Okay, I used a big signal from a clean generator, but the results
were perfect. We knew it would work, because I blocked out the real
world.

If Jon allows himself an IC, I suggest using an FM radio chip to
get a limiter, then the audio path can be all tubes. If he puts in
a decent pre-selector, and perhaps a "strong signal" trap, it can
be made to work.

My lash-up is limited by the fact that I need a big signal to get
the square wave from the CMOS.


My new (1980+) radio books say that this will give superior
performance to an envelope detector, but is too complicated for
"consumer" receivers and therefore, never caught on.

I wonder if you can get enough limiting on local signals to make
this work with all tubes.



Maybe build a local oscillator that runs at the carrier frequency, and
rig it so it
"locks" on the received carrier thru a buffer stage. Color TV set
chroma detector
circuits do something like this to lock a local oscillator to feed a
pair of synchronous
detectors to get the U (in phase) and V (quad) chroma signals.

There's also "phsydo-synchronous" detectors. Those are usually an IC with a
high gain "limiter" to amplify the hell out of the received carrier and
then use
that to run a synchronous detector. RCA used such in TV sets about 20
years ago.
TV signals use a form of AM for the video, and FM for the sound.