An Exceptional Purity of Sound: Noise Reduction Technology and the Inevitable Noise of Sound Recording

Melle Kromhout

1.   Introduction: The Specificity of Different Noises

Critical assessments of the position of noise in sound recording often perform something of a conceptual noise reduction, in which a plurality of different noises is made manageable and defused by a singular concept of noise. But noise, both as a concept and as a phenomenon, consistently resists such overarching theories. As Frances Dyson (Dyson 2009: 189) notes in Sounding New Media: “the limit of noise is its oscillation between singularity and plurality: the difference between ‘a noise’ and ‘a sound’ and ‘Noise’ and ‘Sound.’ This difference designates a filter, what is in fact a noise filter, in the form of a capitalized ‘N’ that enables certain insights but obscures others.” Indeed, many attempts to conceptualize noise serve as such a filter: silencing all specific ‘noises’ almost entirely by the mute abstractness of overarching (or underpinning) concepts of some ontological “Noise.” [1]

By shoving all specific cases into a single conceptual framework, one avoids having to deal with the many different and often incongruous phenomena and concepts that “noise” connotes. This is why, to be able to commence a meaningful assessment of the position of noise in recorded sound, it is necessary to get away from broad conceptualisations and instead focus on the specificity of various noise-phenomena: to deal with plural noises instead of singular Noise. Adopting a technological point of view as the starting point of analysis enables one to do so. The following argument therefore focuses on the conceptual consequences of a specific kind of filter that played a pivotal role in the development of sound recording and its relation to noise: technological noise reduction systems. [2]

A close reading of noise reduction technology shows how it operates exactly at the crossroad of several leading notions of noise: sonic noise (which operates in the domains of sound and acoustics), physical noise (in communication engineering and physics) and communicational noise (in information theory and computer science). Whereas its developers regard it as a wholly neutral procedure that aims to make itself disappear along with the noise it reduces, noise reduction is in fact a much less stable and, actually, an explicitly active gesture. It is not simply a technological filter, but also a conceptual filter in Dyson’s sense, with which information theory and communication engineering perform a conceptual reduction (from many different noises to one concept of Noise) in order to enable the actual technological reduction of sonic and physical noise. As such, noise reduction actually produces its own notion of noise in the very process of reducing it.

Whereas Jonathan Sterne (Sterne 2003: 95) describes how this double reduction has indeed managed to effectively rid sound recording of the problem of noise, Michel Serres (Serres 1995; Serres 2008) consistently argues that the presuppositions underpinning the concepts of “noise” and “signal” in information theory not only provide the necessary framework for noise reduction, but also reveal that noise reduction is never fully complete or successful. Even more so, it transpires that noise reduction not only produces its own notion of noise, but also inevitably influences the signal itself. The output of noise reduction – a particular kind of silence – can be seen as a fake-signal: created out of the reduction of noise and posing as information.

As they are simultaneously involved in the reduction and production of various notions of noise, analysing noise reduction provides a way to disentangle how various notions of noise influence and contradict each other. Through asking what a certain filter tells about, first, the object it filters, and, second, the result of its filtering, noise reduction is a fruitful object to confront some of the conceptual difficulties surrounding the position and importance of noise in sound recording, without losing track of its technological and phenomenological specificities. [3]

2.   Three Noises 

In the second half of the nineteenth and first half of the twentieth century, the notion of noise underwent several profound changes, resulting in the dominance of three major conceptual strands. First, as described in Jonathan Sterne’s seminal book The Audible Past (Sterne 2003), nearly a century of research and experiments into the physical nature of sound and hearing redefined the traditional hierarchy between musical “sounds” and unmusical “noise.” Amidst numerous studies in acoustics, medicine, engineering and physics, renowned German physicist Hermann von Helmholtz applied the Fourier analysis to the study of overtone series and the timbral structure of sounds, coining the term “tone colour” (Evens 2005: 3-6; Kursell 2006: 216-221), but also spurring a new notion of noise. [4]

Secondly, major developments in communication technology, most notably the invention of the telegraph in the first and the telephone in the second half of the nineteenth century, gave birth to another dominant conceptualisation of noise. Musicologist Julia Kursell (Kursell 2006: 221) describes how, in the late nineteenth century, German philosopher and psychologist Carl Stumpf used “what had been the object under investigation in Helmholtz’s experiment” (namely the specific sonic qualities of sound) as “the stimulus for the investigation of a different object: communication.” By studying the human perception of differences between sounds instead of the qualities of individual sounds, Stumpf’s focus shifted from the study of tone colour to the basic question of communication engineering: how well is a signal sent in one place received in another? Next to the sonic “noise of sound,” this latter question introduced the very different “noise of communication engineering”: a random disruption in the transmission of signals.

