Satgen 375  Why FFTs Pt2  by GM4IHJ                    1st June 1996

  As we attempt to probe deeper and deeper into radio noise , in order
to pick out weak signals, it is perhaps useful to look at what we can
do to improve our detection capability. As mentioned in Satgens 373
and 374, we can now use the mathematical technique of Fast Fourier
Transforms for Digital signal processing. This is a recent innovation
consequent on the availability of fast computers.
  Try using AF9Ys FFTDSP42 software on a computer with an Intel 286
central processor chip, and you will find the program runs very
slowly, with the CPU time report, showing overload of its calculation
capabilities. Even the Intel 386 chip is not fast enough for the
number crunching needed, though it will be marginally adequate if
supplemented by a maths coprocessor chip. Go to a computer with a fast
Intel 486 and the program runs easily, and on the still faster Pentium
90 Mhz the CPU shows only about 10% usage, which allows concurrent
operation of other tasks, such as satellite tracking.
  But what advantage do we get from FFT software ? As mentioned in
Satgen 374 the FFT procedure allows examination of the signal spectrum
in a very narrow bandwidth, Something impossible to achieve without
Digital signal processing. The advantage of narrow band filtering is
clear if we examine the performance equation for a space radio link :-

Sig to noise = (Txpower*Txant*Rxant)/(k*T*(R^2)*(wavelength^2)*Bwidth)
      strictly true only where pulse duration matches bandwidth  
 where Txant = transmit antenna effective area metres^2
       Rxant = receive antenna effective area metres^2
           k = Boltzmann's Constant  1.38 * 1E-23
           T = system temperature in degrees Kelvin
           R = path length metres
      Bwidth = bandwidth in Hz
 
  Considerable progress has been made in reducing temperature T. We
can only go so far in terms of antenna size and signal wavelength.
However reduction of bandwidth from say the 100 Hz of a mechanical or
Xtal filter, down to the 2 Hz of FFT processing gives us a lot of
improvement.
   So where are the snags ? How far can we reduce bandwith . In space
the problems of multiple scattering in the interstellar medium,
probably set a minimum usable bandwidth of 0.1 Hz . But that would
require tremendous computer power and takes no account of the problems
created by the ionosphere. It also still leaves us with the problem
that the integration time used by FFTDSP42 software is too long for us
to be able to read signal modulation. Which does not matter if we want
to see the constant carrier of the Mars Probe beacon, but it is no
good for actual communications work.
   It is therefore likely that future progress will centre around
designing digital systems which are optimised for the modulation
characteristics of the signal, both at transmission, and after it has
been modified by its passage through space, the ionosphere, and, in
the case of Moonbounce signals, after non linear reflection from the
lunar surface. To this end , experiments to examine just exactly what
happens to signals from space , are essential if we are to progress
further.