Though I wasn't there for the 2008 pileup contest, I have run Skimmer
against a pileup test tape. Here's what I reported a couple of months ago:
"I just did a series of Skimmer runs against the 1999 KCDXC pileup
recording. The human winner that year was W9WI with 71 calls; The median
score was 38. Skimmer counted 36-7, depending on the
run, but by design it only counts and lists those that it hears twice
identically ("verified calls"). Watching the Skimmer bandmap and the
decoder row below the waterfall, I counted at least 20 more callsigns that
it properly decoded but did not count because it only heard them once.
If all of those were correct, and there were no busts on the ones it heard
twice, then it would have placed 7th or 8th. Not bad."
Judging by the scores this year, the 2008 tape may have been a lot
tougher. I wonder whether anyone made an effort to note callsigns that,
for whatever reason, were decoded correctly but only once.
A final technical point - CW Skimmer relies on frequency to differentiate
between data streams, which is why it is not at its best when "listening"
to a rapid time series of callsigns that are within a couple of hundred Hz
of each other. That's why it is at its best when used with a wide-band
receiver with a fixed center frequency. Spread 20 signals out over 3-4 KHz,
and it will probably copy them all.
73, Pete N4ZR
_______________________________________________
CQ-Contest mailing list
CQ-Contest@contesting.com
http://lists.contesting.com/mailman/listinfo/cq-contest
|