h2g2 The Hitchhiker's Guide to the Galaxy: Earth Edition

"the traditional method of speeding up audio, known as re-sampling,had the side effect of causing a comparative increase in the pitch of the sound, similar to the increase in pitch that comes with speeding up a vinyl or tape."

the traditional method to speed up a sample was to play it at a higher pitch. raising it by an octave will double the speed, lowering it will halve the speed.

re-sampling only re-records the sound back into the sampler along with with any processes or effects to create a new sample.

changing the sample's sample rate (22.05kHz, 32kHz, 44.1kHz, 48kHz etc) without re-sampling will affect the speed of the sample with a negligible change of pitch. increasing the sample rate will speed up the sample and vice versa. decreasing the sample rate can give a nice zingy, choppy effect used on vocal samples in many dance tunes.


-MyC

"the traditional method to speed up a sample was to play it at a higher pitch. raising it by an octave will double the speed, lowering it will halve the speed." - MyC

the process of playing digital audio at a higher pitch is distinguished from more recent frequency independant time streatching by the term 'resampling'. If you look closely the key to this is there in your point about re-recording.

the method, although done algorithmically by the computer, is to 'play back' the sample at a higher pitch (higher clock rate), and encode the resultant data at the original sample rate.

it is important not to confuse sampling rate (the frequency with which incoming or generated audio would be 'sampled' -ie. the rate at which data are placed on the amplitude-time graph, with 'clock' rate, the rate at which the computer chooses to play back the samples.

Playing back the sample with a higher clock rate whilst re-sampling it at the original sampling rate speeds up the audio in the manner of speeding it up on a deck or tape (equation for speed change (%) necessary for pitch change is 2^(x/12) where x is the semitone difference.)

reducing sample rates causes artificial thinning of data on the amplitude-time graph.
Contrary to your assertion, this does not cause frequency-independant stretching, but a reduction in the highest audible frequency (the Nyquist frequency). The rate at which a sound is sampled must be at least twice that the highest frequency component to be captured. (Nyquist's theorem). The results of a sample rate change tend to suffer from quantisation distortion, eg. the sound at the beginning of Aphex Twin's 'girl boy song'.

Pitch independant timestretching is an altogether more complicated affair, and some very clever and complex algorithms have been invented to do it. This is why re-sampling tends to sound the same whatever program you use, whilst dedicated time stretching software can cost up to 600 pounds at the proffessional level.
I'll maybe add another note about time-independant time stretching algorithms when i'm less hungover.


metasystem

ps. i think i remember reading that the on the fly way that samplers do re-sampling across a keyboard is a little more complicated due to the need for cutting corners in a real-time environment. I'll check my audio book when i wake up properly..

meta