Thanks Steve and Glen, IIRC James' use case was a home brew SDR with complex baseband inout, rather than a traditional SSB Tx. Interesting points about modulator distortion Glen.
Another approach might be some optimisation of the code, SIMD vectorisation, gcc tricks etc. Cheers, David On 18/12/18 07:54, glen english wrote: > Hi STeve > > I would think a 40dB stopband would be sufficient. > > bear in mind the modulator distortion in the rig will probably be about > no better than 3% at best. > > IE put a 600Hz tone in your sideband radio and at various drive levels, > have a look at the harmonics... > > there will be harmonics from ahead of the sideband filter, (whatever the > implementation) . > > then have a look at IMD products with say 1000Hz and 500Hz. probably at > best -25 down. > > > > On 18/12/2018 8:14 AM, Steve wrote: >> I have played around by using a smaller number of coefficients. >> Currently the filter uses 160 and I created a filter that uses 63 >> instead. >> The current filter is very sophisticated (designed by James Ahlstrom), >> as it uses complex coefficients (essentially doubling the numbers I >> listed above, one for the I and one for the Q sides. This filter can >> be tuned to other center frequencies. >> >> It's kind of neat how it works, but you need about 2 GHz CPU. It >> provides 100 dB attenuation for a @1100 Hz BW and looks great on the >> SDR spectrum view :-) >> >> My change to 63 filters cut the time in half, but I only used a 40 dB >> attenuation. This left some higher amplitudes on the ends, but >> probably acceptable. I didn't transmit the signal, just looked at it >> from a saved file in Audacity and using FreeDV spectrum view. >> >> I also tried an 81 coefficient filter with 55 dB attenuation, and to >> tell the truth I didn't see much change. >> >> Be a good area to experiment on the air. I believe the filter was >> implemented early on because there was kind of a step in the spectrum, >> so it cleaned that up, but I don't think the step would be a killer, >> and leaving the filter off if you are out of CPU ought to be fair game. >> >> 73, Steve >> >> >> >> >> _______________________________________________ >> Freetel-codec2 mailing list >> Freetel-codec2@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/freetel-codec2 > > > > > _______________________________________________ > Freetel-codec2 mailing list > Freetel-codec2@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/freetel-codec2 _______________________________________________ Freetel-codec2 mailing list Freetel-codec2@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/freetel-codec2
