[Skimmertalk] FW: Skimmer Tests during SS

[Jim Baremore]

Dick answered my queries at 3:20 AM this morning following my 11:30 PM
message last night.  He's still at it almost 12 hours later. I'm beginning
to wonder if Dick ever sleeps HI HI.  We had had a good exchange and I
appreciate his thoughts and comments.  I will just add a few clarifications
to this note and try to stay off the soapbox - note I said 'try.'

The test environment was processing over 1000 spots an hour to verify they
were valid licensed US call signs and it was outputting those at a 990
calls/hour rate.  Unfortunately because of a few mis-thrown software
switches, it was only outputting about 500 calls an hour to the logger,
still enough to keep anyone busy.  As I mentioned in a prior post, the raw
accuracy of CW Skimmer was estimated to be 94%.  The low 30% number is not a
reflection of call accuracy but rather how many of the calls I clicked on as
a spot were really there. I like Dicks' word "Mis-Snagged"

I suspect the comments Dick made about SS 'CQ'ing' in the form of "SS
<callsign> SS" vs. 'CQ' or 'Test' might be the real key to understanding why
CW Skimmer was reporting the worked station more than the running station on
the frequency.

With regards to the use of the Paranoid mode.  I don't think Dick's
assessment is right - yes he has been up too long! -  CW Skimmer generates a
call by some mechanism.  If 'Paranoid' is selected, it takes that call and
checks it against a user selected data base and eliminates non-USA calls.  I
don't see how that would have reduced the number of busted calls that got
mangled into valid FCC calls.  If it was busted by the CW Skimmer decoder,
there is no way the data base checking for USA calls would have known it was
not a US call.  

In my test, every call generated by Skimmer was checked against the FCC data
base and had to be a valid and licensed USA call to get passed as a spot.
If a mangled call turns into another US call, it still didn't get through
unless it is a licensed call.

Finally, note that 8% of the calls I sorted were calls I deemed to be close
to the real call but were off by a dit or a dot somewhere.  This was pretty
subjective and of course the call could have been right after all but I just
heard it wrong myself.  This could be a reflection of calls that were
mangled by CW Skimmer into a US call that was in fact a licensed US call.
However it happed, the data still says the copy was 92% right after
filtering.  That's pretty accurate and consistent with Dick's findings.


73's

Jim  K5QQ



-----Original Message-----
From: skimmertalk-bounces at contesting.com
[mailto:skimmertalk-bounces at contesting.com] On Behalf Of Dick Green WC1M
Sent: Saturday, November 15, 2008 1:42 PM
To: 'Pete Smith'; skimmertalk at contesting.com
Subject: Re: [Skimmertalk] Skimmer Tests during SS

Jim and I have been discussing his report offline. His results don't
correlate with mine in IARU CW, where I often got more spots for CQing
stations than I could keep up with, and accuracy was outstanding.

While Pete is probably correct about the effect of SO2R time delays on
finding CQing stations, another reason for the low hit rate is that it's
quite common for stations CQing in SS to simply send something like "SS
<call sign> <call sign> SS". The words "CQ" and "TEST" are often omitted. CW
Skimmer doesn't have SS-specific code, does it? If not, I'm sure it will
eventually... :-)

I agree on using Paranoid mode. With the right SCP file, that would have
eliminated the non-USA calls, and I suspect it would have greatly reduced
the number of busted calls that got mangled into valid FCC calls.

73, Dick WC1M

> -----Original Message-----
> From: Pete Smith [mailto:n4zr at contesting.com]
> Sent: Saturday, November 15, 2008 6:44 AM