Linear and non-linear processes - Signals, Processes, and Systems - page 183

A complex component: the transistor
The most important component in microelectronics is no doubt the transistor. Unfortu-
nately, transistors have in general non-linear characteristic curves for physical reasons. It
is therefore rather difficult to develop circuits with transistors which work very precisely,
even though the message conveyed in classes and lectures is a different one. It requires
experts to build a linear transistorized amplifier that amplifies in a way which „maintains
the proportions“. Such an amplifier can to a certain extent be linearized by means of
numerous circuit technology tricks - the best trick is called negative feedback . But for the
above reasons no amplifier is completely linear. The so called distortion factor is a
measure of its non-linearity.
First consequences :
The above Illustrations and explanations show in what cases linear and non-linear
processes are generally applied:
If the frequency range of a signal is to be - or can be - changed,
e.g. by relocating it to a different range, it can only be achieved
by using non-linear processes.
If the frequencies contained in a signal are not to be changed or
new frequencies added this can only be achieved by using linear
processes.
This means that linearity and non-linearity are about whether or not new frequencies, e.g.
sinusoidal oscillations, are created in a process.
Unlike mathematics, communications technology deals with linearity and non-linearity
only in the context of sinusoidal oscillations. This is understandable enough as according
to the underlying FOURIER Principle any signal can be regarded as composed of
sinusoidal oscillations.
Almost every signal processing causes a change in the time and frequency domains. The
link between these two changes is (of course) the FOURIER transformation, which is the
only way of getting from the time into the frequency domain and vice versa. All the results
of the signal processing described below will therefore be described in the time and in the
frequency domains.
There are only few linear processes
The situation is quite clear. There is just a total of five or six linear processes and most of
them appear ridiculously simple at first sight. But they are still enormously important and
appear in numerous applications. In contrast to the indefinite variety and complexity of
non-linear processes, we know almost everything about them and there are hardly any
surprises, even when several linear processes are combined to form a linear system.