In his book on the work of Friedrich Kittler, Geoffrey Winthrop-Young (Winthrop-Young 2010: 81) explains how in the second half of the nineteenth century both communication engineers and physicists applied the German word “Rauschen,” connoting the sound of natural “noisy” phenomena like the wind and the sea, to the physical phenomenon bearing some acoustical and physical similarities: a “disturbance variable with a broad-frequency spectrum.”As engineer Robert Höldrich (Höldrich 1995: 128) observes, in English such a disturbance is often just called “noise,” but the more correct description would be “random noise”: the physical phenomenon of irregular “random” movements of molecules or electrons. Somewhere in every signal transmission such random movements creep in and distort the signal.

The reduction of these random disruptions was of huge importance for the further development of telecommunication technologies. Hence, it was in the context of the optimisation of telephone lines that engineer and mathematician Claude Shannon, writing in 1948, treated noise no longer merely as an external nuisance, but placed it at the heart of his “Mathematical Theory of Communication” (Brech 1995: 99-100; Ballard 2008: 10; Sterne 2012: 87-88). After, firstly, the “noise of sound” and, secondly, the “random noise” of physics, this third conceptualisation of noise took centre stage in Shannon’s theory. Shannon not only proved that one ultimately cannot get rid of noise entirely, but also showed how, by statistically calculating the signal-to-noise ratio of a message, the receiver can determine the exact amount of noise that disrupted the signal during transmission and retroactively reconstruct the parts of the message that had been corrupted (Krämer 2006: 102).

The reduction of these random disruptions was of huge importance for the further development of telecommunication technologies. Hence, it was in the context of the optimisation of telephone lines that engineer and mathematician Claude Shannon, writing in 1948, treated noise no longer merely as an external nuisance, but placed it at the heart of his “Mathematical Theory of Communication” (Brech 1995: 99-100; Ballard 2008: 10; Sterne 2012: 87-88). After, firstly, the “noise of sound” and, secondly, the “random noise” of physics, this third conceptualisation of noise took centre stage in Shannon’s theory. Shannon not only proved that one ultimately cannot get rid of noise entirely, but also showed how, by statistically calculating the signal-to-noise ratio of a message, the receiver can determine the exact amount of noise that disrupted the signal during transmission and retroactively reconstruct the parts of the message that had been corrupted (Krämer 2006: 102).

            These three separate but nonetheless connected conceptualisations of noise – sonic, physical and informational or communicational – are important for understanding the complex role of the continuous production and reduction of noises throughout the history of sound recording. Fully congruent with Shannon’s assertion (Shannon 1998: 454) that noise can be calculated and reduced to a certain extent but is nonetheless ultimately inevitable, every attempt to improve recording technologies has been accompanied by its own kind of physical noise. In reaction to this, noise reduction technologies were developed to save innovative technologies like electrical recording and magnetic tape from their own internal noisy enemies.

Firstly, in the 1920s, as Greg Milner (Milner 2009: 50-58) recalls in Perfecting Sound Forever, opponents of brand new electrical recording technologies (most prominently Edison himself) insisted that electrical amplification meant the corruption of the purity of recorded sounds. Indeed, writes media theorist Stefan Heidenreich (Heidenreich 1995: 22), with all the extra wires, tubes, transistors and microphones positioned between the sound source and the listener, electrical recording “never runs without interferences and is therefore susceptible to noise.” Nonetheless, electrical recording became the univocally accepted standard, as its practical advantages – a significant increase in volume and more flexible recording methods – greatly overshadowed the desire for technical purity (Milner 2009: 52-56; Thompson 1995: 161; Malsky 2003: 244). Furthermore, electricity introduced sound recording to the domain of signal transmissions, opening the doors for the implementation in sound recording of earlier developments in noise reduction for telegraph, telephone and radio (Heidenreich 1995: 22).

