Granular synthesis

from Wikipedia, the free encyclopedia

Granular synthesis is a method used in some synthesizers and programs to create artificial sounds .

Similar to a film, which simulates a fluid sequence through individual images, with granular synthesis a continuous sound is simulated, which in truth consists of many individual parts. These individual parts, the grains, are very short, digital sound fragments, the length of which is usually less than 50 milliseconds. If this limit is exceeded, the listener can recognize the fragment as an independent sound event.

The application of granular synthesis for the resynthesis of sampled material is widespread . The recorded sound is analyzed and broken down into precisely those grains in order to put them back together again afterwards.

Compared to the conventional sample playback method, this has the advantage that the speed of the playback process can be changed independently of the pitch. This is achieved by multiplying the grains when the material or the individual grains are stretched ( time stretching ). The formants and the pitch can also be edited independently during granular synthesis ( pitch shifting ).


Theoretical basis

The theoretical basis of granular synthesis is the Gabor transformation and the related theory of sound quanta by Dennis Gábor . Here, an analogy between quantum phenomena in particle physics and the acoustic properties of short sound segments ( Gabor grains ) is used.

Naturally occurring acoustic signals usually have a change in their spectrum over time. In order to be able to describe it exactly, one would have to be able to determine the spectrum of the signal for a certain period of time over any short time interval. However, the Fourier analysis with the help of the continuous Fourier transformation is unsuitable for the investigation of the frequency spectra of short signal segments, since errors that are too large occur when extrapolating to an infinite time range.

The resulting inaccuracy when determining the partial frequencies of the signal is inversely proportional to the duration of the signal segment. Time and frequency (spectrum) are therefore complementary properties of the signal section which cannot be exactly determined at the same time. A similar situation is known in the form of Heisenberg's uncertainty relation in quantum physics.

Gabor takes advantage of this by using mathematical tools in his theory to describe acoustic signals that were originally developed to describe quantum physical phenomena. He considers elementary signals that contain exactly a quantum of information about the spectrum and the course of time. These have the form of a harmonic oscillation that is modulated in the time domain by an envelope curve that indicates the probability of the corresponding spectrum being located on a section of the time axis.

history

The composer Iannis Xenakis developed a compositional theory on the basis of Gabor's theory of sound analysis in the early 1970s. Several of his compositions from this period describe the generation of musical events from a number of elementary sound particles, the composition of which changes over time. Technically, this was initially implemented in a very complex way by cutting up audio tapes into tiny sections and putting them back together according to the specifications of the composition.

In the mid-1970s, the composer Curtis Roads (* 1951) carried out further research on the subject and experimented with computer-generated granular compositions, the calculation of which took a long time to complete. In order to simplify the control of the synthesis, he developed methods for the graphical description of the control parameters.

Barry Truax (* 1947) first developed a process for generating granular sounds in real time in the 1980s.

Granular synthesis was also used by artists. In 1991 the Austrian artist Kurt Hentschläger and the German artist Ulf Langheinrich founded the Duo Granular-Synthesis in Vienna. The name refers to the technique of granular synthesis, which Hentschläger and Langheinrich applied to both sound and visual work. Her installations work with disorienting projections, projected onto large screens, accompanied by a soundtrack created by means of granular synthesis, in order to completely destabilize sensory perception. The duo toured worldwide and won the Nagoya International Biennale competition.

functionality

Grain parameters

Waveform

Basically, any waveform can be used as the basis of granular synthesis, for example sounds generated with additive synthesis or FM synthesis , any type of audio sample , or even white noise .

Grain Duration

Theoretically, the length of a single grain can be chosen between 1 and 100 milliseconds. Usually a length of around 10 to 50 ms is used, as the pitch is barely perceptible below and the individual samples can be heard above. The grain duration can be chosen to be constant or variable for all grains.

Amplitude and envelope

An envelope curve is placed over each grain, which usually regulates the tone volume based on the following parameters: the permitted peak amplitude , the time to reach this (attack), the hold time (sustain) and the time to return to a volume of zero ( release). In the simplest case, such an envelope curve can be implemented as a trapezoidal shape, but more complex curves such as the Gaussian curve can also be used.

Grain density

Grain density, i.e. the number of grains that are played per second, is the factor that has the greatest influence on the texture of the sound that is produced in the end. A low density of less than 30 grains per second creates a slightly rhythmic effect. The more densely the grains are arranged, the more they overlap and the more likely it is that the perception of a rhythm is lost. At the same time, the amplitude and the pitch increase and the overlapping creates a more complex sound texture with more different overtones and formants .

Arrangement of the grains

Pitch-synchronous granular synthesis

The pitch-synchronous granular synthesis is a process for the analysis and resynthesis of a given sound with the aim of reducing data while preserving the formants. Put simply, the time-frequency level is broken down into many small cells, each of which is represented by a grain at the end.

Synchronous and quasi-synchronous granular synthesis

In quasi-synchronous granular synthesis, a stream of grains is formed, which are linked together at certain time intervals. Usually, in addition to a fixed time interval, a random deviation is also selected, whereby the individual grains are played back at irregular intervals, but a specified density (grains per second) is maintained. A special case of this type of synthesis is synchronous granular synthesis, in which the intervals between the played grains are all of the same length.

Asynchronous granular synthesis

The principle of asynchronous granular synthesis breaks away from the concept of a linear flow of grains. Instead, the grains are distributed over the time-frequency level by a stochastic or chaotic algorithm in so-called "clouds". The composer can define various properties of the "cloud", such as the duration, the permitted frequency band, waveforms used for the grains or a spatial distribution of the sounds.

Systems that can work with granular synthesis

Exemplary signal course of a grain

Web links

Individual evidence

  1. ↑ http://de.scribd.com/doc/101365788/Gabor-1946
  2. ↑ Christopher Philips: Machine Dreams . In: Art America . November 1999.
  3. ↑ http://www.granularsynthesis.com/hthesis/grain.html#21
  4. ↑ http://monoskop.org/images/d/d1/Roads_Curtis_Microsound.pdf p. 88
  5. ↑ http://www.granularsynthesis.com/hthesis/envelope.html#221
  6. ↑ http://www.granularsynthesis.com/hthesis/shape.html#2262
  7. ↑ http://www.granularsynthesis.com/hthesis/sync.html#2312
  8. ↑ http://monoskop.org/images/d/d1/Roads_Curtis_Microsound.pdf p. 96