Marquette Seminar Presentation :: Feb. 1st 2008
So it seems my last post was a little premature. I haven’t gotten the correct answer yet but I’ve been working on some stuff that will get me there. I think I can get the single slot application working in the next week or so. And again, I will probably be so happy that I will blog about the results. eesh
Below is my presentation for today. It sums up a rectangular version of the problem and the steps to go about solving my problem. Hopefully my next presentation at the end of the semester will be a completed thesis.
Sidabras, J.W., “Coupling to Cylindrical Cavity Evanescent Modes”, Marquette Microwave Seminar, February 1, 2008.
You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.
February 1st, 2008 at 5:41 pm
Since the field is evanescent inside the guide, aren’t the scattering and incident fields, then, related by a reflection coeff with magnitude 1? The slot is acting as a reactance, since no energy is propagating into the wave. I suppose it is a complex impedance since it is likely some power is absorbed out of the reactive field into the sample in the guide, but otherwise at 100kHz there should just be a phase angle between incident and reflected fields from the slot, forming your H total. As a result, you aren’t really radiating through a slot anymore — I think that only applies if the slot is an appreciable fraction of the incident wavelength.
Also, on slide 16, what is the comparison being made? The Gaussian fields compared to the expoential fall off of the evanescent waves?
February 1st, 2008 at 5:42 pm
Geez, almost forgot to mention - nice presentation!
February 4th, 2008 at 11:15 am
On slide 16 it shows the data from ansoft (dots), where it has some exponential falloff and a “cap” region. Using a Semi-Log graph and curve fitting (solid) you can tell that the “cap” region is gaussian in nature due to its parabolic shape on the Semi-log graph and the curve fitting. Exponential look like straight lines on the semi-log graph.
What I believe this is saying is that the Hinc is creating some kind of gaussian that falls of as e^(x^2) while the evanescent propagation falls of as e^x as you would expect.
Give me references to what you are talking about with the reactance. I don’t really understand.
Solving the problem I use equivalence principle by using Einducced (by Hinc) to find a magnetic current then solving by integrating the green’s function over the source. But that only gets me Hs, its pretty obvious that Hinc is still in the picture, I just can’t figure out how. Huygens’s principle is just a statement on equivalence principle.
February 4th, 2008 at 3:04 pm
Ok, I wasn’t sure of what slide 16 was showing, that makes more sense.
All I meant by the capacitance is that when looking at the problem of radiation through an aperture, in this case with a 100kHz field, the slit is very small and is backed by a guide that will not support a propagating wave at 100kHz wave, so looking into that slot, it is going to look like a reactance, inductive or capacitive, and as such you will have all of the incident power reflected. You will have power stored inside the guide in the evanescent wave at 100kHz, which I mentioned could lead to a real impedance from loss if some of that energy is absorbed by something inside the guide, but otherwise that slot+guide combination will not have a real impedance, no power transfered, so you will have all power reflected with some sort of phase change due to the reactance. Losses are the only thing that are going to change the impedance looking into that slot at 100kHz from purely reactive to having a real part.
I am basing this on just the behavior of cutoff waveguides, the same way that taking a waveguide and terminating it in a waveguide with half the width will essentially be terminating the guide in a reactance. The thing about losing power to something in the guide is similar to the WG attenuators made by placing a small section of skinny waveguide inline with a waveguide, such that the small section is in cutoff. Despite there not being a propagating wave inside that narrow section, there is a reactive field that can have power coupled off of it by terminating the skinny guide with a wide guide once more within a short enough distance that there is power in the exponentially decaying reative field.
February 4th, 2008 at 3:10 pm
Hrm, maybe I am not looking at this the right way though. I am looking at this from the standpoint of radiating through a narrow slot, the problem you mentioned. In your case, the 100kHz wave is a quasi-static magnetic field, right? I guess my point is that it’s not so much propagating, but forming it’s static field around a magnetic source? So maybe looking at this as propagating outside versus nonpropagating inside isn’t appropriate, because the field is not in a propagation mode exterior to the guide.
February 4th, 2008 at 3:22 pm
so i don’t think of it like that. I use equivelence principle to say everything i need is in the slot induced by Hinc. So if there is a reflection, I don’t care. That reflection would translate into power loss from modes that don’t couple. Instead of looking at it from a circuit model, I look at it from a fields model.
So the problem with the evanescent field is all but solved. I know I have a magnetic current and then I use the proper green’s function to figure out what is going on. The confusion comes in because at z=z’ the evanescent mode is discontinuous *and* there is significant Hinc coupled in the waveguide which seems to be done not by equivalence principle (and has a gaussian like dependence).
I am now looking at Near-field diffraction (fresnel diffraction), because it accounts for the curvature of the wavefront.
Its like Huygens’s principle near-field.
February 4th, 2008 at 8:12 pm
Alright, I see what you are going for. You aren’t using th eradiation through a slot then, and I guess I was hung up on that. Fresnel diffraction sounds like a good way to attack this, since you are dealing only with reactive fields.
One question I have, is at 100kHz, you could also have field coupled in through the walls, right? Could that be accounting for the field that isn’t explained by the slot coupling?
- by the way, heh, I am not challenging you on these things, just adding my two cents in hopes that either an idea will be helpful or to at the very least another point of view.
February 4th, 2008 at 9:10 pm
I make the assumption that the walls are thick enough, lets say 10 skin depths with silver. Using HFSS makes the assumption of PEC then.
Maxwell 3D would be a different story, but I can’t use Maxwell3D because of its quasi-static currents aren’t right.
February 5th, 2008 at 10:41 am
How thick is the walls of the guide? 1 skin depth is about 0.2mm at 100kHz with silver.
Maxwell 3D doesn’t do the correct quasi-static currents? That’s good to know! Is it not meant for quasi-static or did you find this out trying to use it to run simulations?
February 5th, 2008 at 11:13 am
Right, so typical thicknesses range from 1-5 skin depths.
So HFSS does not solve inside metal boundaries, so there is no “leakage” from skindepth. I believe HFSS assumes PEC during the calculation then uses the conductivity to reduce power in the structure.
Maxwell3D on the other hand solves inside the metal. During the simulations I’ve done I can watch the exponential decay in the sidewall and none gets in.
To prove this one of my other presentations did a quasi-static Maxwell3D solution (normalized intensity) and a full-wave HFSS solution (normalized intensity) and they lay right on each other.
The problem with Maxwell3D was that if I did a surface current plot along a slot in the side of a waveguide the top of the slot would have a current profile while the bottom would be at zero. This is fine for creating potentials but not the correct full-wave solution.
Maxwell3D is meant for low frequency work, eddy current and electro/magnetostatics and thermal stuff.
If you put a 2 meter loop and have 100GHz source on the loop, all the current vectors will be uniform!! Since quasi-static. Although the object that I use is << wavelength I still found some issues with current profiles.
February 27th, 2008 at 4:53 pm
[...] Grail Master’s. As expected the Holy Grail master’s is awesome. I eluded to it in my seminar presentation. The holy grail would be now to use the first order solutions I have for one slot and extrapolate [...]