In article ,
Jerry wrote:

Thanks Jerry. No you aren't missing anything other than the fact that
my familiarity with Smith Chart analysis is limited to working through
a few exercises in the last chapter of the ARRL Antenna book. I will
be looking for more Smith Chart tutorial info on the web and am
certain I can get myself up to speed enough to start working with
conductance, suseptance, admittance, etc.

I strongly encourage you to work it through from the basic math:

- Impedances and admittances are complex numbers, with both real
(pure-resistive) and imaginary (reactive) components.

- When you put two things in series, add their impedances

- When you put two things in parallel, add their admittances

- Admittance = (1+j0) / impedance, and vice versa.

The math is quite easy to implement in FORTRAN, or any other
programming language which has native complex-number capability. C
doesn't, alas, and you'd have to use a complex-number library or write
your own. I'm sure that there are standard C++ classes and methods
for handling complex numbers.

Although I haven't yet used the capability, it looks to me as if
modern spreadsheet systems (e.g. Excel, OpenOffice Calc) can deal with
complex numbers. You can't use the normal arithmetic operators (at
least, not in OpenOffice Calc) but myst use special complex-number
functions such as IMPRODUCT and IMDIV and IMSUM.

Not too hard to do, though.

I'd suggest starting out with the simplest calculation (e.g. a
inductor or capacitor, in series with a pure resistance). That one's
really easy, you just calculate the impedance of the reactive
component at the frequency of interest, and do a complex addition.

Then, do an inductor or capacitor, shunted in parallel with a pure
resistance... calculate the reactive impedance, invert the impedances
to get admittances, sum them, invert again to get the final impedance.

Once you can do those two basic steps, you can simply repeat them as
necessary to determine the effect of L and T matching networks, tanks,
and so forth.

The Smith chart makes it easy to do this via graphical means - it's
faster than using a slide rule or table of logarithms - but it's very
instructive to do the math youself (with the aid of a spreadsheet).

With the Smith chart, it's usual to normalize all of the impedances to
your reference value (e.g. divide by 50 ohms), do the calculations,
and then denormalize (e.g. multiply by 50 ohms).

Doing so isn't necessary if you do the complex-number calculations
yourself... you *can* normalize/denormalize if you wish, but you'll
get the same results if you work directly with the raw impedances and
admittances.

--
Dave Platt AE6EO
Friends of Jade Warrior home page: http://www.radagast.org/jade-warrior
I do _not_ wish to receive unsolicited commercial email, and I will
boycott any company which has the gall to send me such ads!

On Jun 5, 7:51*am, Roger D Johnson wrote:

Yes! It's called RevLoad and is available free from Tonne Software.

http://tonnesoftware.com/

73, Roger


Many thanks Roger. Exactly what I was looking for.

73

Dykes AD5VS

On Fri, 5 Jun 2009 07:18:17 -0700 (PDT), wrote:

Now, this is the most curious statement of them all. *Every LPF that
is mounted in any Ham grade HF rig is designed with both a 50 Ohm
input Z and a 50 Ohm output Z. *This is easily verified through the
same page that does the calculations, or through trivial math for the
individual components' Z.

Uh huh... and all manufacturers use high precision components, and the
impedance at one end of the filter isn't affected by the impedance at
the other end?

Hi Jim,

I haven't the slightest idea why your objection demanding "high
precision components" is necessary. Do you have anything that is
quantifiable to sustain this concern? Give us a Monte Carlo result of
those quantified precisions and their impact on Z.

My original point is that, barring measurement, you don't KNOW.
(which is sort of your argument too, eh?)

Given decades of lock-step design that conforms to accepted practices,
why would anyone have to measure something to KNOW what it is? These
are exceedingly basic considerations that do not demand differential
calculus to get to the first pass approximation. A simple algebraic
approximation is going to be quite good. Measurement will confirm.

You recite Motorola without context, I will fill that gap:
"The network theory for power amplifier design is
well known but is useless unless the designer
has valid input and output impedance data for
the transistor."

