Content-Transfer-Encoding: 8bit
Message-ID: <7D653C9C42F5D411A27C00508BF8803D55C67F@pdw-mail-r1.dstl.gov.uk>
From: "Talbot Andrew" <ACTALBOT@dstl.gov.uk>
To: rsgb_lf_group@blacksheep.org
Subject: LF: 7 FSK Spacing, PSK,  and simple accurate	, frequency sources
Date: Wed, 2 Jan 2002 11:56:09 -0000
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1; format=flowed
Precedence: bulk
Reply-To: rsgb_lf_group@blacksheep.org
Sender: <majordom@post.thorcom.com>

There has been quite a lot of comment recently about 7FSK and the bandwidth
it takes up.  The only transmissions (other than my initial tests) were by
Jim who seemed to be using quite a wide spacing relative to the dot length.
This is a classic spread spectrum transmission where bandwidth is quite a
lot wider than necessary either to assist in demodulation or to mitigate
certain propagation effects.

If we are to use this mode with visual 'decoding' using Argo then we really
need to do some tests with weak signals and different spacings to see what
is optimum.  A good point to start would be at a tone spacing equal to twice
to three times the reciprocal of the symbol interval.  Thus for 30 second
symbols, one letter per minute, for Argo's 30s dot mode, try a tone spacing
of 0.07 to 0.1Hz, giving a signal bandwidth of 0.4 or 0.6Hz.   It may be
possible to go narrower, but reading will be more difficult as the signal
will now be in adjacent frequency bins.
.....................................

On a different matter, I have been taking part in some 137kHz test
transmissions using very low power to G3PLX who is using coherent reception
and GPS locked signalling.  We are still in the early days yet, and Peter
has been sidetracked into some domestic work for now, but initial results
look promising.   I have been radiating about 1 Watt of RF which, with my
antenna, equates to about 200uW ERP and Peter got perfect copy at 400km
distance.  The signal format was 30 second BPSK, but the results would have
stood much faster than this.   His next stage is to integrate over longer
periods and for me to transmit with even lower power.

The key to these experiments is very accurate signal timing and frequencies,
we require stabilities and accuracies of just a few parts in 10^-10 in order
for the signal to stay within a few degrees of phase over tthe entire
transmitting period.  In fact a DDS source makes life difficult due to my
only having steps of 5MHz / 2^32 !!  Symbol timings are also locked to GPS.
Peter uses a very novel and very simple way of locking a receiver to GPS - I
shan't explain how here, but will leave him to cover it in a subsequent
write up; it is not finalised yet, but is very straightforward and needs
little more hardware than a GPS receiver module with a 1 PPS output.  I have
extremally high frequency stability anyway, but we  need to look at a
simpler route to getting a low cost locked standard from GPS before this
technique can be used with a wider audience.   For LF signalling, phase
noise or slight jitter on the transmission during a symbol interval is not a
problem, meaning that a GPS locked frequency standard using the 1 PPS output
to adjust a crystal oscillator would be adequatem - so generating ANY
frequency that is a multiple of 1 Hz becomes possible.   

My main interest in high stability sources is rather different as I am,
these days, more into high frequencies where the 1 Hz jitter is an absolute
taboo when multiplied up to GHz - it would appear as audible chirp up there.
However, having been nudged in this direction by Peter I did spend time
during a long walk yesterday thinking of sources for LF :
Start with a very simple VCXO using any cheap crystal,assume its frequency
could be up to +/- 30ppm out.  The only requirement on actual frequency is
that it ends up at a multiple of 1Hz.   Continuously clock this into a
binary counter and exactly every second (derived from the GPS receiver)
measure the counter value.  If the counter is long enough it will be able to
resolve all frequency ambiguities (30 ppm at 5 MHz would normally give 150
potential lock points if a simple latch were used).   Use the error between
this count and what is should be, to generate a correction signal to control
the VCXO.  An 8 bit counter would be good enough and a PIC interrupted by
the GPS pulse reading this counter, plus a handfull of Rs and Cs should do
it.  We end up with a strange mixture of PLL and frequency locked loop, but
one that is simple and optimised to long duration signalling on LF.   This
is probably a re-invention of the wheel but I haven't seen exactly these
requirements before; a very high medium to long term stability but short
term / phase noise not too important.

I think this is basically what Brooks Shera does in his GPS frequency
standard mentioned on this reflector recently, but his design requires a
good oscillator source to give a very high performance standard across the
spectrum.  We do not need this kind of short term performance where
signalling is going to use symbol lengths of many seconds.  This techniqe
may also be allied to the old Huff and Puff stabilisation technique, but
extended by the N stage counter rather than a sigle flip-flop to remove
frequency lock point ambiguity.  The original Huff and Puff system was a
little bit before my time in the early 1970's :-)

Andy  G4JNT


-- 
The Information contained in this E-Mail and any subsequent correspondence
is private and is intended solely for the intended recipient(s).
For those other than the recipient any disclosure, copying, distribution, 
or any action taken or omitted to be taken in reliance on such information is
prohibited and may be unlawful.