Oh Wow thanks Steve - yes txt channel was working - but corrupting one of the codec bits! The last stage of the LSP quantiser for 700B. So any time txt was inserted - we were inserting bit errors!
Those using 700B might want to empty the txt string (callsign box) for now, until the next FreeDV GUI release. Or perhaps compare the speech quality - I am not sure how noticeable the issue was. Good catch Steve - thanks. I've checked a fix into codec2-dev and confirmed it works. Open source really is a great way to trap bugs. BTW Jean-Marc - the latest 700B uses a three stage VQ, so could benefit from sorting index assignment as well. It's actually interesting that the VQ performs about as well as the scaler quantiser (700 mode) in the presence of bit errors. I suspect it's because 700B sounds better than 700 with no errors, so has "further to fall" with bit errors. Cheers, David On 06/09/15 06:48, Steve wrote: > In codec2.c the function codec2_get_spare_bit_index() returns "26" for > both modes 700 and 700B. > > I think it should be "26" for mode 700, and 27 for 700B. > > "27" is the spare bit in 700B, while "26 and 27" are the spare bits in 700. > > FYI > > But if this is true, then the data shouldn't be working, so I was > wondering if users are receiving data OK?? > > Steve > > ------------------------------------------------------------------------------ > _______________________________________________ > Freetel-codec2 mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/freetel-codec2 > ------------------------------------------------------------------------------ _______________________________________________ Freetel-codec2 mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/freetel-codec2
