[Lowfer] Testing Modes

[Stewart Nelson]

Hi Lyle and all,

With a little cheating, I managed to solve the CW puzzle.  The actual
solution is not in this post, in case you want to try your own.

First, I decided to use the "Filtered CW", even though some useful signal
was discarded, because an MP3 full of random noise would quantize the
desired band severely.

Then, did some processing with Cool Edit.  Read in the MP3 and converted
to 2000 samples per second (to better see the FFT).  Found a spectral line
at 449.84 Hz.  Modulated with a sine wave at that frequency.  Saw a few
cycles of beat, measured, remodulated at 449.8166 Hz.  Manually played with
the phase to maximize the DC component.  Lowpass filtered to 50 Hz, to
remove the mixer sum terms.  Resampled at 125 Hz.  One of the least corrupt
words, "whereas", can be seen at http://www.scgroup.com/ham/where1.gif .

But most of the message looked pretty bad, so I tried to sample at the
maximum eye opening.  Couldn't think of an easy way to do in Cool Edit,
so wrote it out as ASCII text, processed with a perl script, and read it
back into Cool Edit for display. http://www.scgroup.com/ham/where2.gif .

Still looks pretty bad for machine decode.  Thought about doing a
dictionary based maximum likelihood search, but punctuation, numbers,
and abbreviations make this pretty complex.

Time to cheat.  Manually decoded the next 3 words.  Typed the 4-word
phrase into Google.  Just one hit, Murphy must be asleep.
http://www.brainerdham.org/2003/2-27-03-meeting.html .  Verified that
the beginning and end of the CW matched the document text.
Voila - perfect copy :)

73,

Stewart KK7KA


> Just for kicks, I recorded locally-generated PSK31 and CW signals,
> attenuated both of them by the same amount, and mixed them with white noise
> generated in CoolEdit. MP3 files (unfortunately rather large; about 500 kB
> each) of these simulations are available at
> http://www.qsl.net/k0lr/pskvcw/pskvcw.htm.