On 15 jun, 05:49, K7ITM wrote:
On Jun 14, 7:00*pm, walt wrote:
On Jun 14, 2:56*pm, K7ITM wrote:
On Jun 13, 11:42*pm, Owen Duffy wrote:
Owen Duffy wrote :
...
In the measurements of an IC7000 that I made, the measured output
power on one VSWR(50)=1.5 load was 82.5W when it would have been
104.6W had the source been 50+j0, an error of 0.8dB. I opined that
this test did not support the proposition that Zs was not 50+j0
Too many "nots", isn't there?
It should read:
I opined that this test did not support the proposition that Zs was 50+j0.
Apologies, Owen.
I suppose this will be buried where nobody will read it...
I realized that with the nice instrument-grade directional couplers
that came with a new 100W RF power amplifier, and with the other
equipment on my bench, I can measure RF amplifier/transmitter source
impedance relatively easily and with good accuracy. *I strongly
suspect the accuracy will be limited first by how well the setup of
the transmitter/amplifier can be duplicated, and not by the
measurement instruments.
I won't go through the whole test setup, but just say that
substituting an open or short for the connection to the transmitter
yields the expected amplitude return signal, and terminating the line
in a precision 50 ohm calibration standard yields a 47dB return loss,
for the frequency I was measuring (nominally 7MHz, for this first
measurement). *The measurement involves sending a signal offset from
the nominal transmitter frequency by a few Hertz at about -20dBm
toward the transmitter, and looking at what comes back.
Measuring a Kenwood TS520S, set up for about 70 watts output, ALC
disabled, operating as a linear amplifier somewhat (about 30 watts)
below its maximum output: *result is 56+j16 ohms at the output UHF
connector on the TS520S. *That's about 1.4:1 SWR, and at some point
along a lossless line, that's equivalent to about 70+j0 ohms: *not
terribly close to 50 ohms. *I'm not going to bother with a detailed
error analysis presentation, but I'm confident that the amplitude of
the return loss is accurate within 0.1dB, and the phase angle within
10 degrees, to better than 99% probability.
I may make some more measurements with different amplifier setups and
at different frequencies, but for now, that's it...
Cheers,
Tom
Tom, you stated earlier that you measured the source impedance of a
TS520S transceiver by inserting a somewhat off-resonance signal into
the output terminals when the rig was delivering 70 watts, and the
source impedance was measured as 56+j16 ohms. However, you chose not
to describe the setup or the procedure for obtaining this data.
I'm hungering to learn of the setup and procedure you used, because
I'd like to know what reflection mechanism gave a return signal that
could be discriminated from the 70w output signal from the
transceiver.
In his Nov 1991 QST article Warren Bruene, W5OLY, used what I believe
is a similar procedure, in which he claims he measured the Rs that he
called the 'source impedance' of the RF amp. *He used his measurements
in asserting that because his Rs didn't equal RL there could be no
conjugate match when the source is an RF power amp. I have never
believed his procedure and measurements were valid, and I still don't.
So if your setup in any way resembles what Bruene presented in his QST
article I would like to know how you can justify a procedure that
involves inserting an off-set frequency signal rearward into an
operating RF power amp *to determine the source impedance.
Walt, W2DU
OK...
So let's consider making a load-pull measurement of source impedance.
Since we're trying to resolve both resistance and reactance, we need
to change the load in at least two directions that have a degree of
orthogonality. *But we could also change the load over a range of
values. *For example, we could connect a 51+j0 load directly to the
output port we're trying to measure, and then connect it through
varying lengths of 50.0 ohm lossless coax. *45 electrical degrees of
line would shift the phase of the 51 ohm load so it looks instead like
49.99-j0.99 ohms. *90 electrical degrees shifts the 51 ohm load to
49.02+j0 ohms, and so forth. *Measurements of the varying amplitude
output with those loads will give us enough information to resolve the
source resistance and reactance and open-circuit voltage.
For a 51 ohm load on a 50 ohm line, the reflection coefficient
magnitude is 1/100, so if the transmitter is putting out 100Vrms
forward, the reverse is 1Vrms.
Now consider a method to change the line length that doesn't use
individual sections that have to be patched in and out, but rather
uses a "trombone" section that, in theory anyway, could range from
zero length to essentially infinite length. *Picture that trombone
section getting longer at a fixed rate, so now the load is rotating
around a circle on the linear reflection coefficient plane (which is,
by the way, exactly the same plane the Smith chart is plotted on); the
circle is centered at zero and is a constant 1/100 amplitude, with
linearly varying phase. *So the 1Vrms reverse wave on the line of the
100Vrms forward example arrives back at the amplifier at continuously
varying phase. *Imagine that the phase shift is 360 degrees in 1/100
of a second. *Now note that the reverse wave corresponds _exactly_ to
a wave offset in frequency from the forward wave by 100Hz. *If the
line is continuously lengthening, the offset is negative; if the line
is shortening instead, the offset is positive.
Now, from the point of view of the amplifier, can that scenario be
distinguished from one in which I have a perfect 50 ohm load that
absorbs all the transmitter's output, and a method to introduce a
"reverse" 1.00Vrms wave into the line at a frequency that's offset
from the transmitter's output by 100Hz?
If you believe that the amplifier can distinguish between those two
scenarios, I fear we have nothing more to discuss.
Cheers,
Tom
Hello Tom,
The method of constant varying phase of a mismatch to determine output
impedance I use in my simulation also (because it saves me time). I
use a second source with some frequency offset instead of the trombone
as you described (the trombone I cannot implement easily in my pspice
package).
When you change the phase very slowly, a soft power supply may change
voltage during the slow variation (as a load change may result in a DC
supply current change). When the difference frequency is sufficiently
high (for example 200 Hz or more), the electrolytics will keep the
supply voltage constant during all phase "steps". Therefore there may
be slight difference between measurement with a set of loads with
increasing phase and the RF injection method (for example with vector
analyzer).
I checked the injection method in simulation for linear circuits (with
both real and complex output impedance), linear active circuits, and
power circuits. I couldn’t find any flaw in the method. So I think
your reasoning is valid.
Some results applicable to PAs, inclusive the implementation in
simulation are in:
http://www.tetech.nl/divers/PA_impedance.pdf. Note
that I determine the reflected voltage by means of interference with
the amplifier's output signal (envelope detection) as this saves me
from making a narrow band filter that results in very long run times.
I am now simulating a circuit with 6146 valve and pi-filter output (50
Ohms) and I will add it to the document.
Best regards,
Wim
PA3DJS
www.tetech.nl
without abc, PM will reach me