Lagrange interpolation is a well known, classical technique for interpolation [194]. It is also called
Waring-Lagrange interpolation, since Waring actually published it 16 years before Lagrange [312, p.
323]. More generically, the term polynomial interpolation normally refers to Lagrange interpolation. In
the first-order case, it reduces to linear interpolation.” (Smith 2010).
21
Wouter Snoei is the lead developer behind The Game of Life software end of the system. An alumni
of the Institute of Sonology in Den Haag, the composer still maintains, updates and add new features
to the Supercollider driven system. WFSCollider can be installed from its Sourceforge repository at:
http://sourceforge.net/projects/wfscollider/.
22
The Game of Life system imposes a roll-of of 6dB per distance doubling according to the inverse
square law (l/r) (Snoei 2014).

68
 To conclude, WFS is more accurate than first-order-Ambisonics for moving

sources as the distance cues are clearer. Given that there is no sweet spot, it can

be used for installations and acousmatic performances where listeners can freely move

around to experience a work from different points or perspectives. There is a visual

discrepancy between the output of the system and the location of the speakers as sound

sources can be positioned anywhere – up to 200 meters away in the Game of Life system

– around an audience which is mainly accustomed to listen to stereophonic and surround

setups. In concert situations, I have seen people looking at the speakers surrounding

them “looking for” sounds as if they were moving from speaker to speaker. With a

WFS system we see something but we hear something that does not corresponds with

the visual cues we rely on for spatial processing. I do not think this is a disadvantage

of the system, on the contrary, it opens the doors to a new way of listening that goes

beyond the loudspeaker and beyond the constrains of the physical space.

2.3 The Design of The Portable Game of Life System

Game of Life is a portable WFS system that was located in room ANNA

RW47, at Raamweg 47, The Hague, Netherlands in November 2014 when I visited. It

is maintained by the Game of Life foundation which is subsidized by the Performing

Arts Fund NL (Fonds Podiumkunsten NL). The foundation was founded in September

of 1999 and its purpose is to promote electroacoustic music organizing and curating

concerts in the field of spatial reproduction.23 The GOL was built as a system that can

be easily used by composers with good sound quality and the most important aspect:

portability. The system can be pack by a few people and transport anywhere by land.
23
The site of the Game of Life Foundation is active and contains documentation and how to place a
request to work with the system. It can be accessed at: http://gameoflife.nl/en/.

69
 Figure 2.18: a complex path created with the WFSCollider software

One of the most remarkable aspects of the GOL is its user interface that allows the

composer to interact with the system in an intuitive manner.

The hardware was designed by Raviv Ganchrow accordingly to the following

specifications:

• 192 coaxial speakers (SEAS)

• 12 active subwoofers (Hypex)

• 24 8-channel Class D amps (Hypex)

• 24 Behringer ADA8000

• 8 Motu 2408mk3 / 2 PCle

• 2 Mac Pro 8-core 21.8Ghz

• MacBook with a Motu 828 sound card

• ± 300 meters of cables

70
 Figure 2.19

The system can be configured as a 10x10 meters array. There are 24 mod-

ules; for a square setup there are 6 modules on each side, each containing 8 speakers.

It can also be configured in any shape allowed by the 24 modules including for ex-

ample, 5x7 modules, circular setups, etc. There is a subwoofer every other module.

Each module is driven by one 8-channel amplifier that receive an analog signal from

a Behringer ADA8000.24 The MacBook computer is connected to the Motu828 audio

interface for synchronizing the digital signals with the 8 2408mk3 connected to 3 ADAs,

each 2408mk3 controlling three modules. There is a network switch between the control

unit (MacBook) and a HEAR ADAT ethernet extender between the Motu828 and the

2408mk3. The Mac Pro servers controls 4 of the motu2408s as shown in Figure 2.19

(Snoei 2014).

The open source software fully written in Supercollider features a solid interface

with a live input and control of every sound source’s spatial position. The system is being

expanded to consider directionality 25 and the addition of more effects and synthesis

techniques (Snoei 2014b).

24
A Digital to Analog Converter.
25
Aperture of the virtual speakers such as in IRCAM’s SPAT (Spatialisateur 2012).

71
Figure 2.20: WFS studio at TU Berlin controlled by parametric spatialization tools written by

the author for the Pure Data software.

2.4 The TU Berlin Systems

The Technischen Universität Berlin has two WFS systems. The lecture and

event hall WellenFeld H104 features a system with more than 2000 speakers grouped in

modules of 8 channels with 3 drivers for each channel and two subwoofers. Drivers are

at a distance of 10 centimeters each. The speakers are controlled by a computer cluster

on the left side of the stage with 832 audio channels. The hardware is controller by a

Linux system running the sWonder software written at the TU Berlin.

The other WFS system is located at the 3rd floor in the EN - building of

the Department of Audio and Communication (Fachgebiet Audiokommunikation) and

is mainly used for research and composition. It is possible to compose in the studio and

transfer the work to the larger system. Like the Game of Life, there is a Supercollider

server that is located in a different room used to control the spatialization. The system

features the following hardware and software:

• 24 panels with 8 channels each, 2 woofers connected to 4 channels. 24 tweeters.

• Ethernet/Network based Dante Controller Matrix by Audinate.

• 1 main computer with the matrix software (Mac Pro in different room)

• 2 nodes computers where the spatialization software/interface runs.

72
 (a) 3D view (b) Side view (c) Top view (d) Top view (e) Top view

Figure 2.21: Speaker layouts.

• Xwonder. Custom software built at TU.

• Iannix software for creating algorithmic trajectories with OSC in/out.

The system can be controller by OSC messages from other applications such

as Pure Data.

73
Chapter 3

Spectral Spatialization

This chapter deals with a thorough description of the research conducted at

EARS using its eight channel system. It includes a description of sound morphologies

(Landy 2007), spectral typologies and morphological archetypes used with different spa-

tialization techniques. I emphasize spatialization techniques such as elliptical motion,

paths and spirals that were used for a series of pieces under the name Laberintos in-

spired by Jorge Luis Borges’ story La biblioteca de Babel .1 These etudes explore the use

of spatialization techniques using ambisonics, SPAT and discrete spectral spatialization,

the latter a technique developed at EARS that combines the creation of an isotropic

sound field with directionality and spectral panning.

The EARS studio is a satellite facility of the University of California River-

side. Designed and built by director Paulo C. Chagas for research in experimental

electroacoustic music, it became a flexible spatialization studio that allows for different
1
The Library of Babel. Borges’ short story describes a universe that consists of a vast library con-
taining an undefined number of hexagonal rooms that are identical. The order of books and their
content seem to be random and extend to the infinity. The books contain every possible permutation of
characters, from books with only one letter to meaningless books of random words. Given that all the
combinations of the alphabet are possible, the library holds all possible books in the universe.

74
configurations of speakers which can be arranged very quickly. The following is the

audio equipment used during the course of the research:

• 8 Genelec 8030 loudspeakers

• RMS firewire audio interface

• Mackie Pro FX16 Mixer

• iMac 2011

• ProTools, Max/MSP, Pure Data, Supercollider, Ircam Forum plugins

I have worked with several electronic music techniques in the time and spectral

domains but special attention is given to spectral morphing, a technique for merging

two sound spectra. Basically, frequencies that are common to both sounds are em-

phasized while the rest of the spectrum is diminished. The spectrum of a wave is a

two-dimensional representation of its frequencies and their relative amounts at some

point in time. By morphing two sounds new spectra with some characteristics of both

sounds is created. Morphing between two sounds requires a process that starts with

the Fast Fourier Transform and the technique is particularly interesting when combined

with synchronized sound trajectories that add musical meaning to movement, the main

goal of this research. Two basic procedures can be applied: sounds and musical gestures

can be spatialized to their entire length or can be timed stretch to fit their trajectory’s

length using a one-to-one – injective – mapping technique, the most common approach

to spatialization, or by spectral morphing two sounds – bijective – in a single trajectory

which is a more interesting approach to spatialization.

The main tools used for the design and implementation of spectral spatializa-

tion were the following 2 :

2
Software tools were developed using the Pure Data and Max/MSP programming environments and
sound examples were also generated with Csound, Open Music and/or Apple Logic.

75
 • The Fast Fourier Transform

• Cross-synthesis

• Envelopes

• Panning and Ambisonics

The FFT

3.1 Spectral Domain. The Fourier Transform and its mu-

sical applications

A periodic wave is a wave that repeats itself. Although natural phenomena

seems to be governed by chaos, there is an element that makes it comprehensible for us:

patterns of repetition. Seasons repeat every year, Comet Halley is visible from Earth

every 75 years, water undergoes a cycle from precipitation to evaporation. If we look at

the smaller scale of natural phenomena we find that mechanical waves, oscillations that

travel through space and time, are an essential component of nature given their ability

to transport energy from one place to another.

We have evolved to perceive the types of waves that manifest in the form

of light and sound with our eyes and ears respectively. They differ in the way they

propagate, their frequency content and length. Visible light is electromagnetic radiation

with waves in the frequency range of 405 THz to 790 THz with a wavelength in a

range from 380 nanometers (nm) to about 740 nm. If light travels through a prism, it

decomposes into its constituents colors; it was Isaac Newton who discovered that light

could be reconstituted back into its original form if passed through a prism again. We

cannot use a prism to decompose sound into its sinusoidal components but there is a

mathematical device: the Fourier Transform, which can deconstruct a sound wave and

76
under certain conditions, the same wave can be mathematically reconstructed using the

Inverse Fourier Transform.

There are several terms related to the Fourier Transform and its implemen-

tations: Fourier series, Fourier Transform, Discrete Fourier Transform (DFT), Short

Time Fourier Transform (STFT) and Fast Fourier Transform (FFT). Although they

refer to different things, the latter is the most common term used in audio applications.

Joseph Fourier (1768-1830) was a French mathematician and physicist who, after study-

ing how heat propagates to an object, proposed that a periodic wave could be expressed

as the sum of infinite simple waves. The Fourier series decomposes periodic functions

or periodic signals [sig˜] into the sum of an infinite set of cosine and sine waves

∞
X
f (x) = a0 + an cos nwt + bn sin nwt (3.1)
n=1

or equivalent complex exponential functions of the form

eint (3.2)

where an and bn are the coefficients that represent the amplitude of the waves and nwt

represents their angular velocity.

To analyze music digitally stored as a recorded audio signal it becomes neces-

sary to create software or hardware devices that emulate how our hearing works. The

first step would be to create a tool that can detect when events happen in time and how

fast they repeat. These two elements, time and repetition, appear to be fundamental

parts of any music. For example, we can look at the rhythm cycles of the Indian tala,

the key cycles of a classical sonata or rondo, a melody or motif that is repeated in its

original form, intervallic relationships, etc. If we can count these repetitions and localize

them in time, then we can extract a statistical analysis of the given work. In computer

77
music, it is possible to modify the spectrum of the work and perform an Inverse Fourier

Transform to create new sounds.

The standard Fast Fourier Transform (FFT) is an efficient 3 algorithm that is

used to compute the Discrete Fourier Transform (DFT) which is based on the mathemat-

ical series of Fourier. Jean-Baptiste Fourier formulated a theory stating that periodic

waveforms – for example sound – can be deconstructed into infinite combinations of

sine waves of different amplitudes, frequencies and phases. In order for a computer to

calculate the series it is necessary to discretize it into small segments. Given that the

human hearing range has its limits around 22 KHz, digital sound is sampled at 44.1 kHz

–Nyquist sampling theorem–. By only computing a small range of the real spectrum the

infinite series is transformed into a discrete algorithm. Additional segmentation is done

– windowing – depending on the application, and segment to be analyzed.

In order to extract information about time-allocation of frequencies, it is nec-

essary to use the Short-Time Fourier Transform (STFT) which is essentially a DFT

adapted to perform that task.4 Although there are several algorithms to compute the

FFT, all compute the DFT ; in this chapter the general term FFT will be used to

describe the process of spectral domain analysis, that is the conversion of any periodic

signal into the sum of its infinite sinusoidal (sine and cosine) components obtaining mag-

nitude (energy) and phase for each bin. The FFT can be viewed as a transform that

converts any finite, discrete signal into a finite, discrete sum of discretized sinusoidal

components. With the FFT, it is possible to obtain magnitudes and amplitudes of the

desired frequencies of an entire audio clip and look at differences and similarities across

the spectrum. The output of the FFT is the spectrum. Therefore, by applying a FFT
3
The FFT performs a fast computation of the Discrete Fourier Transform (DFT)
4
The Short Time Fourier Transform splits the signal into equal overlapping blocks and calculates a
DFT for each block which is windowed.

78
to an incoming audio signal we are taking the signal from the time domain to get useful

statistical information in the spectral domain.

Given that any type of Fourier analysis assumes infinite periodic signals, peri-

odicity is obtained by repeating the signal forcing it to become periodic. To obtain a

small segment of the signal we could multiply the original audio by a signal of value 1

– a rectangular window – during the time period of interest. Moreover, by multiplying

the original signal by a window that has a shape of a bell, we minimize spectral leakage

that occurs at the beginning and end of the windowed signal that adds distortion or

noise (Ramirez 1985).

