Audio: Capture (Sampling)
Sampling Theorem (Nyquist)
A signal must be sampled at at least twice the frequency of the
highest frequency to be represented.
quantization
Definition: For a given value x, chose the closest value to x from some
fixed set of possible values. (Fig 4.9, Buford, p.77)
- linear quantization
- map input range to a integer:
- 8 bit unsigned: 0 to 255
- 8 bit signed: -128 to 127
- 16 bit unsigned: 0 to 65535
- 16 bit signed: -32768 to 32728
- non-linear quantization
- eg) µlaw (mulaw)
- quantization noise
- due to quantization errors. sounds worse at low volumes.
reduce by using more bits per sample (noise reduced by 6 dB per bit
of resolution. 8 bits = 48 dB, 16 bits = 96dB, 20 bits = 120 dB)
Bill Thibault <william.thibault ayat csueastbay dawt edu>
Last modified: Sun Oct 15 20:59:08 1995