Wow this is a long thread. Don't really know where I should put my two bits
in, but here it goes.

I have designed several RF PA sections in the past. 500MHz at about 50W.
Pretty easy stuff if you have the right tools and know how to use them. The
tools I like using for matching the power output FET is two triple stub
tuners. One on the input of the FET and one on the output. So it
goes...pre-amp (50 ohm output) - stub tuner - FET - stub tuner - 50 ohm
dummy pad - spectrum analyzer. Then just tune the stubs for the performance
you desire, these include: efficiency (thermal issues), harmonic content,
spurious emissions, load VSWR considerations, cold start, ect. Then remove
the FET and look into the triple stub tuners with the network analyzer.
Model and duplicate the network out of discrete components that can handle
the voltage/power, send the design off to the enviro test lab, and head home
early for the day.

Cheers,
Thomas

"gwhite" wrote in message
...
Richard Clark wrote:


As I've noted in the past, you can fill a library with negative
assertions...

The troublesome assertion is not the negative one. It is that RF PA's are
conjugate matched. Neither you nor Ken has provided a single example of
such a
design that also extracts the maximum amount of "linear" power from a
device and
essentially its power supply (after all, that is what it is: a _power_
amp).
Your example said nothing about output-Z, which suggests you have no clue,
since
you didn't even remotely address the issue.

For Ken's part, he recently obfuscated by dismissing an example that was
primarily intended to be illustrative, but yet holding the salient points.
He
completely ignored (or didn't understand) the clipping issue. Further
obfuscation was provided by talking about "protection circuitry," which
may or
may not exist in a circuit, but adds zero to a discussion regarding how
the PA
is to be loaded. "Protection" is a non-stater because the PA is either
off or
impaired.

Ken's argument is circular. He say's that if a design is done for
conjugate
match,
then it will behave as if it is conjugately matched. Well of course (or
at
least sort of under specific test conditions and circuits)! It is
self-fullfilling prophecy but it unfortunately makes no statement
regarding
obtaining the maximum power out of the circuit in the sense of turning DC
power
into RF power (yes, *extracting* power from the DC supply and transformed
to
RF). This is paramount to PA design. To use the device to maximum
efficacy, as
Cripps puts it, a load-line match is needed. Ken's "conjugate match"
design
won't do that, and that's why PA's aren't designed that way.

The bottom line is that if I design an amp via load line techniques using
the
same device and power supply as Ken (him using conj-match), my amp will
deliver
higher unclipped PEP than his. That is the factual result you resist.
Now if
you want to pay for extra power and big devices, that's your business--go
ahead
and attempt to conj-match your amp--but engineers who design PA's don't do
that.

Another idealized and hypothetical example to elucidate the load-line
principle
is offered.

Let's say we have a 10 W FET we'll build into a class A circuit. An RF
choke is
used to supply drain current. We DC bias it to Vd = 10 V and Id = 1 A.
Just
for argument sake, let's say it has a constant internal resistance of 110
ohms
and the device will break down at 25 V. According to the most idealized
and
standard load-line theory, we should load it to rL = Vd/Id = 10 Ohms.
This
idealization includes the definition of positive and negative clipping --
whichever comes "first" -- of being the operational limit for output
voltage
swing. Clipping is associated with severe distortion.

Since we need rL to be 10 ohms, and Ri = 110 ohms, we need to make the
actual
load resistor equal to: RL = 11 Ohms. Let's check that result and see if
it
meets the clipping constraint for maximum available power.

positive swing = Id*rL = 1*10 = 10 V
negative swing = Vd = 10 V
Power delivered to RL: Pload = 10^2/(2*11) = 4.55 W
The efficiency is a little under 50% because of the internal resistance.
Note
the Load resistance is decidely not the conjugate of the internal
resistance.

Let's spot check the load to see if it at least appears to be the peak
available
power, by testing two loads "immediately" on either side of our optimum 11
ohms.

