Digital Signal Processing Reference
In-Depth Information
7
Threshold-crossing Sampling
In cases where sampling is being randomized, this is usually done purposefully to
take signal sample values at specifically defined time instants in order to achieve
some specific desirable effects. When that happens, the sampling process is ran-
domized and is also nonuniform. In fact, randomization of sampling and nonuni-
form sampling are almost synonymous. However, deliberate randomization of
sampling makes sense only if such a complication of this operation is justified
by the end results. Apparently there has to be a good reason for randomizing the
sampling operation as the equidistant sample value taking procedure has many
outstanding positive qualities and is preferable whenever application of it is ap-
propriate. The motivation considered so far for randomizing of sampling has been
the widening of the frequency range for fully digital signal processing. That surely
is a good enough argument in favour of deliberate direct randomizing of sampling
as it leads to the irregularities of the sample taking process vital for elimination
of aliasing. However, avoidance of aliasing is not the only possible reason why
sometimes it makes sense to use this approach. There might also be some other
motivational factors, like complexity reduction of systems for data acquisition,
transmission and processing. Specific types of sampling randomizing, indirect
randomizing and targeting goals other than suppression of aliases are discussed
in this chapter.
The operational environment of the data acquisition systems often dictates
the requirements and defines the conditions for sampling. Sometimes there are
restrictions imposed on the accessibility to the signal sources. Then the signals
might be observed and their sample values taken only at random unpredictable
time instants. The obtained sequences of the signal sample values under these