Different types of windows are commonly used, for example: Hann (Hanning),

Hamming, Blackman (Roads 2004). These series of windows are overlapped – shifted

by n-samples and summed – for better time localization. The FFT is a function of both

time and frequency. To summarize, the spectrum provides insightful information about

the components of a sound which is commonly obtained in a computer by implementing

a Fast Fourier Transform. With computer languages for real-time audio processing such

as Max/MSP, Pure Data, Csound or Supercollider, it is possible to compute an FFT and

its inverse iFFT with minimum coding. Finally, The FFT differs from its mathematical

counterpart – the Fourier Transform – in three different aspects. Firstly, it applies to

discrete-time sequences such as an audio recording in a digital format. This is useful

for storing and manipulating a spectrum. Secondly, it is a sum rather than an integral

making it easier to implement with software and hardware. Lastly, there is no need to

define the function over time given that the function operates on a finite data record.

79
 Figure 3.1: Spectrogram of an entire performance of Pierre Boulez’s Répons

One of the uses of spectral domain analysis in contemporary music is for finding

changes in texture or the gravitational center of a work in order to understand its form

or key features.5

For the purpose of this research, a Hanning window was used to compute the

FFT. In addition sound buffers containing this type of window were used in replacement

of fading and panning algorithms. Different windows –not limited to the common ones–

were generated using buffers in Pd and Max. Experimentation at EARS have shown that

different windowing techniques affect the resultant spectrum by introducing sidebands

and clutter that are translated into noise which may be desired – as part of the work’s

aesthetics – for some morphing techniques.

Given:

Z∞
x(t) = X(f )ej2πf t df (3.3)
−∞

we can evaluate X(f )from x(t) with the Fourier Transform:

5
See (Jaroszewicz 2013).

80
 Figure 3.2: Hann Window and resultant spectrum

Z∞
X(f ) = x(t)ej2πf t dt (3.4)
−∞

The first equation x(t) refers to a time-domain signal while the second one,

X(f ), represents a magnitude of the Fourier Transform of the preview time-domain

signal.

The Discrete Fourier Transform (DFT)

N
X −1
X[n] = x[k]e−j2πnk/N (3.5)
k=0

n = 0, 1, 2, 3..., N − 1

returns information containing magnitude, frequency and phase of a discrete

signal.

3.1.1 Windowing

The theory behind the FT allows us to analyze any signal or data. In order to

do that, it would be necessary to apply a window the size of the work which in return

will be converted to a spectrum containing information about all the frequencies present

in the piece, which their magnitudes and phase providing statistical information without

telling much about time or when those frequencies are located.

81
 As oppose to real-time analyses using many overlapping windows, conserving

memory and obtaining results quicker, the analysis of an entire piece using long windows

applied to the desired sections could give us more information about the section itself

and the opportunity to compare it to other sections of the same piece.

According to Curtis Roads:

A spectrum analyzer measures not just the input signal but the product
of the input signal and the window envelope. The law of convolution, states
that the multiplication in the time-domain is equivalent to convolution in
the frequency-domain. Thus the analyzed spectrum is the convolution of
the spectra of the input and the window signals. In effect, the window
modulates the input signal, and this introduces sidebands clutter into the
analyzed spectrum.
A smooth bell-shaped window minimizes the clutter (Roads 2004).

As mentioned above, windows such as Hann, Hamming, Gaussian, Blackman-

Harris are commonly used for analysis and resynthesis in music. One can always try

different type of windows and decide what is more convenient for the type of analysis

to be done.

3.1.2 Window creation with Max/MSP

In Max/MSP it is possible to create a window by writing the values of a function

into an object that stores audio samples buffer˜ . In the following example a window

of size 512 values is generated using the Hanning function:

  n 
ω(n) = 0.5 1 − cos 2π (3.6)
N

If the sample rate is 44100 and the FFT size is 1024, then a window that is

half the FFT size (512) creates band sizes of approximately 86 Hz.

44100
= 86.15 (3.7)
512

82
 Figure 3.3: Hanning Window in Max

3.2 FFT in Max/MSP

If a window is created with a function as shown in Figure 3.3, a way to perform

an FFT in Max is to use the object fft˜ and multiply the incoming signal by the

window. Another option is to use a patcher loaded by the pfft˜subpatch object. Inside

the patcher there should be an fftin˜hanning object with an argument specifying the

desired window type. If using the pair pfft −→ fftin , three elements can be obtained

from the signal: a real part, an imaginary part and the FFT bin index. These are the

three outputs of the fftin˜hanning object inside the pfft patch. With the real and

imaginary signals then it is possible to convert to polar form and the opposite, from

polar back to cartesian before the ifft˜ . In polar form the signal becomes magnitude

and phase. The third outlet is very useful for analysis, storing or displaying a signal.

A form of spectral spatialization can be achieved by dividing the spectrum

into many different channels.6 By knowing the FFT bin index, it is possible to gate

the signal to a specific buffer when a particular index has been reached. Moreover, the
6
8 channels at EARS.

83
 Figure 3.4: Dividing a spectrum into 8 different buffers

signal can be stored on different buffers that can be read and output to the time-domain

multiplying them by their respective fftout˜hanning objects, as shown in Figure 3.4

. Each output of the pfft˜subpatch can be sent to an ambisonics encoder for further

spatialization.

Inside the FFT: Cross-synthesis

Cross-synthesis describes a number of techniques that in some way combine

the properties of two sounds into a single one. Convolution is a special case of cross-

synthesis and serves as a bridge between time-domain and spectral-domain. In signal

processing, Convolution is the multiplication of two spectra (Roads 2004, 213).

Several techniques were used and tested at EARS using a wide variety of sounds

with contrasting spectral content.7 For the purpose of spatialization, morphing between

two sounds requires control over their lengths. In addition, expansion and compression

(compand) in the spectral domain was implemented to match the amplitudes of signals
7
see Appendix B.

84
with strong spectral energy.8 Techniques included cross-synthesis by mixing, source-

filter, convolution, cross-modulation, square-root convolution and cross-product.

The following implementations have been applied in the spectral domain in

combination with cross-synthesis:

• Spectral shredding

• Spectral delays

• Spectral compand

• Spectral gates

Spectral shredding is a term applied to the technique of segmenting the spec-

trum into as many parts as channels are available in the system. For example, with an

octagon array the spectrum of a sound is divided into eight equally 9 divided sections.

This is achieved in the spectral domain after performing an FFT. Each segment is then

sent to the time domain after an iFFT for further spatialization. I have found that spa-

tialization at fast speeds creates an effect similar to applying granular synthesis, that

is dividing a sound into particles that are less than 50ms in length. Spectral shredding

in combination with ambisonics is a great technique for creating a virtual space where

more directional sounds can be layered thus creating a complex and rich sound space.

Spectral gates aid in the control of the amount of noise that is filtered. Applying

spectral noise gates before cross-synthesis creates a smoother transition between the two

sounds, something that may be desired or not. Spectral delays can be used for aesthetical

reasons but delaying across the spectrum with different times and feedback amounts can

help perceive a smoother morphing process.

8
Soundhack’s Spectral Shapers including its +spectralcompand are great tools in the form of plugins
for spectral processing. http://www.soundhack.com/spectral-shapers/.
9
Depending on the nature of the sounds, one could emphasize different parts of the spectrum by
dividing it into unequal segments.

85
Spatialization

• Circular and elliptical motion

• Paths, Spirals

• Point Clouds

Different techniques and software for spatialization had been tested at EARS.

For transformations using two, three or four channels in a circular array, logarithmic

panning in combination with Hanning windows for control of fading work well. In

addition, a window and its inverse can be used to cross-fade between original sounds

and morphed spectra. In this case windows are not used for convolution but as a control

function.

Circular and elliptical motion was first tested using IRCAM’s 10 Spat, ICST’s

ambisonics 11 objects for Max/MSP and CICM’s HOA library for Pure Data. Working

with ICTS’s and CICM’s libraries allow for a better control of specific output as opposed

to the more user-friendly Spat. Most importantly, both libraries are open source and

well documented.

Circular motion works very well especially when working with sounds with a

short attack and fast repetition patterns and for applying morphing techniques.

Many of the sounds were previewed by using an ambisonics to binaural object.

Summary of Techniques Applied to Spatialization:

Cross-synthesis

1. Cross-synthesis by mixing

2. Source-filter

3. Convolution
10
http://forumnet.ircam.fr/product/spat/?lang=en
11
http://www.icst.net/research/projects/ambisonics-tools/.

86
 4. Cross-modulation

5. Square-root convolution

6. cross-product

Spectral Techniques with Cross-Synthesis

1. Spectral shredding

2. Spectral delays

3. Spectral compand

4. Spectral gates

Spatialization

1. circular and elliptical motion

2. Free Paths / Spirals

3. Point Cloud

3.3 Cross-synthesis and Spectral Shredding

Cross-synthesis describes a number of techniques that in some way combine

the properties of two sounds into a single one. Convolution is a special case of cross-

synthesis and serves as a bridge between time–domain and spectral–domain. In signal

processing, convolution is the multiplication of two spectra. The Law of Convolution

states the following:

Convolution in the time domain is equal to multiplication in the frequency

domain, and vice versa (Roads 2004).

There are several techniques used to combine aspects of two sounds. For ex-

ample, IRCAM’s Audiosculpt 12 offers three different types:

12
Audiosculpt is a software for viewing, analysis and processing of sounds by IRCAM. Among other
features, it offers cross–synthesis powered by SuperVP, IRCAM’s phase vocoder analysis engine. More
information here: http://anasynth.ircam.fr/home/english/software/Audiosculpt.

87
 • cross–synthesis by mixing

• source–filter

• generalized cross-synthesis

Cross-synthesis S of two sounds S1 , S2 with corresponding spectra M and phase

θ can be calculated in several ways:

1. Summation: S = S1 + S2 Same as cross-synthesis by mixing, while the amplitude

of S1 increases, the amplitude of S2 decreases.

2. Convolution: M = M1 M2 , θ = θ1 + θ2 . Multiplication of magnitudes and summa-

tion of phases after FFT and conversion from cartesian to polar.

3. Cross-modulation: M = M1 , θ = θ2 . Magnitudes of S1 , phase of S2

√
4. Square-root convolution: M = M1 M2 , θ = 21 (θ1 + θ2 )

5. Cross-product #1: M = M1 M2 , θ = θ1

6. Cross-product #2: M = M1 M2 , θ = θ2

Figure 3.5: Generalized cross–synthesis with gen˜in Max/MSP

88
 The first technique: summation or cross–synthesis by mixing, involves mixing

of two sounds, that is cross-changing their gains in the time-domain. A spatialized

summation would involve fading in and fading out the two sounds respectively. Although

there is no spectral processing involved, it can give the listener the impression of sounds

morphing into different sounds and space.

The rest of the techniques are performed in the spectral domain after analyzing

the signal with an FFT and converting the sines and cosines to amplitudes and phases.

In other words, converting from cartesian to polar coordinates. These techniques differ

in the way the amplitudes and phases are combined. Operations between magnitudes

are multiplicative while phases are subject to addition.

Source-filter cross synthesis requires an extra step taking the signal to the

cepstrum domain. The word cepstrum is a rearrangement of the word spectrum.The

following steps are required:

x(n) −→ FFT −→ abs() −→ log() −→ iFFT −→ cx (n)

In Max, and especially with gen˜, 13 it is possible to perform these operations

at the sample-level in real-time. By synchronizing the cross cross–synthesis of two – or

more – sounds as they move from one loudspeaker to another, it is possible to create

“real-time spectral spatialization”.

In addition, any of these techniques can be combined with “spectral shredding”

creating spectral spatialization at different times between portions of the spectrum.

13
Max/MSP’s gen˜ compiles in real time any visually generated code. Code generated with gen˜
can be exported to C++.

89
 Figure 3.6: Source–filter cross–synthesis in cepstrum domain with Max

3.4 Tools and Software for Spatialization

Several aspects of spatialization and parameters can be modified in real-time

using software tools.

• Speed of circular motion of a sound source including Doppler effects

• Spatial trajectories of sound sources

• Behavior of sound source e.g. swarm

• Directionality (ambisonics order)

• The creation of a broad sound image

• The creation of an acoustic environment using filters and delays.

Perceptual parameters include:

• Azimuth and elevation

• Source presence and distance attenuation

• Room presence

90
 Figure 3.7: Swarm behavior of eight sources after spectral shredding

• Reverberation

• Envelopes

DSP parameters include:

• Equalization

• Doppler effect

• Air absorption

• Reverberation

• Directional distribution of various speaker layouts e.g. hexagon

• Delay

Many of these parameters can be set in IRCAM’s SPAT, ICST’s Ambisonics

Tools and the HOA Library.

3.5 Some Remarks

The ability to localize sounds in space is affected by the spectral content of

the sound itself, especially its attack. As lower frequency sounds tend to be harder to

91
localize, sounds with a complex spectrum are preferred and work well for morphing. The

techniques the author has implemented for his own work range from cross-synthesis of

two sounds to spatialization techniques including basic rotation using an array of eight

speakers, swarm behavior, trajectories and combined movements using a set of tools for

parametric spatialization developed in Pure Data. See Figure 3.7.

3.6 Envelopes

From a psychoacoustic perspective, we perceive music as events and flows that

can be continuous or a succession of discrete time intervals, for example: floats and

integers, analog and digital. In music, the analogy corresponds to staccato versus legato

or a scale versus a glissando. Dealing with events or streams can change the way

the micro and macro structures are perceived. If there is a gradual transformation

throughout a piece, changes may not be perceived by the listener but contrasting events

belonging to the macrostructure clearly define sections.

