Hi David I had not read the post <http://www.rowetel.com/blog/?p=3700 > OK, right you have done alot of work on this, much of what I have said is redundant.
as for cnr*BW (I am not nornalizing anything to 3000 Hz, unlike WSJT and Joe Taylor's data systems) so 3dB CNR and 500Hz BW is 2 * 500 = 1000. (the lower the better of course) either way, I reckon you are on a winner. The key have been to get the source information rate down : which you have done so (by virtue of your terrific codec) , allowing the modem to operate in an BW>1/t condition. On VHF, that is spared of selective fading on base station-base station paths , a single carrier technique will work well. On HF, the multicarrier , wideband will be significantly better than the single narrow carrier on many, multi hop paths, as the multipath as I am sure you are aware of will cause half of a SSB channel to disappear at times, especially difficult are paths that are high in vertical incidence (short distance 2-8 MHz) . These paths often have lots of modes, hops, IE lots of multipath, and also high doppler. Take a look at the DRM spec for ES 201980 http://www.etsi.org/deliver/etsi_es/201900_201999/201980/03.01.01_50/es_201980v030101m.pdf Even in Mode D, experience has shown that DRM has difficulty with these high doppler fast changing channels on vertical incidence paths around dawen and dusk (where the ionspheric height is changing rapidly). There is much to learn about learning ES 201980 cheers glen On 6/02/2015 1:21 PM, David Rowe wrote: > Thanks Glen - great to have independent confirmation of our 10dB-ish > system gain results. > > I started with the DStar GMSK pulse shaping coeffs which I think is > BT=0.5? Yes I think it's theortically about 1dB off PSK but I'm not > sure what "alpha" to put in the theoretical BER equation. It's all in > gmsk.m if you can read octave/matlab code. > > Yes I spent a Merry Xmas coding up a digital PLL, used the design from > the original GMSK for telephony paper back from 1981. refs and code in > gmsk.m > > Re better than SSB, lemme run BW*CNR on the scheme in this blog post: > > http://www.rowetel.com/blog/?p=3700 > > That works at -8dB AWGN SNR (3000 Hz), but the QPSK signal is actually > about 500 Hz wide. So the SNR (500 Hz) is -8+10*log10(3000/500)= -0.218 > or linear 0.95. So is that CNR*BW = 0.95*500? > > Not sure if I did that right. Sorry to other readers for all the comms > math! Then again ... this stuff will matter for Ham's in the 21st century. > > Cheers, > > David ------------------------------------------------------------------------------ Dive into the World of Parallel Programming. The Go Parallel Website, sponsored by Intel and developed in partnership with Slashdot Media, is your hub for all things parallel software development, from weekly thought leadership blogs to news, videos, case studies, tutorials and more. Take a look and join the conversation now. http://goparallel.sourceforge.net/ _______________________________________________ Freetel-codec2 mailing list Freetel-codec2@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/freetel-codec2
