Hi Gaƫtan,

On 12/28/2018 4:05 AM, ON4KHG wrote:
So, if I have well understood, as far as AP and Auto Seq are not concerned (and all other things being 
equal), messages like "F6ABC ON4KHG 73" or "73 XYZ TU" (or even "ABCDEFGHIJK") 
should equally well be decoded ?
HNY 2019.

Just a bit more detail, adding to what Bill has written.

In the v2.0 FT8 and MSK144 protocols there are 2^77 = 151,115,727,451,828,646,838,272 (about 1.5 x 10^23) possible messages. If AP is not enabled, the decoder's job of finding the correct message is equally difficult for every one of these integers. The inner-most layer of the decoder knows nothing about what type of user-meaningful message may emerge from the process. At this stage, "F6ABC ON4KHG 73", "73 XYZ TU", and "ABCDEFGHIJK" are all nothing more than different integer numbers between 0 and 151,115,727,451,828,646,838,271.

As a QSO (or potential QSO) progresses, AP decoding can improve the sensitivity for standard messages by up to 4 dB. The decoder at F6ABC can hypothesize, for example, that the first 28 of the 77 unknown bits correspond to his own callsign, F6ABC, and that another 28 bits correspond to ON4KHG. The decoding task is now reduced to finding a much smaller number of unknown bits. Near the end of your QSO the decoder "expects" that the received message is likely to be "F6ABC ON4KHG 73". With this working hypothesis, the task is reduced to determining whether the received symbols corresponding to 83 parity bits are consistent with that message, and that the 14-bit checksum has been satisfied. Thus, with AP enabled your QSO partner can decode "F6ABC ON4KHG 73" down to about S/N = -24 dB, while the limit for "TNX 73 HNY" is about -20 dB.

Auto-sequencing logic comes into play only after a message has been decoded and converted back into human-readable form. As we have defined it in WSJT-X 2.0, the auto-sequencer insists on seeing both the expected callsign and "73" in order to finish a QSO.

        -- 73, Joe, K1JT


_______________________________________________
wsjt-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wsjt-devel

Reply via email to