[Skimmertalk] Skimmer in SSCW - lots of lessons learned

[Pete Smith]

This year I decided to go Unlimited for the first time in over 50 years of 
Sweepstakes, in order to be able to try out CW Skimmer under full-up, 
high-activity SO2R conditions.

For those who will ask - my computer is an oldish Dell 2.66 GHz Celeron 
(single core) with 2 GHz of RAM.  I ran N1MM Logger 8.10.7, and used an 
SDR-IQ receiver for Skimmer.  CW was via Winkey, so there were no problems 
with producing CW via the LPT or serial port.  SO2R control was via a 
MicroHam MK2R+, using the Microham SO2R protocol to handle all band data 
and switching.  Antenna and bandpass filter selection was automated using 
the MK2R+ to control two W9XT band decoders.  WintelnetX was used to mix 
spots from Skimmer and one Telnet cluster.

I wish that I had done this sort of test in one of the earlier contests 
that I don't care about so much, because then I would have had time to fix 
some things, and could have learned to optimize others.  I guess I'm not 
very good at constructing test scenarios in the absence of real 
conditions.  Items:

1.  With this computer, CWSkimmer's demands on the CPU are just too large 
for effective computer-controlled SO2R.  There were small, variable, subtle 
delays in logging program operation that manifested themselves in several 
ways.  For example, if I wanted to interrupt a CQ on the run radio to call 
a station on the S&P radio, normally I rely on the computer to switch focus 
and start transmitting immediately.

With Skimmer running there was a small but perceptible delay in both the 
focus switch and the beginning of CW.  My timing is normally pretty good, 
but with Skimmer running I was consistently late.  Worse yet, focus changes 
in general were often delayed and sometimes out of synch.  N1MM would send 
a command to the MK2R+, so that the audio in the headphones changed, but 
the entry focus would "hang" for a variable period before it followed 
suit.  This led, all too often, to my entering part of a callsign in the 
wrong entry window, which led to as lot of frenzied keyboarding to try to 
get things back where they should be.

There's no question in my mind that CPU loading was the culprit, because as 
soon as I stopped the Skimmer waterfall there was a noticeable return to 
normal behavior.

The solution seems fairly straightforward.  Skimmer should run on a 
separate actual or virtual CPU.  The easiest way might be to network a 
second older computer to the main logging machine, eliminating any sharing 
at all.  Band-changing can be easily managed by sending commands to the 
Skimmer from the Telnet window of the logging software - commands such as 
"skimmer/qsy 14000" can be stored in a single button.

Users of modern computers which have two or more virtual CPUs may be able 
to solve this problem by assigning their logging software to one CPU and 
Skimmer to another.  I have not tried this, but hope that someone will.

What I did, as a work-around for this contest, was to stop the waterfall 
after Skimmer had run for a minute or two, and work my way through the band 
using the S&P radio, making QSOs interleaved with my QSOs on the Run 
radio.  That seemed to work reasonably well.

2.  Other than the computer issue, Skimmer worked very well.  2-3 minutes 
of Skimmer on a given band would post 150-200 calls to the bandmap for the 
band in question.  N1MM Logger would color-code these as new or dupe, and 
would note which ones, based on the call history file, would be new 
multipliers.  The potential advantage on Sunday in a contest like 
Sweepstakes is huge - instead of tuning the band listening to CQers, 95 
percent of whom are dupes, you can jump right to the new multiplier or the 
"fresh meat."

Since no Skimmer spot is more than 10 minutes old, and all represent 
stations that were being heard at my QTH, I found that I was getting a much 
higher percentage of "hits" with Skimmer spots than with "normal" ones.  In 
fact, the next time I use Skimmer in SS, I think I will turn off the Telnet 
spots altogether.

I found that concerns about the run radio interfering with Skimmer's 
ability to copy signals on the S&P radio were largely unfounded, because of 
the bandpass filters I use for SO2R, which also protect the Skimmer 
RX.  Often, I would watch the S&P radio bandmap while the run radio was 
CQing, and see new spots flowing in at a rate of better than 1 a 
second.  Pretty impressive

I will be pulling together an article for NCJ early in the new year, 
dealing with Skimmer experiences in this fall's contests.  If you used 
Skimmer, please drop me a note and let me know how it worked (or didn't), 
as well as what you liked and what you didn't.

73, Pete N4ZR