Starting transient of the flow through an in-vitro model of the vocal folds

An exact theory of voiced sound production can be useless. An exact numerical calculation is, in fact, a detailed experiment, which does not by itself provide any insight. For example a detailed numerical simulation cannot be used to obtain data compression or real-time sound synthesis. By contrast the most simple models available such as source/filter models are already quite versatile tools. As the available computational power increases we would like to use more complex models. Physical modeling is a guide to generate models with a limited number of parameters with limited ranges of plausible values, determined from physiological data. A physical model should have a reasonable balance in the degree of sophistication used to describe the various elements such as: mechanical system, flow, articulation, and so on. The aim of our long-term research is to test the most simple fluid dynamic theories by means of accurate in vitro experiments, providing the specialist with a range of models rather than a specific model. Our research program was triggered by request of colleagues from the Institute for Perception Research anxious to evaluate the relevance of the work of Teager [1,2]. Based on our earlier experience on natural gastransport systems and on musical acoustics we started by a literature study [3]. We certainly agree with Teager that a more systematic description of the flow can be useful. However, some of his research proposals are highly disputable. For example, it may be fascinating to place a hot wire in our mouth and observe the complex flow signal obtained during phonation. However, we will never be able to relate such data to any quantitative theory because we have no information about the flow channel geometry nor on the position of the hot wire. This makes in vivo experiments with hot wires rather frustrating. As fluid dynamicists we were much more impressed by simple in vitro experiments with fixed geometry [4,5] or oscillating models [6,7]. Based on the difficulties that we had already encountered in oscillating valves models [8], we decided to focus on fixed rigid geometries. We decided to introduce an unsteadiness of the flow by driving the model with a valve as we had done earlier for the organ pipe [9,10] . As we knew how difficult quantitative measurements are, we started by considering a smoothly converging two-dimensional flow channel with a sharp-edged termination. Steady and unsteady flow measurements were carried out in this geometry to check the calibrations of our pressure gauges and of our hot wires. (Even steady wall pressure measurements should always be checked because of potential problems with pressure holes.) While the sharp edged geometry is not physiologically relevant it allows detailed two-dimensional point vortex simulation because we know that vortex shedding occurs at the sharp edges and is described by a Kutta condition [11]. We actually are still working on this reference geometry. It is rather obvious that the glottis is smoothly shaped and that we should therefore develop a theory describing the flow through lip-like channels. A first step in this direction was carried out by Belfroid [12] followed by Pelorson etal. [13-15]. Using this set-up we now produced quite realistic flow pulses as shown in Figure 3-1. In the present chapter we give an informal description of the results we have obtained until now. More precise information both concerning the theory and the experiments is found in the papers previously quoted. We focus on the glottis. The interesting subject of the aeroacoustics of the vocal tract is not considered. A discussion of literature available on this subject is given in our earlier paper [3] and the recent review of Davies etal. [16].