Categories
Dictionary

Granular Synthesis

Agostino Di Scipio

In the jargon of electronic and computer music, granular synthesis refers to a vast range of sound synthesis and processing techniques sharing a common basic assumption: in principle, any sound can be modelled by juxtaposing and piling up large amounts of grains, the latter being defined as time-finite signals of given frequency and amplitude, with duration in the order of the centi-seconds. Iannis Xenakis had been a very early proponent of such an approach, and explored its creative implications in the late 1950s. He did not actually use the term granular synthesis (to be coined in the 1970s [Roads 1978]) but defined it as a kind of synthèse du son à base de quanta sonores [sound synthesis based on sonic quanta]. He wrote: “All sound is an integration of grains, of elementary sonic particles, of sonic quanta” and therefore any sound can be analytically modelled and electronically generated as “an assemblage of a large number of elementary grains adequately disposed in time” (Xenakis 1963, 61 & 1992, 43). By assembling – composing, in fact – myriads of sonic grains, one can create interesting and complex sonorities.

The idea was first sketched at the time of the realization of Analogique B (1959), a tape music work Xenakis finalized at the GRM studio (Groupe de Recherches Musicales), in Paris, with sound materials initially produced in Hermann Scherchen’s studio, in Gravesano (Switzerland). Analogique B eventually became the tape component of Analogique A et B (scored for 9 strings and tape, 1958-59). In itself, it was only made of textural sonorities whose granular density (average number of grains per time unit) and variable frequency ranges were regulated according to a detailed compositional plan (detailed surveys and analytical sketches are found in Di Scipio 1998 and 2006, as well as in Hagan 2005). Xenakis used to call them sound clouds, an atmospheric metaphor later taken up also by other composers.

As is known, textural sounds having a kind of tactile, abrasive character, are often featured in Xenakis’s orchestral music and represent a distinctive feature of his electroacoustic and computer music (Solomos & Hoffmann 1998; Solomos 2015). Most emblematic, in this regard, is the little interlude Xenakis sketched for diffusion in Le Corbusier’s and Varèse’s Philips Pavillon (Expo 1958, Brussels), later entitled Concret PH, a short yet outstanding example of musique concrète, only consisting of crackling sound textures made out of the sound of burning charcoals. The composer made a recording of the firing coals, split it into several tape chunks (in the order of few seconds) and mixed these latter together in various fashions. As he did so, he effectively made classical studio techniques (in Pierre Arnaud’s DMS studio, in Paris) act as a crude form of granular processing, i.e. the cut-up and re-arrangement of several minute slices or droplets of sound.

Xenakis was certainly driven by a fascination with natural sounding phenomena (e.g. cicada choirs and various meteorological events) and other phenomena (fireworks, the noise of weapons in battlefields, etc.). However, his goal was not of a naturalistic or mimetic kind: he was interested in the inner details, in the internal sounding dynamics of such events. Aside of that, Xenakis was aware of the research done by Dennis Gabor, a British-Hungarian physicist who – back in the 1940s – had elaborated a “theory of hearing” based on acoustical quanta (Gabor 1947). He had known Werner Meyer-Eppler’s work on “statistic problems of sound” and “aleatoric modulations” (Meyer-Eppler 1958 & 1959). (Xenakis heard Meyer-Eppler lecturing at Gravesano in the late 1950s. Presumably he first heard about Gabor attending one of the Meyer-Eppler’s lectures.) On these matters, he must have had relevant exchanges with Abraham Moles, a physicist-turned-information-theorist and long-time accomplice of Pierre Schaeffer at GRM. Xenakis’s keen curiosity for the new quantum- and information-oriented models of sound testifies to his distrust for models rooted in classic Fourier analysis. He would later denounce “the impasse of harmonic analysis” (Xenakis 1992, 243), suggesting it could be overcome by quantum-based methods and stochastic models. Today Fourier analysis has not been overcome, of course; nonetheless, in many research areas just as in several technical contexts, it is regularly replaced with time-finite representations of various sorts (wavelet analysis and the like).

