The soundcard becomes a liability for coherent signalling
at LF. The fiddly software is all concerned with coping with
the wandering soundcard conversion rate.
> direct I/Q downconversion from LF to baseband, low pass
> filtering then I/Q sampling into an A/D at something like
> 1kHz (or even lower),
Yes, a small amount of hardware replaces a lot of software
once you move away from the soundcard.
And because the signal is shifted to a carrier frequency of
zero, the A/D sampling of the baseband I/Q doesn't need to be
super accurately timed - it merely affects the symbol timing,
which needs only 1/10th second accuracy.
No doubt there is scope for some direct conversion LF
electronics to take over most of the work (certainly the
timing critical part). An opportunity here to revive some
traditional techniques that are simple and ideal for this.
Time to revisit some projects from a couple of decades ago.
Just needs a GPS-locked DDS for the LO and a simple A/D for the
I/Q. Probably not many bits wide - 8 plenty, maybe 2 enough.
100Hz sampling is more than fast enough considering bandwidths
are 1Hz or less.
Even the DDS probably doesn't need to be very good. Needs the
long term stability of a GPS lock but it probably doesn't
matter if it has 5 degrees or so RMS phase noise.
I suppose at LF, you reach a point where your antenna is as
large as it will go and the house lights dim when you key.
How do you go further from there?
One avenue might be this coherent signalling where you try
to maintain an almost constant phase for the duration of a
transmission. The rx does likewise and the decoder expects the
signal to be so weak it can't rely on picking out a reference
phase, so it guesses. Then you're suddenly several dB more
capable. You'll find yourself exchanging messages that are
not just 50dB short of being audible, but can't even be seen
on a spectrogram.
Hmm... Have I nearly talked myself into setting up something