I've been thinking about the maths behind Wolf and it gets quite interesting.
One thing we can say, its definitely a spread spectrum system !!!!!!!!
SRI USA Hams :-(
A message of 15 characters taken from an alphabet of 40 (A-Z 0-9 and a
few punctuation) give 80 source bits of data after compression.
These are expanded to 480 data bits in six parallel tail-biting
convolutional encoders. In pure error correction terms this seems
like mega-overkill, so there's more than meets the eye going on...
Interleaving may or may not then be employed, but it doesn't affect
things.
The 480 bits are then merged with another 480 bits of a PN seqeunce
(in a not totally dissimilar way that I suspect Opera does) to allow
the decoder to lock to the sequence.
The whole lot is then transmitted at 10 Symbols/s BPSK so the whole
message takes 96 seconds
The underlying data is 80 bits, so without any sequential merging, to
a rough approximation it equates to about 1 bit every 1.2 seconds, so
you could say the effective bandwidth is 0.833Hz and the very strong
FEC will allow something like +2 to +3dB S/N for decoding in this
bandwidth. So therefore something like 3dB up on WSPR, and about
the same as OP8 with the same mean power. (I know PSK and FSK don't
have the same noise performance, but just ignore the details for now!
There's probably a couple of dB in it when signals are weak)
But its not this simple. Being a fully coherent system, and unlike
WSPR/JTxx/Opera and any other non-coherent mode, successive messages
can be overlaid. They can even be used as part of a soft decision
decoding tree, and theoretically could go on for ever - providing it
is possible to maintain synchronisation.
So by continuously looking at repeated message, the effective
bandwidth will get lower by the average of the number of times it has
been repeated. Two repeats 3dB or half BW, 4 repeats 6dB , 8
repeeats 9dB etc etc
Does this seem about right ?
Andy
www.g4jnt.com
On 25 January 2012 02:30, John Andrews <[email protected]> wrote:
> Graham,
>
> There is no single s/n figure or specified BW in which to measure it. WOLF
> works by building copy up over a period of time, and assuming good
> frequency/phase stability, a half hour or more is practical.
>
> The data rate is fairly fast, 10 b/s, with a 960 bit message, so that a
> complete frame of data is sent in 96 seconds. You get three quick reports in
> the first 96 seconds, and then decodes every 96 seconds after. Each message
> is 15 characters, with no rules about callsign format.
>
> Tests done some years ago show it roughly equivalent to QRSS60 in terms of
> signal level. It does give some clues as to whether a signal has been
> locked-in, and has the possibility of partial copy, unlike some "all or
> nothing" modes. Fifteen characters in a half-hour is of course much faster
> than QRSS60 would permit. 2-way QSO's are fairly easy with some advance
> agreement on what to do with the 15 characters.
>
> The downsides are the need for a linear transmitter system (as this is PSK),
> receiver/transmitter stability and frequency accuracy. It works at 600
> meters, but is fairly useless at 160 meters and up.
>
> The newest version is by DL4YHF, and may be found at:
> http://www.qsl.net/d/dl4yhf//wolf/ .
>
> Note that this version has 5 b/s and 20 b/s variations. Testing has shown
> that the program tends to provide copy in the same amount of time, so
> there's no overriding advantage to either.
>
> John, W1TAG
>
>
> On 1/24/2012 7:27 PM, Graham wrote:
>>
>> WOLF DATA MODE S/N FIG ?
>> Long time since I used wolf , anyone have the
>> expected decode min s/n level ?
>> and now what s/n was ref to , as changes
>> may of taken place over time
>> Tnx -G.
>>
>>
>>
>> No virus found in this incoming message.
>> Checked by AVG - www.avg.com
>> Version: 8.5.454 / Virus Database: 271.1.1/4162 - Release Date: 01/23/12
>> 19:34:00
>>
>
|
