Digital Signal Processing Reference
In-Depth Information
Figure 4.3
Illustration of randomized quantizing
that signal value might be rounded off to various digital levels. As a result, the
properties of a randomly quantized signal differ significantly from the properties
of a deterministically quantized signal.
Note that there are two seemingly different approaches to randomization of
quantizing. While in the quite popular cases of so-called dithering, mentioned
in Chapter 1, noise is added to the signal at the input of a deterministic quan-
tizer, the quantization operation could also be randomized as explained above by
just forming time variable sets of reference levels and using them for rounding
off the input signal instantaneous values. Electronic implementations of these
randomized quantization schemes are indeed different. However, the essence of
them is equivalent. It does not matter exactly how the randomness is introduced
at quantizing. The results are exactly the same.
4.1.2 Input-Output Characteristics
The essentials of various quantization models can also be illustrated by spe-
cific representation of the quantized signals, including their input-output charac-
teristics, shown in Figure 4.4. While deterministic quantization has one fixed
input-output relationship (Figure 4.4(a)), randomized quantization, in gen-
eral, is characterized by different input-output relationships depending on the