With the technical resources of the late 1950s, the making of Analogique B proved overwhelmingly complicated. In theory, Xenakis’s effort might be described as a granular synthesis implementation with analogue means. In actuality, it was not even that: it was a kind of “magnetophonic” operation of sorts, as it in fact required the prior taping of electronically generated sound grains (approximate duration 40 msec) and the copying and splicing of numerous tape segments (in the order of seconds or tens of seconds). Xenakis devised a clever work-plan, making his task not too complicated and tedious. Yet, bringing the job to completion must have proved difficult and frustrating, and the end result anyway sounded less interesting than the composer expected. Confronting such circumstances, Xenakis even considered resorting to computational means (as far as we know, that must have been the first time he thought of using computers). But time was not ripe. Later in 1962, when he would be granted access to the IBM 7090 mainframe computer of IBM France (for no longer than one hour, anyway), that was in order to achieve algorithmically-composed instrumental music (Keller & Ferneyhough 2004; Grintsch 2009).

Granular synthesis was to (re)surface some fifteen years later in the context of computer music, with Curtis Roads’s experiments in deferred-time digital sound synthesis (Roads 1978). Roads had attended Xenakis’s classes at Bloomington (University of Indiana) in 1972. Two years later he would implement his first computer-based granular synthesis (with Max Mathews’s compiling language Music V, running on the mainframe computers of the University of California San Diego). In the mid-1980s came Barry Truax’s first real-time implementation (with a PDP minicomputer coupled to special-purpose digital signal processors) (Truax 1988). More developments, with a variety of technological configurations, emerged in the mid- and late-1980s, in the work of computer music practitioners – including Horacio Vaggione (in France), Eugenio Giordani (in Italy), and the author (also in Italy). Developments in granular synthesis were soon followed by digital granular processing approaches – i.e., the granular transformation or granulation of given sounds. The latter opened the way to efficient techniques of time-stretching and related signal manipulations in the frequency domain independent of changes in the time domain. Related approaches were to appear across the decades – including FOF synthesis (forme d’onde formantique), concatenative synthesis, particle synthesis and a variety of sound texture synthesis methods (Schwarz 2011) – not to mention several software tools, either freely or commercially available. One way or another, all developments in microsound (overviewed in Roads 2001) have acknowledged Xenakis’s seminal role in the history of granular sound techniques and related musical explorations (Zavagna & Di Scipio 2016).

Any implementation of granular synthesis involves two distinct but interrelated tasks: one addressed to define the grain properties and the other addressed to the ways by which myriads of grains are arranged in the time-frequency domain (often with statistical and probabilistic methods). In 1959, as he pondered the then-impossible recourse to computational resources, Xenakis wrote that he would need two computer programs, one defining the grain waveform parameters, in ways consistent with Gabor’s theory, and one shaping the probabilistic scattering of sound grains (Xenakis 1963, 72). In addition, he had a kind of recursive strategy in mind: he hypothesized that “sonorities of second order and even third order etc.” (Xenakis 1963, 122) would become audible, with higher-level processes rearranging the sounding outcomes of lower-level processes.

With the tools available to him, Xenakis could not eventually materialize that multi-level strategy (Di Scipio 1997 & 1998). As a matter of fact, only in few passages of Analogique B the stream of sonic quanta is so dense as to coalesce into a second-order sonority: more often than not, one can hear the sonic units for what they are – i.e., individual (quasi-)sine wave particles. However, it seems clear that his strategy involved an attempt to link micro-composition (Xenakis’s word for sound synthesis) to macro-composition (the design of articulated, larger-scale structures), in the quest for a tighter compositional integration of sound and music.

After 1959, Xenakis would never go back to granular synthesis. Not even when more advanced sound technologies would have better supported him. Yet, the technical questions and the compositional implications first dealt with in the making of Analogique A et B were never set aside, as Xenakis would creatively re-formulate them in later computer music efforts, from the early 1970s to the early 1990s (Di Scipio 2009). Indeed his path towards a tighter compositional integration of sound and music culminated in the early 1990s, namely in the thoroughgoing approach of automated music synthesis undertaken with the GENDYN computer program (Hoffmann 2002) and related musical output (Di Scipio 1998 and 2015) (see Stochastic Synthesis).

By embracing granular or quantum-based acoustical models, Xenakis evidently meant to pursue a greater freedom in composing: that provided him with a meaningful notion of sound synthesis, understood in fact as micro-composition. After all, the ancient Greek word syn-thesis and the Latin cum-ponere are equivalent. In Xenakis’s work, sound synthesis thus became a crucial compositional task, the domain in which a composer can craft the means necessary to compose sound and music – or better, sound as music. He managed to compose the sound synthesis process itself just as much as he composed with it.

