Bill, Some early observations of 2.2.0-rc1 decoding:
I parsed through my ALL.TXT (375+ MB) counting number of decodes per cycle. Prior to yesterday, my highest number of unique FT8 decodes in one cycle was 63. After installing 2.2.0-rc1, I monitored 14.074 for a couple of hours yesterday afternoon. During that time I logged 10 cycles with >63 unique decodes, the highest being 72. So, it's definitely working well for me. I also monitored 30 meter WSPR for a little while and was regularly getting 10-15 decodes per cycle. Anecdotally that feels like a big improvement but I don't monitor WSPR enough to have a meaningful baseline. I noticed that the final WSPR decode pass extended well into the next cycle, 10 seconds or more, when the window was busy. Granted, this is with a 10+ year old AMD CPU... 73 -- Ray WQ5L On Tue, May 12, 2020 at 3:30 PM Bill Somerville <g4...@classdesign.com> wrote: > > On 12/05/2020 20:45, j...@comcast.net wrote: > > >one feature of the latest decoder is there to yield decodes before all > >transmissions are completed. > > > > Bill, > > > > Was this a user request, or just a result of the new decoding algorithm? > (My curiosity has been piqued.) > > > > Ed N4II. > > Hi Ed, > > because the FT8 protocol includes Forward Error Correction (FEC) information > it is possible to successfully decode a signal before the complete block of > data has been received, this is an opportunity to pick the "low hanging > fruit" as early as is practical, leaving more elapsed time to work on the > remaining signals. The intent is both to deliver some decodes earlier and to > utilize more CPU resource to yield more decodes in a reasonable time frame. > Most users, probably all, on a busy band will see some decodes earlier in the > receive period and also more decodes overall with the last few maybe only > printing some time into the following period. We have not put a number to the > extra decode yield but we have certainly see up to one third more when > processing saved .WAV files from congested bands compared to the v2.1.2 FT8 > decoder. Steve K9AN has a "golden" test FT8 .WAV file that he uses to verify > decoder yield and I believe the current result from that file is 67 unique > valid decodes. > > 73 > Bill > G4WJS. _______________________________________________ wsjt-devel mailing list wsjt-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wsjt-devel
