Designing Dielectric Resonator Antenna (DRA) Fed By Aperture Coupling
Hi there,
I have contacted FEKO support before about modeling a DRA in FEKO and provided their team with the model attached. I got positive feedback, but after I simulated it the model which is available in the literature, I am not getting even close to those measured or simulated results. Can anybody give me some feedback on the model and if there is anything wrong.
I am also attaching the paper in which you will find the complete description of the design.
Cheers
Answers
-
Hello Hamad,
I had a quick look at your model and I see at least one mistake compared to the paper. Face362 should be PEC according to the paper, but it is not PEC in your model. See the image below (selected face, yellow, should be PEC).
<?xml version="1.0" encoding="UTF-8"?>
If you correct the model by changing that face, do you then get the expected results? If you then get roughly the expected results, there are quite a few things that I would check / improve to get a better result. As an example, you are using coarse meshing - consider standard or fine for final results. I would suggest to mesh the slot (the pink face) much finer (much smaller than lambda / 10). I would also mesh the feed line finer, especially close to the slot (under the dielectric). These are just suggestions if you want to improve the result. Below is an example of what I would expect the mesh to look like.
<?xml version="1.0" encoding="UTF-8"?>
You can also consider using symmetry to reduce the required simulation resources and improve the simulation speed (magnetic symmetry at x=0). You can also consider refinement close to the port.
I suspect that just changing the indicated face to PEC will give you the expected results. Unfortunately I don't have time to simulate the model now. Please let us know.
0 -
Hi JIF,
Thanks for having a look at my model and sorry for the delay. I do not get notified when somebody replies to my post. Anyway, I have followed all the steps you have pointed out and still no success. I have also created another model called ' Example_ap_coupled_DRA' in which the port is placed within the dielectric medium of the substrate (one of FEKO examples is done like that), but again I am still not even close. The first model 'RDRA_Poylmer_ApertureCoupled_test2' is basically the same as the original except that the area under the DRA is replaced with PEC, mesh is refined, and magnetic symmetry is placed.
Also, I have found out that when the length of the feed is changed , there is a noticeable change in the input reflection.
Can anybody still help with this problem?
Cheers
0 -
Hi Hamad
I noticed your feed line is lam0/4 wide at the highest frequency. Even mid band this will already significantly support higher order modes (and radiate).
I suggest you modify the design to have a narrower feed line, unless this was intended?
In addition, a rule of thumb with edge ports is that the edge that they are applied to should not be wider than lam0/30. Again at the higest frequency, this is way to wide.
In the attached image, I replaced the via with a triangular section making the length of the port edge only 0.3 mm. At 30 GHz, this is <lam/30.
This should give more consistent results with regards to the feed.
<?xml version="1.0" encoding="UTF-8"?>
0 -
Hi mel,
Thanks for taking the time to look at my model. The model was created based on the paper attached in my original post, so I went with whatever feed's width they proposed.
Would you mind addressing some of the following questions/comments :
Q1) FEKO support told me to connect the two conducting faces which must lie in free space (or homogenous medium), and that's why you see in the little free space region where the port is created. I was not told anything about the feed width in respect to the edge port. In the other model called , the port is immersed within the substrate; I used that technique instead based on one of the FEKO models called which 'ap_coupled_patch_finite' included in the software's directory. So, I am not sure what the difference is between the two techniques and which one is more accurate?
Q2) Is the via (triangular section) connected to the port be lambda/30 or less only, not the edge of the feed?
Q3) I have tried to implement your technique based on your attached picture in which I see that you included that triangular section and got rid of the little free-space region for the port. Unfortunately, I get an error that 'An edge port may not be defined on the surface of a dielectric body, See also message in the output file' ,even though I do not see that is the case. I am attaching the model where I applied that triangular section; would you please let me what I could done wrong here.
0 -
Hi Hamad
Q1) Unfortunately none of the two techniques are absolutely 100% accurate, since this will remain a discrete port that introduces a discontinuity at the feed. In a perfect world the port should be de-embedded, but FEKO does not support this.
The edge width in most microstrip applications does not come into play, probably why it was not mentioned. But in the following post I see it is also mentioned:
https://community.altair.com/community?id=community_question&sys_id=38b680ba1b2bd0908017dc61ec4bcb34
Q2) It is the actual edge that the port lies on, the length of the blue/red line displayed. The via can be any length, assuming you are connecting (will be placing the connector pin) at the bottom of the substrate
Q3) Perhaps the picture was unclear, and it was just a quick mock-up. You still need the free space region. The port edge must either be immersed in dielectric or in free space, but not exactly on the surface of a dielectric. Of course, the FDTD method does not have this limitation for the port.
Trust this helps
Mel
0 -
Thanks Mel for answering my questions,
I am trying to model this while the port is within the dielectric medium; in the model attached
Example_ap_coupled_DRA_testingNewVia.cfx
can you help me with the error I am getting for the port?
0 -
Hello Hamad,
The model that you attached still has an edge port on the boundary of a dielectric surface. To resolve this, split the 'Via' in half and then place the edge port between the two faces. Thus, after the change the edge port will be in the middle of the substrate and not at the top or the bottom.
0 -
Jif and Mel,
Thanks again for taking your time in replying to my inquiry! I have been doing an intensive investigation for months on different DRA examples to make sure I am properly modeling a DRA coupled by an aperture slot feed. In a new example (attached to this reply) taken from 'Dielectric Resonator Antenna Handbook' by Aldo Petosa, the width of the feed is definitely lambda_0/30, even at the upper frequency. I have modeled the DRA in the example using both techniques: having the edge port either within the dielectric substrate or at the edge of the substrate with including that free-space region.
