The fundamental concept is that every symbol is precisely the same length and is time aligned with all the other symbols sent during a QSO by the sender. The sender does not need better than TCXO stability.
This requires :
Implementing a dit frame clock which can be set to various durations. High accuracy not required. Just stability.
Logic to align non-aligned input before output. User must hear side tone without latency. Receiving station can't detect latency.
Logic to set the alignment of the receiver dit frame for the statistical processing
Logic to compute the mean and standard deviation of the absolute amplitude of the receiver audio output during a dit frame.
Computing the FFT of a Gold code sync signal over a 10 s window and then doing a linear fit to the phase delay to align the dit frame
Periodically doing the FFT & linear regression to update the phase delay every Nth character
I *think* everything else is just a combination of the above, but I could easily have missed some minor things.
I think anyone with any embedded experience will see this is not difficult code to write.
Have Fun!
Reg
