Case Study: Granulation

Investigations into the possibilities of granulation experimented with the well- established possibilities of manipulating playback speed without affecting pitch or transposing audio without altering duration366_{. Granular synthesis techniques theorised} by Xenakis367 _{and Roads}368 _{were explored as well, but the concepts of clouds and} synthesis of textures from fragmented and dislocated grains has not been of particular personal interest. However, the possibility of slowing down and freezing sounds and creating textures from reverse playback were immediately seen as a suitable means to overcome the idiosyncrasies of the piano while maintaining its characteristics. The granulation technique resembles the basic approach of the live sampling technique used for drone creation as multiple instances of an audio segment are looped simultaneously. The playback of grains, using tiny segments of audio, are organised out of phase to create the illusion that the sound is continuous. Grain sizes of 80 ms are sufficient to create convincing results retaining acoustic resemblance to the original369_{. As the} sample position and pitch of the playback can be independently controlled, the performer has the possibility of ‘tuning’ into suitable fragments using a sensor or any other direct or indirect controller considered appropriate, e.g. the analysis data of the loudness of the current acoustic sound. The captured material can be reshaped entirely by controlling three parameters: position, pitch and volume of the playback.

Granulation was therefore considered a more satisfactory and practical solution for live performance situations allowing the eventualities of musical situations: ‘inappropriate’370 _{material can easily be changed by ‘tuning’ the playback position to} something more suitable, adapting volume to blend into the current proceedings or generate counterpoint to the acoustic instrument. But most importantly, the process 366 _{The first Granulation patches used Granular2.5 by Sakonda (2000). Further granulators were built by}

me using Nathan Wolek’s Granular Tool Kit (http://www.lowkeydigitalstudio.com/2007/03/granular- toolkit-v1-49/) taking advantage of the grain.phasor~ object, but eventually I have returned to a granulator modeled on Sakonda’s MSP programming approach.

367 _{Xenakis 1992: 237: Grains described as Sonotrons.} 368 _{Roads 2004}

369 _{One problem of granulation is that due to the overlay the resulting signals shows considerable phasing}

artefacts. This has in itself not been considered too much an issue within this project, especially within live situations. Alternative methods would involve phase-vocoding and other spectral synthesis

techniques which are able to use the spectral contours of the captured original in real time. Although the underlying technology differs significant from granular synthesis, the application and considerations described here would be fully transferable.

370 _{‘Inappropriate’ is situational, i.e. a dislike of the sound itself, avoidance of fragments of sound attacks}

allows control within a field of possible treatments. All parameters can be adjusted in a continuum, whether the sample is played in real-time, faster and slower, or the playback is frozen to a fixed position. The same applies to volume and pitch, either to create transpositions (all grains are tuned to the same value) or to produce polyphonic textures (grains are set to different values).

Next to the interesting sonic qualities of granulation the suitability of the process depends on the integration of the necessary controls. Direct control for the recording would be most reliable in terms of ensuring the capturing of the intended material. A tilt sensor to set the granulation position by a left/right rotation and volume control with forward/backward rotation would give overall gestural control. However, in practice it has also proven beneficial to have a way to limit the granulated material to a smaller selection, thus setting a minimum and maximum range to the sensor movement. Being able to influence the area of the buffer (recording) can adapt the precision of the movement through the recorded material or enable calibration of the sensor data to the physical gesture made. Given such adjustable and precise controls, granulation can be used for longer sections: adjustment of range enables the musician to focus into a certain section of material, while the rotation controls the actual playback position within this selections. A process module has been implemented allowing an arbitrary number of sections totalling up to ten minutes of recording time371_.

When an indirect control, such as onset detection, is used to trigger the recording with a significant event372_{, shorter buffer times are more appropriate and settings allow us to} grab these events immediately for processing to sustain or produce irregular repetitions. The repetitions might be perceived as an ordinary delay line. But the result differs as the timing between events varies due to randomised modulation of the set playback position. For the first scenario it is sufficient to map the onset detection to trigger the recording. No other operational tasks are required as the position of the granulator can be set to a constant value in order to freeze a sound. The granulated sound can be enriched by setting the position modulation to a random factor (i.e. within a range of 1% of the set position). Any new onset detected retriggers the recording and exchanges the 371 _{This granulation module with a 10 min recording buffer is not separately displayed in the flow chart of}

figure 4.3.

372 _{Significant has to be understood as a significant change in the spectral composite of the sound not of}

granulated material. A deliberate avoidance of significant attacks enables us to engage in counterpoint with the currently played material (audio example Rapprochement and

Opposition). The process is closely linked to the acoustic activity of the performer, who

can provoke changes and also attempt to keep responses by adapting her/his play. Practical application has shown how diverse the musical results can be although the parameters were not changed or adapted in any way. The process can be integrated to the performance activity in more complex ways by mapping indirect controls to more parameters (e.g. loudness to control granulation position). The increased complexity in control remains closely related to the acoustic activity of the performer, who is controlling all aspects of the process by musical gestures alone.

Despite the possible musical diversity, the last approach cannot be applied universally. Each approach has unique and significant musical potential. In other words: the two described methods form somewhat different poles within the field of possible “horizontal processes”373 _{of the audio modulation. When further mappings are added,} different approaches to the musical activity are possible. Considering the general discussion on parameter mapping in previous chapters, and considering the extensive research output available on this topic, the relation between process, its control structure and the resulting musical output is hardly surprising. But what appears to be absent from existing research are propositions of how it is possible to gradually change the paradigm of control: i.e., enabling a gradual focus on either control method that avoids binary decisions. Plotting the possibilities on a field of possibilities374 _{shows that it} ought to be possible to find any mixture between different methods. The search is for implementations that allow intuitively controllable processes within a true continuum while keeping operational tasks to a minimum. The possibility to change paradigm of controls in such a manner would give a more natural feel to the controls of processes matching qualities of acoustic instruments.

373 _{See Chapter 2 (p. 51) for definition.} 374 _{Equipping the ‘inner space’ of the process.}