Motorola has for years recited at least three different means to
obtain large signal transistor output Z, and has characterized
individual transistors over frequency in charts. I have enumerated
these in the past to no obvious response from other objectors that
Motorola knows what they are doing (even when objectors recited the
exact same references!), but I will proceed once again:
1. Load pull (which satisfies NOT using small signal parameters);
2. Transistor saturation, current, and supply voltage;
3. transition and diffusion capacitances at the collector junction.

This last method responds to your nervous wondering about low
frequency response (UP TO the highest frequency, Z does not undergo
dramatic changes).

I haven't seen a 13.6V, 100W amplifier deck in the last 30 years that
varies one iota from any other - and for good reason: this stuff has
been cookie-cutter design for decades using transistors with known and
optimized characteristics that haven't varied either. Yes, FETs have
emerged in those years, and so have high voltage PNPs and NPNs. Their
contributions have only shifted the design points, not the topology,
nor the matching principles. With absolutely every new component's
introduction, there have been more than sufficient discussion of
matching principles that describe their output Z.

So, looking at things with which I have practical experience and
measurements.. MMIC amps tend to be be pretty flat over octave
bandwidths, but I don't think they're representative of ham rigs with
either FET or Bipolar output stages (which have to cover multiple
octaves, in any case). *

Why not? *

Because the MMICs are a totally different design model. To start
with, they're also Class A, while most ham rigs run Class AB. They
also tend to be "detuned" for broadbanding, at the expense of
efficiency. (not all MMICs are this way.. I'm talking about the MAR-n
series, for instance)

That is not a reason, nor is it different to any great degree. The
same issues dominate both topologies. Further, efficiency is not
always a goal when it contradicts the need for gain; and both may not
satisfy match (as there is more than one form of match: conjugate
match and Z match). I don't suppose you have anything quantitative
here either, do you?

And for those same 30+ years of HF solid state rigs, their power
transistors have had (and still do) output "native" Z of several Ohms.

Would that the active device has a Z that is constant, but it's not.
Sure, the MRF454 data sheet says the output Z is 1+.2j ohms (or
something like that) at 30MHz, but is it still that at 1MHz?

This is another strange objection. You flatly state it is not, but
you are wholly uninformed about it specifically. Your nervousness is
odd with its faltering expectation for performance at what the
industry generally considers to be a very undemanding frequency. Your
question's implied suggestion is: "is the Z for 30MHz the same as the
Z for 29.9MHz?", which from your limited perspective (you are really
reading a rather thin spec sheet) won't answer either. Study
something more encompassing; Motorola has vastly more informed
resources than your meager recitation. A simple browse of half a
dozen transistors from the same production series provides quite
specific details.

My Drake TR-7 uses MRF421s and those transistors are rated at 100W PEP
(my rig is rated at 140-150W CW) where their output resistance varies
from 1.8 Ohm to 1 Ohm over more than 4 octaves. The spec sheets are
more detailed and do not depart from others of that class. As such, I
do not expect your MRF454 would either, but your qualitative
expectations are hard to account for and no parts engineer would cut a
purchase req from them.

Looking at a more modern power FET for amplifier use, the IXZ210N50L..
There's a whole page of S parameters, and S22 goes from 0.88@-51deg at
2MHz to at 14.32 MHz to at 30 MHz... that's
at Ids =200mA.. bump Ids to 500mA, and the magnitudes stay about the
same, but the phases change, by tens of degrees.

Again, Motorola specifically rejects small signal parameterization for
power applications. This is, perhaps, your problem with
characterizing amplifier issues.

Having actually worked on an amplifier design with similar parts, I
can also say that the datasheet is only a "get you in the ballpark for
the design" tool. The "real parts" (especially when packaged on a
board and attached to the heat sink) are substantially different.

I have selected, inspected, and validated transistors to Mil Spec and
found very few wandered from commercial specification. You must
inhabit a very different realm where production lots contain product
that are "substantially different." Do you have some quantification
for "substantially?" Or is this another example of a technician's
shrug?