Let RL = 10 ohms
positive swing = Id*rL = 1*9.17 = 9.17 V
negative swing = Vd = 10 V
Since we positive clip at 9.17 V, we are limited by our design clipping
constraint to only driving the PA such that 2*9.17 V is the maximum
available
voltage swing.
Power delivered to RL: Pload = 9.17^2/(2*10) = 4.20 W

Let RL = 12 ohms
positive swing = Id*rL = 1*10.82 = 10.82 V
negative swing = Vd = 10 V
Since we negative clip at 10 V, we are limited by our design clipping
constraint
to driving the PA such that 2*10 V is the maximum available voltage swing.
Power delivered to RL: Pload = 10^2/(2*12) = 4.17 W

Sure enough, the power peaked at a load of 11 ohms, just like load-line
theory
says it will. Now let's see what the available power hit of conjugate
matching
is.

By definition, conj-match insists RL = Ri = 110 ohms. Again we are
limited in
our clipping constraint by static drain current, and supply voltage,
specifically 10 V.

Our negative swing limit is, as ever, 10 V (the drain voltage).

positive swing = Id*rL = 1*55 = 55 V

This would breakdown the device, but the lower negative swing will force
us to
back down the drive to meet the design defined clipping constraint.

Pload = 10^2/(2*110) = 0.455 W

Conjugate matching resulted in a 10*log(0.455/4.55) = 10 dB available
power
hit. Power amplifiers are not designed with conjugate matching in mind.
You
don't need to re-invent the wheel. Just follow well established
principles when
doing cookie cutter PA design.

The list could go on,...

LOL. Given your pattern, I am sure it will.

You sighed with content at being offered a "relevent
question/statement" Your re-iterative response contains the same (how
could it be otherwise?) slack of precision that started this. Want to
try again?

Not really. The problem isn't precision, it is you can't, or refuse, to
comprehend what is being said, which I presume is why you instead write
with the
most bizarre terms and phrasology that has nothing of import to the topic
at
hand.

You could have as easily expressed what sense they ARE matched,

For what seems like the billionth time now: they are load-line matched.

...but instead this time offer what Basis
of Matching you are attempting to
describe.

I've given a didactic example (actually a couple), you just don't--or more
likely won't--get it. If you don't like my example, you can refer to
Cripps,
who is considered one of the preeminant RF PA experts in the world.

Even more simplistic is Malvino's discussion on pp177-185 of the first
edition
((c) 1968) of "Transistor Circuit Approximations." It is basically a
technician
level description, so perhaps it is well-suited to you. In academics,
load-line
theory is presented down to tech level courses and up across to
engineering.
That some engineers and techs aren't clear on the load-line concept for
PA's (or
*any* circuit needing a wide symmetrical swing) is notwithstanding.

This is the more rigorous approach that eliminates vague
descriptions and uses standard terms. If you have to query about what
"Basis" means (used by professionals - namely metrologists who can
quantify Output Z of all sources) - then we can skip it as a topic out
of the reach of amateur discussion.

I see you still don't know what impedance is. In any case, it doesn't
mean that
looking into a properly designed PA output with a network analyzer
confirms the
conj-match precept, it doesn't.

Impedance is a *linear* conception, a portion of linear theory, and again
by
definition:

Z = V/I

V and I are sinusoids (phasors). But with power amps, substantial
non-linearity
exists (destroying the linearity assumption of impedance), thus applying a
linearly defined concept to a non-linear milieu is a misapplication. You
are
attempting, as is Ken, to stuff a square peg down a round hole. Why?

The concept is even questionable for the most linear of the power amps:
class
A. In any case, given real devices with real supplies, the conj-match
ideal is
next to worthless. While I could agree that the borderline may be fuzzy
regarding where and when to drop the impedance notion, it still stands
that the
concept is not useful in determining how to optimally load an RF PA.

At this point you own the conj-match assertion as much as Ken. Prove it!
You
can't because it is fundamentally incorrect.

Note:
Again, RF PA's should be load-line matched.
Does not qualify as a Basis.

Load-line matching is such a basic electronic concept it is unbelievable
how
oblivious you are to the concept. Read a basic book. Don't rely on me:
look it
up and do your own design!

It is suggestive of one, but because you indiscriminately
mix several Basis within your discussions, it is your
responsibility to be precise.

