Any comments/suggestions:

Analytical:

(1). Working near cutoff. The input impedance is near the edge of Mr. Smith Boundary. How much power is forward, how much is reverse?

(2). Is this mismatch in impedance effected by the location of your Device under Test? Is load pull an issue with your source?

(3). High Pass Filter. When working for the big "M" 20 years ago, the trunked handi-talkies quite level was -105 dbm (CW). Today synthezed signal generators phase noise -105 dBm/Hz @ 5Khz offset.

Now. The low-side phase noise is killed instantaneouly., But the high pass side runs free, with the wind behind it.

     (A.) The impedance mismatch is lower @+df, more power is effectively coupled to your structure.

     (B.) The attenuation constant is lower @+df.

     (C.) Forget -45 dbm spurs @+df, in the pass band, your dead.

Therefore you start @ +10dbm Source, you lose +20dB from upper end mismatch "Gain at +df", therefor your sensitivity is -105+20-10 = -95dbm, Not good enough for 1982, maybe ok today.

(4.) Is sensitivity really much of an issue, I thought tat with CDMA, etc. It is all in the coding, spectral regrowth etc., what good is CW sensitivity anyway?

Practical: Good Idea, Give it a try.

Anadotal:
          (A). The quantum surfers love it near cut-off, they go faster, too bad they crash, burn and die/

           (B). Honest Indian (tecumsian). I actually witnessed an experiment to test for ether, by rotating a precision waveguide near cutoff at a rate, modulation was intrumented to detect a directional change. This profesional and funded test was not conclusive (no-joke).


 

Thanks for the reply. You bring up some interesting points regarding launch impedance.
We won't be using an actual waveguide launch, though, so the load impedance to the test generator won't be a function of the cavity properties vs. frequency.

At this point, we're thinking of using some small-diameter stovepipe capped at both ends. A small loop at one end will serve as the driving element. The other end will likely have to be capped because we anticipate that it will be difficult to place the rcvr/antenna far enough down the length of the pipe to suppress potential outside sources.

It's not a sophisticated CDMA rcvr that we're testing, but rather a fairly wideband OOK AM unit. Threshold sensitivity is relevant in this case.

I'm surprised that one would have to consider uneven suppression of upper and lower sidebands. No reason not to be well into cutoff, I would think.

More comments:

The small loop will have a very low input impedance - approaching short circuit.

Loading the non-driven end with absorber would be the best to snuff out standing waves, given there is propogation taking place.

Am am thoughly confused about the double negative in the last sentence, I can not comment.

Again, thanks for your comments.

Yes, the small loop would present a very low impedance to the generator. Shouldn't be a problem, I think; the idea is simply to source some current to the loop which should produce a rapidly decaying (non-propagating) field in the tube.

If the tube diameter is less then a half-wavelength would propagation be possible?

To clarify my last sentence in the above post: In a conventionally launched waveguide structure used well below cutoff as an attenuator, the attenuation factor becomes independent of frequency. Therefore, upper and lower sidebands would be attenuated equally.

The lowest cut-off for circular guide is TE11 LambdaC = 3.412*dia., for TM01 Lc=2.613*dia & TE21 Lc =2.057*dia Therefor in circular guide at dia=Lambda/2 three modes are possible.

Now, say the diameter is less than 1/3.412 lambda = .293 lambda, then the guide is cut-off. Only evenescent energy is introduced to the waveguide. This energy is for (most) practicle purposes is snuffed out a 0.25 * wavelength (an old rule of thumb).

If your receiver is is within the 0.25 lamda then some energy may be coupled into it by means of an equivalent lumped reactance. This is how evinescent mode filters are employed.

The impedance of reactance is frequency dependent.

Now the distance from the dut to the source changes the value of this reactance, which in turn changes the input impedance seen by the source. This may become a challanging calibration. I have calibrated quazi-optical bridges for s-parameter measurements, this is bad enough without the source matching issue.

My text (Microwave Engineering, P. Rizzi; and just confirmed in Reference Data for Engineers) shows cutoff wavelengths for circular guides to be half the values you gave. That is,
LambdaC (TE11)= 1.706D
LambdaC (TM01)= 1.306D
LambdaC (TE21)= 1.029D

So for 4-inch dia. stovepipe, LambdaC (TE11) = 6.8 in.
Our test frequency will be 916MHZ --> Lambda = 12.9 in.
Therefore no propagating modes will exist.

Using your rule of thumb, if the DUT is greater than
12.9/4 in. from the source loop, we should be free from evanescent mode effects.

Have I made an error here?

Got my radius confused with my diameter again, wrong again, sorry about that.

Below cut-off there is no propagation only coupling.

At >0.25 lambda the coupling is severly diminished, and the slope is steep.


Say you have two loops, coupled by way of the magnetic field. Changing the DUT position changes the flux linkage between the source impedance and the load impedance. This effects both apparent source and load impedances. The power transfer may not be linear with distance, but who cares, try it.

I still do not feel comfortable with this concept. I feel I was somehow convinced by talk. I have done like a zillion HFFS runs is the past, I remember that the field plots into cut-off waveguide was like a fuzz ball near the source, not much else. This system is cut off, all energy of interest is
evenescent, no way to support propogation just an exponetially decaying em blob. After lambda/4 your going to have trouble, my current thought is that it will be binary, Signal There, Signal Gone. Not particulary useful for your application. I need to crack the books, and dig into it. But I am doing something else now, I have gotten a distaste for the funding side of things. No offense, but why try, when your salary = that of 12 foreign grad students that work 14 hour days including christmas. The EE dept. head gets 5%, good for him and his castle.

That should have been castle(s)!

It seems that what you're predicting - an exponentially decaying "blob" - is just what we're looking for. As long as the DUT is well fixtured, there should be acceptable consistency, unit to unit. As you've said, we should just try it. In fact, other people here have done something like this in the past -- apparently with success. Only difference was lower test frequency.