No simple transformer is going to make that look like a constant 50
ohms.

Ah, are we now down to parsing this to "exactly 50 Ohms" where in your
objections you offer few quantifications? Does 49 Ohms invalidate the
premise and score a home run for the opposing team?

I'd love to see some real data for ham rigs.

Mine (Drake TR-7 and Kenwood TS-430s) exhibit values that vary around
50 Ohms with a low of 35 Ohms and a high of 70 Ohms in the margins.
Those rigs also suffer in those margins. *

so the VSWR looking back from the tuner into your transmitter is
1.4:1? A return loss of around 15dB... what's that work out to... an
error of about 10-15% in the "measuring impedance with a tuner"
technique... not bad, but not great, either, especially stacked up
with the other uncertainties..

Not great? You have already suggested it was unknowable, and others
state it was immaterial. It gives me pause to have given a concrete
result to now find what was unknown is now "uncertain" and what was
immaterial now counts for little at "not great."

Measurements were done by
pull, by substitution, by looking into the antenna connector with an
RF Bridge and all confirmed by simple reverse design principles.
Variations between any method rarely departed from one another, and
never from the values above. *

Although you have to admit that a 2:1 impedance variation isn't a
particularly outstanding "constant impedance load"

This characteristic that is "not particularly outstanding" was
formerly deemed impossible to determine and immaterial by others. I
would say that the gulf that I bridged goes a good deal further than
demurrals such as yours. As for the sudden emergence of the
qualification for "outstanding," that can be obtained if you care to
indulge in quantifying what "outstanding" means to you. Am I being
set up with an impossible goal to satisfy your point? This
performance characteristic that I have published has served an active
market for generations; and that in its own right responds to issues
of "outstanding." The market would respond: "sufficiently so."

As to could it be better? Without a doubt, and again, design would
proceed from knowing the transistor source Z, not "by guess and by
golly." This again goes to conventional design considerations that
were enumerated by H.W. Bode in the late 30s.

just because *I* don't want to spend the time measuring
it doesn't mean that the information is of no value to the community.
I would venture that of all the data that hams, collectively, could
measure, this is actually not as useful as some other data.. It just
doesn't have that much impact on day to day operation.

What a tedious reply in the face of correspondence in other topics
that struggle to pass as intelligent discussion. I wonder why you put
such effort to responding here if this has so little impact, no
relative importance, no relation to design, nor application to "using
a tuner to determine Line Input Impedance" (a fully informed topic of
great impact on day to day operation, isn't it?).

Very few hams
adjust their tuner by calculating L and C based on measured data, or
else there wouldn't be a plethora of articles and posts about
"tuning", "pruning", "trimming" and the techniques for doing this, and
arguments about whether a Brand X meter is better than a Brand Y
meter, etc.

Your comments to the original poster, then, could have been reduced to
one response of one line telling him to abandon his quest by this same
logic.

Hams, by and large, adjust their tuners by minimizing the reflected
power, and don't much care what the actual component values are. (e.g.
what ham tuner actually has accurate dial calibrations in pF or uH? )

Do I need to quote that correspondent's post? And if I did, what
objection would fall from that? another step descending into
intellectual entropy.

Professionals, on the other hand, do CARE,

and Hams do NOT. By this standard, our group would dissolve to the
merits of photons bouncing waves off each other.

Every problem is reduced to those two options?

Obviously not, but I'll bet that it covers over 90% of hamdom (and a
lower percentage of the folks reading this thread).

90% of hamdom - leaving some several many thousands who are interested
is sufficient enough. Not for you? Not enough readers here?
Diminishing the level of discourse is not going to build audience to
that several many thousand potential, will it?

On the other hand, I have worked with high power Transistor circuits
that have acted exactly as resistors, inductors, and capacitors and
output Z was exactly like an antenna at a given frequency (or rather
input Z, as one design was an active 100W load).