Secondly, in the 1950s, magnetic tape yielded substantially less surface noise and, like electrical recording, proved far more flexible than any earlier technology (Milner 2009: 116-119). Once again, however, after the surface noise of wax roles, shellac and vinyl records that had always accompanied recording and the transduction noises introduced by the wires, tubes and microphones of electrical recording, magnetic tape added yet another type of noise to sound recording: “tape hiss,” the high-pitched sound caused by the movement of the magnetized particles on tape (Clark 1999: 94; Hawkins Williams, Isom III and Smith-Peaches 2003: 83). The flexibility of tape, however, also provided the solution to this problem, as it was the crucial last step toward a successful marriage between information theory and sound recording.

Ray Dolby explains the invention of Dolby Noise Reduction.

3.   Framing Noise Reduction 

In MP3, The Meaning of a Format, Jonathan Sterne describes the combined efforts of communication engineering and information theory in neutralising the disturbing effects of noise in signal transmission. With the help of successive innovations, writes Sterne (Sterne 2003: 95), “noise could be masked and put in its place; it did not have to be eliminated.” In short, information theory provides two possible strategies to deal with noise: it can either be put to good use by internalising it or be reduced by noise reduction technologies. Contained or manipulable, Sterne argues, noise disappeared as a problem.

The first, internalising, approach takes its cue from the fact that information theory rendered noise fundamentally relative to its context. In many cases, the ways that transmission and storage media change the original signal became part of specific aesthetic approaches. The noises of storage and transmission became part of what Emily Thompson calls the desired “quality of tone” of a specific sound recording (Thompson quoted in Benschop 2007: 499). This essentially means these noises are no longer unwanted by-products of the transmission or recording, but have instead become a crucial part of the musical or sonic message itself: noise becomes information and thereby effectively stops being noise. In case both source and receiver deem the added complexity of the signal disruptive to its transmission, however, noise must be reduced.

Various techniques enable the reduction of noise before, during or after recording takes place. The most sophisticated of these are dual-ended noise reduction technologies. Based on an earlier technique called “pre-emphasis/de-emphasis,” inventor and engineer Ray Dolby developed a procedure called “companding” in the early 1960s: a contraction of first compressing the signal’s amplitude at the moment it is recorded and subsequently expanding it at playback. High frequency-low volume passages are pushed in volume by compression, effectively “masking” the high-pitched noise of tape hiss. When expanded, the amplitude-range is restored to its original level, and together with the volume of the formerly compressed passages, the noise level is lowered significantly. Consequently, most noise is now masked by the louder signal and becomes practically inaudible. In a professional recording process, the compression or “coding” takes place during recording, between the sound source and the tape; the expansion or “decoding” happens during the production of the master tape.

Figure 1: Hyperphysics [n.d.]

The first version of Dolby’s noise reduction system, Dolby A, presented in 1966, reduced noise by up to maximally 10 decibels, especially in the higher frequency ranges, where tape hiss is most prominent (Dolby B, C, and S Noise Reduction Systems [n.d.]). However, as noise is less noticeable when it is equally spread, reducing high frequency noise unfortunately “focuses attention to noise in other spectra,” prompting the need to extend the process to lower frequencies as well (Dolby B, C, and S Noise Reduction Systems [n.d.]). Hence, the last analogue system, Dolby SR from 1986, added sophisticated sliding frequency bands in various ranges to also enable the reduction of “other low-level disturbances by as much as 25 dB” (Dolby^® SR 1987: 2). According to Dolby Laboratories, this last system succeeded in effectively pushing all noise below the level of audibility, yielding a “remarkable clarity of reproduction” (Dolby^® SR 1987: 5). [5]

A demonstration of noise reduction

Sterne (Sterne 2012: 94) calls this combination of either using or reducing noise, making noise effectively harmless, the “domestication of noise.” “Sometimes noise is hidden away,” he writes (Sterne 2012: 126), “sometimes it is endowed with meaning and portent; and sometimes it is simply let be.” Sterne’s analysis, however, does not do justice to the differences and complex interrelation between the sonic “noise of sound,” its transference to a physical phenomenon in communication engineering and its subsequent development into a fundamental, but contextually relative, concept in information theory. Because of its function in Sterne’s larger narrative about the history of perceptual coding, his analysis remains framed by the very discourse it deals with: the history of communication engineering and the conceptualisation of noise in information theory.

