Duplicating the Human Voice
The human voicebox creates sound waves by compressing and decompressing air. The process creates pressure waves that our ears interpret as vibrations and our brains interpret as sound. The production of each pulse of pressurized air can be thought of as a cycle where the pressure exerted by the air goes from highest pressure to lowest pressure. The number of times the pressure level cycles from low to high and back is measured in 'Herz'. The word Herz (abbreviated as Hz) is used as shorthand for 'cycles per second' and is used as the unit of measure for (amongst other things) the pitch of the human voice. The term Herz is capitalized because it was a scientist named Herz who first put down on paper the cycles per second concept.
The the human voice has a range of approximately 100 Hz to 5,000 Hz (5 kHz), but through experimentation, the phone company discovered that the recognizeable frequencies are between 100 hz and 3,300 Hz. The phone company reproduces the frequencies from 30Hz to 3,300Hz.
In order to reproduce the human voice using digital information, you first have to sample the incoming voice information that is transmitted as an analog electrical signal on the subscriber's local loop. A scientist and mathematician figured out that you need a sample rate equal to twice the highest frequency of sound you are trying to duplicate. Since the recognizeable part of the human voice peaks at approximately 3,300 Hz, you would need a sampling of at least twice that rate to reproduce it later from your samples. To keep the voice clean and clear, the phone system needs what is called 'guardband'--a small gap at the top and bottom of the frequencies used for human voice. The frequencies between 30 hz and 100 Hz and between 3,300 Hz and 4,000 Hz comprise the guardband.
By sampling the voice line 8,000 times per second (twice the maximum rate expected on the voice line), the phone company is able to re-create a very life-like copy of the human voice without wasting too much bandwidth.
It is because of this sampling, and the guardband, that people's voices sound the way they do on the phone.