Hello,

Why not try it another way: in software upconvert the audio to 20kHz (or 40kHz if you have a 96kHz sound card), mix up to 136kHz or 472kHz and filter out the remaining sideband which is 40kHz away (or 80kHz). You can play with an image suppressing mixer here if you want, but an LC filter should be sufficient.

And if you want to use this also for receiving, then you can do it in the opposite direction: downconvert to 20kHz, then downconvert to audio in software.

The upconversion/downconversion to 20kHz IF can be easily done in software.

Some programs can be persuaded to just tx at this frequency (so no up/down conversion). Many programs are opensource, so this would be a matter of a simple hack to change the tx frequency.

Others (the ones which can't be easily hacked to tx at 20kHz) can be up/down converted via external software. Under linux i would use gnuradio and pulseaudio (to route the audio from the program to gnuradio input). Under windows the same could probably be done with Spectrum Lab and virtual audio cable.

This leaves us with an NE612, some LC filtering and a crystal (the upconverter can be made tunable, and the bands are narrow, so we can use a fixed frequency source). Seems easier.

VY 73

Jacek / SQ5BPF