Message-ID: <003201bfce81$f1a552f0$0a00a8c0@server1.ThreeLakes.ca>
From: "Larry Kayser" <kayser@sympatico.ca>
To: rsgb_lf_group@blacksheep.org
Subject: LF: a really weak signal in Canada....
Date: Sun, 4 Jun 2000 20:06:39 -0400
MIME-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk
Reply-To: rsgb_lf_group@blacksheep.org
Sender: <majordom@post.thorcom.com>

Greetings:

Bill de Carle has aggressively gone after my ultra weak signal on 137.710
and managed to copy it.  He sent the following email to the LowFer
reflector, I am resending it here for those who might be interested.  With
the antenna finally up here I am working on the test set so I can tune up
the antenna.  The signal is QRS CW at the speed of .4 WPM.

Quote
I don't know if this has been tried before, but I just tried it and it
worked nicely.

The general idea is to record an ultra-slow CW signal, then to play it
back speeded up so the human operator can copy the CW message by ear
at normal speed.  Ham operators have years of experience trying to dig
weak CW signals out of the noise - by ear - not by looking at a picture
on a computer monitor.

I used VA3LK's ultra-weak test signal.
Using FFTZZ, I knew the signal was being received at 803 Hz - after some
minutes of integration, the spectral line came up out of the noise, so I
knew beforehand what the exact frequency was.  Unfortunately the signal
was way too weak to be able to decode by any spectral display technique
available to me, so I hit upon the idea of time-compression.

I recorded some 578 seconds of audio to hard disk at 7200 samples per sec.
Then I post-processed that file as follows:

1.  Run it through a narrow bandpass filter centered on 803 Hz.

2.  Multiply the resulting data with a sinewave at 825.4 Hz - that
    acts like a mixer producing sum and difference frequencies at
    1628.4 and 22.4 Hz respectively.

3.  Run that waveform through a 32-point FIR lowpass filter to keep
    only the 22.4 Hz component.

4.  Keep only 1 resulting filtered sample out of 32 - essentially
    compressing the total recording time by a factor of 32.

5.  Make it into a .wav file specified as sampled at 8000 s/s and
    lasting 16.25 seconds.

When I played the 16.25 second wav file back I could actually hear the
CW at a reasonable speed (about 35.5 times faster than it was transmitted)
and at a reasonable tone (about 800 Hz).  The ident was easily recognized.

It worked on first try.  I will now optimize the filter coefficients and
write a single program to go directly from digitized samples on disk at
7200 s/s to a .wav file time-compressed at 8000 s/s.

Anybody else tried this?

Bill VE2IQ

Unquote

I have the .WAV file and will send to anyone who asks.

Larry
VA3LK