A way to divide sections is to consider the change in direction of a signal. An

envelope or transient generator makes a signal to rise and fall “smoothly” approaching

its limits [0, x], 0 >= x <= 1 without any discontinuity that can cause unwanted clicks

(Puckette 2007, 89). As envelopes can be applied to modulate a variety of musical

parameters, when applying this concept to computer music, the main application is the

modulation of the amplitude of a signal. A typical synthesizer implementation has four

components: Attack, Decay, Sustain, Release (ADSR). Any continuous function can

be used as an envelope and some are designed to emulate the changes in amplitude of

acoustic instruments. For example, Supercollider has built-in envelopes that can be used

for that purpose:

92
  
1 Env.perc.test.plot;

2 Env.triangle.test.plot;

3 Env.linen.test.plot;

4 Env.sine.test.plot;

5 Env.asr.test.plot;

6 Env.adsr.test.plot;
 

Figure 3.8: Percussion, triangle, linear and sine envelopes created in Supercollider

or one can generate a more complex envelop as shown in the code snippet and

Figure 3.8.
 
1 ({

2 var curve = \sine; //or \step \exp \sin \wel

3 var dur = 2;

4 var envgen, trig, output;

5 var env1 = Env([0,1,0,0.5,0.3,0.7,0],[dur/2,dur/2,dur/2,dur/2,dur/2,dur/2],curve);

6 env1.plot;

7 envgen = EnvGen.ar(env1,doneAction:2);

8 output = SinOsc.ar(mul:envgen) ;

9 }.play

10 )
 

93
 Figure 3.9: A more complex envelop applied to a sawtooth oscillator

3.6.1 The Sound Envelope as a metaphor for composition

Envelopes can be used for other purposes than modulating the amplitude of a

sound in time. For example, an envelope could be used to modify parameters such as

density or texture when working with acoustic instruments. The envelope becomes a

composition tool. The following excerpt shows how to use a linear envelope to control

orchestration. For example, measure 46 of Ligeti’s Atmosphères (1961) can be inter-

preted as a Quasi-Gaussian envelope applied to density in the orchestration (see Figure

3.10). Atmosphères originated as an electronic piece (Pièce électronique no. 3 ) after

writing two electronic works (Glisandi, 1957 and Artikulation, 1958) at the WDR studio

in Cologne. The beginning of the piece gradually builds as a timbral changing cluster

without any identifiable melody or rhythm: “The micropoliphonic web enters extremely

softly, increasing in volume only in the last four bars. Each part has a step-wise pitch

sequence except were octave shifts are necessary[..](Steinitz 2003)”.

94
Figure 3.10: The shape of the increasing texture resembles a Gaussian envelope

95
 Figure 3.11: The PD patch showing the creation of an envelope (ADSR) applied to a table

In PD, the object vline˜, Figure 3.11 is used for signal rate control data. vline˜

is a high precision audio ramp generator that can be used to create an ADSR envelope.

As in Supercollider, the object can be programmed with an array of numbers or list.

96
3.7 Parametric Tools for Spatialization

3.8 Parametric Spatialization

The idea of creating PD externals for spatialization came from the difficulties

the composer faces when working with libraries such as Ircam’s SPAT, ICST’s Am-

bisonics Tools and CICM’s Hoa Library. There is no standardized system of coordinates

and/or spaces. For example, the HOA Library includes an object for the rotation of the

ambisonics field. Ircam’s SPAT includes examples on how to achieve different rotational

figures using signals but the composition of paths has to be done with programming.

I created a series of externals for Pure Data to help the composer working with

algorithms simplifying the use of trajectories by using parametric or procedural ideas.

These externals specify trajectories of three different categories: circular motion, recti-

linear motion and a stochastic point cloud. These can be applied to a two-dimensional

or three-dimensional field such as WFS or ambisonics respectively.

The externals work well with the HOA library 14 and can be used in any en-

vironment that receives OSC data as they output coordinates (x,y,z )as floating point

numbers. The system of coordinates is based on a cube of 2 units with (0,0,0) at the

center with a Cartesian system of coordinates (-1..1)

14
The HOA Library is an open source high order ambisonics spatialization tools collection programmed
in C++ with externals for Max/MSP, Pure Data, included in Faust and also available as a VST plugin.
It was developed by the CICM, Center for research in Computer Science and Musical Creation of the
University Paris

97
3.8.1 HOA Library externals

3.8.1.1 hoa.2d.decoder∼

Decodes an ambisonic sound field for several loudspeakers.

Arguments:

1. Ambisonics order

2. Decoding Mode:

• 0 or ambisonics

• 1 or binaural

• 2 or irregular (see below)

3. Number of speakers (min= 2 * order +1)

4. Offset of speakers in degrees) e.g. [@angle] 0 30 60 90

Note that ambisonics decoding is used for a circle of equidistant speakers. For

an irregular configuration there should be delay and gain compensation for each speaker.

At EARS, I used IRCAM’s spat.align∼ for calculating those parameters.15 It is possible

to do it manually knowing the distance to each speaker or measure with a microphone.

Any configuration of speakers can be used if this method is applied. I created two

abstractions that use the hoa.process∼ external for working in PD with the HOA

library:

1. mjDelayCompensation (processes multiple delay lines in parallel)

2. mjGainCompensation (processes the gains so all speakers are in phase according

to a reference position which should be the middle of the ambisonics field)

15
IRCAM SPAT calibration: Ircam’s SPAT (Max/MSP) includes objects for calibration delay and
gain: spat.gaincalibration∼ and spat.delay calibration∼.

98
 Figure 3.12: Speakers configuration at EARS

Both abstractions need a list with floats representing the values for each speaker.

Delay is in milliseconds and gain is a multiplier (0–1). By applying delay and gain com-

pensation we are creating a virtual circle regardless of the layout of the studio allowing

to work with spatialization in different conditions.

mjDelayCompensation processes multiple delay lines in parallel Irregular mode

is used for standard configurations:

1. mono

2. stereo

3. 3.1

4. quad

5. 5.1

6. 6.1

7. 7.1

The hoa.decoder needs phase optimization to avoid artifacts. This can be

achieved by using the hoa.optim∼ external

3.8.1.2 hoa.optim∼

Weights the spherical harmonics for different configurations

99
Arguments:

1. Ambisonic order

2. Optimization type:

• basic (no effect)

• maxRe (when the listener is at the center of the circle)

• inPhase (for an audience that covers the entire circle)

3.9 Parametric Spatialization objects for PD

• mjRotation

An external that can be used for the creation of elliptical, circular or linear

motion paths. The inlet parameters are trajectory(0..1), ellipticalX(0..1) and exllipti-

calY(0..1) and the outlets output (x,y) coordinates in the form of floats. It is possible

to create pendular rectilinear motion over the Y axis and vice versa by sending ellipti-

calX = 0 and ellipticalY = 1 . The trajectory parameter can be driven by a phasor∼

connected to a snapshot∼ thus sending floating point values from 0..1 to the trajectory

inlet.

100
  
1 void mjRotation_float(t_mjRotation *x, t_float f){

2 t_float a1 = x->f_x;

3 t_float a2 = x->f_y;

4 t_float a3 = x->f_z;

5 ;

6 x->f_x = sin(f * x->f_mult)*x->f_elipX;

7 x->f_y = 0;

8 x->f_z = cos(f * x->f_mult)*x->f_elipZ;

9 outlet_float(x->l1_out, a1);

10 outlet_float(x->l2_out, a3);

11 outlet_float(x->l3_out, a2); // This works for HOA Library

12 }
 

• mjRose

A mathematical rose curve.16 This is an expansion of the idea of circular

motion for HOA or WFS. Given that this external outputs (x,y,z) coordinates, it is

suitable for working with a spherical system but can be used with a planar system with

Z = 0. the mathematical rose curve is best know by its polar equation:

r = asin(nθ),

or

r = acos(nθ).

• If n is odd, the rose is n-petalled. If n is even, the rose is 2n-petalled.

16
Also known as Rhodonea

101
 Figure 3.13: Simple and complex paths can be created with mjRose

• If n = r/s is a rational number, then the curve closes at a polar angle of θ = πsp,

where p = 1 if rs is odd and p = 2 if rs is even.

• If n is irrational, then there are an infinite number of petals (Weisstein 2104).

 
1 void mjRose_float(t_mjRose *x, t_float f){

3 double n;

4 double d;

5 modf(x->f_n,&n);

6 modf(x->f_d,&d);

7 int z = x->f_z;

8 int in = n;

9 int id = d;

10 if(in%in != 0 && id%id !=0){

11 x->f_mult = M_PI*2*d;

12 } else {

13 x->f_mult = M_PI*d;

14 }

15 x->f_position = f;

102
16 float position = x->f_position*x->f_mult;

17 float rose = sinf((n/d)*position);

18 float out_x = rose*cosf(position);

19 float out_y = rose*sinf(position);

20 float out_z = rose*sinf(z);

21 outlet_float(x->l1_out, out_x);

22 outlet_float(x->l2_out, out_y);

23 outlet_float(x->l3_out, out_z);

24 if (out_z == 0.0 && out_y == 0.0 && out_z == 0.0){

25 outlet_bang(x->l4_out);

26 }

27 }
 

103
 • mjRandom

Stochastic point cloud generator. The first inlet – as with the other externals

– accepts a float from 0..1 to traverse the cloud. The second inlet takes a value for the

seed of the random number, the third inlet selects the number of points. It generates

a 3-dimensional matrix of unit 1 where (x : 0..1) (y : 0..1) (z : 0..1). Random values

are calculated using the C Macro RAND MAX and the function rand().17

The data provided by these objects is not limited to providing coordinates for

sound sources. It can be used for other purposes. For example, to easily create tables

without writing much code for modulating other parameters.18

17
See code.
18
See Appendix A for source code. The C files can be compile using the makefile provided with Pure
Data’s source, also included in Appendix A.

Figure 3.14: 3-D model of a point cloud at EARS

104
 Figure 3.15

 
1 void mjRandom_float(t_mjRandom *x, t_float f){

3 int index = f*x->i_num_points;

4 post("Incoming float %f with index %d",f, index);

5 outlet_float(x->l1_out, x->a_x[index]);

6 outlet_float(x->l2_out, x->a_y[index]);

7 outlet_float(x->l3_out, x->a_z[index]);

8 if (f == 0.999){ // a phasor never gets to 1 but this does not work

9 outlet_bang(x->l4_out);//change seed here

10 }

11 }

12

13 void mjRandom_bang(t_mjRandom *x){

14

15 float seed = x->f_seed;

16 int points = x->i_num_points;

17 if (points > 100){

18 post("Max number of points is 100");

19 points = 100;

20 }

21 if (points < 0){

22 post("Please provide a value between 0 and 100");

23 points = 0;

24 }

25 int index;

105
26 float m1[points];

27 float m2[points];

28 float m3[points];

29 for (index = 0; index < points; index++){

30 m1[index] = (float)rand() / (float)RAND_MAX;

31 m2[index] = (float)rand() / (float)RAND_MAX;

32 m3[index] = (float)rand() / (float)RAND_MAX;

33 post("x[%d] = %f",index,m1[index]);

34 post("y[%d] = %f",index,m2[index]);

35 post("z[%d] = %f",index,m3[index]);

36 x->a_x[index] = m1[index];

37 x->a_y[index] = m2[index];

38 x->a_z[index] = m3[index];

39

40 }

41 }
 

3.10 Parametric Spatialization using 3D Procedural Ani-

mation Software

Increasingly popular 3D procedural tools for animation are being added as

plugins to animation and SFX applications such as Side Effects’ Houdini and the Open

Source Blender. Blender includes a game engine that can be used in real time to send

– and receive – coordinates via the OSC protocol. Both softwares can be extended by

using Python scripting, Blender being more flexible as it is possible to create a full

application for spatialization using its engine. For algorithmic spatialization, there is

an open source graphical sequencer Iannix ,19 which can be programmed and extended
19
http://www.iannix.org/

106
Figure 3.16: mjRose with different values for n, d and z

107
with JavaScript. With Iannix it is possible to create Bezier and elliptical curves in 3D,

interface with OSC and follow objects in real time.

By using Blender or Houdini, the composer can create or use any imaginable

shape not limited to an algorithm. For example, one can import a 3D model of a building

and use its vertices as (x, y, z) coordinates. In addition, the same algorithmic shapes

can be easily created with some scripting. More complex planes and surfaces can be

used as both softwares include dynamics, simulations, deformations, etc, as shown in

Figure 3.17.

Figure 3.17: A 3D mesh (left) with a wave deformation (right) in Blender.

For non-real time work, vertices and its coordinates can be exported as Wave-

front Object files (.OBJ) or other formats which can be used to populate a list or matrix

in PD or Max (Murray 1996). As these type of files include more information than the

necessary and the format is not readable by Max’s coll or PD’s textfile the format

needs to be converted to something else. For that task, I created a command-line tool

– polyConverter – for Mac that can transform Houdini .poly files to a text file that can

be read with both Max and Pd.

108
 The application polyConverter takes the following arguments source(file)

target(file) max(pd). For example, one can write fileFromHoudini.poly polyForPd.txt pd

 
1 if ok {

2 // ------------------------------ CONVERT ------------------------------

3 var isError : NSError?

