Seminar 4th December 2012 4 p.m. Building 13, Room 3021
Computational acoustics and large scale sound synthesis on GPUs
Stefan Bilbao
University of Edinburgh
- Categories
- Complex Systems
- Submitter
- Luke Goater
Abstract
Physical modelling sound synthesis and audio effects processing is concerned with the emulation (through simulation) of acoustic entities, ranging from standard acoustic musical instruments to room acoustics to analog electronic components and hybrid electromechanical systems, and finally virtual constructions without a real world counterpart. The goals are a) greatly increasing sound quality, b) offering simplified control, and c) allowing flexible new effect/instrument design exploration while retaining sound output with an acoustic character. In many respects it is analogous to similar developments in computer graphics, also based around increasingly sophisticated simulation techniques. In some regards, however, it represents a distinct challenge: among these are the large frequency range to which the ear is sensitive, requiring very small time steps in simulation, and the strongly nonlinear nature of even simple sound producing objects. Another major challenge is that of computational cost, particularly for systems in 3D, requiring HPC solutions using numerical designs constrained by the peculiarities of human audio perception. Sound examples and video demonstrations will be presented.
Speaker Biography
Stefan Bilbao received his BA in Physics at Harvard University ('92), then spent two years at the Institut de Recherche et Coordination Acoustique Musicale ( IRCAM ) under a fellowship awarded by Harvard and the Ecole Normale Superieure. He then completed the MSc and PhD degrees in Electrical Engineering at Stanford University ('96 and '01, respectively), while working at the Center for Computer Research in Music and Acoustics ( CCRMA ). He was subsequently a postdoctoral researcher at the Stanford Space Telecommunications and Radioscience Laboratory, and a Lecturer at the Sonic Arts Research Centre at the Queen's University Belfast.
He currently works on sound synthesis based on physical models of musical instruments, with a particular focus on mainstream numerical simulation techniques, such as, e.g., finite difference methods. Special topics of interest include: Hamiltonian and symplectic methods, distributed nonlinear systems such as strings and plates, estimates of computational complexity, multichannel sound synthesis, and hardware and software realizations.
A sideline is joint work with composers of electroacoustic music. Other interests include virtual analog effect modelling and artificial reverberation design, modelling of electromechanical instruments, shock wave propagation in tubes, numerical absorbing boundaries in acoustics applications, and large-scale audio rendering using graphics processing units (GPUs) in conjunction with the Edinburgh Parallel Computing Centre.