There is no problem with the square ports and the round waveguide. CST calculates these correctly.

There will almost always be a small reflection at the large end of the horn, so S11 will almost never be zero.

Making the target PEC results in the signal being reflected from it, but not all of the reflection goes back into the horn. I'm surprised its as much as 90%!

With a reflection from a flat surface, it is as if there is another horn behind the surface (the image) radiating back.

To get a larger reflection you would have to make the surface concave, with the focus at the phase centre of the horn, which is at the centre of the aperture at long range, but might be further inside the horn at the distance in your arrangement.

The phase centre is the centre of curvature of the waves at any position. It is constant for any long range, but is different for different distances close to the horn.

The waves look spherical, but for the first few aperture diameters of distance the radius of the sphere actually reduces, before increasing again, because at the aperture the wave is nearly flat.

Malcolm White Thank you for your reply.

I should correct my description, actually S11 reaching around 0.985 is the situation where I did't put antennas, but two plane waveguide ports covering the whole boundaries of the Z direction.

I tried to model the scalar horn corrugated antennas and finish the simulation. The results are like in the files.

The specimen is also PEC, but S11 reaches at more than 1 and vibrates between 0 to 1 while S21 is a constant very close to 0.

I can't change the surface of the specimen since it is a simulation which simulates a specimen is exactly a flat surface.

For me I think since S21 is ideal, the reason why S11 is not a constant close to 1 is that parts of the waves leak outside the antenna, in the real experiments we can use calibration of VNA to normalize it, but how to optimize it in CST?

Thank you.

The ripples are due to multiple reflections, probably between the reflector and the horn, or possibly the two horns. The two reflecting objects are half wavelength apart at one ripple period, which is about 454 MHz, I think, so about 33 cm apart, if in air.

1.0006 is so close to 1 the error really is insignificant. I would expect the accuracy would be a bit worse than that.

0.985 is only 1.5% short of 1 and is close enough for most antenna purposes! What were the walls? Are they PEC? Or PMC top and bottom and PEC at the sides? Where does CST say the power went? The tree on the left has a list of where the power went.

Malcolm White Thank you for your reply!

The idea of this simulation validation is that between two ports, put "PEC" first to see if the S11 is around 1 and S21 is around 0 (both in linear); then, put "Normal" in between two antennas to see if S11 is around 0 and S21 is around 1. If so I can say that this simulation is reliable and then simulate the real specimen.

According to your suggestion, I have discussed and decided not to use antennas in this simulation but "plane wave", just delete the antennas and put ports on the surface of vaccum brick (blue bricks), and the boundary condition is attached in the files also.

The result of this simulation is good for reflction, which is PEC in between, S11 is around 1 and S21 is around 0; however, when I changed the material in between as "Normal" (epsilon=1, mu=1.conductivity=0), the result doesn't change a lot, S11 is still close to 1 and S21 is close to 0, which is not expected.

I have attached the results of "full transmission" and where power went in this scenario, could you please also give some ideas of why full transmission doesn't give good result as full reflection?

PS: The way of setting boundary condition as this is from the CST website, which I used as blackbox...

The boundary conditions seen fine, for Y polarized plane waves. They should just go straight through with no reflection.

Try taking out all the material and just having one block of vacuum, and see what you get, then replace it by air- you shouldn't be able to see any difference. Then put in the metal lump. You don't need to do any subtraction or insertion, CST lets conductors override dielectrics by default.

Are your ports facing the right direction?

Malcolm White

Hello,

I have got good result for validation, including metal/air/a PVC (epsilon=2.7) in between two plane waveguide ports.

However when I use real specimen to do the simulation, the result goes wrong again. The specimen composites of two layers with epsilon=5.05 and one layer with epsilon=1.3, according to the experimental, the S21 should be around -19 (dB) within frequency range 50-51GHz, but actually in the simulation it gives only -0.2 (dB). And the simulation needs a lot of time even more than 30hours and still doesn't finish...

I really don't know where to correct this problem, is this possible to add you in some remote contacting software like Teams or Skype? I think it is the easiest way to see the problem...of course if it is in your convinience.

Thank you very much!

If the simulation goes on for ever it is because the power can't get out of the solution space. This can be because there are big reflections somewhere, like a cavity between two big mismatches

I would guess you have one of your materials wrong, or a layer or two of another material in there somewhere. Try adding the layers one at a time. When it takes for ever it usually means there is a lot of high reflection happening. With epsilon 5 the mismatch is only about 2.2 to 1 so about 1/10 of the power is reflected at each boundary so it should fall off reasonably quickly.

Have you got the losses right? Is there a wrong sign in there somewhere?

Because your boundary conditions allow a plane wave, the problem doesn't need to be more than 1/4 or perhaps 1/10 wavelength square in the x-y plane, which should speed things up if things are working right. It won't solve the problem if you have big reflections you shouldn't have, though.

Sometimes you can learn from the port signals, which show you the transient reflection from the first boundary before multiple reflections set in.

Try both ports one at a time and see if they give the same result (and port signals).

With your boundary conditions you shouldn't need to use symmetry planes, but if you are usig them make sure they agree with your boundary conditions.