4 let stringFromFile: NSString = NSString(contentsOfFile: sourceFile, encoding:

NSWindowsCP1250StringEncoding,

5 error:&isError)

6 var copy = stringFromFile.stringByReplacingOccurrencesOfString("POINTS", withString:"")

7 copy = copy.stringByReplacingOccurrencesOfString("POLYS", withString:"")

8 copy = copy.stringByReplacingOccurrencesOfString("END", withString:"")

9 if conversionFormat == "max" {

10 copy = copy.stringByReplacingOccurrencesOfString(":", withString:",")

11 } else {

12 copy = copy.stringByReplacingOccurrencesOfString(":", withString:"")

13 }

14

15 copy = copy.stringByTrimmingCharactersInSet(NSCharacterSet.whitespaceAndNewlineCharacterSet())

16 copy = copy.stringByReplacingOccurrencesOfString("\n", withString:";\n")

17 copy = copy.stringByAppendingString(";")
 

109
 (a) 3D view (b) Side view

(c) Top view

Figure 3.18: Different views of a point cloud. In addition to the creation of curves using param-

eters or manually drawing curves, 3D software allows the composer to visualize the curve in real

time from any angle and modify any vertex or point as desired.

110
 (a) Lines (b) Helix

Figure 3.19: Point Cloud

Figure 3.20: Different paths created with Houdini and Python. See Appendix A for source code.

111
 Part III

Laberintos

112
Chapter 4

Laberintos

This chapter presents s series of works inspired by Borges’ short story The

Library of Babel 1 and the book The Unimaginable Mathematics of Borges’ Library of

Babel by William Bloch (Bloch 2008). In his book, Bloch analyses the possible numerical

outcomes of the story which he uses for the purpose of teaching some mathematical

concepts such as Information Theory, Topology and Combinatorics. Most interestingly

is the chapter Geometry and Graph Theory where he describes how the library in Borges’

story could look like with its hexagonal and spiral shapes.

The next observation is that it’s conceivable the floor plan of a level of the
Library may look like the preceding illustrations, in which the paths run
straight through the hexagons. However, it is consistent with the text – and
the atmosphere of the story – that the corridors weave and spiral around
symmetrically or chaotically (Bloch 2008, 97).

As a composer, the infinite permutations and the paths were the main motivation behind

the music and the research conducted at EARS. Beginning with the hexagonal shapes

of the library, one can create a path that traverses each room as shown in the figures

below
1
La biblioteca de Babel (Borges 1997).

113
 Figure 4.1: A contorted passage and path through the library.

This idea of an algorithmically designed path was the main inspiration that

led me to pursue research in sound spatialization. 2 Firstly, and to be more consistent

with the story, six speakers in an hexagonal array surrounding the listener were used.

The space could be virtually rotated and transformed in real time. I think of parametric

generated shapes and their permutations as a new form of working with space in music

and a way to connect sound spatialization to form thus breaking away from the aesthetics

of circular motion, at least from a conceptual point to view. The paths in Borges’ story

became sound trajectories in an ambisonics space and from there other curves were

created and new ideas were generated.

All the etudes were originally conceived from Borges’ stories and realized us-

ing the theory and technique of spatialization, envelopes, transformations in the spectral

domain and other computer music techniques. Laberinto 5 is an audiovisual algorithmic

composition with the strongest connection to Borges’ story; Laberinto I, for flute and

live electronics uses point cloud spatialization; Laberinto 3, for Clarinet and live elec-
2
The path is a metaphor for the trace left by a librarian in Borges’ story. The laberynthic geometry
of the hexagons and plausible paths are discussed by Bloch (Bloch 2008, 99).

114
 (a) A path (b) Ambisonics trajectory

Figure 4.2: A path applied to an ambisonics field

Figure 4.3: Some trajectories used in Laberintos’ ambisonics field.

tronics, uses envelopes and circular motion with delay lines to create an infinite spiral

of sounds; Laberinto 4, for piano an electronics, uses granular synthesis and algorithms

as a metaphor for the Minotaur and the Labyrinth in Borges’ The House of Asterión .

3 In Laberintos, labyrinths become paths and sound trajectories in which the listener

can discover his own labyrinth or a meaningful connection between movement and the

music.
3
La casa de Asterión.

115
 Figure 4.4: Laberinto V screenshot

4.1 Laberinto 5

Audiovisual composition. 5.1 Surround

Laberinto 5 (2014),4 the last of the pieces in the series Laberintos, is an audio-

visual algorithmic composition that include elements that reflect the beginning and the

last stages of the research conducted at EARS. The beginning, because it incorporates

the metaphor of the Labyrinth with its infinite permutations as imagined from Borges’

story. It is also a work written for 5.1 surround which is somehow a primitive form

of spatialization but developed with the algorithmic tools – computer music and 3D

software – I investigated during the last years.

In the story, the Library of Babel is an endless universe made up of small

hexagonal rooms. On the shelves of these rooms are books and all the books put

together contain every possible permutation of single characters –all the letters of

the alphabet and punctuation marks–. Most books are complete nonsense but some of

the books contain true and accurate descriptions of the universe, including everything
4
Laberinto 5. Stereo. http://vimeo.com/100868756.

116
that has ever happened and everything that will happen. Departing from the idea of

infinite permutations of characters, I used a random algorithm in Houdini to access a

list with Western characters. Those characters were also randomly put in a matrix for

visualization in a 3D space. Visual effects were used to transform each character into

particles representing the starts of the Universe.

The sounds in Laberinto 5 were created using the technique of granular synthe-

sis 5 with Csound, Max/MSP and mixed with ProTools. Firstly, phonemes were recorded

and converted to samples cutting unnecessary silence to make them all the same length,

and to optimize the files for granular synthesis. Using a random algorithm, sections of

each sample were randomly selected using parameters such as the sample index, size and

duration. The small “pre-grains” were then recorded and sent to Csound were granular

synthesis was applied.

One of the interesting aspects of the piece is the camera movement done in

Houdini. The letters only move forward and there is a slow motion rotation that is

more perceivable in the second section. The reason behind such unusual way to move

the camera without having objects moving from left to right –and vice versa–, as one

might see in a film, is to focus the audience’s attention on the movement of the sound

as the spatialization was carefully composed.

The piece is divided in two sections. The first section begins with a short

introduction where vertical stripes of black against a bright background move back and

forward in pendulum motion. The lines dissolve into recognizable characters which move

towards the viewer at a fast rate. The scene is saturated with indistinguishable sounds

which occupy all 5 channels. Although the spatialization is minimum and the visuals are
5
Granular synthesis is a technique that uses sounds that are between 10-100ms and windowed by a
function. See Roads Microsound (Roads 2004).

117
 Figure 4.5: Algorithmically created letters in Houdini.

not synchronized with the sound, the granular synthesis creates a strong connection with

the visuals as the stripes are composed of small particles. The contrasting second section

is dark and slow. Small particles move counterclockwise at a slow rate slowly revealing

as text characters. The movement is synchronized with the sounds and spatialization is

noticeable and surrounds the listener. The visuals accompany the sound and not vice

versa. Grains are of longer duration and although they change fast, one can understand

they are phonemes. As time move forward layers are added and the particles on the

screen reveal as letters as they become closer to the viewer. Towards the end, sounds

and movement –like in the first part–saturate the medium. Finally, sounds abruptly

stop and there is a short fading out of a single character that occupies the entire screen.

118
 Figure 4.6: Laberinto V Frame.

Figure 4.7: Images only move forward while sounds move in all directions.

119
Summary of techniques

• Sampling

• Granular Synthesis

• Stochastic algorithms for both images and sound using /Python

• 5.1 Surround

120
4.2 Laberinto I

for flute and live electronics. 8 channels The first piece of the series and a short

spatialization étude for flute or piccolo flute. It features an assortment of extended

techniques including multiphonics, whistle tones, key clicks, microtonal oscillations and

different sounds produced by variations in the position of the instrument’s embouchure.

The processing and spatialization is in real time requiring a live input from a microphone

to be connected to the sound interface. See Diagram 4.1.

Flute − > Mic − > ADC − > Sound Processing − >

− > Spatialization − > DAC − > Mixer − > Speaker Array (4.1)

A good quality microphone with a cardioid pattern and a frequency range of

20Hz-20KHz is preferred 6 to avoid noise from the performance space and feedback from

the speakers. Several patches were made using PD and Max/MSP to find which one

could do better pitch tracking. Best results were obtained in PD with the sigmund∼

object and Ircam’s yin∼ in Max/MSP.

yin∼ uses the YIN algorithm which is a fundamental frequency estimator for

speech and music based on the autocorrelation method 7 .

The pitch tracing object is used to split the signal in two according to a thresh-

old. Detected clicks and high pitch sounds are sent to a granular process (g) and lower

pitches to a harmonizer (h). If using a flute, frequencies below 500Hz – first octave –

are sent to the harmonizer. The idea is divide the spectrum and apply different synthe-

sis and spatialization techniques (p). An ambisonics field is created with the noise (n)
6
Such as the DPA 4011A. Cardioid microphones have an angle of rejection of 180◦ .
7
The autocorrelation method compares the signal to its shifted self. It applies a Fourier transform
of the power spectrum and can be seen as measuring the regular spacing of zcs within that spectrum.
(De Cheveigne 2002).

121
Figure 4.8: Two excerpts. The first shows micro-glissandi (long sounds); the second, short

sounds. These gestures are spatialized using rotation and point cloud respectively.

generated by the extended techniques. See Diagram 4.2. Both processes are sent to a

spectral delay before the spatialization.

ADC− > sigmund < (p, n) : (p : h <:

(long− > mjRotation , short− > mjRandom ),

n− > g) (4.2)

The spatialization can be done in real time using the hoa.mapl object or algorithmically

with the mj Library if working in PD. For this piece, I have chosen to use rotation for

longer sounds with micro glissandi, and point cloud spatialization for short sounds.

122
Martin Jaroszewicz

!
!
!
Laberinto I
for piccolo flute and interactive electronics

Laberinto I

!
LEGENDA - LEGEND !
!
! !
! !
! !
gradual accelerando
! ! key clicks!
!
!
!
! random/approximate key clicks!
!
!
!
!
! ! pizzicato!
!
!
!
!
!! pizzicato with key click!
!
!
!
!
!
! Microtonal oscillation (more or less fast)
! by embouchure rotation!
!
!
!
!
!
! Embouchure open, close, and tongue in
! mouthpiece!

123
 Laberinto I
for piccolo flute and interactive electronics
Martin Jaroszewicz

q = 40
ª
4 j √
‰ Œ Œ ‰ ¿ Œ Œ ¿ ‰ Œ ∑
&4 ¿
Jw w #¿
f w w

W O #œ nœ µœ n œ OO
œ µœ ® ‰ ‰ n¿ ‰ Œ
nO U
4

œ >#œ bœ
>. . . . . . . . ˜>. . . . . . . nO. O
& ΠΠ3
Jw >bœ nœ O 8
3
sfz p f mf
mp f

>œ œ˘œ >œ >œ œ >œ >œ >œ

œ™ œ ™™
e = 178
n>œ œ œ >œ >œ œ >œ
3 œ™ æ æ ææJ ææJ ææJ ææJ ææJ >œ œ œ
æ >æ #>ææ ææ
>>#> ææ ææ ææ > >
œ >
7

& 8 œ™ b ‰ ∑ ∑ r ≈R > >#>

Ϫ #> >3 #>
sfz
3
f sfz

ª R
œ ˘ µœ ™ n>œ µ>œ µ>œ µœ ˘ nœ Bœ n>œ œ >œ n>œ >œ >œ œ™ >œ™ ™ µœ ™ œ
µ
ææ æ ææ
19 3

& > >J b> > #> b> b > nœ #œ œ 3 TR

> > >>>> O

#O
5 5
6 9 3

W R W R W
31 ˘
n>
# b> > #> bO ™
j j j j #>
3
‰
3
& O
> #> > O > > > #>
bO ™
nO O O O O O
#O nO O O O O O O
p sfz ff

˘
#œ œ œ bœ n>œ O O O >O O ™
>O O #œ O O O œ bœ
& ææ æ æ ææ æ #æœ
nœ #œ #œ œ #œ
37
J J J J J J J J
æ 5
> >
3
3
f 6

#œ µœ n> > B œ œ œ œ b > n œ œ#œ nœ# œ

˜œ ≈ r n ¿
44
#œ œ J J œ
& ≈ #œJ ≈ n¿r ‰ J j R ≈ J J ≈ R #œ
3 3

n¿ #œ J R n ¿w ¿ n ¿ w
3 sfz sfz sfz
3
w w

√
æ æ
nœæ™ œ™æ
51
b¿ ¿
& ¿ w ∑ w ¿
w w #¿
w mf
fff

124
 3

æ æ bœ œ œ œ nœnœ ∑
œ™æ œ™æ
59
œ #œ
& œ
œ #œ #œ œ œ œ œ œ œ œ œ œ œ œ œ nœ#œ œnœ#œ œ#œ œ œ œ œ œ
fff mp fff

66 ˘ √ √
#œ œ ∑ ∑ #œ bœJ œ™ œ™ bœ œ œ œ#œ
& œnœœbœ œ#œœ#œœ nœnœ #œ#œnœ#œ œ œnœ œbœœ#œ nœ ™ œ™ œ™ œ™ œ J
5 6
sfz pp mf p