You just like to hear yourself talk. I've been explicit and precise. You
just
don't know anything about the elementary electronics principle of load
line
matching. I presume this is why your comments have zero substantive
responsiveness.

If you can accomplish this, then we can
proceed to review how little it all matters.

If you keep ignoring what I've written, and that which is written in
elementary
electronics texts, you can remain happily ignorant of understanding the
simple-basic-fundamental concept presented. Your choice.

Barring resolving any of these issues of precise language,...]

The guy ignorant of the definition of impedance and that s-domain theory
*is*
linear circuit theory (and more goodies) is talking about "precise
language."
Amusing.

I notice
that you rather enjoy...

No, I don't enjoy it at all. Your lack of electronic understanding is
dismal,
especially given your tone. It would have been a lot easier for me if Ken
hadn't made the erroneous
statement in the first place and made a correct one instead. That would
have
been my preferance.

..fruitless jousting with them than challenging my
support of Ken's (supposed) statement that you say is your focus:
However, responding to the bald statement, I find nothing
objectionable about it.

That's because you don't understand the difference between impedance
matching
and ac load line matching.
We will leave that as another dead-end.

I suspect you will. I already understand it -- you're the one who
doesn't.


"One of the principal differences between linear RF amplifier design and
PA
design is that, for optimum power, the output of the device is not
presented
with the impedance required for a linear conjugate match. That causes much
consternation and has been the subject of extensive controversy about the
meaning and nature of conjugate matching. It is necessary, therefore, to
swallow
that apparently unpalatable result as early as possible (Section 1.5),
before
going on to give it more extended interpretation and analysis (Chapter
2)." --
Cripps, p1


The quote is on Page 1. Swallow it now. Learn something for a change.

On Fri, 04 Mar 2005 01:13:39 GMT, "Thomas Magma"
wrote:

Wow this is a long thread. Don't really know where I should put my two bits
in, but here it goes.

I have designed several RF PA sections in the past. 500MHz at about 50W.
Pretty easy stuff if you have the right tools and know how to use them. The
tools I like using for matching the power output FET is two triple stub
tuners. One on the input of the FET and one on the output. So it
goes...pre-amp (50 ohm output) - stub tuner - FET - stub tuner - 50 ohm
dummy pad - spectrum analyzer. Then just tune the stubs for the performance
you desire, these include: efficiency (thermal issues), harmonic content,
spurious emissions, load VSWR considerations, cold start, ect. Then remove
the FET and look into the triple stub tuners with the network analyzer.
Model and duplicate the network out of discrete components that can handle
the voltage/power, send the design off to the enviro test lab, and head home
early for the day.

Cheers,
Thomas

Hi Thomas,

Thanx, your two bits were worth more than the academic plug nickel.
This is something that our original poster should hearken to as his
needs were obviously production oriented. Bench experience will trump
cut-and-paste theory in a heart-beat.

However, triple stub is pretty aggressive. How long did it take you
to flatten response?

73's
Richard Clark, KB7QHC

Response is flattened through gain controlling the pre-amp from a look-up
table held in the micro's EEPROM. The alignment procedure is automated using
the HB-IP bus from the spectrum analyzer and a computer. The
computer/analyzer also looks for harmonic content and spurious emissions
during this procedure. Think it takes about ten seconds to do this.

"Richard Clark" wrote in message
...
On Fri, 04 Mar 2005 01:13:39 GMT, "Thomas Magma"
wrote:

Wow this is a long thread. Don't really know where I should put my two
bits
in, but here it goes.

I have designed several RF PA sections in the past. 500MHz at about 50W.
Pretty easy stuff if you have the right tools and know how to use them.
The
tools I like using for matching the power output FET is two triple stub
tuners. One on the input of the FET and one on the output. So it
goes...pre-amp (50 ohm output) - stub tuner - FET - stub tuner - 50
ohm
dummy pad - spectrum analyzer. Then just tune the stubs for the
performance
you desire, these include: efficiency (thermal issues), harmonic content,
spurious emissions, load VSWR considerations, cold start, ect. Then
remove
the FET and look into the triple stub tuners with the network analyzer.
Model and duplicate the network out of discrete components that can
handle
the voltage/power, send the design off to the enviro test lab, and head
home
early for the day.

