(Richard Harrison) wrote in news:20731-
:

Owen Duffy wrote:
"Richard stated "It produces no significant harmonics, so the system is
linear." It is that with which I disagree."

A clear statement. Congratulations. Too bad it is wrong.

Terman wrote:
"Amplitude distortion exists when the modulation envelope contains
frequency components not present in the modulating signal."

Fine.

It is also true that absence of of harmonics is proof of linearity,

That is not a logical implication of your quote from Terman, it is
entirely your statement, and without splitting hairs over the absolute
meaning of 'absence', it is wrong when applied to a Class C amplifier
with pure sine wave excitation and a resonant load.

The converse is not logically equivalent to the original.

...
I challenge you to prove a mistake in Terman`s writings.

I have no problems with the statement you attribute to Terman and haven't
said anything contrary to that during the discussion.

Richard, I accept that you are committed to your view, lets leave it at
that. I don't think your statements on the matter support Walt's
proposition, rather since they are in my view flawed, I think they weaken
the body of evidence.

Owen

"Owen Duffy" wrote in message
...
(Richard Harrison) wrote in news:20731-
:

Owen Duffy wrote:
"Richard stated "It produces no significant harmonics, so the system is
linear." It is that with which I disagree."

A clear statement. Congratulations. Too bad it is wrong.

Terman wrote:
"Amplitude distortion exists when the modulation envelope contains
frequency components not present in the modulating signal."

Fine.

It is also true that absence of of harmonics is proof of linearity,

That is not a logical implication of your quote from Terman, it is
entirely your statement, and without splitting hairs over the absolute
meaning of 'absence', it is wrong when applied to a Class C amplifier
with pure sine wave excitation and a resonant load.

The converse is not logically equivalent to the original.

...
I challenge you to prove a mistake in Terman`s writings.

I have no problems with the statement you attribute to Terman and haven't
said anything contrary to that during the discussion.

Richard, I accept that you are committed to your view, lets leave it at
that. I don't think your statements on the matter support Walt's
proposition, rather since they are in my view flawed, I think they weaken
the body of evidence.

Owen

Owen, as I view your last paragraph above it seems apparent that you do not
believe Richard's and my position that the output of a Class C amplifier can be
linear. We're talking about at the 'output', not the 'thruput'. How can you
refute the evidence of a nearly pure sine wave at the output terminals of the
pi-network?

Walt, W2DU

"Owen Duffy" wrote in message
...
(Richard Harrison) wrote in news:20731-
:

Owen Duffy wrote:
"Richard stated "It produces no significant harmonics, so the system is
linear." It is that with which I disagree."

A clear statement. Congratulations. Too bad it is wrong.

Terman wrote:
"Amplitude distortion exists when the modulation envelope contains
frequency components not present in the modulating signal."

Fine.

It is also true that absence of of harmonics is proof of linearity,

That is not a logical implication of your quote from Terman, it is
entirely your statement, and without splitting hairs over the absolute
meaning of 'absence', it is wrong when applied to a Class C amplifier
with pure sine wave excitation and a resonant load.

The converse is not logically equivalent to the original.

...
I challenge you to prove a mistake in Terman`s writings.

I have no problems with the statement you attribute to Terman and haven't
said anything contrary to that during the discussion.

Richard, I accept that you are committed to your view, lets leave it at
that. I don't think your statements on the matter support Walt's
proposition, rather since they are in my view flawed, I think they weaken
the body of evidence.

Owen

Owen, as I view your last paragraph above it seems apparent that you do not
believe Richard's and my position that the output of a Class C amplifier can be
linear. We're talking about at the 'output', not the 'thruput'. How can you
refute the evidence of a nearly pure sine wave at the output terminals of the
pi-network?

Walt, W2DU

"Walter Maxwell" wrote in
:
....
Owen, as I view your last paragraph above it seems apparent that you
do not believe Richard's and my position that the output of a Class C
amplifier can be linear. We're talking about at the 'output', not the
'thruput'. How can you refute the evidence of a nearly pure sine wave
at the output terminals of the pi-network?

Walt, you have posted this twice.

There are subtle word shifts here, you are saying "a Class C
amplifier can be linear" rather than is (always) linear.