Viewed from within this frame, engineers like Dolby indeed manage to successfully deal with the problem of noise in signal transmission and, by extension, sound recording. As long as one stays within the domain of communication technology, the reduction of noise is a complex, but ultimately manageable affair, as the past century of developments in sound and communication technologies testifies. Although, following Shannon, the goal of complete removal can never be attained, the process, which Dolby Laboratories (Dolby^® SR 1987: 1) claims manages a “practical elimination of the influence of noise and non-linearity,” does effectively amount to less disruptive effects.

But, although the conceptualisation of noise in information theory is, as Friedrich Kittler (Kittler 1993: 161) argues, a consequence of the specific problems of communication technology, the relation between noise as a clearly defined problem in communication engineering and a much more abstract concept in information theory is far from seamless. Whereas the incorporation of noise as an intrinsic part of the system proved vital for dealing with the technical problems that certain kinds of physical noises pose, the context-dependent definition of noise that Shannon created makes the presuppositions of information theory its decisive framework.

Dolby’s description of noise reduction technology confirms this: according to a promotional folder (Dolby^® SR 1987: 2), they claim that “an ideal audio device or system would impose no audible limitation on the signal passing through it.” Although they are of course aware that such an ideal audio device is technologically impossible, Dolby’s systems do promise (Dolby^® SR 1987: 1) a “remarkable clarity of reproduction” and recordings with an “exceptional purity of sound.” Not measured against any objective, external standard, such transparency, clarity and purity can only be reached by removing everything that opposes this clarity, purity and transparency. As Dolby’s system seems to manage this, everything it reduces must therefore qualify as the opposite of transparency, clarity and purity or, in other words, as noise.

This goes to show that, as sociologist Urs Stäheli (Stäheli 2003: 246) observes in a compelling analysis of the noises of the London stock exchange, “classical communication theory always already knows where to locate noise.” Consequently, it also already seems to know how to reduce it fully. The noise that Sterne’s “domestication of noise” rendered harmless or useful is therefore a positivistic account of a phenomenon only negatively defined in the process of its reduction: within the framework of communication technology, noise is described, defined and problematized in order to be able to reduce its effects. The term “noise reduction” presupposes something called “noise” that can – and will – be reduced, without actually defining, or having to define, what this noise is. 

The framework itself is part of the process of reduction as the very idea of reduction is already inscribed in the phenomenon it aims to reduce. Noise reduction only reduces the noise that it defines as noise, exactly by reducing it. By presupposing both its goal through its object and its object through its goal, it is perfectly self-affirming.

4.   Noise Remains Present

In The Five Senses, Michel Serres (Serres 2008: 126) presents the successive journeys of Orpheus and Ulysses past the Sirens as different metaphors for dealing with the problem of noise in communication. The ship is the signal, the sea is the channel and the sirens produce noise. Ulysses’ triumph is based on a clever ruse: with the ears of his men firmly clogged with wax and he himself tied to the mast, he “blocks noise out” and ensures a clear transmission of his ship through the noisy channel. [6] For Serres (Serres 2008:126), this cunning victory is the victory of reason and science, exemplified by the way Leibniz “presupposed a world without noise” in favour of clear and rational information: an orderly world with clear messages. [7] Dolby presupposes such a world as well: the possibility of sound recording with “no audible limitation on the signal passing through it.” With this presupposition, a pure, clear, transparent and rational world is always already within reach for Ulysses, Leibniz and Dolby.

But this rational, scientific ruse of “Ulysses-Leibniz-Dolby” is inherently limited: according to Serres (Serres 1995:131) “the purest is never pure enough to remain the master forever.” Purity is always relative. Prior to Ulysses, Orpheus – the singer and musician – already safely steered the Argonauts past the sirens. Whereas Ulysses-Leibniz and Dolby-Sterne claim an absolute victory over noise, Orpheus does not block or suppress the noise entirely, but merely covers it with something else. He is not tied to the mast, nor did he clog the ears of his companions. Instead, by drowning out the noise of the sirens with his singing, he performed an excellent example of auditory masking avant-la-lettre: forcing siren-noise under singing-signal. This strategy leaves Orpheus, according to Serres (Serres 2008: 126), “open to the risk of collapsing into noise.” Noise is not blocked or removed entirely, but only temporarily suppressed, ready to crop up again at any time. Whilst Odysseus-Leibniz and Dolby-Sterne create a notion of noise that always already produces its own reduction, Orpheus shows what Shannon confirmed: noise remains present and crops up at different places.

