Dear LF Group,
Thanks for the interest in the "Wolf" BPSK trials. To make a start, I
intend to transmit a test signal this evening, in order to make
available a real signal for people to record and experiment with. Mr.
Murphy has intervened as usual, and I have to fix a problem with
my synthesiser, but I hope to be QRV around 2000utc. The format
of the signal will be:
Frequency: 137.5000kHz. If occupied, I will QSY up or down by
50Hz.
Power: about 300mW ERP
ID: Normal CW at 6wpm, approximately on every hour and half
hour, followed by a few minutes of carrier for frequency
measurement purposes.
The rest of the time, the signal will be BPSK in the Wolf format.
This is 10 bits/second, and consists of 96 second frames that
repeat continuously. The Wolf program can operate on a signal of
length up to about 25 minutes, so if you record from 5 minutes past
the hour/half hour, for 25 minutes, you will have a full set of data.
However, if the signal is reasonably strong at your QTH, only a few
minutes of recording is required.
The web pages contain a fair amount of info, but here are some
hints from my experience.
The signal must be recorded at 8000 samples/sec - I used
DL4YHF's Spectrum Laboratory to do this; it also produces a
spectrogram at the same time. Most sound editing software seems
to offer a choice of sample rate too.
The program expects a nominal audio frequency of 800Hz. It does
not work well with a very strong signal; I found that with a local
signal it was neccessary to reduce signal level to 20 or 30dB
below the sound card's overload point to get good results. KK7KA
says that signal level should be -20 to -50dB below full scale. The
noise level should be at least -6dB on full scale. You can alter this
on a recorded signal using sound editing software if neccessary.
It is neccessary to calibrate the soundcard sample rate accurately.
This can be done by feeding an audio signal with accurately known
frequency into the soundcard, recording a few minutes, and then
running Wolf in the -m (frequency measure) mode; the true sample
rate is then (true f / measured f) x nominal sample rate. In my case,
this worked out to 7954.801 samples/sec. It is essential to do this
on receive, although the web page info could be interpreted to
suggest it isn't
You can do a similar thing to determine the frequency error in the
receiver; receive an accurately known frequency, run it through
Wolf in measure mode, which gives you the frequency offset.
Using the Spectrum Lab spectrogram to measure the apparent
audio frequency gave very similar results, but with less resolution
than the Wolf software.
If it does not work first time, you can use the same recorded signal
as many times as you like, until the settings are right. Once I had
level, sampling rate, and frequency offset correctly set, I got
perfect decoding straight away.
Hope this is some help - please let me know if you have queries,
suggestions for improvements, flames, etc.
Cheers, Jim Moritz
73 de M0BMU
|