Yes.. but were those run-of-the-mill amateur transceivers? (the
original question)..

Yes.. but indeed - how tedious. I have responded to the specific
question; I have responded to the specific component; I am responding
to correlating designs. Every response is rooted in published and
quantified characteristics.

I have no doubt that it is possible to build
amplifiers with constant Z (to any degree of constancy desired.. heck,
a 1000W amp and a 60db pad gives you a 1mW amplifier with very good
output Z, regardless of what the amp does).

Why does your imposed solution have to come with a crippling hit to
efficiency or match? The only inhibition against designing an
amplifier to meet your unquantified expectations is the cost involved,
not loss of efficiency, not loss of match. You said you have
experience in amplifier design, this cannot be such an impossibility
can it?

But, does a "designed for mass production and cost target" transmitter
fall into that category?
It's not a published spec

You mean you haven't read the spec. I have seen this objection too.
When I've offered just such specs, objectors have then recursed back
into how output Z is unknowable and immaterial as if the topic had
never been encountered before.

ARRL doesn't measure it when they review rigs

Now THERE's an authority! Do they measure efficiency?

73's
Richard Clark, KB7QHC

On Jun 5, 1:19*pm, Richard Clark wrote:
On Fri, 5 Jun 2009 07:18:17 -0700 (PDT), wrote:

Uh huh... and all manufacturers use high precision components, and the
impedance at one end of the filter isn't affected by the impedance at
the other end?

Hi Jim,

I haven't the slightest idea why your objection demanding "high
precision components" is necessary. *Do you have anything that is
quantifiable to sustain this concern? *Give us a Monte Carlo result of
those quantified precisions and their impact on Z.

I'm not going to bother. If you care, you can do so, or point to a
published summary, rather than spending hundreds of words reciting how
to do something we both know how to do.

Again.. without published data from the transceiver maker (whether
derived by measurement or analysis, it matters not) or data from
somewhere else, I made the initial assertion that since using the
tuner to measure impedance depended on the source being 50 ohms (or at
least, known Z), one should not blindly assume that a rig has a output
Z of 50 ohms. You provided one set of data for your rig (35 to 70
ohms), which does actually bound the problem, assuming that yours is
representative of the general class. It might be, it might not.
There's no "trivial" (as in spending no more than 5 minutes) way to
know.

For myself, I don't care, today, what my rig's output Z is, because
everything I use it with doesn't care much (e.g. the auto matching
network finds a match, and whether it's matching 50, 30, or 70 ohms,
there's not much difference). If I were doing something different
(measuring Z with a tuner) I would care, and I'd measure it.

Given decades of lock-step design that conforms to accepted practices,
why would anyone have to measure something to KNOW what it is? *

Funny thing, then... there's remarkably little published (as in
findable with google) data on the output impedance of solid state
transmitters. Yes, the designs are pretty cookbook, but there's a
dearth of published test or analysis (I maintain, of course, it's
because nobody really cares much in actual application situations).

For other RF power amp applications (like plasma etchers, RF heating,
etc) there IS data, but those devices aren't ham transmitters.

I did find a couple master's theses that have some data (but over a
very small frequency range around 7MHz) because they used a ham rig as
a source for a bridge scheme of some sort.


Motorola has for years recited at least three different means to
obtain large signal transistor output Z, and has characterized
individual transistors over frequency in charts.
Sure..


And for those same 30+ years of HF solid state rigs, their power
transistors have had (and still do) output "native" Z of several Ohms.

Would that the active device has a Z that is constant, but it's not.
Sure, the MRF454 data sheet says the output Z is 1+.2j ohms (or
something like that) at 30MHz, but is it still that at 1MHz?

Just happened to be a data sheet I have handy... As you say, others
have more data.
(for large signals, no less)


Looking at a more modern power FET for amplifier use, the IXZ210N50L..
There's a whole page of S parameters, and S22 goes from 0.88@-51deg at
2MHz to at 14.32 MHz to at 30 MHz... that's
at Ids =200mA.. bump Ids to 500mA, and the magnitudes stay about the
same, but the phases change, by tens of degrees.

