Dear group,
It seems that there are interesting theoretical questions to answer. And we,
the practical interested's, could do this by trying that. Anyway if it is over
10km or 200km... Maybe we'll get some nice surprises...
Stefan/DK7FC
________________________________
Von: [email protected] im Auftrag von Alexander S. Yurkov
Gesendet: Sa 27.02.2010 02:47
An: [email protected]
Betreff: Re: LF: Re: 9kHz noise level
On Fri, 26 Feb 2010, Andy Talbot wrote:
> probably end up appearing quite close to Gaussian. Certainly, on 73kHz
> the impulsive Loran interference and natural spikes appeared completely
> Gaussian when filtered to the sub-Hz bandwidth we used there.
Andy,
certanly AFTER filtering noise became gaussian. But matter of fact is that
if noise is nongaussian then to get optimal reception one should make some
nonlinear processing (limiting in simplest way) BEFORE narrow band
filtering. After filtering when noise is gaussian limiting or someone
else nonliniar processing has no sence. There is exact matematical
theorem: if noise is gaussian then optimal signal processing is linear
filtering (or correlator calculation in coherent case).
> However, for 9kHz signalling, I think you would do a lot better by going for
> coherent signalling.
IMHO coherent signaling improoves SNR by 3 dB. It does not depend of
the fact if noise is gaussian or nongaussian. Anyway this (coherent
signaling) is linear processing nonoptimal when noise is nongaussian.
Interesting quation arises there. Is any optimal form of the signal if
noise is nongaussian? I don't know answer. This is high science. But
if noise is gaussian then form of the signal has no meaning. Only energy
of the signal has meaning if noise is gaussian (with constant spectrum).
If filter is optimal for specific signal of couse.
Regards,
Alexander/RA9MB
<<winmail.dat>>
|
