UvA-DARE (Digital Academic Repository)

Noise resonance Technological reproduction and the logic of filtering Kromhout, M.J.

Publication date 2017 Document Version Final published version License Other Link to publication

Citation for published version (APA): Kromhout, M. J. (2017). resonance: Technological sound reproduction and the logic of filtering.

General rights It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).

Disclaimer/Complaints regulations If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible.

UvA-DARE is a service provided by the library of the University of Amsterdam (https://dare.uva.nl)

Download date:06 Oct 2021 MELLE JAN KROMHOUT

NOISE

RESONANCE

TECHNOLOGICAL SOUND REPRODUCTION

AND THE LOGIC OF FILTERING

MELLE JAN KROMHOUT

NOISE

RESONANCE

TECHNOLOGICAL SOUND REPRODUCTION

AND THE LOGIC OF FILTERING

Printed by Ipskamp Printing

Cover design by Rufus Ketting

ISBN: 978-94-028-0524-6

© 2017, Melle Jan Kromhout. All rights reserved. No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the author. www.mellekromhout.nl

Noise Resonance Technological Sound Reproduction and the Logic of Filtering

ACADEMISCH PROEFSCHRIFT

ter verkrijging van de graad van doctor aan de Universiteit van Amsterdam op gezag van de Rector Magnificus prof. dr. ir. K.I.J. Maex ten overstaan van een door het College voor Promoties ingestelde commissie, in het openbaar te verdedigen in de Agnietenkapel op donderdag 16 maart 2017, te 12:00 uur

door Melle Jan Kromhout geboren te Amsterdam Promotiecommissie:

Promotor: Prof. dr. S.A.F. van Maas, Universiteit Utrecht

Overige leden: Dr. C.J. Birdsall, Universiteit van Amsterdam Prof. dr. R. Boast, Universiteit van Amsterdam Prof. dr. J.J.E. Kursell, Universiteit van Amsterdam Prof. dr. J. de Mul, Erasmus Universiteit Rotterdam Prof. dr. A. Rehding, Harvard University, USA

Faculteit der Geesteswetenschappen

Table of contents

Acknowledgements ...... IX

Introduction: The role of noise in the of the media age 0.1 A dominant discourse on noise in/as music ...... 3 0.2 Sonic, physical and communicational noise ...... 7 0.3 The noise of sound reproduction ...... 11 0.4 Resonating noise between music and media ...... 14

Chapter 1: “How much noise is necessary?” A brief history of sound recording and 1.1 Analogue 1: disc recording ...... 21 1.2 Analogue 2: magnetic tape recording ...... 33 1.3 Digital recording ...... 46 1.4 Dither: fighting noise with noise ...... 62

Chapter 2: Confronting the fuzziness of the real Dolby and dither: concealing and revealing noise 2.1 Noise reduction reconsidered ...... 71 2.2 Things you want and things you don’t want ...... 79 2.3 From ideal models to physical filters ...... 86 2.4 Digital error and analogue noise ...... 92 2.5 Revealing and concealing noise ...... 103

Chapter 3: Infinite time or perfect transience Ideal filters and the temporality of sound 3.1 Toward the limits of representation ...... 117 3.2 Fourier analysis: history and basic principles ...... 130 3.3 The domain of ideal filters ...... 140 3.4 Time and transience: the domain of physical filters ...... 149 Chapter 4: “In the Fourier domain are we immortal” On the pastness and presence of reproduced sound 4.1 The importance of sonic transience ...... 159 4.2 Pastness and finitude in reproduced sound ...... 168 4.3 The presence of reproduced sound ...... 181 4.4 On passed and passing time ...... 187

Chapter 5: The sound of an ‘other music’ Sonic transience and the darker presence of the media age 5.1 A logic of filtering ...... 195 5.2 Producing a new signal ...... 205 5.3 The new sound of music ...... 214 5.4 An ‘other music’ and the presence of the Real ...... 218

Epilogue: Listening for the event On the possibility of an ‘other music’ 6.1 Locating the ‘other music’ ...... 231 6.2 Music hiding in hardware ...... 233 6.3 I hear a new world ...... 236 6.4 Hearing an ‘other music’ ...... 238

Conclusion: Noise resonances Sound in the age of technical media ...... 245

References Bibliography ...... 255 Discography ...... 266 Videography ...... 267

Summary ...... 269

Samenvatting ...... 273

NOISE RESONANCE | ACKNOWLEDGEMENTS IX

Acknowledgements

As people repeatedly warned me in the years before beginning my PhD at the Amsterdam School for Cultural Analysis: writing a dissertation is a lonely job. This is definitely the case, but it does not mean one works alone. Many people have been indispensible for this project. First and foremost, I have to thank my supervisor Sander van Maas. From well before the earliest roots of what became this book, going back all the way to his classes on music at the Amsterdam Conservatory more than ten years ago and subsequently to his supervision of my bachelor thesis in 2006, Sander encouraged and challenged my intellectual development like no one else. Over the course of multiple attempts to secure a PhD-position, it was his often relentless, but always fair and encouraging critical commentary that kept me from giving up altogether. Hence, it is no overstatement saying that without his confidence during these earliest stages of the project and his ability to keep me from settling for anything less than the best, this book would not have turned out the way it did. During the past four and a half years, I have often cursed his incredible eye for detail and conceptual rigour, but also thoroughly enjoyed our stimulating intellectual exchange. Having completed this book, I sincerely hope we will continue to have our long conversations and will keep exchanging ideas and music for many years to come. Next, I want to thank my PhD-colleagues at ASCA for sharing this journey with me and not seldom providing some much needed distraction and comic relief in the form of drinks, parties or a simple coffee. These are of course the people I have shared my office with over the years: Judith Naeff, Mirjam Meißner, Noam Knoller, Thijs Witty, Matt Cornell and Andres Ibarra, but also the rest of the ASCA PhD-family, including Simon Ferdinand, Blandine Joret, Lucy van der Wiel, Melanie Schiller, Annelies Kleinherenbrink, Lara Mazurski, Geli Mademli, Pedram Dibazar, Irene Villaescusa, Jeffrey Pijpers, Eva Sancho, Peyman Amiri, Margaret Tali, Enis Dinc, Selçuk Balamir, Nur Ozgenalp, Asli Ozgen-Tuncer, David Gauthier, Christian Olesen and many others. Of course, at the office, Eloe Kingma and X NOISE RESONANCE | ACKNOWLEDGEMENTS

Jantine van Gogh—the heart and soul of ASCA—continue to make PhD-life easier and much more enjoyable for all of us. Outside of Amsterdam, special thanks goes out to Peter McMurray, who was crazy enough to join my idea for an international online reading group, which resulted in a great online resource with input by a magnificent group of international scholars of which I am still very proud. Also, the organisers and participants of the Sound Studies Summer and Winter Schools in 2014 and the Princeton/Weimar Summer School in 2016 offered very inspiring platforms to learn, discuss, try out ideas and meet many interesting people from across the globe. Although, in the midst of the whole endeavour, it sometimes seems as if the ivory tower of academia is indeed without windows, there was and always is a life outside of these walls that kept me with both feet on the ground. My dear friends in Amsterdam and elsewhere, many of whom I’ve known for over half my life, are like a second family. You know who you are, I would not trade any of you for the world and I could not have finished this book without you. My band members in the now defunct Fata ‘el Moustache’ Morgana and still operative Glice kept me in touch with music as a living, breathing thing instead of just words on paper and offered the sublime outlet of dancing and screaming on stage. Last but not least, my parents showed me the beauty and power of making and listening to music almost from before I was born and taught me the invaluable lesson to always approach the world with open ears and without prejudice. My big sister initiated me in the world of during my teenage year and has been my role model ever since; I hope we will be sharing music for the rest of our lives. Lastly, there is only one person who has truly dealt with all the ups and downs that went into writing this book, from overtly enthusiastic babbling to completely irrational despair. It is one thing to write a book in four years, it is another to build a life together for twelve years and counting. In the end, being home is the only thing that matters; and I am home wherever you are: the purest signal amidst the noise.

Amsterdam, January 2017

If we view ourselves from a great height, it is frightening to realize how little we know about our species, our purpose and our end, I thought, as we crossed the coastline and flew out over the jelly-green sea. - W.G. Sebald – The Rings of Saturn (2002: 92)

Strangely enough, there are signals that exist in all mathematical purity, such as on the one hand the sinus tones that have lasted for eternities and that will continue to sound for eternities, and on the other hand signals like the Dirac Impulse, whose large but finite energy discharges itself in an infinitely short time. Between these two extremes our life is made up, in the words of Sophocles, of ‘mere phantoms, shadows of nothing’. - Friedrich Kittler – “Lightning and Series – Event and Thunder” (2006a: 71)

Das schläft in der Maschine. - Einstürzende Neubauten – “NNNAAAMMM” (Bargeld 1997: 126)

NOISE RESONANCE | INTRODUCTION 3

Introduction: The role of noise in the music of the media age

0.1 A dominant discourse on noise in/as music

More than ten years ago, around 2005, I started researching what was to become my bachelor thesis about the work of German - pioneers Einstürzende Neubauten (Kromhout 2006). I became acquainted with the band in my late teens through the album Silence Is Sexy, released in 2000, and had slowly worked my way back through their catalogue toward the debut album Kollaps from 1981 (Einstürzende Neubauten 2000; 2003). Whereas the work from 2000 consists of experimental pop songs combining metal percussion, melodic bass lines, enticing deep vocals and a fair share of unusual —from self-made instruments and field recordings to experimental studio practices—the earlier material introduces a much more abrasive sound world: screeching, shouting, banging, drilling, manic drumming on metal objects, gut wrenching screams, loud, distorted, metallic guitars, harsh, unwelcoming layers of dissonance. This music, sounding like nothing else, first opened my ears for all the noise that is sometimes called music and all the music that is sometimes called noise. I was hooked. At the time, the questions I asked regarding this sound world were very, in hindsight one might even say naively, simple. Why, I wondered, do people listen to this noise? What makes it appealing or musically attractive? Why are we drawn to it? Why do we like it? In trying to answer these questions, however, what had seemed relatively straightforward became more and more opaque at every turn. Questions regarding the issue of noise invoke a host of disparate academic disciplines—from musicology to physics and from media studies to mathematics, to name a 4 NOISE RESONANCE | INTRODUCTION

few—and among those disciplines, the very definition of noise itself is notoriously unstable. Noise is a sonic object, a social nuisance, a physical phenomenon, a therapeutic background sound, a concept in communication theory, a musical genre, a legislative issue, an obstacle in sound engineering. Most of all, it is, as philosopher Michel Serres writes, “a black thing, an obscure process, or a confused cloud of signals—what we shall soon call a problem” (1982b: 17). Regarding this problem, my attempts in 2005 and 2006 barely scratched the surface, so those somewhat naive questions stayed. The prime reason why they did, however, was my strong impression that other accounts on the role of noise in music did not provide entirely satisfactory answers either. In The Aesthetics of Noise, Danish cultural theorist Torben Sangild interprets the role of noise in music in three ways. Firstly, as “an aesthetization of violence and suffering.” Secondly, as abject sounds “that are discarded as being impure, unmusical.” And thirdly, as a kind of multiplicity that is “often used to express anger, fear and violence” (2002: 21, 25-26). For Sangild, noise is an abject, transgressive, violent, subversive sonic element that forces itself into the domain of music and mainly contributes to it by means of contrast—destroying or disrupting its clarity, and regularity. Noise confronts order with chaos, harmony with dissonance, peacefulness with aggression and social cohesion with disruptive subversion. These interpretations are not uncommon for literature on the role of noise in or as music. More often than not, sound scholar Caleb Kelly writes, noise is considered “a disruptive and excessive area of sound practice” and the source for a kind of “joyful transgression” (2009: 61).1 More often than not, in the context of music, noise is regarded as the ultimate sonic Other. ’s 1913 futurist manifesto The Art of Noise, which is often considered the founding document of both ‘’ as a genre and the theory of its relation to music, already implicitly considered noise to be separate from proper musical sound (Russolo 2004). As film and sound scholar Michel Chion argues in “Let’s Have Done with the Notion of

1 For a concise, although somewhat German-language oriented, overview of academic literature on noise in or as music see Borsche, 2015. NOISE RESONANCE | INTRODUCTION 5

‘Noise’,” notwithstanding Russolo’s call for the inclusion of all kinds of in the music of his day, his manifesto “confirms the idea of an absolute distinction—an essential distinction—between musical sounds and noises” (2011: 244). Heralded as the saviour of music, the breaker of rules or the liberator of sound, noise is defined in contrast to what it saves, breaks or liberates: well-ordered musical sound. Similarly, a book that is still regarded as one of the seminal accounts on the role of noise in the history of music, social economist Jacques Attali’s 1977 Noise. The Political Economy of Music describes this history as a dialectic interplay between musical order and disruptive noise. Although the latter is conceptualised as an emancipatory force, not to be excluded or discarded but to be embraced and nurtured as essential to the advancement of new forms of music, this power is fundamentally based on its ability to transgress and violate musical and social order. The power of noise, Attali argues, is therefore akin to the uncontrolled violence of murder; and like the controlled violence of ritual murder, music serves as its “channelization” (2003: 12-13). Thus, in this continuous cyclical interplay, noise disrupts well-ordered music whereupon the order of music tames disruptive noise. Cultural theorist Paul Hegarty’s more recent Noise/Music. A History is one of the few books that deals exclusively with noise as a musical phenomenon (2008). Ordered thematically along a set of themes relevant to the history of noise as music (such as ‘electricity,’ ‘industrial,’ ‘power’ or ‘sound art’), Hegarty describes its changing role and increasing importance over the course of the twentieth century. Throughout the book, his concept of noise alternates between a primarily metaphorical reading akin to Attali’s take on noise as an agent of change, resistance and subversion, and a more phenomenological approach that considers it first and foremost as a welcome source for new musical sounds and genres. Going back and forth between these two concepts, Hegarty’s historical overview of the emancipation of noise in or as music also affirms the dominant dialectic interpretation of the relation between noise and music, chaos and order, violence and peace. Although his account adds valuable nuances to this history by differentiating between various forms of noise and different types of musical practices, for Hegarty as well, noise remains a principle 6 NOISE RESONANCE | INTRODUCTION

antithesis to the order of music: a marker for failure, incompletion, transgression, disruption and subversion.2 Staying in tune with my initial interest in the appeal of noise as a musical phenomenon, however, I want to ask whether the musical noise practices envisioned by Russolo, conceptualised by Attali and described by Hegarty can really only be understood on the basis of this inherent antithetical character of noise. Can we only account for the proliferation of noise practices in contemporary musical culture within an oppositional framework based on an inherently negative definition of noise? Or is it also possible to consider the musical importance and appeal of noise as something that contributes to the world of sound and music on its own terms—as a sonic phenomenon characterised by randomness and non- periodicity? Crucially, assuming such a different role of noise implies that its presence in music does not signify, or at least not only signifies, failure, transgression and disruption. Taking on this assumption, the key questions are as follows: how would it be possible to identify, and conceptually come to terms with an affirmative, perhaps foundational role for noise in contemporary musical practices; and how would it subsequently be possible to account for the ways in which noise, as an affirmative sonic presence, makes musical sense?

2 Ultimately, Hegarty writes in his conclusion, noise “is not proper, linear, meaningful. But not bad either, as noise transvalues listener and object, noise and music, hearing and listening, perception and its failure, performance and its failure, noise and its failure to be music, noise and its failure to be noise. And the transvaluation itself, only as if it could ever be. As if it really were noise, after or before, all” (2008: 200). Recently, several essays in Resonances. Noise and Contemporary Music, edited by Michael Goddard, Benjamin Halligan and Nicola Spelman intend to break with the conceptualisation of noise as failure, transgression and disruption (2013). Similarly, in Noise Matters. Toward an Ontology of Noise, Greg Hainge uses several case studies—from haikus to horror movies and from David Lynch to Merzbow—to redefine noise as a “relational process through which the world and its object express themselves” (2013: 13). For Hainge, noise is thereby something that appears with every expressive act. Cultural theorist and dubstep-DJ Steve Goodman in his book Sonic Warfare also takes noise beyond what he calls the futurist notions of " & speed." Arguing in favour of "a transdisciplinary concept of noise," he takes noise as a form of "turbulence" and "rhythmic reservoir" (2010: 105-107). Lastly, breaking with “the binary shackles of noise as good or bad” and “the all-too-common belittling of noise as mere epiphenomenon and fleeting byproduct,” Hillel Schwartz’s monumental Making Noise. From Babel to the Big Bang & Beyond traces the importance and significance of sonic noise, including its “metaphorical bearings […] as something other than acoustic,” throughout human history (2011: 25-26, 29). NOISE RESONANCE | INTRODUCTION 7

0.2 Sonic, physical and communicational noise

Because revaluating the role of noise in music means a break with the opposition-driven discourse that remains dominant in most accounts, I suggest that developing such a revaluation could benefit from a methodological step back from the domain of music. Comparing the role of noise in music with its role in domains like information theory, communication engineering and media studies opens up new perspectives on the development of the concept of noise that, as I intend to show, could further the understanding of musical noise as well. In “Concepts and Significance of Noise in Acoustics. Before and After the Great War,” historian of science Roland Wittje describes three concepts of noise that developed in the “German scientific discourse” and roughly correspond to three “notions […] in the German language, ‘Lärm,’ ‘Geräusch,’ and ‘Rauschen’” (2016: 8). The more or less consecutive development of these concepts between the mid-nineteenth and the mid-twentieth century can be described as a series of semantic and conceptual extensions of the notion of noise in the contexts of, first, sound and acoustics, second, communication engineering and physics and third, information and communication theory. Geräusch, firstly, denotes a sonic concept of noise grounded in nineteenth century acoustics, most significantly German physicist Hermann von Helmholtz’s extensive experimental analysis of sound colour. On the one hand, it defines noise more or less objectively as sound consisting of primarily non-periodic frequencies. On the other hand, Wittje writes, because this emphasis on the non-periodicity of noise also marks its fundamental difference from periodic ‘musical’ sound, it laid the basis for the oppositional relation between noise and music that “continued to characterize the discourse of noise into the twentieth century” (11). Secondly, in the last decade of the nineteenth and first decades of the twentieth century, the physical similarities between this non-periodic sonic noise and random physical disturbances that appear in the transmission channels of communication media inspired scientists, inventors and engineers to define a more general physical concept of noise corresponding to the German word Rauschen (24). To clearly differentiate 8 NOISE RESONANCE | INTRODUCTION

this physical concept from the most commonly used notion of noise as any unintended, undefined or disturbing sound, engineer Robert Höldrich notes, the most accurate English translation of this physical Rauschen is “random noise” (1995: 128).3 Thirdly, this commonly used notion of noise as any disturbing, unintended or unwanted sound corresponds to the German word Lärm. As historians of science and technology Emily Thompson and Karin Bijsterveld describe, this notion first rose to prominence in discussions regarding and noise abatement in the early twentieth century, as well as through modernist associations between noise, speed, power and progression—for example in Russolo’s Futurist manifesto (Thompson 2002; Bijsterveld 2008). In the context of communication engineering and signal processing, however, the combination of the objective physical concept of noise as Rauschen (the random movement of fluctuating particles) and this more subjective notion of noise as Lärm (any acoustic nuisance) inspired a communicational concept of noise developed in information theory in the 1930s and 1940s (Mills 2011: 123). Spearheaded by the publication of mathematician and engineer Claude Shannon’s “Mathematical Theory of Communication” in 1948, information theory formalised the operations of communication media based on the statistical calculation of the relation between information and noise (Shannon and Weaver 1964). Significantly, it thereby no longer defines noise on the basis of its sonic or physical characteristics, but redefines it solely on the basis of its role in communication systems. In information theory, every signal that hinders or affects the clear transmission of information constitutes as noise. By treating this relativist,

3 According to media theoretician Geoffrey Winthrop-Young, in the nineteenth century, the German word Rauschen (etymologically related to the English ‘rustling’) changed from exclusively denoting natural ‘noisy’ (and indeed rustling) sounds like the wind in the trees or the surf of the sea to referring to all physical signals characterised by broadband randomness (2010: 81). As physicist Leon Cohen describes in “The History of Noise (on the 100th Anniversary of Its Birth),” parallel to this introduction of the notion of random noise in the field of communication engineering, noise was conceptualised and defined in theoretical physics, beginning with Einstein’s 1905 paper on stochastic processes and so-called Brownian motion—named after botanist Robert Brown’s observation of the stochastic behaviour of small particles under a microscope, which he describes in a paper from 1828 (2005: 22). NOISE RESONANCE | INTRODUCTION 9

statistical noise not as an external thread to the communication system, but as an internal factor that is inherent to it, Shannon’s theory shows how to calculate and potentially reduce its influence (Sterne 2012: 81). Exactly because it is considered internal to the system, however, information theory also proves that the complete reduction of communicational noise is fundamentally impossible. Claude Shannon’s model of communication and its statistical (re)definitions of information and noise are, as his most important commentator mathematician Warren Weaver writes, “exceedingly general in its scope” and their impact can hardly be overstated (Shannon and Weaver 1964: 25). Due to the general applicability of Shannon’s model, Wittje points out, the communicational concept of noise found its way to “virtually all fields of science and engineering, and even the social sciences” (2016: 7). Even beyond the natural and social sciences, I would add, it resonated far and wide. Although Attali’s and Hegarty’s interpretations partly take on the sonic concept of noise as non-periodic sound, the model of a dialectic interplay between the forces of noise and music is clearly indebted to the communicational concept according to which noise can be statistically located, specified and reduced, but never completely eliminated. Because of its fundamentally contextual definition rooted in mathematical statistics, the communicational concept of noise is notoriously fluid. As a consequence, literary scholar Greg Hainge argues in Noise Matters. Towards an Ontology of Noise, it “has been used to apply to everything and nothing at the same time, subject to a whole host of mutually contradictory definitions and usages, its apparently ineffable nature the result of divergent agendas rather than something proper to noise itself” (2013: 8). Hence, on the long run, Goddard, Halligan and Spelman write in their introduction to Resonances. Noise and Contemporary Music, owing to the general applicability of information theory and the communicational concept of noise, exactly the “expansion of noise ‘studies’ into multiple fields [risks] a loss of focus on, or dispersal of the relation between, noise and music” (2013: 9). Taking up on this critique and backtracking on the proliferation and expansion of noise studies into all 10 NOISE RESONANCE | INTRODUCTION

corners of academia, I want to return to the question regarding its role and importance in sound and music. The discursive slippage of the communicational concept of noise notwithstanding, returning to questions regarding its role in music does not imply a return to the sonic concept of noise, as this would mean a return to the definition that inspired the discursive opposition between periodicity and non-periodicity, order and chaos, harmony and dissonance in the first place. As Wittje shows, firmly rooted in nineteenth century acoustics and a supposedly objective difference between periodic and non- periodic sounds, the sonic concept of noise implies a structural opposition between harmonious music and non-harmonious noise (2016: 10).4 Hence, whereas the communicational concept of noise in information theory risks being too general and relativistic, the sonic concept of noise in sound and acoustics is too specific and restrictive. By defining noise as “a vibration, either electrical or mechanical, which cannot be dissolved into periodic harmonic vibrations,” however, the physical concept of noise focuses on the material basis of the physical processes in which random noise emerges (24). As such, it might provide the necessary conceptual middle ground between the relativism of the communicational concept and the essentialism of the sonic concept of noise. As Wittje’s account of the changing concept of noise from the mid- nineteenth century onward indicates, the physical concept of noise originates in the context of late nineteenth and early twentieth century communication engineering. Notably, along with this changing concept of noise, the proliferation of musical noise practices occurred almost simultaneous to the invention and further development of technical media that transmit, record, store and reproduce physical sound. Given the close interconnection between music, media and the changing concept of noise, it seems no more than logical to assume that this contemporaneity of the proliferation of noise in musical practices and the development of technological sound reproduction is not entirely coincidental. On the basis

4 For Helmholtz, Wittje explains, “the sensation of sound and the experience of music were interchangeable,” so the perceived periodicity of sound in contrast to the non-periodicity of noise always also implies the ideal of well-ordered music (2016: 10). NOISE RESONANCE | INTRODUCTION 11

of this assumption, I suggest that a critical analysis of the role of noise in sound reproduction can provide the necessary conceptual groundwork for the revaluation of its role in music.

0.3 The noise of sound reproduction

This combination of noise, music and media brings me to the German tradition of media theory, in particular the work of media philosopher Friedrich Kittler. On the one hand, as media theorist Jussi Parikka writes, Kittler’s media theory and the traditions that followed in its wake “insisted that the ‘founding event’ of modern media is Claude Shannon’s and Warren Weaver’s mid-twentieth-century model of noise” (2011: 256). On the other hand, however, although it is clear that Kittler’s post-hermeneutic discourse analysis of technical media is highly influenced by Shannon’s redefinition of the relation between noise and information, this influence also means that German media theory, as Kittler explains in an interview with Rudolf Maresch in 1992,

[takes] noise […] very serious. We do not just treat it as ‘the Other.’ We try to make truly differentiated statements about specific filterings of noise, not just dealing with the single command that forever abolished all noise to begin with, but with media- and time-specific selections that deal with noise to a greater or lesser extent (1992a).5

5 “Unsere Medientheorie [nehmt] das Rauschen […] sehr ernst. Wir nehmen es nicht einfach als “das Andere”. Wir versuchen wirklich differenzierte Aussagen über bestimmte Filterungen des Rauschens zu machen, also nicht den einen Befehl, der am Anfang alles Rauschen für immer abgeschafft hat, sondern medien- und zeitspezifische Selektionen anzugeben, die mehr oder minder das Rauschen bewältigen.” NB. Translations of otherwise not translated German sources are my own. The original is provided in the footnotes. 12 NOISE RESONANCE | INTRODUCTION

Kittler consistently situates the causes and effects of noise and signals—and the relation between the two—in the physical operations of media technological hardware. This materialist media specificity prevents the kind of conceptual slippage of analyses that apply the concept of noise, as Hainge puts it, “to everything and nothing at the same time” (2013: 8). By accounting for the fact that the mathematical reconceptualisation of the concept of noise in information theory originates in the conceptualisation of physical noise as random disturbances in transmission channels, Kittlerian media theory always approaches the issue of noise in the specific context of media technological processes. Following this media specific approach, I suggest that an analysis of the technological processes through which the specific noise of sound reproduction emerges enables a revaluation of the role of noise as a sonic phenomenon in music without sacrificing the conceptual agility of the concept of noise in information theory. This noise of sound reproduction is what musicologist Stan Link calls “the noise of documentation and transduction” and encompasses all the ways in which the operations of technological sound reproduction affect, change and shape the physical characteristics of reproduced sound (2001: 34). Firstly, it refers to the physical noise (or random disturbances) that appears in electronic circuits and transmission channels, which can be conceptualised as communicational noise and sometimes, but certainly not always, manifests as sonic noise. Secondly, it is linear or non-linear that does not qualify as physical noise because it is not random, but is still considered communicational noise and in some cases manifests as sonic noise as well. Lastly, it can even refer to interferences that do not qualify as random physical noise nor as non-random distortion but still register as communicational noise because they affect the transmission of the signal, even when they are not perceived as sonic noise. From the earliest recordings onward, sound reproduction technologies have been confronted with the introduction of all these types of noise. As a consequence, inventors, sound engineers and musicians developed a refined sense for the many ways in which the noise of sound reproduction is both (and sometimes simultaneously) disturbing or harmful and enriching or desirable. The methodological step back from the NOISE RESONANCE | INTRODUCTION 13

domain of music in favour of a media specific analysis of noise in the context of technological sound reproduction thereby establishes a conceptual middle ground between the sonic and the communicational concept of noise. More specifically, it allows for a conceptual reconfiguration of the relation between the sonic noise that is still of primary concern for questions regarding sound and music, the physical noise that appears in the channels and circuits of technical media and the communicational noise that was conceptualised in the context of information theory. Hence, this analytical approach offers a unique opportunity to develop a different, perspective on the concept of noise in general. However, bringing this media technological reading of the role of noise back to the domain of music requires the development of a common conceptual frame in which the discourse on technical media and the discourse on sound and music can meet. Following media theoretician Berhard Siegert’s concept of “cultural techniques” as an extension of the media theoretical framework focusing on “operative chains that precede the media concepts they generate,” establishing such a conceptual common ground requires an even more radical detour toward the historical assessment and theoretical analysis of the mathematical and physical principles in which both the contemporary discourse on sound and music and the discourse on signal processing and technical media find their origin (2015a: 11). This expansion of the analytical scope to include the assessment of the fundamental scientific framework that shaped the dominant discourse on sound, noise and media facilitates the jump back to the domain of music. Creating a conceptual connection between a media specific analysis of the noise of sound reproduction and the scientific discourse that supports the basic principles of the physics of sound paves the way for a more theoretical and metaphorical reading of the interaction between noise, music and media in the final stretch of this thesis. Ultimately, this theoretical conceptualisation of the role of noise in technological sound reproduction and its relation to the physics of sound enables the development of a more affirmative interpretation of the role of noise in 14 NOISE RESONANCE | INTRODUCTION

contemporary musical practices and the way it resonates in the ears and brains of listeners.

0.4 Resonating noise between music and media

In order to properly identify the dominant discursive framework regarding the role of noise in technological sound reproduction, Chapter One opens with a historical overview spanning from the invention of Edison’s phonograph to the advance of digital recording. Focussing on the interaction between the development of sound reproduction technologies and attempts to prevent, reduce or eliminate the noises and they add to their output, this history traces the emergence of a persistent myth of perfect fidelity based on what sound scholar Jonathan Sterne in The Audible Past calls the ideal of a “vanishing mediator” (2003: 218). By assuming that the output of technological sound reproduction should ideally be identical to its input, this myth suggests that the noise of sound reproduction is the inherent enemy of all reproductive transparency, to be prevented, eliminated or maximally reduced at all times. To develop a better understanding of the way this myth of perfect fidelity shapes the discourse on technological sound reproduction, Chapter Two consists of two case of technologies that explicitly confront the issue of noise in sound technology. Firstly, it deals with inventor Ray Dolby’s analogue noise reduction systems of the 1960s, which are applied to actively reduce the noise of sound reproduction. Secondly, it analyses so- called ‘dithering,’ which is the deliberate addition of small amounts of noise to a digital recording in order to eliminate the effects of certain digital errors. Although these examples constitute seemingly contrasting attitudes toward the role of noise in sound reproduction, a detailed analysis of their operations shows how both the reduction of noise in analogue systems and its intentional addition to digital sound are ways to conceal limitations that are inherent to all forms of sound reproduction. Through this concealment, they uphold the suggestion of the ideal relation between input and output that supports the myth of perfect fidelity. This analysis of the role of noise in sound reproduction and the myth of perfect fidelity in Chapter One and Two thereby shows how a NOISE RESONANCE | INTRODUCTION 15

conceptual logic of noise reduction dominates the discourse on technological sound reproduction and perpetuates the discursive opposition between disturbing and disrupting noise and pure and clear music. Breaking with this conceptual logic of noise reduction, I introduce the alternative concept of the noise resonance of sound reproduction to reveal instead of conceal the effects caused by the structural limitations of technological sound reproduction. This concept thereby emphasises how the inevitable presence of noise indicates that the operations of sound reproduction technologies fundamentally affect and thereby change the sounds they (re)produce. This development of this concept of noise resonance, however, requires a deeper understanding of the idealised relation between input and output suggested by the conceptual logic of noise reduction. This is why Chapter Three leaves the media specific analysis of sound reproduction technologies behind and traces the discursive origins of the ideal of infinitely precise, absolutely clear and completely noiseless reproduction that underpins the myth of perfect fidelity. It does so by focussing on what Kittler calls an early nineteenth century “method of calculation that paved the way not just for thermodynamics but also for all media of technological sound-catching, from Edison’s cylinder phonograph up to the music computers” (2013c: 171). This particular “method of calculation” is the symbolic representation of sound waves as a series of singular frequencies by so-called ‘Fourier analysis,’ including the concept of the ‘sine wave’ as the representational figure for such a singular frequency—often interpreted as the basic element of sound. The historical and conceptual assessment of Fourier analysis in Chapter Three shows how the representation of sound as a series of clear sine waves can be conceived as an entirely periodic and therefore essentially noiseless idealisation of a sonic event. It also shows, however, that this removal of randomness and contingency mathematically requires the symbolic removal of any sense of the temporal duration of a sound signal. Hence, although Fourier analysis is able to symbolically represent completely noiseless sound spectra, the technological reproduction of sound remains inherently limited by the fact that physical sounds extend in space and develop over time. Because this development over time 16 NOISE RESONANCE | INTRODUCTION

introduces small amounts of contingency, randomness, transience and, indeed, noise, physical reproductions by technical media can never achieve the pure, noiseless representational clarity of Fourier analysis. Chapter Three thereby shows how, on the one hand, the conceptual logic of noise reduction is predicated on the representational clarity of Fourier analysis, whereas, on the other hand, the noise resonance of sound reproduction is based on a structural relation between the presence of noise and the temporality of sound. After this analysis of the mathematical and physical principles underlying the symbolic representation and technological reproduction of sound, the argument could return to the media specific analysis conducted in Chapter One and Two, thereby taking up the kind of media archaeological assessment of sound, audio technology and the temporality of technical media developed by German media theorist Wolfgang Ernst. In this work, Ernst argues for the strict epistemological separation between on the one hand analyses that deal with the way human beings try to make sense of the output of technical media, and on the other hand, a type of fundamentally a-historical media archaeology that analyses the operations of technical media without taking this human agency and historicity into account. Because their operations take place on entirely different temporal levels altogether, Ernst argues, the post-human logic of technical media structurally undercuts or surpasses human physiological senses. This is why, for Ernst, their symbolic signification is no longer concerned with human sense-making at all.6 However, contrary to Ernst’s post-humanist interpretation of

6 This thesis is developed throughout most of Ernst’s more recent work, most extensively in Chronopoetik: Zeitwesen und Zeitgaben Technischer Medien and Gleichursprünglichkeit. Zeitwesen und Zeitgegebenheit von Medien (2012a; 2012b). More specifically related to sound media, the argument is put forward in Im Medium Erklingt die Zeit and Sonic Time Machines. Explicit Sound, Sirenic Voices, and Implicit Sonicity (2015; 2016). Regarding the difference between human, culturally conditioned ‘listening’ and the media archaeological study of technical signals outside of this cultural context, he writes in the latter: “If the communicational approach to sound focuses on listening as cultural interpretation […], the media- archaeological understanding assumes an interlaced option. It concentrates neither on the socio- historical, nor the bare psychoacoustic level but on the epistemological dimension that is embedded in sonic articulation. As a result, the study of the sonic signal event can refrain from immediate cultural contextualization without being reduced to mere physical acoustics” (2016: 45). NOISE RESONANCE | INTRODUCTION 17

technical media, the analyses put forward in Chapter One and Two intend to open up an analytical space that connects the communicational concept of noise dominant in Ernst’s, but also Kittler’s, type of media theory with questions regarding the role of noise as a sonic phenomenon in the context of music. This is why Chapter Four consist of a more metaphorical investigation of the way that the suggested relation between the presence of noise and the temporality of sound might contribute to how technologically (re)produced sound makes sense for human listeners. Because the analytical clarity and purity represented by Fourier analysis cannot be achieve by technical media, any process of technological sound reproduction will, as Shannon proved as well, introduce a certain amount of noise and randomness to the signal. Contrary to the ideal filter of Fourier analysis, the physical filtering operations of technical media affect and change the specific sonic characteristics of their output. By combining the analysis of the structural limitations of technological sound reproduction in Chapter Two with the relation between noise and temporality suggested in Chapter Three, Chapter Four argues that the noise of technological sound reproduction simultaneously resonates with our sense of pastness and finitude and our sense of presence in the here and now. Finally, completing the revaluation of the role of noise in sound reproduction and bringing it back to where it originally started—the domain of music—Chapter Five puts forward a primary logic of filtering as an alternative for the idealised relationship between input and output presupposed by the conceptual logic of noise reduction. Similar to how the conceptual logic of noise reduction underpins the myth of perfect fidelity, I suggest this logic of filtering supports the noise resonance of sound reproduction. It emphases how every technical filter in the recording and reproduction chain between sender and receiver shapes the output of technological sound (re)production in specific ways. This specific impact of fundamental filtering operations on the sound of technologically produced music, I claim, produces what Kittler calls an ‘other music.’ Although by the time of his death in October 2011 he had already published the two parts of the first volume of his planned tetralogy called Musik und Mathematik, the importance of music in Kittler’s work is often overlooked (2006b). Nonetheless, music is an infrequent but persistent 18 NOISE RESONANCE | INTRODUCTION

theme throughout his oeuvre; and exactly the decisive role of noise connects his writings on music with his more famous work on technical media. In “Musik als Medium” in 1995, Kittler wrote: “in the nameless name of noise, an other music must […] be invented—a music, the power of which is no longer derived from its relation to the medium of speech and its ‘meaningfulness’: pure media technology, pure control flow” (1995a: 99).7 More than ten years earlier, in “The God of Ears,” Kittler already suggested that this ‘other music’ will ultimately triumph (2015a: 12). In a way, my development of the concept of the noise resonance of sound reproduction should therefore be regarded as an attempt to further Kittler’s intriguing but undertheorised concept of the ‘other music’—the music of the media age. Following Kittler’s assertion that “media studies […] only make sense when media make senses,” I argue that in the age of technical media, studying music only makes sense when sound media make senses. The issue of the role of noise in sound reproduction, the role of all the disturbances, distortions, disruptions and interferences that cling to a signal during its transmission from sender to receiver, exactly touches upon the question how sounds produced by technical media continue to make musical senses (2006c: 55). I suggest that the sound of what Kittler calls the ‘other music,’ produced by technological filtering operations that fundamentally escape our analytical grasp, resonates with human listeners not despite but because of the inevitable noise of sound reproduction. Ultimately, I claim, this noise is not a transgressive disruption or subversive by-effect of sound reproduction technologies. By continuously affecting, changing and shaping the acoustic appearance of the world around us, it constitutes the very essence of the ‘other music’ of the media age. Besides an assessment of the role of noise in music, this thesis can therefore also be read as a more general account of the impact of technical media on the sound of music in the media age.

7 “Im Namenlosen Namen des Rauschens muß folglich eine andere Musik erfunden werden—eine Musik, deren Macht keinen Anleihen beim Medium Sprache und seinen “Bedeutsamkeiten” mehr macht: reine Medientechnik, reine befehlsfluß.”

NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 21

Chapter 1: “How much noise is necessary?” | a brief history of sound recording and noise reduction

1.1 Analogue 1: disc recording a) One hundred and forty years of noise In many ways, the one hundred and forty years of history of sound reproduction is a story of the fight against noise. From the earliest days of sound recording in the later half of the 1870s all to the present day, from the earliest wax cylinders to the most advanced digital equipment, random noise and non-random distortion have affected the recording and reproduction of sound signals; and strategies to prevent, reduce or eliminate these influences of recording, transmission, storage and reproduction media have been a major concern for inventors, developers, recording engineers and musicians. By looking into the many ways noise and distortion has been dealt with and thereby influenced the development of sound reproduction, this first chapter offers a comprehensive historical overview of sound recording technology in relation to the struggle with noise and distortion. As the beginning of the revaluation of the role of noise in recorded sound and music developed throughout this thesis, this historical framework is not meant as an exhaustive analysis of either the history of sound recording or the history of noise as such; others already contributed to both.8 By

8 The conceptual pre-history and early days of sound recording are extensively described in Sterne 2003. A historical overview of the entire history of sound recording can be found in Millard 2005 and Milner 2009. Hegarty 2008 provides a history of noise in and as music; and more or less everything there is to know regarding the conceptual history of noise can be found in Schwartz 22 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

assessing how the development of sound technology affected the conceptualisation of noise and, vice versa, how the conceptualisation of noise influenced the development of sound technology, this chapter aims to show that the discourse on the role of noise and distortion in sound reproduction is framed by concepts of fidelity, mediation and sound definition that took shape alongside the invention and further development of sound media. In broad strokes, this history of noise and its role in sound recording is divided into three phases. In the first phase, spanning the years up to the First World War that roughly correspond to the era of acoustical recording, the prevention and reduction of noise was primarily a matter of improving and refining the technology itself. In the second phase, the introduction of electricity in the 1920s enabled the employment of more specific noise reduction strategies that had already been developed in other fields of communication engineering such as telephone, telegraph and radio technology. This development continued in the 1940s and 1950s, when magnetic tape recording improved the overall quality of sound recordings but also introduced a high pitched noise called ‘tape hiss’ that required even more sophisticated noise reduction technologies. In the third and last phase, from the 1970s and 1980s onward, digital sound technologies seemed to have achieved the ideal of completely noiseless sound (re)production. Even theoretically perfect digitisation, however, ran into physical limitations that causes noise and distortion. Most strikingly, random noise is deliberately introduced to digital recordings in the form of so-called ‘dither’—described in detail in Section 1.4—that serves to remedy exactly these physical limitations of the digitisation procedure. The development of the attitude toward noise and distortion over the course of these three phases can thus be characterised as a shift from, firstly, coping with their inevitable appearance in the formative years of sound recording toward, secondly, the development of increasingly sophisticated methods and strategies to reduce their influence between the

2011. Furthermore, the changing definition of noise in the late nineteenth and early twentieth century is described in Wittje 2006, and an overview of the development of the concept of noise in theoretical physics can be found in Cohen 2005. NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 23

1920s and 1960s and, lastly, an unexpected return of noise in the form of so-called dither toward the end of the 20th century. The story in this chapter thus develops from ways to resist noise, via strategies to reduce noise toward processes that add noise. Along the way, I highlight two exemplary technologies to be analysed in more depth in Chapter Two. Firstly, the last part of Section 1.2 discusses so-called dual-ended analogue noise reduction technologies, exemplified by the systems developed by engineer and inventor Ray Dolby in the early 1960s. Secondly, Section 1.4 discusses the aforementioned practice of dithering in digital recording: the addition of minute quantities of noise to eliminate certain sonic artefacts introduced by digitisation. Overall, this chapter thereby provides a counter narrative to the idea that the history of noise in sound recording is a slow, but steady progression toward the ideal of entirely noiseless, transparent recordings, encapsulated by the myth of perfect fidelity that I will introduce in the next section. In Genesis, Michel Serres writes: “noise is needed for messages, sand is needed for stones, how much disorder is necessary for living beings, how much noise is necessary for history?” (1995: 132). As I understand Serres, he thereby asks how much noise and randomness are required to get a message across, to set the events of history into motion, to make an organism grow. Similarly, I ask the question how much noise is necessary to reproduce a sound, a voice, instruments, music. As a first step toward answering this question, this chapter puts the issue of the continuously appearing and reappearing noises of technical media at the centre of the history of sound recording. b) Noise in the era of acoustical recording Regardless how ground-breaking it was as proof of principle, Edison’s original 1877 phonograph, was, as music critic Roland Gelatt puts it in his 1954 book on the history of sound recording, “an instrument of crude design and dubious utility” (1954: 26). More recently, historian Andre Millard describes in America on Record: A History of Recorded Sound how audiences at early demonstrations of the phonograph “had to pay close attention to discern the faint noises coming from the vibrating diaphragm” 24 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

(2005: 26).9 Although Edison’s device was able to capture and (re)produce physical sound, as these accounts attest it was apparently not very good at it. With the acoustical or mechanical recording device Edison invented and presented to the general public in 1877, sound waves are captured by a horn and transduced into mechanical vibrations by a diaphragm. Following a “fine spiral groove impressed in [the] surface,” a needle attached to the diaphragm (the ‘stylus’) etches these vibrations into “a piece of tin foil” wrapped around a metal cylinder (Gelatt 1954: 20). During the recording process, “the stylus would move vertically, creating a so-called ‘hill and dale’ pattern in the trough of the groove. On replaying, the reproducing needle was to convert these indentations on the tin foil back into sound” by means of a second diaphragm intended for playback (21). The rotating cylinder was operated by a hand crank, which required the operator to keep a steady hand to ensure recording speed remained more or less constant. Hence, although the phonograph was able to record and reproduce sound, it left much to be desired in terms of its ability to reproduce sounds that could be picked out from the noises produced by the machine itself: the stylus on tin foil, the turning of the hand crank and the rotation of the cylinder. Within a few years after the invention of the phonograph and continuing all the way up to the introduction of electrical recording some forty years later efforts to prevent and reduce these noises were primarily aimed at improving and refining the reproduction process itself by changing the materials and improving the design and functionality of recording and reproduction surfaces, cutting stylus and recording and playback horns. The earliest of these improvements were developed by Alexander Graham Bell, who had been the first to patent a working prototype of the

9 Millard’s anecdote confirms Jonathan Sterne’s suggestion in The Audible Past that, at a time “when sound-reproduction technologies barely worked, they needed human assistance to stitch together the apparent gaps in the ability to make recognizable sounds” because, at this earliest stage of sound recording, listeners had to put some trust in the workings of the machine to believe it could indeed do what was claimed it could and be able to classify the faint sounds that emerged from the horn as bona fide sound reproductions (2003: 246). NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 25

telephone one year prior to Edison’s invention of the phonograph in 1876. Together with his cousin Chichester and fellow engineer and inventor Charles Sumner Tainter, Bell worked on improving the phonograph from about 1880 onward. In 1887, they presented a new prototype called the ‘graphophone.’ For the graphophone, Edison’s metal wrapped in tin foil was substituted by “cardboard coated with wax, in which the recording stylus engraved the pattern of its vibrations in narrow grooves” (Gelatt 1954: 34). Furthermore, the needle, which had been fixed directly to the diaphragm of the phonograph, was replaced by “a loosely mounted stylus which could more easily be guided by the record” (34). Because “the use of wax allowed for sharper, better defined recording,” Gelatt writes, these changes meant a first jump in sound quality (35). The newly designed stylus, Millard adds, amounted to “a more intelligible sound and a little less background noise” (2005: 30). During these experiments with the graphophone, Sterne describes in The Audible Past, Bell’s associate Tainter was one of the first to conceptualise the noises of recording and reproduction apparatuses as ‘external,’ separated from the reproduced sound. Tainter, Sterne writes, “had sought a kind of acoustic transparency in sound reproduction: ideally, the medium would disappear, and original and copy would be identical for listeners” (2003: 256). Given the mechanical limitations of the graphophone and the inevitability of noises occurring during recording and playback, Sterne argues this ideal could only be realised if listeners were trained to “separate foreground and background sounds, to treat the apparatus of sound reproduction as merely incidental to the sounds thereby perceived” (256). With this conceptualisation, Tainter introduces one of the most important frames regarding the position of noise in sound recording. By placing the recording and reproduction device “outside the universe of sound reproduction,” he positions the “noises made by the machine [as] ‘exterior’ or ‘outside’ sound” (258). As I will show in the next section, it is this logic of separating interior sound from exterior noise that would later be conceptualised by telephone and radio engineers in the 1920s and 1930s as the ‘signal-to-noise ratio’ and formalised and problematised in Shannon’s information theory in the late 1940s. 26 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

With Tainter’s conceptualisation of the perfect separation of internal sound and external noise, the elimination and reduction of noise became one of the leading concerns for the further improvement of sound reproduction technology. In 1887, Edison responded to the launch of the graphophone with an improved version of the phonograph that used cylinders of solid wax (Gelatt 1954: 37). That same year, inventor Emile Berliner developed a competing format. His ‘gramophone’ used discs instead of cylinders and applied a ‘lateral cut’ instead of the ‘hill and dale’- method of the phonograph and graphophone, etching the sound waves horizontally into the recording groove instead of vertically (Siefert 1995: 426). Initially, however, Berliner’s disc technology with horizontal grooves did not match the sound quality of cylinders and “suffered from a hissing surface noise” (Millard 2005: 47). As a consequence, cultural historian Marsha Siefert writes, the older “cylinders outproduced and outsold discs through 1911” (1995: 421). Nonetheless, after considerable improvement of disc technology due to the introduction of shellac discs in 1897 and the launch of user- friendly and easier to operate Victrola-players in 1901, the disc based format slowly took over the market (Gelatt 1954: 88). Toward the end of the acoustical era, in 1912, Edison introduced the ‘Edison Diamond Disc,’ his final contribution to the history of sound recording and the last major improvement in sound quality before the introduction of electricity. Although the Diamond Disc still relied on the vertical ‘hill and dale’- grooves, it used a stylus with a diamond point and introduced a “new material based on phenol resins […] called condensite”—a “hard, easily molded plastic,” which, according to Millard, “provided the best recording surface to date” (2005: 87). During the thirty five years separating the invention of the phonograph in 1877 and the launch of the Diamond Disc in 1912, historian of technology Susan Schmidt Horning argues, most attempts to improve sound quality and prevent or reduce the noise of recording and reproduction devices focussed on “improving the design of recording and playback equipment” (2013: 34). Besides technical improvements, strategies to reduce noise involved optimising the recording process itself, for instance by experimenting with, as Emily Thompson describes, the NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 27

proper positioning of musicians in the room, “the selection of different sizes and shapes of horns, and […] the arrangement of musicians with respect to the horn” (2002: 294). There were no objective standards to measure and compare recording and reproduction quality, however. In these first decades, recordings and technologies were judged according to a loose and constantly changing set of (aesthetic and technological) values based on technological advancements, the faith of listeners, engineers and musicians in the equipment and their expectation of and preference for certain sonic qualities. Although a concept of fidelity entered the discourse on sound reproduction quite early (according to Sterne it was “first applied to sound in 1878”), Sterne notes that “before sounds could be captured by electric devices for measuring signals,” that is before the start of the age of electrical recording in the late 1910s, “fidelity was an amazingly fluid term” (2003: 216, 221). In the first decade after its invention, when the phonograph was still primarily used as a device to record spoken language for business purposes, the term simply referred to the level of intelligibility of recorded words (Thompson 1995: 137). When, over the last decade of the nineteenth century, the phonograph, graphophone and gramophone were gradually appropriated and marketed for the recording and reproduction of music, fidelity came to refer to what Sterne calls “aesthetic preferences and tonal distinctions among sound reproduction technologies” (2003: 276). These preferences and distinctions were not measured according to some standardised system of reference, but established by means of the more or less subjective comparison, by inventors, engineers and regular listeners, of one recording or one technology to another. Hence, during the acoustical era, rather than the recording’s resemblance to an objective point of reference, fidelity connoted the difference in sonic quality between different technologies and different recordings. Although the prevention and reduction of interfering noise was conceived as a way to improve this quality and achieve a more ‘natural’ sound believed to be closer to ‘reality,’ “the rhetorical question was,” Siefert writes, closer to “which reality?” (1995: 443). In answer to this question, Sterne describes how, although early accounts of sound recording 28 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

show a persistent “desire to capture the world and reproduce it ‘as it really is’,” this desire was driven by an idealistic concept of mediation that presupposes an ontological difference between original and copies based on the idea of the reproduction device as a “vanishing mediator” (2003: 218). The concept of fidelity thereby supports the idea that complete similitude between original and copy can be achieved if every acoustic trace of the medium is eliminated. Despite this underlying idealistic concept of mediation, Sterne argues however, the notion of fidelity in these first decades of sound recording remained “a shifting standard for judging reproduction” that “had little to do with correspondences between reproduced sounds and sounds that existed outside networks of reproduction” (282). Although the concept of fidelity as being ‘true to’ something rhetorically suggests both an intrinsic relation and a fundamental difference between original sound and reproduced copy, in the context of sound recording practices “a set of procedures and aesthetics had to be developed to stand in for reality” in order for such a concept to hold (285). Hence, fidelity did not so much rely on an objective comparison of particular recordings to some original or to ‘reality’ in general but on a subjective comparison of one recording to another according to a changing set of technological, social and aesthetic values. This situation changed, however, with the introduction of electricity. c) New noises and measurements: electrical recording Developed in the 1920s, electrical recording achieved the biggest jump in sound reproduction quality since its invention forty years earlier. Its basic principles, according to which sound waves are transduced into an electrical current before they are etched onto the recording surface, and again before they are send to the loudspeaker, were already conceived and patented in 1903, but, as Gelatt remarks, “without condenser microphones and vacuum-tube amplifiers and adjuncts, its potentialities could not be recognized,” which is why it took more than a decade before more serious experiments were carried out (1954: 219). Because “workable microphones and amplifiers” were developed to serve the urgent need to improve the telegraph and develop a functioning radio system during NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 29

World War I, these technologies became available in the adjacent field of sound recording and reproduction in the years following the war (219). Hence, in 1919, engineers at Bell labs started to draw up “specifications for an electromagnetic recording head […] (based on analogous telephone and microphone research)” that were subsequently “translated into tangible equipment” (220). Although the fundamentals of sound reproduction remained essentially the same (sound waves are etched into a recording surface by a stylus and take the opposite route on playback), Schmidt Horning emphasises that

by using condenser microphones and vacuum-tube amplifiers, and devising an electromagnetic rubberline recorder and reproducer based on the principle of mechanical analogs of electrical filters that had been developed for telephone systems, [the engineers] eliminated most […] of the major problems associated with acoustical recording and reproduction (2013: 35).

After the introduction of microphones, singers and musicians did no longer have to cram around one or multiple recording horns to make sure their playing and singing was accurately captured by the recording device; and with the use of amplifiers, sounds could be recorded and played back much louder than what had been physically possible with acoustical recording (Gelatt 1954: 220). Furthermore, because electrical recording heads are able to pick up a much wider frequency range and thereby enable the reproduction of much lower and higher frequencies, sounds could now be reproduced with much more timbral detail (Malsky 2003: 244). “By 1927,” sound engineer and historian Oliver Read writes in 1952 in The Recording and Reproduction of Sound, electrical recording had almost entirely “supplanted the old method of singing, talking or playing directly into a horn” (1952: 15). Besides enabling more flexible recording methods for singers and musicians, microphones could also capture the “‘atmosphere’ surrounding music in the hall,” which many considered an important asset for the realistic reproduction of sounds 30 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

(Gelatt 1954: 223). Despite these clear advantages, however, there were also objections. “When sound is converted into an electrical signal via the microphone, then back to sound through the loudspeaker,” Schmidt Horning explains, “an inherent coloration, or distortion, takes place, depending on the characteristic of the microphone as well as the other controls and the speakers” (2013: 99). Due to this “inherent coloration or distortion” introduced by the electrical components in the recording chain, music journalist Greg Milner vividly describes, adversaries of electrical recording, among whom Edison himself, argued that the transduction of sound waves into electricity and back into sound puts an unacceptable distance between the sound source and the recording device (2009: 53- 56). These opponents of electrical recording considered the influence of all those microphones, cables, amplifiers, plugs and loudspeakers a detrimental corruption of the purity of the reproduced signal. “Whatever may have been the limitations of the acoustic method,” Milner quotes audio archivist and critic Walter L. Welch, writing in 1933, “the drawbacks consisted largely of what was left out of the recording process, instead of the present defects, which are added, namely excess resonance, distortion, over-amplification, and extraneous noises” (Welch in Milner 2009: 55). According to Welch and other purist, the problem with the new electrical method was not that it lacked something—for instance lower or higher frequency spectra, as with acoustical recording—but what it gained: more noise and distortion affect the output signal and thereby the fidelity of the reproduction in terms of its supposed faithfulness toward the sound source. Sterne argues, however, that the debate on whether acoustical or electrical recordings are closer or further removed from the sound source ultimately “is a purely semantic exercise” (2003: 277). Whereas advertisements at the time “promoted electric recording as progress, getting closer to ‘truth’ in reproduction, Read and Welch read it the opposite way” (278). In both cases, Sterne concludes, the concept of fidelity as a technologically objective factor is “confused with aesthetic preference” of one type of sound over another (278). This is not to say electrical recordings are not affected by distortions and noises introduced by the additional links in the recording NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 31

chain at all. Whereas acoustical recording is affected by surface noise or so- called ‘needle scratch,’ as well as by the influence of the diaphragm, recording horn, recording material and other characteristics of the mechanical apparatus, the transduction from sound waves to electricity and back makes the recording process more susceptible to interferences and new types of distortion and noise. Dealing with these effects of electrical recording, however, its technological origin in telegraph, telephone and radio systems was of considerable importance. With the introduction of electricity, the recording and reproduction process entered the domain of signal processing, which had made significant advances in the preceding decades, especially during the Great War. When electricity became the norm in recording studios, several methods, procedures and measurement standards that had already been developed for the analysis of signal transmission in the context of telegraph, telephone and radio communication—including ways to analyse and measure noise levels—could be applied to sound recording and reproduction practices. Thus, due to the transduction of sound waves into electrical signals, the new noises of amplifiers, microphones, transistors and cables that affect the signal during recording, transmission and playback went hand in hand with attempts to apply and further develop electrical signal processing for containing and reducing the effects of noisy transmission channels. As Mara Mills argues in “Deafening. Noise and the Engineering of Communication in the Telephone System,” it was exactly in the context of developing and improving the telephone and radio network in the 1910s and 1920s that, first, “[the] concept of noise as ‘unwanted’ sound— specifically masking desired communication—entered the scientific lexicon” and second, “electrical ‘perturbations’ in the atmosphere and in vacuum tubes became known as ‘noise’ because they first manifested as static and other such sounds on radio and telephone receivers” (2011: 123). Similar to the way Tainter classified the noise of recording and reproduction media as external, telephone engineers regarded static noise as “intrinsic to the medium but extraneous to the signal” (123). With this redefinition of the concept of noise in electrical signal processing it became possible to develop what earlier concepts of sound quality and recording 32 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

accuracy had lacked: scientific standards that can be technologically measured and objectively compared. In “Improving the Noise Performance of Communication Systems: Radio and Telephony developments of the 1920s,” professor of engineering Mischa Schwartz describes how radio engineers at the time conceived the so-called signal-to-static ratio to formalise the influence of noise on their systems (2009: 17). On the basis of this new ratio, expressing the level of the transmitted signal relative to the level of the background static, they began “to design systems to improve this quantity” (17). As these attempts progressed, the definition of noise in communication systems gradually widened and, Wittje writes, “by the mid-1920s,” a whole range of “random distortions” that occur in the transmission channels of electric systems were labelled as noise (19). Initially connoting specific interfering phenomena, Mills explains, over the course of the 1930s and 1940s, the gradual “redefinition of noise as ‘interference’” ultimately led to “its expansion to all categories of signal” (2011: 136). By the time the signal-to- static ratio was rebranded ‘signal-to-noise ratio’ in the 1930s, it had become a univocally accepted standard for expressing the background noise level in communication systems (Schwartz 2009: 18). Applied to the field of sound recording and reproduction, these new standards enabled better methods for noise reduction. In The Recording and Reproduction of Sound, Read describes various technologies used by recording engineers around and before 1950. In order to reduce “the noise components of disc records” that are “caused mainly by tiny irregularities in or on the record surface in the form of abrasives, grain, dust, etc." and which transmit “scratch vibrations to the stylus of the phonograph pickup,” one can for instance install “scratch filters in reproducing circuits” (1952: 48). Modified “Anti-Noise Modulation styli” are able to eliminate certain “noise patches” in the signal; and a so-called “Variable Reluctance Pickup […] eliminates a considerable portion of the frictional noises which ordinarily are transmitted from the record surface” (115-116, 152, capitals in original). To further reduce noise on playback, it is possible to install a “Noise Suppression Filter” in between the pickup and the preamplifier, which will “attenuate very little, if any, of the music signal while effecting a great reduction in surface noise” (398). Most of these devices are inserted NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 33

somewhere along the reproduction chain, thereby reducing the noise levels of already recorded material. More important on the long run, however, was the 1930s development of equalisation filters applied to deal with noise in specific frequency bands—especially with high frequency noise that became increasingly bothersome due to the improved ability to reproduce exactly these parts of the sound spectrum. This noise can be reduced, it was discovered, by installing an equaliser that increases “the amplitude of the higher frequency undulations in the record in such a manner that they are considerably higher than those created by the tiny irregularities in or on the record surface” (Read 1952: 49). By pushing the amplitude or, in other words, the volume of the higher frequencies, the signal becomes louder than the surface noise and is able to cover or ‘mask’ the noise during recording. Subsequently, to restore the signal to its original amplitude value during playback, one installs “an electrical network in the output of the pickup, which is the reverse of the recording characteristic” (40). According to Read, this method of noise reduction, which he calls “pre- and post-equalization” (also referred to as pre-emphasis/de-emphasis) “results in a substantial reduction of surface noise” (109). Given the importance of this principle for the development of the most effective analogue noise reduction systems in the 1960s, I will discuss this strategy in more detail in Section 1.2c.

