In article , Joe McElvenney
writes:

You are right about AN-393, and the non-linearities, of course. However
when I first posted I needed a ballpark figure and the app-note gave it me.
My
application is a rather mundane, fixed frequency, 10MHz frequency standard
required to drive a counter/signal generator combo and whose output should be
reasonably sinusoidal so as not to spray harmonics around the place. The
buffered or inverted output coming off the VCXO has to supply about 1V pk-pk
into 50 ohms. A 74HC04 or '07 followed by a single-section, image-parameter,
pi -section filter appears to do the job but it was just my day for being
scientific :-(

No problem here. :-)

As a first approximation for the application, it should be intuitive that
the peak-to-peak voltage out of the gate/inverter into a resistive load
would yield the beginning of data on RF power output capability. The
rest can be figured out on the bench.

More eleborate data sheets for particular devices have specs for
driving different loads, usually as the logic 1 level voltage minimum
for a given resistive load. Depends on the manufacture's budget for
data sheet preparation, I would guess, since those vary in content.

If others are trying such an application at maximum power handling
capability and need lowpass filtering, a caution. Some elliptic filter
input impedances can vary widely in stopband frequencies. That
might, but not necessarily, result in edge spikes above breakdown
at the gate/inverter output. Butterworth response filters are the
most forgiving of such but also have the lowest stopband attenuation.
The 74HC family ought to be rather robust on that and I don't think
it is a great concern.

I did encounter some problems with 74S devices driving filters some
years ago when using a gate as a digital mixer for an optical
interferometer phase measurement instrument. That was with old
74S family devices before the newer incorporated protection
devices were introduced a decade or so ago.

"Digital mixers" work very well indeed when both RF inputs are at
similar high levels common to digital circuits. It's quite easy to get
difference frequency output and to lowpass filter it to remove the
input frequency components. Never tried it for sum-output mixing,
probably need a spectrum analyzer handy for such circuits.

Len Anderson
retired (from regular hours) electronic engineer person