Synthesis 2:
Signal models

·   Sampled waveforms

fidelity:	excellent (but.. unnatural repetition?)
controls:	very few (level & resampling rate)
cost:	simple CPU / lots of store

- enhancements to sampling:
+ looped sections for simple 'sustain'
+ mix 2 or 3 samples for timbre 'space'

·   Sinusoid models

- exploit harmonic structure of pitched sounds


fidelity:	good to excellent
controls:	pitch and timescale well separated
cost:	moderate++ CPU / large store

- parameter extraction is straightforward
- additional 'noise' residual for non-pitched parts

·   Nonlinear models (e.g. FM)



fidelity:	pleasant sounds but limited scope
controls:	good range but unpredictable
cost:	moderate CPU / little store