1.2 Analogue 2: magnetic tape recording a) Sound definition: dynamic range and frequency response With the transition to electrical recording and the introduction of ways to measure, quantify and standardise sound recording, transmission and reproduction, the concept of sound fidelity lost some of its most subjective tendencies, while gaining a set of objective norms to measure sound quality and noise levels. Nonetheless, the more ideologically motivated aspects of the concept of mediation, according to which sound reproduction devices function as a “vanishing mediator,” stayed in place. Although signal processing enabled the standardisation of sound properties such as amplitude levels, frequency response and dynamic range, as well as the 34 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

development of verifiable and univocally accepted standards like the signal-to-noise ratio, more relativistic interpretations of the fidelity- concept remained. While the flexibility introduced by microphones, amplifiers and devices such as equalisers opened up a whole new domain of sonic manipulation, they also “changed the concept of ‘original’ or ‘authentic’ performance” altogether (Schmidt Horning 2013: 6). Although newly developed standards supposedly provided more objective measurements of the faithfulness of a recording to the original sound source that existed prior to recording and outside of the sphere of representation, the question arose what this so-called original actually is.10 In Audio-Vision: Sound on Screen, Michel Chion argues that not the subjective concept of fidelity but the concept of ‘definition’ provides an objective point of reference (1994: 98). The sound definition of a recording describes the extent to which objectively measurable parameters of the signal can be substantially differentiated from the background noise. Although, Chion writes, “high definition” is often “(mistakenly) taken as proof of high fidelity,” the (potential) definition of a record or recording system is based on the objective, technologically verifiable parameters of frequency response and dynamic range (98). Frequency response describes the range of frequencies that a system or record is able to reproduce without any distortion. For human listeners, a ‘perfect’ frequency response covers the commonly accepted full range of human hearing from around 20 to 20.000 Hz and allows listeners, Chion writes, “to hear frequencies all the way from extreme low to extreme high” (98).11 The

10 Regarding this difference between objective technological advances and the notion of fidelity, science historian Hans-Joachim Braun writes in his introduction to Music and Technology in the 20th Century: “As to the idea of fidelity and realism in sound recording, advances in the physical reproduction of sound have made it clear that such concepts are in fact chimeras. Indeed, improved sound reproduction technology has rather increased the difference between sound recording and sound reproduction than diminished it” (2002: 22). 11 As Mara Mills argues in “Deafening. Noise and the Engineering of Communication in the Telephone System,” this ‘normal range of hearing’ is itself the product of the pathologisation of and a corresponding concept of standardised ‘normal hearing’ produced over several years of psychoacoustic tests conducted in the context of optimising telephone systems in the 1920s and 1930s (2011: 130-136). NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 35

dynamic range of a system or recording, secondly, is the difference in amplitude (or volume) expressed in between the softest (weakest) and loudest (strongest) signal. According to the Encyclopedia of Recorded Sound, around the time of Edison’s Diamond Disc in the early 1910s, the best acoustical recordings were able to reproduce a frequency range from “about 1.000 to 2.000 or 3.000 Hz” ("Tone Tests" 2005). Electrical recording pushed this to around 8.000 Hz ("Disc" 2005). Commissioned by the British Royal Air Force during the Second World War, the Decca Recording Company developed “full frequency range recording” (ffrr) to enable the production of “a training record to illustrate differences between the sounds of German and British submarines” (Gelatt 1954: 282). Available for the general public in 1949, ffrr-records boasted a frequency range between 50 and 14.000 Hz ("Decca Record Co." 2005). By the 1960s, this range expanded from 30 to 15.000 Hz (Millard 2005: 313). Toward the end of that decade, finally, the combination of the vinyl LP and magnetic tape recording enabled the coverage of the full range of human hearing: from 20 Hz to 20 kHz ("Disc" 2005). The bandwidth of a regular Compact Disc, finally, ranges “well below 20 Hz, on out to 20 kHz” ("Compact Disc" 2005). Sound signals with the weakest amplitude or the lowest volume are drowned by background noise because they do not carry enough energy to be louder than the noise. Hence, the dynamic range of a system is directly related to the background noise level. This means that, in principle, a system’s maximum dynamic range equals its maximum signal-to-noise ratio ("Signal-to-Noise Ratio" 2005). Early equipment and record materials, the Encyclopedia of Recorded Sound writes, “yielded very low ratios,” which, for instance, in the case of “Emile Berliner’s hard-rubber discs […] did not exceed 6 dB” ("Signal-to-Noise Ratio" 2005). For regular shellac recordings during the electrical era, Millard writes, “a ratio of 30 dB was very good” and for vinyl records developed after World War II this range “rose to 55 and 60 dB” (2005: 204). Finally, by the digital age, “a good CD player” showcases “a S/N ratio of 96 dB”, which means the maximum difference 36 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

between the amplitude of the background noise and the loudest reproduced signal is 96 decibels ("Signal-to-Noise Ratio," 2005).12 This concept of definition, as described by Chion, does not contain the idealistic undertones of the concept of fidelity—presupposing the possibility of a “vanishing mediator” and the existence of some intrinsic relation between the output signal and an objective external reality. Instead, definition is based on technologically measurable and objectively comparable standards like frequency response and dynamic range. These are not measured against some supposed original or unmediated sonic reality and they do not refer to any aesthetic judgement. Because aesthetically, Sterne writes, there is no reason to assume “that wider frequency response is a necessarily desirable characteristic of sound reproduction” (2003: 277). Whether a wider frequency response or a larger dynamic range are desirable characteristics of recordings depends on all kinds of relative factors: the context in which the recording is played, the attention of the listener, the aesthetic preferences of musicians, engineers or listeners. In contrast to the objective frequency response and dynamic range, these factors differ per person, group and context and change over time. After the transition to electrical recording, attempts to further increase sound definition continued by developing new ways to reduce noise levels, extend dynamic range and enlarge undistorted frequency response. In the 1930s, Schwartz describes, wideband FM radio transmission achieved a suppression of noise that left most records far behind, proving that a wider bandwidth and the corresponding larger frequency range amounts to lower noise levels (2009: 20). During the same period, the film industry tried to improve the definition of soundtracks, something for which, as Millard writes, “the development of noise- suppression techniques was the critical part” (2005: 277). Using the pre-

12 As I will explain in detail in Section 1.4, the objective dynamic range and signal-to-noise ratio of a digital recording (which is 96 dB for a 16-bit digital file) can differ from the subjectively perceived dynamic range. With so-called ‘dithering’ and digital ‘noise shaping,’ mastering engineer Bob Katz writes, a 16-bit digital recording can “have a perceived dynamic range about as great as 115 dB” (2002: 51). NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 37

and post-equalisation-technique described by Read, “the frequencies where troublesome noise was found” were increased during recording and decreased again at playback (277). Crucially, Millard notes, as the steady increase of frequency response caused higher frequency noise to become more and more noticeable in the decades following World War II, these practices of “equalization and compression were to become important techniques in the reduction of unwanted noise in all types of sound recording” further down the line (277). After the invention of full frequency range reproduction during the war, the last major achievement in the era of disc recording was the development from 1945 onward of the microgroove disc, more commonly known as the LP, by Peter Goldmark at Colombia Records (Millard 2005: 204). In order to improve sound quality, Goldmark, “proposed to change a number of things: amplifier, record material, shape of the groove, cartridge and stylus, method of recording, the turntable drive,” thereby essentially redesigning every aspect of the recording and reproduction chain (Goldmark in Coleman 2003: 38). In the process, he “miniaturize[d] the groove in order to avoid distortion,” “change[d] the radius of the stylus” and switched “from sapphire to diamond” styli; he designed “new motors and drives” and a “new pick up arm,” developed a better recording material called vinylite or vinyl and, lastly, introduced new microphones and loudspeakers for improved recording and playback (Goldmark in Coleman 2003: 39). All in all, Gelatt concludes, the new system, which used “a turntable that revolved steadily at 33⅓ rpm and a specially designed pickup of extremely light weight […] afforded full-range reproduction free from rumble, and distortion” (1954: 292). Upon their launch in 1948, the new vinyl LP and corresponding record players provided the highest sound definition to date. They only rose to their full potential, however, along the 1950s and especially the 1960s in combination with another major innovation: magnetic tape recording. After the advance of electrical recording in the 1920s, the introduction of tape was the second most influential revolution in recording practices prior to the digital age. Not in the least because it also enabled the most effective analogue technological noise reduction systems. 38 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

b) Flexibility and higher definition: magnetic tape “Generally speaking,” Oliver Read wrote at a time when commercial systems for magnetic tape recording were still very new

magnetic recording machines are not particularly complicated. A magnetic material, such as wire or tape, is drawn past a recording head. As it passes through the head, the material becomes and remains magnetized. The amount of magnetization remaining in the material at each instant is governed by the impressed signal upon the recording head. In playing back, the magnetized material is drawn past a playback head. The varying magnetization which remains in the material induces corresponding voltages in the coil of the playback head (1952: 181).

As with electrical recording, the basic principles of magnetic recording were discovered long before their widespread application and commercial exploitation. Already in 1899, Danish inventor Vladimir Poulsen developed the ‘telegraphone’: a device based “on the ability of an electromagnet to create varying magnetic patterns in a piece of steel relative to the varying electrical impulses actuating it” (Gelatt 1954: 286).13 Because recordings by the telegraphone were very quiet, the technology would have required microphones and amplifiers that had not yet been invented. After these became available with the transition to electrical recording in the 1920s, further experiments to improve magnetic tape recording, especially the recording material or so-called ‘base,’ began in earnest. The result of these experiments, the German ‘magnetophone’

13 In “Voices out of Bodies, Bodies out of Voices: Audiotape and the Production of Subjectivity,” media philosopher N. Katherine Hayles notes that even prior to Poulsen, “as early as 1888 Oberlin Smith, at one time president of the American Society of Mechanical Engineering, proposed that sound could be recorded by magnetizing iron particles that adhered to a carrier” (1997: 76). NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 39

introduced in 1935, was, as Gelatt writes, an adequate dictating machines, but still “could not rival the phonograph when it came to reproducing the full spectrum of orchestral sound” (286). Over the course of World War II, however, the potential of magnetic tape came to fruition due to further inventions by German engineers. “With the purpose of reducing distortion and enhancing signal-to-noise ratio,” a so-called ac (alternating current) bias was added to the input signal, which means “a high-frequency alternating current, usually between 75 kHz and 100 kHz” was fed “to the tape record head along with the audio signal” ("Bias" 2005). Because this high-frequency bias-signal corrects “the nonlinearity of the magnetic recording medium,” it achieved a big reduction of noise and a significant increase in sound definition ("Bias" 2005). The improved magnetophones, using coated plastic tape as recording base, were employed by the German Wehrmacht for propaganda purposes, discovered by Allied forces toward the end of the war and subsequently brought back to England and the USA (Read 1952: 190). There, development continued and in 1947, as engineer and former employee Beverly Gooch recalls in his contribution to Magnetic Recording. The First 100 Years, the American Ampex Electronic Corporation introduced a machine whose recordings were able to rival the sound definition of traditional disc recordings (1999: 84). Some more years of development were directed toward determining the best magnetic coating material, optimal tape length and ideal running speed, as well as ways to prevent ‘print-through’—a problematic phenomenon by which “the signal from one layer of tape transferred to the adjacent layer of the tape reel,” irreparably ruining the recording (Schmidt Horning 2013: 107). By the mid-1950s, however, most of these issues were dealt with and the transition from disc to tape was all but completed. By that time, Gooch writes, “Magnetic audio recording had completely revolutionized the record and broadcasting industry. All records were mastered on tape, and radio broadcasters were exclusively using tape as a time-delay and programming tool” (1999: 90). The advantages of tape were many. Because the recording process avoids direct contact between recording mechanics (recording head) and recording surface (magnetised tape), the problem of surface noise all but 40 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

disappears. Furthermore, Schmidt Horning writes, the maximum frequency response of magnetic tape “was not limited by the inertia of mechanical parts [and] dynamic range was not limited by the dimensions of the groove” (2013: 106). Furthermore, the shift to tape meant yet another big leap in the flexibility of recording practices. With an uninterrupted thirty minutes run, recording sessions could last longer; and with the flexibility of the base material, techniques like splicing and editing, changing the recording or playback speed, reversing sounds or creating flexible reverb and echo effects became either possible or much easier (Gelatt 1954: 299). Most strikingly, magnetic tape provided efficient and flexible ways to conduct multi-track recording or so-called ‘overdubbing,’ which enables artists to slowly build a recording layer by layer, taping different parts or various takes of the same part separately and non-simultaneously. This unlocked a whole new set of studio practices and musical possibilities that were to define both avant-garde and popular music from the 1960s onward (Clark 1999: 92). However, with this new recording method and extra flexibility came new noises, interferences and distortions. Firstly, tape recordings still had to be converted to regular discs for commercial release. In combination with the newly developed vinyl LP, the greatest achievements of both formats—increased frequency response—turned into a disadvantage. Both magnetic tape recording and the two vinyl formats (Colombia’s 12” LP, as well as RCA Victor’s competing 7” single) achieved a significantly increased response in the higher frequency domain. Up to the development and widespread introduction of vinyl, so-called ‘tape hiss,’ a type of high- frequency noise caused by the movement of magnetised particles on tape, had gone largely unnoticed because older disc formats did not reproduce the frequency spectrum at which it occurs (Milner 2009: 132-136). With vinyl, however, especially at passages with an amplitude level too low for the reproduced signal to cover or ‘mask’ it, tape hiss became a nuisance (White 1996). Secondly, the creative freedom of multi-track recording was limited by the fact that tape hiss increases with each track mixed into the final recording. This was a minor problem with early two-track stereo recorders developed in the 1950s and early 1960s, but, as historian Mark H. Clark NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 41

writes in Magnetic Recording: The First 100 Years, three-track recorders were already available by the late 1950s and four-track was in general use around 1967 (1999: 95). “By 1968,” Clark continues, “8-track machines had become standard, with 16- and 24- track machines introduced shortly thereafter” (95). Even without eight-, sixteen- or twenty-four-track machines, however, the problem could be significant, because recording engineers and musicians developed the habit to “remix (or ‘bounce’)” a four-track recording down to two or three tracks in order to “free up a track for overdubbing” (Millard 2005: 298). This ‘bouncing’ increases the amount of noise as well. As acoustician F. Alton Everest explains in The Master Handbook of Acoustics, “when two tracks having equal noise levels are mixed together,” the tape hiss is 3,1 dB higher than a single track and increases with that amount with each doubling of tracks (2001: 456). “It is simply a matter of adding noise powers,” Alton Everest writes: with four tracks, the noise is 6,02 dB higher, with eight 9,03 dB and when sixteen tracks are used, the noise level has increased with 12,04 dB (456). Because of tape hiss, Clark explains, “the signal-to-noise ratio of the professional recorders was marginally adequate to meet the requirements for making the highest quality master recordings, particularly if significant rerecording took place” (1999: 94). Hence, despite magnetic tape’s potential increase in sound definition and promise of more recording flexibility, tape hiss posed a serious hindrance toward its general acceptance and further development. After the surface or scratch noise of wax roles, shellac and vinyl records that haunted sound recording from its very beginning, after the many mechanical noises of recording and playback apparatuses that had painstakingly been reduced by subtle engineering, and after all the transmission noises caused by the additional links in the chain of electrical recording, the hiss of magnetic tape introduced yet another noisy problem. “This problem,” Clark continues, “was addressed by Ray Dolby,” who developed a highly successful noise reduction system in the early 1960s (94). Combining the advances in electrical signal processing in the 1930s, which had been conceptually framed by information theory in the 1940s, with the recording flexibility introduced by magnetic tape in the 1950s, Dolby designed a so-called dual-ended noise reduction system—the 42 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

most sophisticated and effective noise reduction technology of the analogue era. c) Companding: dual-ended noise reduction Specialised noise reduction technologies are installed in the sound reproduction chain to prevent, reduce or eliminate the noise and distortion of technical media. They are not necessarily designed to serve the quest for ever higher sound quality and more and more opaque reproduction technologies following the myth of perfect fidelity. Instead, they are developed to deal with specific types of noise that appear along the process of sound reproduction and transmission. Saving sound technologies from their own internal enemies, noise reduction systems are introduced in the reproduction chain to maintain and increase the potential frequency response and dynamic range of a system.14 Devices installed at the beginning or end of the recording chain (just before the input signal is recorded or just before the output signal is played) are called ‘single-ended’ systems. To reduce noises that appear in the chain before recording—for instance the noise of microphones, amplifiers, effect modules, cables or electronic musical instruments—one can install an adaptive filter. Engineers David Miles Huber and Robert E. Runstein describe how such filters analyse and process the dynamic range and frequency spectrum of a signal in order to “break up the audio spectrum into a number of frequency bands, such that whenever the signal [amplitude, MK] level within each band falls below a user-defined threshold” and is drowned out by background noise, the width of that specific frequency band is gradually limited in order to “reduc[e] the noise content” (2010: 517). Similarly, besides such gradual limiting of the frequency range of a low amplitude segment affected by noise, a so-called “noise gate” simply cuts off the signal below or above a given threshold (517).

14 As literary theorist James Steintrager writes in “Speaking of Noise: From Murderous Loudness to the Crackle of Silk”: “the original goal of Ray Dolby was not fidelity of reproduction per se but rather noise reduction using technologies of companding, that is, the compression and expansion of signals” (2011: 257). The principles of this more advanced version of the pre-emphasis/de- emphasis technique described in Sections 1.1c and 1.2a are explained further on in this section. NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 43

At the other side of the chain, just before the output, filters can be installed to partly remove the noise of playback media. Most of the filters mentioned by Read in Section 1.1c—such as scratch filters and noise suppression filters—fall in this category. Similar devices can be applied to reduce noise in already recorded material. Like adaptive filters at the front of the chain, these devices mostly filter out noises by attenuating specific frequency bands. More recently, the advance of digital technologies enabled the quick analysis of entire sound spectra. these digital noise reduction systems, Huber and Runstein write, take “a digital ‘snapshot’ of a short snippet of the offending noise,” which “can then be digitally subtracted from the original soundfile or segment in varying amounts” (2010: 519). Up to a certain extent, such systems are thereby able to clean up old (or new) recordings that are heavily affected by noise.

Figure 1 Noise Reduction with Pre-emphasis/De-emphasis from Rod Nave, “Dolby Noise Reduction,” Hyperphysics, hyperphysics.phyastr.gsu.edu/ hbase/audio/ tape5.html#c1, accessed 15 Apr. 2013.

Contrary to single-ended noise reduction systems applied before or after the recording process, the system developed by Ray Dolby in the early 1960s—initially meant to primarily reduce tape hiss and later extended to deal with other noises as well—is a dual-ended system. Like the principle of pre- and post-equalisation or pre-emphasis/de-emphasis described in Sections 1.1c and 1.2a, dual-ended noise reduction systems work in two steps on both ends of the recording chain: before recording and before playback. As shown in figure 1, pre-emphasis/de-emphasis boosts 44 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

(‘emphasises’) the amplitude of specific (often higher) frequency bands, giving the signal enough energy to be louder than the noise and cover or mask it. At playback or during the mastering process, the same bands are restored to their original, lower, amplitude levels (‘de-emphasised’). Because the amplitude levels of both recorded signal and background noise are lowered or de-emphasised, the noise is significantly reduced and masked by the signal at playback (White 1996). In the early 1960s, Dolby refined this pre-emphasis/de-emphasis- process and developed a more sophisticated procedure generally referred to as ‘companding’ or ‘compansion’—a contraction of compressing and expanding or compression and expansion. In the terminology used to describe the process, like pre-emphasis/de-emphasis, a compander ‘encodes’ the signal at the recording stage and ‘decodes’ it during playback or mastering. Prior to recording, the dynamic range of specific frequency bands (again, often the higher regions of the frequency spectrum—where tape hiss is most prominent) is compressed, increasing the amplitude of these segments to makes sure their level is high enough to mask the hiss. When decoded, the process is reversed: the dynamic range is expanded and the amplitude of the affected signal restored to its original value; the noise levels are reduced together with the signal and the signal masks the noise (Huber and Runstein 2010: 515-16). Dolby’s most significant improvement upon earlier pre- emphasis/de-emphasis-techniques was his “principle of least treatment" (Dolby B, C and S, 2001). As the inventor himself puts it in a United States patent filed in October 1969, this principle means “no attempt is made to establish the required compression or expansion law by operating upon the whole dynamic range of the signal” (Dolby 1969). Whereas earlier forms of pre-emphasis/de-emphasis are applied to the entirety of a recording, Dolby’s system avoids processing unproblematic parts: passages that are already loud enough to mask the noise or frequency ranges at which noise is not problematic. To achieve this, technology journalist Paul White describes in an article on “Tape Noise Reduction,” Dolby’s first system, Dolby A, splits the audio signal “into four separate frequency bands […] each of which is processed independently and then added back to the original untreated signal” (1996). This system uses different frequency NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 45

bands and a dynamic range limiter to make sure the encoding and decoding is not applied to all frequency ranges and amplitude levels equally, but only to those segments that are problematic: low amplitude, high frequency passages. As Dolby Laboratories puts it a 2001 brochure: “least treatment essentially means that if there is no benefit to be gained by changing the audio signal, then don't change it” (2001). Between the mid-1960s and mid-1980s, Dolby developed multiple generations of dual-ended noise reduction systems. Dolby A (introduced in 1966) and Dolby SR (in 1986) were used in professional studios. Dolby B (launched in 1968), C (in 1981) and S (in 1990) were simplified versions using the same principles on consumer cassette tapes (Clark 1999: 106). Along the different versions, most improvements were achieved with the “principle of least treatment” and the use of variable frequency bands. Dolby A consisted of four frequency bands and reduced noise with maximally 10 dB, especially in the higher frequency ranges (Dolby SR 1987: 3). Because noise is generally less noticeable when it is spread out equally across the entire frequency band, however, the reduction of these high frequency noises focuses the listener’s attention on noise in lower areas of the spectrum. Consequently, Dolby Laboratories explains, “as you continue to reduce high frequency noise, the middle and low-frequency components of the tape noise become relatively more significant” (2001). This required the procedure to cover lower frequencies as well. In the last analogue system, Dolby SR, journalist Hugh Robjohns describes, there were “10 bands altogether, some operating at fixed frequencies and others moving automatically to suit the material” (2005). In their 1987 promotional brochure for the SR-system, Dolby Laboratories claims the system reduces noises and “other low-level disturbances” over the whole spectral range “by as much as 25 dB” (1987: 2). According to them, this means the SR-system effectively pushes all noise below the level of audibility, thereby producing what they call a “remarkable clarity of reproduction” (5). Evidently, about one hundred years after the Bell cousins and Tainter presented the graphophone, the latter’s conceptualisation of what Sterne calls a “vanishing mediator” still guided the general discourse on sound recording and noise reduction: “an ideal audio device or system,” Dolby’s brochure reads, “would impose no 46 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

audible limitation on the signal passing through it” (2). A perfect sound system should be fundamentally inaudible and the goal of noise reduction, the brochure suggests, is to approximate this ideal as close as possible. In the first half of Chapter Two, a close reading of the principles of dual-ended noise reduction will reveal how the discursive framework supporting this technology problematises the persistent myth of the vanishing mediator. For the moment, however, the second half of this chapter continues the history of the relation between sound reproduction technology and noise into the digital age. From the 1970s onward, the transition to digital technology seemed to suggest that the ideal of complete noise reduction was reaching its logical conclusion. As I will show in the coming paragraphs, however, on closer examination the digital revolution did not escape the continuous occurrence and reoccurrence of the noise of sound reproduction either.

1.3 Digital recording a) A short history of digital sound, 1926-1982 By the time Dolby Laboratories launched its last analogue noise reduction system in the spring of 1986, the market for sound reproduction technology looked very different from the time Ray Dolby demonstrated his first system in 1965. Four years earlier, Philips and Sony presented the first commercial digital sound medium, the Compact Disc as the latest chapter in a slow but steady takeover of digital sound technologies that began in the 1970s, took up steam in the 1980s and was more or less completed by the 1990s (Morton 2000: 172). The introduction of the CD was the culmination of many years of intensive research into digital sound technology, going as far back as the mid-1920s.15 The first and still most commonly used principle to convert analogue sound into digital signals, pulse code modulation (PCM) was “first outlined in a patent obtained by P. M. Rainey of Western Electric in 1926

15 For a concise history of the development and introduction of the Compact Disc, see Chapter Six in Milner 2009. NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 47

and elaborated by A. H. Jeeves, another telephone engineer, in 1937” (Millard 2005: 347). It was “based on the concept that a continuous signal could be reconstructed from isolated samples” that “could be approximated by discreet numbers” (347). With pulse code modulation, sound is turned “into a pulsating electric current that is measured and expressed as a binary code of digits” (348). This code represents the amplitude values of the pulsating current and is inscribed on a hardware medium, for instance as pits in a surface, spots of magnetic flux or in any other binary codification system. At playback, the code is read by a laser or a magnetic tape head, translated back into electrical pulses and subsequently transduced into sound waves. The first time the principles of pulse code modulation were taken beyond their theoretical conceptualisation was during World War II, when engineers at Bell Labs developed a system to codify telephone messages between the and the United States. As engineer and historian Thomas Fine describes in his article on “The Dawn of Commercial Digital Recording,” this “PCM-based encrypted-transmission system called SIGSALY […] was deployed in 1943” and “grew to 12 terminals before being retired in 1946” (2008: 3). After the war, in 1948, Claude Shannon’s information theory formalised and refined the theoretical principles underlining digital signal processing and extensive research into its possibilities was subsequently carried out throughout the 1950s and 1960s (Roads 1996: 10; Millard 2005: 346). By 1957, researchers at Bell labs conducted the first experiments with digital sound synthesis, proving that, as composer and computer programmer Curtis Roads writes in The Computer Music Tutorial, “a computer could synthesize sounds according to any pitch scale or waveform, including time-varying frequency and amplitude envelopes” (1996: 87). A few years later, in 1962, the first signal transmission system using pulse code modulation was installed by Bell Laboratories (Millard 2005: 347). After these early attempts in digital signal transmission and sound synthesis, the age of digital sound recording started in 1967, when engineers at the Japanese NHK Technical Research Laboratory developed a monophonic PCM recording system using a “videotape recorder as its storage medium,” followed by a stereo version in 1969 (Fine 2008: 3). 48 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

According to Fine, the first commercially available digital recording made with this Japanese prototype was a cover of hit ‘Something’ by musician Steve Marcus, recorded in 1970 and released in January 1971 (4). New and better digital recorders were developed throughout the 1970s. In the UK, the BBC implemented “a 13-channel PCM system” in 1972 (3) and in the US, the 1976 Soundstream digital recorder “allowed mixing of up to eight tracks or sound files stored on computer disk at a time” (Roads 1996: 13). “Two years later,” Roads continues, “the 3M company, working with the BBC, introduced the first commercial 32-track digital recorder […] as well as a rudimentary digital tape editor” (13). According to music historian Thom Holmes, the first digital synthesiser, the Synclavier, appeared on the scene around the same time in the mid-1970s (2002: 265). “By the beginning of the 1980s,” Fine writes, a short forty years after the initial conceptualisation of PCM and a little over a decade after the first successful digital recording in Japan, “all major record companies had embraced digital recording in one form or another” (2008: 14). It was this success of digital recording that motivated a consortium of Philips in the Netherlands and Sony in Japan to develop the Compact Disc as the first commercially available digital sound carrier. In part based on Philips’ earlier laserdisc technology, the CD represents binary code “as minute pits on the underside of a revolving disc, in which 0s were the pits and 1s were the unpitted surfaces of the disc. The laser beam moved across the disc from the center to the edge, and each pit reflected back the light in a different way” (Millard 2005: 350-351). Sony and Philips demonstrated the system in 1981 and released it to the general public in 1982 (351). Although, as Millard writes, until well into the 1990s “most popular recordings were still made on multi-track tape recorders” and only turned into a digital signals at the very last stage of the recording process, with the introduction and commercial success of the CD in the 1980s, the entire chain of sound recording, transmission and reproduction could in principle be carried out in the digital domain (356-357). The age of digital sound went into full swing. NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 49

b) Basic principles of digital sound recording In many ways, the advantages of digital sound recording echo those of electrical recording and magnetic tape: it offers increased sound definition and much greater flexibility for recording and manipulating sound and music. As Millard writes: “turning sound into a string of numbers provides an opportunity to manipulate the data just like a computer can manipulate the binary code of a document typed into its memory. […] It is possible to make hundreds of edits in a few seconds of digital sound” (2005: 356). Especially from the 1990s onward, the development of user-friendly digital recording consoles and accompanying software unlocked the potential of digital sound processing for an increasingly large group of recording professionals and amateurs. Furthermore, whereas analogue technologies always add a minimum amount of noise with every copy, accumulating more and more ‘generational’ noise the further a copy is removed from the master recording, this does not apply to digital copies. When the correct conditions are observed, “digital technology,” Roads writes, “can create any number of generations of perfect (noise-free) clones of an original recording” (1996: 21). In principle, every copy of a digital recording is exactly the same as the master recording. The most important reason for the sustained effort put into the development of digital sound technology from the late 1960s onward, however, was its potential for highly increased sound definition. With most of the noise sources of analogue recording removed, digital sound technology promises a frequency response and dynamic range that outreaches even the most advanced analogue system. With the standards used by the developers of the CD (16-bit files with a sample rate of 44.100 samples per second), digital recordings can easily reproduce the long sought-after frequency spectrum from well below 20 Hz to over 20 kHz, as well as a dynamic range of at least 96 dB ("Signal-to-Noise Ratio," 2005). Digital technology thereby issued in a new era of sound recording in which many noise-related problems of the analogue era seemed to have disappeared. Digitising sound, however, is a highly complex procedure that requires a specific set of engineering skills and the correct use, adjustment and calibration of specialised and sensitive technology. “Digital recording,” 50 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

Figure 2 The Digital Recording and Playback Chain from Curtis Roads, The Computer Music Tutorial, Cambridge, MIT Press, 1996: 23.

NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 51

audio engineers Francis Rumsey and Tim McCormick write, “converts the electrical waveform from a microphone into a series of binary numbers, each of which represents the amplitude of the signal at a unique point in time” (2009: 203). This conversion of an electrical waveform into binary code is carried out in two steps by a so-called analogue-to-digital (A/D) converter, shown in figure 2. Firstly, the signal is sampled into discreet bit of time that can be interpreted as “‘snapshots’ of the analog signal taken many thousand times per second” (208). Secondly, the voltage levels of the electrical signal, corresponding to the amplitude levels of the original sound wave, are measured and quantised. This means, Roads explains, the A/D-converter “converts the voltages into a string of binary numbers at each period of the sample clock” (1996: 22). This twofold operation of sampling bits of time and quantising voltage or amplitude values forms the basis of sound digitisation. Each sample contains a binary number representing the amplitude value of the signal at the moment of measurement. Combined in sequence, these samples—in the case of the CD 44.100 per second—“form a representation of the continuous waveform” (Rumsey and McCormick 2009: 211). At playback, Roads writes, the digitisation process is reversed and the binary numbers “are read one-by-one from the digital storage and passed through a digital-to-analog converter, abbreviated DAC (pronounced “Dack”). This device, driven by a sample clock, changes the stream of numbers into a series of voltage levels” (1996: 24). In short, at the end of the chain, the digital code is translated into an electrical current and subsequently transduced back into sound. Contrary to common belief, the parts of the signal that fall ‘in between’ the samples are not lost or unrecorded. Instead, as Roads writes, “given certain conditions, the missing part of the signal ‘between the samples’ can be restored. This happens when the numbers are passed through the DAC and smoothing filter. The smoothing filter ‘connects the dots’ between the discrete samples. Thus, a signal sent to the loudspeaker looks and sounds like the original signal” (1996: 27). On paper, provided that “certain conditions” are met, the digital procedure is able to create perfect reproductions. However, as shown in the next section, in reality, 52 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

exactly these conditions limit the physical implementation of such theoretical perfection. With digital technology the fight against the noise of material carriers and transmission channels, which marked sound reproduction technology from its inception onward, entered a new era. Because the digital procedure no longer relies on the physical inscription of continuous acoustic or electrical signals on a recording surface, but on the measurement of signal values and their symbolic representation in the discreet form of binary code, issues like surface noise, needle scratch, irregularities of the recording material, tape hiss and nonlinearity are no longer a problem. “When quality is independent of the [recording] medium,” engineer John R. Watkinson writes in “The History of Digital Audio,” “all the problems that beset analog recorders disappear”—that which resists complete reduction in the analogue realm (the physical noise introduced by storage, reproduction and transmission devices) is almost entirely absent in digital media (1999: 114). In 1951, at a time when the question how to define the digital and the analogue and understand the relation between digital and analogue devices was still very much up for debate, mathematician and computer pioneer John von Neumann discussed the issue of noise in relation to digital machines in his seminal paper on “The General and Logical Theory of Automata.” 16 “The real importance of the digital procedure,” von Neumann writes, “lies in its ability to reduce the computational noise level to an extent which is completely unobtainable by any other (analogy) procedure” (1963: 295).17 Like any analogue system, digital computers operate on the basis of circuitry that is susceptible to noise and physical

16 In the lecture “On Codes and Coding,” held at New York University in 2015, Bernhard Siegert describes this moment in the history of digital technology. It is discussed in more detail in Chapter Two, Section 2.5 (2015b). 17 Von Neumann’s article specifically discusses the difference between analogue computers that base their processing on “the intensity of an electrical current, or the size of an electrical potential, or the number of degrees of arc by which a disk has been rotated” and digital computers that are “representing numbers as aggregates of digits” (1963: 293-294). Regarding the issue of noise, however, I argue that this structural difference between analogue and digital computers can be extrapolated to the difference between analogue and digital machines in general. NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 53

interferences. The decisive gesture of digital technology, however, is the symbolical separation of these electronic processes from the calculations carried out on the computational level. This gesture enables the very efficient separation of digital signals from the inherent background noise of their physical carriers and transmission channels (294). Regarding the presence of noise, von Neumann argues, the fundamental difference between analogue and digital technology is therefore not qualitative, but quantitative (295). Although digital operations are based upon analogue processes, on a computational level, noise and signals can be separated to an extent that is structurally unobtainable by any analogue machine. In Cultural Techniques: Grids, Filters, Doors, and Other Articulations of the Real, Siegert compares this noise reducing quality of digital technology with the discreet operations of alphabetic writing. Following Michel Serres’ reading of communication theory, he argues that the “elementary cultural technique” of writing must be understood as “the filtering out of signals from noise” (2015a: 30). Because the material basis of written language (paper, pen, ink, etc.) is symbolically separate from the letters and words that are written down, the noise of these transmission channels does not interfere with the content of the transmitted message. In comparison, the material basis of analogue technical media always affects the inscribed, transmitted and reproduced signals, necessitating the development and implementation of mechanical adjustments, technological interventions and physical filters to prevent noise from distorting or overtaking the stored or transmitted message. The order of digital signals, lastly, returns to the discreet logic of written signs. In digital media, noise reduction is not an add-on filter applied before, during or after the recording process. Instead, the separation of signal and noise in the form of symbolically separating the level of analogue circuits from the level of digital computation is constitutive for the digital procedure itself. Hence, Siegert argues, in “the order of digital signals” the logic of filtering signals from noise “becomes nothing less than systemic” (2015a: 30). c) Digital problems: jitter, aliasing and quantisation error Given this systemic separation of signal and noise, digital technology is often considered to be the final word on the issue of noise in sound 54 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

recording. 18 As Millard writes with a pathos that echoes the jubilant enthusiasm of early advocates of digital sound:

Here at last was a system of recording in which there was no extraneous noise: no surface noise of scratches and pops, no tape hiss, and no background hum. The compact disc has a signal-to-noise ratio of 96 db, which in effect makes it noiseless recording. Audiophiles might have been aghast that music was now sampled rather than being recorded, but the results of the digital system were so astounding that it was difficult to find fault with it (2005: 353).

The physical reality of digital signal processing, however, is not as flawless as the theory suggests. Whereas the mathematical models that constitute the basis for digital technology indicate perfect reproduction, their implementation into physical hardware poses considerable problems, of which, jitter, aliasing and quantisation errors are the most prominent.19 The first of these, jitter, is caused by temporal irregularities in the sampling process. When the sampling of bits of time is not carried out in strictly regular intervals, audio engineer Jay Kadis writes in The Science Of Sound Recording, “small deviations in the timing will result in sample error,

18 In 1985, at the dawn of the age of digital sound recording, engineer Ken C. Pohlmann writes in the first edition of Principles of Digital Audio: “now the wait is over. With digital music one can at last listen to playback and begin to feel as if one is there—at the performance. High fidelity will have to be redefined as higher fidelity” (1985: 266). Ten years later, media scholar Stefan Heidenreich defines the advance of digital technology in his article “Rauschen, Filtern, Codieren– Stilbildung in Mediensystemen” as “das Ende des Leitbegriffs Rauschen” (1995: 24). 19 In Microsound, Roads describes the difference between theoretically ideal digitisation and its physical implementations as follows: “In contrast to the myth of ‘perfect reconstruction’ which pervades the mathematical theory of signal processing, the actual quality of all analysis-resynthesis methods is limited by the resolution of the input signal and the numerical precision of the analysis procedures. Distortions are introduced by numerical roundoff, windowing, peak-tracking, undersampling of envelope functions, and other aspects of the analysis” (2001: 273). NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 55

because the signal is changing with time and measurements made at incorrect times will have different values than samples made at the correct time” (2012: 145). The resulting jitter errors in the analogue-to-digital or digital-to-analogue converter introduce random noise and periodic distortions in the signal that cause considerable loss of sound definition. In most cases, however, careful calibration of the sampling clock keeps the risk of jitter to a minimum. The other two problems, on the other hand, are more structural and difficult to avoid. They are directly related to the basic principles of sampling and quantisation. Aliasing, firstly, touches upon the cornerstone of sampling theory: the sampling theorem, also called ‘Shannon-Nyquist’ or simply ‘Nyquist’- theorem, named after physicist and communication engineer Harry Nyquist who first described its basic principle in the article “Certain Topics in Telegraph Transmission Theory” in 1928, and Claude Shannon, who formalised the theorem in the context of digital sampling in “Communication in the Presence of Noise” in 1949. In short, the sampling theorem defines the minimum number of samples necessary to reproduce a given frequency spectrum without errors. In the words of Huber and Runstein, it states:

In order for the desired frequency bandwidth to be faithfully encoded in the digital domain, the selected sample rate must be at least twice as high as the highest frequency to be recorded (sample rate ≥ 2 × highest frequency) (2010: 204).

This means the minimum sample rate (amount of samples per second) of the digital reproduction must be at least twice as high as the highest frequency (amount of oscillations per second) of the reproduced signal. To adequately represent a signal and turn it back into a continuous sound wave without errors, the system must take at least two samples for each cycle of the highest frequency. Hence, following the Nyquist-theorem, a system using the standard sample rate of a CD (44.100 samples per second) can encode a frequency spectrum up to roughly 22 kHz. Not 56 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

coincidentally, this is some 2000 Hz over the commonly accepted upper threshold of human hearing (20 kHz). If, in defiance of the sampling theorem, signals are digitised with a sample rate less than twice the highest frequency, “the successive samples won't be able to accurately capture these higher frequencies, and will instead actually record false or ‘alias’ frequencies” (Huber and Runstein 2010: 204). Because the sampling process cannot faithfully reproduce the entire frequency spectrum of the original signal, Roads explains, “the reconstructed signal sounds at a pitch different from that of the original signal. This kind of distortion is called aliasing or foldover” (1996: 28). As shown in figure 3, frequencies that oscillate above the so-called ‘Nyquist’- threshold of half the sample rate will ‘fold over’ to lower frequency ranges, introducing frequencies in the output signal that were not present in the input signal. To make sure the Nyquist-threshold is not violated and prevent the occurrence of fold-over frequencies, an anti-aliasing filter is installed to cut off all frequencies above the Nyquist-threshold. In the case of the CD standard, all frequencies above 22 kHz should ideally be filtered out by a ‘brick wall’-filter. However, “in the real world, such a ‘brick wall’ doesn't exist” (Huber and Runstein 2010: 204). For reasons that will be taken up in much more detail over the course of Chapter Three, real-time filters cannot abruptly cut off a signal at some arbitrary threshold. They need time to process the signal, which causes parts of the frequency spectrum to sift through. Due to these limitations of physical filters, “a slightly higher sample rate must be chosen in order to account for the cutoff slope that’s required for the filter to be effective” (204). The difference between the clean cut-off of a ‘brick wall’ and the gradual slope of a real world filter is shown in figure 4. Improving anti-aliasing filters and developing other strategies that prevent aliasing, such as ‘oversampling,’ has been a major concern for engineers and manufacturers of digital equipment. Notwithstanding these efforts, however, Rumsey and McCormick note that even with much higher sampling rates “filter effects are unavoidable to some extent” (2009: 216). As such, the practical implications of the sampling theorem are the first NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 57

reason why the theoretically perfect sampling procedure is always compromised in physical reality.

Figure 3 Aliasing or Fold-Over from Francis Rumsey and Tim McCormick, Sound and Recording, 6th ed., Oxford, Focal Press, 2009: 215.

Diagram (a) shows a correct sampling procedure, with a sampling rate that is twice as high as the frequency rate. In (b) the frequency rate is higher than the sampling rate, so upon reconstruction, instead of the original frequency, a very different ‘fold-over’ frequency is reproduced: the dotted line.

58 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

Figure 4 Anti-Aliasing Filter: Brick Wall and Cut-off Slope from David Miles Huber and Robert E. Runstein, Modern Recording Techniques, 7th ed., Oxford, Focal Press, 2010: 207.

The Nyquist-theorem hereby defines one of the fundamental limits of digitisation. The sampling rate determines the frequency spectrum that a given system can reproduce and is therefore directly related to the maximum bandwidth or frequency response of a digital system. Furthermore, given the physical impossibility of manufacturing filters with an abrupt cut-off frequency, preventing aliasing will always introduce some filter effects in the reproduced signal, however slight. Notwithstanding these issues, however, sampling is, as engineer Ken C. Pohlmann argues, “based on well-understood principles” that yield what he calls “completely NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 59

predictable results” (2000: 47). Quantisation, on the other hand, is not even perfect on paper. Its problems are even more fundamental than the practical limitations imposed by the sampling theorem. Quantisation is the measurement of the voltage of a sampled signal and the translation of these measurements into binary ‘words’ that numerically represent the originals levels. Each binary digit (1 or 0) in such a word constitutes one ‘bit’ of information and the number of available bits, the bitrate, designates the possible length of each word. Hence, a larger bitrate equals more available bits and longer possible word length; and longer word length allows voltages values to be stored with more precision. Because larger bitrates store values with greater precision, the accuracy of the quantisation process increases with each available bit (Rumsey and McCormick 2009: 219). Unlike the sampling process, according to which the signal can be fully restored as long as the sampling theorem is not violated, Roads describes, “the values of the sampled signal cannot take on any conceivable value. This is because digital numbers can only be represented within a certain range and with a certain accuracy” (1996: 33). If, as is often the case, the numerical representation of the voltage level ideally requires a very long, sometimes infinite number value to be stored, the converter rounds off the binary word to make the stored value fit the available number of bits. This difference between actual voltage values and approximate quantised values causes so-called ‘round- off-‘ or ‘quantisation-errors.’ Like aliasing, quantisation errors cause distortion in the digitised signal. In the case of high amplitude signals with a complex frequency spectrum, for instance “something complicated like a symphony,” this error “sounds like noise” and “if the errors are large, then one might notice something similar to analog tape hiss at the output of a system” (34). Something different occurs, however, with relatively simple signals with a low amplitude level. Roads:

At very low (but nonzero) signal levels, quantization noise takes a pernicious form. A very low level signal triggers variations only in the lowest bit. These 1-bit variations look like a square wave, which is rich in odd 60 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

harmonics. Consider the decay of a piano tone, which smoothly attenuates with high partials rolling off— right until the lowest level when it changes character and becomes a harsh-sounding square wave. The harmonics of the square wave may even extend beyond the Nyquist frequency, causing aliasing and introducing new frequency components that were not in the original signal. These artifacts may be possible to ignore if the signal is kept at a low monitoring level, but if the signal is heard at a high level or if it is digitally remixed to a higher level, it becomes more obvious (1996: 36-37).

Thus, similar to tape hiss, the distortion caused by quantisation errors is most apparent in signals with a low amplitude level, but for different reasons. In the case of low-amplitude signals, over the course of many consecutive samples, the distortion caused by quantisation errors becomes something very unlike the more or less evenly distributed noise floor of analogue media. When the amplitude level drops, the error no longer translates into random, uncorrelated analogue-like noise, but into semi-periodic signals that are statistically correlated to both each other and the original signal. These unwanted harmonic or anharmonic artefacts are called ‘harmonic distortion’ (Rumsey and McCormick 2009: 223). Because with higher bitrates more exact values can be stored, the problem of quantisation errors and harmonic distortion decreases when the bitrate is raised. Higher bitrates, however, require more storage space and cannot be raised endlessly. Furthermore, because, as Pohlmann puts it, the “finite number of amplitude levels coded in the binary word can never completely accommodate an infinite number of analog amplitudes,” some portion of the quantised values will inevitably remain “an approximation of the actual” (2000: 32-33). Even more so, Kadis adds, “in theory, the more bits used to encode sample words, the greater our confidence in the NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 61

accuracy of the measurement and the better the sound quality,” in practice, however, not unlike the physical limitations of ideal anti-aliasing filters, “the physical process of conversion limits the real accuracy predicted by theory because real converters fall short of theoretical perfection” (2012: 145). Thus, although the risk of quantisation errors is minimised by higher bitrates, complete precision is structurally unattainable and a solution is required. At first sight, the most common solution to the problem of quantisation error seems entirely counterintuitive. It involves the addition of a minimum amount of noise to the reproduced signal. As Roads describes, in order to counter the negative effects of quantisation errors, sound engineers

introduce a small amount of analog noise—called dither—to the signal prior to analog-to-digital conversion. This causes the ADC to make random variations around the low-level signal, which smooths [sic] out the pernicious effects of square wave harmonics. With dither, the quantization error, which is usually signal-dependent, is turned into a wide-band noise that is uncorrelated with the signal. For decrescendos like the piano tone mentioned previously, the effect is that of a ‘soft landing’ as the tone fades smoothly into a bed of low-level random noise (1996: 37).

In the context of the long standing struggle against the problem of noise in sound reproduction technology and in light of the overall question regarding the role of noise in music, this introduction of random noise to remedy the problem of quantisation error and harmonic distortion is striking. In the second half of Chapter Two, I therefore analyse this strategy of fighting noise with noise in relation to the myth of perfect fidelity. Before 62 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

doing so, however, the final section of this chapter further explains the underlying principles and consequences of ‘dithering.’

1.4 Dither: fighting noise with noise a) Dithering 1: eliminating quantisation errors The story of the invention of dither is somewhat clouded, but both Pohlmann and Watkinson provide origin stories of which Pohlmann’s is the most interesting. According to him, the term dithering goes back to World War II, when “mechanical [analogue, MK] computers” that were used “to perform navigation and bomb trajectory calculations” on bomber planes, seemed to perform “more accurately when flying on board the aircraft, and less well on ground” (2000: 46). This happened because, Pohlmann continues, “the vibration from the aircraft reduced the error from sticky moving parts. Instead of moving in short jerks, they moved more continuously” (2000: 46). To be able to correctly calibrate the computers while the airplanes were on the ground, “small vibrating motors were built into the computers, and their vibration was called dither from the Middle English verb ‘didderen,’ meaning to tremble’” (2000: 46). Regarding the more specific context of quantisation errors, Watkinson describes how “dither was first recognized in connection with video quantizing in the 1950s, but the definitive treatment of dither and audio quantizing is generally considered to be that due to John Vanderkooy and Stanley Lipshitz, published in 1984” (1999: 112-3). ‘Dithering’ is the addition of a small amount of random noise (‘dither’) to a reproduced signal to cancel out the effects of quantisation errors. As recording engineer Nika Aldrich explains and figure 5 illustrates,” this noise “breaks up” and randomises the “statistical determinability” of quantisation errors (2002: 14). Due to this randomisation, the error no longer translates into harmonic distortion in low amplitude signals, but turns into a slight layer of random noise. In most cases, especially with bitrates under 20 bits, the A/D-converter explicitly adds dither noise to the signal prior to digitisation, but occasionally dithering already “happens naturally by means of the thermal noise within NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 63

the converters,” which renders the addition of extra noise unnecessary (9).20

Figure 5 Harmonic Distortion and Dither from Curtis Roads, The Computer Music Tutorial, Cambridge, MIT Press, 1996: 37.

The long spike on the left is a single frequency, quantised without dithering in the upper diagram, where the harmonic distortion caused by quantisation error is visible as shorter spikes on the right. Below, the same signal is digitised with dither: the harmonic distortion is gone and a minor random noise floor is added to the recording.

20 In Mastering Audio. The Art and the Science, mastering engineer Bob Katz nuances Aldrich’s statement: “every well-made 16-bit A/D incorporates dither to linearize the signal. If you were lucky enough to have a 20-bit or 24-bit A/D and 24-bit storage to begin with, then dither is probably not necessary during the original analog encoding. Although the inherent thermal noise on their inputs is not shaped to perfectly dither the source, current 20-bit A/Ds self-dither to some degree around the 18-19 bit level because of this basic physical limitation. Similarly, a transfer from typical analog tape probably has enough hiss to self-dither any transfer to 16-bits, as long as there is no digital processing before storage” (2002: 51). 64 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

However, dither noise is not only added during analogue to digital conversion. Each time the digitised signals is processed while mixing or mastering (editing the recording, changing volume levels, adding sound effects, etc.), the sample values are requantised. Due to the complex calculations required for even the simplest digital processing, these number values rapidly expand and, depending on the bitrate, introduce new round-off errors. Although the CD-standard of 16-bits is still the most commonly used bit-rate for commercially released digital audio, because of these requantisation errors, higher bitrates—up to 24-, 48- and sometimes even 96-bits—are often used for mixing and mastering. The extra bits provide more ‘head space’ to compute new number values without the need for constant dithering. When these 24- or 48-bit recordings are transferred to the 16-bit standard of CDs and other digital formats, however, they are ‘requantised’ or ‘truncated’ to fit the lower bitrate. At that point, dithering is necessary. Instead of the analogue noise added to a signal prior to digitisation, this digital dither used for requantisation consists of randomly generated bits, which means it is not pure random noise, but a noise-like digital signal. It is important to note that dither does not mask or cover harmonic distortions by superimposing higher amplitude noise, as if it were the reverse version of Dolby’s companding system. In the case of masking, the noise is covered by a louder signal or, conversely, a signal is covered by louder noise. The noise is not removed or eliminated, but drowned out and rendered inaudible. In the case of dithering, however, the statistically correlated artefacts of quantisation error called harmonic distortion are entirely eliminated. By decorrelating the error, it does not disappear, but is broken up and redistributed randomly (Rumsey and McCormick 2009: 227). The results in an artefact entirely different from harmonic distortion: a minor random noise floor. Because of its non-periodicity, Rumsey and McCormick argue, this noise floor “allows signals to be faded smoothly down” (226). Because “undithered audio signals begin to sound ’grainy’ and distorted as the signal level falls,” a “small amount of continuous hiss is considered preferable” (226). This random, evenly distributed noise floor of dithered recordings can be reduced further with the use of ‘noise shaping,’ by means of which, NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 65

Rumsey and McCormick write, “noise within the most audible parts of the audio frequency range is reduced at the expense of increased noise at other frequencies” (231-232). Instead of spreading the noise equally over the signal’s entire frequency range, “digital filtering is employed to shape the spectrum of the quantizing noise,” which “usually involves moving the noise away from the 4 kHz region where the ear is most sensitive and increasing it at the high-frequency end of the spectrum: (236). If necessary, companding techniques or single-ended noise reduction filters can be applied to reduce the noise level even further. All these approaches to bring down the noise floor notwithstanding, the occurrence of quantisation errors and the addition of dither puts audiophile dreams of vanishing mediators to sleep. This is confirmed by a second significant effect of dither: its positive influence on the perceived dynamic range of digital recordings. b) Dithering 2: stochastic resonance Analogous to the way that the sample rate of a digital signal determines its maximum bandwidth or frequency range (limited by the sampling theorem), its bitrate determines the precision of the amplitude values that can be stored. The bitrate of a digital system is therefore directly related to the dynamic range of the system. The more bits, the more precision toward low amplitude signals, the larger the recording’s potential dynamic range. As a rule of thumb, every extra bit amounts to approximately 6 dB in increased dynamic range, which is why a 16-bit system boast the aforementioned maximum dynamic range of 96 dB (Rumsey and McCormick 2009: 223). As I explained in Section 1.2a, with analogue media, the dynamic range is limited by the amount of background noise drowning out low amplitude signals, which is why effective noise reduction results in increased in dynamic range. In undithered digital recordings, these low amplitude signals are not affected by background noise, but by quantisation error. The signal-to-noise ratio of analogue recordings therefore turns into a signal-to-error ratio with digital recording. With dithering, random noise is reintroduced and can be referred to as signal-to- noise ratio again, or, as mastering engineer Bob Katz does, signal-to-dither 66 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

ratio (2002: 51). Without dithering, there is no background noise and a maximum signal-to-noise ratio of 96 dB, but harmonic distortion caused by quantisation errors renders the signal-to-error ratio significantly lower. As dither commonly accounts for about 3 to 5 dB of noise, the signal-to-dither ratio of a dithered 16-bit recording ratio is somewhere around 91 to 93 dB. With proper noise shaping, this ratio can be increased. But dither affects the dynamic range of digital recordings in yet another way. In a 1995 article, physicist Luca Gammaitoni explains how this second effect relates to a different problem of digital quantisation: the “loss of signal detail that is small compared to the quantisation step” (1995: 4692). Because digital systems can only represent value numbers “within a certain range and with a certain accuracy,” any amplitude value that falls below the minimum threshold of 0.5 will be rounded off as 0, representing nothing or silence (1996: 33). Amplitudes lower than the minimum quantisation step of one single bit, the so-called ‘least significant bit’ (LSB) (see figure 6), are no longer represented. The larger the bitrate, the smaller

Figure 6 The Least Significant Bit-Step (LSB) from David Miles Huber and Robert E. Runstein, Modern Recording Techniques, 7th ed., Oxford, Focal Press, 2010: 207.

NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 67

the LSB is relative to the total amount of available quantisation values: in a 16-bit system equipped with 65.536 possible binary values, the smallest quantisation step is much less significant than in an 8-bit system, which can only store 256 possible values. Nonetheless, regardless of the bitrate, without dithering, a certain amount of low amplitude detail is inevitably lost. In the case of this least significant bit, which either does or does not register a quantisation value, Gammaitoni writes, dither triggers a so-called “noise activated process” by pushing low amplitude signals that would otherwise go unregistered over the minimum threshold of 0.5 (1995: 4691). Without dither, a signal with an amplitude of 0.2 will be rounded off to 0 and disappear. With the added energy of dither noise, however, it is pushed over the threshold of half the quantisation step and registered by the converter (4692). Quoting a 1995 blog post on dithering by audio engineer Nigel Redmon: “by jiggling a signal that’s not large enough to cause a bit transition on its own, the added noise pushes it over the transition point for an amount statistically proportional to its actual amplitude level” (1996). With dither, more information is stored. Interestingly, the faint signals that dither pushes over the minimum threshold are generally quieter than the noise itself, but still loud enough to not entirely drown in it. In order to drown out a signal completely, engineer, mathematician and philosopher Abraham Moles explains in Information Theory and Esthetic (sic) Perception, “the intensity of the noise must be four to eight times stronger” than the intensity of the signal (1966: 83). Because many low level signals still carry enough energy to be picked up from the noise, however, dither “allows signals to be reconstructed even when their level is below the noise floor of the system” (Rumsey and McCormick 2009: 223). Consequently, Katz sums up, even when the “signal to noise (signal to dither) ratio will only measure about 91 dB,” a 16-bit recording with proper dithering can “have a perceived dynamic range about as great as 115 dB” (2002: 51). By pushing faint signals above a threshold of registerability to be processed by the digital system, dither increases the perceived dynamic range well beyond the maximum signal-to- noise ratio of 96 dB of an undithered recording. 68 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

In 2000, Stanley P. Lipshitz and John Vanderkooy (whom Watkinson credits as the first to systematically describe dither in the early 1980s) together with Robert A. Wannamaker (who wrote a PhD-thesis in applied mathematics on the topic of dithering in 1997) published an article in Physical Review E entitled “Stochastic Resonance as Dithering” in which they note that Gammaitoni’s aforementioned publication had been “the first […] to directly acknowledge the correspondence” between the principle of dithering and a physical phenomenon called “stochastic resonance” (2000: 233). At the outset of the article, this “stochastic resonance” is defined as “an unexpected increase in output signal-to-noise ratio observed in certain nonlinear static or dynamical systems as the noise level at the system input increases” (233). In other words: in specific cases, the addition of noise makes a phenomenon appear more clearly. Examples of this effect have been observed and put to use in fields as diverse as, Gammaitoni lists, “neurobiology (e.g., neuron firing), natural events (e.g., avalanches), laser systems (e.g., laser threshold), complex systems (e.g., bifurcations), chemical systems (e.g., activation threshold), and political sciences (e.g., electoral schemes)” (1995: 4691). In his book Noise, physicist Bart Kosko describes how, like dithering, the basic premise of this “threshold effect” called stochastic resonance comes down to the fact that “a faint signal can be too faint to cross a threshold and so adding noise energy can sometimes boost the signal above the threshold” (2006: 150). He illustrates this with four images of a baboon to which increasing amounts of random noise are added, reproduced here as figure 7. The first, entirely noiseless image (a) is hard to make out, but the addition of a little noise in the subsequent images (b and c) considerably improves the contrast and clarity. However, in the last image (d), the noise begins to overtake the image, indicating that the maximum threshold of beneficial noise has been crossed. This series of images thereby indicates that the positive effect of stochastic resonance in general and dither in particular relies on an ideal window of added randomness: enough noise to trigger a stochastic resonance effect and push the signal over a threshold, but not too much for the noise to overtake the signal. NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?” 69

“How much noise is necessary?” asks Serres. It is this question that both the reduction of the surplus of noise in analogue recording since the earliest days of sound reproduction and the unexpected return of noise in digital recording in the form of dither raise as well. As a first, tentative, step toward answering this question, the history of noise reduction and the continuous reappearance of noise and distortion in different forms and different places described in this first chapter suggests we should not, as the myth of perfect fidelity would have us believe, take noise as a by- product of the recording and reproduction chain, to be eliminated, masked, reduced or filtered out at any costs. Instead, the observation that digital technology, which reduces the material noises of sound reproduction to unprecedented levels, is marked by the return of noise in the form of deliberately added dither retroactively puts any seemingly progressive account of sound recording and noise reduction in a different light.

Figure 7 Stochastic Resonance from Bart Kosko, Noise, New York, Penguin, 2006: 151.

Ultimately, a different conceptualisation of the role of noise in sound recording requires an analysis that is no longer based on the myth of perfect fidelity. Paving the way for such an alternative conceptualisation, Chapter Two presents close reading analyses of the two examples that showcased contrasting attitudes regarding the role of noise in sound reproduction over the course of this first chapter: Dolby Noise Reduction and digital dithering. A critical assessment of these technologies and the conceptual framework that supports their operations provides a closer look at how the myth of perfect fidelity fundamentally shapes the discourse on technological sound reproduction. In the end, this analysis of two 70 NOISE RESONANCE | “HOW MUCH NOISE IS NECESSARY?”

technologies that deal with the presence of noise in seemingly contrasting ways opens a road to developing a different perspective on the relation between noise, sound and technical media.

NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 71

Chapter 2: Confronting the fuzziness of the real | Dolby and dither: concealing and revealing noise

2.1 Noise reduction reconsidered a) Dolby and dither: reducing noise, adding noise The historical overview in Chapter One suggests an on-going interaction between the development of technological sound reproduction and attempts to prevent, reduce or eliminate all the noise and distortion technical media add to reproduced sound signals. Regarding this interaction, narratives based on the myth of perfect fidelity privilege interpretations of noise that uphold its position as a negative factor in comparison to which the ideal of flawless sound reproduction is defined. Even assessments of the role of noise in sound recording that underline its potential for sonic and social subversion and transgression often implicitly assume its position as sonic Fremdkörper that forces its way into and disrupts, subverts or transgresses an ideally noiseless environment. It also suggests, however, that the presence of noise in reproduced sound is not incidental but structural and that the story of its prevention, reduction or elimination is not one of linear progression. On the contrary, the presence of noise, the fight against noise and the importance of noise for shaping the sonic characteristics of specific technologies and specific recordings have been a driving force in the development of sound reproduction technologies. In order to further this claim and assess its consequences for better understanding the role of noise in sound reproduction more generally, this second chapter presents two case studies of technologies that were highlighted in the previous chapter: 72 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

analogue dual-ended noise reduction and the practice of dithering in digital sound. Contrasting the active reduction with the active addition of noise, these two cases critically examine the dominant discourse on the role of noise in sound recording.21 Firstly, based on the ‘companding’ principle explained in Chapter One, Section 1.2c, Dolby’s noise reduction systems compress parts of the signal during recording and expand it at playback. They thereby cover the noise of magnetic tape recording with a louder signal. Whereas commercial developers of noise reduction technologies market it as a neutral procedure that generates “exceptionally pure,” “remarkably clear” and fundamentally noiseless signals, I argue that noise reduction is not only a technical filtering operation but presupposes a conceptual filter that construes what is considered information and what is considered noise prior to the physical operation (Dolby SR 1987: 1, 5). Based on idealised notions of noise and signal developed in the context of communication engineering in the 1930s and information theory in the 1940s, this conceptual noise filter retroactively defines noise in contrast to signal as everything that can and will be reduced. The analysis of the principles of dual-ended noise reduction thereby problematises the discursive framework underpinning this logic of concealing and revealing different forms of noise and signals, or what I call the conceptual logic of noise reduction. Secondly, although the principles of digital sound recording suggest a radical escalation of this conceptual logic of noise reduction, in the practice of dithering noise reappears in a surprising way. Whereas the digital procedure is flawless in theory, its implementation in the hardware of technological media is not devoid of error, distortion and noise. Instead, as I explained in Chapter One, Section 1.4, in order to, firstly, decorrelate quantisation errors that occur during digitisation and cause harmonic distortion in the digitised signal and secondly, trigger a stochastic resonance-effect that pushes low amplitude signals above the threshold of registerability by the converter, a small quantity of dither noise is added to

21 The parts of this chapter dealing with the analysis of Dolby Noise Reduction were published in an earlier version in the Journal of Sonic Studies (Kromhout 2014). NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 73

the signal. Furthering the critical assessment of the conceptual logic of noise reduction, I argue that dither not only points to the structural limitation of digital sound but also suggests that noise is important for the way listeners relate to sound recordings more generally. This analysis of dithering thereby paves the way for a conceptualisation of the role of noise that is not based on the conceptual logic of noise reduction. b) The domestication of noise As described in Chapter One, Section 1.1b, in The Audible Past, Jonathan Sterne debunks the myth of perfect fidelity by analysing the idealistic tendencies underpinning the concept of fidelity itself. From its earliest days, the discourse on sound reproduction has been framed by the idea that a reproduction should somehow be true to the original sound, presupposing a concept of mediation that assumes the intrinsic relation between recorded original and reproduced copy. Within this frame, undesirable differences between originals and copies are the result of forms of noise and distortion that are external to the sound but internal to the reproduction process. In practice, Sterne shows however, this concept of fidelity designates a more subjective set of technological, social and aesthetic values regarding the quality of one sound reproduction in comparison to another and not an objective standard of its level of truth toward a supposed original. Even after electrical recording enabled the development of objective, technologically verifiable standards for measuring the definition of sound technologies and reproductions in terms of their frequency response and dynamic range, attempts to reduce and eliminate the noises and distortions of sound reproduction technology retained some of the more idealistic tendencies of this myth of perfect fidelity. The implicit goal to create a vanishing medium that eliminates all sonic differences between original and copy remained; and in the 1940s and 1950s, this quest for higher definition led to the development of magnetic tape recording, vinyl records, stereo sound, the introduction of transistor amplifiers and Dolby Noise Reduction. By the early 1960s, this technological assemblage produced sound recordings that were able to (re)produce the long sought after standardised full range of human hearing (20 Hz to 20 kHz) while 74 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

maintaining a minimum noise floor that allowed for a dynamic range of 60 dB without major distortions or nonlinearity. In an analytical move that seems somewhat at odds with his analysis of the myth of perfect fidelity, Sterne, in his study of the (pre)history of data compression in MP3. The Meaning of a Format, defines this particular moment in the history of sound media as the “domestication of noise” (2012: 94). For him, the notion of domestication contextualises how a “group of approaches developed over the twentieth century” in psychoacoustics, information theory and communication engineering tackled the quantification, analysis, prediction and control of the noise of sound reproduction (52). Between the early 1910s and early 1960s, Sterne writes, noise in signal transmission was reconceptualised from an external thread to a regular part of “the world of sound and signal” (126). Depending on the context, it “could be masked and put in its place” (94-95). Either rendered useful or made irrelevant, engineers, musicians and listeners began to choose which of “a plurality of relationships to noise” suits their purpose (126). Ultimately, Sterne argues, this taming of noise by psycho-acousticians, computer scientists and communication engineers, ushered in an age of “sound reproduction after noise” (122). I agree with Sterne’s observation that interpretations of noise “far too often […] still fetishize noise as transgression or a challenge,” especially, as I explained in the Introduction, when it comes to its role in musical practices (124). As an alternative interpretation, however, I consider his concept of the “domestication of noise” too limited, because it is too uncritically informed by the information theoretical discourse that inspired its development in the first place. Describing the research on noise in the fields of psychoacoustics, information theory and communication engineering as a process of domestication and ‘taming,’ Sterne metaphorically likens noise reduction to bringing home a wild animal and making it harmless. Framing noise as a structural nuisance to be tamed by sophisticated technology designed by skilled engineers thereby implicitly reaffirms the concept of noise “as transgression or challenge” that Sterne claims to discard. By assuming that the importance of noise as a defining feature of sound reproduction—whether negative or positive—had been played out NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 75

at the moment it became technologically controllable, manageable and sometimes even functional, Sterne’s notion of the “domestication of noise” ultimately inscribes itself in the conceptual logic of noise reduction. As shown in Chapter One, Section 1.1b and c, this logic goes back to Tainter’s separation of the internal and external sounds of the graphophone in the 1880s and the development of the signal-to-noise ratio by telephone and radio engineers in the 1920s and 1930s. In both cases, the conceptual separation of signal and noise and the gradual expansion of the definition of noise beyond random interferences in telephone, telegraph and radio channels were inspired by the practical concerns of communication engineers (Mills 2011: 136; Schwartz 2009: 18). On the basis of these practical concerns, Claude Shannon, working at Bell Laboratories, developed his “Mathematical Theory of Communication,” in which he formalised and generalised the processes of signal transmission by introducing statistically defined concepts of information and noise (Shannon and Weaver 1964). c) Communicational noise: the concept of noise in information theory Published in 1948 by Bell Laboratories, Shannon’s theory mathematically describes the model shown in figure 8. In this model, noise is no longer conceived as an external disturbance but considered internal to the communication system. On the one hand, this repositioning allows Shannon to show how one can calculate the amount of noise that accumulates during a transmission and design more effective ways to deal with its influence. On the other hand, however, this reconceptualisation of noise as an inherent part of every communication system also proofs that complete noise reduction is fundamentally impossible (Sterne 2012: 81). Shannon’s concept of noise is directly related to his concept of information, because both are based on the statistical analysis of information transmission. Rather counter intuitively, Shannon defines information on the basis of the improbability of the content of a message. Because the content of a very complex message is relatively hard to predict and less probable to occur than the content of a simpler message, the amount of information it contains is statistically defined by this level of improbability (Ballard 2008: 84-88). Analogous to the use of the term in 76 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

thermodynamics, where it designates the level of molecular disorder in a system, Shannon calls this level of improbability the entropy rate of a message. The more improbable and harder to predict the content, the higher the entropy rate of the message and the more information it contains. Conversely, the percentage of the message that is more likely to occur, easier to predict and thus lower on information is called its redundancy rate. A high redundancy rate equals a low entropy rate and vice versa.

Figure 8 The Shannon-Weaver Model of Communication from Claude Shannon and Warren Weaver, The Mathematical Theory of Communication, Urbana, University of Illinois Press, 1964, 7.

Significantly, the comparison between Shannon’s use of the concept of entropy for defining information and the thermodynamic definition of entropy as the level of molecular disorder already indicates that messages with a high entropy rate and high information level are more complex, more disorganised and, indeed, more random. In contrast to uneventful, repetitive, periodic events, a random event is hard to predict and thus contains a lot of information. As a consequence, Kittler points out, the statistical definition of information is very similar to the physical definition of random noise (2013c: 167). Because the highest amount of information statistically equals the highest level of unpredictability, random noise— highly unpredictable by definition—can also be interpreted as information. NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 77

Hence, in information theory, noise and signal cannot be unequivocally distinguished. From the perspective of the receiver, Warren Weaver points out, owing to the statistical equivalence between information and noise, any disturbances added to the message during its transmission from a to b increases its complexity and raises the information level (Shannon and Weaver 1964: 18-19).22 Following this logic, information theory no longer defines noise on the basis of physical characteristics or even its internal randomness, but solely as those aspects of a message that were not intended by the sender.23 Everything that changes the message during the transmission and thereby increases its entropy rate constitutes as noise. This renders the concept of noise in information theory fundamentally relative to the context of transmission: following the logic of information theory, noise is defined by the information it disturbs. Depending on the context, everything can be interpreted as noise. On the one hand, this communicational concept of noise originates in the practical concerns and observations of communication engineers dealing with the problem of random interferences in transmission channels. On the other hand, however, the conceptualisation of noise in information theory is put at considerable distance from this straightforward, physical definition of noise as a random physical disturbance. Shannon’s statistical approach to noise and information renders their definition entirely relative to the context of a message and the conditions of its transmission. Hence, although it originates in the context of the concerns of 1930s communication engineering, the concept

22 “Certain things,” Weaver writes, “are added to the signal which were not intended by the information source. These unwanted additions may be distortions of sound (in telephony, for example) or static (in radio), or distortions in shape or shading of picture (television), or errors in transmission (telegraphy or facsimile), etc. All of these changes in the transmitted signal are called noise” (Shannon and Weaver 1964: 8). 23 As Abraham Moles, of whom I will come to talk more in Chapter Three, explains this primacy of the sender in Information Theory and Esthetic Perception: “there is no absolute structural difference between noise and signal [in information theory]. They are of the same nature. The only difference which can be logically established between them is based exclusively on intent on the part of the transmitter: a noise is a signal that the sender does not want to transmit” (1966: 78-79). 78 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

of noise in information theory is no longer confined to phenomena with the random characteristics of sonic or physical noise. Instead, what is to be considered noise and what can be consider information is discursively framed by the statistical logic of information theory itself. In “Financial Noises: Inclusion and the Promise of Meaning,” German sociologist Urs Stäheli therefore argues that, notwithstanding Shannon’s mathematical proof that it can never be fully reduced or eliminated, noise in information theory is still considered a “disturbance of a signal, which, in the worst case, makes it impossible to receive any informational value” (2003: 245). Although Shannon’s reconceptualisation of noise as internal to the model of communication turned it into an intrinsic part of the communication system, this repositioning still served the explicit goal of minimising its influence on the transmitted signal. Exactly by including it in the model of communication, Stäheli argues, “classical communication theory always already knows where to locate noise” (246). Pursuing this same line of reasoning, I argue that information and communication theory always already presuppose the possibility of complete noise reduction. This presupposition is what defines the conceptual logic of noise reduction. In 1964, at the beginning of what Sterne calls the age of “sound reproduction after noise,” Ray Dolby’s noise reduction systems hit the market. The technological and discursive frameworks supporting these systems share several notable points of reference with those that frame the reconceptualisation of the concept of noise in information theory. As shown in Chapter One, Section 1.2c, the development of Dolby’s dual-ended noise reduction is technologically rooted in the turn to electrical recording and the appropriation of signal processing and noise filtering technologies from other communication technologies after World War I. These technologies opened up the possibility to measure signal and noise levels, apply objective standards like the signal-to-noise ratio and develop single- ended and early dual-ended noise reduction technologies. Secondly, following World War II, the introduction of magnetic tape provided the material basis for a more flexible version of the already existing pre- emphasis/de-emphasis method, and information theory subsequently provided the conceptual framework for Dolby’s companding procedure. NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 79

The discursive framework supporting the technological operations of Dolby’s noise reduction systems is therefore exemplary for the more general conceptual logic of noise reduction that dominates the discourse on sound reproduction.

2.2 Things you want and things you don’t want a) Dual-ended noise reduction 1: concealing noise At the start of every sound reproduction chain is the first input: an electrical instrument plugged into an amplifier, a microphone capturing someone singing or playing an acoustical instrument or some electronic device plugged into the recording apparatus. On the other end of the chain, an output signal is physically inscribed on a record, tape, hard drive or some other medium—in most cases first on a master copy and subsequently on whatever format is used for commercial release. Ultimately, the sound stored on these carriers is played through loudspeakers or headphones and reaches its final destination: the listener. Along the way, each link in this technological chain between initial input and listener affects the signal: each device, each cable, each acoustic space, each plug, each instrument, each electrical circuit adds specific characteristics to the signal; sometimes very clearly, sometimes very minute. These effects are either audible changes in the spectral and temporal character of the signal or entirely new sonic objects that appear at the output of the reproduction chain. Analogous to the use of the term in the natural sciences (where it designates experimental results that are produced by the measuring apparatus or test procedure itself) these additions to the signal are called ‘sonic artefacts.’ 24 They are sounds

24 A blog post on the website of the Chicago School of Media Theory cites the following definition of artefact in the English Dictionary: “[i]n technical and medical use, a product or effect that is not present in the natural state (of an organism, etc.) but occurs during or as a result of investigation or is brought about by some extraneous agency” (Zao 2007). In the case of artefacts of sound recording, Huber and Runstein for example write about “artifacts such as tape hiss, hum, obtrusive background ambience, needle ticks, pops and even certain types of distortion” (Huber and Runstein 2010: 518). 80 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

entirely brought forth by and belonging to the links in the recording and reproduction chain itself. In the case of Dolby’s companding (compressing/expanding) process, Kadis describes how the “improper adjustment” of the equipment “will result in artifacts created by the noise reduction systems” (2012: 131). Similarly, Rumsey and McCormick warn for the wrong alignment of the settings used for compression and expansion, because the recording will “sound […] overbright and with fluctuations in HF [high frequency, MK] level” in case they do not match up (2009: 192). Dolby Laboratories describes the same issue in the 2001 instruction manual for the Dolby B, C and S systems (used on commercial compact cassette players). They remind the user that if cassette tapes recorded with Dolby Noise Reduction turned on are played back “without any Noise Reduction decoding,” they will “sound brighter” (Dolby B, C, and S 2001). When they are not expanded and restored to their original amplitude upon playback, the compressed high frequencies will sound louder. When you hear these higher, over- bright frequencies, the brochure states, “you are hearing the encoded sound, not the original” (2001). As critical theorist Mark Nunes argues, errors like these reveal “not only a system's failure, but also its operational logic” (2011: 3). What Dolby Laboratories call the original is not the input signal prior to recording, but the doubly treated signal: an encoded and decoded recording, recorded and played with Dolby Noise Reduction turned on. This act of switching system on and off and the over-bright sound of the erroneously executed or incomplete compansion process, however, are testament to an underlying logic of concealing and revealing signals and noise. When the reduction is turned on during recording and playback, the listener hears what Dolby Laboratories calls the original. When it is turned on during recording but turned off at playback, the compressed, encoded, brighter version is revealed. The very existence of this encoded version, however, changes our understanding of the so-called original—the encoded and decoded version—as well. When noise reduction is turned off at playback, the compressed, over-bright, encoded version of the recording is revealed. When it is switched back on, tape hiss and other background noises are concealed and the so-called original is revealed. Hence, what NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 81

Dolby Laboratories calls the original is a recording produced by the noise reduction system itself. Dolby’s original is not the entirely uncoded signal prior to recording, nor the compressed signal prior to playback, but the encoded and decoded signal. Produced by the noise reduction process itself, this original could therefore be considered an artefact of correctly applied noise reduction. Following Nunes, this begs the question what this ‘correct’ artefact reveals about the operational logic of dual-ended noise reduction and the input and output—noises and signals—of this filtering operation. b) Dual-ended noise reduction 2: revealing silence Long before technological noise reduction filters were available, Charles Sumner Tainter relied on the corrective abilities of listener’s ears and brains for his ideally transparent sound reproduction system to work. Listeners separate the sound on the graphophone record from the sound produced by the graphophone—internal sounds from external noise—and listen through instead of to the noise. As Greg Hainge explains in an article from 2007: “in Tainter’s experiments, the listening subject integrates noise into a successful and seemingly noise-free, high-fidelity sonic expression [and] becomes a kind of agent of noise reduction” (2007: 37). Hence, before the development of technological noise reduction systems, the ears and brains of listeners already functioned as a cognitive noise filter in and of themselves.25 This means that, instead of cleaning or clearing up a sonic environment that listeners experienced as inherently distorted, dual-ended noise reduction systems introduced a new type of sonic environment. In an article about the introduction of Dolby Noise Reduction in cinema soundtracks in the early 1970s, Michel Chion describes how noise reduction created a type of silence that audiences had never experienced or expected (1999: 108). One can assume that listeners at the time initially noticed this silence and recognised the absence of a background noise that

25 As Stan Link puts it: “Listeners learned to ‘hear through’ noise. The dust and nicks on vinyl recordings, amplifier hum, or speed inaccuracies of tape mechanisms produced types of noise that were basically as predictable as potholes on a familiar road” (2001: 36). 82 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

had always been there but which they had routinely filtered out of conscious perception. Once this background noise was technologically reduced and exchanged for a deeper silence, listeners might have retroactively realised the noise had been there all along. As they grew accustom to the silence, however, it no longer signified the absence of noise but only the presence of silence. Generic background noise became generic background silence and although this silence is predicated upon the operations of technological noise reduction, it ceased to signify the absence of noise. Sound scholar Mack Hagood’s describes a similar silence in his analysis of the ‘Bose QuietComfort Acoustic Noise Cancelling Headphones.’ Equipped with “tiny microphones and signal processing to produce an out- of-phase copy of the aural environment in an attempt to negate its phenomenological existence,” the headphones separate, Hagood quotes Bose-founder and inventor Amar Gopal Bose, “things that you don’t want from things that you want” (2011: 573, 575). They thereby enable the user to technologically insert a background silence at places where background noise is abundant. In order to do so, the headphones do “not merely block out the aural world but mediatize[s] it in order to cancel it out” (568, emphasis in original). Noise-cancelling headphones, Hagood argues, do not only shut off the ears from noise. By actively recording the background noise and cancelling it out with an out-of-phase copy (a mirror image of the noise spectrum) they actively block everything that intrudes the sonic space of the listener. Similar to switching-on and -off Dolby Noise Reduction, “the power button” of noise cancelling headphones, Hagood writes, “offers an (imperfect) on-off interface with the soundscape” revealing an attempt to produce an idealised, noiseless sonic environment (575). This on-off gesture pervades the silence created by noise reduction or cancellation with some kind of active agency. The silence is not the result of doing nothing or the absence or suspension of action. The silence of noise cancelling headphones and Dolby Noise Reduction does not signify the absence of sound. Instead, based on the active masking or cancellation of background noise, this silence is carefully constructed and continuously maintained as long as the device is turned on. As soon as it is switched off, NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 83

silence disappears and noise reappears. By switching-on and -off the noise, cultural theorist John Mowitt observes, noise reduction “operate[s] according to a systematic logic that produces information out of suppressed noise” (1987: 194). The silence thereby gains as much significance as the reproduced sound. Hence, by concealing noise and revealing silence, Dolby’s compansion actually produces the so-called original itself. This twofold operation of concealing and revealing noise and silence resonates with Heidegger’s notion of ‘enframing’ (Ge-stell) as laid out in “The Question Concerning Technology.” The ‘enframing’ of technology characterises the way that technology is always, as Heidegger puts it, “challenging forth into the frenziedness of ordering” (1977: 33). In order to enable this process of technological ordering, he argues, nature must be conceived as a “as standing-reserve” (Bestand), a set of resources that are readily available to be formed, used and changed in accordance with the requirements of technological progress (Heidegger 1977: 23; Ruin 2010: 192). In order for nature to be such a “standing-reserve” the science of physics unravels and demystifies the natural world to make it “orderable as a system of information” (23). For Heidegger, this mutual interdependence of technology, nature and physics implies that, instead of technology depending on the laws of nature and physics, nature and physics are framed in such a way as to serve the “challenging forth” of technology (23). Nature serves physics and both physics and nature serve technology as a means for man to control and order the world, turning it into manageable and coherent information (Ruin 2010: 186, 191). The concept of ‘enframing’ sheds light on the way noise reduction is engaged in a process of concealing noise and revealing silence, thereby creating the suggestion of increased information by increasing the signal- to-noise ratio of the recording. Rather than signifying the absence of, to use Bose’s terminology, “things that you don’t want,” noise reduction continuously reveals “things that you want.” Through its twofold operation of compression and expansion, the active process of technological noise reduction conceals what is deemed outside of information and reveals a silence that is said to be part of the original recording. The technological operations of noise reduction thereby presuppose a conceptual filter that 84 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

already define ‘what you want’ and ‘what you don’t want’ prior to the actual filtering operation. This conceptual logic of noise reduction negatively defines noise in contrast to everything that is not reduced, as everything that can and will be reduced. c) Something sticks to the signal: the real of sound recording In the last of three case studies in the article “Die Geburt der Literatur aus dem Rauschen der Kanäle,” Berhard Siegert discusses how directors of German radio plays in the 1950s employed an “on-going exclusion of noise” to rid their plays from any influence of recording and transmission channels (2007: 32). 26 By excluding as much noise as possible, they maximised the signal-to-noise ratio of the sound recording. The resulting silence, Siegert writes, created the possibility to dramatically “stage a silent pause as the expression of an absolute interiority” (34).27 In the absence of noise, the silence in between words and sounds gained greater significance. As such, the “on-going exclusion of noise” in these radio plays (produced around a time that the introduction of magnetic tape recording and the development of more sophisticated noise reduction filters began to enable greater sound definition) prefigures the logic supporting Dolby’s systems in the 1960s. Epistemologically, the silence produced by this exclusion of noise is reminiscent of the inwardness of the Romantic reading subject that Kittler in Discourse Networks 1800/1900 identifies as the leading paradigm for literary reading in the pre-technological era, when the supposedly clear meaning of written words was framed by a practice of silent reading (1990: 161). 28 The technological exclusion of the noise of recording and

26 “Einer fortschreitenden Ausschließung des Geräusches.” 27 “Ein Schweigen inszenieren als Ausdruck einer absoluten Innerlichkeit.” 28 Regarding this practice of silent reading, translator David Wellbery writes in his introduction to Kittler’s Discourse Networks 1800/1900: “Finally, hermeneutics draws on and ratifies a specific rendering of linguistic materiality, the myth of the silent inner voice that Derrida has described as foundational for the modern philosophy of the subject. In Kittler’s analysis, however, this myth appears less as a philosophical hallucination than as a function of instructional practices and technologies. Far from being our natural or human condition, hermeneutics merely results from a specifically trained coordination of children’s eyes, ears, and vocal organs. It is a discipline of the body” (1990: X). NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 85

transmission media in order to frame the meaning of spoken words with a significant silence recalls this symbolically integrated world of nineteenth century literature. Based on discreet alphabetic signs, silent writing and reading filters out everything that does not fit its symbolic framework, which is why, prior to the age of technical media, the reduction of external disturbances was relatively effective and words and their meanings could be considered fixed, lucid and fully knowable. Physical signals processed by technical media, on the other hand, cannot be reduced to such clear signs, because over the course of their transmission something irreducible sticks to the signal or, as Kittler puts it in Gramophone, Film, Typewriter, something “ceases not to write itself" (1999: 3). On the one hand, Kittler paraphrases Jacques Lacan in his essay “Signal-to-Noise Ratio,” “the sign of the sign is that, by definition, it can be replaced” (2013c: 166). Due to their discreet nature, alphabetic signs can remain symbolically separate from the material channels that record and transmit information. On the other hand, Kittler continues, “all that is Real sticks in place” (166). Because signals (re)produced by technical media run in the physical real, something of the irrepresentable Lacanian Real “sticks in place” and affects the signal. Whereas the symbolic bottleneck of writing reduces the randomness and contingency of the Real, with physical signal transmission, they remain in place. Information theory simply calls these random and contingent events that stick in place and affect the transmission ‘noise.’ With the transition from what Kittler calls “discourse network 1800” (ruled by the reductive bottleneck of the alphabet) to what he calls “discourse network 1900” (ruled by media that store and reproduce physical signals) this noise turned from that which is excluded from all symbolic representations into that which remains in place in each physical reproduction (1990: 185- 186).29 This is why Kittler in Gramophone, Film, Typewriter posits that the gramophone is the only medium to store the noise of the Lacanian Real itself. I will get back to this notion in Section 2.5 and again in Chapter Five (1999: 4).

29 It is often noted that the English term ‘Discourse Systems’ does not completely do justice to the original German concept of “Aufschreibsystemen,” which literally translates as something like ‘writing down systems’ or ‘systems for writing down.’ 86 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

The complete reduction of noise presupposed by the conceptual logic of noise reduction would produce an entirely clean output consisting of a meaningful silence framing clearly delineated, unambiguous information. The physical operations of technological noise reduction systems, however, are not as perfect as this ideal suggests. Their output is marked by something that “sticks in place”—the noise of technological transmission channels. The presence of this noise points to a fundamental distance between ideal models according to which noise is always clearly defined and precisely located (as presupposed by the conceptual logic of noise reduction and Sterne’s domestication of noise) and the physical operations of technological systems.

2.3 From ideal models to physical filters a) Ulysses and Orpheus: blocking or masking noise In The Five Senses, Michel Serres highlights the distance between ideal models supporting the idea of perfect noise reduction and the physical operations of technical filters with the metaphor of two mythical Greek ships carrying two Greek heroes—Ulysses and Orpheus—past the deadly Sirens, luring the men and their crew two shipwreck with their singing. For Serres, the myths exemplify two ways to confront the problem of noise in communication systems: the sailing ship is the signal, the sea is the channel and the Sirens provide continuous background noise. In the case of Ulysses’ journey, the ears of his men are clogged with wax and the hero himself tied firmly to the mast, unable to move. With wax and physical restraint, Serres writes, Ulysses “blocks noise out” (2008: 126). He hears the Siren noise, but it does not affect the successful transmission of his signal/ship, because the ears of the men are stuffed. Orpheus, the famous singer and musician, sailed past the Sirens as well, together with the Argonauts on their ship the Argo. He, however, was not tied to the mast and the ears of his companions were not clogged with wax. He did not “block noise out.” Orpheus covered it with singing and playing (126). The cunning, resourceful Ulysses, the great teller of tales, is a man of reason and logic, a man of words. Always looking for practical solutions and unambiguous answers, his triumph over the Sirens is based on a clever NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 87

ruse: he blocks out their noise to ensure the safe and steady passage of his ship, ensuring the clean transmission of his signal through the noisy channel. With less noise, signals are transmitted better and faster, which is why it is “hardly surprising,” Serres writes, “that his messages are heard” (126). Because he manages to successfully sail his ship through the channel and reach the other side unharmed, Ulysses is in the position to tell the story as if the Sirens did not get through to him and his noise reduction strategy worked perfectly. History is written by the victors, which is why the story of Ulysses’ ruse has been told for millennia.30 Ulysses, Serres writes, “mak[es] it through the pass in silence, but cheats by suppressing all noise, danger or temptation” to successively claim absolute victory over the Sirens’ noise (122). Orpheus’ strategy is different. He performed an example of auditory masking avant-la-lettre: drowning out the noise with his music. Covering Siren-noise with singing-signal he proofs that, to use Sterne’s definition of auditory masking, “noise could be masked and put in its place” (2012: 94- 95). Serres argues, however, that this victory is much more precarious than the triumph claimed by Ulysses. Contrary to Ulysses’ ruse, Orpheus’ masking is relative and remains “open to the risk of collapsing into noise” (2008: 126). With Orpheus’ strategy, noise is not eliminated. “Ears open,

30 In the first volume of Musik und Mathematik, Kittler presents an alternative interpretation of Ulysses’ journey past the Sirens (2006b: 56-58). With the Siren song, he writes, “fängt alles Senden in Europa an.” However, as Kittler empirically tested by sailing past the Italian islands Il Gallo Lungo, Casteltuccia and Rotonda while opera singers were singing at shore, contrary to what Homer writes, Ulysses cannot have received the Siren song from his ship as clearly as he claims. “Wir hörten klar und rein […] Stimmlaute strahlen,“ Kittler describes the results of the experiment, “doch von Mit- und Stummlaute nicht die geringste Spur. So hat uns denn kein Wort erreicht.” If Ulysses really stayed on board, tied to the mast, as Homer describes, the transmission of the Siren song would have failed, because the noise of the sea overwhelms the words as only vowels reach Ulysses’ ears. Since these words are nonetheless written down by Homer, Kittler concludes, Ulysses must have lied. He did not sail past the island, but landed and made love to the Sirens. Hence, if we are to trust Kittler and with him the Sirens, although Ulysses claims the Siren song reached his ears loud and clear, he is betrayed by the necessary noise of consonants, without which the words of the song would not have made any sense and which Kittler ‘proves’ must have been lost in the transmission from island to ship. Hence, Kittler warns, “nicht dem grössten Lügner Griechenlands, sondern zwei Sirenen glauben.” 88 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

carrying his instrument before him, waxen heart bared to the winds, Orpheus confronts the chaos” (126). The noise is temporarily suppressed, but can crop up at any time. Orpheus does not adhere to the ideal of complete noise reduction and reveals its relative and temporary basis. Years later, when he attempted to drown noise with singing and playing for a second time to get past the wild Bacchantes, he got ripped apart and died: noise reduction is never complete (126). For Serres, Ulysses’ story, told as if no noise got through to him, is exemplary of the way that “science presupposes a world without noise” (126). Science, logic and reason presuppose a well-ordered world of clear answers, noiseless signals and pure information. The most famous example of this rationalist worldview, he argues, is Leibniz’s Law of Continuity, captured by the famous dictum that ‘nature does not make jumps’ (126).31 According to Serres, with the Law of Continuity in the New Essays to Human Understanding and the concept of the perfectly self-contained monad as elementary ontological unit in The Monadology, Leibniz “presupposed a world without noise” as well (126). This world, as physicists Ilya Prigogine and Isabelle Stengers put it in their Postface to Serres’ Hermes, is a world “without friction or holes” (1982: 155). It is the world of The Monadology, according to which each monad, each element “supposes and translates th[e] system in every detail,” suggesting the possibility of a “full passage between the local and the global” (144). In a world in which each part reflects the whole and vice versa, there is complete continuity from the smallest to the largest element and ambiguity, inextricability and confusion do not exist. A world governed by the Law of Continuity is a world without the randomness and contingency of noise. This is why, for Serres, “Leibniz is bound to Ulysses” (126). By extension, I argue, the ideal of complete noise reduction that conceptually frames Dolby’s technological operation is bound to both Ulysses clever ruse and Leibniz’s noiseless world. By suggesting

31 In the New Essays on Human Understanding, first publishes posthumously in 1765, Leibniz writes: “In nature everything happens by degrees, and nothing by jumps; and this rule about change is one part of my law of continuity” (1996: 473). NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 89

that every signal can get through the channel in full clarity, the conceptual logic of noise reduction assumes that pure, clear and transparent transmission is always possible. Notwithstanding his self-declared triumph over irrationality, inextricability, confusion and noise, however, Ulysses’ rational ruse is inherently limited. Regardless of the heroic claims of Ulysses, the rational system of Leibniz and the technical filters of Dolby, Serres remarks in Genesis, “the purest is never pure enough to remain the master forever” (1995: 131). As Orpheus’ fate shows, the purity ensured by noise reduction is relative and precarious. Ulysses’, Leibniz’s and Dolby’s ideal, noiseless world presupposes the possibility of complete reduction, but as Orpheus’ strategy shows and Shannon’s mathematical model of communication confirms, this world is ultimately impossible, because, regardless of the method of reduction, noise is internal to the system itself. Thus, like Orpheus’ attempts to keep the Sirens and Bacchantes at bay by covering their noise with singing and playing—succeeding the first time and failing the second—noise crops up time and time again. Even more so, not only is the reduction process never complete, it is itself subject to the logic of Shannon’s model of communication: as signals travel through the physical channels of noise reduction systems, they inevitably contract noise.32 Ulysses and Orpheus both apply filters that separate signals from noise and mortal men from murderous Sirens, but their respective filters operate on a different basis. The perfect separation of signal and noise and absolutely smooth sailing of Ulysses’ heroic account would have required a perfectly transparent filter that, like Dolby’s “ideal audio device,” imposes no “limitation on the signal passing through” (Dolby Laboratories 1987: 2). Orpheus’ singing, on the other hand, reveals how every noise reduction filter is applied with specific criteria, using specific standards in a specific context, and no signal passes through a channel without being affected. Some noise gets through, no matter what. Regardless of the fact that Dolby’s most advanced SR system “can create an infinite number of filters through which the signal must pass before it is recorded,” these filters can only be applied to those elements that the system identified as noise in the

32 As Stäheli concludes as well: “every reduction of noise produces a noise of its own” (2003: 253). 90 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

first place (Dolby SR 1987: 5). Instead of the ideal passage or perfect journey suggested by Ulysses’ heroic stories, conceptualised by Leibniz’s Law of Continuity and marketed by Dolby Laboratories’ brochures, noise reduction is an active, on-going and inherently incomplete procedure. It is not an ideal filter that effortlessly separates clearly defined signals from precisely located noise. It is a physical filter that, like Orpheus’ singing and playing, is continuously and precariously masking noise with signal, all the while taking the risk of being affected by the very process of noise reduction itself. Hence, Sterne’s notion of the domestication of noise cannot be the final word on the role of noise in sound recording. Because something always sticks in place, what is received on one end of the channel is not the same as what went in. Even more so, as the concealing and revealing of the companding procedure shows, what comes out as a supposedly noiseless original retroactively shapes our understanding of what ‘originally’ went in. Hence, I propose that not the supposedly inherent connection between input and output or the supposedly clear difference between signal and noise should be the focus of an assessment of the role of noise in sound recording (as the myth of perfect fidelity and the conceptual logic of noise reduction would have it), but the technological operations of the filtering channel itself, which continuously conceal and reveal, configure and reconfigure different layers of and different relations between signal and noise. b) A more rigorous filter: from Dolby to dither The relation between pre-war communication engineering and the postwar concept of noise in information theory, explained in Section 1.1c and Section 2.1b on the basis of analyses by Schwartz and Mills, showed that the operations of dual-ended noise reduction systems are conceptually framed by the assumption that signal and noise can be completely separated. However, as Weaver explains in his commentary on Shannon, Information theory also shows that noise and signal cannot be separated entirely because noise is internal to all transmission channels (Shannon and Weaver 1964: 22). Following Stäheli’s critical reading of this information theoretical frame, I therefore conclude that, firstly, noise NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 91

reduction only reduces what its underlying conceptual logic already marks as noise; and, secondly, that some noise inevitably escapes the filter to ‘stick in place.’ In the digital age, the conceptual logic of noise reduction supporting the notion of the complete domestication or taming of noise in signal processing became, as I cited Siegert in Chapter One, Section 1.3b, “nothing less than systemic” (Siegert 2015a: 30). Returning to the representational logic of discreet signs that had been the dominant mode of representation for pre-technical media, digital media reinstall the complete symbolic separation of channel-noise and signal-information that is characteristic of alphabetic media. In contrast to alphabetic media, however, the operationalisation of this representational logic of discreet signs by digital media also enables the storage, production, reproduction and transmission of physical signals. It can therefore be argued that the operations of digital sound media, as part of a new technological paradigm, are based on a very rigorous noise filter by means of which digital machines are able to separate all things that you want from all things you don’t want. Although the promotional brochure for Dolby SR at the dawn of the digital age in 1987 argues that “a typical digital recording provides performance that is better than unassisted analog tape in several obvious ways” because “the noise level is […] much lower than the noise of analog tape,” it also warns that “the usable improvement in noise level, especially in the presence of a signal, is not as great as theory predicts” (1987: 7). Because Dolby’s analogue noise reduction systems were commercially threatened by the emergence of a technology that operationalised the structural separation of noise and signal, the brochure highlights one of the most pronounced drawbacks of digital sound: quantisation error and dither. Similar to the difference between the conceptual logic of noise reduction and the physical operations of technological noise reduction, the addition of dither to eliminate distortions caused by errors in the digitisation process points at a difference between the seamless operations suggested by analytical models of representation and their physical implementation in the form of digital media. Whereas the case study of dual-ended noise reduction systems problematises the conceptual logic of noise reduction, which takes noise as 92 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

something to be ignored, eliminated or instrumentalised depending on the context of reproduction or transmission, the analysis of dither constitutes a next step. If, as I proposed on the basis of the analysis of the concealing and revealing of noise reduction technologies, the technological operations of the filtering channel itself should be the primary site for assessing the role of noise in sound recording, a close reading of the more rigorous filtering operations of digital media—including the almost complete elimination of the noise of material channels and its return in the form of dither—could open the possibility to begin reading the role of noise in sound reproduction technology differently.

2.4 Digital error and analogue noise a) The limits of digitisation: sampling and quantisation Critics of digital sound media often point to the fact that, following the Nyquist theorem, a digital system can only capture frequencies up to half its sampling rate, which means the frequency response of a digital recording is inherently limited—a system with a sample rate of 44.1 kHz cannot reproduce frequencies above approximately 20 kHz. In “Defining Phonography: An Experiment In Theory,” media theorists Eric Rothenbuhler and John Durham Peters for instance argue that, “the upper limit of fidelity in an analog system is perfection, while the upper limit of fidelity in a digital system is the Sony-Phillips [sic] convention” (1997: 235). Although it is possible to increase the sample rate and record higher frequencies, digital recordings always contain an absolute limit inherent to the basic principles of their mode of reproduction. In contrast, so the argument goes, the upper limit of analogue systems is restricted by the physical variables inherent to a specific medium (wax, tape, vinyl etc.), but not the basic principles of analogue sound reproduction as such. As in Zeno’s paradox of Achilles and the tortoise (figure 9), the infinitely discreet can approximate but never actually attain the status of a continuous signal. Both Zeno’s paradox and the supposedly fundamental opposition between discreet and continuous sound reproduction, however, erroneously represent temporal events as spatial issues, but the contest between Achilles and the tortoise and the sampling of sound waves do not NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 93

Figure 9 Achilles and the Tortoise by Martin Grandjean, “Achilles and the tortoise.” Wikimedia, 19 Mar. 2014, upload.wikimedia.org/wikipedia/ commons/6/66/Zeno_Achilles_Paradox.png, accessed 23 Oct. 2015.

In the paradox of Achilles and the tortoise, Achilles gives the tortoise a lead in a running contest but fails to overtake him because, logically, the distance he has to run before overtaking the tortoise can be infinitesimally divided in half.

94 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

only take place in space, but also over time.33 The reduction to purely spatial terms, according to which the original sound wave can only approximate the original signal because the converter always misses the ‘spaces’ in between the samples is thus based on a crucial misunderstanding of digitisation.34 Although a digitised signal is often visualised as a series of discreet samples with nothing in between, the reproduced signal that comes out of the loudspeakers is entirely continuous, because, as I cited Roads in Chapter One, Section 1.3b, as long as the conditions of the Nyquist-theorem are not violated “the missing part of the signal ‘between the samples’ can be restored [as] the smoothing filter ‘connects the dots’ between the discrete samples” (1996: 27). Owing to a sample rate of tenths of thousands of samples per second, when they are transduced back from digital signs into electric voltage levels and from voltage levels into sound waves, the discontinuity of binary representations turns into entirely continuous signals. Although the output signal is entirely continuous, it is true that a digital system cannot reproduce frequencies above the maximum of half the sampling rate. Contrary to what Peters and Rothenbuhler argue, however, the upper limit of analogue systems is not “perfection” either. Like the frequency response of digital systems is inherently limited by the sampling rate, the bandwidth of an analogue system is inherently limited by background noise, distortion and other interferences that occur during signal transmission. An analogue system with unlimited bandwidth is as hypothetical as an infinitely precise sampling procedure. Furthermore, focussing exclusively on the limitations of the sampling procedure, Peters and Rothenbuhler fail to note that the occurrence of quantisation errors is a much clearer indicator of the preconditions, possibilities and structural limitations of digital technologies and their relation to the analogue domain.

33 For more on (the history of) this fundamental epistemological fallacy of reducing temporal phenomena to spatial terms, see philosopher Milič Čapek’s introduction to The Concepts of Space and Time. Their Structure and their Development (1976). 34 In Sound Ideas, musicologist Aden Even for instance writes: “whatever variation occurs in the time between two samples will be missed” (2005: 12). NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 95

As Weaver explains Shannon’s model of communication, in information theory “all […] changes in the transmitted signal,” whether caused by random noise, by distortion, by errors, by static or by any other interference, “are called noise” (Shannon and Weaver 1964: 8). On the level of the physical characteristics of these disturbances, however, Shannon writes in “Communication in the Presence of Noise,” “noise and distortion may be differentiated on the basis that distortion is a fixed operation applied to the signal, while noise involves statistical and unpredictable perturbations” (1998: 447). Due to this predetermined, correlated and predictable (“fixed”) nature, Shannon adds, distortion “can, in principle, be corrected by applying the inverse operation” (447). Random and unpredictable noise, on the other hand “cannot always be removed, since the signal does not always undergo the same change during transmission” (447).35 Thus, on the level of technological operations, both non-random interferences (distortion) and completely random perturbations (physical noise) cause changes in the transmitted signals, but whereas the first can be eliminated owing to its predictable, fixed nature, the second can only be prevented, masked or reduced. On the conceptual level of information theory, however, all interfering disturbances (random or non-random) are called noise and the changes they cause in the transmitted signal, whether large or small, are classified as errors. In short, Shannon’s model of communication classifies all interferences as forms of noise that cause errors in the transmitted signal. In the case of quantisation errors, however, it is the other way around: a (non-random) error in the digital representation of signal amplitudes introduces (non-random, harmonic) distortion in the digital signal that is eliminated by the addition of (random) noise. As explained in Chapter One, Section 1.4b, in the digital scenario, the non-random

35 Adding to the terminological confusion, in The Recording and Reproduction of Sound Oliver Read writes that “a sound is said to be distorted when the waveform is altered in transmission or when the intensity of any frequency is suppressed or exaggerated out of its natural proportions,” whereas “noise is usually considered to be random sound waves with little or no periodicity” (1952: 3, 5). Thus, Reads definition of distortion comes closer to Weaver’s definition of noise, whereas his definition of noise is the more limited physical concept of noise as random perturbations. 96 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

quantisation error, as the source for non-random harmonic distortion, statistically takes the place of the random background noise of analogue recording: the signal-to-noise becomes a signal-to-error ratio. Furthermore, whereas analogue noise reduction masks noise and other disturbances with a louder signal, the non-random error of digital recording is statistically randomised by noise, changing harmonic distortion in a layer of random or semi-random noise that is similar to an analogue noise floor. Although, analogue background noise and digital quantisation error are therefore statistically similar in their effect on the signal, “the signal-to- error ratio of a digital system,” Pohlmann writes, “is […] not identical to the S/N (signal-to-noise) ratio of an analog system” (2000: 34). Notably, the difference between the two ratios is indicative of the difference between analogue and digital recording more generally. Whereas the analogue signal-to-noise ratio determines the dynamic range (the difference between the minimum noise floor and the maximum amplitude level) of a recording, the digital signal-to-error ratio “indicates the degree of accuracy that’s used to capture a sampled level” (Huber and Runstein 2010: 207). The amount of noise in an analogue recording indicates the bandwidth of the transmission channel and its capacity to transmit information without errors. The amount of error in digital recording, on the other hand, only indicates the precision of the measured sample values. Because this error is not some random but statistically correlated to the digitised signal, Pohlmann argues, “it cannot be described as noise; rather, it must be classified as distortion” (2000: 36). Although Shannon writes that, contrary to random noise, distortion can “be corrected by applying the inverse operation,” however, this does not hold for quantisation error. Although the background noise of analogue sound recording is internal to the communication system, the conceptualisation of the signal- to-noise-ratio allowed engineers to idealise signal and noise as two separate phenomena originating from the same source; a conceptual move that contributed greatly to the development of sophisticated noise reduction technologies (Kittler 2013c: 167). Quantisation errors, on the other hand, cannot be symbolically separated from the signal in the same way. “What [a digital machine] produces when a product is called for,” von NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 97

Neumann writes, “is not that product itself, but rather the product plus a small extra term—the roundoff-error” (1963: 295). Because these errors come into being at the moment of its digitisation, they are inseparable from the digital signal itself. Whereas the random noise in analogue recordings is inherent to the transmission channel, the “small extra” of quantisation error is inherent to the signal. Like the sonic artefacts of analogue recording methods, quantisation errors thereby reveal what Mark Nunes calls the “operational logic” of digital systems. Although, as Nunes continues his analysis of system error, the “logic of maximum performance demands that error is either contained as predictable deviation […] or nullified as an outlying and asignifying event,” quantisation errors cannot be entirely contained or nullified because they occur at the very moment of digitisation itself and are thereby part of its foundational operation (2011: 3). As a consequence of the operational logic of the digital according to which truncated value numbers asymptotically approximate potentially infinitely long value numbers, quantisation errors attest to an inherent problem of digitisation: the impossibility to represent infinitesimally precise values with an inherently finite number of signs. b) Leibniz: the end of classical representation At the outset of Passage des Digitalen, Bernhard Siegert situates this question of representing infinitesimally detailed phenomena with an inherently limited representational systems within a historical trajectory going all the way back to the thirteenth century English “inquisitio” or the attempt to create a complete description of every mobile and immobile thing in the king’s possession (2003: 21). With the emergence of modern mathematical analysis in the seventeenth and eighteenth century and the development of technical media as a result of this emergence in the nineteenth and twentieth century, this history of representation, Siegert argues, is marked by a ‘rupture’ or ‘break’ (ein Riß) in the classical order of representation that had been dominant up to Leibniz’s (and Newton’s) infinitesimal calculus in the late sixteenth century (17). Like the English inquisitio, this classical order of representation assumes the possibility of a complete representation of the world on the basis of writing: “a complete 98 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

description of all things” produced by “a big bureaucracy governed by the Law of Continuity” (16). From Leibniz onward, Siegert argues, mathematical analysis breaks with this logic of full representation. Modern mathematical analysis does not assume the possibility of complete, exhaustive representation. From Leibniz’s infinitesimal calculus in the late sixteenth century, via Leonhard Euler’s discontinuous functions in the eighteenth century, up to Joseph Fourier’s “Analytical Theory of Heat” in the nineteenth century, mathematical analysis and theoretical physics introduced functions that no longer correspond to any observable phenomenon in the physical world. They only exist on paper (17). In contrast to the rationalist logic of classical analysis, this produced, as Siegert argues, “a deterritorialised analysis” that triggered a “drift of the non-representational” and ultimately caused a “removal of the fundaments” of classical representation altogether (16-17).36 Exactly this loss of a firm fundament for the ideal of complete representation, however, enabled the emergence of a conceptual framework on the basis of these discontinuous functions that slowly “open[ed] up a space for technical media” able to technologically reproduce physical phenomena that could not be fully represented (16).37 The first cracks of this rupture appear with Leibniz’s introduction, by means of his invention of infinitesimal calculus, of a thinking of the infinite that sowed the seed for the eventual dissolution of his own Law of Continuity. Siegert’s rupture in the order of the representational marks the end of what Serres calls Leibniz’s “world without noise” (2008: 126). The best example of the way this “world without noise” relates to the Law of Continuity and the development of infinitesimal calculus is Leibniz’s own analysis of the noise of the sea, cited by both Siegert in Passage des

36 “Das moderne Analytische, das heißt die Analysis seit Euler, ist ein deterritorialisiertes Analytisches: “Eine Drift des Nicht-Reprasentierbaren̈ .” “Die technischen Medien sind gegründet im Entzug des Grundes.” Note in this context the multiple meanings of the German word Grund: I translated it as ‘fundaments,’ which is close to its most literal meaning as ‘ground,’ but Grund can also be translated as ‘reason’ (as in ‘the reason for something’) or ‘cause,’ adding to the significance of this loss of a firm fundament for complete representation. 37 “Es ist der Riß einer im Denken der Repräsentation verwurzelten Ordnung der Schrift, der die Passage des Digitalen freisetzt und den Raum der technischen Medien eröffnet.” NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 99

Digitalen and Serres in Genesis. In the context of his philosophy of perception, Leibniz’s example of the noise of the sea illustrates how he interprets human perception of phenomena in the world as the combination of an infinite number of infinitesimally small perceptions (petites perceptions) (Leibniz 1989: 295). Although, Leibniz argues in the “Preface To The New Essays,” the “roar or noise of the sea that strikes us when we are at the shore” sounds inextricable to our ears, it is in fact a combination of infinitely many individual sounds, one for each single wave (295). Although our ears process each of these small perceptions “in the confused assemblage of all the others,” Leibniz suggests that every sound belongs to a separate entity, because a large phenomenon can only consist of many smaller ones: “a hundred thousand nothings cannot make something” (296). As Gilles Deleuze puts it in his lectures “On Leibniz”: “there has to be something simple, if there is something composite” (2007). Together with his concept of the elementary unit of the monad, this notion of the world as a continuum of infinitesimal elements shows that, as Prigogine and Stengers write, “[Leibniz] refused to give up the idea of the rational nature of the real, measured not by the yardstick of man, who observes and generalizes, but by that of God, who, calculating, created the world” (1982: 139). Hence, what humans perceive as the inextricable noise of the sea, an all-knowing God can see as an infinitely complex assemblage of infinitesimally small, but perfectly self-contained, clear sounds.38 Given this rationalist foundation of Leibniz’s system of thinking, Serres argues, “it would no doubt have seemed absurd to the old master, for rationality in its totality not to be rational” (1995: 19, 20). This is why

38 Regarding the role of God, Leibniz writes: “eyes as piercing as those of God could read the whole sequence of the universe in the smallest of substances” (Leibniz 1989: 296). In Allwissen und Absturz: Der Ursprung des Computers, media scholars Peter Bexte and Werner Künzel put forth the thesis that, three centuries after Leibniz, the possibilities of the digital computer come close to those of the supreme being Leibniz considers capable of disentangling the most inextricable noise (1993: 156). It is Bexte as well who, following a lecture by Friedrich Kittler that was published in 2012, made a remark on the similarities between Fourier analysis and Leibniz’s s notion of minute perception, probing Kittler to answer that “God is the big Fourier-analyst” (2012b: 48). I will comment more extensively on this remark in Chapter Four, Section 4.2a. 100 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

“a noisy philosophy would be the shadow of Leibnizianism” (20). These rationalist principles notwithstanding, along with the mathematical invention of infinitesimal calculus and the ontological premise of The Monadology, the notion of the perceptual integration of infinitely many small perceptions undermined Leibniz’s own Law of Continuity. Due to the central place of mathematical infinity at the heart of the concepts of minute perceptions and infinitesimal calculus, Serres argues, Leibniz’s rationalist order, which is based on the complete integration of every unit in a larger whole, “remains inaccessible” (21). In order to uphold the continuity that his rationalism dictates, “Leibniz lumps everything into the differential, and under the numberless thickness of successive orders of integration” that only God can unravel (20). In his reading of this aspect of Leibniz’s philosophy in The Fold: Leibniz and the Baroque, Deleuze argues that this “division of the continuous," based on the mathematical principle of perceptual integration

must not be taken as of sand dividing into grains, but as that of a sheet of paper or of a tunic in folds, in such a way that an infinite number of folds can be produced, some smaller than others, but without the body ever dissolving into points or minima (2006: 6).

This concept of the continuous as a fold that is “folded within a fold, like a cavern in a cavern” takes the very idea of a continuum itself to its utmost limits (6). Because the Law of Continuity does not allow for breaks, holes, jumps or pure contingent events, ideas of the complete integration or infinite folding of entirely singular small perceptions only hold with the introduction of a deus ex machina, a divine being that is able to do what humans cannot do: perceive, calculate and integrate all individual parts of the whole. Human beings can either trust their imperfect senses and put faith in this higher power or use the “numberless thickness of successive orders of integration” to analyse the infinitesimally entangled or confused. Because, musicologist Roger Moseley writes in “Digital Analogies: The Keyboard as Field of Musical Play,” “negotiations between the NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 101

properties of continuity and discreteness shaped Leibniz’s formulation of infinitesimal calculus as well as the metaphysical materialism of his monadology” (and, I would add, the notion of small perceptions), compromises between continuity and discreteness are fundamental to his philosophical system (2015: 168). Over the course of the following centuries, Siegert writes however, this Leibnizian “ontology of the noise of the sea” changed into “a non-Leibnizian order of things,” because infinitesimal calculus and mathematical integration opened the road toward a type of analysis that no longer assumes the possibility of complete representation (2003: 234, 182).39 When the non-representable drifted toward the center of mathematical analysis in the form of discontinuous limit cases and asymptotical approximations, the connection between everything that Leibniz’s Law of Continuity conceptually held together was broken (253). This deterritorialisation of the rationalist ideal of complete representation created a gap between symbolic representations produced by mathematical analysis and the physical phenomena they represent. Description and the descripted, representation and the represented drifted apart (182). c) Fourier: the analysis of the entangled About a century after Leibniz’s death, in 1807, Joseph Fourier’s application of Euler’s discontinuous trigonometric functions to the mathematical analysis of wave phenomena in his “Analytical Theory of Heat” completed this “drift of the non-representational” (Siegert 2003: 192). I will discuss Fourier analysis and its impact on and importance for the modern concept of sound in detail in the following chapter, but for the moment it is important to emphasise that Fourier’s revolutionary analytical method marks the end of the ideal continuity of Leibniz’s concept of noise as an inextricable assemblage of singular small perceptions. After the application of Fourier analysis to the study of sound waves by Georg Ohm and Hermann von Helmholtz in the latter half of the nineteenth century, it became theoretically possible to represent all

39 “Das [...] Überschlagen der Leibnizschen Ontologie des Meeresrauschens in eine nicht- leibnizsche Ordnung der Dinge.” 102 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

individual frequency components of an entire sound spectrum as a series of so-called sine waves. Although only the processing speed of digital computers and the introduction of the Fast Fourier Transform in the latter half of the twentieth century made this mathematical analysis fully operational, Fourier analysis provided the analytical method to extract the inextricable and unravel the noise of the sea into its singular components, creating an analytical order that human senses can never achieve. What remains fundamentally unconceivable in Leibniz’s rationalist order, (perceiving every element of the roar of the sea) is symbolically achieved by Fourier’s analytical method, which separates the assemblage of complex sound waves into all its individual frequency components. Siegert therefore argues that Leibniz’s “ontology of the noise of the sea,” based on the integration of an infinite number of infinitesimal minute perceptions, anachronistically presupposes the mathematical integration of Fourier analysis (2003: 187). Instead of confirming Leibniz’s ideal continuity, however, Fourier analysis hereby confirms the rupture in the order of the representational. Fundamentally based on the asymptotic approximations and discontinuous limit cases that mark modern mathematical analysis, the physical phenomena represented by Fourier analysis only asymptotically approximate but never fully coincide with the idealised representations produced by its analytical procedure. Fourier’s analytical clarity thus comes at the costs of the rationalist ideal of unbroken continuity and complete representation. Physical wave phenomena like sound, heat and light had proven to be fundamentally non-representable in a rationalist representational order. With Fourier analysis, the complexity of sound became symbolically representable, but only through a type of mathematical analysis that lumps their full complexity, in Serres words, “under the numberless thickness of successive orders of integration” and sacrifices the ideal of complete representation nonetheless (1995: 20). Over the course of the nineteenth century, exactly this “drift of the non-representational” caused by modern mathematical analysis opened the road for the emergence of technical media. Based on the asymptotic nature of mathematical analysis, but exchanging symbolic representation in the form of written signs for NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 103

physical (re)production in form of physical signals, technical media do not fully represent, but physically (re)produce signals. What Moseley calls Leibniz’s “negotiations between the properties of continuity and discreteness” therefore marks the beginning of a trajectory that eventually released the potential of technical media to analyse, store, produce, reproduce and manipulate physical signals. This is why Siegert argues that “the technical (which means electrical) media are positioned in the rupture of classical analysis” (2003: 389).40 Borne from the mathematical analysis of the entangled, the inextricable, the confused and the infinite, technical media are part of the “drift of the non- representational” from Leibniz to Fourier and beyond. Hence, the inherent limitations of analogue and digital transmission channels to fully reproduce physical signals must be understood as the product of the asymptotic logic of modern mathematical analysis that enabled the conception of technical media in the first place. Both the presence of random analogue background noise and the impossibility of digital sound media to represent infinitesimally specific amplitude values can be traced back to the rupture in the order of the representational; the rupture between on the one hand Leibniz’s small perceptions, infinitesimal calculus and monadology and on the other hand his Law of Continuity. Marked by this rupture in the classical order of representation through the emergence of modern mathematical analysis, the conceptual basis of both digital and analogue technology is tied up in this long history of analytical representation.41

2.5 Revealing and concealing noise a) The analogue and the digital: the Real and the Symbolic Even if the question of their relation and difference is restricted to the domain of sound reproduction, any simple distinction between the

40 “Die technischen (das heißt elektrischen) Medien sind plaziert im Riß der klassischen Analysis.” 41 As Siegert writes in the introduction to Passage des Digitalen: “Das Digitale und das Analoge sind nicht Episoden einer Geschichte der Medien, vielmehr sind die technischen Medien eine Episode des Digitalen und des Analogen, eine Epoche der graphé” (2003: 15, emphasis in original). 104 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

analogue and the digital breaks down as soon as one tries to locate fundamental differences. Already on a descriptive level, Huber and Runstein for instance remark that “if you think this [digitisation, MK] process of changing one form of energy into an analogous form (and then back again) sounds like the general definition of a transducer—you're right!” (2010: 200). Indeed, Watkinson gives a more detailed rendition of this similarity:

In all audio recording the goal is to preserve one or more electrical waveforms, which are analogs of the motion of a microphone diaphragm. In an analog tape recording, the distance along the tape is an analog of time and the strength of the magnetic flux left on the tape is an analog of the signal voltage. In digital audio two analogs of time and voltage are handled in a completely different way, but they remain analogs and, despite the complexity, digital audio is just an alternative way of preserving a waveform (1999: 110).

