Hi Rick, Yes now we have a modem running it would be good to try alternatives like TCM. Hopefully it won't affect other parts of the modem much like frequency offset or timing estimation.
Can u pls point me some TCM functions for that map roughly to the fdmdv.c bits_to_dqpsk_symbols() & qpsk_to_bits() functions? I am wondering if the 3dB gain in performance occurs at the low SNRs and high _decoded_ BER we can use with the codec (e.g. 1% BER after FEC decoding is OK for Codec 2). For example with TCM and a 3dB coding gain can we operate at 1dB SNR with the TCM decoder having an output BER of 1%? Looking at the mess HF makes of modem signals (e.g. wiping out tones completely) I also think it would be useful to look into interleaving. My gut feel is a few 100ms of interleaving will be OK on half duplex Push to talk systems. However I am not sure if this will be enough to help significantly. But lets experiment and find out. Cheers, David On Sun, 2012-05-20 at 12:47 -0400, Rick Muething wrote: > I took some time this weekend to read about the FDMDV (14 carrier + > pilot, 50 baud QPSK) mod/demod used. I am not sure I have all the > latest documentation. > > This is a good test vehicle for the CODEC2 and should allow some very > useful data to be gathered. > > Longer term I think a slight modification in this could offer a > substantial (2-3 db Power equivalent ) gain. (aside: in the HF modem > world we fight for every fraction of a dB!!!) > > It might work like this. > > 1) Baud rate, user data rate (to/from the CODEC2), carrier frequencies > etc all stay exactly the same. The bandwidth would be the same as the > current 4PSK scheme also. > > 2) The 4 PSK symbols are replaced by Trellis coded 8PSK symbols > carrying one FEC bit per symbol (e.g. code rate of 2/3) > 3) A fairly simple R=1/2 K=7 or perhaps lower K value encoder is used > to create the 8PSK code. Basically called Pragmatic Trellis code > moduation. > 4) Some interleaving is helpful but the tradeoff here is latency. For > example if you interleaved across all the carriers and used a Viterbi > look back buffer of say 28 symbols you would have added about 20 ms (+ > Viterbi decoding time) of latency. That is probably acceptable but > slightly longer latency and look back buffers might give better > performance on certain channel types. Probably would need to > experiment with that. > > This would require more CPU time but the bit rates required (1400 > bits/sec) are certainly possible with even older PCs and easy for > modern ones. > > In a WGN environment 8PSK Trellis should give 3 dB or so improvement > over uncoded 4PSK....equivalent to doubling the transmit power...a > very worthwhile improvement. For HF the number will be different but > still effective. I saw a substantial improvement in WINMOR going from > 4 PSK to Trellis 8PSK but that also had a second outer layer of RS FEC > which would not be needed here. > > Just some thoughts and a relatively easy to implement experiment. > Maybe using a HF Channel simulator would allow capturing enough > consistent data to determine the actual improvement (in dB or BER) > compared to standard non FEC 4PSK. > > Rick Muething, KN6KB > > ------------------------------------------------------------------------------ > Live Security Virtual Conference > Exclusive live event will cover all the ways today's security and > threat landscape has changed and how IT managers can respond. Discussions > will include endpoint security, mobile security and the latest in malware > threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ > _______________________________________________ Freetel-codec2 mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/freetel-codec2 ------------------------------------------------------------------------------ Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ _______________________________________________ Freetel-codec2 mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/freetel-codec2
