• Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

CST MWS boundary conditions problem?

Status
Not open for further replies.

winson

New Member
Hi everybody,

I'm new to CST microwave studio. Just finish constructed a structure of an L-probe patch antenna (from IEEE paper) and just run the simulation by transient time solver, the curve of the return loss(S11) against frequency that i get is different from what showing on the IEEE paper, so is it the boundary condition setting will affect the simulation results? And actually what is the function of setting the boundary condition?

During the simulation, a warning message "some PEC material is touching the boundary" was show. After change the boundary condition setting to "open(add space)" then the warning will eliminate when run again the simulation. But the s11 curve still different from the "actual" results.

Anybody can help?
Any comments will be appreciate.

Thanks.:)
 
Last edited:

JimB

Super Moderator
Most Helpful Member
OK, starting from a position of relative ignorance myself, I dont think that there will be many people here who are familiar with Microwave Studio.

You have simulated some published antenna design and the simulated return loss plot does not agree with that published in the original article.
You dont specify in what way or by how much the two plots differ.

Was the plot in the original article obtained by simulation or by measurement on a practicle example of the antenna?
If it was simulated, did they use the same simulator as yourself?
If it was measured how good was their test equipment?

JimB
 

winson

New Member
Hi JimB,

From the article, i know that they obtain the return loss plot by measurement, but the article does not specify what equipment they using to get the plot.

I have attached my simulation results(the picture no.1), and also the original plot from the article(picture no.2). The shape of my plot is similar to the original curve but my frequency range is from 0 GHz to 10 GHz, which is very much different from the 1.7GHz - 2.7 GHz in the original article. My simulation is obtain with a warning message occur as describe in previous post.

So if we ignore the factor "how good is their equipment", is it possible for both the simulation and measured results same with each other?

I have also attached the article and the structure of the antenna that i draw in CST.

Thanks in advance.
 

Attachments

Papabravo

Well-Known Member
In the world of partial differential equations boundry conditions are EVERYTHING. Without the boundry conditions there are an infinite number of feasible solutions to the PDE. In order to have a unique solution, the problem must be well-posed. It is the description of appropriate boundry conditions that makes the problem "well-posed" and the solution unique.

Here is the general Wiki on boundry value problems
Boundary value problem - Wikipedia, the free encyclopedia

In a Dirichlet problem you need to specify the values of the solution at the boundries.

Dirichlet boundary condition - Wikipedia, the free encyclopedia

In the Neumann problem you specify the derivative of the solution at the boundry

Neumann boundary condition - Wikipedia, the free encyclopedia

There are also lesser known types which are combinations of these two types.

From the plots it looks like you got the dimensions wrong. Did you confuse inches and mm. perhaps?
 
Last edited:

RadioRon

Well-Known Member
Hi everybody,

I'm new to CST microwave studio. Just finish constructed a structure of an L-probe patch antenna (from IEEE paper) and just run the simulation by transient time solver, the curve of the return loss(S11) against frequency that i get is different from what showing on the IEEE paper, so is it the boundary condition setting will affect the simulation results? And actually what is the function of setting the boundary condition?

During the simulation, a warning message "some PEC material is touching the boundary" was show. After change the boundary condition setting to "open(add space)" then the warning will eliminate when run again the simulation. But the s11 curve still different from the "actual" results.

Anybody can help?
Any comments will be appreciate.

Thanks.:)
The boundary condition defines the radiating environment of the model. If you specify OPEN, the simulator effectively places perfectly matched microwave absorber material at the boundary, which guarantees that this plane appears to be open space. If you define the boundary to be an electric (Et=0) boundary, then this is effectively the same as placing a perfect ground plane on that boundary of the rectangle that contains your model. Obviously this would have a severe effect on the model. Other conditions do similar critical things, so it is very important to define the right boundary conditions. As an antenna designer, I have typically used the boundary condition Open (Add Space) in order to insure that my nearest fields are not hitting the boundary as this seems to give me the best results. When I load an example of a patch antenna provided by CST, I note that they just use Open as their boundary for five of the six planes, and place an electric (Et=0) boundary behind their pcb. This would generate the smallest number of meshcells so the simulator would compute more quickly than with my method.

