[Lf] New software available for copying weak QRS-CW]
Andre' Kesteloot
akestelo at bellatlantic.net
Thu Jun 8 17:53:09 CDT 2000
Bill de Carle wrote:
> CRUNCH Version 1.0 from VE2IQ
> -----------------------------
>
> Hams have been experimenting recently with ultra-slow speed CW (QRS-CW).
> One standard is sending at around 0.4 words per minute, or 3 seconds
> per dit. It takes a very long time to send a message and it cannot be
> decoded by ear. Up to this point the technique for recovering the message
> has been to display the spectral output from a FFT program on the computer
> screen graphically, and try to pick out the dots and dashes with the
> human eye. There are some advantages to this method - you don't have to
> know the frequency very exactly and the eye can recognize patterns in
> noise quite easily. There are also some disadvantages, mainly that the
> eye doesn't work very well at discriminating subtle differences in
> brightness or color and sometimes it is very difficult to determine
> whether the key was down or up when all you have to go on is a display
> full of background noise and a very faint trace that may or may not be
> there. Even if we have 16-bits of precision in our processed data,
> the eye can't use it. But how about the ear? High-fidelity buffs
> will tell you they can "hear" the difference between 8-bit audio and
> 16-bit audio.
>
> It occurred to me that hams have been copying CW by ear for a long time
> and we seem to have developed a knack for digging weak signals out of
> the noise by using the processing power between our ears, hi! Some CW
> people are amazing in their ability to "hear" signals way down in the
> noise. The human ear is very good at detecting coherent but very faint
> audio signals in white noise and reading the CW under those conditions.
> I also know that if we compress a slow-CW signal in time about 35 times
> we can effectively reduce the bandwidth by 35 times and increase the
> SNR correspondingly.
>
> I was trying for a day or so to copy VA3LK's ultra-slow CW signal and
> not having much success. My FFTZZ program, when set for a long integration
> period (15 minutes) was able to *detect* the signal and show its frequency
> but I was not able to decode the message by any conventional means.
> So I recorded about 10 minutes of the receiver's output and tried a few
> tricks to process the resulting audio off-line. Eventually I was just
> able to copy the message. I decided to write a program that would do
> all the off-line steps in real-time, so we wouldn't have to tie up huge
> amounts of disk space to store an overnight monitoring session.
>
> The first draft of this program is available on my web page. It's called
> "CRUNCH" - why not download it and try it for yourselves?
>
> Go to http://www.ietc.ca/home/bill/bbs.htm
>
> The basic idea of this program is to time-compress very slow-speed CW
> received off the air so that it can be played back about 35 times faster
> and "heard" with the human ear.
>
> You will need a relatively fast computer (Pentium will certainly work,
> maybe a fast 486?) - and one of my Sigma-Delta interface boards. You
> do *not* need a sound card for recording, only for playback. So you
> may be able to press an old laptop into service. CRUNCH runs under
> DOS. You don't need Windows. It can use any serial port. You can
> specify which port as a command-line parameter when you invoke the
> program. (E.g. "CRUNCH COM=2").
>
> The Sigma-Delta board samples received audio at 7,200 samples per second,
> converts it to 8-bit digitized format, then sends it into the computer
> via a standard serial port. The circuit is described on my web page and
> also in my article "A Receiver Spectral Display Using DSP" which appeared
> in Jan '92 QST. I still have parts kits available: contact me for
> details if you're interested.
>
> DSP software running in the computer performs the following steps:
>
> 1. (optionally) passes the incoming audio through a narrow bandpass
> filter centered on 800 Hz. This is a good idea when receiving slow-speed
> CW because the signal bandwidth is minimal. The filter has a Butterworth
> shape and a width (at the -3dB points) of 50 Hz.
>
> 2. Mixes the signal down to a target frequency of 22.5 Hz.