Both Huber and Runstein and Watkinson hereby suggest that digital and analogue reproductions are equally based on analogous representations of physical sound waves. Neither representation is inherently closer to or further removed from what is reproduced. Neither one is more ‘true to reality.’42 According to Moseley, all “analog modes trace continuity [and] at the same time […] span and measure the gap that separates the resembled from the resembling. They thus represent both continuity within a medium and the rupture of transduction,” (2015: 154- 155). Hence, if both analogue and digital reproductions are analogues of

42 Regarding the supposed difference between analogue and digital media on the basis of the continuity or discontinuity of the representation, Jonathan Sterne remarks in “The Death and Life of Digital Audio” that “analogue tape is just as discontinuous as the 0s and 1s in digital storage” (2006: 341). NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 105

physical processes that represent “continuity within a medium and the rupture of transduction,” the difference between the two forms of representation is in the specific way they “span and measure the gap that separates the resembled from the resembling.” Whereas analogue reproductions “trace continuity,” digital modes are fundamentally based on discontinuity; on discreetness and difference. The difference between continuity and discontinuity (which, as Moseley remarks, already “reflects a binary choice between the digital and the analog” in itself) is constitutive of the difference in relation between analogue or digital media and the phenomena they reproduce (2015: 155). In a lecture at New York University in 2015, Siegert traced the conceptual roots of this difference between analogue and digital modes of representation and reproduction in the proceedings of the famous Macy Conferences on cybernetics held between 1946 and 1953. Because the conferences took place at a time when “the concepts of the analogue and the digital had not yet been clarified at all” and the categories they connote were still relatively unstable and undefined, these proceedings shed light on the most fundamental aspects of their conceptual difference (2015b). Most poignantly, Siegert shows, the attendees of the conferences discussed the question whether the digital “is part of the real or of the symbolic” (2015b). John von Neumann, on the one hand, was “in no way interested in how the digital was implemented in the real” because he considered the domain of the digital to be exclusively restricted to the operations of symbolic machines, or, in other words, computers (2015b). The neurophysiologists at the conference, on the other hand, “tried hard and desperately to localize the digital within the real” (2015b). Trying to unify the two positions, mathematician Norbert Wiener posited that the essence of the digital operation is the symbolic exclusion of the very short time required to switch between the two binary states (1 and 0). “The digital,” Siegert paraphrases Wiener, “is a function of time and its basis was the creation of ‘a certain time of non-reality,’ which lies between two stable states” (2015b). It is exactly on the basis of this symbolic exclusion of a ‘time of non-reality’—which means the exclusion of the analogue switching 106 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

operation itself—that digital machines are able to create symbolic representations of physical phenomena. “The purpose of creating discreet states by forbidden time zones,” Siegert continues, “is to permit only sharp, discreet values and to prevent that inaccuracies are propagated as some kind of noise” (2015b). Since these ‘inaccuracies’ are the random events that occur on the level of analogue circuitry, in other words the random noise of analogue transmission channels, the whole purpose of the discretisation of time and the exclusion of the ‘time of non-reality’ is to prevent analogue noise from propagating in the digital domain. Prior to the conception of this symbolic gesture of excluding the physical moment of switching, Siegert argues, “precisely these transition states had been the main focus of physicist and electrical engineers which endeavoured […] to filter out a signal from the noise of analogue oscillations” (2015b). Whereas engineers in the age of analogue media tried to reduce the noise of the physical operations of analogue machines, the symbolic gesture at the heart of the digital—the radical exclusion of the moment of switching that excludes the noise of analogue operations—is the reason why, as I cited Siegert before, in the digital, the reduction of noise became systemic. Hence, the question of the difference between the analogue and the digital, social theorist and cybernetician Anthony Wilden points out in System and Structure, “is one of relationship, not one of entities” (1980: 188). More specifically, the difference between the analogue and the digital comes down to a difference in the relationship between representations and represented, resembled and resembling. Although both continuous analogue and discrete digital media are based on ‘analogue’ representations of physical phenomena, the fundamental difference in the way they deal with the noise of their most basic physical operations amounts to a fundamental difference in the way they “span and measure,” in Moseley’s words, “the gap that separates the resembled from the resembling.” On the basis of this different relationship to the noise of the physical real, analogue and digital media part ways. Because of this connection between the noise of the physical real and the operations of analogue media Kittler argues that the gramophone is able to store and reproduce that which cannot be symbolically NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 107

represented: the Lacanian Real (1999: 111). Although, according to Lacan, the Real can only be experienced as a lack, as a sudden apparition through a crack in the higher orders of the Imaginary and Symbolic, the phonograph, Kittler writes, “does not hear as [...] ears,” but “registers acoustic events as such” and is therefore able to “record all the noise […] prior to any semiotic order and linguistic meaning” (16, 23). Due to this indiscriminate storage, what had been fundamentally non-representable prior to technical mediation—the continuous noise of the Real—sticks in place with analogue reproduction. Within the limits of available bandwidth and dynamic range, analogue sound reproduction media capture the complete spatio-temporal flow of physical sound waves, including those sounds that are added by the transmission channel itself. In the form of this surplus of noise added to the signal by the transmission channels of analogue media, the excess of the Real does not cease to write itself onto the recording. By symbolically excluding the surplus of noise that occurs on the level of analogue circuitry, digital technology left this reference to its material basis behind. It thereby creates a Symbolic order removed from the operations of analogue media and returned to a representational order that excludes the fundamental randomness of the Real. Whereas analogue media are unable to completely filter out the noise of their physical operations, the symbolic order of digital code, Siegert cites German computer scientist Jörg Pflüger, functions as an “encapsulation of fuzziness” that represents sharp limits and clear differences on the basis of the symbolic exclusion of noise (2015b). By contrast, these sharp contours produced by digital modes of representation thereby emphasise the connection between the noise of analogue recording and the irrepresentable fuzziness of the Real. It can be argued, for instance, that the statistical correlation between the quantisation errors and the amplitude values of the original input signal emphasises the inability of digital machines to capture the fuzziness of the Real with a finite, discreet encoding system. Rounding-off binary number values to the nearest available quantisation step, the A/D-converter introduces a hard limit in place of an infinitesimally more specific value. In the process, Wolfgang Ernst writes in Chronopoetik, “Discreet-time 108 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

sampling sacrifices an essential characteristic of the Real itself: fuzziness— to return again as so-called quantisation noise” (2012a: 274).43 Despite Ernst’s use of the term quantisation noise, however, quantisation errors and harmonic distortions are not random. They are periodic and predetermined; and as I explained in Section 2.5a, because they are produced by the same symbolic gesture that defines the digital itself (the encapsulation of analogue fuzziness by excluding a ‘time of non-reality’) quantisation errors cannot be corrected and harmonic distortion cannot be reduced. Whereas the irrepresentable Real is inscribed on analogue recordings in the form of a surplus of random noise, its irrepresentability manifests in digital recordings in the form of quantisation errors. The equivalence between the two manifestations of the Real is empirically supported by the statistical equivalence between digital quantisation error and analogue noise. Even more so, the introduction of dither to eliminate the statistical correlation transforms non-random harmonic distortion into a sonic artefact very similar to the random noise floor of analogue media. This begs the question how this return of random noise in the digital domain relates to the conceptual logic of noise reduction and the scientific “world without noise” presumed by Ulysses, Leibniz and Dolby. b) Dither: concealing errors and revealing signals Quantisation errors are a reminder of the rupture in the order of the representational, which, from Leibniz onward, marks the impossibility to create complete symbolic representations of physical phenomena and simultaneously laid the scientific foundation for the development of technical media able to reproduce these phenomena. On the basis of the mathematical analysis of the entangled and the confused, technical media (re)produce physical signals, but these reproductions are marked by what von Neumann calls a “small extra” added to the signal during transmission. The channels of continuous analogue media, unable to suppress the

43 “Geopfert wird in Abtastverfahren als zeitdiskreten Funktionen allgemein ein Wesensmerkmal des Realen selbst: die Unschärfe–um dann als sogenanntes Quantisierungsrauschen wieder einzukehren.” Thanks to Anthony Enns for corroborating this translation. NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 109

physical noise of their own operations, add a surplus of random noise to the transmitted signal. The channels of discontinuous digital media, unable to represent measured amplitude values with full infinitesimal precision, add a surplus of non-random distortion caused by quantisation error. Although, at first sight, the practice of adding noise to digital recordings seems diametrically opposed to the decade long attempt to reduce noise, on a conceptual level, dithering shares more than a few characteristics with the logic that supports analogue noise reduction. Similar to how noise reduction systems actively process analogue recordings to conceal noise and reveal silence, thereby ordering and reordering the relationship between signal and noise to suggest an idealised transmission of pure signals, dither conceals sonic artefacts introduced by the recording process in order to maintain the suggestion of an intrinsic relation between input and output and the idea that the latter could hypothetically coincide with the former. Quantisation errors, Rumsey and McCormick write, “may be considered as an unwanted signal added to the wanted signal” (2009: 219). The resonance between this qualification of quantisation errors and Bose’s description of “things that you want” and “things that you don’t want” suggests the relation between wanted digital signal and unwanted quantisation noise is similar to the relation between noise and signal as presupposed by the operations of noise cancelling headphones and Dolby Noise Reduction. In the case of analogue noise reduction, unwanted noise that accumulates during the recording process is masked by wanted signals. In the case of dither, unwanted semi-periodic signals (harmonic distortion) that accumulate during the digitisation process are eliminated by wanted noise. Introducing a layer of random background noise to the digitised signal, dither conceals the periodicity of harmonic distortion and breaks up its statistical correlation. From this perspective, dither is part of “things that you want” and emphasises, firstly, how the definition of noise on the level of information theory is entirely relative to the system: what is considered unwanted in one situation (the random background noise of analogue media) is wanted in another (the random background noise created by dither). The addition of dither as wanted noise to get rid of 110 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

unwanted distortion turns this noise into a preferred (semi-)analogue alternative to unwanted digital artefacts. It thereby changes the perspective on the background noise of analogue media as well. Secondly, however, the notion that dither is ‘good noise’ introduced to fight ‘bad distortion’ also shows how the operation of dithering supports the ideal of perfect signal transmission that is characteristic of the conceptual logic of noise reduction. Similar to how Dolby Laboratories argue that the encoded and decoded version of a recording is in fact the original, an output of a digital recording chain with dithering and without harmonic distortion is generally considered to be closer to the supposed original as an output without dithering and with harmonic distortion. Although dithering is nothing but a trade-off between harmonic distortion on the one hand and a slightly higher noise floor on the other—trading one statistical error for another—recording and mastering engineers argue that this higher noise floor is less disruptive for human listeners than the harmonic distortion caused by quantisation error. Bob Katz, for instance, argues that a properly dithered recording “always sounds better” than recordings without dithering and Rumsey and McCormick describe how “a small amount of continuous hiss is considered preferable to low level distortion” (Katz 2002: 57; Rumsey and McCormick 2009: 226). Even more strikingly, Nika Aldrich contributes the positive effect of dither to the fact that it produces “a true floor that is a natural occurrence, and therefore more pleasing to our ears”; and Watkinson argues that “the harmonics produced” by quantisation error, can be “especially distressing to the human listener because it does not occur in nature” (Aldrich 2002: 8; Watkinson 1999: 112). Hence, although in terms of information theory, dither merely trades one statistical error for another, all these engineers suggest the level of nuisance for human listeners differs greatly because, compared to statistically correlated harmonic distortion, a random noise floor is ‘better,’ ‘more pleasing,’ more ‘natural’ and overall more ‘preferable’ for human listeners, who are more inclined to tolerate a slight layer of random noise than periodic artefacts of the finite encoding process of sound digitisation.44

44 This supposed preference for a uniformly distributed noise floor over semi-periodic harmonic distortion is supported by the psychoacoustic observation that our hearing is generally more NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 111

This discursive frame of natural noise and unnatural distortion shows how dithering serves a purpose very close to that of dual-ended noise reduction: to conceal the artefacts of the recording and reproduction procedure and reveal a recording that suggests a close correspondence to some supposed original. Dither thereby works by going unnoticed. Because random noise seamlessly fades in the background whereas periodic or semi-periodic harmonic distortion attracts the attention of the listener, dither fits the same discourse on idealised representation as the conceptual logic of noise reduction. It creates recordings that tend toward some idealised original signal and suggest a supposedly natural link between input and output. Both noise reduction and dither are therefore intended to conceal the noise of the channel and uphold the suggestion of an inherent relation between input and output. Like noise reduction, dither is part of a chain of technical filters that simultaneously enable the transmission of the signal and affect its characteristics. However, where noise reduction covers bad noise with good signals and introduces a significant silence that supports the myth of perfect fidelity, I suggest the introduction of dither to eliminate harmonic distortion even more strongly emphasises the fundamental idealisation at the heart of the conceptual logic of noise reduction: the assumption that noise is always clearly defined and easily reducible. Leibniz’s rationalism is upheld by the ideal of a fully integrable world with full continuity from the smallest monad or the sound of a single

sensitive to semi-periodic signals. Because of this sensitivity we are able to pick even the faintest signals out of a much louder noise floor, as is the case with the stochastic resonance effect discussed in Chapter One, Section 1.4b. This is also the reason why, as I cited Moles before, a noise floor “must be four to eight times (12 dB to 18 dB) stronger that the fortissimi (loudest passages) of the transmitted message” in order to drown out a signal completely (1966: 83). Furthermore, the fact that a uniformly distributed noise floor is less noticeable than a signal with a narrower frequency band is the reason why, as I described in Chapter One, Section 1.2c, lower frequency noise only became a nuisance when Dolby’s noise reduction system’s began to reduce high frequency noise, causing the lower noise to stand out more clearly. Lastly, as Jonathan Sterne describes in MP3, the Meaning of a Format, in the 1970s, the preferability of a uniform noise floor over more periodic sounds was put to good use in office situations, where a “continuous, not too loud, unobtrusive” noise was used to drown out the sounds of “typewriters, telephones, office machines, or loud conversation” (2012: 120). 112 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

wave to the entirety of the universe, but, as Prigogine and Stengers argue, confronted with “the failure of simple solutions,” modern science has proven this rationalist worldview to be unattainable (1982: 149). The noise of signal transmission emphasises this failure. It reveals the rupture between the models of an ideal “world without noise” and the asymptotical approximation of technological reproductions that, no matter how sophisticated the noise reduction or how precise the analogue to digital converter, change the transmitted signal along their process of transmission. c) The noise resonance of sound reproduction In addition to decorrelating quantisation errors and supporting the conceptual logic of noise reduction by upholding the idea of an inherent relation between input and output, the last section of Chapter One described a second effect of dither. This effect is the “noise activated process” through which the addition of random noise pushes very faint, low amplitude signals over the threshold of registerability by the digital encoder (Gammaitoni 1995: 4961). Besides concealing the sharp contours of digitised signals with a supposedly more natural random noise floor, the introduction of dither during analogue-to-digital conversion thereby also enables the digitisation of low amplitude signals that would otherwise go unregistered. This means the successful reproduction or transmission of these signals entirely depends on the addition of the right amount of noise. Hence, whereas dither’s first role is to conceal the limitations of digital reproduction and transmission by statistically decorrelating their effect on the digitised signal, its second role indicates that a transmission or reproduction has taken place at all. It indicates that the presence of noise in sound reproduction not only demarcates the rupture in the order of the representational and emphasises the inherent limitations of physical reproduction, but also enables the very possibility of transmission itself. The stochastic resonance effect thereby reveals that the inevitable presence of noise in sound reproductions points to the primary role of the filtering operation in shaping the output signal of a recording chain and creating the supposedly inherent connection between input and output in the first place. NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 113

In various publications on stochastic resonance and its relation to dither, Wannamaker, Lipshitz, Vanderkooy, Kosko and Gammaitoni all question whether the term stochastic resonance is an accurate description of the “noise activated process” it connotes (Wannamaker, Lipshitz, and Vanderkooy 2000; Gammaitoni 1995; Kosko 2006). In “Stochastic Resonance as Dithering,” Wannamaker, Lipshitz, and Vanderkooy argue that, notwithstanding the name generally given to the effect, “no true resonance per se exists in such systems” (233). The added noise and the original input signal, in other words, do not resonate with each other in any conventional sense of the word. Instead, the added noise pushes the signal over a threshold. Thus, Gammaitoni writes

it seems reasonable to conclude that SR in the threshold systems considered here, far from being a resonant phenomenon, can be more correctly interpreted as a special case of the dithering effect consisting of a threshold crossing process aided by noise. For this class of effects the name ‘noise induced threshold crossings’ seems more appropriate (1995: 4698).45

Hence, in terms of the physical characteristics of “noise activated” processes, stochastic resonance is somewhat of a misnomer. I want to suggest, however, that despite the absence of any physical resonance effect, the practice of dithering in digital sound media can be interpreted as a more conceptual type of resonance. I propose that the addition of random noise to digital systems to randomise quantisation errors and push faint signals over a threshold of registerability invites the reconceptualisation of

45 Although Gammaitoni claims that in case of the threshold systems discussed in his article, stochastic resonance can be interpreted “as a special case of the dithering effect,” Kosko argues that Gammaitoni’s explanation of these effects “can be only a partial explanation of the SR effect” (Gammaitoni 1995: 4698; Kosko 2006: 217). Whatever is the case, the argument that the stochastic resonance effect is not a true resonance effect holds either way. 114 NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL

the role of noise in sound recording more generally. Playing on the erroneously named stochastic resonance effect, I call this reconceptualised role the noise resonance of sound reproduction. As I explained over the course of these first two chapters, from Tainter’s ideal “acoustic transparency” via Dolby’s “ideal audio device or system” all the way up to the “nothing less than systemic” separation of signal and noise in digital systems, the ideal of perfect reproduction through vanishing mediators comes up time and time again in discourses on sound reproduction. The concept of a noise resonance proposes an alternative view on this presence of noise in sound media: not as a nuisance, but as a crucial, indispensable aspect of technological sound reproduction. “Noise is needed for messages,” I cited Serres at the outset of Chapter One, “sand is needed for stones” (1995: 132). It is this necessity that the role of noise in Shannon’s model of communication reveals as well: information theory no longer treats noise as external to the system of communication, but as internal to the transmission channel. Because no message can be produced and no signal can be transmitted without a physical channel, it can be argued that the transmission channel not only presupposes the presence of noise, but the presence of noise conversely also indicates that a transmission has taken place. If the presence of noise in signal transmission indicates the possibility of transmission itself, this presence must at least partly be responsible for the content of the message and thus, in the case of sound reproduction, the quintessential sounds of the media age. “How are harmony, singing, sound,“ Serres continues, “, and song born from this noise?” (132). The answer that dither provides is: by pushing periodic signals over a minimum threshold of representation, providing enough randomness for the picture of a baboon to appear. While its strategy of fighting noise with noise, firstly, problematises the fundamental premises of the conceptual logic of noise reduction, the “noise activated process” of stochastic resonance, secondly, thereby inspires the more general alternative concept of a noise resonance of sound reproduction. In contrast to the conceptual logic of noise reduction, this alternative concept highlights the foundational role of noise and randomness for the formation of the specific sounds of the media age, NOISE RESONANCE | CONFRONTING THE FUZZINESS OF THE REAL 115

resonating sonic messages that travel through the countless transmission channels of sound media in the receptive ears and brains of listeners. The concept of the noise resonance of sound reproduction will be further developed in the coming chapters, but doing so requires a better understanding of the discursive framework that supports the conceptual logic of noise reduction itself. This is why Chapter Three takes a closer look at the discrepancies between the symbolic idealisations at the heart of the mathematical analysis of sound and the technological processes that implement these idealisation into physical hardware. Extending upon the analyses of concrete technologies that revealed these discrepancies over the course of Chapters One and Two, Chapter Three turns to the mathematical and physical principles of sound analysis that established the scientific context for the emergence of technological sound recording over the course of the nineteenth century. More specifically, in order to trace the discursive origins of the ideas of purity and clarity that define the conceptual logic of noise reduction up to this day, I will look at the history and principles of one of the most ground- breaking mathematical tools for the analysis of sound: so-called Fourier analysis and the corresponding concept of the ‘sine wave’ as the representation of a single, pure frequency, the ‘basic element’ of sound.

NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 117

Chapter 3: Infinite time or perfect transience | ideal filters and the temporality of sound

3.1 Toward the limits of representation a) From technical media to mathematical physics In Chapter Two, the successive analyses of analogue noise reduction and digital dithering showed how the conceptual logic of noise reduction that took hold in the context of communication engineering in the 1920s and 1930s and was consolidated by information theory in the 1940s and 1950s, supports the myth of perfect fidelity. Firstly, the close reading of analogue dual-ended noise reduction systems revealed that according to the conceptual logic of noise reduction, noise is everything that can and will be reduced, which is why their operations can only reduce what is defined as noise prior to reduction. The analysis also emphasised that, because something always sticks to the signal, noise reduction systems affect the transmitted signal itself. Secondly, although the process of digitisation is based upon the strict separation of signals and noise, dither, introduced in digital systems in order to cope with structural limitations caused by this very strictness, suggests that the role of noise in sound recording is more fundamental than the myth of perfect fidelity and the conceptual logic of noise reduction suggest. The necessity of dithering points to a gap between, on the one hand, ideas of ideal representation and perfect reproduction and on the other hand the physical operations of technical filters that always affect the transmitted signal in the process of filtering. 118 NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE

Regarding this gap, the statistical similarity between the small extra that is added to digital signals in the form of quantisation errors and the small extra added to analogue signals in the form of random noise indicates the relation between the random noise of analogue systems and the quantisation errors of digital systems. In turn, this relation is indicative of more fundamental limitations of technological reproduction in general, limiting both analogue and digital media. In order to assess this connection between the limitations of analogue and digital media and gain a better understanding of their relation to the role of noise, this chapter looks beyond the analogue/digital-divide to focus on the conceptual foundations of the analytical representation and technological reproduction of sound as such. From Section 3.2 onward, the analysis thereby focuses on the nineteenth century development of Fourier analysis and the conceptualisation of the notion of the sine wave as the basic element of physical sound. Most significantly, it will look at the curious fact that Fourier analysis produces idealisations of physical sound waves only on the expense of symbolically doing away with the representation of their temporal duration: mathematically, a pure sine wave is by definition infinite. By exploring these conceptual foundations of the dominant model for analysing sound phenomena over the past two centuries and asking how they relate to physical sounds that extend in space and change over time, this chapter introduces the issue of temporality as a fundamental factor for understanding the difference between idealised, mathematical representations and technological reproduction. To be able to better relate this issue of temporality with the main questions regarding the role of noise in sound reproduction, however, the first section of this chapter turns to Abraham Moles’ description of two so- called “uncertainty principles stemming from the nature of things,” which are fundamental in communication engineering and information theory (1966: 5). Besides introducing the issue of temporality, further developed in the discussion of Fourier analysis and the sine wave from Section 3.2 onward, these uncertainty principles also provide a physical explanation for Shannon’s conceptualisation of the role of noise as both a prerequisite for and a disturbance of all successful signal transmission. NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 119

These two uncertainty principles, described in the third chapter of Moles’ Information Theory and Esthetic Perception are, he remarks, analogous to the better-known uncertainty principle in quantum mechanics, first formulated by Werner Heisenberg in 1927. Heisenberg’s principle states that a subatomic quantum particle cannot simultaneously have an exact location and an exact momentum, which means that if we observe where the particle is, we do not exactly know when it is at that position and, vice versa, if we know when the particle is observed, we cannot tell exactly where it is located.46 Similar to Heisenberg’s uncertainty principle, the principles described by Moles constitute trade-offs between the extent to which a communication system can produce accurate representations or reproduction of different parameters of transmitted signals. If the precision of the system toward one of two parameters increases, the precision toward the other decreases and vice versa. The first principle describes such an uncertainty relation between the system’s sensitivity to low amplitude signals and its sensitivity to the frequency spectrum (or bandwidth) of the signal. The second principle indicates an uncertainty relation between the system’s sensitivity to low amplitude signals and its sensitivity to the duration of the signal (Moles 1966: 83-87). Notably, as I will explain in the next section, the causes and effects of both principles are physically related; and it is with the second principle—the trade-off between amplitude and duration—that the issue of temporality comes into play.

46 Regarding the relation between Heisenberg’s uncertainty principle and the uncertainty principles in communication engineering and information theory, Bart Kosko points out that in the context of signal processing “versions of this uncertainty trade-off appeared at Bell Laboratories and elsewhere” already before the publication of Heisenberg’s version in 1927 (2006: 113). In “Spatio- Temporal Continuity, Quantum Theory and Music” (discussed in more detail in Section 3.4), physicist Norbert Wiener shows how Heisenberg’s uncertainty principles and the uncertainty principles in communication engineering and information theory can be explained “through the same harmonic analysis” (Wiener 1976: 545-546). Even more so, mathematician Gerald Kaiser argues that, “contrary to some popular opinion,” the uncertainty principle “is a general property of functions, not at all restricted to quantum mechanics” (Kaiser 1994: 52). 120 NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE

b) Uncertainty principles and the limits of physical filters The first uncertainty principle is a trade-off between the minimum signal amplitude and the maximum frequency range of a reproduced or transmitted signal. It states, in short, that the narrower the frequency bandwidth of a transmission channel is, the more sensitive the channel is to low amplitude signals; and, vice versa, the broader the frequency bandwidth, the less sensitive the channel is to signals with a low amplitude. This means that, in principle, a signal with a broader frequency range will have a smaller dynamic range, and a system with a larger dynamic range will have a narrower frequency range. “The information gained in one direction is lost in the other. What is gained in sensitivity is lost in the variety of elements” (1966: 84). As Moles goes on to describe, the reason for this trade-off between sensitivity for low amplitudes and frequency range is the presence of noise. Random noise, Moles explains, constitutes the necessary background for the emergence of physical signals, because the randomness of noise provides the contrasting ‘ground’ for the ‘form’ of more periodic signals to stand out from (77). The presence of noise is therefore a physical precondition for the emergence of clearly distinguishable signals, which is also what Shannon shows in the “Mathematical Theory of Information” and “Communication in the Presence of Noise”: the noisiness of the channel provides the necessary ground for the signal and is inherent to its transmission.47 Although the statistical approach of information theory might suggest that noise and signals can indeed be separated on the basis of this difference between the randomness of noise and the periodicity of signals—conceptualised as the difference between entropy and redundancy— very low amplitude signals invariably run the risk of drowning in random background noise. Physically, Moles writes, Einstein proved that “in the last analysis,” this “background noise is due to the

47 In “The Mathematical Theory of Communication,” Shannon writes: “ordinarily, channels have a certain amount of noise, and therefore a finite capacity, exact transmission is impossible” (1964: 108). In “Communication In The Presence Of Noise,” he states that “we cannot send continuous information exactly over a channel of finite capacity” (1998: 457). NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 121

agitation of electrons in conductors,” which means that random noise is present down to the level of elementary particles (84). The notion of elementary background noise is therefore closely related to the physical concept of heat as the random movement of molecules. This makes it less surprising that the most influential mathematical theory on the analysis of sound waves, the early nineteenth century “Fourier analysis” (of which I will come to speak in more detail in the next sections) was not conceived in the context of acoustics but as an analysis of the propagation of heat waves. Like random noise, Serres writes in Genesis, the notion of heat “covers […] none other than the mythical concept of chaos”; and like noise, heat is a necessary prerequisite for signal transmission (1995: 103). Without heat, at the absolute thermodynamic zero, there is no movement of electrons and no signal transmission can take place; too much heat, conversely, can cause a system to overload and breakdown. “Heat a little,” Serres writes in The Parasite, “I hear, I send, I pass; heat a little more, everything collapses”—every music lover or musician who ever drove his amplifier beyond its maximum capacity knows what this collapse sounds like (1982b: 194). As repeatedly showed in 1967, the only way to drive up the heat even more is to set your guitar on fire. Thus, both the random agitation of heat waves and the physically related randomness of noise are agents of signal transmission, but also of system breakdown. Random noise provides the physical ground for the emergence of signals, but simultaneously threatens to overtake and destroy them (Moles 1966: 76). If the energy level of the background noise is substantially higher than that of the signal, it will drown in noise and be lost; and the weaker the signal or the lower its amplitude level, the higher this risk is. To prevent this drowning of the weakest parts of reproduced or transmitted signals, electronic amplification can raise the energy level of the signal above the minimum background noise. Because the energy of the noise floor is amplified together with that of the signal, however, and because this amplification process requires energy and therefore generates noise itself, all amplification eventually reaches a point at which, as Moles puts it, the entire signal “drowns in erratic background noise [and] all 122 NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE

further amplification increases the noise and the signal simultaneously” (1966: 85). Beyond this point, further amplification is useless. The opposite strategy is to minimise the background noise, the level of which, Moles explains again following Einstein, is determined by two factors. Given the aforementioned relation between random noise and the random movement of molecules, the noise level is, firstly, proportional “to the absolute temperature” (84). When system temperature approaches the absolute thermodynamic zero, the lowest possible noise level is achieved. Secondly, background noise is proportional “to the frequency band considered” (84). The narrower the bandwidth, the less potentially noisy frequencies it contains. Because amplification reaches a point of saturation and lowering the temperature of sound equipment is only possible within reasonable limits, this last scenario—reducing noise by narrowing the bandwidth—is the most practical way to make the system more sensitive to lower amplitude signals. It can be implemented by introducing the kind of adaptive filters or noise gates described in Chapter One, Section 1.2c, which cut off or limit the signal’s frequency range and filter out part of its spectrum. When these filters are narrowed, they focus the channel capacity on an increasingly small band of frequencies and filter out more and more random noise. As the filter becomes narrower and narrower, however, the risk of filtering out frequencies that belong to the signal instead of the noise increases as well. By narrowing the pass band of the channel, one runs the risk of losing part of the signal itself. Imagine, for example, you want to remove the noise of an old 78 rpm recording of a symphonic piece of music. By applying a noise filter to filter out the parts of the spectrum most heavily affected by noise, the bandwidth of the signal is narrowed and the information (symphonic music) is transmitted with less disturbing noise. The frequency range of symphonic (and, in fact most) music, however, is generally quite broad and filtering out the parts affected by noise will in most cases also affect the signal, filtering out parts of its low- and high-end. When the filter is narrowed, larger portions of the musical signal will be lost, up to the point where valuable musical information, for instance the low rumble of or the higher parts of the string section, are sacrificed. Hence, although a smaller frequency range increases the NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 123

sensitivity for capturing or transmitting low amplitude signals within this given frequency range, it also decreases the maximum bandwidth of the transmitted signals, thus running the risk of throwing out the signal-baby with the noise-bathwater. The symbolic limit case of an increasingly narrow noise filter would narrow the channel’s bandwidth to a single frequency (Moles 1966: 87). Such an ideal filter is so narrow it would block or filter out all frequencies—noise or otherwise—but one. By that point, all characteristics of the signal, defined by the specific composition of its frequency spectrum and its development over time, would be filtered out; the spectral richness of the symphonic music on the noisy 78 rpm record would be reduced to one perfectly unambiguous frequency. By lack of any spectral information about the music, “the only question which could be posed” of such a signal, Moles explains, “would be whether or not the message was there”; and the only answer the single frequency output of the filter would provide, would be a binary, one bit piece of information: yes or no, the signal is either present or it is not (85). Because an ideal filter would remove all information that enables the receiver to determine what kind of message it receives, it has become impossible to tell whether the recording contains symphonic music, noise, speech, a rock performance, a field recording or anything else. What is gained in sensitivity and precision by applying an ideal filter that removes all potential noise and ultimately produces a completely unambiguous single frequency, is lost in information carried by the other elements of the signal’s frequency spectrum that are all filtered out. Variety is traded for sensitivity. Although one can be certain no noise affects the ideally filtered version of the recording on the noisy 78 rpm record, along the process of filtering all relevant information—musical or otherwise—is lost. The single frequency that gets through the filter only tells whether the message is there or not. Hence, neither signal amplification nor any form of noise reduction gets around the trade-off between sensitivity to low amplitude signals and the reproduction of a broader frequency spectrum. When no filter is applied, the frequency band is limited by physical background noise that drowns out all signals below a minimum amplitude threshold; and 124 NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE

although the energy level of low amplitude signals can be amplified, too much amplification will raise the noise floor up to a level at which it threatens to overtake the signal entirely. Conversely, installing a noise filter to narrow the frequency spectrum and reduce the noise floor introduces the risk of filtering out parts of the signal itself, up to a point at which only a single frequency remains, carrying as little information as the remaining noise in the amplification scenario. The trade-off of the first uncertainty principle, summarised in table 1, thereby shows how the physical presence of noise, down to the most elementary level, simultaneously provides the necessary background for periodic signals to appear and fundamentally limits the maximum capacity of any transmission channel.

Table 1 The First Uncertainty Principle.

Moles: “Error in amplitude ! error in frequencies = constant” o System is more sensitive to lower amplitudes = narrower bandwidth = less noise = narrower frequency spectrum o Broader bandwidth = more frequencies are transmitted = more noise = system less sensitive to lower amplitudes

Source: Abraham Moles, Information Theory and Esthetic Perception, Translated by Joel E. Cohen, Urbana, University of Illinois Press, 1966: 85.

The second uncertainty principle follows directly from the first. Besides involving the spatialised parameters of amplitude and frequency, it consolidates the physical limitations set by the first principle by including the duration of the signal in the equation, thereby addressing the complete spatio-temporal reality of physical signals. To explain how this second trade-off relates to the first uncertainty principle, Moles describes a thought experiment involving a device that would hypothetically solve the trade-off between amplitude and frequency range. Such a device would consist of a large number of ideal filters, lined up in sequence and each attuned to a different frequency. Hypothetically, this machine would enable the transmission of a many single frequencies without any interfering NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 125

background noise (87). It is, however, both logically inconsistent and physically impossible. Firstly, determining to which frequency each of the ideal filters should be tuned requires complete knowledge about the frequency spectrum of the signal one intends to transmit as well as the noise that has to be reduced. Because this unambiguous information about the signal’s frequency spectrum is exactly what is unavailable, this is poses a logical inconsistency: if it were possible to unambiguously determine the frequency spectrum of both signal and noise, one could achieve their complete separation without the need for Moles’ hypothetical machine (87). Dolby’s noise reduction systems, for example, might be very efficient in attenuating noisy passages by creating, as I quoted Dolby Laboratories in the Chapter Two, a potentially “infinite number of filters,” these filters can only be applied to those parts of the signal that have already been identified as ‘noise’ in the first place (Dolby SR 1987: 5). Because such complete information about frequency spectrum is not available prior to the filtering operation itself, Moles writes, transmission of the complete signal without any noise would require “an infinite number of arbitrarily narrow filters” (87). Due to the physical impossibility to construct “an infinite number of filters,” the problem of determining which frequencies belong to the signal and which to the noise remains. Besides this logical inconsistency, however, the very concept of an ideal filter attuned to one single frequency runs into an even more fundamental limit due to the fact that no physical filter responds instantaneously to its task (86). Instead, every filter requires a minimum amount of time to process the input signal. This necessary response time causes a slight delay in the production of the output. Similar to the first uncertainty relation between the bandwidth of the channel and its sensitivity to low amplitude signals, Moles explains, the delay caused by the response time is “proportional to the narrowness” of the filter (86). The narrower the filter, the more frequencies it filters out, the more time it needs to produce an output. This time increases and decreases proportionally in relation to the signal’s bandwidth. Consequently, “instead of overturning the [first, MK] uncertainty principle,” the theoretical device 126 NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE

consisting of a large number of ideal filters “only provides a different formulation of it by changing the magnitudes to which it applies” (87). Because the frequency components of most natural signals tend to change very rapidly, they often change faster than the minimum response time and corresponding delay of a filter. This causes inaccuracies in the output signal. If the frequency spectrum of the input changes before the filter produced an output, the filter does not register this change and the duration of the filtered frequency components will not be registered correctly. Although a narrower channel will reduce more noise and transmit lower amplitude signals, a narrower filter requires more time to filter out more frequencies, which causes a longer response time and longer delay. Following the second uncertainty principle, described in table 2, the gain in sensitivity to low amplitude signals goes at the cost of the sensitivity to the duration of these signals and vice versa.

Table 2 The Second Uncertainty Principle.

Moles: “Error in amplitude ! error in duration = constant” o System more sensitive to low amplitude = narrower pass-band (= less noise) = longer delay = more uncertainty about duration o Measure precise duration = shorter delay = pass band wider (= more noise) = system less sensitive to low amplitude

Source: Abraham Moles, Information Theory and Esthetic Perception, Translated by Joel E. Cohen, Urbana, University of Illinois Press, 1966: 87.

Most significantly, in the case of the hypothetical ideal filter with an output of one single frequency, the time required to produce the output mathematically tends toward infinity. This means that filtering out one single frequency would require an infinite amount of time. Because the filter would never stop filtering, the perfectly unambiguous output is never produced. Hence, an ideal filter is mathematically perfect, but physically impossible to implement. In theoretical physics and communication engineering, such ideal filters are used to model physical processes or the operations of technical NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 127

media. On the basis of mathematical analysis, the models thereby reinforce the ideal of a fundamentally noiseless world.48 Nonetheless, as I explained following to Siegert’s Passage des Digitalen in Chapter Two, Section 2.4c, such idealisations have been decisive for the further development of theoretical physics and technical media. The question is, therefore, how these ideal models relate to the physical filtering operations of technical media; or how what Serres calls the idea of a “world without noise” nonetheless shapes our understanding of the physical world in which noise is always present (2008: 126). c) Models and idealisations: ideal and physical filters In Into the Universe of Technical Images, philosopher Vilém Flusser suggests that practices of modelling do not objectively picture or represent the physical world, but order and structure it (2011: 170). The “so-called natural laws” of physics, he argues, are not objective descriptions of external physical processes, but ways to decode the "gigantic quantity of indications, signs, clues” we are confronted with on a daily basis (2011: 46). Models are not neutral. Applying idealised versions of complex processes to symbolically create order and reduce complexity, they introduce new conceptual categories that shape our perspective on and approach to physical phenomena themselves.49 As our attempts to represent and reproduce physical processes run into the complexity of these very processes, idealised models are introduced to impose order, regularity and linearity. Rather than creating

48 "Mathematics,” Serres writes, “is the kingdom that admits only the absolutely unavoidable noise, the kingdom of the quasi-perfect communication, the manthánein” (1982a: 69). 49 "Models,” Flusser writes, “give form to a world and a consciousness that has disintegrated; they are meant to ‘inform’ that world. Their vector of signification is therefore the reverse of that of earlier images: they don't receive their meaning from outside but rather project meaning outward. They lend meaning to the absurd" (2011: 170). Similarly, in his contribution to the essay collection Ways of Thinking, Ways of Seeing. Mathematical and Other Modelling in Engineering and Technology, engineer John Monk writes: “it is tempting to imagine that a model or theory is an accurate reflection of what takes place in reality; however, prominent nineteenth century physicists and latterly pragmatist philosophers have insisted that our descriptions of reality are of our own making and are a product of our institutions and customs. Models as part of our descriptive practices, therefore, make a contribution to the of reality” (2012: 2). 128 NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE

exact representations of natural phenomena, they break up complex processes and bring them down to manageable proportions, thereby enabling a conceptual understanding of the processes at hand and in some cases their technological reproduction or transformation. This means that without the mathematical conceptualisation of ideal filters to enhance our understanding of their operations and make them manageable, the technological development of physical filters would not be have been possible. In the physical reality of technical media, the seamless output of ideal filters used to model their operations is challenged by the physical limits imposed by the uncertainty principle. This applies as much for digital media as it applies to analogue media. Due to the physical uncertainty relation between accuracy in amplitude and accuracy in frequency or duration, even technical media that are able to process signals spectacularly more precise than what our natural senses can achieve eventually run into the physical limits posed by these trade-offs. As explained in the discussion of dithering in the second half of Chapter Two, a higher bit rate decreases the number of quantisation errors and allows a more precise representation of the amplitude value of each sample. Higher bitrates thereby reproduce lower amplitude signals and achieve a larger dynamic range, but, following the uncertainty principle, measuring and processing these more accurate values requires filters with a longer response time and corresponding delay, introducing errors with respect to the duration of the signal. Conversely, a higher sample rate enlarges the bandwidth and allows the reproduction of broader frequency spectra, but as higher sample rates increase the number of samples per second, they also require shorter samples with a shorter response time. This shorter response time, in turn, decreases the sensitivity to low amplitude values. Hence, the presence of background noise or its digital equivalence in the form of quantisation errors limits a system’s sensitivity to low amplitude signals. Narrowing the bandwidth increases this sensitivity, but excludes parts of the frequency spectrum, like increasing the bitrate increases this sensitivity, but allows more inaccuracies in terms of signal duration. Conversely, enlarging the bandwidth or raising the sample rate NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 129

increases the system’s sensitivity to broader frequency spectra, but decreases its sensitivity to low amplitude values. There is no way out. On the basis of the effects of the uncertainty principles on the transmitted signal, it can be argued that the difference between the ideal model and the physical system is characterised by the presence of noise—random physical noise in the case of analogue media, and communicational noise in the form of error and distortion in digital media. This means that the distance between the ideal models of mathematical analysis and the operations of technical media is marked by the limits posed by the uncertainty principle and the physical presence of noise. Conceived in the context of twentieth century communication engineering and put in the terminology of information theory by scholars such as Moles in the decades after World War II, the uncertainty principle of signal processing maps the distance between the idealised domain of mathematical models, which I will call the domain of the ideal filter, and a contrasting domain of physical filters that operate in physical reality. This distance is an example of what Siegert calls the rupture that appeared with the development and consolidation of modern mathematical analysis between Leibniz’s invention of infinitesimal calculus in the late sixteenth century and Joseph Fourier’s application of trigonometric functions to the analysis of complex wave phenomena in the early nineteenth century. With Fourier analysis, the symbolic idealisations of mathematical analysis produced representations of complex physical processes that would eventually enable the autonomous (re)production of some of these processes by technical media. The invention of Fourier analysis thereby marks a crucial moment in the rupture between representation and represented and the emergence of a new order of, on the one hand, analytical idealisations based on mathematical models and, on the other hand, physical reproductions produced by technical media. Although the mathematics underpinning the uncertainty principle can ultimately be traced back to Fourier’s work as well, his spectral analysis of complex wave phenomena and the resulting figure of the sine wave as the mathematical description of a single frequency are conceptually constitutive of what I call the domain of the ideal filter. Furthermore, Fourier analysis also marks the origin of the ideas of infinite 130 NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE

precision and maximal purity characteristic of the conceptual logic of noise reduction up to the present day. Understanding what the noiseless domain of the ideal filter tells about the role of noise in the domain of physical filters therefore requires a closer look at Fourier analysis and the figure of the sine wave.

3.2 Fourier analysis: history and basic principles a) Fourier’s analytical theory of heat From its official publication in 1822 onward, Fourier’s method for the analysis of physical wave phenomena became one of the most important analytical tools of modern times.50 As historian of Olivier Darrigol claims in "The Acoustic Origins of Harmonic Analysis," part of this tremendous success was due to the fact that Fourier’s work marks the final transition from a type of physics in which mathematical analysis was of secondary importance to a modern kind of theoretical physics that makes extensive use of mathematical models (2007: 397).51 This supports Siegert’s claim that Fourier’s work is a decisive step in the development of modern mathematical analysis and the completion of the rupture in the order of the representational (2003: 192). The origins of Fourier analysis, however, are more modest. It was developed in order to analyse, as physicist I. Grattan- Guinness describes in his survey of Fourier’s life and work, “the diffusion, or propagation, of heat in continuous bodies (1972: vii). Jean-Baptiste Joseph Fourier had been a prodigious mathematician from early childhood, but his experiences in the hot desert sun of Egypt, where he worked in the service of Napoleon during the last years of the seventeenth century, provided him with a livelong obsession with heat. Because, as biographer John Herivel describes, Fourier was unable “to acclimatize himself to the change from Egypt” upon his return to France

50 In the words of mathematician T.W. Körner, cited by science journalist Barbara Burke Hubbard, Fourier analysis is “built into the commonsense of our society” (Körner in Burke Hubbard 1996: 8). 51 As Fourier’s biographer John Herivel writes: “the whole subject of the propagation of heat [was] reduced to a matter of mathematical analysis which Fourier then proceeded to apply to one case after another” (1975: 152-3). NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 131

around 1801 or 1802, “the question of heat, its loss by propagation in solids and radiation in space, the problem of conserving it […] can never have been out of his mind for long” (1975: 99). From 1804 at the latest, Fourier, by now appointed prefect of the newly created département Isère in Grenoble under Napoleon, spend most of his time off to develop his theory on the propagation of heat. The first version of Fourier’s treatise “On the Propagation of Heat in Solid Bodies” (generally referred to as the Draft Paper) was completed in 1807 and met with considerable resistance from several leading physicist of the time, most notably Siméon Poisson, Jean-Baptiste Biot and Pierre- Simon Laplace. In 1811, Fourier entered a revised version of the treatise in the contest for that year’s grand price in mathematics of the Institut de France, the topic of which—‘the propagation of heat in solid bodies’—not entirely coincidentally suited Fourier’s project like a glove. By correcting several mistakes and amending and extending upon the Draft Paper, he won the contest, but the committee was still under the impression that his solutions were “not exempt of difficulties” and left “something to be desired” (committee report cited in Herivel 1975: 103). Due to this persistent opposition and his on-going and quite turbulent political career, the final version of the “Analytical Theory of Heat” was only published in 1822, after Fourier was appointed permanent secretary for the mathematical sciences at the Académie des Sciences in Paris and eight years before his death in 1830. The solutions Fourier develops in the “Analytical Theory of Heat” mark the transition from a concept of physics primarily based on empirical observations and experimental set-ups to a type of physics that relied more and more on purely mathematical concepts; and much of the professional resistance he faced in the fifteen years between the Draft Paper in 1807 and the publication of the final treatise in 1822 was due to this radical approach. Steeped in a deep knowledge of mathematical and physical work conducted in the years preceding his treatise, Fourier was especially good at, Herivel writes, taking “an essentially complex problem and make it amenable to mathematical treatment while simultaneously providing a solution yielding a good approximation to the actual physical situation in a wide range of cases” (1975: 213). He thereby introduced a level of 132 NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE

mathematical abstraction that would quickly become standard practice in the years following its publication, but was still highly debated at the late eighteenth and early nineteenth century. b) Explaining the Fourier transform Fourier analysis mathematically transforms the function !, representing the development of a complex waveform (heat, sound, light, radiation, etc.) over time into a series of sine and cosine values corresponding to the amplitude, phase and frequency of every individual wave in that waveform. The outcome of this transformation is a mathematical representation of the frequency spectrum of the complex waveform, consisting of many sinusoidal components called ‘sine-waves.’ This procedure is called the Fourier transform and its simplest rendition, which strictly applies to periodic signals, is the Fourier series.52 By definition, a periodic signal repeats identically over and over again. It might be a complex waveform composed of many individual waves that oscillate at different frequencies, but for a signal to be purely periodic, each complete cycle must be exactly the same. Given this periodicity, every single cycle contains all information on the frequency spectrum of the signal and determining this spectrum therefore only requires the analysis of one complete cycle. Consequently, the timeframe necessary for the analysis of a strictly periodic signal is exactly the time it takes to complete one cycle (Tempelaars 1996: 129). The resulting Fourier series represents the sine and cosine values corresponding to each individual frequency in each cycle of the periodic waveform. The sum of all these values expresses the total amplitude of the original waveform. By transforming a periodic signal into its Fourier series, Fourier analysis treats the signal as if it repeats identically without beginning or end. This idealisation is a consequence, as Gareth Loy explains in the first volumes of his Musimatics, of Fourier’s use of the ‘trigonometric’ or

52 Some suggest the Fourier series applies to periodic functions and the Fourier transform to non- periodic or quasi-periodic functions. Engineer Stan Tempelaars, however, argues that “the Fourier series […] is the Fourier transform of a periodic function,” which means that the Fourier series is a particular form of the more general Fourier transform (1996: 142). NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 133

Figure 10 Harmonic Circles from Mehmet E. Yavuz, “Fourier Series Animation Using Circles,” 4 Mar. 2014, Youtube , uploaded by Meyavuz, www.youtube.com/watch?v= LznjC4Lo7lE&feature=youtu.be, accessed 1 Aug. 2016.

These video stills illustrate the relation between a circle and the sinusoids represented by Fourier analysis. The upper image shows the representation of four individual sine waves, the second shows the summation of these four waves into one complex, periodic waveform.

134 NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE

‘circular’ sine and cosine functions to represent the values of each wave (2006: 140). As illustrated by the harmonic circles in figure 10, the mathematical origins of the sine and cosine function are the representation of the ratios between the sides of a triangle (hence the name ‘trigonometric’), which are in turn geometrically derived from the relation between a triangle and a circle (hence the name ‘circular’). Because of this relation with the circumference of a circle, which is mathematically infinite by definition, Loy writes, “sinusoids […] have no beginning and no end” (2006: 140). As a consequence of this symbolic infinity of sine waves, the symbolically infinite repetition of a periodic signal in the Fourier domain is an analytical given. A Fourier series therefore represents the signal and all of its individual waves as if they have oscillated and will oscillate forever. This infinite repetition poses little problems for the Fourier transform of periodic signals, because the fact that they are periodic already implies that they theoretically repeat endlessly; otherwise they would not be purely periodic. Non-periodic signals, on the other hand, do not repeat cyclically; they change, often rapidly, over time. Because of this, a regular Fourier series cannot represent non-periodic signals as the sum of its individual sine and cosine values. Analysing non-periodic signals requires a trick. To transform the function of the development of a non- periodic signal over time into a representation of its frequency spectrum, Fourier analysis applies a second idealisation: it treats the non-periodic signal as if it is periodic. The Fourier transform of a non-periodic signal treats the entire signal as one cycle of some imaginary periodic signal. In order to achieve this, the analysis represents the temporal factor ! , representing one full cycle of a signal, as if it is infinitely long. Effectively, this means the Fourier transform of a non-periodic signal renders the factor time in terms of duration, that is the time of things that begin and end, no longer relevant (Tempelaars 1996: 129). The infinite factor ! is the idealisation required to crack the problem of analysing the frequency spectrum of non-periodic functions. In order to subsequently derive the frequency spectrum of this symbolic infinite cycle, the Fourier transform of non-periodic signals replaces the sum, adding up all sine and cosine functions of a Fourier series, with an integral that integrates all sine and cosine values within the NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 135

timeframe of the new, artificially imposed ‘period’ of the non-periodic signal (Tempelaars 1996: 129). According to Wolfgang Ernst, this Fourier integral can therefore be considered a conceptual “bridge” between the relatively straightforward Fourier transform representing periodic signals as Fourier series, and the mathematical ingenuity necessary to analyse non-periodic signals (2012a: 40). The integral, Ernst quotes engineering pioneer Ralf Heartley, “may be thought of as a mathematical fiction for expressing a transient phenomenon”—unpredictable, non-periodic waves that constantly change over time—“in terms of steady state phenomena”— periodic signals that repeat infinitely and unchanged (Heartley in Ernst 2012a: 40).53 The analytical idealisations of the Fourier transform thereby create a perspective from which all ambiguity is expelled and through which perfect clarity appears. This type of clarity goes, Siegert argues, at the costs of the concept of complete representability that had been the ideal up until and including Leibniz’s Law of Continuity (2003: 246). As shown in the Chapter Two, Section 2.4c, what Siegert calls Leibniz’s “noise order” was based on the assumption that human subjects unconsciously process many infinitesimally small perceptions—like the sound of every individual wave that makes up the roaring noise of the surf—and integrate these into one coherent impression, similar to, as Deleuze puts it, a fold that is continuously “folded within a fold” (Siegert 2003: 187; Deleuze 2006: 6). Although, Leibniz argues, the continuity of natural phenomena thereby remains inextricable and forever out of reach for human perception, God perceives every infinitesimal perception and guarantees the fundamental continuity of a world in which ‘nature does not make jumps.’ With Fourier’s talent to come up with, in Herivel’s words, “just the right sort of idealisation,” the Fourier integral introduces a crucial gap between the mathematical representation and the physical reality of wave phenomena. Although the representation closely approximates the

53 In the “Variables, Abbreviations, Definitions”-section of Tuning, , Spectrum, Scale, music theorist and electrical engineer William A. Sethares defines a “steady state” as “the part of a sound that can be closely approximated by a periodic waveform” and a “transient” as “that portion of a sound that cannot be closely approximated by a periodic signal” (2005: XVIII). 136 NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE

properties of the physical phenomenon, there is no exact correspondence. The output of the mathematical analysis always represents a hard limit to which the physical phenomenon infinitely converges. According to Siegert, the graphs produced by Fourier’s analysis of heat therefore represent “the seemingly sharp contours of surfaces, which are actually just the infinitely fine heat-shimmering of these surface themselves” (2003: 246).54 Hence, what appears in all sharpness and clarity in the Fourier domain is an idealised ‘steady state’ representation of constantly changing transient phenomena. c) From heat waves to sound waves: Ohm and Helmholtz Although Fourier’s own work was not concerned with the possible application of his analytical method in the field of acoustics, as Darrigol describes in his extensive article on the mathematical history of Fourier analysis, he did draw upon an acoustic problem widely discussed in the last decades of the eighteenth century. 55 This so-called ‘problem of the vibrating string’ (regarding the issue how to represent the vibrations of a string) occupied many of the mathematicians and physicists whose work inspired Fourier’s: Jean le Rond d’Alembert, Leonhard Euler, Joseph Louis Lagrange and Daniel Bernoulli (2007: 344). At the heart of the problem of the vibrating string lies the issue of the existence of partial tones ringing together with the fundamental frequency of a sound or, in other words, the question whether sounds consist of a single wave or a multitude of waves. This specific issue was not directly at stake in Fourier’s analysis of heat, but his application of trigonometric functions to the analysis of the physical behaviour of complex wave phenomena does mark the decisive step between Bernoulli’s work in the late eighteenth century arguing in favour of the hypothesis that sound consists of a large number of

54 “Es sind scheinbar scharfe Konturen von Flächen, die in Wahrheit nur das unendlich feine Hitzeflimmern der Fläche selber sind.” 55 Fourier’s only reference to sound in the published version of the Analytical Theory of Heat is the following passage in the Preliminary Remarks: “The problems of the theory of heat present so many examples of the simple and constant dispositions which spring from the general laws of nature; and if the order which is established in these phenomena could be grasped by our senses, it would produce in us an impression comparable to the sensation of musical sound” (2009: 8). NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 137

superimposed frequencies and Ohm’s and Helmholtz’s ground-breaking work in acoustics in the second half of the nineteenth century corroborating this hypothesis (2007: 401). Via Ohm and Helmholtz, Fourier’s analytical approach, partly based on Bernoulli’s solutions to the problem of the vibrating string, was brought back into discussions on the physical nature of sound and, by extension, music.56 In his mid-nineteenth century work on acoustics, musicologist Julia Kursell describes in “Experiments on Tone Color in Music and Acoustics: Helmholtz, Schoenberg, and Klangfarbenmelodie,” Georg Simon Ohm intended to “formalize” all available physical knowledge “about the movement of sound waves” by replacing the physical and physiological approaches common at the time with a mathematical approach largely informed by Fourier’s theorem (2013: 194). In order to prove the longstanding hypothesis that sound vibrations can be represented as a complex periodic waveform consisting of a multitude of smaller vibrations, Peter John Blamey writes in his dissertation on sine waves in , Ohm turned to Fourier’s recent analytical solution to the problem of the propagation of heat (2008: 29). As Kursell describes, in a famous dispute between Ohm and acoustician August Seebeck concerning the properties of the sound of a siren, Seebeck based his findings on “the sounds he had heard,” whereas Ohm “aimed at supplementing a mathematical theory for the periodic sound of the siren” (2013: 195). Thus, whereas Seebeck still relied on empirical observations made by his own faculty of hearing, “Ohm sought to illuminate physical phenomena with the help of their mathematical formalization” (2013: 196). Through the application of “the most recent knowledge of periodic phenomena” provided by Fourier analysis, Kursell argues, “mathematics even replaced the ear for Ohm” as his preferred tool for sound analysis (2013: 195). Furthermore, historian of science R. Steven Turner writes, although

56 Concerning Fourier’s role in the discussion on the properties of the vibrating string, Grattan- Guinness also notes that “Fourier 's work did not resolve the discussion but his results on trigonometric series had implications for the generality of that [analytical, MK] type of solution” (1972: 243). 138 NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE

Ohm “seemed not to have fully realised this until somewhat later,” his analytical approach to sound also pointed toward a new perspective on “the physiology of auditory perception” (1977: 7). Further establishing Fourier analysis as the quintessential theory on the nature of sound, the physical law that was to be called ‘Ohm’s law’ (or ‘Ohm’s acoustic law’ to differentiate it from his more famous work on electricity) laid the foundation for the hypothesis that, when listening, the ear itself performs some kind of Fourier analysis (1977: 5). Seebeck repeatedly challenged Ohm’s findings on the basis that, as Turner writes, his “theory failed in that it could not correctly predict the intensities of these components as they are actually heard” (1977: 7). Despite the fact that Ohm’s calculations were not entirely correct, however, the work of physicist Hermann von Helmholtz ultimately ensured the lasting success of his application of Fourier’s method to the analysis of sound. In his influential book On the as a Physiological Basis for the Theory of Music, published in 1862, Helmholtz set out to collect and further all available knowledge about the physical nature of sound and the physiology of hearing into one overarching theory, combining the mathematical analysis of Ohm with empirical experiments. In the process, he corroborated, corrected and expanded Ohm’s hypothesis that Fourier’s theorem proves the physical existence of partial tones (1977: 20). In an attempt to experimentally verify the outcomes of Fourier analysis and empirically proof the physical difference between musical (harmonic) sounds and non-musical sounds or noise, the strict periodicity that Fourier analysis mathematically represents encouraged Helmholtz to take, as Kursell puts it, “the steady, internal repetition of periodic sound waves” as the starting point for his experiments (2013: 192). In order to physically replicate the infinite vibrations suggested by Fourier’s mathematical model, Helmholtz tried to approximate “the mathematical description of a periodic wave as closely as possible" (205). To achieve this, he created an experimental set-up that produced sounds, which, like sine

NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 139

waves, “avoided any characteristic beginnings and endings” (192).57 The artificial sine-like sounds produced by these experiments, Kursell argues, effectively constituted a new type of sound for which no physical referent had existed up to that moment (192). With this combination of Fourier analysis and its application to experimental acoustics, Helmholtz forged a connection between the mathematical idealisations of Fourier analysis and the acoustic phenomena under investigation. Through this connection, the idea of the sine wave as an acoustical phenomenon appeared. Never explicitly mentioned as such by Fourier himself, the strictly periodic and ideally infinite sine wave was therefore not so much the subject of experimental analysis as it was produced by the analysis itself. Reinterpreted as an acoustic, instead of a purely mathematical object, the sine wave subsequently took hold of our sonic imaginary, because it provided a seemingly empirical basis for the age-old discursive connections between music, harmony and regularity that go back as far as Pythagoras’ theories on celestial harmony in the sixth century BC (Blamey 2008: 36, 62).58 Ultimately, the mathematical-acoustic figure of the infinite sine wave also constitutes the foundation for the technological ideal of perfect, noiseless signal transmission and the conceptual logic of noise reduction that dominates the discourse on sound recording from the late nineteenth century onward.

57 Kursell writes: “Helmholtz’ approximation of a mathematical model that implied the infinite duration of its elements could only succeed because the vibrations of the electromagnetically driven tuning forks reduced the habitual components of the beginnings and endings of sound. The synthesized sounds had no specific envelope—that is, they almost did not change in time—and the same held for their compounds. Such temporal aspects had to be ignored in Helmholtz’ approach" (2013: 206). 58 As Douglas Kahn writes in the introduction to Noise, Water, Meat: “The figure of vibration was upheld by the Pythagoreans, refurbished by neo-Platonic and neo-Pythagorean thought centuries later, and invigorated by scientific, Eastern and spiritist thought in the West in the nineteenth century. The monochord–the technology that underscored the harmonic totality of Pythagorean thought, the vibrating string structuring the cosmos–was so overcoded by the late-nineteenth century locus of vibrations in the synesthetic arts that it was functionally nonexistent, although the connections between acoustics, music, and mathematics, not to mention certain ambitions toward the cosmos, remained strong” (2002: 16). 140 NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE

3.3 The domain of ideal filters a) The ideal sine wave: infinite periodicity With Ohm’s application of Fourier’s theorem to the analysis of sound and Helmholtz’s subsequent experimental verification of its principles, the sine wave, a purely mathematical limit case that draws seemingly sharp contours around infinitely fuzzy sound waves, became something of a Platonic idea: an ideal, fixed, undividable point of reference for sonic purity toward which all physical sounds tend. Following Siegert’s reading of Serres in the first chapter of Cultural Techniques, the relation between this idealised mathematical object of the sine wave and real acoustic phenomena that approach but never attain its ideal parameters is a relation between a “symbol, as defined by logicians, and signal, as defined by information theorists” (2003: 20, emphasis in original). A sine wave is not a signal in the physical sense, but an analytical symbol representing the discontinuous limit case of vibrational phenomena. According to Serres’ concept of communication, Siegert writes, rendering such an unambiguously clear symbol requires a conceptual “cleansing of the ‘noise of all graphic form’” to separate it from all material references (2003: 20). “The mathematician,” Serres explains in “Platonic Dialogue,” “does not see any difficulty on this point”; on the basis of the mathematical manipulation of written signs, the conceptual reduction of the noise of their physical foundation is perfectly possible (1982a: 68). This is a matter of mathematical abstraction, or “a process by which one passes from concrete modes of thinking to one or several abstract forms” and thereby “means to eliminate noise as well, in an optimal manner” (68). Thus, Serres argues, “to isolate an ideal form is to render it independent of the empirical domain and of noise,” which comes down to the removal of any reference to its physical production and transmission as signal (70). This means the reduction of any reference to the material transmission channel and the complete reduction of noise. Hence, the conceptualisation of the sine wave as a purely mathematical symbol presupposes a process of abstraction that separates the symbol from its material reference. Subsequently reconceptualising the sine wave as an acoustic object turns it into an ideal signal, discursively NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 141

positioned in between the purely symbolic domain of mathematical analysis (the domain of the ideal filter) and the physical domain of acoustics (the domain of physical filters). As an ideal signal, the sine wave presupposes complete noise reduction for which, as Serres writes in The Parasite, “the gate can be, must be, maximally narrow ” (1982b: 153). This shows that the representational model that is required for conceptualising the sine wave as an ideal acoustic object resembles an ideal filter. It now becomes clear that the operations of Fourier analysis and the concept of the sine wave are both subject to the logic of the uncertainty principle. If the concept of the ideal sine wave requires the analytical removal of any reference to the physical channel, the operation of Fourier analysis itself, by representing sound in terms of a series of sine waves, can be interpreted as an ideal, infinitely accurate, spectral filter, incredibly sensitive to all frequencies. Following the uncertainty principle, the closer a physical filter comes to this ideal state, the narrower its bandwidth and the more time it needs to complete the operation. As I described in Section 3.1b, at the analytical limit case of this process, the response time of the filter mathematically tends toward infinity. At that point, the physical signal turns into an ideal signal: a pure sine wave. Hence, the temporal infinity of the sine wave, which is the analytical ground for its absolute periodicity and a mathematical consequence of its relation to the circle, follows directly from the uncertainty principle: only when the factor ! (in mathematical terms) or the response time of the filter (in engineering terms) is infinite, an ideal filter symbolically produces an ideal sine wave. b) Event and series: extremes of the uncertainty relation Combining Serres’ reading of the model of communication with Moles’ description of the uncertainty principles in signal processing shows how the ideal of a perfectly noiseless signal is based on the symbolic cornerstone of the mathematical analysis of sound: the sine wave. This sine wave is first and foremost a mathematical symbol that exists solely in the domain of the ideal filter. Noise, Serres writes on the other hand, is “the empirical portion of the message” tied to the domain of physical filters that is defined by the impossibility of the analytical purity of Fourier analysis (1982a: 70). The most significant part of this connection between the uncertainty principle, the noise of physical systems and the idealised 142 NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE

symbol of the sine wave is the role of temporality. The symbolic representation of a sine wave implies the complete removal of any reference to a channel. The complete removal of any reference to a channel would physically require the complete reduction of noise; and the complete reduction of noise symbolically requires the removal of the factor of time. In his 2003 essay “Lightning and Series – Event and Thunder,” Kittler conceptualises this relation between the mathematical idealisations of Fourier analysis and the removal of the temporal factor (2006a, 63-74). In his argument, ranging from Greek mythology, via early modern times to our current age, he sets out to explain how the nature of any (ideally infinitesimally) short event, like the very brief energy discharge of a lightning bolt, can be analysed in terms of a series, like the acoustic reverberations of thunder. Because of its briefness, the only information to be derived from a lightning flash is its “thatness” (dass es ist) or the simple fact that it took place (70). Understanding its “whatness” (was es ist), on the other hand, requires some form of repetition (70). In the case of lightning, this repetition comes in the form of thunder, which, contrary to the lightning flash, “can be heard and pondered” (65). 59 Through this repetition, or indeed this “frequentia, the return” in the form of a series, there is time to analyse the event and acquire some sort of stable knowledge about what took place (69). This, Kittler argues, is what Fourier analysis does: it transforms a brief, random, constantly changing signal into a series of repetitions, or frequencies. The frequency domain created by Fourier analysis thus relates to the original signal as the thunder relates to the lighting, or a series relates to an event. For Wolfgang Ernst, this means that the “analytical procedure” of the Fourier transform, representing transient wave phenomena as sets of infinitely oscillating sine waves, “assumes the world is a sinusoid harmony” (2012a: 40). 60 His use of the word harmony, however, is somewhat misleading. The Fourier transform represents brief, non-periodic events as series of periodic frequencies that are endlessly repetitive, but they are not

59 The German terms are from a video of Kittler’s original lecture from 2003, uploaded to Vimeo in 2011 (Kittler 2011). 60 “Das analytische Verfahren […] unterstellt der Welt eine sinusoide Harmonie.” NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 143

harmonic in the spectral or acoustical sense of the word. The Fourier domain is not necessarily harmonic in the sense of a harmonious organisation of its frequency spectrum. Instead, the Fourier transform implements some kind of temporal, even rhythmical order in the form of a strict, infinite repetition of its elements. Via the Fourier transform, Kittler writes, “everything that is going on can be transferred into the frequency domain” (2006a: 69). Everything that happens in time can be taken out of time and the Fourier domain thereby represents a fundamentally a- temporal world in which everything always returns and nothing ever changes. To further assess the opposite of this perfectly noiseless world and the connection between temporality and noise that it suggests, I turn my back to the static, repetitive world of Fourier analysis to face the moving, changing world of physical signals, thereby going from the frequentia of the series to the flash of the event. On one extreme of the uncertainty relation between amplitude and temporality, squeezed through an impossibly narrow filter, without any reference to the temporal flow of things and cleansed of all possible noise, the ideal sine wave represents a timeless series. Widening the bandwidth of that ideal filter allows a larger frequency spectrum to seep through, consequently, following the uncertainty principle, shortening its response time and its delay: the temporal factor ! turns from the idealised ! = ∞ back into a finite, physical timeframe. Continuing this process, ever further widening the bandwidth, and ever further shortening the response time and delay, one ultimately arrives at the other extreme of the uncertainty relation, where another analytical idealisation represents the exact opposite of the ideal series. With a symbolic delay time of 0, it represents a timeframe reduced to an infinitesimally short moment. This idealisation at the temporal extreme of the uncertainty relation is either called Dirac impulse (or Dirac impulse function), after British physicist and mathematician Paul Dirac, or the delta function (also Dirac-delta), after the sign representing the function, the !. It represents the radical instantaneity of something that happens in less than a flash, the ultimate transient phenomenon: the ideal event. 144 NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE

c) The Dirac impulse: the ideal event Although Dirac formally defined the delta function and suggested its standard notation in his Principles of Quantum Mechanics in 1930, the concept itself already appears with mathematician and physicist Oliver Heaviside’s step function in the late nineteenth century, and, as mathematician Jesper Lützen notes in The Cambridge History of Science, ultimately follows from results that Fourier himself “was first confronted with” in the early nineteenth century (2002: 479-481).61 The Dirac-delta thereby originates in the same discursive context as the mathematical conceptualisation of the sine wave. It is a peculiar function, for which the value of ! is always 0, except when ! = 0. This means that, contrary to the infinite timeframe of the sine wave, the timeframe of a Dirac impulse is infinitesimally short, with nothing happening except when it is exactly zero. In terms of the uncertainty principle, such a function implies a filter with an instantaneous response and a delay of 0. Because the response time of a filter affects the precision of its operation (the longer the response time, the more precise the filtering; with the infinite delay necessary for a pure sine wave as its limiting case) a hypothetical filter with response time 0 will not filter anything; all frequencies pass through unfiltered. Consequently, as illustrated by figure 11, the frequency spectrum of a Dirac impulse is infinite. At one, infinitesimally short moment, when ! = 0, an infinite number of frequencies occur; in all other cases, when ! ≠ 0, nothing happens. The Dirac impulse is therefore the exact inverse of the sine wave. Whereas the latter contains one frequency that lasts infinitely, the former contains all frequencies in an infinitesimally short time. “Strictly,” Dirac himself wrote in 1927, “!(!) is not a proper function of x, but can be regarded as the limit of a certain sequence of

61 As Graham Farmelo writes in his biography of Dirac: “In an interview in 1963, [Dirac] remarked that it was his study of engineering that led him to his new function: […] But Dirac’s recollections may have been wrong. It may well be that he first read about the delta function from Heaviside, who introduced the function with his customary belligerence in one of the books Dirac read as an engineering student in Bristol” (2009: 113). NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 145

Figure 11 The Dirac Impulse or Delta Function from Omegatron, “Dirac distribution PDF,” adapted by Qef, Wikimedia, 1 Jul. 2008, commons.wikimedia.org/ wiki/File:Dirac_distribution_PDF.svg, accessed 23 Oct. 2015.

