Automated Turbulence

In wind instrument models, turbulence is generally modeled using ad hoc filtered noise injection. It would be considerably more convenient, and presumably more accurate, to generate the turbulence automatically using a model for turbulence generation. For example, given air velocity and mouthpiece geometry, the noise spectrum can sometimes be predicted fairly well from theory. Early ideas along these lines were pursued in articulatory speech synthesis [23]. A recent turbulence model for flue pipes based on the Lighthill theory is described in [73], and more extensively in [72, Chapter 5]. Another based on sound produced by vortex shedding was proposed in [10]. An advantage of automatically generated noise due to turbulence is that pulsed modulation at the pitch rate falls out automatically [9].