Michael,
have you looked at the hint code that was added ... its a bunch of 'if
then' decision trees, which can explain the 'extra processing'.
As a user, I'd rather trust the 13K WSJT-X decodes to be REAL, rather
than the 15K of which 2 or 3K may be spurious.
Neil
KN3ILZ
On 6/11/2017 3:49 PM, Black Michael wrote:
I see why people are claiming JTDX decodes better.
I ran a 24-hour test side-by-side on the same audio stream and got the
following numbers.
JTDX average dB on matched QSOs is ~1dB better (not a big deal)
JTDX got 15,690 decodes
WSJTX got 13,692 decodes
JTDX got 2,144 decodes that WSJT-X did not
WSJTX got 146 decodes that JTDX did not.
So JTDX got ~15% more decodes...nothing to sneeze at. Can't say I
really see many false decodes but I guess really can't know for sure.
Here's a good example of a noticeable difference in processing this
wav file.
https://www.dropbox.com/s/zzddsaqop94bfc9/170504_1159.wav?dl=1
<https://www.dropbox.com/s/zzddsaqop94bfc9/17504_1159.wav?dl=1>
I minimized the values in the Advanced tab for JTDX but it still gets
12 decodes and WSJT-X gets 7.
It takes a bit over 5 seconds for JTDX to do its thing and a bit over
2 seconds for WSJTX.
So there is some extra processing going on for sure. Either in JTDX's
"subpass 1" or "subpass 2"?
Are sure we don't want to borrow something back from JTDX?
We had a set of test files before that were used for timing and
getting some params for false alarms and such.
Perhaps we should run those again and compare?
Inline image
Inline image
-----------------------------------
Michael D. Black
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
wsjt-devel mailing list
wsjt-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wsjt-devel