A graphical representation of the Dirac-delta: the upward arrow, the width of which should ideally be infinitesimally small, indicates its frequency spectrum is infinite. This figure illustrates why the Dirac impulse is also referred to as the ‘needle function.’ functions” (Dirac in Nahin 2006: 191). Mathematically, the delta function is therefore an ‘improper function,’ and, like the sine wave, it is a symbolic limit case that does not represent anything in the physical world. 62 Physically, it can only be approximated. The closer an actual signal approximates the infinitesimally short spike of a Dirac-delta, the more

62 Reportedly, it was his training as an engineer that led Dirac to ‘tolerate’ these approximations. Working in engineering made him realise the practical value of mathematical idealisations that the physical phenomenon only tends toward: “the pure mathematician who wants to set up all of his work with absolute accuracy,” he said, “is not likely to get very far in physics” (Dirac in Nahin 2006: 192). 146 NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE

frequencies it contains. Kittler’s choice of a bolt of lightning is therefore not coincidental: its discharge of energy contains a very large frequency spectrum that approximates a Dirac impulse. Musically, one can compare a bow gently striking a violin string and slowly setting its semi-periodic vibrations in motion with to a short smash on a , offsetting a sudden sonic blow. Whereas the first is a long gesture producing a primarily harmonic frequency spectrum tending toward the limiting case of ideal sine waves, the second is very short and produces a large, complex frequency spectrum tending toward the limiting case of the Dirac impulse Moles defines the infinitesimal timeframe of a Dirac impulse as the idealised representation of an infinitesimally short “pip” (1966: 80). In engineering, this impulse function is used to model the frequency response of a system. Because an ideal, infinitesimally short Dirac delta contains an infinite number of frequencies and its physical real-time approximation in the form of a very (but not infinitesimally) short “pip” contains a large (but not infinite) number of frequencies, feeding an impulse to a system’s input is a efficient way to measure which frequencies resonate within the system itself and determine which elements, materials or room acoustics affect, shape or dampen the output. In Kittler’s terms, this means the “thatness” of an approximate Dirac delta is used to measure the characteristic “whatness” of an (electro)-acoustic system and test how these characteristics might affect any signals fed to its input.63 Like a sine wave can be interpreted as the product of an infinitely accurate spectral filter that filters a completely unambiguous signal out of

63 A similar principle is central to pulse code modulation (PCM), the most common method for sound digitisation, discussed in Chapter One, Section 1.3a. In order to determine the amplitude values of each sample, an analogue-to-digital converter runs a series of sampling pulses each of which, Kadis writes, “approximates an impulse, an infinitely narrow pulse” (2012: 149). Before they are modulated, Rumsey and McCormick explain, “all these pulses have the same amplitude (height), but after modulation the amplitude of the pulses is modified according to the instantaneous amplitude of the audio signal at that point in time” (2009: 211). Hence, by modulating a regular series—in most cases 44.1000 per second—of approximate Dirac impulses with an analogue audio waveform, the impulses are scaled to match the amplitude of the waveform at that particular point in time, resulting in a time-limited sample of a band-limited signal, which is the basis for all digital audio. This also shows that the ideal timeframe of a digital sample would be the infinitesimally short timeframe of a Dirac impulse. NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 147

an infinite sea of frequencies, a Dirac impulse can be interpreted as the product of an ideal temporal filter, filtering one infinitesimally short instance out of the infinite flow of time. Whereas the infinite temporal factor ! of an ideal sine wave effectively comes to a standstill and turns a changing sound into an endlessly repetitive rhythmical order, a Dirac impulse does not ‘last’ any amount of time, because time reduced to its absolute zero value can no longer be understood in terms of duration. A Dirac impulse has no duration; it is a pure, point-like present. Without any connection to past or future, it focuses on an infinitesimally short temporal instance, an impossibly exact moment in time, a pure now. Understood through Kittler’s metaphor of the almost instantaneously appearing and disappearing flash of lightning, the upwards pointing needle of the Dirac impulse represents a completely singular event, radically contrasting and interrupting the repetitive predictability of periodic signals. Diametrically opposed to the complete stasis of the Fourier domain, the Dirac impulse represents pure transience.64 A technical device capable of perfect accuracy and unlimited resolution would be able to capture and analyse the spectrum of such a singular event at the very moment it occurs; it would instantaneously register, to use Kittler’s vocabulary, ‘that’ the impulse took place, exactly pinpointing its “thatness” in time, and ‘what’ it is composed of by defining its “whatness” in terms of frequency spectrum. Thus, infinitely perfect sound analysis theoretically requires a combination of the temporal precision of Dirac impulses and the spectral clarity of sine waves. However, the pure, infinitesimally small event can never be analysed completely, because it contains an infinite number of frequencies; and the pure series can never be fully registered, because it would require an infinite amount of time. Kittler therefore calls the combination of these two limiting cases “a crux, a cross, or, in more mathematical terms: a dilemma” (2006a: 71). This dilemma is Kittler’s philosophical rendition of the uncertainty principle: whatever we gain in one domain, we lose in the other. Perfect representation in either one or the other is possible within the confines of symbolic analysis and the abstraction of mathematical idealisation, which

64 Similarly, Moles calls the Dirac impulse a “perfect perturbation” (1966: 80). 148 NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE

separates “thatness” and “whatness” by trading the ideal of perfect representation for infinitesimal asymptotic approximation. Without this “mathematical outlook” of analytical models, Kittler writes, sonic phenomena “would not have appeared in all their clarity and sharpness” as sets of frequencies that can be measured, analysed, calculated and, eventually, technologically reproduced and synthesised (2012b: 53). Thus, he argues, our “acoustical Real,” which is the Real of technologically (re)produced sound and music, is “at least partly elicited by […] these mathematics” (53).65 Contrary to Kittler’s infatuation with the purity of the Fourier domain and its ability to unravel the noise of the Real, however, at the other side of the rupture between representation and represented, the physical reality of physical signals is as much ‘elicited’ by the effects of the uncertainty principles and the impossibility to choose between pure clarity in frequency and pure accuracy in time. Kittler’s take on the crux or dilemma of the uncertainty principle emphasises the Fourier domain and the way it enables analyses of natural phenomena that far outreach the capabilities our own senses, especially because they enabled the development of technical media that no longer produce asymptotic representations but real signals in all their physical complexity. Putting more emphasis on the other side of the dilemma (on the lightning instead of the thunder; on the event instead of the series), however, opens up a different perspective. “In that ideal world experienced only by the gods of mathematics,” Roads argues, “the delta function ! ! breaks the monotony with an instantaneous impulse that is born and dies within the most infinitesimal window beyond point zero” (2001: 40). Analytical emphasis on this infinitesimal window of the Dirac impulse therefore shifts the focus from the ideal purity and clarity of the Fourier domain toward all transience that escapes the spectral filter of Fourier analysis.

65 “Das akustisch Reale [ist] von dieser Mathematik aber doch mit hervorgelockt worden. […] Ohne diesen mathematischen Blick [wären] die Dinge nicht so in ihrer Klarheit und Schärfe aufgetreten.” NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 149

3.4 Time and transience: the domain of physical filters a) A clean cut and the noise of sound reproduction Governed by the laws of the uncertainty principle, physical signals strike a balance between precision in time and precision in amplitude. In “Spatio- Temporal Continuity, Quantum Theory and Music,” Norbert Wiener recounts a talk in Göttingen in 1925 at which he, two years prior to Werner Heisenberg’s conceptualisation of the uncertainty principle in quantum mechanics, uses the example of sound and music to explain the physical consequences of this balancing act. Physically, he writes, every waveform takes place within a certain timeframe. By definition, this timeframe cannot be infinitesimally short, because it minimally has to cover the duration of one complete cycle. In the case of complex harmonic waveforms like most musical sounds, the characteristics of the frequency spectrum (or in Kittler’s terms, the whatness) of the sound can generally be identified when its base frequency, which is the foundation of the harmonic structure of the sound, finishes one cycle.66 “If you take a note oscillating at a rate of sixteen times a second,” Wiener explains, “and continue it only for one twentieth of a second, what you will get is essentially a single push of air without any marked or even noticeable periodic character” (1976: 545). Because of this, Wiener illustrates, it is impossible to “play a jig on the lowest register of the organ”: the tempo of a properly played jig is faster than the time it takes the basic frequencies at the lowest register of an organ to finish one cycle

66 In “Mineral Sound or Missing Fundamental. Cultural History as Signal Analysis,” Bernhard Siegert describes how, for a long time, bell sounds puzzled acousticians because they are perceived as having a clearly distinguishable pitch (the so-called “strike-note”) although this fundamental frequency seems to be absent from its overall spectrum. This absence was confirmed when it became possible “to subject the acoustic signal of church bells to an objective pitch analysis” via digital technology. It thus became clear, Siegert notes, that contrary to Ohm’s Law, when the ear “perceives a sound that consists of various harmonics,” it does not perform a Fourier transform, “but only transmits the impression of a sound of a certain pitch and color.” In other words, because in these cases the “identity of acoustic color overrules the ear’s ability to decompose the tone into its overtones,” we tend to “ascribe to a tone the pitch of its fundamental even if that fundamental does not exist within the spectrum” (2013: 113-114). 150 NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE

(545). As a result, the oscillations of these lowest frequencies, which are essential for the identification of the spectrum, are cut short by the performance and the sound produced by the organ pipes “will not sound to the ear like a note,” but come across as a short, transient noise; a blow or an impulse (545).67 This is the uncertainty principle at work. At a certain limit, tending toward, but never reaching the infinitesimally short timeframe of a Dirac impulse, it becomes physically impossible to shorten a sound without losing its whatness, its identifiable frequency spectrum, in the process. Beyond this threshold, the analytical clarity of the Fourier domain gives way to the instantaneity of a Dirac impulse, clearly definable sine waves disappear into fuzzy, undefined spectra and all that remains is a transient blow, pip or, indeed, a noise. Besides demarcating the musical limits of church organ performances, these fuzzy, non-periodic transients are an indispensable aspect of every sound. Physical sounds do not exist as either one of the limiting cases of the uncertainty relation. Positioned midway between the two poles of infinite sine waves that displace time by endless repetition and infinitesimal Dirac impulses that displace any sense of duration by a point-like present, the domain of physical filters is ruled by the crux or the cross that is the uncertainty principle. Every signal has a beginning, a longer or shorter duration, and an end. Even an almost entirely periodic signal does not continue forever; at some point it will stop. The absolute purity of symbolic sine waves requires a timeframe that stretches infinitely into the past and the future. When the timeframe is not infinite, the filter is not ideal and the signal is not absolutely pure. Thus, because the series is always shot through with events, transients negate the infinite periodicity of the Fourier domain. The starts and stops of every signal cause, as Wiener writes, “an alteration of its frequency composition which may be small, but which is very real” (544-545). This means their beginning and end—or, in musical terms, their ‘attack’ and ‘decay’—add elements of non-periodic transience to even the most periodic sounds.

67 In “Lighting and Series – Event and Thunder,” Kittler writes: “Before a deep organ tone can turn into an event, many high trebles have already been recognized” (2006a: 71). Although he does not credit Wiener, it is most likely that he drew the example from his paper. NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 151

Exactly these “small, but very real” random alterations of the frequency composition give each sound its uniquely identifiable timbral quality. Whereas periodic frequencies are responsible for determining the pitch and overall harmonic composition of a sounds, their tone colour largely depends on what composer calls the “noise element in the very tone itself”—the non-periodic elements caused by their attack and decay (2004: 23). This is why, as musicologist Max Peter Baumann explains, our sound recognition is “partially based on noise,” (1995: 33).68 Exactly these traces of attacks and decays mark the difference between the idealisations in the domain of the ideal filter on one side of Siegert’s rupture, and the domain of physical filtering operations that define the operations of technical media, on the other side. The symbolic gesture that defines the domain of the ideal filter, of symbolic objects like the sine wave and the Dirac impulse, can be characterised as a clean cut that seamlessly removes a singular event from the natural flow of time and turns it into an infinite series—temporal infinity in the case of an ideal sine wave and spatial or spectral infinity in the case of an ideal impulse.69 The clean cut is the ultimate symbolic noise

68 “Das Erkennen des Tones erfolgt partiell mit dem Geräusch.” Regarding the importance of this noise element, Julia Kursell explains: “A sound’s individuality lies to a great extent precisely in those noises that mark its beginning. The longer a sound continues, the more difficult it is for the ear to recognize it as specific. It is equally difficult for the ear to identify the sound if the beginning is not heard” (2013: 193). See also Brech 1995, Höldrich 1995 and Hilberg 1992. 69 Derrida, in his discussion of Kant’s aesthetics in The Truth in Painting, writes about the “sans of the pure cut” [Le ‘sans’ de la coupure pure]: “So it is the without that counts for beauty; neither the finality nor the end, neither the lacking goal nor the lack of a goal but the edging in sans of the pure cut, the sans of the finality sans-end” (1987: 89, emphasis in original). Conceptually, this paradox of the finitude that is inherent to the application of a cut and the ideal infinity of a pure cut that leaves no traces of its cutting, bears similarities to my concept of the clean cut. In a different context, in Meeting The Universe Halfway, Karen Barad describes the problem of separating subject and object, observer and observed in quantum mechanics (which, given the importance of the uncertainty principle, is not that far removed from the present discussion): “So the question of what constitutes the object of measurement,” she writes, “is not fixed: as Bohr says, there is no inherently determinate Cartesian cut. […] What constitutes the object of observation and what constitutes the agencies of observation are determinable only on the condition that the measurement apparatus is specified. The apparatus enacts a cut delineating the object from the agencies of observation. Clearly, then, as we have noted, observations do not refer to properties of observation- independent objects (since they don’t preexist as such)” (2007: 114). In this example, as with my 152 NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE

filter: the Fourier transform cuts signals from the flow of time and represents their frequency spectrum as infinitely oscillating sine waves and the impossible temporal exactitude of the infinitesimally short Dirac impulse is what Siegert calls a “cut that freezes the movement” and represents the event as an infinite series of frequencies (2003: 251).70 Both idealisations are representative of the scientific rationale to create clear- cut objects of analysis without any transience and contingency. Without temporal or spectral randomness, the domain of the ideal filter does not contain any noise. Because technical media emerged on the basis of the rupture between such idealisations and the phenomena they model, a rupture that is itself defined by the symbolic cut of mathematical analysis, the concept of the cut appears throughout discourses on the operations of technical media; and sound media are no exception. Although the common expression to ‘cut’ a record is a literal reference to the way grooves used to be ‘cut-out’ of the recording material (wax, acetate, vinylite), this cutting extends to a more metaphorical dimension. When, for example, Oliver Read recommends the use of recording styli “that produce quiet, clean cuts,” the ‘cleanliness’ of this cut does not only refer to the technological procedure of cutting grooves, but also to the sound quality of the recording itself, which should be ‘cut,’ as it were, from the flow of time with as few acoustic traces of the technical cut as possible (1952: 46). This goes to show that the clean cut of the ideal filter is constitutive of the myth of perfect fidelity—both notions refer to symbolic operations that are not supposed to leave behind any physical trace.71 concept of the clean cut, the cut itself is (at least partly) constitutive for establishing a more or less unambiguous object of observation and analysis. 70 “Der Schnitt, der die Bewegung einfriert.” 71 Of course, in the context of sound recording, the metaphor of the ‘cut’ is not limited to cutting grooves in a surface. In musicologist and record producer Albin Zak’s book The Poetics of Rock. Cutting Tracks, Making Records, the ‘cutting’ in the title (a common expression in sound engineering) extends the trope to magnetic tape recording and can refer to the literal ‘cutting’ of tape and (again) the metaphorical ‘cutting’ of a piece of music or part of a piece of music from its sonic flow, crafting a more or less clearly delineated musical object, like a song. The cut thus separates one song or ‘track’ from the other. Magnetic tape also enables cutting and pasting as a aesthetic tool, both as a way of separating and ordering sonic material and as a means of disruption, as in William Burroughs’ famous ‘cut-ups’ in the 1960s “where he,” as N. Katherine NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 153

In contrast to the clean cut of an ideal filter, real-time filtering operations apply physical cuts that require a compromise between the spectral domain and the time domain, between sine waves and Dirac impulses. As soon as a signal starts, the non-periodic oscillations of transient events introduced by the attack continuously negate the heavenly periodicity of ideal sine waves. They add an amount of randomness, unpredictability and complexity that is characteristic of the pip of the Dirac impulse, which, as I cited Roads, “breaks the monotony with an instantaneous impulse that is born and dies within the most infinitesimal window beyond point zero” (2001: 40). In sharp contrast to the frequentia or infinite repetition of sine waves, such transience makes each moment sonically different from the next. Only an ideal filter would be able to produce or transmit pure, unaffected ideal signals; only Moles’ theoretical device consisting of an infinite number of ideal filters would produce completely unambiguous frequency spectra under the purely symbolic condition that ! = ∞. Due to the uncertainty principle, something happens between the input and output of every physical channel; transient elements cling to the transmitted sound, changing it by adding a certain amount of fuzziness. A system might approximate absolute linearity with an output that exactly matches the input, but it is never entirely linear. Input and output will differ ever so slightly. In terms of both information theory and thermodynamics, this difference can be expressed as an increase in entropy. “In the sense of thermodynamics,” Einstein puts it in a 1949 letter to Gödel, “any signal transmission is an irreversible process, a process which is connected with the growth of entropy” (Einstein in Gödel 1976: 459). Due to this irreversibility, the difference between input and output is not only a difference in frequency, in sonic character; most of all, it marks a difference in time. All “periodic events,” Ernst argues, “reveal their actual Being-in-the-

Hayles describes, “physically cuts up previously written narratives and arbitrarily splices them together,” a technique that was and still is heavily employed in both avant-garde and pop musical practices (1997: 88). These practices, in turn, take their cue from modernist cut-up and cut-and- paste practices in early twentieth century avant-garde movements. 154 NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE

world [ihrem realen In-der-Welt-Sein] by carrying with them the influence of their attack and decay, that is to say their moment of occurrence [Ereignismoment]” (2012a: 39). 72 The influence of transient events distinguishes a physical signal from the ideal signal it analytically approximates. In other words, the distance between an ideal system and its physical implementation is characterised by the introduction of random noise. Any signal transmission, whether natural or technical, whether through air, copper, glass fibre or any other medium, is subject to this logic of filtering, the effects of the cut that are dictated by the uncertainty principle. In the case of sound recording, all partial channels in the chain from sender to receiver, like the one pictured in Moles’ Information Theory and Esthetic Perception reproduced as figure 12, form a large technological assemblage to transmit sounds to many listeners; and each link in the chain adds transient noises to the signal. b) Transient traces of physical filters The uncertainty principles and the symbolic limit cases of sine wave and Dirac impulse show how the conceptual logic of noise reduction described on the basis of technological filtering operations in Chapters One and Two presupposes the clean cut that defines the domain of the ideal filter. Although the myth of perfect fidelity assumes the possibility of complete representation or reproduction, technical media operate on the basis of physical filtering operations that require a negotiation between accuracy in time and accuracy in frequency; a negotiation between the purity of sine waves and the exactitude of Dirac impulses. After this analysis of the idealisations at the symbolic side of the rupture between representation and represented in the domain of the ideal filter, Chapter Four will therefore leave the domain of the ideal filter and focus on the implementation of the physical cut of technical filters. The contrasts between the idealisations in the domain of the ideal filter and operations of in the domain of physical filters supports the claim that, as the analysis of noise reduction in Chapter Two revealed, technical filters not only filter signals from noise but add noise to these signals as well,

72 “Periodische Ereignisse wiederum zeichnen sich in ihrem realen In-der-Welt-Sein dadurch aus, daß sie die Wirkung ihrer Ein- und Ausschwingphase, also den Ereignismoment mitschleppen.” NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 155

Figure 12 An Analogue Recording Chain, Anno 1958 from Abraham Moles, Information Theory and Esthetic Perception, Translated by Joel E. Cohen, Urbana, University of Illinois Press, 1966: 10-11.

As Moles remarks: “Considering this dizzy chain of transformation, it seems remarkable that what remains at the end has some similarity to the original signal.”

156 NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE

thereby problematising the supposedly inherent connection between input and output. The turn to the operations of physical filters therefore constitutes a return to analysing real-time signals that extend in space and change over time; a move away from symbolic representations and toward technological reproductions. The analyses of dual-ended noise reduction and dithering already showed how the conceptual logic of noise reduction implies a presupposed notion of what constitutes noise and what constitutes signal. By conceptually concealing the first and revealing the second, this logic assumes completely unambiguous notions of noise and signal that would require the clean cut of a perfect symbolic noise filter. It thereby supports the assumption that perfect sound reproduction is indeed possible. My competing concept of the noise resonance of sound reproduction, however, suggests a more fundamental role for noise in the sound of the media age. In the domain of physical filters, the randomness and ambiguity of noise are the traces of physical filtering operations applied by all the technological links in the recording and reproduction chain. As such, noise signifies the impossibility of a clean cut. Any transmitted sound contains traces of everything it encountered (be it acoustically, electro-acoustically, electronically or digitally) from source to receiver. These irregular, unexpected, random elements, which are not representable in terms of clear frequency spectra but tend toward the impossibly specific instantaneity of the Dirac impulse, provide crucial sonic information. In contrast to the steady state of the Fourier domain, which is removed from the natural flow of time and therefore fundamentally a-historical, Ernst argues that transients “always already start to vibrate as temporal contamination” (2012b: 446).73 Hence, the sonic events that Wiener describes as “small, but very real alterations” of the frequency composition and von Neumann called the “small extra” added to a system’s output, change the sound signal with every step along the transmission chain. This, as Ernst calls it, “time critical” character of transients (of which no acoustic sound and no physical filter is entirely devoid) is what distinguishes the domain of the ideal filter on one side of the rupture from

73 “Der Anklang (Transienten) schwingt als temporale Kontamination immer schon mit.” NOISE RESONANCE | INFINITE TIME OR PERFECT TRANSIENCE 157

the domain of physical filters on its other side (2012b: 39-43). Adding a historical index to the transmitted signal, these transients thereby constitute a temporal trace. As I will argue in Chapter Four, because information about a sound’s identity and past is not revealed through the repetition of periodic frequencies but through the passing of random transients, exactly this sonic singularity of physical signals produces the multi-layered temporality of technological sound that resonates with human listener.

NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL” 159

Chapter 4: “In the Fourier domain are we immortal” on the pastness and presence of reproduced sound

4.1 The importance of sonic transience a) On internal and external noises “Just as the gods confined us to finite lives in the temporal domain,” Kittler writes ominously in “Lightning and Series – Event and Thunder,” “our bodies restrict us to a limited spectrum in the immeasurable range of frequencies” (2006a: 72). Restricted by these limitations, we designed technical media that are able to go beyond these limited capacities of our senses and capture and analyse frequencies that our ears cannot hear and our eyes cannot see. These media produce signals that are physically real and bear an empirically identifiable and statistically non-arbitrary resemblance to the input signals they are said to reproduce. The physical limitations posed by the uncertainty principle, however, assure that even these machines cannot instantaneously process infinitesimally detailed, real-time events, because the more they are the one, the less they are the other. In Chapter Three, my analysis of the uncertainty principle and its idealised extremes (infinite repeating sine waves on one side and infinitesimally Dirac impulses on the other) showed how this supposed insufficiency or lack of technologically reproduced sounds and their unmistakable presence in the physical real are related to the transience of random noise and distortion. Furthermore, given the fundamental a- temporality of the domain of the ideal filter and the (re)appearance of both noise and temporality in the domain of physical filters, the seemingly 160 NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL”

paradoxical combination of the impossibility of perfect technological reproduction and the unmistakable presence of technological signals in the physical real is indicative of the importance of this temporality for understanding the noise resonance of sound reproduction. Whether recorded years, decades or more than a century ago, acoustic signals stored on material hardware in one form or another have to be actualised with complicated playback procedures in order for them to return to the acoustic here and now. In purely physical terms, sound vibrations only exist in the present because, like everything else in the physical world, no sound can be ‘in’ the past.74 Signals that are acoustically, electro-magnetically or digitally stored on some kind of storage medium are merely grooves, magnetised particles or pits in a surface until they are transduced back into physical sound waves at the moment of their technical playback. This is why Wolfgang Ernst argues that sound recording effectively achieves “the cancellation of the distinction between past and present” (2012b: 22).75 For Ernst, the primary temporality of technical media is that of their technological execution in the present: “on the basis of the temporality of their processual nature,” he writes, “technical media negate the past” (2012b: 51).76 The “material implementation” of these processes, on the other hand, inscribes a “historical index” onto the recording (2012b: 51). 77 Although they are acoustically present and entirely real, Ernst argues, this index of sonic artefacts of reproduction technology reveals to the listener that the perceived sound waves are technological reproductions of an acoustic event that took place at an earlier moment in time. The historicity of gramophone scratches, magnetic tape hiss or quantisation noise, however, is not inherent to these sonic artefacts themselves. They are, Ernst argues, as acoustically present as the reproduced signal and the fact that they came to signify the historicity of

74 As Heidegger writes in The Concept of Time: “Everything that is encountered in the world is encountered by Dasein as residing in the now” (1992: 16E). 75 “Die Aufhebung der Unterscheidung von Vergangenheit und Gegenwart.” 76 “Technische Medien verneinen von der Zeitweise ihres prozessualen Wesens her die Vergangenheit.” 77 “Qua materialer Implementierung aber haftet ihnen ein historischer Index an.” NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL” 161

the recording is due to a discursive association between these sounds and the supposed pastness of the recording (2012b: 398). Owing to this combination of the perceived acoustic presence of technological sounds that include sonic artefacts produced by sound reproduction itself, and a discursive association between these artefacts and the historicity of the recording, Ernst argues, listeners experience what he calls “an aesthetic irritation of the human sense of time” (Eine aisthetische Irritation im menschlichen Zeithaushalt) (2012b: 45). Regarding these sounds, Ernst argues that

contrary to the kind of additional noises [Nebengeräuschen] generated during musical live performance (for instance the breathing of the singer and instrumentalists and the sound of gripping the violin), which are directly connected to the way of performing, this distortion [the scratches and noises from sound carrier and reproduction device] only holds an arbitrary connection to the sonic content (2012b: 57).78

This distinction between the additional, but supposedly relevant noises of musical live performance and the additional but arbitrary noises of technical media—which is reminiscent of Tainter’s separation of internal and external sounds discussed in Chapters One and Two—assumes the existence of some original sound event, a signal that existed prior to and separate from the arbitrary noises added by recording and reproduction media. As I argued in Chapter Two, however, this distinction between relevant internal noise and arbitrary external noise presupposes the

78 “Im Unterschied zu jener Art von Nebengeräuschen, die sich in der musikalischen live-Darbietung einstellen und unmittelbar verkoppelt sind mit der Weise der Aufführung (etwa das Atemholen der Sänger und Instrumentalisten, das Greifgeräusch an der Geige), steht diese Störung, steht dieses Rauschen [das Kratzen und Rauschen von Tonträger und Reproduktionsapparat] in einem arbiträren Verhältnis zum klanglichen Inhalt.” 162 NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL”

conceptual logic of noise reduction that assumes the possibility of an unambiguous separation between clear signals and disruptive noise. Contrary to Ernst’s implicit assumption that such a separation is possible, I claim that, when it comes to the artefacts of recording and reproduction media, there are no such things as external noises. "We can model music,” write Barry Blesser and Linda-Ruth Salter in their book on aural architecture, “as a sonic energy package that progressively passes through a series of passive acoustic objects, each of which then radiates and couples energy to other acoustic objects, and eventually to listeners" (2007: 150). By the time they reach the listener’s ears at the end of the recording chain, these “sound energy packages” travelled through and bounced off many such “passive acoustic objects”— walls, air, furniture—and travelled through many active technological components. Each element in this assemblage of microphones and walls, amplifiers and furniture, cables and air, compressors, effect modules, loud speakers, human beings and every other object it might encounter, adds specific physical characteristics to the signal; not only by raising the noise floor or adding arbitrary noises to the final output, but by its influence on the transient elements that shape the acoustic characteristics of the sound signal. Each link in the chain changes the frequency composition of the signal and thereby shapes the sound colour itself.79 Hence, the influence of every microphone, transistor, compressor, cable, amplifier or speaker is as relevant for the tone colour of the output sound as the specific resonance of a violin, the exact tuning of a piano or the breathing technique of a singer. What is presenced at the moment of playback is not the original sound signal combined with some arbitrary noises of recording and playback media, but a signal shaped by all the sonic traces of all transmissions and manipulations it encountered. Most importantly, these transmissions and manipulations are not only signified

79 In Earth Sounds, Douglas Kahn also writes about the accumulation of such characteristics in every sound: “Sounds can be heard as having acquired their character through the course of their propagation, acoustically and electromagnetically. In this way, a sound is as much of the intervening space as it is from the source. I use the term transperception to denote the perception of those characteristics […] along with the source” (2013: 62, emphasis in original). NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL” 163

by all the supposedly extraneous noises that are discursively associated with the historicity of the signal, but as much by the specific sonic quality of the output signal itself. The sonic transience of acoustic attack and decay affects the frequency composition and temporal development of all natural (not technologically mediated) sounds. It thereby fundamentally denies the strict periodicity of ideal sine waves. Creating an irreducible acoustic difference between consecutive temporal moments, this transience sonically emphasises the irreversible flow of time. In the case of technologically reproduced sounds, each small extra added to the signal along the chain of transmission changes the frequency composition of the sound as well, thereby marking the signal both sonically and temporally on its route from the moment of recording to its potentially many moments of physical reproduction. Ernst’s separation of external noise that discursively signifies pastness and internal sound that is always entirely present implies the existence of an original sound event that constitutes the actual ‘content’ of the recording and physically shines through in the here and now. This original would remain separate from the arbitrary external noises of recording and reproduction media, which are equally present but discursively associated with the pastness of the recording. Contrary to this distinction, I suggest that both the listener’s sense of pastness and his or her experience of presence are evoked by the sound of the recording itself, which is shaped by the transient elements that affected its sonic characteristics along its journey from sender to receiver. b) The transmission channel as parasitic third My analysis of technological noise reduction in Chapter Two already showed how the kind of distinction between arbitrary and relevant noises made by Ernst is based on idealised notions of signal and noise that were formalised by information theory and enabled the conceptualisation of the symbolic separation of the two in the first place. Serres argues that this clear-cut logic is symptomatic for the perspective of science, exemplified by Ulysses’ passage past the noisy Sirens and Leibniz’s Law of Continuity. Firmly tied to the mast and with the ears of his men plugged with wax, Ulysses always already seems to know where noise is located and how it 164 NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL”

can be reduced. This strategy enabled him to arrive at the other side of the channel and tell the story as if he made the journey completely unharmed and unchanged. When Helmholtz in the mid-nineteenth century tried to empirically approximate the purity of ideal sine waves, he inserted the narrow end of a spherical glass resonator in one of his ears and used a piece of warm wax to seal off that ear completely (Blamey 2008: 46). To shut his ears from all other acoustic disturbances, the other ear was stuffed with wax. This experiment is reminiscent of Serres’ reading of Ulysses’ ideal noise filter. Both Helmholtz and Ulysses plug ears with wax, the first his own and the second those of his men; and both intend to tame the noise to ensure a clear journey of the ship through the channel or the sound waves through the air and the resonator. During the experiment, Helmholtz is like Ulysses men, who do not hear the noise and only have one goal: to “sail […] past the obstacle of noise, neither transmitting nor receiving, and cancel […] out the Sirens” (Serres 2008: 122). Analogous to Ulysses’ heroic account of his dangerous journey, Helmholtz’s experiment assumes the possibility of a noiseless clean cut by an ideal filter.80 Contrary to this ideal filter, the cuts of physical filters presume a method that follows what Serres describes as the tactic of Orpheus, for whom the noise of the Sirens bleeds through and affects the singing with which he desperately tried to drown out the Siren voices in order to make it through the channel. Whereas “Ulysses succeeds, making it through the

80 In “Echoes. Ein Prolog,” published in 2005, Kittler draws a similar comparison between Helmholtz’s experiments and the journey of Ulysses and his men past the Sirens. “It is the same experimental set-up with Ulysses and Helmholtz,” he writes. “The Siren sings and people filter something out.” The complete passage is as follows: “…und deshalb wiederholt Hermann von Helmholtz am eigenen Leib die bittere Erfahrung von Odysseus’ Gefährten vor den Inseln. […] Wenn es dann bei 40 Grad gerade noch an den Fingern und an den Innenohren erfüllbar war, presste er sich in ein Ohr das dünne siegellackumhüllte Ende des Resonators, während er das andere Ohr völlig dicht machte. So präpariert, hat er dann seiner Sirene, die ihm aus Paris zugekommen war und erstmals ein kontinuierlich durchstimmbares Frequenzverhalten aufwies, also einer Sirene im modernen Wortsinn, gelauscht. Es ist dieselbe Versuchsanordnung bei Odysseus und Helmholtz. Die Sirene singt, und die Leute filtern etwas aus, mit dem Resultat, Dass, wenn die Sirene nicht die Resonanzton traf, Helmholtz so gut wie nichts hörte. Werin sie aber die Resonanz, die Eigenfrequenz traf, wollt‘s ihm schier das Ohr zerreißen” (2005: 22-23). NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL” 165

pass in silence, but cheats by suppressing all noise, danger or temptation,” Orpheus, who is “all ears, his lyre or cithara held before him,” is “confronting the problem and resolving it with music” (2008: 122). In his approach, the signal remains “open to the risk of collapsing into noise” and he cannot pretend to have arrived at the other side of the channel entirely unaffected (122). The distance between Serres’ account of Ulysses’ and Orpheus’ journey is the distance between the conceptual logic of noise reduction—assuming a clear separation between noise and signal—and the physical reality of the inseparability of transmitted signals from the random influence of transmission channels. Plugging his ears with wax and designing resonators to maximally reduce the transience of attack and decay, Helmholtz tried to approach the ideal sine wave, the perfect signal, as close as possible. Nonetheless, no matter how much wax one plugs in one’s ears, signals still have a beginning and an end. They cannot achieve the purity of timeless sine waves. The cuts applied by physical filters introduce transients that affect the frequency spectrum of the signal. Following this scenario and in accordance with Shannon’s model of communication, signal, noise and channel are parts of the same system. They are fundamentally inseparable. The notion of external noises assumed by Ernst’s analysis of the temporality of sound recording is therefore a product of the rationalist outset of science, which assumes clear separations, demarcated categories and self-contained models. Only through the conceptualisation of a symbolic filtering operation, the distinctions between, firstly, signal and noise and, secondly, between signals themselves are introduced. This makes filters an exemplary case of what Serres conceptualises as the ‘parasite,’ the element that is positioned ‘in the middle’ of a system: “what is between, what exists between. The middle term” (1982b: 65). The parasite is a “third” element in a system, in relation to which the positions of the other elements are defined (19). In the language of communication technology, the parasite could be the channel that facilitates and affects the transmission. On its journey through a channel, “the message,” Serres writes 166 NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL”

is changed. It arrives neither pure nor unvarying nor stable. […] The message is burdened and arrives thus burdened. To speak correctly, it is parasited. The parasite has placed itself in the most profitable positions, at the intersection of relations (43)

Serres thus calls the channel in between input and output the parasitic third, but he describes noise itself as a potential parasite as well.81 As I cited Moles in Chapter Three, Section 3.1b and as Serres writes as well, noise constitutes the necessary background for all communication but also denies the signal its absolute symbolic purity (167). Noise both creates and interrupts the relation, because a “parasitic intervention in the middle of a channel can help and block at the same time” (188). This is why the establishment of successful communication requires the reduction of the influence of parasitic background noise; for instance, as Moles describes in his explanation of the uncertainty principle, by installing a filter to reduce the bandwidth of the channel. “One parasite chases out the other, as one disorder chases out the other” (Serres 1982b: 88). The influence of one parasite—background noise—is dealt with by the introduction of another parasite: a filtering channel to narrow the noise spectrum. By shaping the relation between the different elements in a system, defining which constitutes the channel, the signal or the noise and creating new conditions for potential communication, every parasite introduces a new type of order, but also issues a cut or disruption. Hence, the recording chain is a parasitic chain (172). Each partial channel in between sender and receiver, in between those who record the sound and those who listen to it, constitutes a filter that issues a physical cut. Each is a parasite, the influence of which must be repressed to enable the signal to get through, but which is nonetheless indispensable for the transmission to take place at all. Each filter is a parasite that cuts the sound; and as each cut adds new transients, each cut changes the signal in

81 Indeed, Serres’ use of the French ‘parasite’ plays on its double meaning as both the word for a parasitic animal and the term for interference, static or unwanted noise (bruit parasite). NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL” 167

specific ways. Hence, as Serres writes, “the attribute of the parasite is its specificity” (230). The parasite changes the general and non-specific (for instance, a frequency spectrum composed of highly predictable, semi- periodic sine waves) into the particular and specific (for instance, a more irregular, less predictable, non-periodic signal). Sonically, each parasitic cut affects the noise of sound reproduction by shaping the transient elements and the unique spectral character of the signal. The parasitic filter thereby introduces crucial differences between signals and determines the degree of organisation and internal differentiation of the system as a whole. With every cut of every filter, the system is organised and reorganised, ordered and reordered. Through this reshuffling of the system by the introduction of sonic specificity on the basis of transience, a parasitic filter fundamentally negates idealisations that assume the clean cut of ideal filters. “The channel,” argues Serres, “carries the flow, but it cannot disappear as a channel, and it brakes (sic) the flow, more or less. But perfect, successful, optimum communication no longer includes any mediation. And the canal disappears into immediacy. There would be no spaces of transformation anywhere” (79). When a signal physically exists, it takes a certain amount of time that is neither infinitesimally short nor infinitely long; and because it takes a certain amount of time it must begin and end, introducing transient elements and a level of randomness that differentiates it from other signals. This randomness should not be described as the disruption of or addition to some originally pure signal. The purity only exists and will only exist in the realm of symbolic idealisations—the domain of the ideal filter—and the residue of attacks and decays, the transience caused by the cut of the technical filter, signifies the physical existence of the signal itself. Hence, Serres argues, it is the presence of the parasite that changes the ideal unchanging system into a physical system that irreversibly develops over time. In an ideal system, like Leibniz’s system described in Chapter Two, “every parasite is reduced to almost nothing in it, a grain of sand or of salt, a seventh” (44). Contrary to the infinite purity of ideal models, however, “we know of no system that functions perfectly, that is to say, without losses, flights, wear and tear, errors, accidents, opacity” and the parasitic transience caused by these losses, tears, errors and accidents 168 NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL”

imply the start and stop of signals and a sense of duration that is characteristic of everything that not only exists in the frequency domain but develops in the time domain as well (13). Because the transient elements introduced by the process of filtering create spectral and temporal differences between signals, these noise elements in each sound sonically indicate temporal development. The temporality of reproduced sound and the role of sonic transience must therefore be understood in relation to this sonic-temporal differentiation, which exists because of and in spite of the relation to the parasitic role of the filter. Firstly, I suggest, by emphasising the impossibility of infinite sine waves and the clean cut in time and space that would be necessary to capture a signal in its entirety, the transient elements that sound technologies add to the signal signify the inherent pastness of any reproduced sound. Secondly and complementary to this pastness, these same elements reinforce the physical presence of technologically (re)produced sound, as they unfold in the temporal here and now. This combination of sonic presence and pastness is reinforced by the transient, non-periodic noise elements of sound. In the interplay between these two levels of temporality, the noise resonance of sound recording emerges.

4.2 Pastness and finitude in reproduced sound a) We are immortal: infinite sine waves During the after talk following a lecture on music and mathematics in Cologne, given eight months before his death in October 2011 and published in 2012 under the title Und der Sinus Wird Weiterschwingen. Über Musik und Mathematik, Kittler responded to a remark by media scholar Peter Bexte on the anachronistic conceptual similarities between Fourier analysis and Leibniz’s account of the perception of the noise of the sea. Based on the observation that the mathematical integration of the Fourier transform approximately achieves what Leibniz’s considered only God to be capable of (that is, to extract the inextricable and unravel the noise of the sea into all its individual—sinusoid—components), Kittler said, it can NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL” 169

be argued that the “God” of Leibniz is “the big Fourier-analyst” (2012b: 48).82 For Kittler, this association between Fourier analysis and the Leibnizian concept of divine omniscience (which Siegert, cited in Chapter Two, Section 2.4c, notes as well) suggests a conceptual connection between the operations of mathematical analysis and their importance for the development of technical media and the human desire to attain full analytical grip on reality. Regarding the symbolic representation of the domain of the supra- or superhuman, Kittler repeatedly argued that the operations of technical media took the place of the divine.83 Extending upon this idea of the connection between the symbolic power invested in mathematical analysis and media technology on the one hand and a sense of divine capabilities on the other, Kittler continues his description of Fourier analysis by stating that

in the time domain are we mortal and in the frequency domain, in the Fourier domain are we immortal. [...] It is the essence of the sine and cosine that they do not have a beginning or an end and are therefore immortal (48).84

By translating the strictly mathematical idealisation of an infinite timeframe into a discourse that is as close to theology as it is to mathematics, Kittler’s conceptualisation of the Fourier domain offers a different perspective on the implications of the infinite timeframe of sine waves for understanding the temporality of technologically reproduced

82 “Gott ist der große Fourier-Analytiker.” 83 On this connection between technical media and the divine in Kittler’s thinking, see for instance the early text “The God of Ears,” which will be discussed in more detail in Chapter Five, the late text “Lightning and Series – Event and Thunder,” discussed in Chapter Three, and “Preparing the Arrival of the Gods,” cited in the Introduction and Epilogue (2006a; 2015a; 2015b). 84 “[…] im Zeitbereich sind wir sterblich und im Frequenzbereich, im Fourierbereich, sind wir unsterblich. […] Es ist im Wesen des Sinus und des Kosinus angelegt, dass sie keinen Anfang und kein Ende haben, also unsterblich sind.” 170 NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL”

sound. Conceptually, Kittler’s use of the word ‘immortal’ instead of the mathematically more correct ‘infinite,’ I argue, points in two directions that seem contradictory at first, but prove to be complementary on closer examination. On the one hand, Kittler’s rhetorical paring of the all too human dream of immortality and the symbolic idealisations necessary for the mathematical analysis of physical signals suggests a connection between the symbolical purity of sine waves and spiritual motifs of heavenly purity, infinite bliss and eternal life. On the other hand, by means of contrast, the use of the word immortality focuses attention on the physical impossibility of these mathematical idealisations and thereby on their fundamental absence in the physical world. Regarding the first perspective, Serres argues as well, only the symbolic suspension of “change, time, and history” applied by mathematical analysis can “chase out the parasite in us. Paradise then is there” (1982b: 183). In Kittler’s interpretation, this paradise is the Fourier domain. With his application of trigonometric functions for the analysis of wave phenomena, Fourier created a symbolic domain that can be interpreted as the discontinuous limit value of our everyday experience of space and time. Only “when we measure frequencies,” that is when we apply a Fourier transform, Kittler writes in “Lighting and Series – Event and Thunder,” “we are on the other side of death, in an immortality that has replaced the old gods” (2006a: 69). Measuring frequency spectra requires the symbolic suspension of temporality by introducing an infinite timeframe. Kittler’s reading of this infinite timeframe of the sine wave as ‘immortality’ reinterprets the mathematical ‘infinite’ as the more theological ‘eternal.’ The fundamental a-temporality of the Fourier domain is turned into an eternity—the traditional temporality of the gods. Ohm’s and Helmholtz’s application of Fourier’s method for the study of heat to the analysis of sound transformed the initially purely symbolic mathematical object of the sine wave into one of the cornerstones of the analytical representation of sound. With the subsequent invention of technological machines able to physically reproduce sound, “speech has become,” as one of the first reviewers of the phonograph famously noted, NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL” 171

“as it were, immortal” (cited in Kittler 2015b: 105). 85 Following this narrative on the power of sound recording, the technology ultimately strives for the infinite clarity of the Fourier domain and the perfect, infinite repetition of pure sine waves that would enable entirely seamless sound reproduction. “As long as a turntable is spinning or a CD is running,” Kittler imagines as well, “an old magic emerges despite the fading of years, hair and strength. Time stops, what more do hearts want?” (2006a: 68). Hence, as long as one assumes the possibility of a clean cut that achieves transcendental clarity by presupposing an infinite timeframe, the magic of media technology contains the promise of immortality. Time, however, does not stop. That which, under flashy theatre lights and accompanied by dramatic music, looks or sounds like magic is in the end always revealed as mere illusionism, simple trickery or, like Ulysses’ strategy, a clever ruse. Besides techno-religious dreams of immortality, Kittler’s reconceptualisation of the infinite sine wave as immortal therefore also points in another direction. As it is much more emphatic than infinity, the word immortality implicitly connects the a-temporal purity of the Fourier domain to its conceptual opposite. Because, by means of contrast, the exclusively symbolic status of the ideal sine wave also carries with it a sense of the temporal finitude of the physical world; the world in which, as Kittler said in Cologne, very much in the face of his own imminent demise, “we are mortal.” b) The ‘Endlichkeit’ of physical sounds After noting that “it is the essence of the sine and cosine that they do not have a beginning or an end,” Kittler continues by saying how “this property is quite annoying, as we do not only want to know frequencies, but events as well; for instance, when something has taken place” (2012b: 48-49).86 In spite of all its analytical precision, pure frequency analysis provides knowledge of the properties of the signal in one domain at the expense of

85 Kittler and many others attribute these famous words to Edison himself, but as Jonathan Sterne notes in The Audible Past, they actually appeared in an 1877 editorial comment on Edison’s invention in The Scientific American (2003: 298). 86 “Das ist übrigens ärgerlich, diese Eigenschaft, weil wir ja nicht nur Frequenzen wissen wollen, sondern auch Ereignisse, wann etwas passiert ist, zum Beispiel.” 172 NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL”

knowledge of its properties in the other. As physicist Dennis Gábor explains in his 1947 paper on “Acoustical Quanta and the Theory of Hearing,” the fact that “sound has a time pattern as well as a frequency pattern finds no expression either in the description of sound as a signal ! ! in function of time, or in its representation by Fourier components ! ! ” (1947: 591). Spectral analysis is only one half of the puzzle and a complete representation also requires information about when the signal occurred and how long it lasted. Gábor’s paper describes a possible solution to this problem. By chopping a signal into very small bits or so- called ‘windows’ and plotting the frequency information of these windows on a temporal axis, the representation can account for both time and frequency. Although such a windowed approach enables more accurate time-frequency analysis, however, due to the uncertainty principle, the accuracy remains limited in both domains. Gábor’s analytical circumnavigation of the uncertainty principle thereby acknowledges the necessary negotiation between representing complete spectra and exact durations that is required for any representation to represent the properties of any physical signal with a fair amount of precision. Following Gábor’s time-frequency analysis, Kittler’s conceptualisation—first linking the infinity of the Fourier domain to the eternity of Paradise and subsequently stating that this infinity is “quite annoying”—thus combines the promise of immortality in the frequency domain with the limitations of the time domain. In the domain in which phenomena have a limited duration, a beginning and an end, biological life is, indeed, mortal. The uncertainty principle is, as I quoted Kittler in Chapter Three, Section 3.3c, “a crux, a cross, or, in more mathematical terms: a dilemma” that limits the possibility to completely represent or reproduce signals as they occur in the flow of time (2006a: 71). The limit cases of phenomena in the Fourier- or time domain (ideal sine wave and Dirac impulse) thereby reveal how the operations of modern mathematical analysis always already produce their own negation. Despite their analytical clarity, these idealisations find their mathematical origin in Leibniz’s concept of infinity and thus inherently express their own physical impossibility. They can only exist in the symbolic domain, by virtue of their very impossibility in the NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL” 173

physical real. Whereas the model suggests that a physical signal endlessly tends toward the perfectly smooth limit of its analytical representation, exactly the signal’s failure to coincide with this limit is proof of its physical existence in space and time. Put like this, Kittler’s conceptualisation of the supposed immortality of sine waves cannot be properly understood without taking into account that the idea of immortality also evokes its opposite: mortality. In the first volume of Time and Narrative, Paul Ricœur discusses Augustine’s take on the problem of eternity in book eleven of the Confessions. For Augustine, Ricœur writes, divine eternity is a “limiting idea,” the absence of which in everyday human experience of time “is not simply a limit that is thought, but a lack that is felt at the heart of temporal experience” (1984: 26). By its absence in our everyday experience and in contrast to the finitude of all events, this “limiting idea then becomes the sorrow proper to the negative" (26). Our experience of temporality, Ricœur paraphrases Augustine, is therefore “permeated through and through with negativity” (26). For Augustine, the observation that eternity can be thought but never experienced is therefore fundamental to our experience of time as limited and finite. By analogy, Kittler’s conceptualisation of the immortality of the sine wave in the Fourier domain can be interpreted as a limiting idea that only becomes meaningful in contrast to the mortality that characterises everything in the time domain. Although Fourier analysis approximates the omniscience of Leibniz’s God when it comes to the analysis of full frequency spectra, it did not unlock the gates to divine eternity. Only when time is infinite and transience is fully reduced, the “paradise” that Serres associates with the complete removal of the parasite, “then is there.” By contrast, as long as transients are sonically present, the timeframe is not infinite, sine waves are not immortal and the eternity of this paradise remains forever out of reach. Physical signals decay and ultimately die out; and time flows irreversibly in one direction. If the symbolic sine wave is infinite and thereby signifies the domain of the Big Eternal Fourier Analytic in the sky, the transient noises of sound (re)production—a sonic marker for the physical cut that produces every physical signal—signifies finitude. 174 NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL”

Contrasting the heavily purity of immortal sine waves, transience marks the physical absence of immortality and divine eternity and signifies the finitude of all that exists in the time domain. The sonic traces of transient elements in the attack and of a sound always permeate throughout the signal and, as Ernst writes, “reveal the ‘Endlichkeit’ (the temporal limit) of any physical event” (2009: 12). This is why not only the arbitrary noises of the operations of recording and reproduction media, but also the relevant noises and distortions added to the signal by every link in the recording chain produce the perceived temporality of reproduced sound; and why Ernst’s strict separation between the two does not hold. Traces of the impossibility of the clean cut, the elements of non-periodic randomness that are so decisive for the specificity of sonic identities, negate the possibility of divine eternity, heavenly immortality and temporal infinity. They imply that a signal at some point began and at some point will end. As they thereby confirms the irreversible flow of physical time and the fundamental inaccessibility of eternity, these noise elements mark the Endlichkeit of all physical phenomena. In the final analysis, this includes our own finitude. In The End of Certainty: Time, Chaos and the New Laws of Nature, Prigogine recounts how, after a lecture on irreversible thermodynamics he gave at a conference in 1946, “the greatest expert in the field of thermodynamics” remarked that “irreversible processes are transient,” so why not “wait and study equilibrium as everyone else does?” (1997: 62). At the time, Prigogine recounts, taken aback by the criticism, he “did not have the presence of mind to answer” (62). Looking back on the incident, however, he describes what he should have answered: "but we are all transient. Is it not natural to be interested in our common human condition?" (62). Judged at the appropriate time scale, everything is transient. Kittler knew this as well. Positioned at the intersection of two idealised extremes—the immortal sine wave that resides in the infinite Fourier domain, and the transient Dirac impulse that disappears as soon as it appears—“our life,” he writes, “is made up, in the words of Sophocles, of ‘mere phantoms, shadows of nothing’” (Kittler 2006a: 71). Measured against geological timescales of tens or hundreds of millions of years, each of our lives is nothing but a flash, an impulse or a burst of random noise. NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL” 175

The transience of signals resonates with this transience of our own existence. If the infinity of ideal sine waves promises immortality, the ever- present noise element puts our feet back on the ground. Exactly the contrast between the idealised domain of symbolic representations and the physical domain in which “all things,” to quote the late George Harrison, “must pass,” was heightened by the advance of technical media that, as I quoted Siegert in Chapter Two, Section 2.4c, “are positioned in the rupture of classical analysis” (Harrison 1970; Siegert 2003: 389). Sound recording did not make speech, as it were, immortal, but it considerably prolonged its potential lifespan. By virtue of a digital reconstruction produced in 2008, the oldest known recording of a dates back to 9 April 1860 (“The Phonautograms of Édouard-Léon Scott de Martinville" 2008). Pre-dating Edison’s invention of the phonograph by seventeen years, it consists of French inventor Édouard- Léon Scott de Martinville singing the song ‘Au clair de la lune’ into his so- called ‘phonautograph,’ a device intended for the inscription but not for the reproduction of sound waves. It is somewhat of a happy coincidence that the first more or less intelligible recording of a human voice has now been revealed to be Scott’s twenty second rendition of ‘Au clair de la lune’ (‘by the light of the moon’) and no longer Edison’s alleged singing of ‘Mary Had A Little Lamb.’ Ever since the rediscovery of Scott’s phonautograph recording of 1860, the age of recorded sound and music starts with a reference to moonlight.87 In The Parasite, Serres argues that rationalist philosophies always assume a “world where there is only one system […] an ideal world of light and dark where there is only one exterior and one interior, only one shadow and one light” (1982b: 70). This ideally clear-cut, unambiguous world, he writes, can only be a place "without any atmosphere,” without “the air, the milieu (the medium)” to “make light diffuse”; which means

87 Edison singing ‘Mary Had A Little Lamb’ has for a long time been considered the first or at least one of the first recordings the inventor made on his newly invented phonograph in 1877. The actual wax role, however, was lost and the only sounding ‘version’ of this historic event is a more recent recording from 1930 on which Edison sings ‘Mary Had A Little Lamb’ once again: a sonic “ re- enactment,” as Ernst puts it, “to preserve the scene for cultural memory” (2016: 122). 176 NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL”

that, Serres concludes, “this imaginary world is on the moon” (1982b: 70). Singing about the "light of the moon” in order to be able “to write a word” (“pour écrire un mot”) in the dark of night, Scott’s recording implicitly invokes this play of light and dark that came to define technical media: contrasting the rationality of a clearly illuminated, rational world, a world beyond the fuzzy borders of this earth, with the continuous diffusion of such clarity in our physical, earthly life world. As I will go on to describe in Chapter Five, in the conclusion of “The God of Ears,” Kittler also describes the sonic Real produced by technical media with a reference to the moon. Playing on the final words of Pink Floyd’s 1973 blockbuster record The Dark Side of the Moon (“there is no dark side of the moon, really; matter of fact, it’s all dark”), he argues that “electronic media might yet invoke a still darker presence” (2015a: 16). This darkness at the heart of electronic media negates the “ideal world of light and dark” that is suggested by analyses of technical media that do not take the rupture between representations and represented into account. Setting the stage for this focus on the darkness of electronic media in the Chapter Five, I will continue the analysis of the pastness and presence of sound recording with a more nuanced reading of Kittler’s notion of the immortality of the sine wave in relation to Heidegger’s notion of “being- toward-death.” c) Being-toward-death and the transience of life Whether naturally produced or technologically reproduced, any sound, as Ernst argues, physically unfolds in the here and now. Nonetheless, as the residue of the sound’s trajectory through space and time from one medium to the next, the transient noise element acoustically emphasises that this radical presence, which is entertained throughout the entire discursive chain that runs from fundamental mathematical-acoustic theory to techno- commercial practices of entirely adequate signal reproduction, remains elusive. In between sender and receiver, this addition of transient elements as the parasitic effect of the many channels in the recording chain results in subtle, non-periodic changes to the signal itself. For the listener, despite its physical presence, these changes, firstly, create an aural awareness of the pastness of reproduced sound. Secondly, as sonic markers for the complexity of a world in continuous flux and the Endlichkeit of the signal, NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL” 177

the added transients emphasise the unidirectional flow of time as it unfolds in the here and now. Through this combination of the pastness and presence of reproduced sound, the sonic transience resonates with a sense of finitude that recalls Heidegger’s notion of being-toward-death in the first chapter of the second division of Being and Time. As soon as Dasein is thrown into the world, Heidegger argues, being-toward-death is part of its being-in-the-world and its ending is included in its Being as a potentiality to be fulfilled at some point in the future. It is important to note that, as Carol White explains in Time and Death. Heidegger's Analysis of Finitude, Heidegger’s concept of death “is not grounded in the biology of our bodies but rather in Dasein's relation to being. Death, as the 'possibility of the impossibility of any existence at all', is the 'conclusive' possibility which closes Dasein's being” (2005: 72). The concept of being-toward-death, in other words, does not so much deal with physical death, but with the existential death of Dasein as such, which means the end of “its existence as an openness toward being” (White 2005: 66). Hence, the questions of being-toward-death and the finitude of the being of Being are existential questions. Although, Heidegger argues, it is uncertain when Dasein will come to an end, it is an “indefinite certainty” that there will come a time when Dasein will no-longer-be (1962: 310). This indefinite certainty of death “belongs essentially to Dasein's thrownness” in the world and is inseparable from its Being (295). “As soon as man comes to life,” Heidegger cites Der Ackermann aus Böhmen (The Ploughman from Bohemia), “he is at once old enough to die” (289). Exactly because Dasein contains this “possible impossibility of its existence,” the absence of the end and the anticipation of death as the uncertain, but definite limit of Dasein enable an understanding of “one's ownmost and uttermost potentiality-for-Being” (310, 307, emphasis in original). Hence, the certainty of the finitude of Dasein is a confirmation of its being. Exactly by facing the possibility of non-existence and finitude, the being of Being can be open to the future. White illustrates this “notion of limitation” that lies at the heart of Dasein’s being-toward-death with the “metaphoric description of Dasein as a 'clearing' or 'lightening'” in Heidegger’s reading of Heraclitus (2005: 74). This clearing, White writes, “is the realm of possibilities which are revealed 178 NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL”

to Dasein by being” (74). In Heidegger’s analysis, death poses the utmost limit of this realm of possibilities because, White writes, at “the 'other side' of the unconcealedness of the open is the concealedness of death” (74). As such, White concludes, “as Heidegger's seminar partner Eugen Fink indicates, death refers us to the night which surrounds the open” (74). Included in the lightening that reveals the unconcealedness of possibilities is the indefinite certainty of death, which remains concealed. This play of light and dark, concealing and revealing and above all the notion of lightening [Lichtung] bears more than a superficial connection with Kittler’s use of the metaphor of lightning as the sudden appearance of the event, as a burst of information which can be pondered upon in hindsight by analysing its thunderous reverberations, thereby unconcealing the being of its Being. All the more because, at the outset of “Lightning and Series – Event and Thunder,” Kittler references Heidegger’s seminar on Heraclitus as well (2006a: 64).88 The relation between this notion of darkness and Dasein’s being-toward-death shall be further developed in Chapter Five. For the moment, however, Heidegger’s notion of being-toward-death shows how Kittler’s immortal sine wave should not only be understood as an Augustinian “sorrow proper to the negative” that is always reminiscent of its inherently mortal opposite.89 Instead, philosophy professor Charles E. Scott writes, “Heidegger emphasizes that always coming to the end defines life. It defines life’s incompleteness at any given moment, its potentiality, its power yet to be. The absence of perfectly completed life means possibility and futurity as

88 “Others may have encountered, as I did,” Kittler writes, “a sentence that Martin Heidegger threw into his 1966/67 seminar on Heraclitus: ‘I remember an afternoon during my journey in Aegina. Suddenly I saw a single bolt of lightning, after which no more followed. My thought was: Zeus’” (2006a: 64). 89 Regarding Heidegger’s notion of immortality, White writes: “In his 1966/67 Heraclitus seminar, Heidegger commented that, in the terminology of Being and Time, 'immortality is no category, but rather an existentiale, a way that the gods relate themselves to their being. Not a matter of just living forever, thus Heidegger's notion of immortality, too, denotes a way of existing not the extent of a lifespan, but one appropriate to the gods, not to finite Dasein. Only humans are mortal in Heidegger's sense, certainly not gods and not even animals: 'Only man dies.' Mortals, as who they are, are ‘present in the shelter of being' in a way that other domains of what-is are no” (2005: 86-87). NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL” 179

long as being is happening” (2010: 63). As long as death is still outstanding as a possibility, Dasein is still in existence. With this in mind, I argue that transient sound unfolding in the acoustic present not only resonates with the inherent transience of human existence. Following the notion of Dasein’s being-toward-death, the immortality in the domain of the ideal filter and the awareness of the “indefinite certainty” of death in the sublunary also highlight the not-yet of not having died yet. Hence, the transient presence and inherent finitude of the signal resonate with both the finitude and the being-alive of Being. This means that the stasis, clarity and lack of transience and change created by the infinite timeframe of Fourier analysis do not only refer to immortality or the complete absence of death. Quite the contrary, as Serres writes, the periodic “order from which parasitic dissonance is chased as much as possible” may as well be regarded as “an antechamber of death” itself (1982b: 138). When everything stays forever the same (like a Fourier series repeating infinitely from the distant past to the furthest future) it can be argued that nothing ever dies, but also that nothing is truly alive. “Heaven,” David Byrne sings in the eponymous song by the Talking Heads, “is a place where nothing ever happens” (Talking Heads 1979).90 Kittler calls the infinite sine wave immortal, because the mathematical idealisations at the basis of the Fourier domain represent a clarity and purity that is reserved for immortal deities. A domain that is bound to remain unchanged forever cannot support any mortal living. Hence, only not-living can achieve immortality and the infinite stasis of a domain called heaven or any other name; which is why, for John Durham Peters, mathematics, “studies how to die” (2008: 19). Focussing on symbolic clarity and purity, mathematics studies things that exist in endlessly clear stasis. In sound analysis, everything that tends toward the stasis of the Fourier domain is trapped in what Serres calls the

90 In light of the infinite repetition of ideal sine waves and Kittler’s old magic that emerges with the technological repetition of music, it is worth mentioning that Byrne also sings: “The band in Heaven plays my favorite song. They play it once again, they play it all night long” (Talking Heads 1979). He thereby confirms that heaven is nothing but endless repetition; a single song playing over and over again. 180 NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL”

“antechamber of death.” In contrast, exactly because they mark finitude, the introduction of randomness, transience and noise points to the absence of stasis and the presence of life. Contrary to the stasis of heavenly immortality., they constantly push time forward. They signify a world in which time flows, matter changes and nothing is without end. As the opposite of infinite repetition, stasis and the absence of change, the randomness of noise signifies life; but because it inherently implies finitude and the impossibility to truly store, let alone halt time, it always does so in the sense of being-toward-death, which means it signifies life because it includes the indefinite certainty of death. The possibility of the event of death, which, like a strike of lightning, remains beyond representation, is what separates the domain of technical media that operate in the domain of physical filters, from the domain of the ideal filter marked by a clean cut. The symbolic immortality of the sine wave and the Fourier domain are one side of the coin. By representing the reverberations of the event, they grant clarity and offer insight into the being of Being, but only by turning all transients into steady states, sacrificing the sheer singularity of the event in exchange for complete analytical repeatability, which is not coincidentally the prerequisite for proper scientific falsification. Transients that tend toward the Dirac impulse, on the other hand, mark the singularity of the event itself, which, as Jacques Derrida argues, “implies surprise, exposure, the unanticipatable” (2007: 441). The representations produced on the basis of an ideal filter suggest a completely static world in which nothing ever happens—a heavenly world. Contrary to this horizontal representation of infinite repetition, Derrida writes, “the event falls on me because I don’t see it coming. Like the arrivant, the event is something that vertically befalls me when I didn’t see it coming” (2007: 451-452). Such verticality is what defines the needle shaped Dirac impulse and marks the transience produced by physical filters. Whereas written representation, including the sound spectra represented by Fourier analysis, “always comes after the event” and thus “misses the singularity of the event,” technical media are capable “to intervene, interpret, select, filter, and, consequently, to make the event happen” by (re)producing physical signals themselves (2007: 446-447). NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL” 181

Hence, to be able to fully account for this instantaneity of the event and further assess the multi-temporality of sound media, the final stretch of this chapter puts Kittler’s reading of the “immortality” of the sine wave in dialogue with the instantaneity of the Dirac impulse and Derrida’s concept of the delay.