References

Di Scipio, Agostino.1997. “The problem of 2nd-order sonorities in Xenakis’ electroacoustic music”, Organised Sound, 2 (3), 1997.

Di Scipio, Agostino. 1998. “Compositional models in Xenakis’ electroacoustic music”, Perspectives of New Music, 36 (2), 1998.

Di Scipio, Agostino. 2006. “Le nubi di suono ed i loro meccanismi. Uno studio di Analogique A et B”, in Iannis Xenakis. Musicista scienziato architetto (Alessandro Melchiorre, ed.), Edizioni Scuole Civiche di Milano, 2006.

Di Scipio, Agostino. 2009. “The notion of synthesis in Xenakis’s music”, in Iannis Xenakis: Das elektroakustiche Werk (Ralph Paland and Christoph von Blumenröder eds.), Der Apfel Verlag, 2009.

Di Scipio, Agostino. 2015. Agostino Di Scipio, “Stochastics and granular sound in Xenakis’ electroacoustic music”, in Xenakis. La Musique Électroacoustique (Makis Solomos, ed.) L’Harmattan, 2015.

Gabor Dennis. 1947. “Acoustical quanta and the theory of hearing.” Nature 159, 1947.

Grintsch, Jan Simon. 2009. “Random control. Xenakis and early mainframe computers”, in Iannis Xenakis: Das elektroakustiche Werk (Ralph Paland and Christoph von Blumenröder, eds.), Der Apfel Verlag, 2009.

Hagan, Kerry. 2005. “Genetic analysis of Xenakis’ Analogique B”, in Proceedings of the Electroacoustic Music Studies Network, Montreal, 2005 http://www.ems-network.org/spip.php?article150

Hoffmann, Peter. 2002. “Music Out of Nothing? The Dynamic Stochastic Synthesis: a Rigorous Approach to Algorithmic Composition by Iannis Xenakis”,PhD, Technische Universität Berlin, 2002.

Keller, Damien & Brian Ferneyhough. 2004. “Analysis by Modeling: Xenakis’s ST/10-1 080262”, Journal of New Music Research 33 (2), 2004.

Meyer-Eppler, Werner. 1958. “Statistic and psychologic problems of sound”, Die Rehie 1, 1958.

Meyer-Eppler, Werner. 1959. Grundlagen und Anwendungen der Informations Theorie. Springer-Verlag, 1959.

Roads, Curtis. 1978. “Automated Granular Synthesis of Sound”, Computer Music Journal, 2 (2), 1978.

Roads, Curtis. 2001. Curtis Roads, Microsound, MIT Press, 2001.

Schwarz, Diemo. 2011. “State of the Art in Sound Texture Synthesis”, in Proceedings of the International Conference on Digital Audio Effects (DAFx-11), Paris, 2011. http://recherche.ircam.fr/pub/dafx11/Papers/30_e.pdf

Solomos, Makis, ed. 2015. Iannis Xenakis, la musique électroacoustique. L’Harmattan, 2015.

Solomos, Makis & Peter Hoffmann. 1998. “The electroacoustic music of Iannis Xenakis”, in Proceedings of the Symposium in Music and Computers, Ionian University, 1998.

Truax, Barry. 1988. “Real-time granular synthesis with a digital signal processing computer”, Computer Music Journal, 12 (2), 1988.

Zavagna, Paolo & Agostino Di Scipio, eds. 2016. Grani sonori e textures sonore. Special issue of Musica / Tecnologia, vol.10, 2016 https://oajournals.fupress.net/index.php/mt/issue/view/284

Xenakis, Iannis. 1963. Musiques formelles. Special issue of La revue musicale, 253/254, 1963.

Xenakis, Iannis. 1992. Formalized Music. Thought and Mathematics in Music, Pendragon Press, 1992. Revised and expanded revision of Formalized Music. Thought and Mathematics in Music, Indiana University Press, 1971.

How to cite

DI SCIPIO, Agostino. 2023. “Granular Synthesis.” In A Xenakis Dictionary, edited by Dimitris Exarchos. Association Les Amis de Xenakis. https://www.iannis-xenakis.org/en/granular-synthesis