• FM vocal model places a carrier near each formant
  resonance, and modulates all carriers with a
  common oscillator at the voice fundamental pitch.
  
  
• Instruments with a noisy attack and exponential
  decay of sinusoidal partials are modeled well
  with one simple FM pair for the attack, and
  another pair for the decay. The nature of FM lends
  itself to the higher harmonics or partials dying
  away more rapidly than the lower spectral components.