It is true that a Class C amplifier with resonant load and a constant
amplitude sine wave drive may appear linear when comparing Vout to Vin.

But, as I explained earlier, if you vary the drive amplitude, it is
clearly not linear... in typical cases output will cease below about 25%
of the drive level required for maximum output.

Further, if you drive it with a complex waveform, it is clearly non
linear at any drive level.

Richard's solution to detecting RF PA distortion by monitoring harmonics
is an interesting one, because it suffers the disadvantage of output
filtering masking the harmonics (unless the monitor point was prior to
filtering).

The most widely accepted test for linearity (Vout/Vin) of an RF PA is the
'two tone test', where the drive is a complex waveform (the sum of two
equal amplitude sine waves quite close in frequency) and at least some of
the distortion products due to third order and fifth order etc transfer
terms appears in-band in the output after all output filtering, and where
they can be reliably compared in amplitude to the desired signals. A
Class C RF PA will not appear to be linear under such a test at any drive
level.

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.

Owen

"Owen Duffy" wrote in message
...
"Walter Maxwell" wrote in
:
...
Owen, as I view your last paragraph above it seems apparent that you
do not believe Richard's and my position that the output of a Class C
amplifier can be linear. We're talking about at the 'output', not the
'thruput'. How can you refute the evidence of a nearly pure sine wave
at the output terminals of the pi-network?

Walt, you have posted this twice.

There are subtle word shifts here, you are saying "a Class C
amplifier can be linear" rather than is (always) linear.

It is true that a Class C amplifier with resonant load and a constant
amplitude sine wave drive may appear linear when comparing Vout to Vin.

Owen, with a Class C amplifier biased beyond cutoff the grid is never going to
see a constant amplitude sine wave, even if the constant amplitude sine wave
were impressed on the grid. How then can the transfer linearity ever occur under
these conditions? I maintain that it cannot.

But, as I explained earlier, if you vary the drive amplitude, it is
clearly not linear... in typical cases output will cease below about 25%
of the drive level required for maximum output.

Further, if you drive it with a complex waveform, it is clearly non
linear at any drive level.

Richard's solution to detecting RF PA distortion by monitoring harmonics
is an interesting one, because it suffers the disadvantage of output
filtering masking the harmonics (unless the monitor point was prior to
filtering).

The most widely accepted test for linearity (Vout/Vin) of an RF PA is the
'two tone test', where the drive is a complex waveform (the sum of two
equal amplitude sine waves quite close in frequency) and at least some of
the distortion products due to third order and fifth order etc transfer
terms appears in-band in the output after all output filtering, and where
they can be reliably compared in amplitude to the desired signals. A
Class C RF PA will not appear to be linear under such a test at any drive
level.

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.

Owen

Owen, you are either twisting my words, or you're not listening. I've made it
very clear that I'm NOT talking about 'transfer linearity', and never have. My
position is only that the OUTPUT of the pi-network is linear. The linearity at
the output is irrelevant to the waveform at the input of the tank circuit in
Class C amplifiers. I don't even understand why the discussion concerning
'transfer linearity' with respect to Class C amplifiers should have come up.

Walt, W2DU

PS--I didn't send two identical emails--something must have happened at the
server to have caused it.

Owen Duffy wrote:


The most widely accepted test for linearity (Vout/Vin) of an RF PA is the
'two tone test', where the drive is a complex waveform (the sum of two
equal amplitude sine waves quite close in frequency) and at least some of
the distortion products due to third order and fifth order etc transfer
terms appears in-band in the output after all output filtering, and where
they can be reliably compared in amplitude to the desired signals. A
Class C RF PA will not appear to be linear under such a test at any drive
level.

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.



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.


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.




Owen

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

"Owen Duffy" wrote in message
...
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

Owen, on whether my observations apply in general, if you re-read the
summarizing paragraph on my Chapter 19A you'll see that I've made measurements
of the source impedance of two different xmtrs with several different complex
impedance loads. All measurements showed the source impedance equal to the load
impedance when all available power is delivered to the load.

As to the explanation, Richard H said it well. When all available power is
delivered, according to the maximum power transfer theorem the source impedance
equals the load impedance. My measurements have proved this to be true in
determining the source impedance of the xmtrs I measured.

Walt, W2DU