Further than this, Stäheli (Stäheli 2003:253) writes that “every reduction of noise produces a noise of its own.” All sound recording creates sounds that are themselves the result of the process of recording. Each procedure, each technology, each material, each device produces its own specific disruptions, caused by either an audible change in the quality of the original sound or the introduction of sounds produced by the recording device or sound technology itself. In the terminology of engineers, these disruptions and distortions are called “artefacts.” [8] Indeed, some of the sounds of sound technology are entirely its own, created in the activities of recording and manipulation themselves. Noise reduction is susceptible to this as well: signals have to pass through a material channel to arrive at the decoder, inevitably contracting noise.

Contrary to what Dolby wants us to believe, even the correct use of noise reduction generates such artefacts. On commercial compact cassette players equipped with Dolby B or C, the user can switch the noise reduction on and off: “you may have noticed that Dolby B tapes [recorded with noise reduction switched on. MK] sound brighter when played without any Noise Reduction decoding,” writes Dolby Laboratories (Dolby B, C, and S Noise Reduction Systems [n.d.]). “Now you know why!” the folder continues, “you are hearing the encoded sound, not the original.” According to Dolby, switching on the decoder returns the signal to its “original” state, but that very act, the switching-on and off, demonstrates this is not the case: it shows the simultaneous concealing and revealing that takes place in all noise reduction technologies. When the device is switched off, one hears the coded version, which retroactively reveals the coding to have happened. Hence, when the device is switched on, noise is concealed by what noise reduction must logically reveal: what Dolby calls “the original,” but what is in fact the artefact of noise reduction.

Hence, instead of the seemingly neutral procedure portrayed by Ulysses-Leibniz and Dolby-Sterne, noise reduction is an active filter applied with specific criteria and specific standards. Surely, “the Dolby SR control circuit can create an infinite number of filters through which the signal must pass before it is recorded” (Dolby^® SR 1987: 5), but it can still only create filters that are to be applied to what they presuppose to be noise in the first place. In its attempt to produce clear signal transmission, implying a well-defined objective and reducing everything that falls within its equation, the process of noise reduction creates its own artefact in the form of a peculiar kind of silence.

The engineer in this video uses Dolby Noise Reduction as a specific effect: by encoding the signal with Dolby but not subsequently decoding it, he deliberately uses the ‘brighter’ sound of a Dolby-encoded recording.

5.   Silence As Fake-Signal

From the earliest days of sound recording, listeners have been trained to separate foreground from background sounds, or, in terms of information theory, separate signal from noise. [9] With this listening “through” the noise, Greg Hainge (Hainge 2007: 37) explains, “the listening subject integrates noise into a successful and seemingly noise-free, high-fidelity sonic expression in much the same way as a visitor to an art gallery will deliberately look through and not at the glass in front of a painting in order to see what he or she desires.” Before the interface of adequate technological noise reduction was invented, the audience thus functioned as a noise-filter in and of themselves.

Hence, rather than freeing a sonic environment that had always been consciously experienced as distorted, noise reduction revealed a new and unknown sound world. As Michel Chion (Chion 1999: 108) argues regarding the introduction of Dolby noise reduction in cinema soundtracks in the early 1970s, it especially created a kind of silence that audience had never experienced in, nor expected of sound recording. This novelty silence might have initially been perceived as the “absence of something,” the absence of the noise floor that had always been there, which listeners remembered and now suddenly noticed by its very absence. But as people grew accustomed to the new sonic standard, the silence no longer signified the absence of a previous sound, but the presence of silence: a silence that is not “left,” not even as a break between two instances of sound, but actively created or infused.

Sound scholar Mack Hagood (Hagood 2011) describes a similar silence in his analysis of the Bose QuietComfort Acoustic Noise Cancelling Headphones. Much like noise reduction these headphones separate, in the words of Bose-founder Dr. Amar Gopal Bose (Bose quoted in Hagood 2011: 575), “things that you don’t want from things that you want.” In order to do so, they actively create silence at places where noise is in fact abundant. Both headphones and noise reduction function as a medium, reducing noise and producing a reconstructed sonic image of an imagined original, with silence artificially induced. This pervades the silence with a particular significance and sense of agency: it is carefully constructed and as potentially significant as any sound.

