Page 33 - RTH3BA Preview

R���� T����� �������� - B������� �� A������� 3 �� E������.
DIGITAL SIGNAL PROCESSING (DSP)

WHAT IS DSP?

Digital Signal Processing (DSP) takes real-world signals like voice from a
microphone, video or images from a camera, ASCII keystrokes from a computer
keyboard and digitises them. Analogue signals are digitised by an analogue-to-
digital-converter (ADC). When these signals have been digitised, they can be
mathematically manipulated by software and/or hardware. When we have
processed the digital signal we convert it back to analogue using a digital-to-
analogue-converter (DAC).
Digitised signals could be displayed and analysed in the time or frequency domain
(like on a bandscope). We could convert from one signal protocol to another. For
example, keystrokes could be converted into one of the many variations of FSK.
Digital signals can be compressed for more efficient transmission. Analysed to
identify noise and remove it. We can create many digital protocols (modes) and
transmit them via radio. At the receiver, we generally need, if human ears are being
used, to convert the digital signal back to analogue or display them text or images
on a screen.

The principles of DSP have been around and understood since the 1960s. However,
it wasn’t till recent times that the computer or hardware processing speed has
become fast and cheap enough to make DSP viable.

SAMPLING

An analogue signal is a “continuous” signal. If we want to convert an analogue signal
to a digital signal, we have to sample the signal against time at regular intervals so
that we can store those discrete measurements in a digital form. An analogue signal
has an infinite number of points. There is no way an infinite number of points can be
stored in digital media.

Signal sampling is the process of taking discrete measurements against time of a
continuous analogue signal.

Figure 35-18 show an analogue signal that has been sampled (discretised). This
signal can be totally reconstructed if the sampling rate is sufficient.

What if the sampling rate is too low?

In Figure 35-19A we see a sinewave being sampled once every three quarters of a
cycle. Figure 35-19B shows the sampling points with the sine wave removed. In
Figure 35-19C when we reconstruct this signal, it is very obvious that the
reconstructed signal is nowhere like the original. This error in reconstruction is called
aliasing. It is called aliasing because we say the signal at C is an alias of signal A.

To avoid aliasing we need to ensure the sampling rate is fast enough.
PREVIEW

440