Hi Jean-Marc, For the VQ-ed 1200 bit/s mode - no not yet. But happy if some one wants to take a look at that.
The 1300 and 700 bit/s modes in common use with FreeDV use scalar quantisers. I did study the effect of gray versus natural binary last year: http://www.rowetel.com/blog/?p=3079 Cheers, David On 05/09/15 07:26, Jean-Marc Valin wrote: > Hi David, > > Just one thing that came to mind because you're talking about FEC... did > you ever look at optimizing index assignment in codec2 to minimize the > effect of bit errors? > > Cheers, > > Jean-Marc > > On 09/04/2015 05:42 PM, David Rowe wrote: >> Hi Thomas, >> >> It's great to see you have been thinking about these issues. As Codec 2 >> is open source, you can experiment and try them out. >> >> Some thoughts: >> >> + FreeDV does not have a hard threshold like other DV systems. It uses >> "brain FEC" - this is part of the design Mel Whitten has been advocating >> for many years. >> >> + FreeDV 700B uses a coherent QPSK modem (rather than DQPSK) and >> diversity - rather than reducing the bandwidth it is spread out in >> frequency to combat fading. This has given significant gains in >> simulations and now proven real world performance exceeding SSB. >> >> + FreeDV 1600 uses partial FEC of selected bits, however we are getting >> equivalent or better results with no FEC with our latest 700B waveform. >> >> + FEC, in general, has problems when applied to narrow band HF DV: >> >> http://www.rowetel.com/blog/?p=4147 >> >> + I don't think dynamic reduction of the rx bandwidth alone will help. >> The SNR of the remaining carriers will be the same, each carrier has the >> same power as before and the noise power in each carrier is the same. >> >> Cheers, >> >> David >> >> On 04/09/15 19:11, Thomas Schiller wrote: >>> Hi David, hello List, >>> >>> thank you for your work on FreeDV and Codec2. I'm strongly >>> interested in a better HAM-DV which must be solve different >>> demands rather than commercial DV. One major DV-drawback is >>> the hard decision cutoff at low s/n. While in SSB and even >>> FM the demodulated s/n is rather proportional of the HF s/n >>> all DV-systems have a stong threshold. This is a good approach >>> for commercial but a poor behavior to HAM. We like to use our >>> brain to listen inside of the channel noise to get the last >>> sample biological-approximated to what we often already expect >>> to know or hear (for example in a contest). >>> >>> Your attempt to use a 2nd carrier points in the right direction. >>> I've written a suggestion to a HAM-defined DV-system in the german >>> amateur journal FUNKAMATEUR ("Analoge und digitale Sprachsysteme >>> für den Amateurfunk"; FA12/12, Page 1260-1263). >>> >>> http://archiv2.funkamateur.de/search >>> >>> The idea is to use at least four D-QPSK carriers shifted to minimize >>> interference. The inner should be modulated with the basic parameters >>> and a strong FEC. The carriers with more distance to the central >>> should carry information to enhance the quality and could get a weaker >>> or at maximum center-distance may be no FEC. The distribution of >>> parameter bits to the carriers must be carefully choosen out of a list >>> of significance of your CODEC2 or any other codec. As far as I know >>> nobody has followed this idea before. >>> >>> Doing so, the RX then has the chance to dynamically reduce the received >>> bandwith if the s/n is getting weak. Using only half of the carriers >>> could enhance the s/n up to 6dB which is very much! Well of cause it >>> will sound then more "spacy" below the decision point. But this is >>> much more similar to what we hear on analog SSB or FM if the HF-signal >>> becoms weak. At least one can listen to the behavior of the cannel and >>> it's a nice gift to enhance the covering range at the same tramsmitting >>> power! This is amateur's most wanted feature at all ;-) >>> >>> This is very different to commercial DV-systems which we currently use >>> to "recycle" or adapt to an HAM application. Commercial DV needs a >>> fixed s/n decision due to their cell-design and the quality necessities >>> of the ordinary telephone user. In fact cell-design is not really >>> compatible to what amateurs need on short wave. Hence we will never get >>> below this interesting cutoff level if we don't put dynamic bandwith >>> adaption to the HAM-DV functional specification. Do you agree? >>> >>> It would be very nice to do some experiments to find out what one can >>> win in practice on a real HF-channel. >>> >>> 73, Thomas - DC7GB >>> >>> ------------------------------------------------------------------------------ >>> _______________________________________________ >>> 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 >> > > ------------------------------------------------------------------------------ > _______________________________________________ > 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
