Tim Shoppa wrote:

In 1930's QST's it's not too uncommon to see neophytes warned that
crystals will often oscillate on something other than their marked
frequency. They didn't call this overtone operation, though.

They /might/ be suggesting that frequency can change with loading. I find
that some of the reference crystals I use are quite some way off their marked
frequency when given capacitive loading that differs from that recommended by
the manufacturers!

BC-604's (WWII era) start with a ridiculously low crystal (400ish kHz)
frequency and multiply up but I think the reason for this is more to
do with FM deviation than anything else. ("Armstrong method"?) For
many decades, broadcast FM stations similarly started with low crystal
frequencies and multiplied up.

Some of the broadcast transmitters I worked on 25 years ago used this method
for FM, and were /really/ difficult to line up! They also included complex
circuitry for the required "pre-distortion" of the audio to compensate for
the non-linear deviation you got out of a crystal oscillators. Some
manufacturers tried to overcome the distortion issue by using phase
modulation and the "right" audio curves, but these required even more stages
of multiplication!

One of my earliest jobs as a broadcast transmitter engineer was to develop a
PLL to replace the horrible multiplier chains in some of these transmitters.
I used (normally) either half or quarter frequency generation, and used the
last one or two multiplier stages. The CMOS PLL circuitry could be prone to
bizarre effects with high field strengths, so they were built in sealed
diecast boxes, and the lower frequency generation meant that the high power
output stages were unlikely to couple back into the oscillator!

Bob