Cheers,
Thomas

Hi Thomas,

Thanx, your two bits were worth more than the academic plug nickel.
This is something that our original poster should hearken to as his
needs were obviously production oriented. Bench experience will trump
cut-and-paste theory in a heart-beat.

However, triple stub is pretty aggressive. How long did it take you
to flatten response?

73's
Richard Clark, KB7QHC

On Fri, 04 Mar 2005 16:47:05 GMT, "Thomas Magma"
wrote:

Response is flattened through gain controlling the pre-amp from a look-up
table held in the micro's EEPROM. The alignment procedure is automated using
the HB-IP bus from the spectrum analyzer and a computer. The
computer/analyzer also looks for harmonic content and spurious emissions
during this procedure. Think it takes about ten seconds to do this.


Hi Thomas,

10 seconds to adjust all 6 stubs?

73's
Richard Clark, KB7QHC

No the triple stub tuners are only for development. Production boards have
discrete components to form the match network. Power levelling or
"flattening the response" is computer adjusting the output power to
compensate for the reactive components to ensure a constant output power
over the entire band of the radio. We also put in a small temperature
compensation coefficient into the EEPROM because the PA tends to put out
more power when it is cold.

"Richard Clark" wrote in message
...
On Fri, 04 Mar 2005 16:47:05 GMT, "Thomas Magma"
wrote:

Response is flattened through gain controlling the pre-amp from a look-up
table held in the micro's EEPROM. The alignment procedure is automated
using
the HB-IP bus from the spectrum analyzer and a computer. The
computer/analyzer also looks for harmonic content and spurious emissions
during this procedure. Think it takes about ten seconds to do this.


Hi Thomas,

10 seconds to adjust all 6 stubs?

73's
Richard Clark, KB7QHC

On Fri, 04 Mar 2005 17:39:04 GMT, "Thomas Magma"
wrote:

No the triple stub tuners are only for development.

Hi Thomas,

I thought 10 seconds was awful quick. How long would it take to
flatten the response when manually adjusting the triple stub tuners?

What merit did you find with triple that could not be found with
double stub tuners?

73's
Richard Clark, KB7QHC

If your amp has to operate over a wide frequency range it is not likely that
you can flatten the response just with stubs. Stubs should be looked at as
more single frequency devices than broadband networks. But you can use the
stubs to plot out the appropriate impedance curve on the Smith Chart to
ensure a flat response when you model in the discretes.

I usually just try to get the flatness of the response as close as possible
and rely on a software calibration routine to flatten it off. Saves a lot of
time.

It's my understanding that a triple stub tuner of the right length can reach
anywhere on the Smith Chart where as a double stub can not.

"Richard Clark" wrote in message
...
On Fri, 04 Mar 2005 17:39:04 GMT, "Thomas Magma"
wrote:

No the triple stub tuners are only for development.

Hi Thomas,

I thought 10 seconds was awful quick. How long would it take to
flatten the response when manually adjusting the triple stub tuners?

What merit did you find with triple that could not be found with
double stub tuners?

73's
Richard Clark, KB7QHC

Thomas Magma wrote:
No the triple stub tuners are only for development. Production boards have
discrete components to form the match network. Power levelling or
"flattening the response" is computer adjusting the output power to
compensate for the reactive components to ensure a constant output power
over the entire band of the radio. We also put in a small temperature
compensation coefficient into the EEPROM because the PA tends to put out
more power when it is cold.

Richard was asking how long it took you to tune the triple stub filters
during devolpment.

I am curious about the exact nature of the impedance transmformation
these devices provided.

jk

"Richard Clark" wrote in message
...

On Fri, 04 Mar 2005 16:47:05 GMT, "Thomas Magma"
wrote:


Response is flattened through gain controlling the pre-amp from a look-up
table held in the micro's EEPROM. The alignment procedure is automated

using

the HB-IP bus from the spectrum analyzer and a computer. The
computer/analyzer also looks for harmonic content and spurious emissions
during this procedure. Think it takes about ten seconds to do this.


Hi Thomas,

10 seconds to adjust all 6 stubs?

