Didge(R)Evolution    
Computer Aided Didge Sound Design
Frank Geipel, 21th September 2008

After 6 years of private theoretical and experimental (low budget) research driven by curiosity and the longing after individually playable desired instruments a highlight of the common Didjeridu research project (Geipel / Reimer) is reached now.

Many Didjeridu players from all over the world want to know whether it is possible to create an instrument after its individual desired conceptions. Since some years that is possible with our Computer-Aided-Didjeridu-Sound-Design-tools (CADSD), but required much experience and simulation time, since many interior forms had to be adjusted and varied by hand, to get the desired relevant impedance and sound spectra.

So far it was not possible to find the interior forms for desired spectra in a more efficient and automatically way. In order to reach this, I started at the beginning of 2008 the project „Didge(R)Evolution“- the application and advancement of nature modelled evolution algorithms for the generation of desired forms from practically immense, infinite large variety of possible forms.

The current project highlight is a simulation system, which is able to generate desired instruments on basis of an advanced high performance simulation model of the Didjeridu acoustics and using new Didjeridu specific directed-evolution-methods. It runs like a “living evolution system” in which special instruments with many specific parameters can quasi “breed”. The so created interior forms often looks similar the cross-sectional contours of good termite-carved instruments, with the difference, that the method is able to generate also forms which do probability not occur in nature.


 

w a t c h    v i d e o s

Simplified description of the CADSD based Didge(R)Evolution software

Since only few Didjeridu building interesting people have experiences with mathematical modelling, programming and physical theories, here a simplified description of the Didge(R)Evolution software:

Before the program can take up its work, a goal of a sound- and an impedance spectrum exact as possible must be defined. That can be momentary up to 20 different characteristics.

Imagine you then, we create (program) virtual termite swarms, whose single termites have individual abilities to form Didjeridus with special characteristics. In the next step the size and number of this virtual termite swarms have to be fixed - in our animated example is it only 1 swarm with 512 termites.

With the start of the program each individual termite takes up its work and eats (first times completely randomized) its own Didjeridu within the spatial delimitation of the virtual blank tree. After some seconds the first 512 instruments are finished.

Now each individual instrument is tested how well it fulfils the given goal characteristics. The termites, which were most successful, may reproduce and pass their genetic information on to a new termite generation. The others become extinct. Additionally, similarly as in nature, coincidental mutations are produced, which change the individual characteristics of individual termites. Thus the first generation is terminated.

In the second generation the descendants as well as its relatively successful parents and mutated brothers and sisters eats out again 512 Didjeridus from virtual blank trees again randomized however with improved individual characteristics to reach the goals. Afterwards again all instruments are tested. The successful termites of the swarm may reproduce again, while the remainder becomes extinct.

This evolutionary cycle is repeated until a virtual Didjeridu was produced with all defined goals in the best possible way fulfilled and no evolutionary progress is reached - in the example visibly as the red line. Usually after 150 to 400 generations this is the case.

In order to produce outstanding and special instruments with this directed evolution method, the know-how of the “Didjeridu breeder” to formulate the goals is very important. Extensive knowledge about the interactions of most diverse intrinsic resonance pattern with different play techniques on the effects of the sound spectra are of crucial importance, but also extensive building and playing experiences. The results depend on the quality of the defined goals.

back to top