4.3 The presence of reproduced sound a) Always already too late: the delay of the physical cut In his essay Athens, Still Remains, originally published in 1996 as an accompaniment to a set of black-and-white photographs of Athens by Frenchman Jean-François Bonhomme, Derrida thematises and problematises the philosophical “tradition of being-for-death” in the form of an extensive meditation around two themes (2010: 59). Firstly, based on an analysis of the brief click of the photographic shutter that freezes an image in time and space, the essay discusses the multifocal temporality of the photographic image. Secondly, it revolves around a sentence that suddenly revealed itself to (or struck) the author on a bright, warm summer day near Athens:

Nous nous devons à la mort. We owe ourselves to death. It was this past July 3, right around noon, close to Athens. It was then that this sentence took me by surprise, in the light—“we owe ourselves to death"—and the desire immediately overcame me to engrave it in stone, without delay: a snapshot [un instantané], I said to myself, without any further delay (1).

In these opening sentences, in only a few strokes, Derrida sketches the main concern of Athens, Still Remains. The semi-exact date, time and place (July 3, right around noon, close to Athens) reference both the accuracy and inaccuracy attributed to an event. The surprise emphasises 182 NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL”

the instantaneity (like being struck by lightning) of that event; and the snapshot (engraved in stone without delay) expresses the desire and the impossibility to capture it exactly as it is at the very moment it happened, on a summer day near Athens. The seemingly tossed-off mention of the light (which, based on the date, time and place, must have been the bright, torrid sunlight of a Greek summer afternoon) encapsulates the irreproducibly singular experience of being here, now, at the present moment, but also the attempt, through the light-capturing medium of photography (literally “the writing of light,” as he remarks a few pages later) to capture that moment nonetheless (19). Ultimately, all these references to the singular experience of a moment in time and the fruitless attempt to capture that moment entirely spiral back to the event itself: the sudden arrival—without explanation or context—of the sentence “we owe ourselves to death.” In his efforts throughout the rest of the essay to (quite literally) make sense of this sentence, Derrida begins by focussing on its last word: death. Implicitly following Barthes’ analysis of photography and the concept of time as punctum in Camera Lucida, Derrida argues that, when it comes to the photographic capture of a moment in time we are always ‘too late.’ At the very instance the picture is taken, when the shutter opens and closes almost, but not entirely, simultaneously, the moment already passed. As soon as the photographic image has come into existence, the moment itself is gone. Invoking Heidegger’s being-toward-death, Derrida argues that photography, in the attempt to capture what is always already gone, confronts the viewer with a realisation that life is nothing but a “temporary reprieve” from the time to come when one is no-longer living (29). By producing images that inherently depict things that have already happened in the past and are no longer happening in the present, the ‘lateness’ of photography is a confrontation with the transience of life. Looking at a photograph is the closest we can get to looking at the past. Photography, Derrida writes, confronts us with the realisation that every moment one is alive “suspends the coming due,” but at the same time, it “signs [...] the verdict” (27). Because the things or people depicted on a photograph might have already vanished from the face of the earth, their photographic image—made when they were still physically present and NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL” 183

visible in the here and now, sometime, somewhere—implicitly tells us “this will have to die” (57). What the picture shows, will perish; and perhaps already has. Through its frozen capture of what is always already gone, the photograph is reminiscent of the catastrophe of death and invokes the not- yet of being not yet dead. This decisively melancholic side of photography underlines how we, as Derrida puts it, always exist in relationship to a fundamental delay that occurs between the moment something takes place and our ability to process it (17). A soon as we can grasp it, the present is already the past. Unable to represent the here and now at the very moment it happens, we can only reflect upon or represent what has already passed. The click of the photographic shutter, which is the moment of making the cut, covering the short time between pressing the release and the closing of the shutter, represents this delay. As I describe in Chapter Three, in “Lighting and Series—Event and Thunder,” Kittler explains that such a short moment of capture tends toward the impossible instantaneity of the Dirac impulse, the ideal now. Compared to the logic of the uncertainty principle, Derrida’s fundamental delay is analogous to the technological delay caused by the response time required for any technological filtering process to complete its operation. Because the uncertainty principle postulates that the response time of a physical filter can never be zero, only in the ideal instance of a Dirac impulse, when the timeframe is infinitesimally small and the number of frequencies infinite, the delay is completely absent. “In order to know what something is, we need time to recognise it, thus we always miss when it happened,” Kittler writes; “if, conversely, we want to know when something happens, there is no time left to say what it was” (2006a: 71). Once it is processed and defined, captured and reproduced, the event itself is gone. The now can only be stored as the no-longer-now and we are always already too late to grasp it, left with either an impression in our memory or an incomplete representation inscribed on some kind of hardware medium. b) Nowness and the destruction of the delay Whether one interprets the infinite timeframe of the ideal sine wave positively as a marker for immortality (as Kittler does) or negatively as a 184 NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL”

marker for mortality (as Peters’ qualification of mathematics, which “studies how to die,” suggests), the idealised spectra produced by Fourier analysis invite conceptualisations that relate sound in general and reproduced sound in particular to finitude and death in the sense of Heidegger’s being-toward-death. Derrida’s analysis of the click of the photographic shutter initially heads in a similar direction. It is clear from the start, however, that his goal in Athens, Still Remains is not to come up with a media specific rendition of Heidegger’s being-toward-death. Despite the initial focus on that tendency, he does not want to read “nous nous devons à la mort” “in the sense of the great post-Socratic and sacrificial tradition of being-for-death” or the inevitable coming due of death and the inescapable delay (2010: 59). Combining the analysis of the historic and discursive layers in Bonhomme’s photos of Athens with the question of the possible meanings of “we owe ourselves to death,” Derrida’s thematises a different interleaving of past and present—of has been, being, and will no longer be. He thereby sets out to conceptualise what he calls “the at-present of the now” by rethinking “instantaneity on the basis of the delay,” which is to say he aims to rethink the irrepresentable experience of the transient presence of the present in relation to its pastness (2010: 17, emphasis in original). Contrary to the infinite timeframe of ideal sine waves, the moment of photographic capture tends toward the opposite of infinity: the radical infinitesimal ‘now’ of a Dirac impulse. Thinking about this instantaneity shifts the focus from Kittler’s preoccupation with the a-temporality of sine waves (which is only able to grasp the whatness of signals: their being expressed in series of periodic frequencies) to an analysis of the presence of signals and their development over time: the continuous thatness of signals in the here and now. Taken on its own, a Dirac impulse encapsulates the promise of an impossibly pure, non-reflexive moment— an ideal event—at which, however unfathomably brief, the (necessarily infinite) entirety of the signal is present and the inherent transience of life does not-yet make itself felt. In the final sections of Athens, Still Remains, Derrida thematises this promise encapsulated in the instantaneity of a Dirac impulse by no longer focusing on the delay to which the inherent lateness of photography draws NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL” 185

attention. Instead, he argues, there is something else to be found in the very short instance between the opening and closing of the shutter. A focus on the moment of the click itself, the incredibly short moment of the photographic cut, opens a radically different perspective that is diametrically opposed to the conceptualisation of the static infinity of sine waves. By ignoring the delay and the impossibility of the clean cut and zooming in as close as possible on the moment of capture itself, we can try to, Derrida writes, “refuse [the] debt” that the Heideggerian echo of “we owe ourselves to death” imposes. If only ever so briefly, we can try to shift the focus from the delay contained in the lateness of the photographic image to the moment of the click itself, which, tending toward the infinitesimally short timeframe of the Dirac impulse, represents the almost unimaginable nowness of the present (63). To conceptually approach this nowness of the being “at-present of the now,” Derrida no longer highlights the last word of “nous nous devons à la mort” (death), but the first word of the sentence: we. In French, this ‘we’ sounds twice: nous nous—we (...) ourselves. Grammatically, the second nous is the object of the sentence (ourselves) (61). As that which is due or owed to death, this second nous gains its meaning in relation to Dasein’s being-toward-death and confirms Derrida’s notion of life as nothing but a reprieve of what is due. Through the self-reflexive gesture of owing oneself to something or someone, the second nous always exists in the delay. Always already defining itself in the face of the outstanding debt, it is always already owed to death. The first nous, however, is different. As the subject of the sentence (we) it does not refer to what is owed, but to the one who owes: we or, in the singular, I. Taken on its own, without the semantic relation to the second nous, this first ‘we’ is not due or owed to death. It simply is. From the semantic position of this first nous, it is possible to refuse the debt and ignore what is owed. This is not only a matter of semantics or Derridean wordplay. Stressing the agency of the first ‘we’ opens a route away from the “great post-Socratic and sacrificial tradition of being-for-death” toward thinking and conceptualising the “at-present of the now.” Taken separately from its grammatical context, the subjective first nous implies the possibility of an existence in the pure now, without reference to the 186 NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL”

disappearing past or the impending future. Conceptualising this instantaneity emphasises the singularity of each moment; for instance, the singularity of experiencing the light and warmth of the summer sun, somewhere close to Athens, on July 3 of a certain year in the late twentieth century, right around noon. When, at such a very specific, unrepeatable moment in time and space, the first nous ignores its debt by eschewing the reference to that second nous, it postpones, if only temporary, what is due to death. For an infinitesimally short moment, lasting no longer than the non-duration of an ideal Dirac impulse, the end is kept at bay and we are nothing but, as Derrida describes, “an innocent living being who forever knows nothing of death” (63). This, then, is the perfect transient experience. It comes and goes as instantaneously as a flash of lightning and is radically unconnected to any past and future, including its own. Like a Dirac-delta, it contains an infinite amount of information, too much to actively process or filter in any limited amount of time. It cannot be captured, nor reproduced; it only exists, can only exist, in the radical present. This “destruction of the delay,” assumed by imagining such a perfectly transient moment, Derrida writes, is “the very desire of philosophy” (51). It is the dream of immediately and completely capturing and processing all things in the world at the moment they unfold, giving anything its proper time and place and warding off the constant, unpredictable imminence of death. It is also the rationalist desire of positivist science as Serres understands it: the creation of models that perfectly fit the things they represent, suggesting “an ideal world of light and dark where there is only one exterior and one interior” (1982b: 70). This is the attraction of the limit cases at the extremes of the uncertainty principle, the infinity of the Fourier domain and the idealised moment at which time is constricted to the perfect ‘pip’: to ban all fuzziness, uncertainty, lack of clarity and randomness. However, mathematics and philosophy alike are aware that this ideal of a non-self-reflexive subject outside of the delay can only exist as symbolic idealisation. It is sought after like the fountain of youth, but can only be found on paper, in either our mathematical formulas or our philosophical stories.

NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL” 187

4.4 On passed and passing time a) We are infinite: the sonic present At the infinitesimally short moment of the ideal click with no connection to past or future, the debt can be refused and the not-yet of not yet being dead can be ignored. Freed from the burden of death, at such an idealised moment, “we are,” Derrida writes, “infinite […]. We are infinite, and so let's be infinite, eternally” (2010: 63). This “we are infinite” echoes Kittler’s “we are immortal” (or, remaining faithful to the irreversible arrow of time, Kittler’s ‘we are’ echoes Derrida’s), but this does not mean Derrida’s infinity is the same as Kittler’s immortality. It refers to the opposite extreme. According to the uncertainty principle, if time contracts to the most infinitesimally short instance, the corresponding frequency spectrum becomes infinite. Hence, when the purest, exact present of the ideal click of the camera is understood as the perfect Dirac impulse, its frequency spectrum must be infinite, encompassing everything from the slowest to the fastest (from the lowest to the highest) wave. Consequently, Derrida’s infinity is not temporal, but spectral. Like lightning and thunder, event and series, the Dirac impulse and the sine wave, Derrida’s “we are infinite” and Kittler’s “we are immortal” are idealisations at opposite extremes of the uncertainty principle. Reconceptualised in terms of the temporally uncorrelated transience of the Dirac impulse, Derrida’s analysis of the click of the camera’s shutter and his call to be infinite reveals an aspect of the temporality of technical media and the technological reproduction of sound signals that Kittler’s emphasis on the immortality of ideal sine waves does not and cannot reveal. Applied to sound technology, the combination of Derrida’s analysis of transient temporality and the photographic cut and Kittler’s conceptualisation of the infinite sine wave emphasises the relation between the simultaneous physical presence of reproduced sound and its experienced historicity. Besides the melancholic pull of pastness inherent to the necessary cut of physical filters and the impossibility of pure sine waves, the randomness of non-periodic transients also produces an experience of the continuously developing sonic present.

188 NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL”

Figure 13 Dirac Comb from Krishnavedala , “Dirac Comb.svg.” Wikimedia, 17 Jul. 2012, upload.wikimedia. org/wikipedia/commons/4/49/Dirac_comb.svg, accessed 23 Oct. 2015.

As I argued in Chapter Three, Section 3.3a, the sine wave can be conceptualised as the product of an ideal spectral filter of completely discrete frequencies and the Dirac impulse as the product of an ideal temporal filter that turns continuous time into discrete, infinitesimal windows. A hypothetical succession of such impulses, one after another, represents the idealised discretisation of continuous time, with each impulse representing a perfect infinitesimally short sample with no gaps in between. In the language of engineers, such a succession of impulses is called Dirac comb, impulse train or sampling function. As shown in figure 13, serialised in the form of a Dirac comb, the “purely discreet entity” of the Dirac delta turns into, as Siegert calls it, the “basic element of digital signal processing” (2015b). This is an idealised model of a sampling procedure: cutting infinitesimally small bits out of the time continuum, with each sample containing the complete frequency spectrum. As such, an ideal Dirac comb represents a continuous succession of, in Moles’ terminology, infinitesimally short pips, each of which is completely unrelated to its NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL” 189

neighbours. As every pip is a unique snapshot of a singular moment in time, a perfect Dirac comb is renewed at every infinitesimally short instance and irreversibly pushes time forward from one pip to the next. An ideal Dirac Comb is completely random and infinite in both time and frequency. Moles describes the “more or less continuous noise” it represents as an “erratic repetitions of elementary shocks […] without any correlation of either amplitude or interval of succession” (1966: 81). This description perfectly fits Kosko’s definition of white noise: a series of signals with a frequency range that runs, as he puts it, “the whole infinite length of the frequency spectrum” (2006: 66). No instance, however small, is correlated to the next. In the case of ideal white noise or the ideal Dirac comb, each sonic instance is therefore a random, irreducibly different, transient pulse.91 b) Sonic presence: sound unfolding in the here and now Pure white noise is as symbolic as a pure sine wave. It requires infinite time and an infinite number of completely uncorrelated frequencies, which implies infinitesimally short Dirac impulses with infinite energy (Kosko 2006: 66-67). On the opposite side of the uncertainty principle, the supposed regularity of completely periodic signals is evenly idealised; their regularity tends toward, but never achieves the completely predictable repeatability of ideal sine waves. Hence, as no signal is completely periodic or completely transient, no physical sound coincides with either one of these extremes. Physical sounds exist in between periodicity and aperiodicity, sine wave and Dirac impulse, infinite timeframes and infinite frequency spectra. Immortal sine waves, Pythagorean celestial harmony and Kittler’s old magic that emerges “as long as a turntable is spinning or a CD is running” all tend toward the purity of perfectly periodic signals, marked by absolute repeatability, a maximum degree of autocorrelation. The noise of sound (re)production that affects technological sounds from their onset

91 It is not for nothing that Kittler in “Signal-To-Noise Ratio” writes that “Lacan’s concept of the Real refers to nothing but white noise”—white noise is nothing but pure, fundamentally irrepresentable randomness (Kittler 2013c: 175). 190 NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL”

onward, however, tends toward the infinite spectral complexity of a Dirac Comb. Contrary to Ernst’s interpretation of the external noise of reproduction media, this influence of the channels in between the moment of the first attack and the moment of the final decay is part of the signal itself and emphasises its historicity and Endlichkeit. Because the physical conditions of each of these channels, including the amplifiers, speakers or headphones that are used for playback, might be very similar, but are never entirely the same, every recording and playback of sound is irreversibly singular. Thus, the accumulative transience of reproduced sound not only emphasises its historicity and finitude, but also produces its unmistakable sonic presence. Following philosopher Hans Ulrich Gumbrecht’s argument in the Production of Presence, such sonic presence should first and foremost be understood in spatial terms. It describes the physical presence of signals as they are, to use Gumbrecht’s description, “in reach of and tangible for our bodies” (2003: 17). In the case of sound, vibrating airwaves physically touch our ears and, depending on the frequency range, other parts of our bodies. In temporal terms, this physical presence is profoundly transient; always fleeting and “in constant movement” (17). It comes and goes like a strike of lightning or the overwhelming light and warmth of the Greek summer sun. Mathematically represented by the pure transience of Dirac impulses, the presence of the present knows nothing of past and future. For Gumbrecht, this production of presence is the residue of the materiality of communication media. It is the resonance of the materiality of each element in the transmission chain in the ears of the listeners. On the one hand, as the result of the delay required to make a physical cut in space and time, these transients elements are reminiscent of the impossibility to capture this presence exactly as it happens, time and time again, at some specific moment in some specific place. On the other hand, these transients are physically reproduced in and inevitably affected by the specific conditions of the present moment. They thereby confirm the radically contemporaneous passing of all reproduced sound through the now. c) The multi-layered temporality of technological sound Taking into account this multi-layered temporality of technological sound, the noise resonance of reproduced sound turns out to be a double-edged NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL” 191

sword. As the result of a long series of intermittent filters, the sonic transience of noise, distortion and randomness is a marker for passed time and passing time. Passed time in the sense that it affirms the Endlichkeit of the signal as indexical trace that resonates with both the temporal irreversibility and finitude of all transient phenomena. In this respect, reproduced sound and music emphasise how we are always running out of time. It marks passing time, on the other hand, in the sense that these traces caused by the materiality of communication media confirm the constant flow of time through the present and produce the experience of a continuously renewed sonic presence. In this respect, reproduced sound and music emphasise how we are always inside time. The uncertainty principle is the physical basis for a fundamental delay in capturing and retaining the full temporal and spatial presence of any event. Derrida describes a similar delay as constituent for our temporal experience and the impossibility of writing the event. By making a conceptual link between the physical uncertainty relation and Derrida’s concept of the delay, I intend to emphasise how the delay in the response time of technical filters is something more—and something more fundamental—than an objective technical limitation of media, which may or may not be ‘overcome’ at some point. Instead, the delay postulates, as Peters puts it, “an ontological point about the nature of things and an ethical point about the uniqueness of every act” (2008: 11). Due to the absolute physical limit imposed by the uncertainty principle, “all empirical representation,” Peters writes, "both depends on and crashes into the wall of finitude” (19). If we were able to wallow in the eternal presence of an infinitesimally short instance, there would never be any time to pinpoint what was what, which means to analyse, process and define the event. Defining the being of Being requires a cut that turns events into series, transience into steady states and vibrations into standing waves. As the sonic residue of this cut, noise and randomness mark passed time. From this perspective, noise is, as Serres argues, “a straightener” (1982b: 185). By changing the signal, it directs it in one direction, turning what an ideal model represents as something reversible into something that, as Ricœur cites Augustine, “has a before and an after [and] is subject to ‘change’ and to ‘variation’” (1984: 23). Surely, ever since the invention of 192 NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL”

magnetic tape, the temporal order of sound signals can easily be turned around. In the twenty-first century, any child can digitally reverse the flow of sound; but because it turns the attack into a decay and vice versa, this reversal leaves an audible mark. Hence, notwithstanding what Peters calls the “symbolic reversibility provided by media,” without a single exception “everyone still dies” (2015: 311). Even if you hear the reversal of the temporal flow of a signal, time itself is still experienced as irreversibly flowing in one direction. Parasites, writes Serres, are “filtering a meaning, creating a meaning” (1982b: 185). When a recorded sound arrives at the ears of listeners, it includes all the sonic traces and alterations it accumulated; all the parasites it encountered and incorporated. On the one hand, encapsulated in the specific frequency composition of the signal, we hear its journey over space and time. It reveals that this signal, which is physically present in the here and now, travelled over time and through many different places. It was captured and by that very act, it was cut from its moment of origin and changed from its original form. On the other hand, the same random singularities added to the sound by this physical cut also clearly resonate with the listener in the present. Semi-periodic frequencies enable the identification of the whatness of a sound, including some of the sonic traces of its own historicity engrained in the frequency alterations caused by every medium it encountered. The transient events that trigger these ‘small alterations,’ however, cannot be described in terms of whatness. They are constitutive of its thatness. With their brief appearance and disappearance we experience the passing of time and the continuous production of the present. Despite their formative role, we are always too late to process and capture these events at the moment they happens, but although this moment of production is always already gone, their being “at-present” in the moment of playback is constantly emphasised. The specific singularity and temporal unfolding of sound have always been fundamental to the appeal of music, but the invention of recording pushed the contingencies of the entropic side of sound or, in other words, the noise element in each sound, to the centre of creative production. Sound recording is based on a series of physical filters that cut NOISE RESONANCE | “IN THE FOURIER DOMAIN ARE WE IMMORTAL” 193

the flow of events and thereby, on the one hand, change the frequency composition of signals and evoke a sense of pastness and, on the other hand, enable the repetition of transients that (re-)produce a sense of sonic presence. Evoking a sense of presence whilst simultaneously reminding us of the delay, the noise resonance of sound reproduction thereby encapsulates our biggest dream and our biggest anxiety: to annul the delay and halt time, and the knowledge that this is never entirely possible.

NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’ 195

Chapter 5: The sound of an ‘other music’ | sonic transience and the darker presence of the media age

5.1 A logic of filtering a) Turning toward the parasitic middle The physical effects that are caused by the uncertainty principle emphasise the impossibility of complete symbolic representation and perfect technological reproduction. Putting analytical emphasis on this impossibility, however, implies one still takes the perfect clarity and transparency of the domain of ideal filters as a primary point of reference, if only by means of contrast. No matter how much one embraces the imperfections of the physical world, emphasises unpredictability and randomness or glorifies the grittiness of noise, any analysis that implicitly or explicitly relegates noise and distortion to the domain of the abject— celebrating the transgression of discursive borders, subversion of musical or cultural norms or disruption of social order—inadvertently confirms the idealist position it intends to reject. In order to get beyond this antithetical relation between the presence of noise and the domain of ideal filters, I propose to no longer take the symbolic output of ideal filtering operations as the primary point of reference for understanding the output signal and instead focus on everything that emerges on the crossroads between the two extremes of the uncertainty principle. Doing so means to leave the idealist figures of sine wave and Dirac impulse for what they are and embrace the complexity of things that exist in the middle, things that extend in space and change over time. It means to strike a balance between whatness and thatness, 196 NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’

between the finitude of signals and the eventful strike of their phenomenological presence, thereby acknowledging the continuous push and pull of pastness and presence that characterises technologically (re)produced sounds. As shown in Chapter One, over the span of their hundred and forty year history, sound media developed from relatively simple machines that could transmit, store and replay sound signals with limited spectral resolution and limited dynamic range to highly complex technologies that are able to record, transmit, analyse, manipulate, produce and reproduce sounds with a definition that equals the spectral and temporal complexity of any non-technological sound. This development, I argued, should not be interpreted according to the myth of perfect fidelity or the idea of a continuously growing resemblance between reproduction and supposed original. Instead, as I explained following Sterne and others, the concept of fidelity is defined by what art historian Ruth Benschop calls the specific “quality of tone” of a recording (2007: 489). It refers to a subjective, context determined assessment of the sound qualities of reproductions in comparison to those of other reproductions. These qualities are not measured against the objective characteristics of the input signal or the intentions of the sender but evaluated according to the requirements and expectations of listeners regarding certain types of music or technology and certain social or aesthetic contexts. Nonetheless, technologies designed to remove what Bose, cited by Mack Hagood in Chapter Two, calls “things you don’t want” discursively follow a conceptual logic of perfect filtering that goes back to the application of Fourier analysis to sound waves in the mid-nineteenth century (2011: 575). Like the clearly delineated spectra of infinitely oscillating sine waves, these technologies adhere to the ideal of an entirely rational world in which every part of every sound has its proper and unchanging place. By assuming a clean separation of signal and noise, noise reduction technologies presuppose a conceptual filtering operation. Such a conceptual filter is similar to the one on which Western classical music theory has always maintained its separation of music and noise, as well as the one on which Helmholtz based his separation of musical sound NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’ 197

composed of periodic vibrations and unmusical noise composed of non- periodic vibrations. In the 1930s and 1940s, this conceptual separation of noise and signal was further formalised in information theory. On the one hand, information theory perpetuates the idealist discourse of clarity and purity characteristic of scientific modernity—presupposing ideal filters, perfect transmission and unambiguous signal-to-noise ratios. It thereby firmly cemented the idealist horizon of perfect reproduction in the DNA of twentieth century communication media. On the other hand, however, Shannon’s mathematical model of communication shows that every transmission channel adds a minimum amount of noise to the transmitted signal in ways no filter can entirely predict or prevent. As I cited Abraham Moles’ analysis of the role of noise in communication systems in Chapter Three, Section 3.1b, information theory provided a mathematical representation of the physical uncertainty principles that limit every non-symbolic attempt to produce ideal filters and noiseless signals. As Moles explains, the introduction of a secondary filter in the transmission chain can reduce these artefacts added to the signal by primary filtering operations. Because such a second parasitic filter will affect the signal as well, however, this amounts to a situation in which “one parasite chases out the other” (Serres 1982: 88). Noise reduction systems are such secondary parasitic filters. They are installed in the recording chain to reduce the artefacts that the primary filters of sound reproduction add to the signal; but, in the process, they inadvertently affect the signal itself. To escape this vicious cycle of primary and secondary filters on their futile quest toward complete transparency, I propose a break with the dominant conceptual logic of noise reduction. Assuming the primacy of the input signal, this logic relies on the symbolical suppression of the channel in order to connect a supposedly original input signal with the reproduced output signal. It thereby discursively construes the relationship between the two in terms of original and copy, represented and representation, complete and incomplete, and upholds the status of the input as normative point of reference. By conceptually replacing the ideal symbolic filter with a physical parasitic filter, however, I suggest to foreground the filtering 198 NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’

operation itself as the primary point of reference for understanding how technological sound (re)production shapes the sonic identity of its output. b) Beyond input and output: the filtering channel As the analysis of dithering in Chapter One and Two showed, the way the conceptual logic of noise reduction conceals the imperfections of sound reproduction to uphold the suggestion of an inherent relation between the domain of the ideal filter and the domain of physical filters is more clearly illustrated by the operations of digital technologies. Because they are based on symbolic filters that assume the complete reduction of all material noise the separation of signal and noise assumed by analogue noise reduction technologies became, as I cited Siegert, “nothing less than systemic” in digital systems (2015a: 30). On the basis of streams of discrete signs processed on micro- temporal levels, digital media not only represent, but actually (re)produce signals within and beyond the range of human hearing, including intricate manipulations of sonic time and frequency as well as the creation of completely new sounds. For Kittler, with this detailed, real-time sound processing, digital technology finally provided a viable “language for sound, that is for unforeseeable, unthinkable, unimaginable acoustic events” (2013a: 40). 92 However, whereas Fourier analysis mathematically encapsulates the potentially infinite real number values of physical amplitude levels with a Fourier integral, computers still need to transcribe these amplitude values in finite natural numbers (1993b: 199). As explained in Chapter One, Section 1.3c, no matter how many bits there are available to represent the amplitude values of the sampled signals, the representation of these potentially infinitely large number values with a finite set of symbols inevitably leads to quantisation errors. An analysis that assumes the conceptual primacy of sine-like clarity will interpret the distortion caused by these errors as an acoustic reminder of the insufficiency of digital systems to completely reproduce signals in space and time. Similarly, such an analysis will interpret the addition of

92 “Eine Sprache für Sound, also für unvorhersehbare, unausdenkbare, unvorstellbare akustische Ereignisse.” NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’ 199

dither noise to randomise quantisation error and prevent harmonic distortion as a psycho-acoustic stopgap to disguise the fundamental insufficiency of digital sound processing. According to this logic, by trading statistically correlated harmonic distortion for what engineers consider to be a more natural and less obtrusive noise floor, dither fools the listener into believing the sound coming out of the speakers is a faithful reproduction of the original signal. The identification of the digital procedure with an idealised model of representation thereby assumes the original input signal is the primary point of reference and the perfect copy is the ultimate goal for all technological sound reproduction. Exactly the symbolic rigor of the digital procedure highlights the fallacy of this idea. What digital sound technologies reveal more sharply than their analogue counterparts is that the clarity of ideal representations, compared to which any technological reproduction is always already insufficient, is not a property of the input signal at all. Instead, the smooth limit cases that digital representations and reproductions tend toward are brought forth by the mathematical idealisations underpinning the digital procedure itself. Neither original input nor reproduced output showcase the smoothness suggested by these ideal limit cases because this clarity belongs to the model and to the model only. By assessing the fundamental principles of the mathematical and physical representation and reproduction of sound, Chapter Three and Four showed that, in contradiction to the clarity of ideal filters, signals produced in the domain of physical filters require an unavoidable negotiation between time and frequency; a physical cut that introduces transient randomness and distortion in the (re)produced signal. Extending upon this analysis of the cut of physical filters, this last chapter sets out to argue that the noise resonance of sound reproduction is produced on the basis of a logic of filtering shaping all sounds produced by technological media. Such a logic of filtering does not adhere to the idea of a clean input and clear output compared to which the influence of the filter is always the interfering by-product of signal transmission. Instead, it focuses on the parasitic middle and emphasises how neither sender nor receiver, but the operations of the parasitic third, of the channel itself, are primary. 200 NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’

c) The irrepresentable moment of filtering "It is precisely under mediatic conditions,” Kittler states in an interview with Artforum in 1992, “that what cannot be processed, what is impossible, is brought into ever sharper focus” (Kittler 1992b: 68). Exactly this sharper focus on that what cannot be processed is what a logic of filtering reveals. Elemental processes of filtering, Siegert explains in Cultural Techniques, form the basis of all signification: from symbolic signification based on ideal filters to the operations of technical media based on physical filters. As I cited Siegert’s argument on the conceptual relation between “symbol, as defined by logicians, and signal, as defined by information theorists” in Chapter Three, Section 3.3a, the symbols created by ideal filters assume a complete reduction of all noise of their material production as physical signals (2003: 20, emphasis in original). The filtering operations of technical media, in contrast, rely on the basic rules of signal processing. They do not apply idealisations to create symbolic signs, but use technical filters to produce physical signals. On the basis of the almost complete separation of symbolic signs and the noise of their material production as signal—or what Siegert calls the “noise of all graphic form”—pre-technical media hide the symbolic filtering operations that produce supposedly ideal signs (2003: 20). Disappearing behind the seeming clarity of its output, the filter itself becomes almost completely transparent. Something similar can be said of the ideal spectral filter of Fourier analysis. The introduction of the Fourier integral for analysing non-periodic signals creates an absolutely clear, discontinuous limit case representing all frequencies within a given time- interval as a set of infinitely oscillating sine waves. Although this mathematical integration brings the spectral complexity of sound waves into greater focus, Fourier analysis also removes all non-periodicity and temporal transience. Hence, as soon as the ideal filter disappears behind its clear output, the smooth limit case of the symbolic representation conceals the fundamental fuzziness of physical signals. The implementation of these symbolic idealisations into the material hardware of technical media, however, brings the moment of filtering itself into greater focus. Helmholtz’s experimental setup to physically approximate ideal sine waves did not reveal a physical object NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’ 201

that had always existed. As I cited Kursell in Chapter Three, Section 3.2c, acoustically approximating a mathematical object on the basis of the symbolic logic of Fourier analysis, Helmholtz’s experiments created a new kind of sonic object. As much as it provided insight into the physical nature of sound, this reconceptualisation of the sine wave as a sonic object also shaped the framework of sonic transparency and acoustic purity that became dominant in the discourse on sound reproduction and representation. With his resonators as one of the earliest examples of technologies that explicitly follow this discourse, Helmholtz turned the ideal filter of Fourier analysis, producing ideal symbols, into a technical filter producing physical signals. Hence, as this conceptualisation of the sine wave as a sonic object implicitly suggests, the difference between the ideal filters of pre-technical media and mathematical analysis and the physical filters of technical media is their relation to noise. Whereas the first assume the possibility of perfect representation by excluding all noise of their material production, the operations of the second always add the noise produced by their physical cuts. The difference between ideal signals (such as sine waves and Dirac impulses) and physical signals is manifest in the form of the random traces caused by the uncertainty relation between time and frequency. As they cannot be represented by anything but themselves, this spatial and temporal transience caused by the uncertainty relation constitutes what Kittler calls “that what cannot be processed.” What the mediatic conditions of sound technology therefore bring “into ever sharper focus” is the fundamental fuzziness caused by the uncertainty relation between space and time. Complete similarity between input and output can only be expected of the symbolic operations of ideal filters. Beyond their domain, the non- periodic transience that appears at the moment of filtering itself cannot be separated from the signal that it produces. By adding the artefacts of their filtering operations to the output signal, technical filters continuously produce sonic differences. In the domain of physical filters, these sonic differences are discursively construed as the difference between input and output, original and copy. However, by accepting that input and output are idealisations that straighten the complexity of physical signals by assuming 202 NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’

the representational clarity of discontinuous limit cases, I do away with the idea that this noise can only be classified as interference. Instead, it acknowledges that the difference between input and output is always produced by the filtering operation itself. In pre-technical media, this difference is suppressed almost entirely in favour of the clarity of closed systems in which everything is presented as if it has its own unchangeable time and place. Based on the principle of mathematical integration, Fourier analysis puts such pre-technical clarity in a new light. On the one hand, its discontinuous limit cases suggest an even brighter clarity that reaches beyond the whimsicalities of time and transience. On the other hand, as I argued in Chapter Four, Section 4.2b, the idealisations produced by Fourier analysis always already contain their own negation in the form of an inherently asymptotic logic: although the logic of the domain of ideal filters assumes that physical signals infinitely tend toward the limit case of analytical representation, exactly their failure to fully coincide with these limits, is testament of their physical existence. Marking the distance between the idealised representations of mathematical analysis and the fuzzy phenomena they represent, this fundamental asymptotic status of physical signals is the quintessential figure of the rupture between representation and represented, reproduction and reproduced. In the domain of physical filters, the uncertainty principle rules supreme. Although the sheer presence (or thatness) of a technological signal confirms that a physical filtering operation has taken place, as soon as its spectral identity (or whatness) can be determined, its temporal presence, appearing and disappearing like a flash of lightning, already vanished. Hence, because the thatness of physical filtering operations always escapes our perception, the random sonic traces that shape the whatness of their output signals are the only way to deduce they happened in the first place. The physical cuts of technical filters therefore constitute the irrepresentable foundation of all technological signals. From an idealistic perspective, the output of a filtering operation asymptotically tends toward the limit case of an ideal model: frequencies tend toward the infinite oscillations of ideal sine waves and temporal measurements tend toward NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’ 203

the infinitesimal precision of Delta functions. From this point of view, the asymptotic status of technological reproductions always negates their ideal representation. This interpretation conceals, however, that the representation of these discontinuous limit cases constitutes a jump from the material real of physical phenomena to the symbolic domain of ideal filters. When the output of a technological filtering operation is considered on its own instead of in relation to a model of perfect representation, its asymptotic status confirms the primacy of the irrepresentable moment of filtering. Acknowledging this primacy means a shifts from the periodic, steady state aspects of a signal—passing through the filter unchanged—to those aspects that do not pass through unchanged and force the signal to “bifurcate, to take another appearance, another direction, a new order" (Serres 1982b: 160). It focuses attention on the traces of physical filtering operations that cannot be represented beyond their moment of production and cannot be reduced by a secondary filter without changing the signal, if only in the slightest, once again. It focuses, in short, on the transient small extra that is always added to the signal by the cut of a physical filter. Normally, a signal is considered ‘asymptotic’ because of its never- coinciding with the ideal limit case it tends toward. Assuming the primacy of physical filtering operations, on the other hand, reveals that this asymptotic status is precisely what defines the physical signal and sets it apart from symbolic idealisations in the domain of the ideal filter. Because the traces of the filtering operations that produce the signal in the first place are inseparably inscribed in its frequency composition, no signal ever matches the normative standard assumed by the ideal model. When compared to this standard, the filter seems to produce imperfect asymptotic signals, but taken on its own in the moment of its physical unfolding in space and time, the output is always entirely itself and, if ever so subtly, inherently different from all other signals, original input included. Hence, the output of technological sound reproduction is not an incomplete representation of some ideal model or an insufficient reproduction of some original input. Singularly shaped by the transient traces of the moment of physical filtering, technological signals can be 204 NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’

interpreted as inherently asymptotic. Acknowledging this inherent asymptotic status contributes to our understanding of the fundamental role of the parasitic middle—the filter—in shaping the output of a recording chain. Exactly the sonic characteristics produced by the physical cuts of technical filters make each sound singularly different from the next and reveal that technological sound reproduction and analysis does not bring the idealised model into sharper focus, but the irrepresentable moment of physical filtering itself. This primacy of the moment of filtering is particularly audible in practices where the normative logic of input and output, original and copy is overturned entirely and replaced by a circular logic, which is the logic of feedback. When the output signal becomes the input signal and starts to resonate with itself in a closed loop, the filter reveals itself as the primary source for sound production. In a feedback loop, the operations of the channel spin out of control and begin to produce sounds that are entirely brought forth by the channel itself. Even more strikingly, the practice of so- called ‘no-input mixing,’ for instance by Japanese sound artist Toshimaru Nakamura, uses only the circuitry of a sound board to generate an output. By “connecting the input of the board to the output” and “manipulating the resultant feedback,” this music does away with any sense of an original input and consists solely of the transient traces of filtering channels (“Toshimaru Nakamura/Sachiko M” 2001, Nakamura 2001).93 This shows how the filtering channels in between always shape the output of the sound reproduction chain. Whether heard in the control room of the music studio, in the comfort of one’s own living room, while driving a car or dancing in the club, the signal at the narrow end of the chain of filters, I argue, is both radically different and fundamentally the same as the signal that originally went in. It is radically different in the sense that its spectral contours and temporal flow are singularly unique in comparison

93 Similarly, in “Tactility, Traces, Codes: Reassessing Timbre in Electronic Media,” composer and sound scholar Gabriel Paiuk discusses other musical practices in which “the use of lo-fi devices, circuit-bending, cracked electronics and a resurfacing of older technologies is coupled with digital technology in a process which emphasises the devices characteristic modes of sound production and artefacts” (2013: 306). NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’ 205

to those of the input; and fundamentally the same in the sense that the output signal is no more or less physically real than the input signal.

5.2 Producing a new signal a) The real of technological sound signals The logic of filtering belies the idea that sound recordings are incomplete or flawed reproductions of some original sound event, which should ideally approximate the physical characteristics of that event as close as possible. Instead, it emphasises how the output of any reproduction chain is singularly shaped by the random sonic traces added to the signal by each irrepresentable moment of filtering between input and output. Acknowledging this primacy of the moment of filtering for shaping the characteristics of the output signal asks for a more thorough revaluation of the status of the output and its relation to the original input and the filtering channel. I propose that such a revaluation can draw from a critical look at Kittler’s famous conceptualisation of gramophone, film and typewriter in terms of the Lacanian Real, Imaginary and Symbolic, already briefly mentioned in Chapter Two, Sections 2.2c and 2.5a (Kittler, 1999). In his 2011 essay “Symbolizing Time: Kittler and Twenty-First- Century Media,” media philosopher Mark Hansen critically revisits Kittler’s psychoanalytic interpretation of technical media, specifically focussing on his claim that the gramophone processes acoustic data without any symbolic encoding and is therefore the only medium that operates directly on the level of the Lacanian Real (Hansen 2015). For many centuries, Hansen cites Kittler’s basic argument, all symbolic notation fundamentally relied on the “presupposition that contingency was removed in writing, noise in music, and entropy in order” (Kittler in Hansen 2015: 228).94 In the last decades of the nineteenth century, however, the hegemony of this form of noiseless, symbolic representation by written signs was broken by the emergence of technical media that were able to directly inscribe physical signals onto material hardware.

94 Original citation in Kittler, 1993: 195. 206 NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’

Emphasising this structural difference between symbolic representation on the basis of written signs and the physical inscription of signals by technical media, Hansen continues, enabled Kittler to conceptualise sound recording “as an inscription and manipulation [...] of the real—the physical flow of time—itself” (220). In the case of sound media, Kittler thereby understands the transition from symbolic manipulation to physical manipulation as a transition from pre-technical media that operate in the Lacanian Symbolic to technical media that operate on the level of the Lacanian Real. Hansen, on the other hand, argues that this fundamental transition never took place (220). To enable the transmission and storage of information, Hansen argues contra Kittler, all mediatic representation and reproduction requires a symbolic reduction of the physical complexity of signals in space and time. Hence, technical media are as dependent on symbolic and physical filtering operations to process raw physical signals as pre- technical media. Thus, in Hansen’s analysis, all media, whether pre- technical, analogue or digital, produce a “physical or material symbolization” of the real that is “generated through the processing of the real as number” (230). Produced on the basis of a symbolic logic, such as the logic of Fourier analysis, technological reproductions can “only get asymptotically closer to the real they symbolize” (233). According to Hansen, both technical and pre-technical media filter symbolic representations out of a much more complex physical reality, which is why he argues that the difference between the operations of technical media and those of pre-technical media is only a matter of degree. However, by emphasising the fundamentally symbolic nature of all media technological operations, Hansen’s reinterpretation of Kittler adheres to the idea that technical reproductions are fundamentally imperfect, because they continuously tend to but never coincide with the “real they symbolize.” What sound reproductions lack, he argues, is exactly what Kittler claims it captures: all the continuous and contingent characteristics of original input signals that produce their irrepresentable flow in the Lacanian Real. To Hansen, the filtering operations of media— whether symbolic or physical—thereby assure that neither sound NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’ 207

reproduction nor any other medium can actually store, reproduce or manipulate the Real. Because their output signals are always only a symbolically processed version of the input, technological filtering operations produce what Hansen calls the “symbolic of the Real”—a symbolised version of the irrepresentable Real of the original signal (233, emphasis in original). With this line of reasoning, Hansen reinstalls the idea of the primacy of an ideal model according to which the output is always an incomplete version of the input. Relegating all media to the domain of the Symbolic, he upholds the idealist connection between input and output that is characteristic of the myth of perfect fidelity. Hansen thereby reinforces the notion that all technological reproductions asymptotically tend toward a discontinuous limit case and are thus inherently inadequate. I argue, however, that this asymptotic status of the output signal is not the result of the incapacity of technological reproduction to fully represent or reproduce the input signal. Following the logic of filtering that emphasises the primacy of the parasitic middle, the limit case toward which all reproductions in Hansen’s interpretation tend is not the original input at all, but the smooth limit represented by the ideal model itself. The limit to which Hansen claims the output of technical media is getting “asymptotically closer” is not and can never be the irrepresentable Lacanian Real of the input signal. It is the Symbolic limit case of the representational model itself. Hansen is therefore correct to argue that sound reproduction does not store or reproduce the Real of the original sound event, because the sonic Real of any spatiotemporal event remains fundamentally irrepresentable and inaccessible, regardless of the form of mediation or representation. The sonic traces of filtering operations that are hardwired in the output signal, however, are not related to the input signal either. They originate in the filtering operations itself. Whereas Hansen’s alternative category of the “symbolic of the Real” implies a fundamental relation between input and output according to which the output symbolically represents the Real of the input, I argue that these traces reveal that, although the one does not fully reproduce or represent the other, input and output both constitute the irrepresentable sonic Real. 208 NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’

No matter to what extent the output compares to the input, shaped by the moments of physical filtering in between the two ends of the reproduction chain, the signals produced by technological sound reproduction are complete sonic events in and of themselves. Contrary to words in books, notes on paper or even images on canvas or celluloid, the sounds that emanate from speakers or headphones are as non-symbolically complex and physically real as the signals at the beginning of the chain. They might be physically dissimilar in terms of frequency spectrum and temporal development, but the sound that comes out of speakers and headphones is in no way less physically ‘real’ than the sound of so-called acoustic or ‘live’ music. Hence, on the one hand, Hansen’s reinterpretation of Kittler’s qualification of sound recording as the Lacanian Real justifiably points out the structural impossibility to store or process the Lacanian Real. After all, according to Lacan himself, the Real cannot be experienced, expressed or represented, but only appears through the cracks of the Imaginary and Symbolic. Like Kittler’s flash of lightning, its presence is experienced as a sudden shock that can only be grasped after it took place by means of analysing its semi-periodic traces through the mirror of the Imaginary and the grid of the Symbolic. On the other hand, however, by denying this relation between sound recording and the Real altogether and introducing the idea of “the symbolic of the Real,” Hansen throws out the baby with the bathwater. Whereas Hansen’s interpretation precisely reintroduces the type of idealist perspective on technical media that Kittler’s media materialism does away with, I suggest there is a close connection between sound reproduction and the Real, albeit a connection that is somewhat different from the one that Kittler suggests as well. b) Completely transparent windows Kittler’s reputation as a radical post-humanist notwithstanding, Hansen begins his article by confessing that he cannot help but see “signs of a distinct, if subterranean, ‘humanism’ everywhere at work in Kittler’s method” (210). For Hansen, Kittler’s move to generalise Lacan’s concept of the Symbolic to include not only alphabetic but also numerical symbolisations means that all technological signal processing implies some kind of “human-machine co-functioning” according to which both human NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’ 209

subjects and technical media operate on various levels of the Symbolic (212). Although I support this call for the rehumanisation of some of Kittler’s core ideas, contrary to Hansen, I claim it is not in Kittler’s thinking of the Symbolic but in his conceptualisation of the relation between technical media (and especially sound media) and the Real that this humanism comes to the fore. In his reading of Kittler, Hansen erroneously equates the symbolic analysis on which the operations of technical media are based with their physical implementation in material hardware. Contrary to what Hansen claims, only digital media apply the kind of “physical or material symbolization” that he attributes to all technical media; and even in the case of digital media, this symbolisation “generated through the processing of the real as number,” is based on, as I cited Siegert’s rendition of the discussion on the status of the digital in the 1950s Chapter Two, Section 2.5a, the exclusion of a time of non-reality, which means the exclusion of the noise of the analogue switching operations that support all digital processing (230). Notwithstanding the observation that digital technology indeed “allows for a symbolization of the temporal flow on a far finer scale” than alphabetic symbolisations or human senses, when transduced from digital symbols to electrical currents, and from electrical currents back to acoustic signals, these digital symbolisations of the temporal flow of signals re-enter physical reality and are subject to the irrepresentable contingency of the Lacanian Real (228). Crossing this threshold from digital representation to physical sounds, for instance, turns symbolic quantisation errors into acoustic harmonic distortion—the digital equivalent of the random noise of analogue channels—which occurs because digital filters can only process sounds in the form of samples that Kittler describes as “windows that depict the changing world in standing waves” (Kittler 2006a: 69). The application of dither, in turn, disguises this rupture between the ideal domain of perfectly smooth limit cases and the domain of physical filters that is revealed by harmonic distortion. These “windows that depict the changing world in standing waves” thereby turn the thatness of events into the whatness of objects, turn change into stasis and randomness into periodicity. In concordance with 210 NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’

Fourier analysis, the windows or samples of digital sound technology store the spatial contours of physical signals on dispense of their temporal development. This process is supposed to be completely transparent—the window itself should not be looked at, because the filter is to remain concealed by the clarity of its input and output. Hansen is correct to heighten the fundamentally symbolic basis of such concealing and revealing, which is not only at work in digital media and the operations of analogue noise reduction, but also in electro-acoustic technologies like the analogue synthesiser (Kittler 2000: 146). If, however, our only sense of the Real is through a sudden apparition or a crack in the tissue of the Symbolic, and if this crack is like a flash of lightning that briefly brightens the dark sky during a thunder storm, the thatness of this event and the whatness of its analysis relate to each other like the Real and the Symbolic. Hence, although the models through which the analytical representation or physical reproduction of the whatness of sound became possible belongs to the Symbolic and the Symbolic only, the thatness of the signals going into and coming out of the filter fundamentally belongs to the Real. This means the visceral impact of the sounds that musicians, producers and engineers listen to in the studio; the sounds that come out of speakers and headphones all over the world, which listeners relate to and cherish; the sounds that went through a long chain of parasitic filters that physically implement a series of symbolic operations; the impact and significance of these sounds cannot be conceptualised or explained, as Hansen suggests, within the framework of the “symbolization of the temporal flow” or the idea of the “symbolic of the Real.” Instead, I argue, a successful analysis of these sounds has to take into account the transience and randomness that is not part of the original input signal nor of the ideal model, but is the sonic trace of the irrepresentable moments of physical filtering that produced these sounds in the first place. Hansen’s interpretation of Kittler’s analysis of technical media remains rooted in an account that takes the absolute clarity and transparency of ideal filtering operations as its ultimate point of reference. Kittler himself, however, acknowledges that the operations of technical media are not concerned with such symbolic limit cases and operate on a different plane altogether. With their discreet symbolic representation of NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’ 211

physical phenomena in binary code, digital media no longer take “into account philosophical dreams of infinity,” because their mode of representation is structurally finite (Kittler 1999: 15). In terms of the physical processes they represent or reproduce, on the other hand, digital media are no different from their analogue counterparts in producing or reproducing entirely real, physical sound signals. Hence, regardless of what happens in the analogue or digital black boxes in between input and output, technological sound media always produce streams of sound that exhibit the same periodic and non-periodic spectral and temporal qualities as any other. c) Beyond symbolic representation: the Real of reproduced sound The age-old symbolic code of music notation, Kittler told the audience at an award show in honour of British musician and producer Brian Eno in Berlin in 1998, “granted European music the unique opportunity to escape passing time” and compose increasingly complex musical structures (2013a: 36).95 Whereas classical western music notation suspends the contingencies that are inherent to the material basis of symbolic representation by means of writing notes on paper, the music of the media age takes shape through a chain of filtering operations and therefore no longer escapes time. With this transition from symbolic code to physical inscription, those aspects of sound that have always escaped and will always escape the control of human subjects became the core of musical development. Music produced by technical media is still directed at human ears, but, as Kittler argues in “Musik als Medium,” it is no longer concerned with intrinsic meaning and the interpretation of symbolic sign systems (1995a: 99). Instead of an always already intrinsically meaningful connection between music and listeners based on a delineated code and its congruent meanings, the relation between technological sound and human listeners is formed on the basis of the physical flow of sound signals themselves. Without concern for symbolic signification, music in the age of technical

95 “[…] gewährte der europäischen Musik die einmalige Möglichkeit, aus der vergehenden Zeit zu springen.” 212 NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’

filters deals with physical surfaces and raw sound data that take shape and exist independently of human musicians or listeners. When actual sounds, voices and noises—physical and audibly real—can be recorded, reproduced and manipulated, symbolic meaning and human agency (those cornerstones of the nineteenth century paradigm of written music) become increasingly arbitrary.96 At the moment of capture, sound recording cuts complex waves from the flow of time and stores them acoustically, electro-magnetically or digitally on some kind of hardware medium. At the moment of playback, these data are turned back into sounds that include the sonic traces of the chain of interconnected filters in between human musicians and human listeners. Equipped with a wide range of machines that shape and manipulate whatever they record, musicians, engineers and producers in the media age compose, as Kittler puts it in “The God of Ears,” “walls of sound” consisting of “innumerable tracks, one upon the other” that fill the “ears and brain” of listeners (2015a: 11). Singularly shaped by the chain of filters, these concrete, physical, manipulable, technological sounds do not represent anything they not already are.97 In contrast to symbolic representations that assume some inherent meaning is hidden in or behind the music, listeners to music shaped by technical filters do not listen behind the sound to discover forms, structures and deeper meanings beyond the sound itself. The meanings that listeners attribute to music therefore no longer bear a necessary or

96 On the increasing influence of non-human agents and the growing importance of the physical sound of music in the age of technical media, Peter Szendy argues in Listening that “operations ‘external’ to the musical (to so-called ‘pure’ music) are now endowed with the ability to create signifying segments in the course of the music’s flow” (2008: 135). In “Waves, Flows, Streams: Die Illusion Vom Reinen Sound,” Rudolf Maresch writes: “richtig ist, dass der Sound erst mit der elektronischen Revolution entsteht. Erst mit ihr, die vor etwa einem Jahrhundert in Alt-europa einsetzt, wird er am Rande der Notenschrift für das Menschenohr überhaupt wahrnehmbar” (2003: 205). 97 As Rudolf Maresch argues: “Der Klang ist bereits sein Inhalt. Sound vermittelt nur sich selbst" (2003: 206). On a less philosophical note, in Any Sound You Can Imagine, musicologist Paul Théberge writes that in popular music "the term 'sound’ has taken on a peculiar material character that cannot be separated either from the 'music’ or, more importantly, from the sound recording as the dominant medium of reproduction” (1997: 191). NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’ 213

intrinsic relation to the sounds that originally went into the chain, let alone to a metaphysical musical referent beyond physical sound altogether. Shaped by filters that bring forth the singular specificity of technological signals in the first place, Kittler argues in McLuhian fashion that under technological conditions, “sounds speak of what is done by sounds” (2015a: 13). Owing to this emphasis on physical sound, there is something compulsive about the music of the media age, something that goes beyond the symbolic ordering processes carried out by human subjects, something that works on us and controls us instead of the other way around. In terms of Kittler’s conceptualisation of sound media, I suggest, this compulsive aspect of technological sound points to the irrepressible presence of the Real. Hansen opens the last paragraph of his article with a lengthy quote from the Artforum interview I cited in Section 5.1c. In the interview, Kittler poses the question whether the Real is “what Turing says it is or [...] what Lacan says it is?” (Kittler in Hansen 2015: 233).98 This is the question, in other words, whether the Real (or “to use the old-fashioned term, nature itself”) is potentially computable by a Turing machine or not (Kittler in Hansen 2015: 233). Kittler tentatively answers his question by pointing out that “physicists at Livermore [Lawrence Livermore National Laboratory, MK] are tending to side with the Lacanian view of the real as the impossible in relation to our machines and systems” (Kittler in Hansen 2015: 233). Hence, if these physicists are correct and the Real is indeed “what Lacan says it is” (the impossible and irrepresentable), the randomness caused by the dilemma of the uncertainty principle cannot and will never be fully analysed or processed and noise still rules supreme. Hansen sees in the incapacity of technical media to fully represent “the real they symbolize” a confirmation of the inherently Symbolic status of their operations. I argue with Kittler, however, that exactly this supposed incapacity, which results in the continuous production of new signals, supports a deeper or darker conceptualisation of the impact of technical media. For this conceptualisation, noise, randomness and the Real are crucial.