>
> 3. Lowpasses the output from the above mixer stage, then reduces the
> sampling rate by some 32 times. Since there are no high frequency
> components left at this point, we can completely capture the processed
> audio tone with 225 samples per second, well over the Nyquist rate.
>
> 4. Saves the new sampled waveform into a 16-bit PCM (.wav) file,
> stating in the .wav header the sample rate was 8000 samples per second.
>
> Saying that the sample rate was 8000 samples per second when in fact
> it was only 225 results in two things: 1) a 35.55 x time compression,
> and 2) multiplies the nominal 22.5 Hz derived carrier frequency back up
> to 800 Hz, where it can easily be heard with the human ear.
>
> Even though the Sigma-Delta board is only an 8-bit ADC we still generate
> a 16-bit .wav file because the 32x time compression plus the DSP filtering
> makes 16 bits of useful resolution available at the lower sampling rate.
> Sometimes a QRS-CW is so weak it can only be heard with 16-bit audio.
>
> The program may optionally capture data from the SD board in real time
> (without compression) - but it will produce an 8-bit .wav file only in
> this mode. Users are cautioned that leaving it running all night will
> use up gobs of megabytes of hard disk space.
>
> How to set up an overnight monitoring run...
> --------------------------------------------
>
> The most usual operation will be to leave it running with the time
> compression mode enabled. That way if you record 8-hours worth of
> slow CW you will be able to play it back in the morning in just 13.5
> minutes! And you can be sure that if there's any CW in there you won't
> miss a single word. You will of course have to be able to copy CW at
> around 14 wpm by ear but that shouldn't be any real problem. Right?
>
> Since we *mix* the incoming audio down to 22.5 Hz (a frequency trans-
> lation process) then *multiply* it back up to 800 Hz, it is relatively
> important to know the exact frequency of the QRS-CW signal we're trying
> to copy, even though it won't be audible in most cases. I strongly
> recommend running FFTZZ for a few minutes to detect the presence of the
> signal. If FFTZZ can't detect it after 15 minutes of integration
> you're not going to be able to copy it anyway. You should set up FFTZZ
> in its 8192-point FFT mode, with a frequency range of 700 - 900 Hz, and
> integrate with the software AGC enabled. Any narrow audio filtering
> you can do in the radio before the signal gets to the Sigma-Delta board
> will only help - the bandwidth of the QRS-CW signal is essentially
> nothing. After a while you will see a single spectral line making its
> way up out of the background noise. That's the slow CW carrier. You
> can now position the mouse cursor on that spectral line and read out
> its frequency to the nearest Hz. Write the frequency down, you will
> need it as an input to CRUNCH. The freq will have to be as close to
> 800.0 Hz as possible - a few Hz either side is OK as long as you know
> what the value actually is. You can adjust the audio level while
> running FFTZZ - set it as high as you can without getting any of the
> red "CLIPPING" messages.
>
> Bring up CRUNCH - and ask for the 800-Hz bandpass filter (default) and
> the time-compression (also default). The program will ask you to enter
> the exact frequency you wish to monitor (nominal 800.0 Hz).
>
> When it runs it will tell you each time it writes a 32K block of filtered
> data to the output file. Each 32K block has 16384 16-bit audio samples
> obtained at a rate of 225 samples per second, so it will write a new
> block about every 73 seconds.
>
> You can let it run all night and stop it in the morning.
> The resulting .wav file can be played back through any computer with
> a sound card and the necessary software. And you can play around with
> the audio - use your own noise reduction techniques or do additional
> filtering. The Cool Edit program is real neat for that sort of thing.
>
> My own experience has been that the .wav file generated by CRUNCH is
> just fine - and is easy to copy by ear. I have been able to get solid
> copy on VA3LK's QRS-CW signal that I could not copy by any other means.
>
> Try it, you'll like it!
>
> Bill VE2IQ
>
> To unsubscribe, send to MAJORDOMO at qth.net "unsubscribe lowfer" (Do not
> send to list!!) Send on list submissions to lowfer at qth.net
More information about the lf
mailing list