73's
Richard Clark, KB7QHC

Richard Clark wrote:

On Fri, 04 Mar 2005 01:13:39 GMT, "Thomas Magma"
wrote:

Wow this is a long thread. Don't really know where I should put my two bits
in, but here it goes.

I have designed several RF PA sections in the past. 500MHz at about 50W.
Pretty easy stuff if you have the right tools and know how to use them. The
tools I like using for matching the power output FET is two triple stub
tuners. One on the input of the FET and one on the output. So it
goes...pre-amp (50 ohm output) - stub tuner - FET - stub tuner - 50 ohm
dummy pad - spectrum analyzer. Then just tune the stubs for the performance
you desire, these include: efficiency (thermal issues), harmonic content,
spurious emissions, load VSWR considerations, cold start, ect. Then remove
the FET and look into the triple stub tuners with the network analyzer.
Model and duplicate the network out of discrete components that can handle
the voltage/power, send the design off to the enviro test lab, and head home
early for the day.

Cheers,
Thomas

Hi Thomas,

Thanx, your two bits were worth more than the academic plug nickel.
This is something that our original poster should hearken to as his
needs were obviously production oriented.

I doubt you understand what he wrote. I can't fathom why you would be concerned
with the OP when your own difficulties are so acute.

Bench experience will trump
cut-and-paste theory in a heart-beat.

How would you know?

However, triple stub is pretty aggressive. How long did it take you
to flatten response?

How long will it take you to figure out that he wrote not a wisp of a word on
what the "output-Z" of the amplifier is? He did write that he determines how
the amp was loaded to acheive power, something I've been saying is the prime
concern.

Hi gwhite,

I would have to agree with you on most everything you have said through this
thread. I once saw my boss (with his "PHD") try to model and match a power
amp based on the small signal parameters off the datasheet. He insisted that
the stated input and output impedances were characteristic parasitics of
that amp and wouldn't change between a small or large signal. It was kind of
pathetic to watch him struggle for over a month on the matching network, and
I think he had resorted to guessing in the end.

I've often questioned why manufactures put small signal parameters on their
datasheets? Makes no sense to me. Even if they do publish some large signal
parameters it is unlikely to be the exact same mode of operation that you
need for your project.

Playing with triple stub tuners on PA's has shown me that there are many
combinations of input and output impedances that appear to give similar
results at any one frequency, but give different results at others
frequencies. So it's a matter of finding the input and output impedance that
give you adequate performance over the entire scope of your project.

Thomas

"gwhite" wrote in message
...
Richard Clark wrote:

On Fri, 04 Mar 2005 01:13:39 GMT, "Thomas Magma"
wrote:

Wow this is a long thread. Don't really know where I should put my two
bits
in, but here it goes.

I have designed several RF PA sections in the past. 500MHz at about
50W.
Pretty easy stuff if you have the right tools and know how to use them.
The
tools I like using for matching the power output FET is two triple stub
tuners. One on the input of the FET and one on the output. So it
goes...pre-amp (50 ohm output) - stub tuner - FET - stub tuner - 50
ohm
dummy pad - spectrum analyzer. Then just tune the stubs for the
performance
you desire, these include: efficiency (thermal issues), harmonic
content,
spurious emissions, load VSWR considerations, cold start, ect. Then
remove
the FET and look into the triple stub tuners with the network analyzer.
Model and duplicate the network out of discrete components that can
handle
the voltage/power, send the design off to the enviro test lab, and head
home
early for the day.

Cheers,
Thomas

Hi Thomas,

Thanx, your two bits were worth more than the academic plug nickel.
This is something that our original poster should hearken to as his
needs were obviously production oriented.

I doubt you understand what he wrote. I can't fathom why you would be
concerned
with the OP when your own difficulties are so acute.

Bench experience will trump
cut-and-paste theory in a heart-beat.

How would you know?

However, triple stub is pretty aggressive. How long did it take you
to flatten response?

How long will it take you to figure out that he wrote not a wisp of a word
on
what the "output-Z" of the amplifier is? He did write that he determines
how
the amp was loaded to acheive power, something I've been saying is the
prime
concern.