98 Original citation in Kittler 1992b: 68. 214 NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’

5.3 The new sound of music a) Wagner’s acoustic machine With the implementation of ideal filtering processes in the physical hardware of acoustic media, a new modality of sound took hold of our musical culture. A modality that shattered traditional concepts of musical signification and representation because it no longer adhered to the discursive categories that had determined Western musical culture for centuries. Shaped by the primary operations of physical filters, the musical sensibility that emerged with the development of sound reproduction is as much informed by communication engineering as it is by musical aesthetics, as much by acoustics as it is by music theory and as much by the physical filters of technical media as by the symbolic filters of human subjects. Over the course of most of his oeuvre, from the early eighties to some of his last writings in the first decade of the twenty-first century, Kittler refers to this emerging musical sensibility as the ‘other music.’ Shaped by the physical operations of sound media, Kittler writes,” this music “no longer derive[s] its power from alliances with the medium of language and its ‘meanings’” (1995a: 99). Slowly but steadily leaving behind the anthropocentric, subject-based Western concept of music that developed in close alliance with an increasingly sophisticated form of musical notation, the ‘other music’ is governed by what Kittler calls “pure media-technology, pure control flow” (99). Operative on a plane between the autonomous processes inside technical media and the aesthetic signification of human listeners (a plane that Wolfgang Ernst calls sonicity), this ‘pure control flow’ of the ‘other music’ is no longer based on symbolic reductions by ideal filters, but on the production of physically present, complex signals by technical filters (Ernst 2016: 21-33). Historically, Kittler sees the first signs of this ‘other music’ emerge in the second half of the nineteenth century, not long after Ohm and Helmholtz’s brought Fourier’s analytical breakthrough to the field of acoustics. Even before it came into its own with the invention of the physical inscription of sound waves onto the hardware of technical media, however, the shift toward the physicality of sound that defines the ‘other NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’ 215

music’ began to influence and replace age-old representational concepts of music. For Kittler, as always the quintessential German, it first became apparent in the “pure dynamics and pure acoustics” of Richard Wagner’s music dramas (1994: 224). Whether one agrees with Kittler’s singling out of Wagner’s work or not, the underlying argument highlights various key aspects of the concept of an ‘other music’ that reveal how Kittler conceives the changing relation between musical sound and human listeners in the age of technical media. Although still transcribed in the representational system of discreet notes on paper, Wagner’s music dramas show the influence of the new physical concept of sound that emerged with Fourier, Ohm and Helmholtz. According to Kittler. after Helmholtz’s analysis of tone colour, which defines the difference between musical sounds and unmusical noise as the difference between periodic and non-periodic frequencies, the erstwhile “clear division” between “orchestral sounds” that represent something outside of themselves (whether mimetically or in reference to some musical ideal) and “the noise of nature” that is always physically entirely its own, becomes increasingly arbitrary (1997: 7).99 Ever since the invention of Fourier analysis, Kittler writes, all music can be conceived as a set of periodic and non-periodic frequencies; a “subset in a global noise- spectrum” (7).100 Instead of adhering to the traditional system of symbolic representation, Wagner was “the first who truly wrote out the noise-source called nature” by trying to approximate acoustical sounds in all their spectral and temporal complexity (7).101 Influenced by developments in physical acoustics and early sound technologies, this new sound of music no longer produces acoustic representations that support a dramatic narrative or symbolically refer to something outside of the musical domain. Although still translated into a written score that uses the Western diatonic scale to transcribe approximations of physical sounds to be performed by human musicians

99 “Wenn zwischen Orchesterklängen und Naturgeräuschen prinzipiell keine klare Grenze mehr besteht.” 100 “Wenn die Musik, mit anderen Worten zur Untermenge eine weltweiten Rauschspektrums geworden ist.” 101 “Der erste, der die Rauschquelle Natur auch wahrhaft auskomponiert hat.” 216 NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’

on acoustic instruments, Wagner’s singers and , Kittler argues, operate as “a machine” to generate acoustic effects that reproduce rather than represent the sounds of nature (7).102 As Foley sound avant-la-lettre, this acoustic machine reproduces the physical properties of natural sounds. From an approximate Fourier series on E-flat major in the overture of Das Rheingold to the “pure noise” that accompanies the downfall of the gods at the end of Götterdämmerung, Kittler argues, the sounds in Wagner’s music dramas, represent nothing but themselves (1995a: 96). When sounds simply are what they are, music no longer requires hermeneutic interpretation. With this shift toward the physicality of sound itself, the difference between art and nature and between representation and represented, which underpinned the representational logic of written music, was effectively cancelled. The ‘other music’ that Wagner’s music dramas prefigure is not the physical instantiation of a perfect, metaphysical piece of music that remains hidden behind the normative notes of the score. With every performance, the ‘other music’ is completely itself. Wagner first conceived the The Ring of the Nibelungs about a decade before Scott de Martinville sung ‘Au Clair de la Lune’ into the phonautograph; and only one year after the first full staging of the cycle in Bayreuth in August 1876, Edison allegedly sang Mary Had A Little Lamb into his newly invented machine. In between those events, the ‘other music’ that we hear all around us every day a hundred and forty years later, came into its own. b) Sound buzzing seductively in every ear Inspired by a passage in Nietzsche’s Beyond Good and Evil, Kittler’s concept of an ‘other music’ first appears in the “The God of Ears,” written in the early 1980s (2015a: 3). The title character of this piece is the goat-legged demi-god Pan, protector of shepherds and flocks, whose love interest, the nymph Syrinx, turned herself into a reed whilst desperately trying to escape Pan’s grasp. To keep his lost love close, Pan made a flute out of the reed and, Kittler writes, “with no warning whatsoever, when shepherds dreamt and the midday silence was overwhelming,” Pan “buzzed in every

102 “Wagners Musik ist also eine Maschine.” NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’ 217

ear” (3). Although Plutarch wrote in “On the Disappearances of Oracles” in the second century A.D. that the great Pan is one of the few gods of whom it has been reported that he died, Kittler begs to differ because “gods of ears cannot pass away” (3). Instead, the powers of this god that buzzed in the ears of shepherds and nymphs and whose manic shouting caused panic amongst everyone who heard it recently returned in the form of modern communication technology. With “amplifiers and PA systems” that produce and reproduce previously unheard sounds, the seductive powers of the god of ears are casted into material hardware (3). Abruptly breaking the overwhelming midday silence, sounds that forcefully flow from loudspeakers buzz in the ears of listeners all over the world. Sound technologies are different from traditional musical instruments because the manual labour of human agents, which even Wagner still required to turn his silent notes into audible sounds, becomes subordinate to technical filters that bring forth and shape sounds without the symbolic filtering operations of human subjects. When sound media are switched on and plugged in, the chain of filters autonomously produces sounds. In many cases, human musicians still play instruments, either acoustic, electronic or digital, and human engineers meticulously place microphones, pull cables, turn knobs and switches; more recently, such sound production and manipulation even happens entirely in the digital domain. However, regardless of the specific technical setup and the amount of human actors involved, everything that happens in the black boxes along the parasitic chain maintains a level of contingency that is courtesy of the irrepresentable moment of filtering. Hence, the sound that flows from the speakers is no symbolic representation created by and for human subjects. When sounds are recorded and subsequently pressed on vinyl, burned on CD or stored in the cloud, the listener only has to plug in the speakers or earphones, press play and the ‘old magic’ that Kittler describes in “Lightning and Series – Event and Thunder” unfolds: physical sound flows from the speakers as non- periodically and spectrally complex as any non-technological signal, to be repeated time after time (2006a: 68). With this blunt physical presence that has no need for human symbolisation, the ‘other music’ rings the end of the idea that musical sounds represent anything that they not already 218 NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’

are. Produced by machines that create singular sounds all by themselves, the music of the media age “acts,” as Kittler puts it, “as a song from beyond mankind,” a song that is sung for no one and nothing in particular because it only consists of sound for the sake of sound (2015a: 12). This makes the irrepresentable moment of technological filtering all the more crucial for our understanding of the ‘other music.’ The symbolic idealisations of the domain of the ideal filter suggest the clarity of completely noiseless signals. The transient nature of the moment of the physical cut, however, fundamentally negates such clarity and introduces infinitesimal fuzziness. Furthermore, the uncertainty principle that marks the rupture between ideal representations and physical signals reveal that we do not and will never have complete control over the moment of filtering itself. Similar to the brief click of a photographic shutter, the transience of the moment of filtering prevents the filter from producing unambiguous information about what the signal is at that precise moment it takes place. In technological sound, the sonic traces of this transient moment of filtering reveal that the filter continuously produces new sounds that singularly flow through the present.

5.4 An ‘other music’ and the presence of the Real a) “A still darker presence” The relation between the clarity of scientific idealism and Kittler’s darker, existentialist inclination is what separates my reading of the concept of an ‘other music’ from Hansen’s intended ‘rehumanisation’ of Kittler’s media theory, as well as from musicological or cultural historical accounts that interpret the emancipation of noise in musical culture either as part of subversive agendas or as a growing reservoir for ‘new’ sounds. Instead of considering the media technological revolution as one of the crown achievements of human intellect and engineering skills, what Kittler’s media philosophical reading of the musical development of the past two centuries underlines is that humanity did not leave the darkness of NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’ 219

existential ignorance behind. 103 Quite the opposite: as the increasing autonomy of technical media undercuts stable human agency, because we are increasingly unaware of the technological processes that shape, affect and determine the world around us, the grasp of human agents diminishes and the darkness is only deepening. Whilst technical media enable ever more advanced possibilities for symbolic representation and technological reproduction, the more advanced its filters become, the sharper they bring the fundamental irrepresentability and inaccessibility of the Real into focus. This condition is foreshadowed in the passage from Beyond Good and Evil that inspired Kittler’s notion of the ‘other music.’ In his discussion of the European spirit in the section entitled “On Peoples And Fatherlands,” Nietzsche writes about ‘the south in music’ (2009: 173). Although Kittler claims that “of the other music” Nietzsche “knew only Wagner,” Nietzsche especially recognises the ‘southern aspect’ of music, as described in more detail in Nietzsche Contra Wagner, in Bizet’s Carmen (Kittler 2015a: 12; Liébert 2004: 198-202). In stark contrast to “the ‘late Wagner’ and his Parsifal music,” Nietzsche writes, the ‘southern music’ of Bizet incites a dream of a

more profound, more powerful, perhaps more evil and more mysterious music, a supra-German music which does not fade away, turn yellow, and grow pale at the sight of the blue voluptuous sea and the brightness of the Mediterranean sky (2009: 173).

103 This darkness and ignorance invoked by Kittler clearly has Heideggerian overtones to it. As Carol White writes regarding Heidegger’s analysis of the relation between light and darkness as the unconcealing of Being: “Ordinarily we think of shadows and darkness as indicating a lack of light or its complete denial. But, Heidegger argues, the shadow is a testimony to the concealed emitting of light. It is the incalculable, the unpredictable and unthinkable, which lies beyond our capacities to represent, yet points us to being as the locus of the light which may illuminate the dark comers one day” (2005: 74). 220 NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’

This more profound, powerful, evil and mysterious music, shaped by the duality between southern and northern or light and dark tendencies, is what Kittler calls the ‘other music.’ By the time he wrote Beyond Good and Evil, Georges Liébert writes in Nietzsche and Music, “Apollonian images and visual metaphors become more and more frequent” in Nietzsche’s writing and “we find him looking, contemplating, open to the forms that the light reveals to him against an azure background or sharp shadows” (2004: 197). Against the dark, northern, Dionysian inclination of “the ‘harmonic fog’ of Wagner’s orchestra,” which he once admired above all else, Nietzsche now puts, as Liébert cites Ecce Homo, Bizet’s "absolute transparency of his woven counterpoint, the utilization of each instrument in terms of its specific coloration, in the voice that is most natural and fitting to it (Wagner does violence to every instrument), his most economic use, delicatesse instead of a dark subterranean stimulation of our instincts" (Nietzsche in Liébert 2004: 200). Taking this context into account, the imagery of a “more profound, more powerful, perhaps more evil and more mysterious music” in Beyond Good and Evil suggests that Kittler’s concept of the ‘other music’ contains a similar Apollonian emphasis on form, clarity and order. Indeed, without the need for subjective human agency, the ‘other music’ does not rely on imitations and representations and does not “not fade away, turn yellow, and grow pale” in comparison to the sounds of the natural world. Inscribed on the hardware of technical media that operate directly in physical reality, the sound of the ‘other music’ can be repeated over and over again; and every time it is played, it sound “just as rich, colourful and bright as nature itself.” Hence, it does not symbolically represent but physically produces the inextricable, inaccessible Real of the infinite and continuous “noise source” called nature (Kittler 1997: 7). On the other hand, however, this reproduction of the “noise source” called nature runs counter to the Apollonian tendency of the ‘other music’ and bends toward a darker, uncontrollable, Dionysian side of Nietzsche’s musical philosophy that takes Dionysus, as Liébert puts it, as “a symbol of the most affirmative will to life" (2004: 166). In a way, Nietzsche’s “brightness of the Mediterranean sky” is reminiscent of Derrida’s moment NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’ 221

in the Greek summer sun described in Chapter Four, Section 4.3c—the ideal, infinitesimally short moment at which Derrida felt that “we are infinite, […] eternally” (2010: 63). The fundamental inaccessibility of this moment is the dark shadow cast by the light of that summer sun. It emphasises the inherent irrepresentability and pure transience of the infinite presence of the present. Like the “noise source called nature” cannot be captured or reproduced in its entirety, the colourful and bright sound of the ‘other music,’ which Kittler first recognises in the northern sounds of Wagner’s acoustic machine, can only be produced by filters that do not capture the inextricable complexity of the event at the very moment it occurs but always bring forth new sounds that, in turn, keep slipping from our control. Hence, the concept of the ‘other music’ is split between the Apollonian sharper focus on the physical nature of sound signals that the combined forces of mathematical analysis, theoretical physics and communication engineering achieved, and the Dionysian irrepresentability of the very moment that enables this focus to appear in the first place—the inaccessible moment of filtering marked by the uncertainty relation between time and frequency. In the closing paragraph of “The God of Ears,” Kittler describes this double-faced character of the ‘other music’ as follows:

Even a heart attached to contact microphones and oscilloscopes becomes still. And when, with loud and quiet, light and dark, Heaven and Hell, all differences disappear, another realm (possibly known as Satori by other cultures) is coming closer. The media explosion of our days, therefore, should not only be heard in the media-theoretical manner of its prophets. According to Marshall McLuhan, the message of the synthesizer is simply the synthesizer. But even if the darkness is so overwhelming that no dark side of the moon exists, electronic media might yet invoke a still darker presence. Original sound—Waters: ‘the medium is not 222 NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’

the message, Marshall … is it? I mean, it’s all in the lap of the fucking gods … (Pause for laughter)’ (2015a: 16).

The peculiar “Original sound—Waters” preceding the words of Pink Floyd bassist and songwriter Roger Waters that close the essay is a translation of the German term “O-Ton” (1993a: 148). In journalism, the O- Ton or so-called Originalton connotes, as media scholar Jörg Häusermann explains, a “small sound document” to make a radio or other kind of report more “lively” and “authentic” (2007: 25). Although seemingly insignificant, the use of this term thus highlights the apparent contradiction between a sound that is both unique and original as well as technologically recorded and reproduced. As such, Kittler’s reference to Water’s quote as an “O-Ton” indicates the duality that marks the ‘other music.’ Created by recording and reproduction technologies that supposedly reproduce singular acoustic events, in the process of this reproduction, these singular acoustic events are affected by the operations of the irrepresentable parasitic third of signal processing that make the sound of the ‘other music’ singularly original in itself.104 With technological sound reproduction, physical sound becomes the centre of musical invention, because technical media do away with the representational logic according to which musical sounds represent anything that they not already are. Besides the Apollonian positivism of technological operations, however, the “still darker presence” that technical media invoke goes beyond the media theoretical dogma that “sounds speak of what is done by sounds” or “the message of the synthesizer is simply the synthesizer.” Instead, the darker presence of the music of the media age is caused by the inescapable singularity or, as Kittler puts it “unfathomable stochastics,” produced at the moment of the cut of the physical filter (Kittler 2015a: 15). Tending toward the

104 Notably, whereas Paul Feigelfeld and Anthony Moore stay as close as possible to the original by translating the quintessentially German “O-Ton Waters” as "Original Sound—Waters,” Erik Butler’s translation leaves it out entirely and simply writes “Waters” (Kittler 2013b, 45-56). As will be clear from my analysis, I think Feigelfeld and Moore’s solution does more justice to the original. More on the O-Ton and its interpretations in: Wegmann et al 2007. NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’ 223

infinitesimal point-like presence of Dirac impulses and impossible to capture with the ideal spectral filters of Fourier analysis, technological filtering operations continuously produce the Real of reproduced music. b) Asymptotic listening to transient sounds “When, with loud and quiet, light and dark, Heaven and Hell, all differences disappear,” Kittler prophesises, “another realm (possibly known as Satori by other cultures) is coming closer” (2015a: 16). What does this mean? According to Serres, differences appear with the introduction of a parasite in the system and following this logic, “all differences disappear” when the influence of such parasites is symbolically annulled, for instance by applying an ideal filter that symbolically reduces it to zero. Sonically, such a complete reduction of noise by an ideal filter would produce an ideal, infinite sine wave. Hence, if difference disappears in the absence of the parasitic influence of physical filters, a realm where “all differences disappear,” toward which Kittler suggests the “media explosion of our days” is tending, bears a strong resemblance to the Fourier domain, in which strictly periodic sine waves oscillate unchanged into eternity. Processing signals on a level that goes “beyond mankind,” the output of sound technologies continuously tend toward the clarity of the domain of the ideal filter, reaching for something beyond the clouds and beyond the inescapable transience of life. This tending toward a realm where everything has its proper and unchangeable time and place is in keeping with the profoundly human attempt to either fully capture or halt the flow of time, to go beyond the threshold of the uncertainty relation between time and frequency and overcome the gap between representation and represented. This is the attempt to inhabit a universe in which every copy fully coincides with its original, every output is identical to the input. It goes hand in hand, however, with the daily experience that signals only come asymptotically close to their so-called originals, because both copy and original, representation and represented, only infinitely approximate the idealised purity of the Fourier domain. This asymptotic logic of technical media harbours the “darker presence” that defines the ‘other music’ (Kittler 2015a: 16). Conceptualising the moment of physical filtering as a primary ‘third’ that directs and shapes our understanding of the ‘first’ and ‘second’ 224 NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’

(input and output, represented and representation) puts emphasis on the middle ground in between the extremes of Kittler's “we are immortal” and Derrida's “we are infinite.” In the domain of physical filters, asymptotic subjects never fully coincide with their imagined self. They continuously tend toward the extremes at the symbolic limit of Being: the temporal extreme of a pure undividable point-like present at which “we are infinite” and the spectral extreme of the great repetitive beyond in which “we are immortal.” Without ever fully coinciding with these limit cases, we envision ideal technical media to overcome the dichotomy: media that enable real- time signal transmission in perfect, infinitesimally detailed resolution and media that process signals without temporal delay whilst capturing every single detail of each transient event. The sound of a “heart attached to contact microphones and oscilloscopes” that Kittler describes at the beginning of the final paragraph of “The God of Ears” bookends Pink Floyd’s multiplatinum record The Dark Side of the Moon, one of the most successful albums in the history of recorded music. Ever since they were recorded at some time in 1972, these approximately hundred beats sounded billions of times, emanating from speakers all over the globe again … and again … and again … and again. Although the sound of a beating heart usually signifies a foreboding of its inevitable silence (or, in Heideggerian terms, the “indefinite certainty” of its silence), with the aid of sound technology even this most life-confirming and precarious of all bodily sounds can be repeated over and over again, extending its potential lifespan well beyond the average 2.5 billion beats of a biological human heart.105 Every time a sound is repeated, Kittler writes in “Lightning and Series – Event and Thunder,” “time stops, what more do hearts want?” (2006a: 68). Given the chance, hearts probably want to beat forever and because sound recordings often outlive the things or people whose sounds they record, potentially tending toward near-infinite

105 Since its release in 1973, The Dark Side of the Moon sold approximately 50 million copies. With a hundred heartbeats per copy, this amounts to 5 billion heartbeats. If, over the course of the past 43 years, each of these records has been played a minimum of five times (not taking into account illegal copies, online streaming services, radio plays and other sources) it already equals the average 25 billion heartbeats of a human life. Given the lasting popularity of the record, I would argue it safe to assume the actual number is much higher. NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’ 225

repetition, it seems as if the age of technical media indeed delivers on Kittler’s techno-religious dream of immortality. In contrast to what Kittler’s description suggests, however, the sound at the beginning and end of The Dark Side of the Moon is not a recording of an actual human heart “attached to contact microphones and oscilloscopes.” According to music journalist John Harris’ book on the making of the album, the sound is a “simulated heartbeat, looped from a recording of [Pink Floyd drummer, MK] Nick Mason’s ” (2005: 141).106 The conceptual distance between this loop of a specially treated bass drum and the recording of an actual, transient human heart is illustrative of the relation between the ideal of technologically escaping the inevitable delay and the ‘darker presence’ of an ‘other music’ singing “a song from beyond mankind.”107 Although sound technologies allow the endless repetition of any sound, whether “attached to contact microphones and oscilloscopes” or not, the hearts of human beings, musicians and listener alike, will ultimately become still. The ideal realm where all difference disappears is only reached when the finite temporality of life is left behind altogether. Technological signals asymptotically tend toward a clarity that would make one master over time and death, but the cuts of physical filters that produce these signals in the first place continuously escape our control. Every time we replay sounds, the transience of their physical unfolding in time remains absolute. This transience marks the ‘other music,’ producing the noise and distortion that causes reproduced sound and music to resonate in the ears and brains of human listeners over and over again, thus confirming the

106 Concerning the recording of the bass drum ‘heartbeat’ for The Dark Side of the Moon, drummer Nick Mason recalls in his autobiography Inside Out. A Personal History Of Pink Floyd: “Initially we had tried creating the heartbeat that opens the piece from hospital recordings of real pulses, but all of them sounded far too stressful. We returned to the possibilities of musical instruments, and used a very soft beater on a padded bass drum, which strangely sounded far more lifelike, although the average heartbeat rate of 72 bpm was too fast and we slowed it down to a level that would have caused any cardiologist some concern” (2005: 169). 107 In “Positive Feedback: Listening behind Hearing,” David Wills also writes about listening to the sound of one’s own, transient heart: “In every nonbeat or offbeat of the heart, in every flutter or murmur, there resides, if we listen, the irreducibly necessary possibility of stopping beating” (Wills 2015: 87). 226 NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’

fundamental asymptotic nature of both. “The only place where I think about bodies [has been] expressed through the keyword ‘noise,’” Kittler said to Rudolf Maresch around 1992; “the body definitely belongs to these types of beings that are not written in natural, but in real numbers” (1992a: 104).108 Listening to the noise that marks the ‘other music’ means listening to this irrepresentable and irreproducible complexity of the Real, which shapes the temporal transience and spectral singularity of all sonic events as much as it defines the fundamental transience of our physical bodies. c) “Unthinkable, unimaginable acoustic events” “Sound,” Kittler writes in “The God of Ears,” “is the unwritable in music and is immediately its technology” (2015a: 15). Colloquially used to describe the instantly identifiable sonic qualities of songs, artists, records, genres or sound systems, understood through a logic of filtering, this notion of ‘sound’ comes to signify the resonance of the irrepresentable Real in reproduced music. Although, in Kittler’s analysis, it has a pre-history dating as far back as the sixteenth century, the kind of media specific sound that birthed the ‘other music’ really came unto its own with the nineteenth century reconceptualisation of sound in terms of waves and frequencies, culminating in the late twentieth century in the vast sonic possibilities of digital media.109 Digital sound technology enabled the storage, transmission, manipulation and production of sound events in all their spectral and temporal complexity. However, exactly the tempo-spatial aspects of signals that define the Real of technological sound escape its grid as well. Although digital technologies can almost simultaneously tackle physical signals in both their spectral complexity and their temporal development, the

108 “Die einzige Stelle, wo ich über Körper nachdenke, [hat] vorhin schon einmal beim Stichwort ‘Rauschen’ angeklungen […] Der Körper gehört bestimmt auch zu dieser Gattung von nicht in natürlichen, sondern in reellen Zahlen verfaßten Wesen.” 109 In “Wie der Sound von heute vor 300 Jahren entstand," Kittler traces the ‘Sound’ of the media age from Stevin’s sixteenth century conceptualisation of equal temperament, along Mersenne’s discovery of acoustical waves, Saveur’s conceptualisation of overtones and d’Alembert, Euler and Bernouilli’s debate on the problem of the vibrating string, all the way up to Fourier, Ohm and Helmholtz (Kittler 2000). NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’ 227

singular qualities of the ‘other music’ are determined by those elements that cannot be processed or represented beyond the moment of their unfolding in time, because they are produced in and belong to the transient moment of physical filtering itself. As it brings forth the signal in the first place, this moment cannot be processed, represented or analysed beyond its physical unfolding in time. It perpetually escapes our grasp. Hence, exactly because the noise of material channels never ceases not to inscribe itself on the physical signals they transmit and produce, and exactly because this noise and randomness continuously escape our analytical grasp, their non-periodicity and transience mark the ‘other music.’ As part of the acoustical flow of the present, they remain fundamentally inaccessible and irrepresentable. Noise defines the singular temporal and spectral complexity of all sonic events. Even the most fine- grained digital sieve cannot entirely process things that “are not written in natural, but in real numbers,” because with each transduction from physical sound to digital data and back into physical sound, the noises and distortions produced at the many moments of analogue and digital filtering rear their irrepresentable and irrepressible head and new, singular sounds come into their own. At the end of this recording and reproduction process, the last filter, the listener’s brain, “transforms,” as Serres puts it, “hard into soft” (1995: 115). It transforms physical sounds into emphatic music and along that process, as Brian Eno describes in his famous talk on “The Studio as Compositional Tool” in 1979, exactly those “details that weren't intended by the composer or the musicians” are what listeners “become very fond of” (Eno 1983). Exactly the random and uncontrollable sonic singularities that parasitic filtering operations inevitably add to all technological sounds are what draws listeners to the ‘other music.’ Regardless of the method of production, recording, storage, transmission or playback, musical sounds that come out of speakers or headphones are as physical and non-symbolic as any other sound. Their impact does not rely on the ideal purity of perfectly delineated representational models or codes, but on the sheer physical presence of sound waves that are always inherently more complex and contingent than any symbolic representation. This is why Kittler argues that sound media 228 NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’

do not represent nor reproduce the so-called input signal, but produce “unforeseeable, unthinkable, unimaginable acoustic events” (2013a: 40). Singularly marked by the traces of physical filters that do neither belong to the original input nor to the idealised representation of the signal, technologically produced acoustic events are unforeseeable in the sense that they showcase the same amount of temporal contingency as any other sound; unthinkable in the sense that they do not result from any rational compositional process; and unimaginable in the sense that they do not require the creative impetus of human subjects to impact the listener. The transience of acoustic events and the singularity of sound have been defining features of the inherent temporality of music since time immemorial. However, whilst technological sound reproduction enabled the repetition of these sonic singularities over and over again, the primary filtering operations that underpin all recording, manipulation, transmission and production of sound signals remain in the dark and beyond our reach. Contributing to the singular noise of sound reproduction, the audible traces of these operations highlight the fundamental irrepresentability of their moment of production. If ever so subtly, instead of enabling greater control over musical sound, every cut of each physical filter changes the output in ways that cannot be captured or represented by any symbolic logic. “The event,” I cited Derrida in Chapter Four, “is something that vertically befalls me when I didn’t see it coming” (2007: 451-452). The ‘other music’ is a name for acoustic events that vertically befall a listener when he or she did not hear it coming. It designates those moments when the inconceivable presence of the Real cracks through the surface, as the sonic traces of the transient events produced by technological sound (re)production physically resonate in the receptive ears of human listeners. At these moments, the randomness of noise constitutes the resonance of a ‘darker presence’ of the irrepresentable Real in the music of the media age. It heightens the Real of media technological sounds, not by providing access to it, but by bringing its irrepresentability into ever sharper focus. Physically unfolding in the acoustic present, but discursively connected to the past, the traces of the transient moment of technological filtering haunt the ‘other music.’ Every time a record runs, unforeseeable, unthinkable, unimaginable sounds flow seamlessly from speakers and headphones, NOISE RESONANCE | THE SOUND OF AN ‘OTHER MUSIC’ 229

singularly unfolding in acoustic time and space to confront the listener with his or her own transient being.

NOISE RESONANCE | EPILOGUE 231

Epilogue: Listening for the event | on the possibility of an ‘other music’

6.1 Locating the ‘other music’

Throughout this thesis, I wrote about phonographs and gramophones, magnetic tape recording and analogue-to-digital converters, dual-ended noise reduction and dither, sine waves and Dirac impulses; in short, about chains of recording technologies that affect the sound of their output with each irrepresentable technological filtering operation. Only toward the end, this analysis of the way technology shapes the sound of the media age and resonates with listeners returned to what got me started on these issues in the first place: music. For a thesis that started out as a project about the role of noise in recorded music, that last term has been notably scarce. Nonetheless, I am convinced that the media specific analysis of the role of noise and distortion in sound recording technology developed over the course of these five chapters establishes a theoretical point of departure on the basis of which it will be possible to further the conceptual understanding of the way technologically (re)produced music resonates with listeners in the age of technical media. This is why the final chapter concluded with Kittler's intriguing notion of the ‘other music.’ In the way I read Kittler’s work, the appearance of this concept in his early writings (most notably in “The God of Ears”) and occasionally throughout his oeuvre points to a tentative conceptualisation of the status of sound and music in relation to technical media that has been a constant, but understudied force in his thinking, which never developed into a fully formed argument. For more than 30 years, he told Frank M. Raddatz in a conversation published in 2012, Kittler planned to describe 232 NOISE RESONANCE | EPILOGUE

how modern [neuzeitliche] music, with the invention of equal temperament and the development of orchestral music, sonata and symphony with it, slowly turned into the kind of music that excited me: Wagner, Mahler, Stravinsky. And how, due to technological modifications and innovations, from this ecstatic late Romantic music, there arose the British invasion—the Beatles, Rolling Stones, , Pink Floyd and all the other Brit-groups (2012a: 62).110

Such a project, would he have ever pursuit it, might have resulted in a book on the ‘other music,’ but it was eventually abandoned when he, as he concludes these remarks, “encountered the ancient Greek” and began working on what would become his final, largely unfinished, multi-volume project: Musik und Mathematik—a tetralogy in which he set out to rewrite Heidegger’s Seinsgeschichte on the basis of the conjoined history of music and mathematics since ancient Greece. Perhaps, the final of the four planned instalments of this grandiose project (tentatively entitled Turing Zeit) would have returned to the issue of the ‘other music’; and perhaps the eventual publication in his Collected Writings of the notes, lectures and interviews currently kept at the German Literature Archive in Marbach will shed some light on this part of the project.111 In the meantime, however, the argument put forward in this thesis

110 “Mein Plan war zu erzählen, wie die neuzeitliche Musik mit der Erfindung der gleichmäßigen Temperatur und damit der Orchestermusik, der Sonate und der Sinfonie langsam bis zu den Musiken führt, die mich begeistert haben: Wagner, Mahler, Strawinsky. Und wie aus dieser ekstatisch spatromantischen̈ Musik durch technische Umstellungen und Innovationen die British Invasion hervorgegangen ist, also die Beatles, Rolling Stones, The Who, Pink Floyd und all die anderen Brit-groups.” 111 According to the website of Wilhelm Fink Verlag, the notes, lectures and interviews on volume III and IV of Musik und Mathematik will be published as volume 11 of the monographs [Monographien] section of Kittler’s Collected Writings (Gesammelte Schriften) ("Friedrich Kittler. Gesammelte Schriften" 2016). NOISE RESONANCE | EPILOGUE 233

is an attempt to turn the notion of the ‘other music’ into a conceptual tool that might help us understand how the music of our time takes shape in between the irrepresentable operations of technical media and the receptive ears of human listeners. Shaped by the filtering operations of technological sound reproduction, the output signal of the reproduction chain triggers the noise resonance between reproduced sounds and its listeners. Characterised by this noise resonance that originates in the irrepresentable moment of technological filtering, the possibility of the ‘other music’ must be understood as a musical modality that defines our contemporary musical culture. Hence, both in conclusion of this thesis and explicitly looking forward toward the possible continuation and deepening of the issues it addresses, I want to ask: what is the other music? Can we hear it? And if so, where?

6.2 Music hiding in hardware

Given my strong emphasis on the friction between ideas of perfect fidelity and complete technological reproduction on the one hand and the randomness introduced by the physical cuts of technical filters on the other, one might assume that the ‘other music’ cannot be heard at all, because the irrepresentable moments of physical filtering that produce the transience of reproduced sound cannot be perceived as such. Constituting the irrepressible Real of reproduced sound, these moments only reach our conscious perception through the filtering operations of the Imaginary and the Symbolic, which, as I quoted Serres in Chapter Four, are “filtering a meaning, creating a meaning” out of the continuous fuzziness of the Real (1982b: 185). Such a fundamental impossibility to hear the ‘other music’ in itself could be a first possible answer to the question were it might be (or not be) located. If one were to pursue this suggestion, it could be argued that the ‘other music’ hides in the hardware of sound technologies, in the many transmission channels and filtering circuits in between sender and receiver where the noise of the Real interferes with the transmitted signal but disappears once it is stored in the grooves of vinyl records, pits of CDs and the magnetised surfaces of tapes or hard drives through which it becomes 234 NOISE RESONANCE | EPILOGUE

repeatable over and over and over again. Kittler might have recognised the potentiality of an ‘other music’ in the sounds of the Beatles, the Rolling Stones and, above all, Pink Floyd, but he did not hear it, as the sounding music itself can only infinitely tend toward the ‘other music.’ Similarly, he might have recognised this potential in the circuitry of the synthesiser he built (or to anglicise a more suitable German word, ‘bastled’) in the late 1970's, which now rests in the archive in Marbach. Piecing together an electronic , Kittler perhaps realised for the first time that technical media that produce, manipulate and reproduce physical sounds all by themselves introduce what he would come to call the ‘other music’ into the basic grid or mainframe of our musical culture.112 Or perhaps—a second suggestion—the ‘other music’ does not hide in the circuitry, but cannot be heard by human ears either. Maybe the sonic singularities of the ‘other music’ can only be recognised by devices that operate on the basis of the same technical filters that were applied by the machines that produced these sounds in the first place. Such is the case with popular smartphone applications like Shazam, designed to ‘recognise’ songs that users record and upload with their smartphone. As I argue more extensively elsewhere, no music or musical parameters are part of Shazam’s operations: the app processes physical sound waves that are recorded by a microphone, turned into digital data, sent to a server, compared with samples in a database, and returned to the user as an artist’s name and a song title.113 Shazam does not ‘hear’ or listen to any music, but analyses data that are fed back (via computer hardware and interface software) to the user. Turned into binary data, the music becomes coded information that can be analysed, synthesised and resynthesised like any other data set, without being dependent on any meaning beyond the logic of binary representations and digital algorithms. What Shazam deals

112 Many thanks to Moritz Hiller for suggesting this idea to me during the Princeton-Weimar Summer School for Media Studies, Princeton 2016. Significantly, in support of the argument that the ‘other music’ cannot be heard, but only hides in the circuitry of technical (sound) media, Kittler’s synthesiser currently cannot produce any sound; and it is unclear whether it ever actually did ("Apparatus Operandi1" 2012; Gringmuth 2012; Steinfeld 2011). 113 For this analysis of Shazam in relation to Kittler’s concept of the ‘other music’ see Kromhout, 2015. NOISE RESONANCE | EPILOGUE 235

with is not addressed to human ears at all. Only at the very last step, when the result of Shazam’s data processing returns to the user, the waves (re)gain their cultural meaning as musical sound. In our technology-saturated world, we are surrounded by devices that shape our sonic environment and listen for the result in ways our ears can never achieve; devices that cut out the biological middleman. Without media that record, store, transmit, produce and manipulate acoustic signals, the sounds of our age would simply not exist, but most of the filtering processes that produce them occur well before the signals reach our ears. We analyse and model these operations and build machines to execute them, but as we cannot perceive them as such, only picking up acoustic traces after the fact, the transient moments of this execution itself are lost to our ears. When it comes to identifying such moments, Shazam is faster, more accurate, and more reliable than any human agent. As co-founder Avery Li- Chun Wang wrote in 2003: Shazam’s “algorithm can pick the correct [one of different versions of a song, MK] even if they are virtually indistinguishable by the human ear” (2003: 7). The proposition that digital audio media have surpassed human hearing, is therefore not only theoretical, but is experienced every day by millions of users. If this is indeed all there is to it, the noise resonance of sound reproduction does not apply to human ears at all. This is what Wolfgang Ernst’s media archaeological reading of sonic media suggests. For Ernst, human hearing and interpretation have become entirely secondary to the technological processes that undercut and surpass and thus determine and shape this hearing and interpretation in the first place. However, keeping with my analysis of “The God of Ears” in Chapter Five, Section 5.4b and Mark Hansen’s suggested ‘rehumanisation’ of Kittler’s intellectual legacy, I claim that the concept of the ‘other music’ runs deeper than these media archaeological accounts. It does not only apply to a class of technological processes to which human senses do not have access. Although the ‘other music’ might be elusive, hard to pin down and contingent by its very nature, I do not subscribe to a reading that renders it fundamentally inaccessible or inaudible. Instead, I suggest that the sonic traces creating the possibility for the ‘other music’ to appear 236 NOISE RESONANCE | EPILOGUE

produce a musical modality in the sense that, beyond the realm of technologically processed, physical sound waves that the algorithms of Shazam deal with, the experience of the ‘other music’ penetrates the cultural sphere of human signification.

6.3 I hear a new world

Kittler heard the ‘other music’ or at least recognised its contours in the same set of examples over and over again: “from Wagner to Hendrix,” he said in 2008 in “Preparing the Arrival of the Gods,” “from Hendrix to Waters, it is the same music” (2015b: 104). Anticipated by the ‘acoustic effects’ of Wagner’s Gesamtkunstwerken, which I described in Chapter Five, hi-fi stereo recordings of the ringing feedback of Jimi Hendrix’s electric guitar or the cosmic echoes on early Pink Floyd records are defined by the logic of filtering that underpins the age of technical media. As Kittler describes the sounds that began to emerge from jukeboxes in the 1960s: “an echo that lasts for ten seconds and keeps returning,” a cosmic echo, “cannot be implemented anywhere on earth. One needs magnetic tape to play [with] the cosmos—here and now” (2005: 25).114 The operations of technical media produce sounds from beyond. Technologies that initially only reproduced acoustic events subsequently began to produce sounds that break with the basic laws of physics and could have only originated in these machines. As the transient moments of physical filtering that produce these sounds are lost to our ears, the ‘other music’ appears when their sonic traces resonate this transience in our ears. They resonated, for instance, in the ears of sixties record producer Joe Meek, who met a tragic end when he shot his landlady and himself only a few months prior to the Summer of Love in February 1967. Seven years earlier, in 1960, Meek recorded an album that was only released in full in 1991. It is called I Hear A New World and consists of thirty-three minutes of spaced-out sound effects, excessive reverb, weirdly pitched vocals and

114 “Ein Echo, das zehn Sekunden lang anhält und immer wiederkehrt […]. Nirgendwo auf Erden lasst sich das implementieren, dazu braucht man ein Tonband, dann aber können wir den Weltraum spielen - hier und jetzt.” NOISE RESONANCE | EPILOGUE 237

heavily treated instrumentation: a close sonic equivalent to science fiction. “I hear a new world calling me,” the singer sings on the title track, “How can I tell them what's in store for me?” (Meek 1991). Meek heard the calling of another world ringing with the music of the future and the music of space, and used all the technology available to him to put it to tape. This is music that could have only been created by means of that technology; and it still sounds as simultaneously real and otherworldly—in time and out-of- time—fifty years after it was first recorded. Does this mean that, in Eno’s words, using “the studio as a compositional tool” is the key toward creating an ‘other music’ (Eno 1983)? Is all recorded music thereby classifiable as ‘other music’? If so, ‘other music’ would be nothing but a different name for what is commonly called ‘sound’; a term that, as musicologist Paul Théberge writes, has taken on a peculiar material character that cannot be separated either from the 'music’ or, more importantly, from the sound recording as the dominant medium of reproduction” (1997: 191). However, this is not what I want to argue either. What Kittler may have heard while listening to Pink Floyd and Jimi Hendrix and what Joe Meek may have heard and tried to capture as well were the traces of a musical event, the possibilities of which came into being over the course of the development of sound recording technology but which are not a feature of sound reproduction per se. “The event,” says Derrida, “is that which goes very quickly; there can be an event only when it’s not expected” (2007: 443). The possibility of the occurrence of the event called the ‘other music’ relies on a shift of musical agency from human agents toward technical media that produce and reproduce non-symbolic, physical signals. These signals are produced in the physical real, but their moment of production, the moment of filtering, remains fundamentally inaccessible; and because these moments of production are purely transient and thus constitute the sonic Real of technological sound, they remain irrepresentable as such. The random, noisy, transient traces of these moments, however, sonically resonate with listeners and can sometimes suddenly (like a shock, like a strike of lighting) hit upon unexpecting ears. Some of these traces are hardwired in the information stored on the recording medium, but they do not originate in the recording process only. 238 NOISE RESONANCE | EPILOGUE

They are produced throughout the entirety of the recording chain. From first input to final output, from the first sound played and recorded in some recording studio to the moment of its playback through speakers or headphones directed at the listener. The entire chain of technological sound (re)production produces what Kittler calls “a single and positive feedback between sound and the listener’s ear” (2015a: 13). Within this feedback chain, sonic traces of physical filtering operations can produce the singularity of the present in the repetition of the past, triggering our aesthetic sense in unexpected ways, ways that draw us in and hit us in the gut. As is well known, feedback can be instrumentalised and put to good use, but only up to a certain level and only with the risk of, as Serres writes of Orpheus’ masking strategy cited in Chapter Two, falling into noise: feedback can also cause a short circuit breakdown and ear-splitting cacophony.115 Hence, creating the possibility of the ‘other music’ to appear might imply letting go of control, taking the risk of exploring the unknown and let machines do the work.

6.4 Hearing an ‘other music’

In the early 1950s, Belgian composer Karel Goeyvaerts and his German colleague kept a correspondence in which they discussed the possibility of a music made from pure sine waves. They wanted to produce, Stockhausen recalls, “pure, controllable sounds without the subjective emotional influence of ‘interpreters’” (1971: 649). As musicologist Richard Toop describes in an article on “Stockhausen and the Sine-Wave,” both composers were devout Catholics and considered the “notion of purity” so perfectly captured by the ideal sine wave to be “not just musical, but theological” (1979: 383). Without disrupting attacks and decays, sine waves and Fourier series seem to approximate divine perfection; a representation of heavenly harmony. Although Stockhausen, as Toop concludes on the basis of the correspondence, remained

115 “Jimi Hendrix,” Kittler writes in “Lightning and Series – Event and Thunder,” “only needed to hold his electric guitar as close as possible to his good old nonlinear Marshall amplifier until guitar and amplifier exploded in an endless rumble of thunder” (2006a: 71). NOISE RESONANCE | EPILOGUE 239

ambivalent toward the idea of composing with nothing but sine waves and eventually sought to combine them with other electronic and instrumental sounds, Goeyvaerts stuck to it and developed the idea of “dead sounds”: “sounds which would have absolutely no unpredictable ‘inner life,’ but would be identical at any moment in time, and therefore detached from time itself” (1979: 386). As musicologist Herman Sabbe explains it, by technologically erasing “all these impulses” introduced by the attack and decay of sound, Goeyvaerts tried to create music that could “lift the sense of time” (1994: 76). Using magnetic tape to create the most static sounds possible, the “dead tones” of Nr. 4 met Dode Tonen (N°.4 with Dead Tones) (1952) minimalises sonic transience to approximate the characterless immortality of ideal sine waves (Goeyvaerts 1998). At the outset of the age of magnetic recording, in the years leading up to what Kittler calls the “media explosion” of the 1960s that produced the spaced-out sounds of Joe Meek’s ‘new world,’ Goeyvaerts’ composition for dead tones constitutes what Sabbe calls “the most radical pretension of totality and positivity ever” (Sabbe 2005: 244). This attempt to make music with ideal filters can be seen as a final, inevitably doomed, attempt to wield total control over musical sound by transcending the material basis of sound production and media technology. Goeyvaerts’ theological program notwithstanding, I argue it is not in how the piece succeeds in these goals, but in how it fails to achieve the eternal stasis of the domain of the ideal filter that it hints at the promise of a new sonic world. Infinitely tending toward one extreme of the uncertainty principle, Goeyvaerts’ dead tones aspire for the clarity suggested by the mathematical operations of Fourier analysis; a clarity that, as I argued in Chapter Four, goes beyond the transience of earthly life altogether and indeed reaches for the heavens, but, as the argument continued in Chapter Five, a clarity that also brings into ever sharper focus the impossibility to achieve eternal stasis and heavenly purity. The traces of the material production of sound signals, the moment of physical filtering, always already disrupt the purity; if only because sounds simply cannot go on for ever, but have to start and stop, thereby introducing, as I 240 NOISE RESONANCE | EPILOGUE

quoted Wiener in Chapter Three, alterations of the frequency composition that “may be small,” but are still “very real” (1976: 544-545). Exactly this tension between clarity and diffusion, between sharpness and fuzziness, between the ideal of perfect reproduction and moments of physical filtering, create a possibility of the ‘other music’ to appear: sounds that shimmer in the moonlight, but, as moonlight itself is only a reflection of the light of sun, also beam further into the dark universe that lies beyond. There is no perfect formula for striking this possibility called ‘other music,’ no unequivocal answer to Serres’ question as to “how much noise is necessary?” On the other extreme of the uncertainty principle, as far removed from Goeyvaerts’ dead tones as possible, the vast musical output of Japanese noise artists like Masami Akita, alias Merzbow, for example on the album Pulse Demon, consists of a succession of near infinitesimal impulses that tend toward pure white noise, creating a continuous sonic difference that overloads our senses with acoustic information (Merzbow 1998). In between these sonic extremes that reach for the heavens by electronically approximating (but never actually producing) pure sine waves, or create dense and highly complex sound spectra that approximate (but never actually are) pure white noise, a continuum of noises and distortions cling to sound signals that travel along the reproduction chain. Subtle and near inaudible or harsh and almost overtaking the signal itself, this noise of sound reproduction shapes the music in ways that cannot be completely controlled nor predicted. Six years after the otherworldly sounds of Joe Meek’s ‘new world’ came the four Californian minutes of “Good Vibrations” that Brian Wilson pieced together out of more than ninety hours of tape recordings in 1966 (Beach Boys 1994). Kittler recognized the ‘other music’ in the dreamy sound effects of Pink Floyd’s aptly names “Echoes” in 1971 and the hi- fidelity sonic splendour of their “Great Gig In The Sky” in 1973, but around the same time, on the other side of the Atlantic, Stevie Wonder also began programming otherworldly sounds on his synthesisers, which he used to great effect on albums like Talking Book and Innervisions (Pink Floyd 2001a; Pink Floyd 2001b; Wonder 1986; Wonder 1990). Toward the end of the decade, the possibilities for such techno-sonic explorations had become almost limitless and people like Brian Eno explored the full potential of the NOISE RESONANCE | EPILOGUE 241

“studio as compositional tool,” for instance on David Bowie’s Low in 1977 or his own Music For Airports in 1978 (Eno 1990; Bowie 1999). By the 1980s, the Fairlight synthesiser sampled and sequenced digital sounds on albums like Kate Bush’s Hounds of Love from 1985 (1995). The Synclavier at the start of Michael Jackson’s “Beat It,” which sounded a few years earlier in 1982, might as well be regarded as the opening for the digital age (2008). The music in these examples takes the technological possibility of creating unforeseeable, unthinkable, unimaginable acoustic events to its logical conclusion by showcasing rich spectral sonorities that aspire, like the heartbeat on The Dark Side of the Moon, to ring forever. They are presenced over and over again in all their spectral clarity every time the record spins. As with Goeyvaerts’ dead tones, however, it is exactly due to the contingent traces of their material production by irrepresentable filtering operations and thus to the extent in which they do not belong to the present but also signify the past, that these sounds can trigger the emergence of an ‘other music.’ On the other side of the spectrum and in contrast to the sonic clarity of pop musical splendour, Lou Reed explored the noise spectrum created by the accidental, contingent, erroneous processes at the heart of technological filtering operations on the hour-hour long feedback assault called Metal Machine Music, released in 1975. In its wake, in the late seventies and early eighties, bands like Throbbing Gristle and Einstürzende Neubauten began to use and record all the hums, buzzes, feedback, overdrive, distortion and tape saturation produced of technological sound reproduction on albums like The Second Annual Report and Kollaps (Throbbing Gristle 2011; Einstürzende Neubauten 2003). 116 Hip-hop sampling, the synthetic sounds of electronic dance music, distorted rock guitars, lo-fi recording strategies, the musical use of digital glitches, the Autotune-effect: these are all strategies that take the irrepresentable cut of technical filters as the starting point for sonic invention, thereby producing

116 I already mentioned my bachelor thesis on the work of Einstürzende Neubauten in the introduction (Kromhout 2006). Furthermore, I discussed the relation between the role of noise and its relation to sound reproduction technology in the work of Throbbing Gristle in Kromhout 2011. 242 NOISE RESONANCE | EPILOGUE

music that consistently and explicitly negates the myth of perfect fidelity and sonically emphasises the transience and finitude of all signals.117 However, like the splendour of hi-fi recordings evokes its own opposite, exactly because of the sonic singularity of sounds shaped by the irrepresentable moments of physical filtering, their negation of representational clarity also produces something else: each time these transient traces of filtering operations are presenced by loudspeakers and repeated over and over again, they gain in significance and begin to shine all the brighter. It would therefore be wrong to think of these examples as inherent opposites that tend toward both extremes of the uncertainty principle—the clarity of the sine wave and the instantaneity of the Dirac impulse, the thunder or the lightning. Rather, I suggest they represent two sides of the same coin, located at different points in the noise continuum. What they show together is that the possibility of a noise resonance of sound reproduction to create the sonic conditions for the ‘other music’ to appear is not restricted to specific musical genres. It can, but not necessarily does, appear in all technologically (re)produced music. “The song sleeps in the machine,” sings Einstürzende Neubauten frontman Blixa Bargeld (1997: 126).118 This song is Kittler’s “song from beyond mankind” and it can be awoken each time someone presses play. Its conditions appear when the contingencies of technical media meet the organised sound we call music, but they are perhaps most apparent, most nakedly unconcealed in some forms of contemporary . By virtue of digital sound technology, providing what Kittler calls the first viable “language for sound,” this music leaves the connection with non- mediated or so-called natural or acoustic sound behind entirely and wakes the vast reservoir of songs that sleep in the machine (2013a: 40). The grainy sound textures construed out of vinyl samples and analogue synthesisers on Music Has the Right to Children by British duo

117 On the aesthetics of sampling, see De La Motte-Haber 2000 or Rodgers 2003. On the affective sound of the machine in electronic music production, see Maresch, 2003. On glitch music, see Hainge, 2007. On guitar distortion, see Poss, 1998. On the aesthetics of lo fi recording methods, see Kromhout, 2012. On the use of auto-tune, see Marshall, 2014. 118 “Das Lied schläft in der Maschine.” NOISE RESONANCE | EPILOGUE 243

Boards of Canada, for example, invokes the clear melancholic pull of pastness courtesy of the cuts of technical filters (Boards of Canada 1998). All characteristics of these sounds evoke the many channels they passed before flowing from the speakers. At the same time, however, they clearly resonate in the here and now, with the full sonic depth and spectral richness of physical, singular sounds. One could argue that these sounds speak of nothing but sound, but it might be more accurate to say they speak of filters. By emphasising there is no way to tell what these sounds are or how they came to be, they speak of the logic of filtering that invokes the still darker presence of the media age. Hence, the message of the synthesiser is and is not the synthesiser. In the dense, scattered and often heavily distorted sound worlds on the recent album Mutant by Venezuelan musician Arca or the fittingly titled Sirens by Chilean-American artist Nicolas Jaar, one hears in the most literal sense of the word, the traces of the entire analogue and digital circuitry that produced this music (Arca 2015; Jaar 2016). This music of digital ticks and glitches, low basses and whirling clouds of synthesiser chords consists entirely of the non-symbolic sound of technological (re)production, shaped by the noisy traces of irrepresentable filtering operations. Often drenched in layers of heavy reverb that infuse the listener with a sense of sonic dread, such music is as much borne from the complete malleability of sound as it is borne from the fundamental irrepresentability of its moment of production. It therefore continuously brings into sharper focus, as all technical media do, that which is irrepresentable and slips from our control.119 Ultimately, oscillating between the dream of perpetuating the magic and halt time and the attempt to create a new and unpredictable sonic presence—between uncanny sine-like purity and the noisy materiality of transmission channels—the noise resonance of the ‘other music’ thereby emphasises that the future remains, as Kittler has Roger Waters say at the close of “The God Of Ears,” “in the lap of the fucking gods” (2015a: 16). As a modality of technologically (re)produced music, to which some music, some musicians and some composers aspire, the noise

119 On the concept of sonic dread, see Goodman 2010. 244 NOISE RESONANCE | EPILOGUE

resonance of sound reproduction brings the event called the ‘other music’ into being. By listening for its resonance, being open for the potentiality of lightning to strike and something to stick out of the continuous flow or stream of background music that surrounds us every day, I hear a new world calling me.

NOISE RESONANCE | CONCLUSION 245

Conclusion: Noise resonances | sound in the age of technical media

When, over the years, someone asked me what this thesis was about, I used to answer it was set to analyse and revaluate the role of noise in contemporary musical practices. Because the growing importance of noise in our musical culture on the one hand and the emergence of technological sound reproduction on the other, occurred more or less simultaneously, it seems no more than logical to assume that the key to understand this proliferation of noise practices lies with the machines that increasingly came to dominate musical creativity over the past hundred and forty years. It was this hypothesis that first drew me to the work of Friedrich Kittler as a thinker who, perhaps more than anyone, conceptualised the relation between the rise of technical media, the growing importance of noise and the emergence of new types of sound, new kinds of music and, above all, a new kind of musical sensibility. He called this new sensibility the ‘other music’—the music of the media age. Although, in recent years, the body of literature on noise has grown substantially, most analyses that specifically deal with its role in music continue to classify noise as a transgressive or subversive force, something that might be sonically or musically relevant but only in contrast to the presumed clarity and purity of musical sound. Often, these interpretations take their cue from the conceptualisation of noise in information theory as developed in Claude Shannon’s mid-twentieth century “Mathematical Theory of Communication.” On the basis of information theory, it can be argued that what is classified as noise on first hearing stops being noise as soon as it is incorporated in the system of communication and henceforth turned into information. Such an essentially dialectic interpretation of the relation between noise and information, noise and signal, noise and sound or noise and music for instance characterises Jacques Attali’s seminal book 246 NOISE RESONANCE | CONCLUSION

on the history of noise. After it forces its way in from the outside and disrupts or renews the (musical) status quo, Attali argues, noise is disciplined into coercion, incorporated in the system and turned into music (Attali 2003). This is also the basis for Paul Hegarty’s more recent argument that all musical noise practices inevitably lose their subversive potential as noise because a process of normalisation strips them of all transgressive connotations (2008: 146).120 Entirely defined by its musical and cultural context, it follows from these arguments that noise, as Greg Hainge puts it, “in and of itself is nothing, for it arises only in the relational process through which the world and its object express themselves in an infinite number of possible relations, assemblages or expressive forms” (2013: 15). According to Hainge’s “ontology of noise,” as a fundamentally relational phenomenon that emerges in the context of all creative acts, noise is inherently elusive and “remain[s] out of reach” (2013: 273). It disappears as soon as you put your finger on it. Kittler’s work on sound and music, on the other hand, considers noise as a, or perhaps even the fundamental aspect of contemporary music and media alike. As with the aforementioned examples, Kittler’s concept of noise is heavily informed by information theory. However, as N. Katherine Hayles notes in How We Became Posthuman, “Shannon himself frequently cautioned that [his] theory was meant to apply only to certain technical situations” (1999: 19). Some of the conceptual overload of the concept of noise might indeed be due to the proliferation into all corners of the academy of information theoretical analyses that fail to take the media technological specificity of its terms into account. Kittler’s approach, however, exactly takes “certain technical situations” as its starting point. His media specific discourse analysis always takes into account the specificity of technological hardware as the material basis for the operations of technical media. Hence, as an essential feature of both music and media, a media specific concept of noise provides a perfect basis for rethinking their relation.

