Frequency Hopper Synchronization

[James Wolf]

How about this.
  
Sample a small number of frequencies and find the noise floor; disregard any
that appear to have a signal.   Store the average result.  
Continually sample one frequency (any in the sequence) that initially has no
signal.
Once it is found to have a signal (n)dB above the noise level, follow the
preset sequence and see if all channels have a similar signal level.
If not, (since it may have an undesired signal), go to the next frequency in
the sequence and repeat until the sequence is found.

Jim, KR9U    

-----Original Message-----
From: tacos-bounces+jbwolf=comcast.net at amrad.org
[mailto:tacos-bounces+jbwolf=comcast.net at amrad.org] On Behalf Of Frank Eliot
Sent: Sunday, March 23, 2014 4:46 PM
To: <tacos at amrad.org>
Subject: Frequency Hopper Synchronization

Gents -
	On Thursday, Andre and Maitland presented a way to synchronize a
frequency hopper, using the 1 pps output from a GPS board. This pulse is
accurate to one millisecond, which is good enough for a hopper that hops 128
frequencies/second. This solution should work, but isn't very elegant, and
is sorta brute force. After thinking about it, I suggest trying to synch
without the requirement of having a GPS decoder and being able to hear a GPS
signal.

	Here's a thought experiment. Sample the signal level from the
receiver at all possible 128 frequencies, at a rate of 128 samples/second,
and store the sampled levels in memory for several seconds. This data rate
is 128 x 128 samples/second, so you don't want to store much. The receiving
site knows the sequence of the 128 frequencies - it just doesn't know where
it starts. In software, sum the signal levels of the 128 known frequencies
in the known sequence starting arbitrarily at sample zero. Then do the same
sum for the 128 frequencies in sequence starting at sample one. Do this sum
for all 128 starting points. One of the sums should be noticeably larger
than the others, and that defines the correct starting point. Feed this
starting point to the decoder. During a QSO, you could continue running the
correlator at plus or minus one count, and switch when the adjacent count
rises. Except for the initial sampling, all the heavy lifting is performed
in a computer, working on stored data. This is where the work should be
done.

	This tradeoff wasn't discussed on Thursday. Maybe it is out of the
question. We have nifty spectrum analyzers. Can we sample 128 points on the
spectrum at 128 times/second? That seems like the most difficult thing to
do. The correlation calculation seems doable, as it is mainly summing, and
doesn't have to be done in quite real time.

	This synchronization scheme appears substantially more complex to
implement than the one I suggested a few months ago for autonomously
phase-locking a synchronous detector to a received signal. That technique
would buy a few dB's S/N on a PSK31 type of signal by means of signal
processing.

	Just trying to start a discussion. Is this doable??

Frank
W3WAG
	

	
_______________________________________________
Tacos mailing list
Tacos at amrad.org
https://amrad.org/mailman/listinfo/tacos