James Courtier-Dutton wrote: > With a bit of clever tweaking, (i.e. in Linux) one could pick a 64byte > boundary within the period and start outputting samples, and they will > reach the speakers, thus not having to wait for the period to end. It > would be a sort of buffer TX cut in feature.
Again, if you have hardware loopback, you can have effectively zero delay. That's not the issue in general. The question is, do you want to monitor input to the radio, or do you want to monitor from someplace within the signal processing chain? If the latter, you have no choice but to go along with the DSP buffering size. > The FFT issue has nothing to do with sound card capabilities, but there > is no reason why the FFT window should have to be in sync with audio > buffer periods. It isn't, even now. > I don't understand why or if FFT is used for CW TX anyway. I would have > thought that FFT would only be needed on CW RX. As presently set up the same DSP buffer configurations are used for RX *and* for generation of the CW signal in the audio passband. That's because the audio subsystem doesn't know from TX or RX. It's just moving sample buffers around. The only way that would change would be to run separate RX and TX processes, which is the way things are headed. For the time being, however, since the SDR-1000 is half-duplex, the DSP software is also. FWIW, the latency issue is somewhat minimized under Linux since the keyer (and CW tone generation) live in a separate process from the DSP, and the sidetone is routed directly to the output without delay, while the actual transmitted tone is routed *through* the DSP. The buffer sizes are still consistent, but the sidetone arrives one buffer earlier than the transmitted CW. 73 Frank AB2KT _______________________________________________ FlexRadio mailing list FlexRadio@flex-radio.biz http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz Archive Link: http://mail.flex-radio.biz/pipermail/flexradio_flex-radio.biz/ FlexRadio Homepage: http://www.flex-radio.com