Here are some observations:
1) Both techniques yield approximately the same input reflection coefficient values; however, they are slightly off from the results found in the book. The dip is at 5.96 GHz whereas it is around 5.7 GHz in results found in the book (for the case when ls=10 mm and s=5mm, page 64) . In addition, the input reflection coefficient found using FEKO models is way off at frequencies off that dip. If you believe that the FEKO models are defined properly, then do you think that the results in the book might not be as accurate as FEKO models?
2) I have noticed as the length of the feed is shortened, there is a change in the input reflection coefficient. This can be observed in the following attached models:
Example_ap_coupled_DRA.cfx: the feed line length is 20 mm + stub length of 5 mm (ground plane size 40 mm x 26 mm)
Example_ap_coupled_DRA3.cfx: the feed line length is 15 mm + stub length of 5 mm (ground plane size 40 mm x 26 mm)
Example_ap_coupled_DRA4.cfx: the feed line length is 12 mm + stub length of 5 mm (ground plane size 40 mm x 26 mm)
If you make the feed's length 10 mm, the dip actually vanishes in that frequency range. I have even seen cases in which making the length longer can drastically influence the outcome of the input reflect coefficient. Such thing does not agree with the theory of transmission line, so is there any explanation for this? By the way, the edge port in the above models was placed in the dielectric medium, so the feed's length is varied without changing the size of the ground plane.
Your comments are highly appreciated about the issues of discrepancies between the book's result and FEKO models and the effect of the feed's length.
Example_ap_coupled_DRA2.cfx is a model where the a larger ground plane is used .
Example_ap_coupled_DRA_FreeSpacePort.cfx is a model where the edge port is placed in a free-space medium.
All FEKO models are contained in the folder attached.
Cheers,
Hamad
0 -
The DRA does not present a constant impedance over the band and as a result, the length of the feed line is critical. But the length of the feed line (substrate dimensions) are not given. This is critical in order to try to exactly match the results.
I would however mesh the aperture with 4 triangles across the width of the aperture. I also attach an image of a suggested feed - place the edge half way between the bottom and top. It is more 'symmetric' with respect to the bottom and top conductors.
But again the substrate dimensions are what's needed.
0 -
Mel,
The quarter-wavelength stub length is fixed, so I thought that's what matters based on my understanding to the theory of transmission line. In respect to the substrate and ground length, yes it is not specified in the book example; however, it is assumed that length is made large enough for the DRA to work.
I have tried to place the edge port half way between the bottom and top part of the via and compared to the technique I have used before, and that do make a difference in the result. However, meshing the aperture (do you mean the via) with four triangles across the width of the aperture is something I have not attempted before and I am not sure how it can be done. Can you possibly show me how it is done? Other than that, do you think the model is flawless.
0 -
I don't see any other problems with the model.
For the aperture mesh size, set a local mesh size on the face in the details tree. I see you have used local meshing on other faces already.
0 -
Further to my previous point, I compared the input impedance on a Smith chart of two different substrate/ground sizes.
In the larger version I set the grnd_w = 26*2 and grnd_l = 39*1.5
One can clearly see the effect of this below.
The larger ground size seems to have a double resonance, pretty much centered around the origin on the Smith chart, but this is not always clear on a Cartesian graph. So maybe a good idea to play around with the size.Side note - if you use magnetic symmetry on Y=0 axis you can save resources.
0 -
Altair Forum User said:
I don't see any other problems with the model.
For the aperture mesh size, set a local mesh size on the face in the details tree. I see you have used local meshing on other faces already.
Yes, do you keep entering a number that would result in four triangles along the width?
0 -
Altair Forum User said:
Further to my previous point, I compared the input impedance on a Smith chart of two different substrate/ground sizes.
In the larger version I set the grnd_w = 26*2 and grnd_l = 39*1.5
One can clearly see the effect of this below.
The larger ground size seems to have a double resonance, pretty much centered around the origin on the Smith chart, but this is not always clear on a Cartesian graph. So maybe a good idea to play around with the size.Side note - if you use magnetic symmetry on Y=0 axis you can save resources.
<?xml version="1.0" encoding="UTF-8"?>
Mel, it is not the size of the ground plane that is causing the impedance to change: it is actually the feed length. As a proof, if you go back to my models Example_ap_coupled_DRA and Example_ap_coupled_DRA2 (see above), where the edge port is within the dielectric substrate and hence the feed length does not change as you change the ground plane's size, you will see that there is barely any difference between the two models when you keep the feed's length the same.
<?xml version="1.0" encoding="UTF-8"?>
0 -
Excuse for not being clear. What I said in an earlier post, we need to know the feed length, and it's not given in the example from the book.
0 -
Hey Mel,
This is the response of the author on the size of the ground plane and feed length:
If the ground plane dimensions are fixed, then I would think that
the feed length would only change the phase of the reflection coefficient,
but not the amplitude (assuming a lossless feed).
However, if the ground plane size is changing along with the feed length,
then there could well be a change in the reflection coefficient magnitude
as well as phase.
0 -
I have reconsidered your previous comments, Hamad and your previous assessment on the feed line length is correct. Adding a phase offset to a longer feed line on the Smith chart in POSTFEKO leads to a near identical response as the original line length.
I have meshed this model super fine and the response we get in FEKO is virtually unchanged.
I am not sure if there is some other parameter perhaps missing from the description (I see they don't mention losses, but this would be very minor I think) but I cannot find anything else wrong with the model.
0
Altair Employee
