Hello Vojtech, RR for all.
>As far as I know there is a metrics returned by the viterbi decoder. I >still wonder what happens if the metrics is averaged in every channel >and after a while the best channel is selected. You have more You are right. It is returned an error metric by the Viterbi decoder (a sort of Hamming distance). At the maximum, it could be possible to use for the AFC, but not for discriminate channels as the correlation is too loose. >I would like to add the RS ID feature to PocketDigi someday. This >would mean to rewrite the algorithm to fixed point and optimize and I >would need your source code or at least good description of the >algorithm. Would that be possible, please? I take contact with you. 73 Patrick ----- Original Message ----- From: Vojtech Bubnik To: digitalradio@yahoogroups.com Sent: Wednesday, December 27, 2006 11:25 PM Subject: [digitalradio] Re: MFSK16 tuning improvement, was:best mode to use for weak signal HF work and Hi Patrick. Thanks for an answer. > It works on Olivia because this mode uses a transform (Walsh Hadamard), an interleaving and a pseudo-random function: so within a pack of bits defined in time and in frequency there is a very strong correlation and outside the correlation decreases rapidly. This permits to define the best band of frequencies (maximum correlation). This seems to explain the meaning of the pseudorandom sequence in Olivia to me. I am working on port of Olivia to PocketDigi, which means to rewrite everything to fixed point arithmetics again. I wonder if my old iPaq 3630 will be fast enough. > This does not exist in MFSK16 where a convolution code is used (so there is no such correlation). Moreover, the S/N measure is too much imprecise, the selection would be almost random on weak signals. As far as I know there is a metrics returned by the viterbi decoder. I still wonder what happens if the metrics is averaged in every channel and after a while the best channel is selected. You have more experience than me to guess, but I hoped you love experiments and will give it a try :-) > If you use RS ID (RX and TX), the tuning is automatic. I would like to add the RS ID feature to PocketDigi someday. This would mean to rewrite the algorithm to fixed point and optimize and I would need your source code or at least good description of the algorithm. Would that be possible, please? 73, Vojtech OK1IAK
