Jim Lux wrote in
:

Owen Duffy wrote:

....
Actually, in modern systems with very complex signals, there are more
meaningful tests like noise power ratio with a notch that look for
spectral regrowth. The two tone test has the advantage of being
moderately easy to perform for middling performance
amplifiers/devices.
But if you're looking for very high performance, such things as
generating the two tones without one generator interfering with the
other get to be challenging.

Noted.


I suspect that the issue of transfer linearity is a red herring to
your proposition about the Thevenin equivalent of an RF PA, but if
you do depend on arguing that the transfer characteristic of a Class
C RF PA is linear, I think you are on shaky ground.

I don't know that the concept of a Thevenin equivalent (a linear
circuit theory concept) really has applicability to "box level"
models, except over a very restricted range, where one can wave one's
hands and ignore the nonlinearities as irrelevant to the question at
issue. Sure, over a restricted dynamic range and bandwidth and
restricted class of input signals, a Class C (or class E or Class F or
E/F1, or a fancy EER system) can be adequately modeled as a linear
ideal amplifier.

I agree with you. I am not implying that you cannot design a PA with
controlled equivalent source impedance, but you don't do they way most
ham PAs are designed.

As I understand it, Walt's proposition is that the Thevinin equivalent
source impedance (at the device terminals) of the PA is equal to the
conjugate of Zl (at the device terminals) as a consequence of adjustment
of the PA for maximum power output, a twist on the Jacobi MPT theorem.
For that model to be generally useful in explaining behaviour of the PA
in the presense of 'reflections', it would need to be true for a wide
range of load impedances.



The real question is what is the value of that model. If the model
provides conceptual understanding of some underlying problem, great.
For instance, it might help with a link budget. If the model helps
design a better amplifier, great. The model might allow prediction of
behavior; so that you can, for instance, detect a fault by the
difference between model and actual observation, as Richard mentioned
with the harmonic energy detector.

I think it goes to whether Walt's proposition and observations apply in
general, and then a valid explanation for what happens.

Owen