It's all about the Software

Er, Simon... "Shout here when you have any DSP-related questions." I've just discovered that KiwiSDR has implemented something John Seamons calls "SAS".
My dream AM-DX demodulator, it uses the phasing method for sideband selection to reconstruct LSB+USB and route to L+R audio channels (not I-Q only "pseudo-stereo" ).
The only DSP needed (that even I can sort-of understand) is a Hilbert transform to get a relative 90 degree "compensating" phase shift between I and Q.
Question: is that something you might think about one day?
And OK yes, I'm thinking I should go and buy a KiwiSDR.

I've also added this, it works but it's not available yet as it uses some GPL code. I want to work something out from first principles, then add this.

Well shiver me timbers! Are you interested in a refinement?

pauldbnut wrote: ↑Mon Nov 30, 2020 2:04 pm Well shiver me timbers! Are you interested in a refinement?

Well,

Yes but I have to define / construct the algorithm from first principles and recode the filtering.

Sure, no worries.
This thread really says it all, at your leisure: http://forum.kiwisdr.com/discussion/comment/11092.
All the best, and thanks again.

Well congratulations Simon, and thank you very very much! Tested your binaural AM on an old translatlantic MW recording, and it blows my experiments out of the water. No surprise there . As far as I can tell your sideband separation is much better than mine, giving better imaging when tuned off-centre (e.g. 1 kHz to the side).

Also unsurprisingly, results are better with IF bandwidth greater than about 2 kHz. This means asymmetric passbands are not the ideal way to get rid of heterodynes from adjacent channel carriers (1 or 2 kHz away). Have you thought about a simple audio notch for those two frequencies? My guess is an auto-detect 1 or 2 kHz notch might work OK?

Otherwise my only suggestion would be an additional unlocked mode to try and separate (mentally focus on) co-channel stations with small offsets (about 1 to 50 Hz).