Content area
Abstract
Physical modeling of musical instruments has been a growing field of research since the early 1980s. At the time, the computational costs of physical models did not allow for real-time sound synthesis. Today, computational costs for real-time synthesis are not very prohibitive, and it is possible to implement more complex models with enough efficiency for real-time synthesis. Digital waveguides provide an efficient way of modeling one-dimensional wave propagation and implementing such physical models for sound synthesis.
While physical models of well-known instruments are ubiquitous in the literature, models of some lesser known traditional instruments from different cultures around the world are not. This thesis aims to provide a model of one of such instruments, the Colombian gaita. In this thesis, the acoustic response of a pair of gaitas was studied and a basic physical model of a gaita was developed and implemented for real-time synthesis using digital waveguide techniques found in the literature.
In order to determine if a cylindrical model would be an accurate representation for the instrument, impedance measurements were made on a pair of gaitas after their heads were removed. These measurements were then compared with the theoretical input impedance and reflection function results obtained using a transfer matrix model with the physical dimensions of a gaita. The results show the cylindrical approximation to be reasonable. While the results showed a similar frequency response and reflection function, they also showed some discrepancies in the amplitude of harmonic peaks, and the end reflection of some open hole configurations of the model.
The digital waveguide model was developed and implemented in MATLAB in order to compare its response with the measurements and theoretical results. The reflection function comparison showed, as expected, some discrepancies between the model and the measurements, particularly in the end reflection of certain open tonehole configurations. The model was then implemented for real-time sound synthesis with a graphical user interface using the JUCE framework and the Synthesis ToolKit. A time- and frequency-domain comparison of the sound synthesized by the model and recordings of the instrument show that while the model conveys the basic attributes of the sound of the instrument, it fails to accurately capture some other sonic qualities that are characteristic of the instrument. The most prominent differences between the model and the real instrument sounds are found to be the magnitudes of the higher-frequency harmonics and the overall noise content.