John Mowitt (Mowitt 1987: 194) observes how noise reduction thereby “operate[s] according to a systematic logic that produces information out of suppressed noise.” This double gesture of concealing and revealing that noise reduction performs, can be understood in terms of what Heidegger (Heidegger 1977: 33) called the “Enframing” of technology: the way it is constantly “challenging forth into the frenziedness of ordering.” To enable such ordering, Heidegger (Heidegger 1977: 23) argues, nature should always be readily available “as standing-reserve” – to be formed, used and changed according to the needs and goals of technological developments. Hence, physics, the science that unravels and demystifies the workings of the natural world to make them suitable for production purposes, requires nature to be “orderable as a system of information.” Nature, in short, should be available as standing reserve to be orderable in a system of information by physics.

As nature and physics operate according to the demands of technologies, they both are “dependent upon the technical apparatus” (Heidegger 1977: 14) rather than, as it is commonly understood, the other way around. For Heidegger, nature serves physics serves technology. This “Enframing” – the process of ordering and arranging in order to create and maintain useable information – underpins the relation between technology, physics and nature. It also underpins the logic of the engineer, the logic of communication technology and the logic of information theory: serving the needs of a technology, they produce useable and saleable information through the reduction of noise. Adjusting the signal-to-noise ratio means arranging and rearranging the disposition of signals and noise into useful information.

Technological noise reduction conceals noise to reveal information. Rather than signifying the absence of “things that you don’t want,” it constantly produces “things that you want.” In information theory, the former is noise and the latter is signal. As noise is produced and defined retroactively by the process of noise reduction, the signal is as well. When noise is “everything that is reduced,” the signal must be “everything that is not reduced.” By its dual operation of compression and expansion, the supposedly neutral, but effectively active operation of noise reduction reduces what is deemed “outside” of information and reveals a silence that would seem to part of “what you want”: a fake-signal, posing as part of the useful information.

James Steinträger (Steinträger 2011: 257) paraphrases Michel Chion when he says that Dolby’s noise reduction filter “has enhanced the use of silence as itself message or expressive force.” Likewise, Heidenreich (Heidenreich 1995: 22) confirms that from the age of electrical recording onwards, filtering conceals noise, reveals silence and creates supposed significance. Epistemologically, this significant silence harks back to the discrete meaning and silent inwardness that, according to Friedrich Kittler (Kittler 1990: 70-176), had been the leading trope for literature in the pre-technological era: the meaning embedded in the silent reading of discreet signs. As if nothing slips through the cracks.

6.   Conclusion: The Medium Of Noise Reduction

But something will always stick to the signal, or, as Kittler (Kittler 1999: 3) writes, at some point something always “ceases not to write itself.” One could call this “something,” as Kittler does, the Lacanian Real, but in the language of information theory it is simply called “noise”: those random and contingent signals that continuously corrupt every attempt to establish clear communication. In the era when the alphabet was the dominant medium, the reduction of this corrupting force had been relatively effective, but with the invention of the phonograph, Heidenreich (Heidenreich 1995: 20) reminds us, noise could “claim maximal significance.” This “significance,” however, is not the same as what Kittler (Kittler 1990) calls literature’s discrete signs and its link to a silent inwardness of human souls.

Noise reduction is a strategy for impregnating sound recording with a meaningful fake-silence, suggesting orderly, delineated, clear significance. But, following Serres, given its idealized status, it remains a fundamentally temporary, unstable and, above all, precarious order, exemplified by Orpheus’ singing rather than Odysseus’ ruse. Or, as Stäheli writes (Stäheli 2003: 244), “it is the very operation of order(ing) which produces a noise of its own, making it impossible to ever establish […] order.” Information theory itself undercuts the clear difference between signal and noise. No matter how desperately noise reduction claims it cunningly reduces a clearly defined object and produces a supposed clean “original,” this object is always only negatively defined by what it is not.

Shannon’s purely contextual definition of noise introduced a fundamental ambivalence as to what noise is or can be. After all, Shannon defines the highest amount of information as the highest amount of improbability or disorder. Random, disorderly, contingent, technological media are always able to turn signals into noise and noise into signals. Signal and noise are fundamentally unstable. This, Kittler (Kittler 1993: 165) notes, blurred their distinction. Instead of an inherent property of the transmitted message itself, the separation of noise and signal as distinct properties of messages is the very result of the filter that the receiver applies (Rautzenberg 2009: 12), and the product of noise reduction always runs the risk of being interpreted as noise itself (Kittler 1993: 165).Sense can become nonsense, information can become noise and vice versa (Kittler 1995: 97).

