On Thu, Jul 3, 2008 at 10:43 AM, Brian Lloyd <[EMAIL PROTECTED]> wrote:
> Once you have the signal in the digital domain it makes NO sense to > convert back to analog. Except when there's no alternative ;-) Some of the modes will suffer worse than others. One thing to keep in mind is that many of the soundcard digital programs are built around voice-channel bandwidth and common consumer-grade soundcards. You have to resample the...ahem...high-quality SDR audio anyhow. Even a weak-signal program like WSJT is doing some serious mangling to the signal in software on the radio side of the software codec. As it is right now, almost no common digimode program is really set up to take advantage of what all-SDR can provide. The digital paths are there for convenience, mostly, not performance. > > In the first case, a third party program called VAC is used to > > create a virtual audio cable (VAC) between PowerSDR and the digital > > mode sound card program. In this configuration the AF between the > > two programs is kept entirely in the digital domain eliminating any > > noise or distortion introduced in the multiple A/D and D/A > > conversions. > > That sounds like the right answer. Under Linux and OS X this is easy since the "modern" audio subsystems are designed with full-duplex patching among audio apps in mind. However, it *does* make some sense ultimately to put codecs into the SDR DSP, for a variety of reasons. But this transposes the problem to a different area. On the application side of the codecs, the data will be bits, not samples. What's the transport yoga for these bits? If they represent varicode, where does the corresponding binary codec live? Furthermore they could be soft and not hard bits. Is the idea to put every variety of Viterbi decoder into the SDR? If not, what's the protocol? None of these questions has even begun to be asked yet, much less addressed, and it's lunacy to plunge ahead with a one-off solution based on our own idiosyncratic needs. It's a further safe bet that the existing digimode programs are far, far away from being restructured to handle binary data I/O rather than digitized audio, if indeed they can be restructured at all. Very few of the ones we've been able to examine at the source level (far too few of them BTW, since so many of the authors lock up the source) are designed for the kind of concurrency that would make that job palatable. Also BTW, these are exactly the issues that arise in building a systematic solution for digital voice, but in spades, since in that arena it will be almost essential to have a fairly tight feedback loop between the codecs and the DSP, irrespective of what the audio payload might be. The current generation of digimode programs, by and large, isn't up to that job in particular. 73 Frank AB2KT -- "Sapristi nabolis!" -- Count Jim Moriarty _______________________________________________ FlexRadio Systems Mailing List FlexRadio@flex-radio.biz http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz Archives: http://www.mail-archive.com/flexradio%40flex-radio.biz/ Knowledge Base: http://kb.flex-radio.com/ Homepage: http://www.flex-radio.com/