˘
œ œ œ™ œ™ >
b¿ b¿ œ ™ n>œ j >œ œ >œ œ >œ œ œ
78

& #œ œ ¿ b¿ #¿ J ¿ J J b¿ ¿j
J
‰ R J J J
mf 3
5 mf 9 p

nœ œ >œ >œ
3

86
œ >œ j >œ œ œ >œ œ œ œ œ >œ nœ œ œ
& ¿ ¿ J J J ≈ #¿ #¿ n¿ ¿ > #¿ ¿
3
#¿ n¿
3 5

œ˘œ œ œ œ œ œ œ™ œ
4
e=160
œ œ >œ œ œ # œ. œ. . . # œ. . . n œ. œ. .
œ œ œ œ™ œ œ œ œ™ œ œ œ œ œ ææ æR œ. œ. œ . œ œ. . œ œ. . œ . œ œ. .
93

≈ ≈ œ . œ œ . œ œ
& #œ #œ
6 6 6
fff p

. . œ™ B œ ™ # OO ™™ # OO ™™ # OO ™™ # OO ™™
œ#œ. >œ œ˘
œ. œ. œ. œ. œ œ. œ. œ œ. .
nœ˜œBœ B O ™ B O ™ B O ™ B O ™
ritardando e=140

œ. . œ. œ. #>œ >œ œ #>œ œ #˙ ™ µ˙ ™

103

æ ææ ææ ææ æ ææ ææ
.
& #œ #œ j‰
¿
w 3
f
fff

125
4.3 Laberinto III

for Bb clarinet and live electronics. 8 channels The piece is divided into two

sections: A, a meditation with long notes that resembles the playing of a shakuhachi

flute 8 ; B is a contrasting section with long chromatic lines of fast motion interlocked

with multiphonics and Flatterzunge. The processes are simple, there is a slow rotation

in the first section with each long note transposed at the micro level using envelopes con-

nected to a real-time pitch transposition module that uses a vocoder. The fast section is

a dense spiral of sound that moves upwards, an effect that is achieved by using spectral

delays and the coordinates of a spiral curve exported from Houdini 9 and spatialized

using ambisonics. The ideal format for this pieces is a high order ambisonics field with

a hemispherical dome or multilevel rings.

Figure 4.9: Envelopes used for micro-pitch modulation

8
A shakuhachi is a Japanese flute made of bamboo introduced from China in the 8th Century. It
was used by the Fuke monks of Zen Buddhism.
9
This can also be achieved by using mjRotation with an offset on the y axis driven by a phasor and
changing the radius of the circle as the piece progresses.

126
Figure 4.10: A spiral curve used in Laberinto III

127
!
Martin Jaroszewicz

!
!
!
!
!
!
!
!
Laberinto III
for clarinet in Bb and interactive electronics

128
 Laberinto III!
!
!
!
!
!
LEGENDA - LEGEND!
!
!
!
!
!
n.v.! ! ! ! non vibrato!
!
v.! ! ! ! vibrato!
!
! ! ! ! progressive transformation!
!
s.g.! ! ! ! slow glissando!
!
!
1!! ! 2! ! 3! ! 4!
!
!
!
1. three quarter flat!
2. quarter flat!
3. quarter sharp!
4. three quarter sharp!
!
!
!
Multiphonics (in parenthesis) are notated as sounding.
!
The piece starts with a mediation in the style of traditional Japanese Shakuhachi music. The durations are not strict
and more or less related to the breathing of the performer.
!
All the chromatic passages should be executed legatissimo.
!
The gestures in the second part of the piece marked floatingly, ethereal, are analyzed and transformed by the
electronics creating a series of spectral delays that follow the performance. Gradual accelerandi and decrescendi
and gradual changes in dynamics are recommended to create a sense of a timeless moving object.

129
 Laberinto III
Martin Jaroszewicz

q = 45 meditativo

n.v. v. n.v. v n.v v n.v v

, , ,
Bœ
U
Bœ n˙ ™ Bœ µ˙ ™
s.g s.g s.g s.g

& ˙™ nœ j bœ Ó bw
#œfi #œfi œ J
Cl. Sib j j
pp p

<n> œ
accel.

> œ #œ > œ #œ > œ #œ

Bœ nœ bœ ‰ <n>œ
7
U #œ >
<n>œ
& œœ bw œ œ œ
œ œ œ
3
mp

tempo giusto

œ , , ,
> œ œ > œ #œ n>œ ˜œ ˜w
U
10 s.g s.g

& #œ œ œ ∑ w j nœ œj #œ ‰
nœ œ œ #œ J

n.v v.

˜œ µœ
16 s.g. s.g s.g
œ
& #w nœ
œ#œ
#w
œbœ
#w µœ
#œ J ‰

q = 96

,
rit. come un eco scompariendo tempo giusto
22
> >œ >œ >œ 5
œ #œ
& œJ J J J œ
œ œ œ #œ bœ œ # œ nœ b œ
œ
œ œ
œ
œ
œ
œ
œ
œ œ
mp p pp ppp mf

tempo primo

24
,
‰
&
w µw nw œ
j #w µw œ #œ
ff ppp pp

Ϫ Ϫ Ϫ
30 pitch noise

& œ™ œ™ œ™ ∑
µœ µœ
j j j j j
œ #œ œ #œ #œ
pp
mf

130
2

q = 96 , ,
µw Y <n>
34 pitch noise

& w nw ‰ #œ œ #œ œ #œ
nœ œ œ
N mf p (audible) f mf f mf

n œ b œ nææœ w
œ <n>œ
,
bœ #œ æ
39

& #œ œ œ #œ œ nœ #œ bœ #œ
b œ
f mf f mf
mp mf

˙ w
æ w √
42

& #Ɯ
J
#œ
j
∑
bw
f fff

tempo primo , q = 96
45
5
∑ #œ Œ ‰ Ó nœ #œ ‰
& #œ. #œ. œ. .
œ œ#œ œ#œnœ œbœ œ #œ#œ w µw
˙ œ nœ#œbœ
mp p mf

nœ
53

nœ nœ œ #œ
& œ #œ œ œ #œ œ nœ œ
nœ nœ #œ

,
~~ Ÿ~~~~~~~~~~~~
accel.

#Ÿ˙ æ nœ Ó nœ ˙ U nœ. #œ. .

œæ ‰ #œnœ ™
54
#œ Œ #œnœ. œ œnœ #œ œ nœ#œ œ#œ
& #œ. #œ. œ. . ææ æ
mf æ #w œ nœ bœ
ff ppp mp
mp

æj nœ ™ œ œ ™#œ œ. #œ. #œ. . # œ œ. #œ

. ,
tempo giusto rit.

n œ #œ #œ.
œ.æ ‰ ææ ææ æ
60

& #œ#œ œ#œnœ#œnœ #œ J

n œ. œ. œ. œ. œ. œ. œ. œ. œ. œ. œ. œ. œ. œ. œ. œ. ˙æ
œ #œ
p f mp mp pp

œ ™#œ. nœ. #œ.

#œ √
œ. œ. œ. œ. œ # œœ
ææ Œ œ Œ
bw œ œ. œ. œ. œ.
U
tempo giusto
w
w æ
64
.
æ œœœnœ#œnœ . Œ
7

& #œ. œ
bw b œ. œ. œ. œ. µœ . . . . . . . .
œ œ œ œ œ œ œ œ ˙
mf mf p pp
f f ffff ff

131
 3
Floatingly, ethereal

œ bœ
70 q = 86
bœ œ œ.
& œ nœ bœ nœ œ œ œ bœ nœ bœ œ œ bœ nœ œ nœ œ œ nœ œ nœ bœ nœ Œ 41
nœ
5 7 7 6
mp mf mp

bœ n œ. b œ.
œ #œ œ #œ nœ nœ bœ œ bœ œ .
œ nœ œ nœ œ #œ œ bœ. œ.
72

& 41 nœ œ
7 6 5
6 mf

nœ œ bœ #œ œ nœ œ œ nœ œ nœ nœ
74

nœ bœ nœ bœ œ nœ bœ œ
&
6 7 7

5 nœ bœ œ nœ bœ œ bœ
#œ nœ #œ
75
œ bœ œ
& bœ nœ œ œ nœ
7
6
f mf

,
76
nœ bœ nœ œ #œ œ bœ œ œ œ nœ b œ n œ bœ #œ n œ œ œ nw
w
& #w
7 7
fff
mp

, U
6
79
#œ n œ œ bœ œ bœ nœ bœ bœ œ nœ # œ œ bœ œ b œ n œ n œ ##w
w
& œ nœ bœ nœ nœ #w
5 6 ff
6

mf f
(as loud as possible)

132
4

accel.

nœ #œ bœ #œ # œ
7
#œ nœ nœ
82

œ nœ œ bœ œ bœ bœ nœ œ bœ nœ bœ nœ #œ œ œ
& œ nœ #œ œ bœ nœ
5 f 7
mf 6 5

æ
n œ nœ #œ
accel. giusto

nwæ
84
nœ œ #œ nœ #œ nœ #œ œ nœ #œ nœ œ bœ nœ bœ nœ
7 7

& #œ
6 mf

accel.

æ
7

#wæ
bœ nœ bœ nœ nœ #œ œ œ nœ œ bœ œ bœ œ
86

& œ
bœ
#œ œ bœ nœ bœ nœ œ
7 7
f p fff

bœ nœ bœ nœ nœ #œ œ œ nœ œ bœ œ bœ nœ œ nœ œ œ bœ œ œ
88 3

& œ #œ œ #œ nœ #œ
3

7 7 7
mf mp mf mp

æ 3æ æ æ3 æ ææ æ æ æ æ æ
giusto

ææ œæ ænœæœææ æbœæœ æœæ ænœæ#œæ bœ nœ

3

œnœnœbœnœ œæ˙æ
3
90
bœnœbœœbœœœbœœbœ
& bœnœœbœnœ
3 3
5
#œ bœ nœ nœ œ œ bœ œ# œ n œ bœn œ œ #œ
7 p
8
mf mp pp mp

flessibile accel.

æ æ
& œæ æ nœææ ææ nœ bœ nœ nœ#œ bœ nœæ æ ææ ææ ææ
flessibile
94

ææ
6
7 7 3

#œ #œ œ b œ n œ n œ bœ nœbœ nœbœ œbœ œ œ b œ n œ b œ nœ n œ w

p ff ppp mp p ff

œ # œ n œ œ # œ n œ #œ
accel.
97
7 7
&
œ œ #œ œ #œ œ #œ nœ œ nœ œ œ #œ œ #œ œ œ #œ 7

ppp mf

flessibile accel.
98
nœ #œ nœ œ bœ œ bœ nœ œ nœ œ #œ
nœ bœ nœ nœ #œ #œ nœ #œ nœ nœ #œ nœ
7
&
5 7
5
f

99
7
& #œ nœ
nœ #œ nœ #œ nœ #œ œ bœ nœ bœ œ œ nœ œ.
fff

133
4.4 Laberinto IV

for piano and live electronics. Stereo Laberinto IV is based on the piece Laber-

into Borgiano inspired by Borges short story La casa de Asterión 10 and the formation

of globular clusters in the universe.Laberinto Borgiano is a metaphor for a unicursal

labyrinth, a type of Greek labyrinth that hold the Minotaur and a metaphor for the

creation of the universe and the composer’s creative labyrinths. These two elements:

the man and the bull, are developed and expanded. The first one is gradually trans-

formed and leads into the B section of the work where the sense of a path is lost. The

second short motif (the entrance to the labyrinth) becomes the labyrinth itself which

reveals itself at the last section of the piece gradually moving towards the center of the

labyrinth where the universe ends.

The piece was constructed through the concept of duality which is found at

every level in its form. The first two measures contain the material that is developed

in different sections of the work. At the immediate surface level, there are two motives

that are expanded and transformed into sections respectively. There is an introduction,

where the material is presented, and there are short transitions to the main sections of

the piece. The concept of duality is furthered explored by trying to create independence

of the hands and the feeling of more than one piano, or two hands, being heard.

Laberinto borgiano is a work that constantly juxtaposes two different ideas

without trying to create a sense of linear counterpoint or polyphony. The idea of duality

is present throughout the work and is presented at the beginning with two different

motifs that become the pillars of the work.

10
Asterión is the mythological creature that was half man and half bull. According to Greek mythology
the name Asterión means “ruler of the stars”or “starry”.

134
 Figure 4.11: First measure of Laberinto Borgiano

Figure 4.12: Main motifs in Laberinto Borgiano

There are four main motifs (Figure 4.12). The first two, M1 and M2 are

developed in each section of the work respectively. M3 gradually becomes and important

element in A at the end of the section. The addition of M4 in A creates the material for

the transition between A and B. M1 is deconstructed at the end of the work intersecting

with M2 chord come un’eco lontano.

In Laberinto IV another element was added to incorporate the electronics in a

meaningful manner. A short introduction based on the beginning of Beethoven’s piano

135
 Figure 4.13: Time envelopes.

sonata Op. 100, III movement, alludes to the bebung technique .11 The electronics

extend the piano and play upon the notions of vibrato and the use of repeated notes

and/or slurs.

For each section different time envelopes were used to mix the sound of the

piano and the electronics. Combining pitch tracking and time envelopes, a cheap form

of “score following” was implemented. The performer also controls time by manually

advancing sections – and triggering the envelopes – with a ”user friendly” interface.

