There ain’t no limit to how impressive computer simulations of the real world can be. Ya only gotta switch on an SP3 or an Xbox 360 to see how far thing have come since since Pong hit the small screen in the 70s as a poor excuse for tennis.
But there are still plenty of simple things that even the most powerful computers get overheated about. And we’re not just talking nuclear stockpile management and the physics of supernovas.
Taizo “Taser” Kobayashi at Kyushu University in Japan says that the way musical instruments generate sound is one of these complicated things that computers just can’t get their microprocessors around.
And here’s the reason. The physics at work operates over many orders of magnitude so any direct numerical simulation from first principles soon explodes into a godforsaken mass o’ messy details. For example, the energy in the turbulent airflow within a wind instrument is 10^5 times greater than the energy of the sound field it radiates. Try simulatin’ that on a supercomputer and it’ll grind to a stutterin’ halt, a-spewin’ out steam ‘n’ flames.
But there’s another way, says Taser Kobayashi. Try dividing up the physical processes by scale into easily simulated packages. Then all yer supercomputer has to worry about is where these packages meet at the edges. It’s called a multiphysics simulation.
That kind of dividin up ain’t always possible: physics just don’t always divide easy. But Taser says it works fine for simulatin’ the sound a flute makes when you blow into it. He reckons it reduces the computational time by two orders of magnitude compared to simulatin’ from first principles. That’s impressive.
Where else could this work? Taser tentatively suggests simulating the behaviour of ion channels in biological cells in which the physics can be easily divided into atomic, molecular and cellular packages.
Any other suggestions?
Ref: arxiv.org/abs/0709.0787: Sound Generation by a Turbulent Flow in Musical Instruments – Multiphysics Simulation Approach