Envelope-to: dave@picks.force9.co.uk
Delivery-date: Tue, 24 May 2005 09:53:49 +0100
Message-ID: <003001c5603e$24e42ee0$1101a8c0@w2ksn>
From: "Stewart Nelson" <sn@scgroup.com>
To: rsgb_lf_group@blacksheep.org
Date: Tue, 24 May 2005 01:54:06 -0700
MIME-Version: 1.0
Subject: LF: reception with multiple antennas
Content-Type: text/plain; charset=iso-8859-1; format=flowed
Content-Transfer-Encoding: 8bit
Sender: owner-rsgb_lf_group@blacksheep.org
Precedence: bulk
Reply-To: rsgb_lf_group@blacksheep.org

Hi all,

I've been dreaming about a system tentatively named MARS (Multi
Antenna Reception System).  The basic concept is very simple.  LF
signals are simultaneously captured at several stations in the same
general area.  A client application running at each station sends the
demodulated output via the Internet for storage on a server.  Upon
request from a user (who knows the time, frequency and other
parameters of an expected DX transmission), the server combines the
signals to enhance the S/N, and the result is displayed or played as
audio.

For common signal formats such as CW, QRSS, and WOLF, it should be
easy to coherently combine the signals, even when individual capture
S/N values are several dB below the readability threshold.  I also
believe that noise received at stations tens or hundreds of kilometers
apart is largely uncorrelated, so combining four stations should
improve S/N by almost 6 dB.  I think that such a system would be
workable for up to ten stations.  That should be enough gain to pull
off a number of tricks, such as decoding a message from ZL, receiving
a T/A signal sent with Part 15 rules, or letting Mal hear normal CW
from across the pond.

For signals to be coherently combined, the client systems must be on
exactly the same frequency and time.  This can be accomplished without
any special hardware, by using the LORAN "QRM" present in the received
signal.  I've suggested that before, but it wasn't taken seriously.
Here is some evidence that the scheme is workable: Look at
http://www.scgroup.com/ham/6731.gif .  This plot was made from a file
Markus DF6NM sent me.  He was tuned to 135.500 kHz, USB.  I used only
100 seconds of audio.  The input was simply mixed with a BFO near 1500
Hz, chosen such that the RF frequency of zero-beat was the nearest
multiple of twice the Lessay GRI.  The complex result was low pass
filtered to remove the sum components and integrated with a period of
2*GRI.  The plot has arbitrary phase and timing.  For reference, see
chain information at http://www.megapulse.com/chaininfo.html .
The LORAN code:

Group  Master        Slave
=====  ==========    ========
  A    ++--+-+- +    +++++--+
  B    +--+++++ -    +-+-++--

Starting from the left, you can see the group B pulses from Lessay
(master, 916 km distant), the somewhat weaker group from Soustons
(1144 km), a gap where the never-built Loop Head station would go,
then the strong pulses from Sylt (630 km).  The A groups follow.  If
this picture seems noisy, note that the real PLL would coherently add
the 16 pulses from both groups, for a gain of 12 dB.  If needed,
another 3 dB could be had by adding the transmissions on GRI 7499.
And, most users would be closer than 630 km from the nearest LORAN
station.

I hope that on most computers, the client software could run in the
background, and not interfere with running ARGO, WOLF, etc. on the
same audio stream.  Perhaps one of the Windows sound experts on this
list can verify if this is possible.  Of course, if you wanted to tune
to another band, use a narrow IF filter, or transmit, you would need
to stop running the client.  However, we would hope that there would
be other stations remaining online and capturing.

There are many interesting ways to combine signals.  Where coherent
combination is not possible or not desired, the system can function as
a simple diversity receiver, selecting the highest quality (if format
permits), strongest, or quietest source.  One could also present the
data for human combination, e.g. for QRSS three sources could control
the intensity of the primary colors, or CW from two sources could be
played as stereo.

IMO, it's likely that if stations are fairly close together, say,
within 100 km, then the received relative phases are predictable,
based on the direction of the signal source.  Of course, this system
would allow us to determine if that is true.  If it is, we would have
a long baseline array that could be "aimed" in a desired direction,
without needing phase information from the signal.  Stations such as
DF6NM with direction finding ability could send two streams, providing
an even tighter beam.  The array would be especially useful for Jason,
DFCW, and other formats with limited "carrier" energy.

A server-based system offers some advantages even when multiple
sources are not available.  For example, if a QRSS plot is unreadable
because of improper settings, one can just change the speed, contrast,
frequency, etc., and run it again.  A typical plot should take only
about two seconds to generate, and about one second to send over a 512
kb/s connection.  One can store lots of history.  A full-band capture
is only about 5 kB/s.  If there are an average of four stations
reporting, two months of data is just over 100 GB.  If someone invents
a better noise blanker, we can then compare its performance on some
old messages.  The server could also run processes that watch for
certain events, e.g. a signal on a certain frequency exceeding some
level, which would ease the search for good propagation.

Some questions for the LF community:

What problems do you see in implementing such a system?
Would you make use of services that this system would offer?
What signal formats or features would you like to be supported first?
Would you run the client at your station?
This will be an open source project.  Would you help with coding
  and/or testing?  Wolfgang DL4YHF has already offered help with
  the GUI, many thanks.
Would you provide recordings from your station for debugging?
Can you suggest a better name for this project?

73,

Stewart KK7KA