Time resolution and frequency resolution in windowed analysis are inversely constrained

Gabor formalized an uncertainty relation for acoustic analysis analogous to Heisenberg’s in quantum physics: the product of time resolution delta-t and frequency resolution delta-f cannot be less than 1. A short analysis window gives good time resolution (you know precisely when an event occurred) but poor frequency resolution (you cannot resolve close frequencies). A long window gives good frequency resolution but blurs the time location of events. This is not a limitation of technology but a mathematical fact - it applies to any windowed transform including FFT, wavelets, and Gabor analysis. In granular synthesis, grain duration directly embodies this tradeoff: short grains are temporally precise but spectrally broad; long grains are spectrally narrow but temporally smeared.

Examples

A 1ms grain at 1000 Hz has a bandwidth of ~1000 Hz (very broad spectrum). A 100ms grain at 1000 Hz has a bandwidth of ~10 Hz (narrow, well-defined pitch).

Assessment

A producer wants to design grains for a granular synthesizer. If she wants each grain to sound like a recognizable pitched tone, should she use 1ms or 50ms grains? Explain using the Gabor uncertainty relation.