Gendy1:
Filter:
Classes | UGens > Generators > Stochastic

Gendy1 : UGen : AbstractFunction : Object

Dynamic stochastic synthesis generator.
Source: Gendyn.sc

Description

An implementation of the dynamic stochastic synthesis generator conceived by Iannis Xenakis and described in Formalized Music (1992, Stuyvesant, NY: Pendragon Press) chapter 9 (pp 246-254) and chapters 13 and 14 (pp 289-322).

The BASIC program in the book was written by Marie-Helene Serra so I think it helpful to credit her too.

The program code has been adapted to avoid infinities in the probability distribution functions.

The distributions are hard-coded in C but there is an option to have new amplitude or time breakpoints sampled from a continuous controller input.

Technical notes

X's plan as described in chapter 13 allows the 12 segments in the period to be successively modified with each new period. Yet the period is allowed to vary as the sum of the segment durations, as figure 1 demonstrates. We can setup some memory of n (conventionally 12) points, or even simply vary successively a single point's ordinate and duration. There are thus various schemes available to us. In one, fix period T and only move the (ti, Ei) within the period. In another, have a memory of 12 segments but allow continuous modification of the inter point intervals and the amplitudes. In yet another, just have one point and random walk its amplitude and duration based on the probability distribution. In this implementation I allow the user to initialise a certain number of memory points which is up to them. To restrict the period to be unchanging, you must set rate variation to zero (dscale=0).

SuperCollider implementation by Nick Collins.

Class Methods

Gendy1.ar(ampdist: 1, durdist: 1, adparam: 1.0, ddparam: 1.0, minfreq: 440, maxfreq: 660, ampscale: 0.5, durscale: 0.5, initCPs: 12, knum, mul: 1.0, add: 0.0)

Gendy1.kr(ampdist: 1, durdist: 1, adparam: 1.0, ddparam: 1.0, minfreq: 20, maxfreq: 1000, ampscale: 0.5, durscale: 0.5, initCPs: 12, knum, mul: 1.0, add: 0.0)

Arguments:

ampdist

Choice of probability distribution for the next perturbation of the amplitude of a control point.

The distributions are (adapted from the GENDYN program in Formalized Music):

0:LINEAR.
1:CAUCHY.
2:LOGIST.
3:HYPERBCOS.
4:ARCSINE.
5:EXPON.
6:SINUS.

Where the sinus (Xenakis' name) is in this implementation taken as sampling from a third party oscillator. See example below.

durdist

Choice of distribution for the perturbation of the current inter control point duration.

adparam

A parameter for the shape of the amplitude probability distribution, requires values in the range 0.0001 to 1 (there are safety checks in the code so don't worry too much if you want to modulate!).

ddparam

A parameter for the shape of the duration probability distribution, requires values in the range 0.0001 to 1.

minfreq

Minimum allowed frequency of oscillation for the Gendy1 oscillator, so gives the largest period the duration is allowed to take on.

maxfreq

Maximum allowed frequency of oscillation for the Gendy1 oscillator, so gives the smallest period the duration is allowed to take on.

ampscale

Normally 0.0 to 1.0, multiplier for the distribution's delta value for amplitude. An ampscale of 1.0 allows the full range of -1 to 1 for a change of amplitude.

durscale

Normally 0.0 to 1.0, multiplier for the distribution's delta value for duration. An ampscale of 1.0 allows the full range of -1 to 1 for a change of duration.

initCPs

Initialise the number of control points in the memory. Xenakis specifies 12. There would be this number of control points per cycle of the oscillator, though the oscillator's period will constantly change due to the duration distribution.

knum

Current number of utilised control points, allows modulation.

mul
add

Discussion:

All parameters can be modulated at control rate except for initCPs which is used only at initialisation.

Inherited class methods

Instance Methods

Inherited instance methods

Examples

WARNING: if you have lots of CPs and you have fast frequencies, the CPU cost goes up a lot because a new CP move happens every sample!