Image Processing Reference
samples. The time between each subsequent sample determines the highest recordable
frequency. The Nyquist sampling rule dictates that a sample rate is able to record frequencies
up to half the sample rate. Any higher frequencies will not be properly resolved.
At the sampling frequency, a peak and a trough may be recorded with two adjacent
samples. With no intervening values the frequency is captured but the wave shape is noth-
ing like what it originally was. The interesting wave shape is created by even higher-
frequency overtones and harmonics that combine. So even though we can sample that
very high frequency, we destroy the timbre.
When CD recording was first introduced, a sample rate of 44.1 KHz was thought to
be adequate since the human ear cannot detect frequencies higher than 22 KHz. Higher-
quality recording is achieved by sampling at 96 KHz and indeed sample rates of 192 KHz
are used in very high-quality production.
Movie sound tracks on DVDs benefit from the slightly better sample rate of 48 KHz,
and you will immediately see the relationship between 48, 96, and 192 KHz sample rates.
Conversion among them uses an integer-based averaging so if the production system runs
at 192 KHz, creating a 48 KHz track should be computationally easy.
Converting between 44.1 KHz and 48 KHz is not an ideal scenario, and a lot of work
has been done on the conversion algorithms to reduce the unpleasant artifacts.
Word Sizes (8, 10, 12, 16, 20, and 24 Bits)
The word size describes the number of bits used to describe a sample. The more bits you
use, the more accurate the sampling can be. The limited number of discrete levels or steps
A minimum of 2 samples
are required. More would
Figure 7-6 Harry Nyquist's sampling proposition circa 1920.