Digital Signal Processing Reference
In-Depth Information
2
Randomization as a Tool
To achieve flexibility of the digitization techniques essential for
matching
them
to the given signal processing task conditions, these operations should be made
variable. The problem was to find out how to achieve this. Techniques existed
that were applicable for variations of signal processing procedures. It simply took
some time to realize that introducing an element of randomness into signal con-
version procedures, an approach initially developed for other purposes, also leads
to variations in the basic digitizing functions. Once that was understood, it became
clear that randomization might be used as a tool for making signal conversion
operations, including sampling and quantizing, more flexible and more adaptable
to specific signal processing needs. The application specifics and potential of this
instrument are discussed in this chapter.
However, the applicability and efficiency of this tool is arguable. Indeed, intro-
duction of randomness into signal processing operations might also have negative
consequences, as it usually leads to increased statistical errors. The end result
might therefore not be acceptable. In fact this often happens whenever random-
ization is carried out in a way that is not sufficiently skilful. The problem of
obtaining the targeted positive randomization effect without degradation of the
whole signal conversion process is crucial and is not easy to resolve. Work on the
development of new alias-free signal processing techniques should always be car-
ried out with considerations of this kind taken into account. Various approaches
used to obtain possible solutions of this problem are discussed in the following
chapters. This is the central issue in this topic.
Gradually the approach to randomization of the sampling and/or quantizing
operations and for taking care of the potential negative consequences of random-
ization has been improved to the extent that little remains from the techniques
