In vlfrx-tools I use

 I = s(t) * cos( 2pi * Fo * t);

 Q = s(t) * sin( 2pi * Fo * t);

where s(t) is the incoming signal and Fo is the
local oscillator frequency.   Is that right, or
should there be a minus sign in there somewhere?

If s(t) is VO1NA at 137370.00000 and I set Fo to
137369.9, then when I put I and Q into EbNaut, it
needs an offset of +0.1 Hz to decode as you would
expect with the LO being 0.1Hz on the low side.

For more general mixing and frequency shift,
to select the upper sideband of the mixer output
I use

   I - j * Q

is that the right way round?   When I take I/Q data
from an RTL2832 I find I have to invert Q to get the
correct sideband, ie I + j * Q.   That's the only
independent source of I/Q signal that I have available
to try.

--
Paul Nicholson
--