Again, Motorola specifically rejects small signal parameterization for
power applications. *This is, perhaps, your problem with
characterizing amplifier issues.

Those are actually large signal parameters.. that's a 150V transistor
running at several amps drain current.

Again, it just happens to be a datasheet I had laying around.



I have selected, inspected, and validated transistors to Mil Spec and
found very few wandered from commercial specification. *You must
inhabit a very different realm where production lots contain product
that are "substantially different." *Do you have some quantification
for "substantially?" *Or is this another example of a technician's
shrug?

Substantially, as in Output C being off by a factor of more than 2.
But that could also be packaging effects, which are easier to quantify
by experiment than analysis. That's what breadboards are all about.

I would imagine that for parts used in amateur radios, this is all
thoroughly thrashed out, and there would be no big surprises. Just
that the "as implemented" data isn't readily available in 10 minutes
of searching.

No simple transformer is going to make that look like a constant 50
ohms.

Ah, are we now down to parsing this to "exactly 50 Ohms" where in your
objections you offer few quantifications? *Does 49 Ohms invalidate the
premise and score a home run for the opposing team?

Hmm depends on what sort of accuracy you want in your impedance
measurment, eh? If all you care about is 15-20% accuracy, a pretty
big variation will be ok.


I'd love to see some real data for ham rigs.

Mine (Drake TR-7 and Kenwood TS-430s) exhibit values that vary around
50 Ohms with a low of 35 Ohms and a high of 70 Ohms in the margins.
Those rigs also suffer in those margins. *

so the VSWR looking back from the tuner into your transmitter is
1.4:1? * A return loss of around 15dB... what's that work out to... an
error of about 10-15% in the "measuring impedance with a tuner"
technique... *not bad, but not great, either, especially stacked up
with the other uncertainties..

Not great? *You have already suggested it was unknowable,

I never said it was unknowable. I said it wasn't readily available or
known. Clearly one can measure it, and then know it. And now we do,
at least to 1 sig fig sorts of accuracies.. which is better than we
were 24 hours ago.

and others
state it was immaterial. *It gives me pause to have given a concrete
result to now find what was unknown is now "uncertain" and what was
immaterial now counts for little at "not great."

Immaterial in the usual amateur application (feeding a tuner which
feeds a transmission line which feeds an antenna).. not immaterial
when using the tuner to measure Z.



Although you have to admit that a 2:1 impedance variation isn't a
particularly outstanding "constant impedance load"

This characteristic that is "not particularly outstanding" was
formerly deemed impossible to determine and immaterial by others.

You're confusing "data not easily available in 5 minutes on the web"
with "impossible to determine".




Your comments to the original poster, then, could have been reduced to
one response of one line telling him to abandon his quest by this same
logic.

yes, probably.

But, does a "designed for mass production and cost target" transmitter
fall into that category?
It's not a published spec

You mean you haven't read the spec.

--- Hmm. don't see any tolerance on the output impedance spec on my
IC-7000.. page 150 of the manual: Specifications.
All it says is:
Antenna Connector: SO-239x2/50 ohm.

Page 11, where it describes the back panel
Antenna Connector [ANT1][ANT2} Accepts a 50 ohm antenna with a PL-259
connector.

Page 15, provides a recomendation that the load impedance have a SWR
1.5:1, and a boxed warning that at SWR higher than approximately
2.0:1 it drops power.

The service manual isn't much better, although it does have a
calibration procedure for the built in SWR meter, where you attach a
50 ohm dummy load and set to SWR=1, and then 100 ohms and set SWR=2.
That just calibrates the meter, though, it doesn't imply that the
actual output impedance is 50 ohms.


*I have seen this objection too.
When I've offered just such specs, objectors have then recursed back
into how output Z is unknowable and immaterial as if the topic had
never been encountered before.

ARRL doesn't measure it when they review rigs

Now THERE's an authority! *Do they measure efficiency?