120 “Noise,” writes Hegarty, “will always fail, as noise at least,” because “any success means it has failed” (2008: 146). NOISE RESONANCE | CONCLUSION 247

In support of this argument, Chapter One shows how the history of sound reproduction from phonograph to digital sound file is in many ways a history of the continuous introduction of and fight against the noises, distortions and interferences that stick to the output of recording and reproduction media. On the one hand, this history emphasises how the relation between the development of new technologies and the appearance of new noises and distortions has been one of mutual interaction. On the other hand, however, it shows how the ways in which inventors, engineers and musicians continuously attempt to prevent, reduce or eliminate these noises have been framed by a myth of perfect fidelity. Assuming perfect sound reproduction and a one-on-one correspondence between input and output, this myth is based on what Jonathan Sterne calls the idea of the “vanishing mediator”: a medium that ensures the complete reduction or removal of its own influence on the reproduced signal. This myth of perfect fidelity began to take hold in the earliest days of recording technology with the conceptual separation between sounds that are internal and noises that are external to the recording. It subsequently consolidated over the course of the ensuing decades, as the concept of noise itself changed from a primarily sonic concept in acoustics and music theory in the nineteenth century into a physical concept influenced by communication engineering in the early twentieth century and ultimately into a communicational concept developed by information theory in the 1930s and 1940s. Over the course of this redefinition, I argue, the idealised notions of noise and signal that support the myth of perfect fidelity and, by extension, the idea of noise as an unwanted, external and disruptive force, took hold. The framework supporting this notion of noise as the seemingly unambiguous antithesis to pure and clear signal transmission, to be removed, eliminated or reduced at all cost, is what I call the conceptual logic of noise reduction. In Chapter Two, the media specific analyses of Dolby’s analogue dual-ended noise reduction systems and the addition of dither noise to digital sound reproductions further define and problematise this conceptual logic of noise reduction. In the final analysis, I argue, both Dolby’s companding procedure and the elimination of digital quantisation errors via dithering are ways to reinforce the suggestion of an inherent 248 NOISE RESONANCE | CONCLUSION

connection between input and output. Instead of signifying fundamental differences between analogue and digital recording methods, these strategies uphold the idea that the output of any reproduction process should always approximate the input as close as possible. This means that both analogue noise reduction and digital dithering confront the inherent limits of symbolic representation and technological reproduction. The actively created silence introduced by analogue noise reduction and the slight layer of random background noise created by dither are supposed to conceal what Bernhard Siegert identifies as the ‘rupture’ between the asymptotic idealisations represented by modern mathematical analysis and the physical signals they represent. This rupture between representation and represented and between reproduction and reproduced is exactly what the myth of perfect fidelity, supported by the conceptual logic of noise reduction, conceals. Although this logic assumes a perfectly unambiguous separation between noise and signal, such a separation implies a clearly definable and thus inherently limited concept of noise that cannot account for the fact that the operations of technical media always run into fundamental limitations. Contrasting this idealised separation between noise and signal upheld by the conceptual logic of noise reduction, I therefore introduce the alternative concept of a noise resonance of sound reproduction to conceptualise how the sound of technical media does not take shape despite of the noise of sound reproduction but exactly because of the way that this noise inherently affects and changes all output signals. In order to further problematises the conceptual logic of noise reduction and embed the concept of a noise resonance of sound reproduction in the historical development of the discourse on sound and sound reproduction, Chapter Three traces the discursive roots of the modern representation and reproduction of sound back to the early nineteenth century mathematical invention of Fourier analysis and the corresponding conceptualisation of the figure of the sine wave as the representation of a pure, single sound wave. Because the symbolic representation of physical sound as sets of endlessly repeating sine waves produced by Fourier analysis seem to empirically confirm ideals of musical harmony and regularity that go back all the way to the Pythagorean NOISE RESONANCE | CONCLUSION 249

Harmony of the Spheres, I suggest that its symbolic representation of a “world without noise” constitutes the modern, scientific origin of the ideals of infinite clarity and maximal purity that define the conceptual logic of noise reduction (Serres 2008: 126). Sine waves and Fourier analysis belong to a domain of entirely clear and pure signals; a domain from which all randomness, ambiguity and noise have been symbolically removed by perfectly seamless filters. I call this the domain of the ideal filter. Whereas sine waves symbolically represent infinite, periodic frequencies, their symbolic opposite are Dirac impulses: infinitesimally short, completely a-periodic, transient events. Both sine waves and Dirac impulses are idealisations suggesting a level of spectral clarity and temporal exactitude that no physical filtering operation can possibly achieve. In contrast to such idealisation and following the fundamental uncertainty principle at work in communication engineering and information theory, signals in the domain of physical filters are produced on the basis of a negotiation between the static noiseless purity of sine waves and the temporal exactitude of Dirac impulses—a negotiation between frequency and time. Although, as Kittler argues, exactly the symbolic idealisations courtesy of the domain of the ideal filter enabled the “clarity and sharpness” of mathematical analysis and technological (re)production, the operations of technical media in the domain of physical filters are always limited by this negotiation between time and frequency (2012b: 53). Whereas the purity and clarity represented by ideal filters requires a symbolic clean cut that leaves behind no traces, the physical cuts of technical filters are subject to the limitations of the uncertainty principle. As a consequence, physical filters always leave behind traces of their own operation in the form of transient noises and distortions added to the output signal. Hence, whereas the conceptual logic of noise reduction presupposes the clean cut of an ideal filter, the noise resonance of sound reproduction is based on the primacy of the physical cuts of technical filters. Further assessing the consequences of this primacy of physical filtering operations for the role of noise in technological sound reproduction, and ultimately, in music, required a shift from the media 250 NOISE RESONANCE | CONCLUSION

specific (or media archaeological) analyses conducted in Chapters One and Two to an analysis of the primary logic of filtering itself in Chapters Four and Five. Because the transient noises and distortions that are caused by the cuts of physical filtering operations shape the singular sound of music in the media age, I argue that the assessment of the logic that supports these filtering operations helps our understanding of the way the output of technical media continues to arrive in the ears and brains of human listeners as sound or, even more extraordinary, as music. The nature of the noise resonance of sound reproduction that shapes the relation between technologically (re)produced music and its listeners, I claim in Chapter Four, is inherently temporal. Because the logic of filtering emphasises how media in the domain of physical filters always (re)produce signals that extend in space and change over time, it draws attention to all the transient events that escape the clean cuts of ideal filters. In the domain of physical filters, any signal has a beginning and an end—an attack and decay—that cause what Norbert Wiener calls “small, but very real,” random alterations to its frequency spectrum (1976: 544- 545). Contrary to clean cuts that leave no trace whatsoever, a technological recording and reproduction chain must be understood, following Serres, as a series of parasites (1982b: 172). Each parasitic filter affects the characteristic noise of sound reproduction and changes the output of the chain. These changes are not external to the signal; they are not an intrusion or disruption. They are as much part of the reproduced sound as all the frequencies that pass through the channel unaffected. They are therefore part of the system (as information, signal and sound) but they also still count as noise: random, transient, unpredictable signals. As biophysicist Henri Atlan—whose work greatly influenced Serres—writes in “Noise as a Principle of Self-Organization”: even when “the effects of noise become events in the history of the system and its process of organization,” they nonetheless “remain […] effects of noise inasmuch as their occurrence was unforeseeable” (2011: 112). Caused by the physical filtering operations of technological sound reproduction, the spectral and temporal changes to the noise of sound reproduction are physical traces of NOISE RESONANCE | CONCLUSION 251

a journey over space and time, hardwired in the frequency composition of the output signal. These transient traces, these effects of the “noise of documentation and transduction” emphasise the multi-layered temporality of technological sound. (Link 2001: 34). They signify passed time and passing time. On the one hand, as an indexical trace of the operations of recording and reproduction technologies, they resonate the temporal irreversibility and finitude of physical signals. Signifying how we, as listeners, are always running out of time, they emphasise we can never achieve the immortality that Kittler associates with the Fourier domain. On the other hand, due to their fundamental irregularity and transience, these traces resonate with the continuous flow of time through the present. Signifying how we are always inside time, they tend toward the irrepresentable short energy discharge of a Dirac impulse and emphasise a sense of being, as Derrida puts it, infinite. Throughout the ages, music has always been a balancing act between such periodicity and non-periodicity, between change and repetition, redundancy and entropy, static states and transience. Although the singularity of the unfolding of sound in time has thus always been an important aspect of musical appeal, I suggest it became even more fundamental with the advance of technological sound reproduction.121 Because technical media create the possibility to endlessly repeat sound again and again and again, by virtue of this repetition the sheer transience of everything that escapes the symbolic bottleneck of music notation can be scrutinised and weighted. This is how the transient traces of physical filtering operations gained in significance. However, as I conclude in Chapter Five, the physical presence of these sounds unfolding in time also continuously emphasises how the moment of their production—the

121 In “Vers Une Musicologie Concrète. Bemerkungen zu Richard Voss,” Kittler remarks how physicists Richard F. Voss and John Clarke, by measuring the frequency of frequencies in different types of music, proof that the statistical distribution of frequencies in almost all music adds up to pink noise. This means that “Rauschen […], wenn ein Signal als Musik wahrnehmbar sein soll, muß in der Mitte zwischen Ordnung und Chaos liegen, also nicht weiß, sondern rosa sein” and that, on the basis of the statistical frequency of frequencies, “Stockhausens kühnste Experimente sich von Bachs Brandenburgischen Konzerten kaum unterscheiden” (1995b: 112). 252 NOISE RESONANCE | CONCLUSION

physical cuts of technical filters—still fundamentally escapes our control. Hence, noise does not remain “out of reach” because, as Hainge puts it, it is nothing “in and of itself” (2013: 273). The inevitable presence of noise, I argue, remains fundamentally irrepresentable because, following Kittler’s conceptualisation of sound recording, the physical filtering operations that produce the signal in the first place constitute the fundamentally irrepressible Lacanian Real itself. As the sonic traces of the cuts of physical filters, noise and distortion reveal how technical media do not generate ever greater clarity and sharpness, but always also produce what Carol White in her study of Heidegger calls the “incalculable, the unpredictable and unthinkable, which lies beyond our capacities to represent” (2005: 74). Noise does not remain out of reach because it is nothing. It remains out of reach because it is the basis for everything. “I do not know if talking of filters,” Serres writes in The Five Senses, “will help us understand how thunder, noise, the vibration of sound waves […] subtly become meaning” (2008: 115). Connecting the logic of filtering with the noise resonance of sound reproduction, I argue that talking of filters indeed helps us understand how sound technologies create what Kittler calls “unforeseeable, unthinkable, unimaginable acoustic events”— singular sound waves, fundamentally shaped by the filtering channels through which they travel. Resolutely doing away with the conceptual logic of noise reduction and the idea that sound recordings are incomplete reproductions of some original source, the noise resonance of sound reproduction emphasises how a logic of filtering produces the sonic Real of technological sound. It emphasises that technologically (re)produced sounds are no symbolic representations created by and for human subjects, but physically present, complex signals shaped by irrepresentable moments of technological filtering; and the more autonomous and advanced these filters become, the more their operations slip from our control. The transient noise of sound reproduction therefore emphasises the inaccessibility and irrepresentability of all technological sound. Produced by the autonomous operations of technical media, the sound of the ‘other music’ continuously escapes our grasp. This is what Kittler calls “pure media technology, pure control flow” (1995: 99). Hence, although the NOISE RESONANCE | CONCLUSION 253

notion of an ‘other music’ marks the end of all anthropocentric interpretations of music, talking of the way that filtering operations produce the noise resonance of sound reproduction nonetheless helps us understand the impact of the “media explosion of our day” and the music it created (Kittler 2015a: 16). This is why a project that set out to revaluate the role of noise in musical practices turned into an analysis of the way technological sound reproduction fundamentally changed the sound of music. As a sonic trace of fundamental filtering operations, the noise of the ‘other music’ resonates the irrepressible presence of the Real in the receptive ears and brains of listeners. As such, I argue, the noise resonance of sound reproduction constitutes an essential element of the sound of music in the media age.

NOISE RESONANCE | REFERENCES 255

References

Bibliography

Aldrich, Nika. “Dither Explained. An Explanation and Proof of the Benefit of Dither.” Cadenza Recording, qwest.net, 25 Apr. 2002, www.users.qwest.net/ ~volt42/cadenzarecording/DitherExplained.pdf. Accessed 20 Sep. 2013. Alton Everest, F. The Master Handbook of Acoustics. 4th ed., New York: McGraw- Hill, 2001. “Apparatus Operandi1: : Anatomie // der Synthesizer des Friedrich A. Kittler.” Institut für Musikwissenschaft und Medienwissenschaft, 11 Apr. 2012, www.medienwissenschaft.hu-berlin.de/de/medienwissenschaft/ medientheorien/projekte/think-analogue/apparatus-operandi-1. Accessed 5 Sep. 2015. Atlan, Henri. 2011. “Noise as a Principle of Self-Organization.” Selected Writings: On Self-Organization, Philosophy, Bioethics and Judaism, edited by Stefanos Geroulanos and Todd Meyers. 2nd ed., New York: Fordham University Press, 2011: 95-113. Attali, Jacques. Noise. The Political Economy of Music. Translated by Brian Massumi. Minneapolis: University of Minnesota Press, 2003. Ballard, Susan Patricia. “Out Of Order. Explorations In Digital Materiality.” Dissertation, University Of New South Wales, 2008, handle.unsw.edu.au/ 1959.4/42596. Accessed 15 Feb. 2013. Barad, Karen. Meeting the Universe Halfway. Quantum Physics and the Entanglement of Matter and Meaning. Durham: Duke University Press, 2007. Bargeld, Blixa. Text fur Einstürzende Neubauten/Text for Collapsing New Buildings. Berlin: Gestalten Verlag, 1997. Baumann, Max Peter. “Rauschen im Kopf.” Das Rauschen, edited by Sabine Sanio and Christian Scheib. Hofheim: Wolke Verlag, 1995: 27-42. Benschop, Ruth. “Memory Machines or Musical Instruments? Soundscapes, Recording Technologies and Reference.” International Journal of Cultural Studies, vol. 10, no. 4, 2007: 485-502. Bexte, Peter, and Werner Künzel. Allwissen und Absturz. Der Ursprung des Computers. Frankfurt am Main: Insel Verlag, 1993. “Bias.” Encyclopedia of Recorded Sound, 2nd ed., 2005. Bijsterveld, Karin. Mechanical Sound. Technology, Culture, and Public Problems of Noise in the Twentieth Century. Cambridge: MIT Press, 2008. Blamey, Peter John. “Sine Waves and Simple Acoustic Phenomena in Experimental Music with Special Reference to the Work of La Monte Young and Alvin 256 NOISE RESONANCE | REFERENCES

Lucier.” Dissertation, University of Western Sydney, 2008, handle.uws.edu.au:8081/1959.7/25659. Accessed 28 Apr. 2015. Blesser, Barry, and Linda-Ruth Salter. Spaces Speak, Are You Listening? Experiencing Aural Architecture. Cambridge: MIT Press, 2007. Borsche, Dahlia. “Geräusch, Musik, Wissenschaft. Eine Bestandaufnahme.” Geräusch – Das Andere der Musik. Untersuchungen an den Grenzen des Musikalischen, edited by Camille Hongler, et al. Bielefeld: Transcript Verlag, 2015: 33-46. Braun, Hans-Joachim. Introduction. Technology and the Production and Reproduction of Music in the 20th Century. Music and Technology in the Twentieth Century, edited by Hans-Joachim Braun. Baltimore: John Hopkins University Press, 2002: 9-32. Brech, Martha. “Rauschen: Zwischen Störung und Information.” Das Rauschen, edited by Sabine Sanio and Christian Scheib. Hofheim: Wolke Verlag, 1995: 99-107. Burke Hubbard, Barbara. The World According to Wavelets. The Story of a Mathematical Technique in the Making. Wellesley, MA: A K Peters, 1996. Čapek, Milič. Introduction. The Concepts of Space and Time. Their Structure and Their Development, edited by Milič Čapek. Dordrecht: D. Reidel Publishing Company, 1976: XVII-LVII. Chion, Michel. “Let's Have Done with the Notion of Noise.” Translated by James A. Steintrager. Differences, vol. 22, no. 2-3, 2011: 240-248. ———. “Silence in the Loudspeakers. Or – Why, with Dolby Sound in Films, It Is the Film Which Is Listening to Us.” Framework: The Journal of Cinema and Media, vol. 40, Apr. 1999: 106-110. ———. Audio-Vision. Sound on Screen. Translated by Claudia Gorbman. New York: Colombia University Press, 1994. Clark, Mark H. “Product Diversification.” Magnetic Recording. The First 100 Years, edited by Eric D. Daniel, et al. New York: The Institute of Electrical and Electronics Engineers, 1999: 92-109. Cohen, Leon. “The History of Noise (On the 100th Anniversary of Its Birth).” IEEE Signal Processing Magazine, vol. 22, issue 6, 2005: 20-45. Coleman, Mark. Playback. From the Victrola to MP3. 100 Years of Music, Machines, and Money. Cambridge, MA: Da Capo Press, 2003. “Compact Disc.” Encyclopedia of Recorded Sound, 2nd ed., 2005. Cowell, Henry. “The Joys of Noise.” Audio Culture. Readings in Modern Music, edited by Christoph Cox and Daniel Warner. New York: Continuum Group, 2004 (1929): 22-24. Darrigol, Olivier. “The Acoustic Origins of Harmonic Analysis.” Archive for History of Exact Sciences, vol. 61, no. 4, 2007: 343-424. “Decca Record Co.” Encyclopedia of Recorded Sound, 2nd ed., 2005. De La Motte-Haber, Helga. “Soundsampling: An Aesthetic Challenge.” Music and Technology in the Twentieth Century, edited by Hans-Joachim Braun. Baltimore: John Hopkins University Press, 2002: 199-206. Deleuze, Gilles. “On Leibniz.” Lectures by Gilles Deleuze, Feb. 2007, deleuzelectures.blogspot.nl/2007/02/on-leibniz.html. Accessed 14 Feb. 2014. NOISE RESONANCE | REFERENCES 257

Deleuze, Gilles. The Fold. Leibniz and the Baroque. New York: Continuum, 2006. Derrida, Jacques. Athens, Still Remains. The Photographs of Jean-François Bonhomme. Translated by Pascale-Anne Brault and Michael Naas. New York: Fordham University Press, 2010. ———. “A Certain Impossible Possibility of Saying the Event.” Critical Inquiry, vol. 33, no. 2, 2007: 441-461. ———. The Truth in Painting. Translated by Geoff Bennington and Ian McLeod. Chicago: University of Chicago Press, 1987. “Disc.” Encyclopedia of Recorded Sound, 2nd ed., 2005. Dolby B, C, and S Noise Reduction Systems: Making Cassettes Sound Better. Dolby Laboratories, 2001, www.dolby.com/uploadedFiles/English_(US)/ Professional/Technical_Library/Technologies/Dolby_A-type_NR/212_ Dolby_B,_C_and_S_Noise_Reduction_Systems.pdf. Accessed 3 Jan. 2013. Dolby® SR. Dolby® Spectral Recording. Dolby Laboratories. 1987, www.dolby.com/uploadedFiles/English_(US)/Professional/Technical_ Library/Technologies/Dolby_Spectral_Recording_(SR)/33_SpectralRecord ingPaper.pdf. Accessed: 14 Jan. 2013. Dolby, Ray M. Signal Compressors and Expanders. Patent 3.631.365, filed 1969, and issued 28 Dec. 1971, patentimages.storage.googleapis.com/pdfs/ US3631365.pdf. Accessed 19 Jan. 2016. Eno, Brian. “Pro Session: The Studio as Compositional Tool,” lecture delivered at New Music New York, the first New Music America Festival sponsored in 1979 by the Kitchen, excerpted by Howard Mandel, Down Beat 50, Jul. 1983 (Part 1) and 51, Aug. 1983 (Part 2), music.hyperreal.org/artists/ brian_eno/interviews/downbeat79.htm. Accessed: 29 Oct. 2010. Ernst, Wolfgang. Sonic Time Machines. Explicit Sound, Sirenic Voices, and Implicit Sonicity. Amsterdam: Amsterdam University Press, 2016. ———. Im Medium Erklingt die Zeit. Berlin: Kadmos, 2015. ———. Chronopoetik. Zeitwesen und Zeitgaben Technischer Medien. Berlin: Kadmos, 2012a. ———. Gleichursprünglichkeit. Zeitwesen und Zeitgegebenheit von Medien. Berlin: Kadmos, 2012b. ———. “‘... Else Loop Forever.’ The Untimeliness of Media, or: Does Technology Have a Sense of Ending?” Il Senso Della Fine, Università degli Studi di Urbino, Centro Internazionale di Semiotica e Linguistica, 10-12 Sep. 2009, Conference presentation, www.medienwissenschaft.hu-berlin.de/ medientheorien/downloads/publikationen/ernst-else-loop-forever.pdf. Accessed 30 Jan. 2015. Evens, Aden. Sound Ideas. Music, Machines, and Experience. Minneapolis: University of Minnesota Press, 2005. Farmelo, Graham. The Strangest Man. The Hidden Life of Paul Dirac, Mystic of the Atom. New York: Basic Books, 2009. Fine, Thomas. “The Dawn of Commercial Digital Recording.” ARSC Journal, vol. 39 no. 1, 2008: 1-17. Flusser, Vilém. Into the Universe of Technical Images. Translated by Nancy Ann Roth. Minneapolis: University of Minnesota Press, 2011. 258 NOISE RESONANCE | REFERENCES

Fourier, Jean Baptiste Joseph. The Analytical Theory of Heat. Translated by Alexander Freeman. Digitally Printed Version. New York: Cambridge University Press, 2009 (1822), dx.doi.org/10.1017/ CBO9780511693205. Accessed: 22 Dec. 2014. “Friedrich Kittler. Gesammelte Schriften. Herausgegeben von Martin Stingelin.” Wilhelm Fink Verlag, www.fink.de/katalog/reihe/friedrich _kittler_ gesammelte.html. Accessed 2 Sep. 2016. Gábor, Dennis. “Acoustical Quanta and the Theory of Hearing.” Nature, vol. 159, issue 4044, 1947: 591-594. Gammaitoni, Luca. “Stochastic Resonance and the Dithering Effect in Threshold Physical Systems.” Physical Review E, vol. 52, no. 5, 1995: 4691-4699. Gelatt, Roland. The Fabulous Phonograph. From Tin Foil to High Fidelity. New York: J.B. Lippincott Company, 1954. Goddard, Michael, et al, editors. Resonances. Noise and Contemporary Music. New York: Bloomsbury, 2013. Gödel, Kurt. “Static Interpretation of Space-Time with Einstein's Comment on It.” The Concepts of Space and Time. Their Structure and Their Development, edited by Milič Čapek. Dordrecht: D. Reidel Publishing Company, 1976: 455-461. Gooch, Beverly R. “Building on the Magnetophon.” Magnetic Recording. The First 100 Years, edited by Eric D. Daniel, et al. New York: The Institute of Electrical and Electronics Engineers, Inc. 1999: 72-91. Goodman, Steve. Sonic Warfare. Sound, Affect and the Ecology of Fear. Cambridge: MIT Press, 2010. Grattan-Guinness, I. Joseph Fourier 1768-183. A Survey of his Life and Work. Cambridge, MA: MIT Press, 1972. Gringmuth, Sven. “‘Aber Dieses Leben Reicht Dafür Nicht…’ Notizen zu ‘Götter und Schriften Rund Ums Mittelmeer—Symposion in Memoriam Friedrich Kittler.’” ZKM, Oct. 2012, blog.zkm.de/blog/einblick/aber-dieses-leben- reicht-dafur-nicht. Accessed 5 Sep. 2016. Gumbrecht, Hans Ulrich. Production of Presence. What Meaning Cannot Convey. Stanford: Stanford University Press, 2003. Hagood, Mack. “Quiet Comfort: Noise, Otherness, and the Mobile Production of Personal Space.” American Quarterly, vol. 63, no. 3, 2011: 573-589. Hainge, Greg. Noise Matters. Towards an Ontology of Noise. New York: Bloomsbury, 2013. ———. “Of Glitch and Men: The Place of the Human in the Successful Integration of Failure and Noise in the Digital Realm.” Communication Theory, vol. 17, issue 1, 2007: 26-42. Hansen, Mark B. N. “Symbolizing Time. Kittler and the Twenty-First-Century Media.” Kittler Now: Current Perspectives in Kittler Studies, edited by Stephen Sale and Laura Salisbury. Cambridge: Polity Press, 2015: 210-237. Harris, John. The Dark Side of the Moon: The Making of the Pink Floyd Masterpiece. Cambridge, MA: Da Capo Press, 2005. Häusermann, Jürg. “Zugespieltes Material. Der O-Ton und Seine Interpretationen.” Original/Ton. Zur Mediengeschichte des O-Tons, edited by Nikolaus Wegmann, et al. Konstanz: UVK Verlagsgesellschaft, 2007. NOISE RESONANCE | REFERENCES 259

Hayles, N. Katherine. How We Became Posthuman. Virtual Bodies in Cybernetics, Literature, and Informatics. Chicago: University of Chicago Press, 1999. ———. “Voices out of Bodies, Bodies out of Voices: Audiotape and the Production of Subjectivity.” Sound States: Innovative Poetics and Acoustical Technologies, edited by Adalaide Morris. London: University of North Carolina Press, 1997: 74-96. Hegarty, Paul. Noise/Music. A History. Cambridge: MIT Press, 2008. Heidegger, Martin. The Concept of Time. Translated by William McNeill. Oxford: Blackwell Publishers, 1992. ———. “The Question Concerning Technology.” The Question Concerning Technology and Other Essays. Translated by William Lovitt. London: Harper & Row Publishers, 1977: 3-35. ———. Being and Time. Translated by John Macquarric and Edward Robinson. Oxford: Blackwell Publishers, 1962 (1927). Heidenreich, Stefan. “Rauschen, Filtern, Codieren – Stilbildung in Mediensystemen.” Das Rauschen, edited by Sabine Sanio and Christian Scheib. Hofheim: Wolke Verlag, 1995: 17-26. Herivel, John. Joseph Fourier. The Man & the Physicist. Oxford: Clarendon Press, 1975. Hilberg, Frank. “Geräusche. Wider die Geringschätzung eines Kompositorisches Materials.” Positionen, vol. 10, 1992: 36-39. Höldrich, Robert R. “Auf der Suche nach dem Rauschen.” Das Rauschen, edited by Sabine Sanio and Christian Scheib. Hofheim Wolke Verlag, 1995: 126-148. Holmes, Thom. Electronic and Experimental Music. Technology, Music, and Culture. 2nd ed., New York: Routledge, 2002. Huber, David Miles, and Robert E. Runstein. Modern Recording Techniques. 7th ed., Oxford: Focal Press, 2010. Kadis, Jay. The Science of Sound Recording. Oxford: Focal Press, 2012. Kahn, Douglas. Earth Sound Earth Signal. Energies and Earth Magnitude in the Arts. Berkeley: University of California Press, 2013. ———. Noise, Water, Meat: A History of Sound in the Arts. Cambridge: MIT Press, 2002. Kaiser, Gerald. A Friendly Guide to Wavelets. Boston: Birkhäuser, 1994. Katz, Bob. Mastering Audio. The Art and the Science. Stoneham, Massachusetts: Focal Press, 2002. Kelly, Caleb. Cracked Media. The Sound of Malfunction. Cambridge: MIT Press, 2009. Kittler, Friedrich. “The God of Ears.” Translated by Anthony Moore and Paul Feigelfeld. Kittler Now. Current Perspectives in Kittler Studies, edited by Stephen Sale and Laura Salisbury. Cambridge: Polity Press, 2015a: 3-21. ———. “Preparing the Arrival of the Gods.” Kittler Now. Current Perspectives in Kittler Studies, edited by Stephen Sale and Laura Salisbury. Cambridge: Polity Press, 2015b: 95-112. ———. “‘Bei Tanzmusik Kommt Es Einem in die Beinen.’ Auditive Medienkulturen. Techniken des Hörens und Praktiken der Klanggestaltung, edited by Jens Schröter and Axel Vollmar. Bielefeld: Transcript, 2013a: 35-42. 260 NOISE RESONANCE | REFERENCES

———. “The God of the Ears.” The Truth of the Technological World. Essays on the Genealogy of Presence. Translated by Erik Butler. Stanford: Stanford University Press, 2013b: 45-56. ———. “Signal-To-Noise Ratio.” The Truth of the Technological World. Essays on the Genealogy of Presence. Translated by Erik Butler. Stanford: Stanford University Press, 2013c: 165-177. ———. “Dionysios Revisited. Vom Patriarchalischen Ideenhimmel und dem Reich der Irdischen Liebe. Friedrich Kittler im Gespräch mit Frank M. Raddatz.” Das Nahen der Götter Vorbereiten. München: Wilhelm Fink Verlag, 2012a: 62-87. ———. Und der Sinus Wird Weiter Schwingen. Über Musik und Mathematik. Köln: Verlag der Kunsthochschule für Medien Köln, 2012b. ———. “Lightning and Series – Event and Thunder.” Theory, Culture & Society, vol. 23, issue 7-8, 2006a: 63-74. ———. Musik und Mathematik. vol. 1, part 1. München: Wilhelm Fink Verlag, 2006b. ———. “Number and Numeral.” Theory, Culture & Society, vol. 23, issue 7-8, 2006c: 51-61. ———. “Echoes; ein Prolog.” Hörstürze. Akustik und Gewalt im 20. Jahrhundert, edited by Nicola Gess and Florian Schreiner. Würzburg: Königshausen & Neumann, 2005: 13-27. ———. “Wie der Sound von Heute vor 300 Jahren Entstand.” Weltbürgertum und Globalisierung, edited by Norbert Bolz, et al. München: Wilhelm Fink Verlag, 2000: 145-155. ———. Gramophone, Film, Typewriter. Translated by Geoffrey Winthrop-Young and Michael Wutz. Stanford: Stanford University Press, 1999. ———. “‘Vernehmen, Was Du Kannst.’ Über Neuzeitliche Musik als Akustische Täuschung.” Neue Zeitschrift für Musik, vol. 158, no. 5, 1997: 4-9. ———. “Musik als Medium.” Wahrnehmung und Geschichte. Markierungen zur Aisthesis Materialis, edited by Bernhard Dotzler and Ernst Martin Müller. Berlin: Akademie Verlag. 1995a: 83-99. ———. “Vers Une Musicologie Concrète. Bemerkungen zu Richard Voss.” Das Rauschen, edited by Sabine Sanio and Christian Scheib. Hofheim: Wolke Verlag, 1995b: 111-112. ———. “World-Breath. On Wagner's Media Technology.” Opera through Other Eyes, edited by David J. Levin. Stanford: Stanford University Press, 1994: 215-235. ———. “Der Gott der Ohren.” Draculas Vermächtnis. Technische Schriften, Leipzig: Reclam, 1993a: 130-148. ———. “Real Time Analysis, Time Axis Manipulation.” Draculas Vermächtnis. Technische Schriften, Leipzig: Reclam, 1993b: 182-207. ———. “Wenn Die Freiheit Wirklich Existiert, Dann Soll Sie Doch Ausbrechen,” by Rudolf Maresch. Rudolf-maresch.de, 04 Apr. 1992a, www.rudolf- maresch.de/interview/16.pdf. Accessed 18 Mar. 2013. ———. “Spooky Electricity. Lawrence Rickels Talks with Friedrich Kittler.” Artforum, Dec. 1992b: 66-70, monoskop.org/images/2/29/Rickels_ NOISE RESONANCE | REFERENCES 261

Laurence_Kittler_Friedrich_1992_Spooky_Electricity.pdf. Accessed 17 Jul. 2015. ———. Discourse Networks, 1800/1900. Translated by Michael Metteer and Chris Cullens. Stanford: Stanford University Press, 1990. Kosko, Bart. Noise. New York: Penguin, 2006. Kromhout, Melle Jan. “‘Antennas Have Long Since Invaded Our Brains.’ Listening to the ‘Other Music’ in Friedrich Kittler.” Thresholds of Listening. Sound, Technics, Space, edited by Sander van Maas. New York: Fordham University Press, 2015: 89-104. ———. “‘The Exceptional Purity of Sound.’ Noise Reduction Technology and the Inevitable Noise of Sound Recording.” Journal of Sonic Studies, Volume 7, 2014: www.researchcatalogue.net/view/84544/84545. Accessed 30 Jun. 2014. ———. “As Distant and Close as Can Be. Lo-fi Recording: Site-Specificity and (In)authenticity.” Soundscapes, Journal on Media Culture, vol. 15, 2012, www.icce.rug.nl/~soundscapes/VOLUME15/Lo-fi.shtml. Accessed 30 Oct. 2012. ———. “‘Over the Ruined Factory There’s a Funny Noise.’ Throbbing Gristle and the Mediatized Roots of Noise in/as Music.” Popular Music and Society, vol. 34, no. 1, 2011: 23-34. ———. “‘This Was Made to End All Parties.’ An Investigation of the Work of Einstürzende Neubauten between 1980 and 1990.” BA thesis, Universiteit van Amsterdam, 2006. Kursell, Julia. “Experiments on Tone Color in Music and Acoustics: Helmholtz, Schoenberg, and Klangfarbenmelodie.” Osiris, vol. 28, no. 1, 2013: 191-211. Leibniz, Gottfried Wilhelm. New Essays on Human Understanding. Edited by Peter Remnant and Jonathan Bennett. Translated by Peter Remnant and Jonathan Bennett. Cambridge: Cambridge University Press, 1996 (1765). ———. Philosophical Essays. Edited by Roger Ariew and Daniel Garber. Translated by Roger Ariew and Daniel Garber. Indianapolis: Hackett Publishing Company, 1989. Liébert, Georges. Nietzsche and Music. Translated by David Pellauer and Graham Parkes. Chicago: University of Chicago Press, 2004. Link, Stan. “The Work of Reproduction in the Mechanical Aging of an Art: Listening to Noise.” Computer Music Journal, vol. 25, no. 1, 2001: 34-47. Loy, Gareth. Musimathics. The Mathematical Foundations of Music. Vol. 1. Cambridge: MIT Press, 2006. Lützen, Jesper. “The Delta Function.” The Modern Physical and Mathematical Sciences, edited by Mary Jo Nye. Vol. 5. Cambridge: Cambridge University Press, 2002: 479-481. Malsky, Matthew. “Stretched from Manhattan's Back Alley to MOMA: a Social History of Magnetic Tape and Recording.” Music and Technoculture, edited by Rene Lysloff and Leslie Gay Jr. Middletown: Wesleyan University Press, 2003: 233-263. Maresch, Rudolf. “Waves, Flows, Streams. Die Illusion vom Reinen Sound.” Soundcultures. Über Elektronische und Digitale Musik, edited by Marcus S. 262 NOISE RESONANCE | REFERENCES

Kleiner and Achim Szepanski. Frankfurt am Main: Suhrkamp, 2003: 194- 217. Marshall, Owen. 2014. “A Brief History of Auto-Tune.” Sounding Out! Sound Studies Blog, 21 Apr. 2014, soundstudiesblog.com/2011/09/12/in-defense-of- auto-tune. Accessed 20 Nov. 2016. Mason, Nick. Inside Out. A Personal History of Pink Floyd, edited by Phillip Dodd. San Francisco: Chronicle Books, 2005. Millard, Andre. America on Record. A History of Recorded Sound. 2nd ed., New York: Cambridge University Press, 2005. Mills, Mara. “Deafening: Noise and the Engineering of Communication in the Telephone System.” Grey Room, vol. 43, 2011: 118-143. Milner, Greg. Perfecting Sound Forever. An Aural History of Recorded Music. New York: Faber and Faber, 2009. Moles, Abraham. Information Theory And Esthetic Perception. Translated by Joel E. Cohen. Urbana: University of Illinois Press, 1966. Monk, John. “Creating Reality.” Ways of Thinking, Ways of Seeing. Mathematical and Other Modelling in Engineering and Technology, edited by Chris Bissell and Chris Dillon. Berlin: Springer, 2012: 1-29. Morton, David. Off the Record: The Technology and Culture of Sound Recording in America. New Brunswick: Rutgers University Press, 2000. Moseley, Roger. “Digital Analogies: The Keyboard as Field of Musical Play.” Journal of the American Musicological Society, vol. 68, no. 1, 2015: 151-228. Mowitt, John. “The Sound of Music in the Era of Its Electronic Reproducibility.” Music and Society. The Politics of Composition, Performance and Reception, edited by Richard Leppert and Susan McClary. New York: Cambridge University Press, 1987: 173-197. Nahin, Paul. Dr. Euler’s Fabulous Formula Cures Many Mathematical Ills. Princeton: Princeton University Press, 2006. Nietzsche, Friedrich. Beyond Good and Evil. Prelude to a Philosophy of the Future. Translated by Ian Johnston. Arlington: Richer Resources Publications, 2009. Nunes, Mark. Introduction. Error, Noise, and Potential: The Outside Of Purpose. Error: Glitch, Noise, and Jam in New Media Culture, edited by Mark Nunes. New York: Continuum, 2011: 3-23. Paiuk, Gabriel. “Tactility, Traces, Codes: Reassessing Timbre in Electronic Media.” Organised Sound, vol. 18, issue 03, 2013: 306-313. Parikka, Jussi. “Mapping Noise: Techniques and Tactics of Irregularities, Interception, and Disturbance.” Media Archaeology: Approaches, Applications and Implications, edited by Jussi Parikka and Erkki Huhtamo. Berkeley: University of California Press, 2011: 256-277. Peters, John Durham. The Marvelous Clouds. Toward a Philosophy of Elemental Media. Chicago: University of Chicago Press, 2015. ———. “Resemblance Made Absolutely Exact. Borges and Royce on Maps and Media.” Variaciones Borges, vol. 25, 2008: 1-24. Pohlmann, Ken C. Principles of Digital Audio. 4th ed., New York: McGraw-Hill, 2000. ———. Principles of Digital Audio. 1st ed., New York: McGraw-Hill, 1985. Poss, Robert M. “Distortion Is Truth.” Leonardo Music Journal, Vol. 8, 1998: 45-48. NOISE RESONANCE | REFERENCES 263

Prigogine, Ilya. The End of Certainty. Time, Chaos, and the New Laws of Nature. New York: The Free Press, 1997. Prigogine, Ilya, and Isabelle Stengers. “Postface: Dynamics from Leibniz to Lucretius.” Hermes. Literature, Science, Philosophy, by Michel Serres. Edited by Josue V. Harari and David F. Bell. Translated by Josue V. Harari, David F. Bell. Baltimore: Johns Hopkins University Press, 1982. Read, Oliver. The Recording and Reproduction of Sound. A Complete Reference Manual on Audio for the Professional and the Amateur. 2nd ed., Indianapolis: Howard W. Sams, 1952. Redmon, Nigel. “What Is Dither?” Earlevel Engineers. 1996, www.earlevel.com/ main/1996/10/20/what-is-dither. Accessed 16 Dec. 2013. Ricœur, Paul. Time and Narrative. Translated by Kathleen McLaughlin and David Pellauer. Vol. 1. Chicago: University of Chicago Press, 1994. Roads, Curtis. Microsound. Cambridge, MA: MIT Press, 2001. ———. The Computer Music Tutorial. Cambridge: MIT Press, 1996. Robjohns, Hugh. “Q. What Is Different about the Many Varieties of Dolby Noise Reduction?” Sound On Sound, Aug. 2005, www.soundonsound.com/sound- advice/q-what-different-about-varieties-dolby-noise-reduction. Accessed 30 May 2013. Rodgers, Tara. “On the Process And Aesthetics of Sampling in Electronic Music Production.” Organised Sound, vol. 8, issue 3, 2003: 313–320. Rothenbuhler, Eric W., and John Durham Peters. “Defining Phonography: An Experiment In Theory.” Musical Quarterly, vol. 81, no. 2, 1997: 242-264. Ruin, Hans. “Ge-stell: Enframing as the Essence of Technology.” Martin Heidegger. Key Concepts, edited by Bret W. Davis. Durham: Acumen Publishing, 2010: 183-194. Rumsey, Francis, and Tim McCormick. Sound and Recording. 6th ed., Oxford: Focal Press, 2009. Russolo, Luigi. The Art of Noise (Futurist Manifesto 1914). Translated by Robert Filliou. Ubuclassics. 2004 (1913), www.ubu.com/historical/gb/russolo_ noise.pdf. Accessed 15 Jul. 2007. Sabbe, Herman. “A Paradigm of 'Absolute Music': Goeyvaerts's N°.4 as ‘Numerus Sonorus.’” Revue Belge De Musicologie/Belgisch Tijdschrift voor Muziekwetenschap, vol. 59, 2005: 243-251. ———. “Goeyvaerts and the Beginnings of 'Punctual' Serialism and Electronic Music.” Revue Belge De Musicologie/Belgisch Tijdschrift voor Muziekwetenschap, vol. 48, 1994: 55-94. Sangild, Torben. The Aesthetics of Noise. Copenhagen, Datanom, 2002, www.ubu.com/papers/noise.html. Accessed 6 Dec. 2005. Schmidt Horning, Susan. Chasing Sound. Technology, Culture and the Art of Studio Recording from Edison to the LP. Baltimore: John Hopkins University Press, 2013. Schwartz, Hillel. Making Noise. From Babel to Big Bang and Beyond. New York: Zone Books, 2001. Schwartz, Mischa. “Improving the Noise Performance of Communication Systems: Radio and Telephony Developments of the 1920s.” IEEE Communications Magazine, vol. 47, no. 12, 2009: 16-20. 264 NOISE RESONANCE | REFERENCES

Scott, Charles E. “Care and Authenticity.” Martin Heidegger. Key Concepts, edited by Bret W. Davis, Durham: Acumen Publishing, 2010: 57-68. Sebald, W.G. The Rings of Saturn. Translated by Michael Hulse. London: Vintage Books, 2002. Serres, Michel. The Five Senses. Translated by Margaret Sankey and Peter Cowley. London: Continuum, 2008. ———. Genesis. Translated by Geneviève James and James Nielson. Ann Arbor: University of Michigan Press, 1995. ———. Hermes. Literature, Science, Philosophy. Edited by Josue V. Harari and David F. Bell. Translated by Josue V. Harari, David F. Bell. Baltimore: Johns Hopkins University Press, 1982a. ———. The Parasite. Translated by Lawrence R. Schehr. Baltimore: John Hopkins University Press, 1982b. Sethares, William A. Tuning, Timbre, Spectrum, Scale. 2nd ed., London: Springer, 2005. Shannon, Claude. “Communication in the Presence of Noise.” Proceedings of the IEEE, vol. 86, no. 2, 1998: 447-457. Shannon, Claude, and Warren Weaver. The Mathematical Theory of Communication. Urbana: University of Illinois Press, 1964. Siefert, Marsha. “Aesthetics, Technology and the Capitalization of Culture: How the Talking Machine Became a Musical Instrument.” Science in Context, vol. 8, no. 2, 1995: 417-449. Siegert, Bernhard. Cultural Techniques. Grids, Filters, Doors, and Other Articulations of the Real. Translated by Geoffrey Winthrop-Young. New York: Fordham University Press, 2015a. ———. “Mineral Sound or Missing Fundamental: Cultural History as Signal Analysis.” Osiris, vol. 28, no. 1, 2013: 105-118. ———. “Die Geburt der Literatur aus dem Rauschen der Kanäle. Zur Poetik der Phatischen Funktion.” Electric Laokoon: Zeichen und Medien, von der Lochkarte zur Grammatologie, edited by Michael Franz et al. Berlin: Akademie Verlag, 2007: 5-41. ———. Passage des Digitalen. Zeichenpraktiken der Neuzeitlichen Wissenschaften 1500-1900. Berlin: Brinkmann & Bose Verlag, 2003. “Signal-to-Noise Ratio (S/N Ratio).” Encyclopedia of Recorded Sound, 2nd ed., 2005. Stäheli, Urs. “Financial Noises: Inclusion and the Promise of Meaning.” Soziale Systeme, vol. 9, no. 2, 2003: 244-256. Steinfeld, Thomas. “Nachruf auf Friedrich Kittler – Spekulant über die Zeichen.” Süddeutsche Zeitung, 18 Oct. 2011, www.sueddeutsche.de/kultur/ nachruf-auf-friedrich-kittler-spekulant-ueber-die-zeichen-1.1168023-2. Accessed 5 Sep. 2016. Steintrager, James A. “Speaking of Noise: From Murderous Loudness to the Crackle of Silk.” Differences, vol. 22, no. 2-3, 2011: 249-275. Sterne, Jonathan. MP3. The Meaning of a Format. Durham: Duke University Press, 2012. ———. “The Death and Life of Digital Audio.” Interdisciplinary Science Reviews, vol. 31, no. 4, 2006: 338-348. NOISE RESONANCE | REFERENCES 265

———. The Audible Past. Cultural Origins of Sound Reproduction. Durham: Duke University Press, 2003. Stockhausen, Karlheinz. “The Origins of Electronic Music.” The Musical Times, vol. 112, no. 1541, 1971: 649-650. Szendy, Peter. Listen. A History of Our Ears. Translated by Charlotte Mandell. New York: Fordham University Press, 2008. Tempelaars, Stan. Signal Processing, Speech and Music. Lisse: Swets & Zeitlinger Publishers, 1996. Théberge, Paul. Any Sound You Can Imagine: Making Music/Consuming Technology. Hanover: Wesleyan University Press, 1997. “The Phonautograms of Édouard-Léon Scott de Martinville.” First Sounds, 2008, www.firstsounds.org/sounds/scott.php. Accessed 13 Nov. 2015. Thompson, Emily. The Soundscape of Modernity. Architectural Acoustics and the Culture of Listening in America, 1900–1933. Cambridge, MA: MIT Press, 2002. ———. “Machines, Music, and the Quest for Fidelity: Marketing the Edison Phonograph in America, 1877-1925.” The Musical Quarterly, vol. 79, no. 1, 1995: 131-171. “Tone Test.” Encyclopedia of Recorded Sound, 2nd ed., 2005. Toop, Richard. “Stockhausen and the Sine-Wave: The Story of an Ambiguous Relationship.” The Musical Quarterly 65 (3), 1979: 379-391. “Toshimaru Nakamura/Sachiko M. Press release from Erstwhile.” Erstwhile Recordings, Mar. 2001, www.efi.group.shef.ac.uk/labels/erstwhil/ erst013.html. Accessed 15 Nov. 2016. Turner, R. Steven. “The Ohm-Seebeck Dispute, Hermann Von Helmholtz, and the Origins of Physiological Acoustics.” The British Journal for the History of Science, vol. 10, no. 1, 1977: 1-24. Von Neumann, John. “The General and Logical Theory of Automata.” Collected Works. Edited by A. H. Taub. Vol. 5. Oxford: Pergamon Press, 1963: 288- 326. Wang, Avery L. “An Industrial-Strength Audio Search Algorithm.” 4th Symposium Conference on Music Information Retrieval, Washington, DC, 26-30 Oct. 2003, Conference presentation, www.ee.columbia.edu/~dpwe/papers/ Wang03-shazam.pdf. Accessed 13 Sep. 2012. Wannamaker, Robert A., et al. “Stochastic Resonance as Dithering.” Physical Review E, vol. 61, no. 1, 2000: 233-236. Watkinson, John R. “The History of Digital Audio.” Magnetic Recording. The First 100 Years, edited by Eric D. Daniel, et al. New York: The Institute Of Electrical And Electronics Engineers, Inc. 1999: 110-123. Wegmann, Nikolaus, et al., editors. Original/Ton. Zur Mediengeschichte des O-Tons. Konstanz: UVK Verlagsgesellschaft, 2007. Wellbery, David. Foreword. Discourse Networks, 1800/1900, by Friedrich Kittler. Translated by Michael Metteer and Chris Cullens. Stanford: Stanford University Press, 1990: VII-XXXIII. White, Carol. Time and Death: Heidegger's Analysis of Finitude. Edited by Mark Ralkowski. Aldershot: Ashgate, 2005. 266 NOISE RESONANCE | REFERENCES

White, Paul. “Tape Noise Reduction Exploration.” Sound On Sound, Jan. 1996, www.soundonsound.com/sos/1996_articles/jan96/reducingnoise.html. Accessed 30 Dec. 2012. Wiener, Norbert. “Spatio-Temporal Continuity, Quantum Theory and Music.” The Concepts of Space and Time. Their Structure and Their Development, edited by Milič Čapek. Dordrecht: D. Reidel Publishing Company, 1976: 539-546. Wilden, Anthony. System and Structure. Essays in Communication and Exchange. 2nd ed., London: Tavistock Publications, 1980. Wills, David. “Positive Feedback: Listening behind Hearing.” Thresholds of Listening. Sound, Technics, Space, edited by Sander van Maas. New York: Fordham University Press, 2015: 70-88. Winthrop-Young, Geoffrey. Kittler and the Media. Cambridge: Polity Press, 2010. Wittje, Roland. “Concepts and Significance of Noise in Acoustics: Before and After the Great War.” Perspectives on Science, vol. 24, no. 1, 2016: 7-28. Zak III, Albin. The Poetics of Rock. Cutting Tracks, Making Records. Berkeley: University of California Press, 2001. Zao, Yi. “Artifact.” The University of Chicago Theories of Media Keywords Glossary, Winter 2007, csmt.uchicago.edu/glossary2004/artifact.htm. Accessed 5 Jul. 2015.

Discography

Arca. Mutant, Mute, 2015 Beach Boys, The. “Good Vibrations.” Smiley Smile, Capitol, 1994. Boards of Canada. Music Has the Right to Children, Warp, 1998. Bowie, David. Low, EMI, 1999. Bush, Kate. Hounds of Love, EMI, 1995. Einstürzende Neubauten. Kollaps, Potomak, 2003. ———. Silence Is Sexy, Mute, 2000. Eno, Brian. Music For Airports, Editions EG, 1990. Goeyvaerts, Karel. “Compositie Nr.4.” The Serial Works [# 1-7], Megadisc Classics, 1998. Harrison, George. “All Things Must Pass.” All Things Must Pass, Parlophone, 2001. Jaar, Nicolas. Sirens, Other People, 2016. Jackson, Michael. “Beat It.” Thriller, Epic, 2008. Nakamura, Toshimaru. No Input Mixing Board, Zero Gravity, 2000. Meek, Joe, and The Blue Men. I Hear A New World, RPM Records, 1991. Merzbow. Pulse Demon, Relapse Records, 1998. Pink Floyd. “Echoes.” Meddle, EMI, 2011a. NOISE RESONANCE | REFERENCES 267

Pink Floyd. “Great Gig In The Sky.” The Dark Side of the Moon, EMI, 2011b. Reed, Lou. Metal Machine Music, Buddha Records/RCA, 2000. Talking Heads. “Heaven.” Fear of Music, Sire, 1984. Throbbing Gristle. The Second Annual Report of Throbbing Gristle, Industrial, 2011. Wonder, Stevie. Talking Book, Motown, 1986. Wonder, Stevie. Innervisions, Motown, 1990.

Videography

Kittler, Friedrich. “Blitz und Serie – Ereignis und Donner.” 2003. Vimeo, uploaded by Formatlabor Berlin, 28 Mar. 2011, vimeo.com/21605213. Siegert, Bernhard. “On Codes and Coding.” The LeBoff Public Lecture, Department of Media, Culture and Communication, New York University, 9 Apr. 2015b. Vimeo, uploaded by Media, Culture, Communication, 21 Apr. 2015, vimeo.com/125610691.

NOISE RESONANCE | SUMMARY 269

Summary

Noise Resonance | Technological sound reproduction and the logic of filtering

This thesis develops a revaluation of the role of noise and distortion in recorded sound and music. Given its growing importance in musical practices over the past century, it asks what it is about noise that musicians and listeners are drawn to over and over again. Contrary to common conceptualisations of noise that define its musical role as a sonic marker for failure, violence, excess, transgression or subversion, this thesis argues that an analysis of the operations of technical media shows how noise and distortion have been fundamental in shaping the singular sound of music in the media age. Chapter One opens with a historical overview of the continuous efforts to prevent, reduce and eliminate the noises of recording and reproduction technologies from the invention of the phonograph in 1877 to the dawn of the digital age in the late twentieth century. Along this history, it traces the emergence of a myth of perfect fidelity based on the discursive separation between sounds considered internal to a recording and noises that are introduced by the reproduction procedure itself. On the basis of this discursive separation, the concept of noise changed from a sonic concept in mid-nineteenth century acoustics to a physical concept in communication engineering in the early twentieth century. Finally, in the mid-twentieth century, information theory reconceptualised noise as anything that interrupts a clear signal transmission. I call this discursive framework supporting the opposition between wanted signals and unwanted noise (to be removed, eliminated or reduced at all cost) the conceptual logic of noise reduction. 270 NOISE RESONANCE | SUMMARY

To be able to develop an alternative conceptualisation of the role of noise in sound reproduction, Chapter Two focuses on two technologies that exemplify and problematise this conceptual logic of noise reduction. Firstly, the analogue noise reduction systems of the 1960s illustrate the attempt to optimally reduce the noise of sound reproduction. Secondly, the practice of ‘dithering’ in digital recording is an example of the deliberate addition of noise to remedy shortcomings of the reproduction process. Although seemingly contrasting operations, I argue that both strategies ultimately support the myth of perfect fidelity. Like noise reduction conceals noise and reveals silence to suggest the perfect transmission of pure signals, dither conceals digital artefacts in order to maintain the suggestion of an intrinsic relationship between input and output. Both strategies thereby confront the inherent limitations of symbolic representation and technological reproduction. To further assess these limitations of sound reproduction and their relation to the inevitable presence of noise, I introduce the concept of a noise resonance of sound reproduction. Before assessing the implications of this noise resonance of sound reproduction, however, Chapter Three further problematises the conceptual logic of noise reduction by looking at the most important analytical tools for the representation of sound: the representation of sound waves as a series of individual frequencies by Fourier analysis and the corresponding figure of the sine wave as the representation of such a single frequency. Seemingly providing a scientific basis for age-old ideas of musical harmony and regularity, Fourier analysis marks the discursive origin of the clarity and purity that define the conceptual logic of noise reduction. As such, the clean cuts applied by the operations of Fourier analysis represent what I call the domain of the ideal filter. In contrast to the symbolic purity of this domain, the operations of technical media are subject to a fundamental physical uncertainty principle that limits every non-symbolic attempt to produce such ideal purity. Due to this uncertainty principle, the physical operations of technical media always require a negotiation between accuracy in time and accuracy in frequency. Because the physical cuts applied by technical media in the domain of physical filters are affected by this negotiation, they add randomness, transience and noise to their output signal. NOISE RESONANCE | SUMMARY 271

This uncertainty principle and the physical limitations it puts on the operations of technical media suggest some kind of relation between the introduction of noise and the factor of time. Chapter Four therefore continues the conceptualisation of the noise resonance of sound recording with an assessment of the multi-layered temporality of technological sound. Contrary to the clean cut of ideal filters, which leaves no trace whatsoever, the operations of physical filters shape their output in specific ways. However, the changes they add to the sound are not, as the myth of perfect fidelity suggests, external to the reproduced signal. They are the traces of the signal’s journey through space and time that changed its frequency composition. These traces, I argue, randomly shape the signal’s sonic contours and produce the multi-layered temporality of technologically (re)produced sound. On the one hand, their transience and randomness emphasise the impossibility to fully capture a signal, thereby signifying the pastness of reproduced sound and resonating with a sense of finitude. On the other hand, because of this random transience, they produce a sense of the continuous flow of time, thereby signifying the radical presence of reproduced sounds as they resonate in the here and now. Following this assessments of the way technological sound reproduction shapes and changes all output signals, Chapter Five puts forward the notion of a logic of filtering as the basis for the noise resonance of sound reproduction. This logic emphasises that neither input signal nor output signal, but the operations of the filtering channels in between the two are primary for understanding the sound that comes out of the recording chain. Doing away with the idea that technological sound reproductions are an incomplete version of some original source, the logic of filtering accounts for the fact that the noisy traces of the physical cuts of technical filters are constitutive for the singularity of the sound of music in the media age. Exactly because these cuts continuously escape our analytical grasp, they produce what media philosopher Friedrich Kittler calls “unforeseeable, unthinkable, unimaginable acoustic events.” Ultimately, I argue, these noisy traces of irrepresentable filtering operations resonate technologically (re)produced sound in the receptive 272 NOISE RESONANCE | SUMMARY

ears of human listeners. They thereby constitute an essential element of the sound of music in the media age.

NOISE RESONANCE | SAMENVATTING 273

Samenvatting

Ruisresonantie | technologische geluidsreproductie en de logica van het filter

Wat is ruis, en welke rol spelen ruis en vervorming in het denken over en het maken van muziek? In het licht van het groeiend aantal ‘ruispraktijken’ dat in de muziek van de afgelopen eeuw zijn intrede deed, dient zich de vraag aan waarom dergelijke ‘ruizigheid’ sonisch en muzikaal relevant is. In tegenstelling tot gangbare interpretaties die muzikale ruis associëren met mislukking, geweld, overdaad, transgressie of subversie—en ruis daarmee buiten de muzikale en maatschappelijke orde plaatsen—betoog ik aan de hand van een analyse van de operaties van technische media dat ruis en vervorming juist fundamenteel zijn voor de specifieke klank van muziek in het mediatijdperk. Hoofdstuk 1 opent met een historisch overzicht van de vele pogingen om het ontstaan van ruis en vervorming door opname- en weergavetechnologieën te voorkomen, te verminderen of te elimineren, vanaf de uitvinding van de fonograaf in 1877 tot het digitale tijdperk in de late twintigste eeuw. Het schetst de ontwikkeling van een mythe van perfecte getrouwheid (fidelity) die is gebaseerd op een discursief onderscheid tussen, enerzijds, klanken die worden beschouwd als onderdeel van het opgenomen geluid en, anderzijds, de ruis en vervorming die wordt veroorzaakt door het reproductieproces zelf. Onder invloed van dit onderscheid tussen interne ‘klank’ en externe ‘ruis’ onderging het begrip ruis in de loop van de tijd diverse transformaties. In de eerste plaats veranderde ruis van een sonisch concept in midden-negentiende-eeuwse akoestiek in een natuurkundig concept, ontwikkeld in de context van vroeg twintigste-eeuwse communicatie- 274 NOISE RESONANCE | SAMENVATTING

technologie. Vervolgens herdefinieerde de Informatietheorie in het midden van de twintigste eeuw ruis als alles wat een perfecte signaaloverdracht in de weg staat. Het discursieve kader dat aan de basis staat van de tegenstelling tussen gewenst signaal en ongewenste ruis (die koste wat kost verwijderd of verminderd moet worden) noem ik de conceptuele logica van ruisonderdrukking. Om een alternatieve interpretatie van de rol van ruis te kunnen ontwikkelen, bestaat Hoofdstuk 2 uit twee analyses van technologieën die deze conceptuele logica van ruisonderdrukking zowel illustreren als problematiseren. De analoge ruisonderdrukkingssystemen uit de jaren zestig zijn een voorbeeld van de poging om de ruis van geluidsreproductietechnologieën maximaal te verminderen. De toepassing van zogeheten dither in de digitale opnamepraktijk is daarentegen een manier om door middel van het bewust toevoegen van ruis specifieke tekortkomingen van het reproductieproces te verhelpen. Hoewel zij elkaars tegenpolen zijn, betoog ik dat deze processen in gelijke mate de mythe van perfecte getrouwheid bekrachtigen. Met het verbergen van ruis en het introduceren van stilte wekt technologische ruisonderdrukking de suggestie dat er sprake is van een geslaagde overdracht van zuivere signalen. Op soortgelijke wijze benadrukt dither-ruis door middel van het verbergen van specifieke digitale artefacten het intrinsieke verband tussen invoer en uitvoer (oftewel origineel en reproductie). Beide strategieën wijzen daarmee op beperkingen die inherent zijn aan iedere symbolische representatie of technologische reproductie van geluid. Om te benadrukken dat deze beperkingen in verband staan met de onvermijdelijke aanwezigheid van ruis, introduceer ik het concept van de ruisresonantie in geluidsreproducties. Om dit concept van ruisresonantie uit te werken, wordt in Hoofdstuk 3 de conceptuele logica van ruisonderdrukking verder geproblematiseerd door middel van een analyse van een van de belangrijkste natuurkundige instrumenten voor de analytische weergave van geluid: de representatie van geluidsgolven als een reeks individuele frequenties door middel van een zogeheten Fourieranalyse, en de figuur van de sinusgolf als de representatie van een dergelijke individuele NOISE RESONANCE | SAMENVATTING 275

frequentie. Aangezien de Fourieranalyse in de negentiende eeuw eeuwenoude opvattingen over muzikale harmonie en regelmaat empirisch bevestigde, staat deze aan de basis van latere ideeën over helderheid en zuiverheid die de conceptuele logica van ruisonderdrukking kenmerken. De zuivere snede (clean cut) die de Fourieranalyse suggereert, is kenmerkend voor het domein van ideale filters. In tegenstelling tot de absolute, maar puur symbolische, zuiverheid in dit analytische domein worden de concrete operaties van technische media bepaald door een natuurkundig onzekerheidsprincipe dat iedere niet-symbolische poging om dergelijke absolute zuiverheid te bereiken fundamenteel beperkt. Vanwege dit onzekerheidsprincipe moeten technische media altijd een compromis sluiten tussen hun nauwkeurigheid in de tijd en hun nauwkeurigheid in het frequentiespectrum. Dit compromis beïnvloedt de fysieke sneden (physical cuts) die worden toegepast door technische media in het domein van fysieke filters. Derhalve voegen fysieke filteroperaties altijd een zekere hoeveelheid willekeurigheid, vluchtigheid en ruizigheid toe aan het uitvoersignaal. De fysieke beperkingen in de operaties van technische media, veroorzaakt door het onzekerheidsprincipe, suggereren een relatie tussen het optreden van ruis en de factor tijd. Om deze relatie te onderzoeken, wordt in Hoofdstuk 4 het concept van de ruisresonantie in geluidsreproducties verder ontwikkeld op basis van een analyse van de gelaagde temporaliteit van technologisch ge(re)produceerd geluid. In tegenstelling tot het symbolische product van de zuivere sneden van ideale filters—die geen sporen nalaten—wordt het uitvoersignaal van fysieke filters altijd op specifieke manieren vervormd. De elementen die door deze vervorming aan het signaal worden toegevoegd kunnen echter niet, zoals de mythe van perfecte getrouwheid suggereert, los worden gezien van het signaal. Ze zijn het resultaat van de reis door tijd en ruimte die het frequentiespectrum en daarmee de contouren van het signaal op willekeurige wijze verandert. Deze klinkende sporen creëren de gelaagde temporaliteit van technologisch ge(re)produceerd geluid. De inherente vluchtigheid en willekeurigheid van de sporen bevestigen enerzijds de onmogelijkheid om een signaal volledig te vangen. Zij wijzen op de voorbijheid (pastness) van ieder technologisch 276 NOISE RESONANCE | SAMENVATTING

gereproduceerd geluid en resoneren met ons gevoel van eindigheid. Anderzijds bevestigen deze vluchtigheid en willekeurigheid de constante aanwezigheid van het geluid in het heden. Zij wijzen daarmee op de fysieke aanwezigheid van technologisch ge(re)produceerd geluid in het algemeen en doen het telkens resoneren in het hier en nu. Aan de hand van deze analyse van de manier waarop technologische geluidsreproductie alle uitvoersignalen vormt en vervormt, introduceert Hoofdstuk 5 tenslotte de logica van het filter als conceptuele basis voor de ruisresonantie van geluidsreproductie. De logica van het filter benadrukt dat noch de invoer, noch de uitvoer van een reproductieproces, maar de operaties van het filter (het kanaal) zèlf leidend zijn voor ons begrip van technologisch ge(re)produceerd geluid. Het rekent af met het idee dat een technologische geluidsreproductie een incomplete versie van de oorspronkelijke bron is, en benadrukt dat de manier waarop de fysieke sneden van technische filters het signaal beïnvloeden bepalend zijn voor het specifieke geluid van de uitvoer. Juist omdat deze fysieke sneden van technische media nooit volledig met een analytische blik zijn te doorgronden, produceren technische filteroperaties dat wat mediafilosoof Friedrich Kittler “onvoorziene, ondenkbare, onvoorstelbare akoestische gebeurtenissen” noemt. Ik stel op mijn beurt dat juist de aanwezigheid van de ruizige sporen van deze niet-representeerbare filteroperaties maakt dat dergelijke mediatechnologische geluiden weerklank vinden in de ontvankelijke oren van menselijke luisteraars. Ze vormen daarmee een essentiële component van de klank van muziek in het mediatijdperk.

What is it about noise that attracted musicians and listeners over the past century? Noise Resonance: Technological Sound Reproduction and the Logic of Filtering sets out to answer this question through an extensive conceptual revaluation of the role of noise and distortion in sound and music. The book traces the issue of noise in a detailed media archaeological analysis of analogue and digital sound technologies.

Noise Resonance does away with the idea that sound reproductions are incomplete copies of some original source and thereby challenges more common conceptualisations that define noise as a marker for failure, violence, excess, transgression or subversion. Instead, on the basis of an assessment of the history of acoustics and the development of sound technology from the nine- teenth century onward, it repositions noise as essential for the singular sound of music in the media age.

Noise Resonance shows how noise and distortion, introduced by the operations of technical media, have been fundamental for shaping the specific sound of technologically repro- duced music. Drawing from disciplines like musicology, media theory, sound studies and contemporary philosophy, it ultimately suggests a way to rethink the relation between music and listeners in the age of technological media.

ISBN: 978-94-028-0524-6 ©2017