There are three basic envelopes that trigger eight sections of the piece. The

first one has an exponential value of 1.09, the second one and exponential value of 1.036

and the last one, a linear envelope –with a value of 1.0–as shown in Figure 4.13. The

sections trigger different settings for a granular synthesizer with a smooth overlap and 12

time stretching/freeze capabilities. The signal is divide into two different streams at

middle C with a crossover filter at 261 Hz. This allows for processing the left and right

hands separately, especially in the last section of the piece where the resonances of high

pitches are extended by a “virtual tremolo”.

11
A type of vibrato executed on the clavichord by applying force to a depressed key. Although not
often written, performers apply the technique as ornamentation.
12
IRCAM’s sogs∼ .

136
Martin Jaroszewicz

!
!
!
!
!
Laberinto IV
for piano and interactive electronics
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!

137
Requirements and Setup
!
!
• Accompanying standalone Max/MSP patch. The patch
requires the sogs granulator from IRCAM’s SoundBox
• Two condenser microphones and cables
• An audio interface with at least two inputs/
outputs. Direct monitoring through outputs must be
avoided.
• Mixer
• Two speakers / PA system
• The Max/MPS patch contains eight buttons the
performer has to click as notated in the score.
The buttons’ numbers turn green when the mouse is
over for better visibility and precision.
• Balanced should be achieved between the sound of
the piano and the electronics.

Mixer

Computer/Max

Audio Interface

Mic Mic

Piano

Speaker Speaker

Audience

138
!
• Double click the adc~ object to configure inputs
and outputs.

139
 Laberinto IV
U
Martin Jaroszewicz
q = 50

44 1 ¿

{
Ped.
sff (pigiare il pedale con tutta la forza)

œ œ ™ œ œ nœ nœ œ œbœ œ œ œbœ œ U
2 A
b nnnnn
Recitativo 6 più adagio
b
&b b b nœ œ ˙
œ œ œ
5 nœ

? bb b
b b œ nœ œ œ œ nnnnn
?

{
™ ™
2
œ œ œœ œœ œœœœœœœœ er œJ œ œ œ œ œR œRÔ œ œ œ œ œ œ œ œ œ œ œ œ
4

& #œ
pp

? n##w
w
w
w
6

œ #œ
œ

{
Ped.

6
œ œ œ œ œ œ œ œ#œ œ œ œ œ œ œ œ
& ≈ ≈ ≈ J
#œ œ
? Œ ‰™
nœ nœ nœ œœ ™™

{
bœ œ œ œ œ œ œ œ œ œ œ œ œ œ œ œ œ œ œ œ œ œ œ œ œ œ œ œ œ œ
7 be
R
&

? nww
*) The use of dotted barlines is evident at rehearsal C. The performer can repeat and randomize this section
ad libitum keeping the transition to section D connected. Pitches in measure 17 can be altered on every repetition.

140
2

“”
B

{
bœ √
*) 9
8
œ œ
& #œ #œ nœ #œ #nœœ #œœœ #œœœ #˙˙˙
?
mp
∑
Ped.

{
bb œœœ ˙˙˙
9 nœ ™ 3
œ bœ ˙
œ
&

{
pp come un'eco lontano
? ∑ &
U
Ped.

##n ˙˙˙˙ 3
œ
#œ bœœ œ nœ
10 9
œœ œœ œœ ˙˙ Ó
& œœ˙
pp

#w
mp
& Œ #œœ #œ Œ Œ b>œ Œ w
b œœ >
mp p
? œ#œ œ nœ
ø ø
œ œ #œ œ œ# œn œ
“‘

{
œ w
ø ø
1/2 (Ped.)

“” >œ n œ œ #œ
œ œ œ nœnœ œ#œ nœ nœ#œ œ
12

& #œ nœ nœ nbœœ œœ 42 ∑ 42
œ # œ # n œœœ œœœ
mp
mf
4
f

& ‰ #œ 3œ nœ nœ #œ Œ ? ‰ 42 ∑ 42
5
#œ œ nœ œ # œ n œ œ
1/2 (Ped.) ø “‘
ppp ff

141
 3

{
C q = 165
q=120 poco accel.
3 3
# œœ
Meccanico e secco

? 42 bœ œ œ
5
j œ
14

nœ j ‰ #œ #œ nœ ‰ ‰ & nœ>œ
3
?
#œ œ œ nbœœ # œœ
f ff

?2 Π3
‰ j‰ ‰ ‰ ‰ j ‰
5
Ó œœ œœ
martellato
4
œ #œ
salvo indicazione “‘
senza pedale Ped.

{
b3œ bœ œ 5 œ #œ
18 q=120
? ‰ ‰ œ œ ‰ J ‰& Œ ≈ nœ bœ
œ‰ ≈
nœ œ #œ
f
? Œ Œ
œ œ#œnœ œ j ‰ ∑ ‰
œ œ#œnœ
‰ ‰
“‘ “‘ “‘
œ

{
œ 3 #œ
‰ j ‰ œj ‰ j
22
3 3
≈≈ nœ J ‰ ‰ Œ
3 3

nœb œ œ œ bœ
bœ #œnœ nœ
nœ # œ œ œ #œ nœ
? ‰ ‰ j ‰ ‰ bœ ‰ ‰ nœ ≈ ‰ j #œ
3

J nœ
œ‘ nœ #œ
“

{
n>œ3 œ 3
#œ ‰ J ‰ J ‰ bœ ‰ ™ œ œ bœ œ#œ œ&
26 3
‰ R ≈ œ ‰ #œr ≈ ≈ j #œ
#œ ™
3 3 3

#œ œ J J
? #œ ‰ j ‰ ‰ bœJ ‰ ‰ Œ
3
‰ ‰ œ bœ j #œ
œ œ J

{
œ
>œ >œ3
#œb œ œ b œ œ œ#œ
30 5

œ œ ≈‰ ≈J ≈ Œ ‰J ‰ ‰ bœJ ‰ ≈
3

& ≈#œ ‰

j ‰ bœ œ œ # œ #œ œ# œ n œ nœbœ œJ nœ bœ œ nœ 3œ #œ
3 5

? nœJ ≈ ‰
3 3
‰ œ

142
4

{
b3œ œ œ
#œ ‰#œJ ‰ ‰ 3 j ‰ ≈ J ≈ ≈ œ#œnœnœbœ ≈ R
34
œ ≈ ‰ J ‰ ‰
3

& ≈J J bœ
b œ œ # œ5 n œ nœ
? œ #œ 3
j
3
j #œ nœ #œ
3

J& nœ #œ nœ
J œ # œ œ œbœnœ œ œ b œ œ

{
#œ nœ#œnœ œ œ#œnœ bœ œ#œ nœ
38
R ≈ ≈ nœR œR ≈ ≈ J > >
& ≈ ≈ ≈bœR œJ ‰ ≈ ≈ œJ ≈

j œ bœ œ #œ bœ nœbœ œ
3

& œ bœ œ bœj #œnœ nœ œ œ #œ œ #œ #œ nœ

3 3 3

{
gradualmente legatissimo e leggiero

D
42 >œ œ b >œ >œ
& ≈ #œ œ ‰ J J J
3

œ nœ bœ œ #3œ nœ nœ #œ œ n3œ bœ œ 3œ bœ œ #3œ nœ #œ n3œ œ

& &

{
44
œ #>œ œ n>œ ™
& J ‰ ‰ J J ‰ ≈ J

nœ #œ nœ nœ #œ nœ #œ #3œ nœ bœ #3œ œ œ #œ nœ œ
3 3 3 3 3

#œ nœ nœ #œ nœ #œ nœ #œ &
3
&

{
#>œ ™ œ >
≈ nœJ œ
46

& Œ ≈ J R J ‰

3 3 3 3 3

& nœ œ œ #œ nœ #œ #œ nœ #œ nœ œ nœ œ bœ œ #œ nœ œ œ #œ œ œ # œ
3 3 3

nœ

143
 5

{
n>œ ™ b>œ
48
œ #œœ
& ≈ J
3 ¥al fine >œ
3
? #œ nœ nœ nœ b3œ œ #œ nœ3
3
3 3 3

& #œ nœ #œ nœ œ bœ œ nœ bœ œ #œ
#œ nœ œ bœ œ
ø ø

{
“”
& <#>œœœ
50

Œ #˙
˙
? 5 5 5 5
?
nœ #œ nœ n œ b œ œ # œ n œ œ #œ nœ #œ nœ n œ b œ
n œ b œ “‘ œ #œ nœ
ø ø
√

{
##n ˙˙˙˙ 5
E
52 q = 50

& Ó
44

√
ff mp
? 44
nnnn ˙˙˙˙˙ ˙˙˙˙˙
<“> ˙˙ ˙˙

{
Ped.

44
54 q = 128
4
&4 ∑ ∑
sim
3 3 3 3 3 3 3 3

?4
4 bœœ œœ
œœ n œœ n œœ œœ
# œœ Ped. œœ >
ff
> sim *
Ped. alzare il pedale gradualmente

144
6

{
56 q = 50
˙ ™™ √
w
6
b˙˙ ™
& ∑ ∑ w
w
√
p
?
nww ww ww ww
nw
ww
ww w
ww
ww w
ww
ww w
ww
ww
ffff

“”
Ped.

{
nœ
b œ nœ n œ √
q = 50 9
60

& Œ #œ #œ nœ Ó
œ bœ
mp
? ∑

{
Ped.

q = 128
61 3 3 sim 3 3 sim
? >
bœœœ bœœœ bœœœ bœœœ bœœœ
œ œ œ œ œ
?
œ œœ œœ œ œœ
# œœœ œ œœ # œœœ œ
>œ >œ

{
sim
*
√
sim
Ped. alzare il pedale gradualmente
n˙ w
63 q=50 # ˙˙ w
w
6 7
& ∑ ∑ Ó
√
#˙˙˙ w
p

∑ ∑ Ó w
w ?
& &
?
nw
ww w
ww w
ww w
ww
ww
ww ww
ww ww
ww ww
ww
ffff
Ped.

145
 7

{
F
(q = 50)
œ™ 3n œ #œ #˙ bœ
67
œ nœ œ
& Œ ‰ œ #œ #œ J ‰
bœ
p
mp
?
w
ww w
ww
ww
ww ww
ww

{
“”
8 nœ œ œ œ
3

bœ R ≈ ‰ œ œ
69 3 3
b œ
3

& Ó Œ ≈ nœ #œ nœ ≈ bœ œ œ # œ bœ #œ
mp sempre legatiss. sempre, quasi sempre Ped.
? ∑ ∑

{
<“>
nœ
b3œ nœ 3 n œ œ 3 n œ n œ #œ œ œ 3 #œ œ n œ œ3
5

nœ #œ nœ
71
r
& nœ œ #œ nœ bœ ≈nœ œ bœ œ #œ
œ# œ œ nœ œ b œ œ

? ∑ ∑ ?

{
“”
œ b œ n œ œ“œ”
5
3 nœ nœ
73
nœ
nœ ≈ R bœ œnœbœ ≈nœ ≈ œ
5

& ≈ #œ# œ nœ ≈#œbœnœ ≈

3
#œ œ
nœ

? nœ #œ ∑ nœ #œ nœ #œ ∑

146
8

{
”
n“œ
#œ œ
3
œ œ œ
75 3
œ œ #œ
5
œ
3
nœ
≈ bœ nœ
3
& œ bœ n œ #œ nœ nœ bœ

? ∑

{
nœ nœ
5
nœ
3 nœ

≈ nœ œ œ #œ
76
bœ
5

&
3
#3œ bœ
bœ nœ # œ n œ n œ œ
? #œ œ #œ

{
n œ3 bœ 5
œ
77 b œ n œ œ 5œ nœ
bœ œ
& nœ
3
3

? nœ #œ nœ nœ bœ
leggierissimo cresc
nœ nœ #œ œ œ

{
“” “”
“œ”
5

#œ 3
œ 5 œ
78 #œ œ œ œ
#œ nœ œ nœ
& bœ

? œb œ
œ œ #œ

147
 9

{
“”
poco accel

œ #œ nœ nœ
5
œ
5

nœ
3
œ
nœ
79

& nœ nœ #œ ≈ ≈
R

?
decresc
œ
œ #œ œ bœ #œ

“”

{
“œ” bœ
80 tempo
J ‰ Œ √
& Ó Ó ‰ ‰

? √ nœ œ
9

œ œ # œ n œ œr
≈‰ Œ bœ bœ œ Ó
œ nœ bœb œ
≈≈≈
“‘ #œ

{
◊ÿ
82 nœ œ U
& ‰ ≈ J ≈ ‰ Œ ∑

bœ U
9
? Ó Œ nœ œ ∑
n œ œ b œ nœ #œ #œ
p mf p
Ped.

◊ÿ ◊ÿ
# w√
ritardando come un'eco lontano

84 ˙ ˙U
˙ w
w
&
ppp

148
4.5 9

electroacoustic. Quadraphonic This work uses the technique of micro montage

with pseudo random algorithms and composed spatialization. In the context of computer

and electroacoustic music, a sound collage or montage is a technique where a work is

made by combining a collection of sound objects in a pseudo random manner. Each

object and morphology contributes to the whole resulting in a work with its own unique

identity. The end result is intrinsically related to the time scale of each individual

object. As the time scale gets smaller, the whole acquires a stronger morphological

identity as perception in the micro sound 13 scale –less than 100ms–. This does not leave

enough time for the listener to aurally discriminate sound morphologies and envelopes,

especially sounds in which the attack has been masked by a window as in granular

synthesis. If the time scale varies or fluctuates from the microtime domain to a bigger

scale, microsounds can become a portion of an envelope of a sound that will be created

and become an entity of the multiscale system escaping from the microstructure. For

the work to be considered a montage or micro-montage, the composer must not create

these macro morphologies intentionally.

