[Skimmertalk] I/Q Balance and CQWW Skimmer observations

[Pete Smith]

Comments interspersed.

At 01:27 AM 12/2/2008, Dean R. Madsen wrote:
>Pete & the group:
>
>I have tried the transmitting in the dummy load, but it did not solve the 
>problem.  The signal from the clock of another soft rock was probably 
>stronger than the signal from the dummy load.  Thus the idea of editing 
>the data.

See below for more on this.


>I did run the skimmers in the CQ WW contest.
>
>I live on an older 50ft wide city lot in Des Moines, IA.  I am fortunate 
>to have a 40 ft tower with tri-bander, GAP Titan DX vertical, 80/40 dipole 
>and Alpha Delta DX-LB Plus dipole for 160.
>
>Unfortunately, this means that all my antennas are all very close to each 
>other.  I can't really operate the contest and have the skimmers running 
>at the same time.

You might be surprised.  For a while this weekend, I was running my Skimmer 
on 40M using my shunt fed tower as an antenna, while S&Ping at over 100/hr. 
on my 40M yagi, on the same tower.  I was getting useful skimmer spots 
between S&P QSOs.  Not as many as if I let Skimmer run for a minute or two 
without local QRM, but still useful.  One of the neat things about Skimmer 
is that you can run it for a minute or two and then go work the spots it 
has generated.  Even that way, they will be fresher than cluster spots (and 
all audible at your own location, which is a big plus for me).


>I did briefly provide skimmer spots for a multi-op effort..  This caused a 
>black blur on the band map (probably need to turn "show unworkable spots" 
>to "off").  The rate at which the skimmer spots came in far exceeded 
>regular packet cluster.  Unfortunately, many spots were flipped on the 
>wrong side of the soft rock center frequency adding to the deluge.  There 
>was also a big difference in antennas between that station and my station 
>that most of my skimmer spots (especially on 40-160) were zero pointers.

With SoftRocks you've gotta get I/Q balance corrected and suppress the 
local oscillator signal feeding back to the antenna.  With good I/Q balance 
the images will almost all disappear, and a good buffer amp will suppress 
the local oscillators by 70-90 dB.  K8ZOA makes a single radio buffer kit 
for $25 and is working on one with multiple buffered outputs that might be 
just the ticket for your setup.  I would drop him a note and ask how it's 
coming - use my name and call as your source.  You could wind up being his 
beta tester.  Jack's a good guy.


>It was decided that it was more useful to be able to watch the DX cluster 
>spots as they come in than have the influx of mostly useless skimmer 
>spots.  The skimmers were disconnected from the logging software only a 
>few hours into the contest (but remained reporting to skimmer.dxwatch.com).
>
>I found the contest to be a good test of the skimmer computer.
>
>I am using a quad core 2.83GHz Intel processor over clocked to 2.91GHz 
>with 4 GB (3.5 GB) RAM under Windows XP Pro.  Three M-Audio Delta 44 sound 
>cards, Softrock Lite 6.1 for 15m, 20m, 40m, 80m and 160m, a Soft  Rock 
>Lite-Xtall (dip switch version) for 10m.  The five Softrock Lite 6.1 
>receivers are mounted in a small Radio Shack aluminum project box (this 
>seemed like a good idea at the time and was a nice compact package, but it 
>appears the receiver clocks interfere with other receivers).  The 
>Lite-Xtall is hanging outside the box as it was just built.  10m, 15m and 
>20m share the tri-band beam.  40m has the vertical and 80 & 160 have the 
>appropriate dipole. CW Skimmer is installed in six sub-directories, one 
>for each band.

Sounds like the big thing you need is to shield between the receivers and 
find a way to contain the local oscillator bleed-through.  In the meantime, 
you can set the I/Q balance for each Softrock by cutting off the power to 
the others and using the transceiver/dummy load technique.  You need to 
given Skimmer a signal with signal to noise ratio of at least 25 dB or so 
before the I/Q calibration will work at all, but once it is calibrated it 
should not need to be changed until you change something about the setup.


>Normally Windows task manager CPU utilization is about 40% on a week night 
>with all 6 copies of CW Skimmer and the aggregators running.  I expected 
>the increased activity during the contest would bring the computer to a crawl.
>I was surprised to see CPU utilization was only around 60% during the 
>contest.  Summing the CPU % reported by the individual instances of CW 
>Skimmer would have been greater than 100%.


This is because Skimmer is reporting near-instantaneous CPU demands, while 
Task Manager is averaging once a second or so.  If the value reported by 
Task Manager is *ever* higher than that reported by Skimmer, then you're 
experiencing the occasional bug that I have found.  You don't mention what 
sound card sampling rate you were running - 96 KHz?


>All copies were set for aggressive callsign validation and from what I 
>could tell seemed to reasonably keep up (new stations spottable about the 
>same time) with other skimmers on dxwatch.com.
>
>Next time I integrate skimmer into a logging program I will make sure 
>"show unworkable spots" is turned off and then band map is zoomed 
>appropriately.  For a DX contest it would be worth trying to filter the 
>skimmer spots through something like VE7CC's ARUSER program to eliminate 
>the spots that aren't worth points and to slow the cluster window scroll rate.

Programs such as N1MM Logger have a variety of filtering options, as does 
AR Cluster.  I agree that turning off "show unworkable spots" in N1MM is 
essential.  My experience here was the opposite of yours, though - I wound 
up turning off the DX Cluster because I was getting too many spots from it 
that were not audible here - no doubt I could have filtered them 
better.  Meanwhile, Skimmer was turning out 10X as many spots as the 
cluster.  I ran either Aggressive or Paranoid validation all weekend - and 
used the masterdx.dta file with Paranoid so that I did not get US spots.


>I have noticed that this release of CW Skimmer is very stable.  I have not 
>had any crashes.
>
>I would find the option to start skimming upon program startup helpful.


Thanks, Dean!  Very helpful report!

73, Pete