Dear LF Group,
Thanks for the reports and comments. Some pointers to receiving
WOLF mode BPSK-
The receive audio must be recorded at 8000 samples/sec., mono.
DL4YHF's Spectrum Lab can do this, and generate a spectrogram
at the same time. WOLF can operate on up to about 25mins of
sampled data; at about 1 megabyte per minute, the files are quite
large. With a good SNR, only a few minutes recording is required,
but more data is needed for low SNR.
In order to decode successfully, 3 parameters need to be set up
correctly:
-The level: The signal must be about 20-40dB down on the full
scale level of the soundcard- WOLF simply does not work properly
with a strong signal. If everything else is OK, what you see is that
the first 3 lines of output are correct, but the rest is gibberish. The
low level can be achieved by reducing the RX gain, or post-editing
with sound recorder software. If the signal is down in the noise,
KK7KA says you should set the level so that the noise peaks are
6dB below full scale.
- The soundcard sampling rate correction (-r switch); The
frequency error due to the soundcard is often the biggest
frequency error in the whole system (about 4.5Hz in my case). It
varies with the particular soundcard, and the particular sample rate
it is set to. It also upsets the timing of the decoding process. One
way to determine the actual sample rate is to inject an accurately
known audio frequency into the sound card, and measure the
apparent frequency using the Wolf -m option; the sample rate is
then (8000 x real frequency)/apparent frequency. Once
determined, the correction seems to be stable. KK7KA describes
the neccessary accuracy as being in the parts per million range for
optimum performance, ie you are aiming for better than 0.1
samples/second accuracy.
-The frequency offset (-f switch): This corrects for frequency errors
in TX and/or RX tuning; it can be determined by receiving a signal
of known frequency, and measuring the received audio frequency
using the Wolf -m option again. This is perhaps less critical than
the soundcard sampling rate error; about 0.1Hz or so is tolerable.
The obvious problem is how to determine the soundcard error, if
you don't have a accurate audio frequency source. I used the
reference output of my synthesiser divided doown to 1kHz (and
filtered; a fairly clean sine wave seemed to be neccessary to avoid
aliasing problems with the soundcard). You could use the RX audio
out, but only if you knew what the RX frequency error was to within
a small fraction of a Hz. The effect of having the wrong sampling
rate is that the first few lines of decoding work OK, but after that
you get gobbledygook, so trial and error is a possible, if time
consuming method. I think a technique where you measure the
difference between two audio output frequencies might work OK,
but this needs some further work to check it out.
All this is a bit of a pain, but once completed, Wolf seems to work
straight away all the time.
I'm not sure about the effects of QRM on this signal - obviously, it
has been designed to work with low SNR, but does it's
performance with gaussian noise also apply to QRM such as
carriers, or pulsed waveforms like mains noise and Loran?
I intend to run some more tests this evening, starting about
2000utc, at about the 30mW ERP level for an hour or so. Later , I
will be running the full 1W ERP for possible transatlantic reception,
between at least 0000 - 0200utc. As before frequency will be
137.500kHz. If anyone would like a different frequency, power
level, time etc. please let me know.
Cheers, Jim Moritz
73 de M0BMU
|
