Notes · Speech from first principles
Glottal source
also: glottal flow · the source
The stream of air puffs produced by the vocal folds opening and closing ~100 times a second: a volume-velocity waveform, and on its own a buzz rather than a vowel. Two properties matter downstream.
Its spectrum is a comb — energy only at multiples of the fundamental f₀, the spacing carrying the pitch. And its envelope is smooth and boring, falling at about −12 dB/octave, set mostly by how abruptly the folds slam shut. What it does not carry is any structure distinguishing “ah” from “ee”: that lives entirely in the vocal tract.
This clean division of labour is the content of the source–filter model, and it is why a recogniser can afford to discard the source and keep only the formant envelope. Fricatives break it — their source is turbulence at a constriction partway up the tube, not a periodic glottal buzz.