As the process of noise reduction, like noise itself, is continuous and never sufficient, the “domestication of noise” is not and can never be the final word on the position of noise in recorded sound. As the channel will never cease to distort the signal, the received message cannot and will never be the same as the message that was sent. Although noise reduction reduces the effects it defines as noise to the level of “practical elimination,” the narrow end of the filter is always, with Serres, susceptible “to the risk of collapsing into noise.” Although the silence it produces claims a position as fake-signal, noise reduction can never fully conceal that it is actually the product of yet another channel: the medium of noise reduction.

A close reading of noise reduction therefore shows how the technology already signifies its own ultimate failure. It reveals noise as rarely singular and straightforward, but mostly plural and always operative at different levels: as a sonic phenomenon, as a physical disturbance and as an important, but highly ambivalent, concept in information theory. Somewhat paradoxically, the analysis of technologies for its reduction reveals that noise, in its many guises, is in fact an inevitable and even intrinsic aspect of all forms of sound recording.

 Acknowledgements

I would like to thank Niall Martin and Simon Ferdinand for proofreading this article at various stages of its development, as well as the peer reviewers and editors for their valuable commentary.

References

Anonymous (n.d.). “Dolby B, C, and S Noise Reduction Systems: Making Cassettes Sound Better.” Retrieved January 14, 2013.

Anonymous (n.d.). “Dolby Noise Reduction.” Hyperphysics. Retrieved April 15, 2013.

Anonymous (1987). “Dolby® SR. Dolby® Spectral Recording.” San Francisco: Dolby Laboratories. Retrieved January 14, 2013.

Attali, Jacques (2003). Noise: The Political Economy of Music (trans. Brian Massumi). Minneapolis: University of Minnesota Press.

Ballard, Susan Patricia (2008). “Out Of Order: Explorations In Digital Materiality” (Doctoral Dissertation). Sydney: The University Of New South Wales.

Benschop, Ruth (2007). “Memory machine or musical instruments.” International Journal of Cultural Studies 10/4: 485-502.

Brech, Martha (1995). "Rauschen: Zwischen Störung Und Information." In Sabine Sanio and Christian Scheib (eds.), Das Rauschen (pp. 99-107). Hofheim: Wolke Verlag.

Chion, Michel (1999). "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 40: 106-110.

Chion, Michel (2011). "Let's have Done with the Notion of Noise." Differences 22 2/3: 240-248.

Clark, Mark H. (1999). “Product Diversification.” In Eric D. Daniel, C. Denis Mee and Mark H. Clark (eds.), Magnetic Recording: The First 100 Years (pp. 92-109). New York: The Institute of Electrical and Electronics Engineers, Inc.

Dyson, France (2009). Sounding New Media: Immersion and Embodiment in the Arts and Culture. Berkeley: University of California Press.

Evens, Aden (2005). Sound Ideas: Music, Machines, and Experience. Minneapolis: University of Minnesota Press.

Hagood, Mack (2011). “Quiet Comfort: Noise, Otherness, and the Mobile Production of Personal Space.” American Quarterly 63/3: 573-589.

Hainge, Greg (2007). “Of Glitch and Men: The Place of the Human in the Successful Integration of Failure in the Digital Realm.” Communication Theory 17: 26-42.

Hainge, Greg (2013). Noise Matters: Towards an Ontology of Noise. New York: Bloomsbury.

Hawkins Williams, Pamela, Dotcy Isom III, and Tiffini D. Smith-Peaches (2003). "A Profile of Dolby Laboratories: An Effective Model for Leveraging Intellectual Property." Northwestern Journal of Technology and Intellectual Property 2/1: 81-98.

Hegarty, Paul (2008). Noise/Music: a History. Cambridge: MIT-press.

Heidegger, Martin (1977). The Question Concerning Technology and other Essays (trans. William Lovitt). London: Harper & Row Publishers.

Heidenreich, Stefan (1995). "Rauschen, Filtern, Codieren – Stilbildung in Mediensystemen." In Sabine Sanio and Christian Scheib (eds.), Das Rauschen (pp. 17-26). Hofheim: Wolke Verlag.

Hilberg, Frank (1992). "Geräusche. Wider Die Geringschätzung Eines Kompositorisches Materials." Positionen 10: 36-39.