My approach to creating a macro structure with microsounds is using a pseudo

random sequence to preserve what we might expect from a random selection of objects.

14 The production of microsounds is, in my compositional work 9, based on pseudo

random operations on a single sound object containing rich and varied material. This

object is accessed randomly at different points in time providing the micro material

for building the macrostructure. More than 5000 “particles” were generated with this
13
For a discussion on time scales and granular synthesis see Roads (Roads 2004).
14
A quasi random algorithm will be more appropriate for this purpose than a “random” or “pseudo-
random” one, which has the tendency to clump objects together instead of being scattered somehow
uniformly.

149
 Figure 4.14: A random or ”pseudo random” distribution in PD

technique for later construction of the montage. The use of particles in granulation

differs from a montage in many respects, the most important of which is the process by

which these particles become part of the work. In granulation, particles are generated

by a computer as part of an automated process, while a micro-montage is the work of

a composer “painting” as in the works of pointillists such as the French artist Seurat

(Figure 4.15). The composer not only adds particles, he also adds silence, which could be

compared to the non-intentional white space between the dots on the canvas which seems

to be a by product of the painter’s technique. The distance between sound particles is

then carefully planned as an integral part of the montage.

The durations in 9 are very short, and silences of different lengths were ran-

domly introduced to avoid saturation of density; other elements pertaining to a different

time scale were also added. These layers that are not part of the random micro-structure

add contrast to the work creating two independent macro-structures that are balanced

with carefully composed quadraphonic spatialization. The addition of a multidimen-

sional sound space to a somewhat linear composition, a characteristic of sound collages,

creates a sense of direction that is perceivable by the listener as a formal structure cre-

150
Figure 4.15: Georges Seurat - ”Circus Sideshow” (1889) - detail showing pointillism technique.

Public domain image.

ating “something to hold on to” (Landy 2007). Silence creates the time scale for the

microsounds to move in the quadraphonic space.

Spatialization at the microsound level has a powerful impact on the listener.

As sounds move in a swarm-like pattern, similar to that of insects, the listener begins

to perceive them in an almost tangible manner. Microsounds become dense creating

a spatial effect in a process that can be thought of as forming a cloud in space. This

manifests a vivid three-dimensional spatial morphology that can be also perceived in

the stereo reduction of the work.15

15
The stereo version of the piece was mixed creating a virtual quadraphonic stage with four speakers
panned in front of the listener at 30,60,90,120 degrees

151
Chapter 5

Conclusion

In this thesis, I presented spectral spatialization, its background, applications

and implementations. In addition, several musical examples that summarize many of

the techniques discussed were analyzed. Most importantly, we looked at composition

in space and the use of technology as a creative tool. Developed by the author to find

solutions to his own compositional strategies, these parametric and algorithmic tools can

help the composer of contemporary music to focus on the composition of sound spectra

as it changes over time.

As part of the research conducted at EARS, and given the portability and

flexibility of the studio, I took advantage of every possible configuration of loudspeakers

allowing me to experiment with different spatialization techniques. This experimental

approach was essential to the development of the collection of works Laberintos and their

metaphors. As the apparatus becomes part of the composition in many different ways,

a studio that is more flexible opens an opportunity for the composer to develop new

musical ideas, overcoming the limitations of the past and finding new ones that in turn

will be part of the techné that will shape future interactions. The mj Library for PD

152
is a small contribution to the integration of spatialization in the compositional process,

but it triggered new ideas for future work and the way the composer can interact with

sound trajectories. The creation of a tool for spatialization that can work in tandem with

synthesis techniques allowed me to reflect on what it means to work with sound–objects

that can transform their spectral content as they move along a composed path. The

research conducted at EARS suggested that a new understanding of the sound-object is

needed, one that considers new philosophical ideas that are current and can contribute

to the understanding of new technological developments as they shape the work of the

composer.

5.1 Contribution

The contributions of the research conducted at EARS include the following:

• A series of objects for the Pure Data programming language to aid the composer

with the design of parametric curves and random algorithmically driven paths.

• A command-line application that converts points from Houdini to coordinates that

can be used in Max/MSP and PD.

• Different approaches to cross-synthesis – implemented in PD and Max/MSP – that

can be synchronized with spatialization curves.

• A series of études that range from stereo to 8 channels using ambisonics and dif-

ferent computer music techniques including granular synthesis in the time domain

and several techniques in the spectral domain.

Much of what I have discussed rests on the different compositional approaches

to spatialization in music, sound imaging, sound fields and the aesthetics of circular

motion. Most systems share a fundamental limitation: they are determined by the

153
location of the listener, with some systems relying more on the sweet spot. There is

a need of more tools to deal with movement in space as most libraries only provide

encoding and decoding with few or no composition tools. The Game of Life is the only

system that allows the composer to “compose” the spatialization along with the music

but is still limited to basic curves and hand drawn paths.

I proposed a system that can be compatible with current 3D software such as

Blender where any shape can be created and exported as coordinates. I also proposed the

use of algorithmically created curves that can be synchronize to a cross-synthesis system

for a spectral spatialization approach that is more meaningful in terms of composition

and form. Spectral spatialization by changing spectra as sound moves is the correct

approach to add meaning and clarity to electroacoustic music, whereas the creation of

sound fields and diffusion creates an immersive experience that is more related to the

space than the work. While spatialization focuses on the movement of sound objects,

spectral spatialization focuses on the object as part of the work, that is, it focuses

on form, thus movement becoming a structural part of it. There is a need for an

integrated system that can allow the composer to work with spatialization in tandem

with morphing.

The contorted passages and the hexagonal shapes in Borges’ La biblioteca de

Babel and the labyrinth in La casa de Asterión were the sources of inspiration to create

a series of works where spatialization played an important role in defining form.

154
5.2 Future Work

At the time of this writing, a small contribution to the groundwork for specify-

ing a framework for composers to work with spatialization in a meaningful manner was

given. Tools for the – open source – Blender 3D software are being developed to be able

to communicate in real time using the OSC protocol. Composing with the space in a

intuitive manner or by using algorithms can benefit from a “user-frienldy” 3D interface

where the composer can transform non-musical ideas into meaningful spatialization.

155
 Bibliography

Aristotle. 2001. Problems. Translated by R. Mayhew. Cambridge, MA: Harvard

University Press.
Barrett, Natasha. 2010. “Ambisonics Spatialisation and Spatial Ontology in an Acousmatic
Context.” Paper presented at the Sixth Electroacoustic Music Studies Network
Conference. http://www.ems-network.org/ems10/ (accessed February 12, 2015).
Baumann, Dorothea. 2011. Music and Space: A Systematic and Historical Investigation
Into the Impact of Architectural Acoustics on Performance Practice Followed by a Study
of Handel’s Messiah. Bern: Peter Lang, AG Internationaler Verlag der Wissenschaften.
Berkhout, A.J, D. de Vries, and P. Vogel. 1993. “Acoustic Control by Wave Field
Synthesis.” Acoustic Society of America 93: 2764-78.
Berlioz, Hector. 1994. The Art of Music and Other Essays. Edited by Elizabeth
Csicsery-Ronay. Bloomington, IN: Indiana University Press.
Bloch, William. 2008. The Unimaginable Mathematics of Borges’ Library of Babel.
Oxford, NY: Oxford University Press.
Bole, Tiziano. 2008. “Faust Tutorial 2”. Grame. http://faust.grame.fr/ (accessed
February 12, 2015).
Borges, Jorge Luis. 1997. Ficciones. Madrid: Alianza Editorial.
Boulez, Pierre. 1986. Orientations: Collected Writings. London: Faber.
Boulez, Pierre and Andrew Gerzso. 1988. “Computers in Music.” Scientific American 258.
http://articles.ircam.fr/textes/Boulez88c/ (accessed February 12, 2015).
Brant, Henry. 1967."Space as an Essential Aspect of Musical Composition".
InContemporary Composers on Contemporary Music, edited by Elliott Schwartz,
Barney Childs, and Jim Fox, 221–42. New York, NY: Da Capo Press.
Brant, Henry. 1977. Antiphony One: For Symphony Orchestra. New York, NY: Fischer.
Chagas, Paulo. 2014. Unsayable Music: Six Reflections on Musical Semiotics,
Electroacoustic and Digital Music. Leuven: Leuven University Press.
Chagas, Paulo. 2008. “Composition in Circular Sound Space: Migration 12-channel
electronic music 1995-97”. Organised Sound 13 (3): 189-98.
De Cheveigne, Alain, and Hideki Kawahara. 2002. “YIN, a Fundamental Frequency
Estimator for Speech and Music.” The Journal of the Acoustical Society of America
111: 1917-30.
Clozier, Christian. 2001. “The Gmebaphone Concept and the Cyberphone Instrument.”
Computer Music Journal 25: 81-90.

156
Daniel, Jérome. 2001. “Représentation de champs acoustiques, application à la
transmission et à la reproduction de scènes sonores complexes dans un contexte
multimédia.” PhD dissertation. Paris: Université Paris 6.
Declercq, Nico F. 2007. “Acoustic Diffraction Effects at the Hellenistic Amphitheater of
Epidaurus: Seat Rows Responsible for the Marvelous Acoustics.” The Journal of the
Acoustical Society of America 121: 2011-22.
Decroupet, Pascal, Elena Ungeheuer and Jerome Kohl. 1998. “Through the Sensory
Looking-Glass: The Aesthetic and Serial Foundations of Gesang der Jünglinge.”
Perspectives of New Music 36: 97-142.
Van Dorp Schuitman, J. 2005. “Wave Field Synthesis Using Multi Actuator Panel
Loudspeakers: Design and Application of Various Digital Filtering Algorithms.” Phd
dissertation. Delft: Delft University of Technology.
Emmerson, Simon. 2007. Living Electronic Music. Hampshire: Ashgate Publishing
Limited.
Eno, Brian. 2004. Audio Culture: Readings in Modern Music. Edited by Dan Warner.
New York, NY: Continuum.
Farnell, Andy. 2010. Designing Sound. Cambridge, MA: MIT Press.
Fletcher, Stephen. 1992. A Biographical Memory. Washington, D.C.: National Academy
of Sciences.
Garcia, Tristan. 2014. Form and Object. Edinburg: Edinburg University Press.
Gerzon, Michael. 1974. “What’s Wrong With Quadraphonics?” Studio Sound.
http://audiosignal.co.uk/Resources (accessed February 12, 2015).
Harman, Graham. 2011. “Realism Without Materialism.” SubStance 40: 52-72.
Holman, Tomlinson. 2008. Surround Sound: Up and Running. Boston, MA: Focal Press.
Ives, Charles. 1908. The Unanswered Question. New York, NY: Southern Music
Publishing Co. Inc.
Jacquemin, Guillaume, Thierry Coduys and Matthieu Ranc. 2012. “IANNIX 0.8”. Paper
presented at des Journées d’Informatique Musicale (JIM 2012), Mons, Belgium, 9-11
May, 2012. http://jim.afim-asso.org/jim12/ (accessed February 12, 2015).
Jaroszewicz, Martin. 2013. “Pierre Boulez’s Répons: A Virtual Cathedral”. Paper
presented for PhD qualifying exam. University of California Riverside.
Landy, Leigh. 2007. Understanding the Art of Sound Organization. Cambridge, MA: MIT
Press.
Manning, Peter. 2006. “The Significance of Techné in Understanding the Art and Practice
of Electroacoustic Composition.” Organised Sound 11 (1): 81-90.

157
Miller, Paul. 2009. “Stockhausen and the Serial shaping of Space”. PhD diss. Rochester,
NY: Eastman School of Music, University of Rochester.
Morton, Timothy. 2013. Hyperobjects: Philosophy and Ecology after the End of the World.
Minneapolis, MN: University of Minnesota Press.
Murray, James and William van Ryper. 1996. Encyclopedia of Graphics File Formats.
Bonn: O’Reilly and Associates.
Pérez-López, Andrés. 2014. “Real-Time Audio Spatialization Tools for Interactive
Performance.” Master thesis, Universitat Pompeu Fabra, Barcelona.
http://mtg.upf.edu/ (accessed February 12, 2015).
Pollio, Vitruvius. 2008. Los diez libros de arquitectura. Barcelona: Linkgua Ediciones.
Puckette, Miller. 2007. The Theory and Technique of Electronic Music. Hackensack, NJ:
World Scientific Publishing Company.
Rabenstein, R., S. Spors and P. Steffen. 2006. “Wave Field Synthesis Techniques for
Spatial Sound Reproduction.” In Topics in Acoustic Echo and Noise Control, 517-45.
Berlin: Springer.
Ramirez, Robert. 1985. The FFT: Fundamentals and Concepts. Englewood Cliffs, NJ:
Prentice-Hall.
Roads, Curtis. 2004. Microsound. Cambridge, MA: MIT Press.
Sascha Spors, Rudolf Rabenstain and Jens Ahrens. 2008. “The Theory of Wave Field
Synthesis Revisited”. Presented at the 124th Convention of Audio Engineering Society
http://www.aes.org/events/124/ (accessed February 12, 2015).
Schaeffer, Pierre. 2003. Tratado de los objetos musicales. Translated and edited by A. C.
de Diego. Madrid: Alianza Editorial.
Scipio, Agostino Di. 1995a. “Inseparable Models of Materials and of Musical Design in
Electroacoustic and Computer Music.” Journal of New Music Research 24 (1): 34-50.
Sèdes Anne, Pierre Guillot and Eliott Paris. 2014. “The HOA Library, Reviews and
Prospects.” Paper presented at ICMC|SMC|2014. http://www.icmc14-smc14.net/
(accessed February 12, 2015).
Shankland, Robert S. 1972. “The Development of Architectural Acoustics: The
Application of the Science of Acoustics to Architectural Designs Has Produced Greatly
Improved Halls for Both Speech and Music.” American Scientist 60: 201-09.
http://www.jstor.org/stable/27843021 (accessed February 12, 2015).
Slonimmsky, Nicolas. 1998. Webster’s New World Dictionary of Music. Edited by R.
Kassel. New york, NY: Macmillan.

