Curious whether there's a way to measure impedance of tube final...

[K8DI]

I am wondering if one could rig up a test that measures the complex actual impedance of an output stage. We use pi networks to match them to 50 ohm loads. We use formulas to figure out the approximate values of the network. But can we measure the impedance? Say I have an ordinary single tube class C final, and a VNA. If I just hook it onto the output of the pi network, with the amplifier off, I get some result. If I turn the amplifier on, the VNA will be somewhere between broken and on fire...

I know I can stick a resistor in place of the plate, and then use the VNA on it, but that assumes the plate is a known pure resistance. I could lash up a temporary adjustable bias supply and run the tube with no RF drive, and the bias cranked until the DC operating point matches the tube under drive/load, but will that give an accurate reading? Or would the VNA output turn into fed-back parasitic amplifier output that would blow it up?

Anybody try anything like this before?

Ed

[W1ITT]

There is a way, but it isn't easy.  In the field, in the presence of other high power broadcast antennas, I used to employ an old HP network analyzer that brought the R, A and B connections to the front of the unit..a 3577 works fine, followed by a broadband amplifier, usually an AR unit, anything from 10 to 150 watts, depending on what I was doing.  This was followed by a high power directional coupler.  I used this to overcome the RF coming onto my curtain antenna from adjacent arrays that had to remain on the air.  I ran in a swept frequency mode ordinarily, but fixed mode is fine too. I plotted return loss, but I could also look at a Smith Chart and get R+j for fun.
And the same technique can be used to determine the input impedance of an amplifier.  Cathode driven amps in particular change impedance with drive level.  We guess and build an input network.  I suspect that grid driven amps change by a lesser amount. 
The point is that modern network analyzers with their QRP power levels don't cut it.  Many run at minus 10 dbm.
Another way to skin the cat would be to use one of the Delta Operating Impedance Bridges, sort of like the old GR916 but built for a bit of power.  You adjust for a null on the detector port and there you are.  I've used them at AM sites in the distant past.  They are hard to find and I can't recall if they are specified for anything above the AM broadcast band.  I don't know how much power one could safely push through a GR916 but I'm not about to find out on mine, even though it sits on the retirement shelf in favor of network analyzers.
73 de Norm W1ITT

[W4AMV]

Oh boy... This is a topic that has been visited several times and finds a convoluted path addressing conjugate match vs. not and the concepts of load pull and on and on.

You might want to read:

http://www.w1npp.org/ARES/REFLEC~1/REFLEC~2/BRUENE~1.PDF

as well,  some of the QEX papers on the subject authored by Warren and his trials and tribulations in attempting to find this "dynamic output impedance" of the power amplifier.

I will spare you the details, but in solid state power amplifier design, ONE method of characterizing this so called device output Z is the load pull method. When discussing other classes of amplifier operation such as D, E, F, inverse F and some other alphabet soup classes of amplifier operation, you might want to look at the work of Dr. Steven Cripps. Just google his name along with power amplifiers. There is a wealth of information on what you would think is a simple topic. It is in fact far more involved.

Alan