The biggest trouble I usually get into with my models is inappropriate meshcell sizes. Often, I find that the automatic mesh generator doesn't provide accurate results and I have to tune my mesh. This is a difficult process for me because I don't understand all the variables but trial and error plus understanding what mesh sizes I really want seems to give a good result most times. A beginner will probably find that the process of tuning the mesh sizes for fewest meshcells vs most accurate results is the most laborious and sometimes hard to understand part of using a tool like CST.
 

RadioRon

Well-Known Member
Hi JimB,

From the article, i know that they obtain the return loss plot by measurement, but the article does not specify what equipment they using to get the plot.

I have attached my simulation results(the picture no.1), and also the original plot from the article(picture no.2). The shape of my plot is similar to the original curve but my frequency range is from 0 GHz to 10 GHz, which is very much different from the 1.7GHz - 2.7 GHz in the original article. My simulation is obtain with a warning message occur as describe in previous post.

So if we ignore the factor "how good is their equipment", is it possible for both the simulation and measured results same with each other?

I have also attached the article and the structure of the antenna that i draw in CST.

Thanks in advance.
I agree with papbravo that perhaps the first thing to double check is your dimensions. The paper shows a fairly narrowband antenna while yours appears to be quite a bit broader and much higher frequency. The higher frequency of your curve is a giveaway that something is the wrong size.

For more help, can you provide some sort of dimensioned image or perhaps email me a file to look at? (contact via PM)
 

RadioRon

Well-Known Member
I made up a quick model to reflect the dimensions in the paper, but did not tune it carefully. The attachment shows that the result was roughly similar to the paper's results, that is, good return loss from around 1.8GHz to around 2.4GHz.



By far the most critical elements to this design lie closest to the feedpoint. It is critical that you copy their geometry near the feedpoint as closely as possible. When I compare my model to yours, the first obvious difference is that you have cut away the ground plane around the feedpoint and presented the thick metal edges to the feedpost. This will add a lot of capacitance resulting in a very unusual characteristic impedance through that cut. In my model, I have used a waveguide port that is circular and passes through the metal of the groundplane almost exactly like the way they have built their prototype. They mention feeding using a TNC connector, which is round, not rectanglar and which maintains a 50 ohm characteristic impedance until it reaches the topside of the gound plane. Your launch does not mimic this.

It is also worth mentioning that my model also shows another mode at around 7GHz that results in some radiation and return loss around 7 dB. Your model has this too. This is just another mode of radiation from the patch and can be ignored. The big problem is that your patch is not resonating near 2GHz.

edit after some tuning: another critical point is that their prototype has the ground plane with larger dimensions than the patch itself. I found that when i also enlarged my ground plane, the patch resonance became much better defined, with very high return loss and very narrow bandwidth. This is critical, because the way this antenna achieves broader bandwidth is that it starts with a narrowband patch, and then simply introduces an additional series LC circuit in the feed to create another resonance at the lower frequency. You can't really tune the series LC unless you find the patch resonance first. The larger ground plane brings this to light.
 

Attachments

Last edited:

Papabravo

Well-Known Member
Give Radio Ron a round of applause. Truly he has gone above and beyond the call of duty for the OP. Way ta' go RR!
 

JimB

Super Moderator
Most Helpful Member
Give Radio Ron a round of applause.
Yes RR is The MAN on this one.

Regarding the OPs simulation giving responses at significantly higher frequencies, I was wondering if he was using the wrong dielectric constant for the patch substrate?

JimB
 

RadioRon

Well-Known Member
Thank you for those kind words. We don't get many questions on antenna simulation, and this is a minefield for the beginner.

On further reflection, I believe that some of those higher frequency responses are legitimate secondary or higher order modes whose frequencies are dominated by the dimensions of the patch. I have them in my model too. Then again, the length of the feed going up to the patch is quite long too so some higher frequency resonances may be coming from the feed structure. The paper that the OP references only reported results up to 2.7GHz so who knows how messy their response might have been at 6GHz and up.

I don't think the dielectric constant will have a large effect at 2.4 GHz since the dielectric is outside the volume of the patch, not inside. However, if the feed structure is resonating then the dielectric constant becomes a bigger issue since it surrounds a good part of that feed. It would certainly affect the capacitance between the feed strip and the main patch. In any case, Rogers Duroid 5880 has a dielectric constant of about 2.2 which is what I used.
 

Papabravo

Well-Known Member
I learned most of what I know about antennas from EZNEC, a free download. Never worked on anything at that GHz. frequency range, but I know a bit about inverted V's from 80M to 10M.
 
