[Moon] WSJT/CW

[Klaus DJ5HG]

Dear Bernd and all,


let me again comment a mail of Bernd:


> Dear Hakan and all,
> 
> Just for clarification:
> 
> (1)
> JT65 does communicate everything via the air.

JT65 indeed transmits the full information. The RS and 
the KV-decoders can retrieve this information, DS cannot.

 
> (2)
> In JT65 DS the callsigns, i.e. "SM7WSJ DF2ZC", you are looking for are
> detected by means of correlation (mathematically comparing the weak signal
> in the noise with signal forms constructed by the help of the call3.txt
> database). Of course the information you are looking for is already present
> on your hard disk, though "SM7WSJ DF2ZC" is transmitted and received over
> the air. The only previously unknown information you gather is that the weak
> signal indeed is one of the constructions with the help of call3.txt, so one
> of the some 1000 combinations. This equals ld (1000) = some 10 information
> bits. The amount of information gathered roughly is as high as it is when
> you receive two arbitrary alphanumerical characters, such "A7" for instance
> as the ld (36*36) = 10,34 bit. If it were 8000 possible combinations, it
> would equal some 13 information bits, with 16000 possible combinations it
> would equal some 14 information bits, with 64000 combinations some 16
> information bits. So with increasing the call3.txt database the information
> bits of a DS decode rise.
> 
> However, everything is still transmitted over the air! Only, the information
> you gather is that the weak signal in the noise is one of the callsign
> contructions with the help of call3.txt or not, so one of the roughly 1000
> possibilities. This correlation is a well established mathematical procedure
> used, similar to a "matched filter" which only delivers the output one is
> looking for (if that signal is indeed present on the receiver) or nothing at
> all. Again: The information gathered is "Yes, it is one of the 1000
> callsigns." - or as DJ5HG calls it: an index.


Nobody questiones that JT65 encodes two callsigns at 
the transmitting end. But it is to the receiving end 
what it really receives. It's the same with SSB and 
CW. What is a measure for a communication path, the 
information that is sent or the information that 
actually can be read from the received signal? 

The DS problem is that it is NOT a decoder which 
decodes the Reed-Solomon codewords. A very small 
subset of the RS codewords are chosen (by the list) 
to generate the corresponding signals that are to 
be correlated with the received signal. The crucial 
point is that the DS decoder only can make it's final 
decision on which of the generated signals fits best, 
not on the symbols which form the callsigns. The actual 
index of the best fitting signal then simply is used 
to address the associated message. So the result of 
DS primarily is based on a decision out of restricted 
number of choices.

I just tried to do the same with a small simulation 
program with CW. CW was binary coded by '10' and '1110' 
being the representations of dots and dashes. Then a 
selected callsign was corrupted by 8 different binary 
errors in each simulation. This makes the call completely 
unreadable. In a million simulations with "decoding" by 
correlation with signals generated from a subset of the 
call3 list, 999981 receptions resulted in the correct 
callsign. So DS by no means does concern the question 
of WSJT vs. CW.  It only concerns the fundamental question 
whether we are interested in COMMUNICATION via the radio 
path or simply in detecting signals while communication 
goes other pathes.

 
> This is very much similar to the well aknowledged (and never discussed!!!)
> sked gain on CW of some 3 dB, or to the fact that when you are called by a
> weak station you nearly always copy your own callsign better than the
> caller's callsign. That is what I meant with the memory between the ears. It
> is always easier to copy "SM7WSJ DF2ZC" in a CW sked at a marginal signal
> strength than on random an arbitrary combination such as "SM7WSJ AZ4GOR". We
> all know that for decades and never put in in question.

The sked never needed such a discussion because there was 
a consensus that the callsigns must have been exchanged 
as if they were unknown. In a HSCW meteorscatter contact 
you must receive enough fragments such that the puzzle 
only has one solution. A single fragment received on a 
lonely frequency is not sufficient even though it may be  
absolutely confident that it came from your sked partner 
who sent both calls.
 
 
> BTW This means in a sked (CW and JT65 as well) your personal call3.txt is
> just one callsign, "DF2ZC". Therefore you gain only ld (1) information,
> which equals 0 information bits! This applies also if you'd copy me with
> 599, the amount of information bits cannot be increased because "DF2ZC" is
> already present on your personal hard disk in the head. Actually, you gain
> not 0 but 1 bit of information: You now know that DF2ZC is really there for
> the sked which is the only thing you did not know in advance.

Yes that's correct. What we demand for is that the deciding 
stage at the receiving end does not make use of such information. 
The KV-decoder really is independent of the call3 list. If you 
use FSK441 in meteor scatter the decoder also is independent of 
any list. But the operator could cheat. Even if he does not, he 
has the advantage that he may be satisfied by a single reception 
of the calls, while in random QSOs this usually is not confident. 

Indeed, assisting the decision process by prior knowledge does 
work very much better with coding. But that should not be an 
argument for really doing it systematically.

Finally, I again want to point out that this discussion is not on 
digital vs. CW.  The question is:  What is demanded for a valid 
contact? Is there a demand to communicate the callsigns (or only the 
unknown callsign) via the radio path or is it entirely sufficient to 
get the knowledge via other ways?  In other words:  
Is the main request of ham radio the simple physical detection of 
signals, or is the main request to establish bidirectional radio 
contacts that communicate a specified minimal amount of information?
   
 
> Actually the underlying information theory is even a little more complicated
> as the amount of information gathered is related to likeliness. When a
> rather active station such as W5UN, KB8RQ or S52LM is copied the information
> content is less than for instance when SV1BTR is copied calling you on JT65
> ;-) 
> I'll leave that out to make understanding easier and assume all callsigns
> are equally likely.
> 
> 
> (3)
> If you understood (2) correctly you will easily see that unlike some guys
> permanently try to make us believe JT65DS does not communicate only two
> letters of both callsigns via the air to have a decode and the rest of the
> callsigns is added with the help of call3.txt. That is rubbish! Those guys
> such as HB9..., SM2...and others do apparently not have the slightest clue
> how DS works. Yet, that does not keep them from repeating incorrect
> statements.

If they say "the DS decoder receives information EQUIVALENT to two 
characters" then all is correct.  The information in JT65 is coded 
anyway. So if they really suppressed the word "equivalent" there 
is no possible misunderstanding into a wrong direction. 

The simulation mentioned above really shows that it even works if 
all received CW-characters are wrong.
  
 
> (4)
> IMPORTANT! I did not and I will never say that this means that people using
> CW don't receive all characters of a callsign!!! Whoever accuses me of doing
> so does not at all understand what I'm saying. I repeat: I confirm all CW
> friends always copy all dits and dahs in a QSO!!! This has no influence on
> what I'm saying above. 
> Sadly, there are some guys who are not serious with the QSO rules, but in
> both modes. The cheat on themselves at the end. Also I do not say it is
> acceptable if people exchange QSO details on chat/telephone etc before the
> QSO is complete. I regard a chat as a online contact room to solicitate
> short notice skeds.
> 
> I do not intend to go on replying replies to this post in public. Should you
> (or anyone else) have any further questions I'm happy to discuss that with
> you in private emails and leave this off the reflector.

That's a basis we can agree on. Ok, from now off the reflector.

 
> vy 73 Bernd DF2ZC (JO30RN)



The WSJT/CW thread also concerns some other aspects which I 
do not comment here. My main concern is to make clear that DS 
creates a general problem. There may be arguments to separate 
CW vs. digital modes. But from the basics, it is even more 
important to separate DS from all other modes, JT65 with KV 
included.  


73,

Klaus, DJ5HG