Roy Lewallen wrote in message ...
... I've found that good paper designs,
often aided by fundamental knowledge gleaned from books, lead to good
hardware results, rather than being an opposing and somehow inferior
method. And they have the advantage of being well understood,
predictable, and repeatable.

Indeed. Occasionaly new not-yet-understood phenomena are discovered
on the bench, but the art benefits greatly from a detailed
understanding of the underlying mechanisms. Coincidently, I was
browsing "Inventions of Opportunity" this afternoon and stumbled
across an article about how in the late 1950's a newly-developed high
speed sampling scope aided in understanding harmonic-generation
mechanisms in diodes, which apparently helped a lot in the development
of step recovery diodes. Before that, apparently there wasn't good
understanding about why some diodes generated lots of harmonics and
others didn't. Step recovery diodes are optimized for fast turn-off
of the reverse recovery, and are used in generating a "comb" of
harmonics. It's not uncommon to pick off the desired harmonic with an
appropriate filter, up to beyond the tenth harmonic. Seems like step
recovery diodes are not in as great favor as they once were, since
there are generally better ways to generate higher order harmonics.

With a little understanding of the spectrum of a non-symmetrical
square (or trapezoid) wave, it's not hard to come very close to an
optimum bias and drive for a given harmonic output in an amplifier
stage. If you do it just by experimentation, you're liable to find a
local optimum that's quite a bit worse than the global optimum. Same
with the output coupling/filtering network.

Cheers,
Tom