Höldrich, Robert R. (1995). "Auf Der Suche Nach Dem Rauschen." In Sabine Sanio and Christian Scheib (eds.), Das Rauschen (pp. 126-148). Hofheim: Wolke Verlag.

Huber, David Miles and Robert E. Runstein (2010). Modern Recording Techniques. Seventh Edition. Amsterdam: Elsevier.

Kittler, Friedrich (1990). Discourse Networks, 1800/1900 (trans. David Wellbery). Stanford: Stanford University Press.

Kittler, Friedrich (1993). Dracula’s Vermächtnis. Technische Schriften. Leipzig: Reklam.

Kittler, Friedrich (1995). "Musik Als Medium." In Bernhard Dotzler and Ernst Martin Müller (eds.), Wahrnehmung Und Geschichte: Markierungen Zur Aisthesis Materialis (pp. 83-99). Berlin: Akademie Verlag.

Kittler, Friedrich (1999). Gramophone, Film, Typewriter (trans. Geoffry Winthrop-Young and Michael Wutz). Stanford: Stanford University Press.

Krämer, Sybille (2006). “The Cultural Techniques of Time Axis Manipulation: On Friedrich Kittler’s Conception of Media.” Theory Culture Society 23/7–8: 93–109.

Kursell, Julia (2006). “Shaping Differences: Hermann Von Helmholtz’s Experiments on Tone Colour.” Preprint 318. Conference: The Shape of Experiment, Berlin, 2–5 June 2005 (pp. 215-224). Berlin: Max-Planck Institute. Retrieved October 2, 2013.

Leibniz, G.W (1989). Philosophical Essays. Eds. Roger Ariew and Daniel Garber (trans. Roger Ariew and Daniel Garber). Indianapolis: Hackett Publishing Company.

Malsky, Matthew (2003). “Stretched from Manhattan’s Back Alley to MOMA: A Social History of Magnetic Tape and Recording.” In Rene Lysloff and Leslie Gay Jr. (eds.), Music and Technoculture (pp. 233-263). Middletown, Connecticut: Wesleyan University Press.

Milner, Greg (2009). Perfecting Sound Forever: An Aural History of Recorded Music. New York: Faber and Faber.

Mowitt, John (1987). "The Sound of Music in the Era of its Electronic Reproducibility." In Richard Leppert and Susan McClary (eds.), Music and Society: The Politics of Composition, Performance and Reception (pp. 173-197). New York: Cambridge University Press.

Rautzenberg, Markus (2009). Die Gegenwendigkeit der Störung Aspekte einer Postmetaphysischen Präsenztheorie. Zürich: Diaphanes.

Schwartz, Hillel (2011). Making Noise: From Babel to the Big Bang & Beyond. New York: Zone Books.

Serres, Michel (1995). Genesis (trans. Geneviève James and James Nielson). Ann Arvor: University of Michigan Press.

Serres, Michel (2008). The Five Senses (trans. Margaret Sankey and Peter Cowley). London: Continuum.

Shannon, Claude (1998). “Communication in the Presence of Noise.” Proceedings Of The IEEE, 86/2: 447-457. Retrieved September 27, 2013.

Shannon, Claude and Warren Weaver (1949). A Mathematical Theory of Communication. Illinois: The University of Illinois Press.

Stäheli, Urs (2003). “Financial Noises: Inclusion and the Promise of Meaning.” Soziale Systeme 9/2: 244-256.

Steinträger, James A. (2011). “Speaking of Noise: From Murderous Loudness to the Crackle of Silk.” Differences 22/2-3: 250-275.

Sterne, Jonathan (2003). The Audible Past: Cultural Origins of Sound Reproduction. Durham: Duke University Press.

Sterne, Jonathan (2012). MP3: The Meaning of a Format. Durham: Duke University Press.

Thompson, Emily (1995). “Machines, Music, and the Quest for Fidelity: Marketing the Edison Phonograph in America, 1877-1925.” The Musical Quarterly 79/1: 131-171

Weibel, Peter (1995). “Geräusche, Rauschen, Schall und Klang.” In Sabine Sanio and Christian Scheib (eds.), Das Rauschen (pp. 91-98). Hofheim: Wolke Verlag.

Winthrop-Young, Geoffrey (2011). Kittler and the Media. Cambridge: Polity Press.