Last edited:

RadioRon

Well-Known Member
EZNEC is a surprisingly good tool, I have used it many times. There are some antennas that are so much easier to build and simulate in eznec when compared to the more complicated 3D EM simulators that I prefer to fall back to EZNEC in some cases even today. I think this is true of all antenna engineers.
 

winson

New Member
Hello,

Thanks everybody for giving all the information to me:), i'm just finish spending some time to understand all that.

From the plots it looks like you got the dimensions wrong. Did you confuse inches and mm. perhaps?
..................................................................................................
I agree with papbravo that perhaps the first thing to double check is your dimensions. The paper shows a fairly narrowband antenna while yours appears to be quite a bit broader and much higher frequency. The higher frequency of your curve is a giveaway that something is the wrong size.
Regarding this, i think i should not make a mistake here because in CST there is a status bar located at the bottom right corner and showing the setting is "mm GHz ns.... etc". So all dimension will be in mm and i double check the paper all the dimension given also in mm range. I have also check that the dielectric constant of the substrate is set to 2.2


After reading all the comments from RadioRon, i think i should:
1. Increase the size of the ground plane to very large dimension in order to increase the return loss.
2. Change the waveguide port from rectangular to a circular shape and the size (diameter) just slightly bigger the vertical feedpost so that the capacitance there can decrease and maintain the characteristic impedance at 50ohms.
3. Change all the 5 boundary conditon to "open (add space)" instead of "open" and set electric (ET=0) boundary behind the ground plane.
4. Reduce the thickness of the ground plane. (so what is the suitable thickness for ground plane, also what is the thickness for patch and horizontal strip feed?)
5. Move the contact between the vertical feedpost and the horizontal strip to the edge of the feed strip (as in the model of RadioRon)
6. Extend the substrate in the feeding part so that the substrate intersect with the vertical feedpost (like what RadioRon done)

So is it correct?:)

Thanks!
 

winson

New Member
The biggest trouble I usually get into with my models is inappropriate meshcell sizes. Often, I find that the automatic mesh generator doesn't provide accurate results and I have to tune my mesh. This is a difficult process for me because I don't understand all the variables but trial and error plus understanding what mesh sizes I really want seems to give a good result most times. A beginner will probably find that the process of tuning the mesh sizes for fewest meshcells vs most accurate results is the most laborious and sometimes hard to understand part of using a tool like CST.
__________________
RadioRon
Just have another doubt, as automatic mesh generator doesn't provide accurate results and we need to manually tune the mesh, so after some tuning will be the new results is totally different with the one using automatic mesh generator?
 

RadioRon

Well-Known Member
Your list looks good. I would use 0.1 mm for the thickness of the metal plating on the substrate. The ground plane can be a bit thicker, maybe about 0.5mm.

The circular opening through the ground plane that is provided by a TNC connector is 5.3 mm diameter, so you should use this size of round hole. The center conductor of this type of connector is about 2 mm, but I varied this dimension during my tuning of the model.

The paper does not specify the ground plane size exactly, but the diagram seems to show that it is approximately 5 to 10 mm larger on each edge when compared to the patch dimensions. I tried 5 mm and the results were good, then I tried 10 mm and things shifted a bit, but the patch resonance was clear in both cases, much clearer than when I had the ground plane the same size as the patch.

To try and answer your original question about whether it is possible to simulate and match their results, the answer is yes but it requires very careful attention to matching exactly how their antenna is built. This is quite difficult since they don't provide enough details or a photo. It is also necessary to optimize the mesh of the simulation which includes a lot of details and can take hours to do.

Another important factor is that their test setup can affect the S11 plot and they don't explain their test setup. For example, if the antenna is not well decoupled from the feedline to the VNA, the S11 will be significantly affected by the length of their feedline and how it is oriented and what metal objects are nearby.

It can be a frustrating exercise to try and match their results exactly, but you might learn something in the process.
 

RadioRon

Well-Known Member
Just have another doubt, as automatic mesh generator doesn't provide accurate results and we need to manually tune the mesh, so after some tuning will be the new results is totally different with the one using automatic mesh generator?
Perhaps. It depends on your idea of "totally". My experience is that the results change, but the major characteristics don't change. I mean, for example, that big resonances are still there and might shift in frequency a bit. Smaller details can change a lot though.
 
Status
Not open for further replies.

Latest threads

EE World Online Articles

Loading
Top