CwGet is ok, but try MRP40.? It's the best CW copier around.
From: Mark Amos <mark@...
Sent: Thu, 30 Apr 2009 5:16 pm
Subject: [dxsoft] CwGet "accuracy"
I did a completely unscientific measurement of CwGet "accuracy" and thought
I'd share my results here. Overall, the results were about 98% accurate
I copied W1AW code practice at 20 WPM (maybe 18?) and compared the results
to CwGet and have detailed the results below. Also, at the bottom is the
text I copied and the text copied by CwGet.
For the "test", the radio I used was an HPSDR and a 1/4 wave vertical on 40
meters running PowerSDR software feeding the audio to CwGet. The signal
strength varied from S5 to S9+10 over an S2 noise floor (on a calibrated S
meter.) The missed characters were during dips in the signal. In all
fairness, I had trouble with them too (in two of the drop outs), but could
guess the words based on context. There were only three of these during the
I was using a 25Hz filter (in the Radio) and a 140Hz filter in CwGet with a
500 lpm burst filter.
By my count (which is probably off - I didn't count punctuation errors,
since I don't generally include punctuation in my copy) it appears that
there were 3 word mistakes out of 105 words. The letter error rate was about
9 characters mistaken out of 550 characters. I didn't bother calculating the
element/bit error rate - but I suspect it was less than 1%. Obviously there
is no error correction with CW so you get what you get...
This is an insufficient test to conclude much about the accuracy of CwGet
(an N of 550 is pretty small and the conditions were good.) With fading,
QRM, a bad fist, etc., etc. the accuracy of the program would suffer (as
would mine.) I have a better processing engine for syntax and grammar and
error "correction" (my brain) which gives me an advantage in filling in
missing letters/words. On the other hand, CwGet can copy accurately at 50
WPM (at that speed my accuracy is about 20% comprehension - I can't copy
accurately over 30 WPM no matter how good the conditions...)
In any case, I think it's a pretty good tool. I've tried a few others and
CwGet seems to be head-and-shoulders above the others I've used.
FYI, there's a neat article in HAM-MAG this month on a "hardware" decoder
(uses PIC software) that looks like it would be fun to play with. I haven't
looked at the code, but I suspect it won't be much competition to CwGet. I
wrote some CW software in 8080 assembler back in the 80s and while it copied
machine sent code (via a keyer) at 100%, it could only do it at 12 WPM and
it barely worked at all on-the-air...
Anyway, I just thought some of you might be interested in my "unscientific"
county search and rescue in washington state I didnt have their number so I
called my local 911 dispatcher all they had was the info for king county in
washington so I called them and they gave me the number for snohomish. when
I got hold of snohomish county search and rescue they asked me to obtain
additional info from russ such as the color of his tent and if he was in a
clear or wooded area and remain in contact with him as long as possible
williams said russ and I were able to maintain contact until about 8 pm on
county search and rescue in washington state. i didnt have their number, so
i called my local 911 dispatcher. all they had was the info for king county
in washington, so i called them and they ga e t me the number for snohomish.
when i got a hold of snohomish county search and rescue, n e tw asked me to
obtain additional info from russ, such as the color of his tent e nd if he
was in a clear or wooded area, and remain in contact with him as long as
possible, williams said. russ and i were able to maintain contact until
about 8 pm on sunday,
[Non-text portions of this message have been removed]