Dear Andy, LF Group,
At 14:54 24/01/2002 +0000, you wrote:
All three of these modes use the same basic technique, looking for the
presence or absence of a signal in a given bandwidth - in the case of JASON
doing this 17 times in parallel. The bandwidth that matters is not the
total bandwidth of the whole signal, just that occupied by one of the tones,
and that is defined in the decoding software...
So, in principle, if one was looking at a spectrogram of a signal made up
of dots of a certain length, with evenly distributed noise, it would not
matter from the viewpoint of detecting them whether they were all in a line
(QRSS) or distributed over 17 different frequencies (Jason), provided you
knew what the 17 frequencies were. The Jason signal is equivalent to 17
separate QRSS stations on different frequencies, using a funny type of code
arranged so that only one station is transmitting at any one time. So the
ability to detect this kind of signal with a given signal power and noise
power spectral density only depends on the length of the dots and not at
all on the number of frequencies used. But using a larger number of
frequencies enables faster transmission of data because then a larger
number of different codes can be made up from a given number of dots. So
Jason should work with the same signal and noise levels as QRSS using 12s
dots, but at the same time be 5 times quicker - is this correct?
I suppose there are practical restrictions on this - for example, you have
to find 4Hz of bandwidth with no narrow band noise in it, like Loran lines,
in order to make the scheme work.
A machine can decide on the presence or absence of a signal in a defined
bandwidth a lot better than a human eye can on a waterfall - it most
definitely can, believe it, its true, even if this fact makes you feel
uncomfortable :-((
I can believe that machines are better at detecting a small change in level
of signal/noise than a human operator, but with the simple encoding schemes
we are using, the human operator has the advantage in that they can quite
often "guess" what the data should be when it is corrupted. But then you
could argue that error correcting codes of one sort or another achieve
similar results for machine decoding.
Cheers, Jim Moritz
73 de M0BMU
|
