Newsletter

If you want to be informed on any new release or update, please subscribe to our newsletter :










Physical Modeling according to Devine Machine

Unlike other kinds of synthesis (additive, subtractive, FM...), Physical Modeling based synthesis is subject to interpretation and a great deal of ambiguity.

What is often called ‘Physical Modeling’ is a specific use of additive synthesis (based on standing waves theory) and parameter labeling that follow a physical logic.

What we call physical modeling at Devine Machine is the reproduction of physical reactions to create the sound.

To understand the difference, a comparative analogy can be made to image reproduction.

  • Sampling can be compared to photography.
  • Physical Modeling with additive synthesis can be compared to painting.
  • Physical modeling as a reaction can be compared to 3D.

 

Sampling

Additive synthesis
used for physical modeling

Physical modeling
based on reactions

like photography :

accurate, but only from one point of view, at a given moment, in a given place.

like painting :

color flexibility but limited realism and needs to be done over to change point of view or environment.

like 3D :

can be very realistic and let you change any aspect but requires more programming and computation.

with sampling, you have the reproduction of a sound in a specific recording context (gear, room) and under specific performance conditions (player, pickup position for a guitar, etc).

Additive synthesis recreates a sound by adding sine waves at different frequencies just as you would add colors when painting.

First step is the analysis of a given natural sound generator and the physics involved and how they affect the final sound.

Second step is the recreation of this reaction with programming.

+ accurate soundwise on single shots

+ no much programming involved

+ light on cpu

- not accurate on reproducing expressive behaviours

- heavy on hard drive space and ram

- poor dynamics

- sounds can't be changed, appart from filtering and distortion.

+ can sound very convincing

+ light on hard drive & ram

+ seamless dynamics

- inherent lack of flexibility, requires "tricks" to fit with physical parameters.

- some expression traits related to pitch alteration (bends) and other oddities (guitar squeaks) can't be realistic

- can't accurately match a given real sound reference

- dynamics require tricks, making it complicated to reproduce real behaviours

+ can match a given real sound reference (sounds like a sample)

+ parameters corresponding to the reality

+ seamless and realistic dynamics

+ realistic expressivity

+ light on hard drive & ram

- can be heavy on cpu

- requires specific researches


For those reasons, physical modeling as a reaction hasn't been much developed, but today's computers being more powerful, a whole new horizon is opening.

Here is an example of something only physical modeling as a reaction can let you do: speak through the model.