At least they do some measurements and they publish their results.

They do measure efficiency, in a round about way (e.g. they measure
output power into a dummy load and they measure DC input power).

jim

On Jun 7, 10:56*am, Richard Clark wrote:

Do you use the MFJ-259B at the load, or through the line?

Through the line Richard. The way I take the measurements is by
replacing my transceiver with the 259B with the tuner in "bypass"
mode.

73
Dykes AD5VS

"dykesc" wrote in message
...
On Jun 7, 10:56 am, Richard Clark wrote:

Do you use the MFJ-259B at the load, or through the line?

Through the line Richard. The way I take the measurements is by
replacing my transceiver with the 259B with the tuner in "bypass"
mode.

you might want to try the mfj reading again at the coax going into the
tuner. the length of the line and the impedance lumps from the relays in
the tuner will transform the impedance going through the tuner even in
bypass mode. while the difference may be small on the low bands it could be
significant on the higher bands.

dykesc wrote in news:0eccead5-4881-47c5-8881-
:


I am trying to validate impedance values I am measuring with my
MFJ-259B. I attempted this by using my MFJ-993B auto tuner. The tuner
uses a simple L network to create the conjugate match. I took

You may or may not be creating a *conjugate match*, probably not, it is
unimportant. What you are doing with the tuner is adjusting it so that
the network *input* impedance is 50+j0 or VSWR(50)=1.

Think about it.

the final inductance, capacitance values from the auto tuner
digital display after achieving a 1.0 swr match and back calculated
the transmission line
input impedance that the tuner is seeing. I used the RevLoad load
program (link supplied
by Roger in an earlier post) to determine Tuner Z.

For what it is worth, here is some data.


Frequency MFJ-259B Z Tuner Z (calculated)

3.5Mhz 15-j36 27-j42
3.98Mhz 34+j42 32+j28
7.15Mhz 48-j18 46+j2
14.175 Mhz 87+j0 53+j7
28.350Mhz 28-j19 65-j16

I am not surprised, in fact, I would have been surprised if you got good
agreement.

You have two many unknowns. Do you have something that is known, like a
good dummy load, a good VSWR meter? Check that the VSWR meter reads 1:1 n
the 50 ohm load. Then, if you adjust the tuner with a known 50 ohm load
and a known 50 ohm VSWR meter, do the tuner settings reconcile with your
expectations?

This experiment might give you some feeling for the scope of error using
the method.

Owen

On Jun 7, 1:15*pm, "Dave" wrote:

you might want to try the mfj reading again at the coax going into the
tuner. *the length of the line and the impedance lumps from the relays in
the tuner will transform the impedance going through the tuner even in
bypass mode. *while the difference may be small on the low bands it could be
significant on the higher bands.

Thanks Dave. I completely missed that. New data set follows with
transmission line directly connected to MFJ-259B. Maybe a little
better agreement, but still too many variables and uncertainties in
the instruments I have to expect much better. Owen, Cecil, Richard et.
al. pointed this out when I started down this trail. Still, it was a
good exercise and I learned some things along the way


Frequency(Mhz) MFJ-259B ZL Tuner ZL (calculated)

3.50 16-j52 27-j42
3.98 25+j31 32+j28
7.15 39+j11 46+j2
14.17 42-j22 53+j7
28.35 68-j37 65-
j16

On Sun, 7 Jun 2009 10:54:46 -0700 (PDT), dykesc
wrote:

Do you use the MFJ-259B at the load, or through the line?

Through the line Richard.

OK. Then you need to consider that the line is transforming the load
Z to the Z measured (by both methods) at the line input. I realize
you already appreciate this. However, the length of the line in
wavelengths also casts more exaggerated results into your computation
when that length is not a multiple of odd eighth wavelengths. This
suggestion comes from Walt Maxwell's own work and sidenotes to his
measurements as published in any of his several releases of
"Reflections." Read his commentary on this for more detail.

73's
Richard Clark, KB7QHC