Rik,
probably your idea would work, but I think that if you can achieve
phase coherence then the
best technique would be to compute the two correlations between the
received signal and
what the signal should be in case either of a "1" or of a "0", then
select the symbol that
gives the highest correlation value. And. if you have CPU cycles to
spare, which probably is
the case for low signaling rates, then you could even correct for small
phase errors by
computing several correlation sums, each shifted by a few degrees,
finding then the peak.
And adding a convolutional FEC coding with a soft-decision Viterbi
decoder would certainly
be of no harm...
73 Alberto I2PHD
-------------------------------------------------------
Rik Strobbe wrote:
Hello Alberto,
phase coherence maybe a bit tricky.
I am not a mathematic expert, but just intuitive I would say that
knowing the exact time of begin and end of a dot could be used to
improve SNR.
One possibility that comes to my mind :
Assume
* a series of dot sequence (let's say 5 = .....)
* a dot length of 1 minute
* each dot starts exactly at the full minute
Now take one FFT every second, with the samples from the last 60
seconds. the subsequent FFT's should show a triangular wave that has a
period of 2 minutes and is in phase to the minute rytme.
Maybe using the right technique (kind of FFT of the FFT ?) one could
detect this 2 minute triangular wave at signal levels that are to weak
to bring up the dots with ARGO or similar software.
Makes this sense or it is just a stupid idea ?
73, Rik
|
