[Laser] Troposcatter and Information Theory-2
Andrew T. Flowers, K0SM
aflowers at frontiernet.net
Fri Jun 25 01:44:55 EDT 2004
Time to get out the envelope.....
I'll start by going down the garden path...I naively asked myself,
"Okay, if I send the data in durations but loop repeat them 82 times, I
end up with the same amount of information. But what prevents me from
sending the data 83 times in that minute, or 1000 times, thus improving
throughput?" That would make life wonderful and we wouldn't have a need
for information technology classes if that were the case :-) , but
unfortunately that doesn't take into account of the bandwidth required
for transmission. If you shorten the duration of each individual tone
you need to increase the bandwidth necessary for transmission. This is
afterall, why you can't have a 500kbs modem on a regular phone line.
The maximum frequency resolution you can have is equal to 1/t, where t
is the duration of the bit. You can think of S/N as the amount of
energy in one bin relative to all the other bins. An increase in speed
means the bins get wider and fewer, assuming a fixed bandwidth. At the
faster speed, each bin contains a larger amount of noise along with the
signal. That decreases S/N. That's not something that can be sacrificed.
The assumption here is that there will be no overhead--in fact,
8.1s/char isn't all that great. We're talking about decoding signal
levels in the ballpark of -30dB S/N (2500Hz BW) and 1.5db better for
each doubleling of the averaging time thereafter, theoretically. That's
just extrapolated from the numbers for K1JT's JT44 code. I
Of course, this all assumes that the path is perfectly stable--on an
unstable path that has brief, unpredictable periods of stronger signals,
redunancy allows for each bit to have a "better chance" of getting
through. Compare the JT44 and FSK441 modes in WSJT for extreme examples
of this contrast. I'll also have to see what propagation really looks
like in the phase domain, as it may be possible to encode in that domain
as well, or at least make use of it's stability in the averaging
algorithm....
Given the signal levels you are going to get with troposcatter you
aren't going to be going fast. It's just not possible....Then again you
are talking to a guy who's yelled into a mic for 90 min in 2m SSB meteor
skeds. Ten minutes...forty minutes.... to exchange calls sounds just fine!
Does that make sense? Or have I gone woefully astray?
Andy K0SM/2 (ready for field day!)
TWOSIG at aol.com wrote:
>Andy K0SM/2 had a recent question to help setup a troposcatter system that
>got me puzzled about coding information. He described the system "uses an 8.2
>second transmission length per character and incoherent averaging of the
>message as it loops
>around. "
>
>I know that there has been success with moonbounce using long characters, but
>I don't understand the advantage. Assume a 10 character message, that takes
>almost a minute and a half to transmit. Doesn't information theory say that
>if you use 0.1 second character time and repeat the message 82 times you get
>the same benefit?
>
>Is there something about troposcatter that would benefit from such long
>characters?
>
>For a communications system, I would think that using short characters would
>allow you to count the number of repetitions needed to "detect" the correct
>value, and thus have a dynamic measure of the data rate that the channel will
>provide. If the system is stable with excess channel capacity, it could then
>adapt by increasing the message rate ( send fewer repetitions ) or lowering the
>transmitted power
>
>It could be that I just have a personal bias for faster communications. I
>think PSK31 with every character repeated three times is as slow as I would want
>to try to work.
>
>
>Andy, I wish you success.
>
>
>James
>N5GUI
>_______________________________________________
>Laser mailing list
>Laser at mailman.qth.net
>http://mailman.qth.net/mailman/listinfo/laser
>
>
More information about the Laser
mailing list