Hi Paul and all,
Thanks for the notes about the importance of the impulsive noise blanker.
Since there are still a few windows users on the group, here's how to
activate the noise blanker in spectrum lab,
assuming that one of the "very slow" morse modes has already been picked
from the 'Quick Settings' menu:
Select 'Components' / 'Circuit window',
click on the 'DSP blackbox' between circuit nodes L4 and L5,
select 'Noise blanker',
make sure to connect the spectrum analyser to 'L5' (not 'L4', which
would be "before" the noise blanker).
The NB defaults should be a good starting point, but as Paul pointed out
we may need more aggressive blanking.
This is controlled by the "Trigger Peak/ Average" settings (10 dB by
default, "the less, the more aggressive").
Details about the noiseblanker settings:
http://www.qsl.net/dl4yhf/speclab/circuits.htm#noiseblanker
To remove signals which may irritate the NB (like strong MSK
transmitters or local QRM), the FFT-based filter can be switched into
the signal path between 'L3' and 'L4'.
For a quick start, begin with a simple bandpass, say 3 kHz bandwidth,
centered around 29.5 kHz.
Details about that filter:
http://www.qsl.net/dl4yhf/speclab/filters.htm#fft_filter
As soon as my work on the WSQ program has finished, I will make a
configuration file which works "reasonably well" for this location, and
post it on a website. But I think the settings used below 9 kHz will
work reasonably well, but don't forget you need soundcard providing at
least 96 kSamples/second (select that sampling rate for input in the
audio settings).
Good luck everyone,
Wolf .
Am 03.03.2014 18:54, schrieb Paul Nicholson:
It is very important at VLF to use effective sferic blanking
when looking for weak amateur signals. There are 10 or 20 or
more sferics per second and the wanted signal will be lost in
the noise when the energy of all those sferics is distributed
across your Fourier bins.
Let me illustrate with a plot. Here is a one hour spectrum
of Bob's signal with and without sferic blanking,
http://abelian.org/vlf/tmp/29499_140302n.gif
As you can see, without the blanking even this strong signal
is completely buried by the noise.
Effective blanking will discard between 25% and 35% of the
incoming signal but will boost the S/N ratio by 20dB or more.
The blanking threshold must be set very low to achieve this,
roughly 1.2 to 1.5 times the mean level is the optimum.
The mean noise floor is conveniently tracked by an exponential
moving mean of the absolute signal value (I use time constant
of 1 second or so but longer is fine).
Some important caveats:
First, the input to the blanker must be free of MSK signals, hum
harmonics, and any other continuous signals when viewed in the
time domain. Two bad things happen otherwise: the blanker mean
level tracking will be foiled by the continuous signals, and
the chopping action of the blanker will spread the continuous
signals across the band to contribute to the noise floor.
So, precede the blanker with a filter not wider than say 3 or
4kHz and include notches for any significant mains harmonics
or MSK signals that remain within the passband. If you can
see them against the noise in the time domain then they need
to be notched out.
Second - the blanker must see a clear sferic in order to work,
so the preceding filter should not be too narrow. 2kHz to 4kHz
is enough. 1.5 kHz is starting to get too narrow. The passband
doesn't have to be centered on the rx frequency, it can be
offset to avoid including some inconvenient continuous signal.
While a bit of lightweight sferic blanking is a nicety at
higher frequencies, at VLF some serious blanking is essential
for detecting weak signals. Nearly every amateur signal I
receive is well beneath the un-blanked noise floor.
--
Paul Nicholson
--
|
