|
I have been working for some time with DSP and
have a development that may be of interest to the group, and especially those
using slow CW with FFT waterfall displays.
I am sampling the audio input continuously at 11025 Hz in 8192
sample blocks. I lose every other sample and do FFT's on the 4096 samples
remaining clocking through 2048 samples each time giving a bin size of 2.6Hz. I
use the fruits of the FFT output and plot the intensity as a pixel in a classic
Hellschreiber display, by which two identical displays are positioned one above
another and synch. problems are avoided because if the output scrolls off the
top display it scrolls on to the bottom. Also speed discrepancies can be
tolerated as these result in a display that slopes up or down but is still
readable. Most importantly for our purpose is the fact that the eye interpolates
and can discern characters that may have parts missing through QSB or are
corrupt due to QRM and I feel that the method may have some advantage where slow
CW is at its limit.
I have three speeds at present, 2 chars/min, 1
char/ min and 0.5 char/min.
I have got over the problem of tuning in to and holding a
channel 2.6 Hz wide by plotting ALL of 10 adjacent bins, so if you can tune to
within 26 Hz the display will decode somewhere.
If anybody would like to try the program I would be glad to
email them a beta copy. The program in written using Windows95 with C++ Builder
and uses a PC with soundcard. I use a Pentium 133 but there should be some
latitude as I am not making outlandish demands on the CPU. The output is audio
for an SSB TX or PTT/Keying using DTR and RTS of a COM port. I have run it with
a colleague locally on 28MHz but not pushed it to the limit. If the technique
shows promise there is some potential for further
development.
Ideally, we could do with somebody to beacon
whilst a number of others listen. (I am not equipped to transmit on LF at
present).
73
Lionel
G3PPT
| 