The Theory of Sampling
This help page deals with the process of converting audio
signals present in an analog form to a digital stream of
data which a computer can handle. There are some aspects
of this conversion that should be known if you are planning
to do some work with audio processing.
From Analog to Digital
On the output from an analog audio source (tape recorder,
microphone, or even the standard output from CD players)
the audio signal is represented by an alternating electrical
voltage. The voltage is analog to the air pressure changes
in the air (which we perceive as sound), hence the term analog
signals. This constantly changing electrical voltage is
changed to a stream of numbers at fixed rate intervals by
sampling and quantisation.
Sampling
The conversion from an continuously changing measure to a
series of measured values at discrete time instances is called
sampling. The rate (number of measurements per second) of which
the sampling is done, is along with the quantisation depth the
most important quality factor of digital recording equipment.
If it is set too low, disturbing artifacts occur. In fact, all
frequencies above half the sampling frequency (known as the
Nyquist frequency) are substituted by frequencies below the
Nyquist frequency. This effect is called aliasing. To avoid
aliasing, a sampling system contains of a low pass filter,
ideally filtering out all frequencies above the Nyquist frequency
and leaving all frequencies below unaffected. A CD quality recording
is recorded with a frequency of 44,1 kHz.
Quantisation
The conversion from a measurement described with a continuos variable
to described with a discrete variable is called quantisation. The
number of bits used to describe one measurement is proportional to the
highest achievable signal to noise ratio, which is a measurement for
the noise present in the system.