158
Smalley, John. 2000. “Gesang der Jünglinge: History and Analysis.” Columbia University
Department of Music.
http://www.music.columbia.edu/masterpieces/notes/stockhausen (accessed
February 12, 2015).
Smith III, Julius O. 2010. “Physical Audio Signal Processing: For Virtual Musical
Instruments and Audio Effects. JOS On-Line Publications, Stanford University.
https://ccrma.stanford.edu/˜jos/pubs.html (accessed February 12, 2015).
Snoei, Wouter. 2014. “Wave Field Synthesis”. Personal Powerpoint.
Snoei, Wouter and Martin Jaroszewicz. 2014b. Personal conversation.
Spatialisateur. 2012. User Manual. Paris, France: IRCAM.
http://forumnet.ircam.fr/product/spat/ (accessed February 12, 2015).
Steinitz, Richard. 2003. Gyorgy Ligeti: Music of the Imagination. Boston, MA:
Northeastern University Press.
Stockhausen, Karlheinz. 1958. “Musik im Raum.” Die Reihe 5: 152-175.
Weisstein, Eric. 2014. “Rose.” Wolfram Mathworld.
http://mathworld.wolfram.com/Rose.html (accessed February 12, 2015).
Xenakis, Iannis. 1992. Formalized Music: Thought and Mathematics in Composition.
Hillsdale, NY: Pendragon Press.

159
 .1 Appendix A - mj Library Source Code

• mjRotation
 
1 //

2 // mjRotation.c

3 // mjRotation

4 //

5 // Created by Martin Jaroszewicz on 7/20/14.

6 // Copyright (c) 2014 com.martinjaroszewicz. All rights reserved.

7 //

9 #include "m_pd.h"

10 #include <math.h>

11 #include <string.h>

12

13 static t_class *mjRotation_class;

14

15 typedef struct _mjRotation {

16 t_object x_obj;

17 //local variables

18 t_float f_count;

19 t_float f_x;

20 t_float f_y;

21 t_float f_z;

22 t_float f_elipX;

23 t_float f_elipZ;

24 t_float f_input;

25 t_float f_mult;

26 t_outlet *l1_out, *l2_out, *l3_out;

27 } t_mjRotation;

28

29

30 void mjRotation_float(t_mjRotation *x, t_float f){

31 t_float a1 = x->f_x;

32 t_float a2 = x->f_y;

160
33 t_float a3 = x->f_z;

34 ;

35 x->f_x = sin(f * x->f_mult)*x->f_elipX;

36 x->f_y = 0;

37 x->f_z = cos(f * x->f_mult)*x->f_elipZ;

38 outlet_float(x->l1_out, a1);

39 outlet_float(x->l2_out, a3);

40 outlet_float(x->l3_out, a2); // This works for HOA Library

41 }

42

43 void *mjRotation_new(t_floatarg f){

44

45 t_mjRotation *x = (t_mjRotation *)pd_new(mjRotation_class);

46 x->f_mult = 3.14159 * 2;

47 x->l1_out = outlet_new(&x->x_obj, &s_float);

48 x->l2_out = outlet_new(&x->x_obj, &s_float);

49 x->l3_out = outlet_new(&x->x_obj, &s_float);

50 floatinlet_new(&x->x_obj, &x->f_elipX);

51 floatinlet_new(&x->x_obj, &x->f_elipZ);

52

53 return (void *)x;

54 }

55

56 void mjRotation_setup(void) {

57 mjRotation_class = class_new(gensym("mjRotation"),

58 (t_newmethod)mjRotation_new,

59 0, sizeof(t_mjRotation),

60 CLASS_DEFAULT,

61 A_DEFFLOAT, 0);

62 class_addfloat(mjRotation_class, mjRotation_float);

63

64 }
 

• mjRose
 
1 //

161
 2 // mjRose.c

3 // mjRose

4 //

5 // Created by Martin Jaroszewicz on 7/20/14.

6 // Copyright (c) 2014 com.martinjaroszewicz. All rights reserved.

7 // // This works with HOA Library

9 #include "m_pd.h"

10 #include <math.h>

11 #include <string.h>

12

13 static t_class *mjRose_class;

14

15 typedef struct _mjRose {

16 t_object x_obj;

17 //local variables

18 t_float f_position;

19 t_float f_mult;

20 t_float f_n;

21 t_float f_d;

22 t_float f_z;

23 t_outlet *l1_out, *l2_out, *l3_out, *l4_out;

24 } t_mjRose;

25

26

27 void mjRose_float(t_mjRose *x, t_float f){

28

29 double n;

30 double d;

31 modf(x->f_n,&n);

32 modf(x->f_d,&d);

33 int z = x->f_z;

34 int in = n;

35 int id = d;

36 if(in%in != 0 && id%id !=0){

162
37 x->f_mult = M_PI*2*d;

38 } else {

39 x->f_mult = M_PI*d;

40 }

41 x->f_position = f;

42 float position = x->f_position*x->f_mult;

43 float rose = sinf((n/d)*position);

44 float out_x = rose*cosf(position);

45 float out_y = rose*sinf(position);

46 float out_z = rose*sinf(z);

47 outlet_float(x->l1_out, out_x);

48 outlet_float(x->l2_out, out_y);

49 outlet_float(x->l3_out, out_z);

50 if (out_z == 0.0 && out_y == 0.0 && out_z == 0.0){

51 outlet_bang(x->l4_out);

52 }

53 }

54

55 void *mjRose_new(t_floatarg f){

56

57 t_mjRose *x = (t_mjRose *)pd_new(mjRose_class);

58 x->f_n = 1;

59 x->f_d = 1;

60 x->f_z = 0;

61 x->l1_out = outlet_new(&x->x_obj, &s_float);

62 x->l2_out = outlet_new(&x->x_obj, &s_float);

63 x->l3_out = outlet_new(&x->x_obj, &s_float);

64 x->l4_out = outlet_new(&x->x_obj, &s_bang);

65 floatinlet_new(&x->x_obj, &x->f_n);

66 floatinlet_new(&x->x_obj, &x->f_d);

67 floatinlet_new(&x->x_obj, &x->f_z);

68 post("mj Library by Martin Jaroszewicz.");

69 post(" 2014 UCR Riverside");

70 post("Version Beta 0.1 (August 9 2014) for PD-extended v.0.43.4");

71

163
72 return (void *)x;

73 }

74

75 void mjRose_setup(void) {

76 mjRose_class = class_new(gensym("mjRose"),

77 (t_newmethod)mjRose_new,

78 0, sizeof(t_mjRose),

79 CLASS_DEFAULT,

80 A_DEFFLOAT, 0);

81 class_addfloat(mjRose_class, mjRose_float);

82

83 }
 

• mjRandom
 
1 //

2 // mjRandom.c

3 // mjRandom

4 //

5 // Created by Martin Jaroszewicz on 7/20/14.

6 // Copyright (c) 2014 com.martinjaroszewicz. All rights reserved.

7 //

9 #include "m_pd.h"

10 #include <math.h>

11 #include <string.h>

12 #include <stdlib.h>

13

14 static t_class *mjRandom_class;

15

16 typedef struct _mjRandom {

17 t_object x_obj;

18 //local variables

19 t_float i_num_points;

20 t_float a_x[100];

21 t_float a_y[100];

164
22 t_float a_z[100];

23 t_float f_seed;

24 t_float f_input;

25 t_outlet *l1_out, *l2_out, *l3_out,*l4_out;

26 } t_mjRandom;

27

28

29 void mjRandom_float(t_mjRandom *x, t_float f){

30

31 int index = f*x->i_num_points;

32 post("Incoming float %f with index %d",f, index);

33 outlet_float(x->l1_out, x->a_x[index]);

34 outlet_float(x->l2_out, x->a_y[index]);

35 outlet_float(x->l3_out, x->a_z[index]);

36 if (f == 0.999){ // a phasor never gets to 1 but this does not work

37 outlet_bang(x->l4_out);//change seed here

38 }

39 }

40

41 void mjRandom_bang(t_mjRandom *x){

42

43 float seed = x->f_seed;

44 int points = x->i_num_points;

45 if (points > 100){

46 post("Max number of points is 100");

47 points = 100;

48 }

49 if (points < 0){

50 post("Please provide a value between 0 and 100");

51 points = 0;

52 }

53 int index;

54 float m1[points];

55 float m2[points];

56 float m3[points];

165
57 for (index = 0; index < points; index++){

58 m1[index] = (float)rand() / (float)RAND_MAX;

59 m2[index] = (float)rand() / (float)RAND_MAX;

60 m3[index] = (float)rand() / (float)RAND_MAX;

61 post("x[%d] = %f",index,m1[index]);

62 post("y[%d] = %f",index,m2[index]);

63 post("z[%d] = %f",index,m3[index]);

64 x->a_x[index] = m1[index];

65 x->a_y[index] = m2[index];

66 x->a_z[index] = m3[index];

67

68 }

69 }

70

71 void *mjRandom_new(t_floatarg f){

72

73 t_mjRandom *x = (t_mjRandom *)pd_new(mjRandom_class);

74 x->l1_out = outlet_new(&x->x_obj, &s_float);

75 x->l2_out = outlet_new(&x->x_obj, &s_float);

76 x->l3_out = outlet_new(&x->x_obj, &s_float);

77 x->l4_out = outlet_new(&x->x_obj, &s_bang);

78 floatinlet_new(&x->x_obj, &x->f_seed);

79 floatinlet_new(&x->x_obj, &x->i_num_points);

80

81 return (void *)x;

82 }

83

84 void mjRandom_setup(void) {

85 mjRandom_class = class_new(gensym("mjRandom"),

86 (t_newmethod)mjRandom_new,

87 0, sizeof(t_mjRandom),

88 CLASS_DEFAULT,

89 A_DEFFLOAT, 0);

90 class_addbang(mjRandom_class, mjRandom_bang);

91 class_addfloat(mjRandom_class, mjRandom_float);

166
92

93 }
 

• Python script that generates a point cloud in Houdini.

 
1 import hou

2 import math

4 # Creates a sine wave path and exports to file

6 def createGridGeo():

7 # Creates main obj GEO

8 aGeoNode = hou.node(’obj’).createNode(’geo’,run_init_scripts=False,node_name = "Random_points"

9 # Creates a grid node with 1 row and 180 columns

10 aGridNode = aGeoNode.createNode(’grid’)

11 aGridNode.parm(’cols’).set(180) # The number of points

12 aGridNode.parm(’rows’).set(1)

13 aGridNode.parm(’sizex’).set(2) #size 2 (from -1 to 1)

14 aGridNode.parm(’sizey’).set(1)

15 # Creates a point node and adds a formula to create a sine wave

16 aPointNode = aGeoNode.createNode(’point’)

17 aPointNode.setFirstInput(aGridNode)

18 aPointNode.parm(’tx’).setExpression("$TY + rand ($PT) * 0.2") #c

19 aPointNode.parm(’ty’).setExpression("$TZ + rand ($PT*3) * 0.2")

20 aPointNode.parm(’tz’).setExpression("$TX + rand ($PT*4) * 0.2")

21 # Creates a delete node to delete everything but the points

22 aDeleteNode = aGeoNode.createNode(’delete’)

23 aDeleteNode.setFirstInput(aPointNode)

24 aGeoNode.layoutChildren()

25 aDeleteNode.parm(’pattern’).set("0 $N")

26 aDeleteNode.parm(’keeppoints’).set(1)

27 aDeleteNode.setDisplayFlag(1)

28 # Creates a file node to export

29 aFileNode = aGeoNode.createNode(’file’)

30 aFileNode.setFirstInput(aDeleteNode)

167
31 aFileNode.parm(’filemode’).set(2) #write

32 aFileNode.parm(’file’).set("test.poly")

33 aFileNode.setDisplayFlag(1)

34

35

36 createGridGeo()
 

168
.1 Appendix B - Spectrograms

(a) A chord played by a string ensem- (b) Granular texture. Dur: 14 sec-

ble. Dur: 7 seconds. onds.

(c) General Cross-Synthesis between (a) and (b)

(d) Source Filter Cross-Synthesis between (a) and (b)

Figure .1: Different Cross-synthesis techniques.

169
(a) Male voice speaking in French. (b) Granular glass sounds. Dur: 16

Dur: 15 seconds. seconds.

(c) Organ C chord. Dur: 12 seconds.

(d) Latin Percussion. Dur: 29 seconds.

Figure .2: Sounds with contrasting spectrum.

170