Timbral Composition with Granular Synthesis
Mara Helmuth
Columbia University Texas A & M University
Music Department Music Department
New York, NY 10027 College Station, TX 77843
mara@woof.music.columbia.edu, mara@silvertone.princeton.edu
Abstract:
Granular synthesis programs written in the Cmix programming language,
and compositional methods using this software were discussed. Two
compositions show the use of these programs: one for tape alone, and a
concerto for eight instruments and computer. Both displayed use of
contour shaping to form timbral and cross-parametric gestures. The
timbral evolution throughout the composition is an important structural
element. The Cmix and NeXT granular synthesis programs used which
involve both synthesis and sampling are described. A real-time
implementation of granular synthesis using MAX on the IRCAM
Workstation, and a live performance were also discussed.
Introduction
I have approached writing software from the point of view of a composer who writes
music that is not easily expressed in traditional music notation. The techniques that have
seemed most conducive to expressing the ideas musically have been those that allowed
precise control over timbre and how it unfolds over time. Constructing complex sounds from
grains, or smaller units of sound, has been called granular synthesis [Roads, 1988]. This
technique, a form of additive synthesis, allowed vast timbral variety because any resulting
waveform could be created using the correct grain parameters. Because the units are
usually simple and easily modified, precise control over parameter change was possible.
Visual analogies were useful in thinking about the event characteristics, and manipulatable
graphically in StochGran, a NeXTstep interface to a Cmix granular synthesis program.
Dragon of the Nebula, for tape, was created with StochGran and displays contours shaped
graphically. Work with sampled sound as the basic grain was done with another Cmix
instrument, which incorporated in-phase correlation. The composition Evolutions, for
computer and eight instruments, used this granular sampling instrument as well as some
real time MAX patches on the IRCAM Signal Processing Workstation (ISPW) [Puckette,
1988].
The Software: StochGran
To implement granular synthesis, the Cmix programming language [Lansky, 1990] was
chosen for its flexibility. Programs, or instruments, could be written in C with as much
complexity as desired. The original version of these programs was described in Helmuth
[1991]. The parameters for the instrument include minimum and maximum values for grain
rate, duration, location and frequency. For each of these, four values were selected which
represent the probability distribution. The low and high values delineate the range, while the
mid point designates the value around which to cluster by the amount specified in the
tightness value. Four functions specified the shape of change throughout the event
between the minimum and maximum values. To facilitate composing with this instrument, a
NeXTstep interface was constructed. The composer manipulated graphical objects to
design the sound. The break points for the four functions were created or moved with the
mouse (Fig. 1).
In order to use sampled sound in the grain, similar programs were written which read
through a sound file. Since grain phases that are not aligned could cause unwanted noise,
in-phase correlation was used [Jones, et al., 1988]. As the grains are overlapped, the start
point of the second grain was chosen so that the grains will be at the most similar point in
their waveform found in the space of one period. The period, specified by the user, was the
period of the fundamental, or most prominent lowest pitch in the sound.
In addition to the Cmix/NeXTstep programs, MAX patches were created to do real time
granular synthesis. Frequency, duration and rate of the grains changed according to the
selected probabilities as described above. Performance with instruments was made more
spontaneous as the computer performer could trigger events in reaction to the live
conductor. Further description of these patches will appear at a later point, with multiple
voices and expanded control.
Contour Shaping to Form Gestures
Timbre can be perceived on several levels. One can often hear the grain rate and other
parameters, and be aware of how they change in time. On a higher level, one may also
have associations with some previously heard sound. Both of these levels may be used
compositionally, simultaneously.
In shaping the parameter changes with StochGran, one builds complex timbral entities
that unfold over time. The parameters may evolve similarly or inversely, or in a more
complex relationship. For example, frequency of the grain may rise as the grain rate is
dropping, creating a particular effect. If both changes proceed slowly and then very quickly
with an abrupt reversal, a clearly defined punctuating effect is produced at the end of the
sound. This musical phrase may be called a gesture, which implies the organization or
connection between its component movements. If similar gestures occur at several climaxes
after slow builds, this gesture can be seen as a structural element of the piece.
In addition to these types of relationships, granular sounds that I found most interesting
sometimes resembled natural or mechanical sounds, such as voices or water. These effects
could act as recognizable symbols to the listener. Because the grain parameters were
entirely manipulatable by the composer, however, one could control the relationships
between the sounds, and move from any sound to another sound by moving from one set of
parameters to another. This flexibility has not usually been possible using sampled sound in
musique concrete composition.
Two Musical Examples
Dragon of the Nebula, for tape, was created with granular synthesis. Algorithmic
programs generated a large number of parameter settings. From these sometimes
unexpected results, appropriate sounds were selected for the composition. Gestures were
shaped and layered with the intention of producing an interesting, continuously changing, but
unified sound.
Evolutions, for eight instruments and computer, made use of the granular sampling
programs. The grains were composed of segments of sampled instrumental sound. Some
of the timbre classes created for the piece were: vocal, natural and mechanical. Within the
vocal class, for example, were sounds I named "oo", "ah", "midvocal monastery" or
"chorusy". In the natural category were "thunder", and "water", and some mechanical
sounds were "grindy" or "mass" sounds. The composition began with long, sustained vocal
sounds created from the sampled cello sound in the computer part. Each instrument came
in alone for a duet with the computer, with the computer exploring some of the sounds
created from that particular instrumental sound. During these duets, the gestures formed
between the computer and instrumental part migrated among the categorized sounds and
transitional timbres.
Summary
Granular synthesis has proved to be a flexible compositional technique for timbral
control. Several aspects of composition with granular synthesis, including parametric
continuity and interaction, and contour shaping have been discussed here. Classes of
granular synthesis timbres have be identified, developed and inter-related compositionally.
Dragon of the Nebula, for tape, and Evolutions, for eight instruments and computer,
illustrated continuous timbral and textural change by manipulating grain and source
parameters. The software used consists of Cmix programs written to do granular synthesis,
and a NeXTstep interface for graphic control. In addition, the second piece involved ISPW
Max patches for real time granular synthesis.
References
[Helmuth, 1991] Helmuth, Mara. Patchmix And StochGran. Proceedings of the
International Computer Music Conference. Montreal: McGill University. 1991.
[Jones, et al., 1988] Jones, D. and Parks, T. Generation and Combination of Grains for
Music Synthesis. Computer Music Journal 12(2):27 - 34, 1988.
[Lansky, 1990] Lansky, Paul. The Architecture and Musical Logic of Cmix. Proceedings of
the International Computer Music Conference. Glasgow: ICMC Glasgow 1990, 1990.
[Puckette, 1988] Puckette, Miller. Combining Event and Signal Processing in the MAX
Graphical Programming Environment. Computer Music Journal 15(3):68 - 77, 1988.
Cambridge, MA: Massachusetts Institute of Technology.
[Roads, 1988] Roads, Curtis. Introduction to Granular Synthesis. Computer Music Journal,
12(2), 1988. Cambridge, MA: Massachusetts Institute of Technology.