RE: [Discuss-gnuradio] PMR data decode

[Johnathan Corgan]

On Tue, 2002-07-09 at 08:32, Matthew Burnham wrote:

> Okay, I've been playing around today. Have managed to get a simple data
> slicer working and got some plots which I'll pop on the web in case
> anyone's interested (or can tell me the difference between the signal
> types). They're at http://www.btinternet.com/~mat_burnham/. 2sine.png
> looks to me like a sine wave, but I don't think I've got enough
> bandwidth to sample it properly.

1.png appears to be 4FSK.  If you quantize it to 4 levels instead of
two, the structure will be apparent.  The smaller peaks and troughs of
the waveform in the very center are the two middle levels.

The harder part then is figuring out what the two bits per transition
mean.  For example, in FLEX, these are two multiplexed data streams
rather than a single "double rate" one.

2sing.png appears to be AFSK, that is, Audio Frequency Shift Keying. The
modulation is one audio frequency for one symbol and another audio
frequency for the other symbol (remember, what your are seeing is the
baseband modulating signal, not the original RF).  However, the sine
wave isn't very clean; in fact, it looks more like a dirty triangle
wave.

Anyway, if you multiply the waveform by a one quarter cycle delayed
version of itself, and lowpass filter it, you'll recover the original
highs and lows.  Look up "non-coherent FSK demodulation" for a more
rigorous explanation. (And of course apply the algorithm to the baseband
audio, not the original RF as you would if this were true FSK.)

Alternatively, you can bandpass filter the two audio frequencies and use
an envelope detector to recover the 1's and 0's.

-Johnathan, KG6KWT