1.1.3.

Sound
The students should be able to:
✗ show understanding of how sound is represented
and encoded
✗ use the associated terminology: sampling, sampling
rate, sampling resolution
✗ show understanding of how file sizes depend on
sampling rate and sampling resolution
✗ show understanding of how typical features found in
sound editing software are used in practice

2
 1.
Sound
Natural sound consists of variations in pressure which are detected by the
human ear. A typical sound contains a large number of individual waves each
with a defined frequency. The result is a waveform in which the amplitude of
the sound varies in a continuous but irregular pattern.
If there is a need to store sound or transmit it electronically the original
analogue sound signal has to be converted to a binary code. A sound encoder
has two components. The first is a band-limiting filter. This is needed to
remove high-frequency components. The ear would not be able to detect
these and they could ca use problems for the cod ing if not removed. The other
component in the encoder is an analogue-to-digital converter (ADC).

4
The method of operation of the ADC is described
with reference to Figure 1.08. The amplitude of the
wave (the red line) has to be sampled at regular
intervals. The blue vertical lines indicate the sampling
times. The amplitude cannot be measured exactly;
instead the amplitude is approximated by the closest
of the defined amplitudes represented by the
horizontal lines. In Figure 1.08, sample values 1 and
4 will be an accurate estimate of the actual amplitude
because the wave is touching an amplitude line. In
contrast, samples 5 and 6 will not be accurate because the actual
amplitude is approximately halfway between the two closest defined values.

5
In practice, for coding sound, two decisions have to be made. The first is the
number of bits to be used to store the amplitude values, which defines the
sampling resolution. If only three bits are used then eight levels can be
defined as shown in Figure 1.08. If too few are used there will be a significant
quantisation error. In practice 16 bits will provide reasonable accuracy for the
digitised sound.
The other decision concerns the choice of the sampling rate, which is the
number of samples taken per second. This should be in accordance with
Nyquist's theorem which states that sampling must be done at a frequency
at least twice the highest frequency in the sample.
Once again file size can be an issue. Clearly an increased sampling rate and an
increased sampling resolution will both cause an increase in file size.

6
Simply recording sound and storing a digital representation is not enough
for many applications. Once a digital representation of the sound has been
stored in a file, it can be manipulated using sound-editing software. This will
typically have features for:
X combining sound from different sources
X fading in or fading out the sound
X editing the sound to remove noise and other imperfections.

7
Thanks!