Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

How to define coaxial lumped Ports when simulating the quarter of a geometry

Please login with a confirmed email address before reporting spam

Hi all,

I am simulating a TEM (Transverse electromagnetic) cell. The geometry have two PMC (Perfect Magnetic Conductor) symmetry planes, so I simulate only the quarter of the geometry. The two planes of symmetry bisects the coaxial lumped port, then I adjust the settings by multipling the impedance by for 4. I have strange results so I think I am not doing it in the right way. My question is how to define the coaxial lumped port . Should it be defined by user ? If yes how to specify 'h_port' 'w_port' and 'a_h' ?

Atatched a figure to explain more my question.

Thank you for your help, Emna



8 Replies Last Post 25.04.2019, 08:50 GMT-4
Robert Koslover Certified Consultant

Please login with a confirmed email address before reporting spam

Posted: 5 years ago 30.03.2019, 12:11 GMT-4
Updated: 5 years ago 30.03.2019, 12:12 GMT-4

Interesting question, so I just played with this a bit. My suggestion is to not even bother with a LUMPED port, in this case. Consider using a "coaxial port." (Boundaries --> Port --> Coaxial). Yes, I know, but it's only a quarter of a coax, right? Well, despite the fact that your port is not the full pair of concentric circles, Comsol Multiphysics seems to understand that it is a legitimate coaxial port (aka, a coaxial RF port) presumably based on the existence of the PMCs and concentric circular arcs. I was actually (and pleasantly) surprised it understood my specification from that perspective, but it did. Anyway, via this approach, it looks as if you can proceed much as if you had started with the full model, and then extract S11, VSWR, etc.

-------------------
Scientific Applications & Research Associates (SARA) Inc.
www.comsol.com/partners-consultants/certified-consultants/sara
Interesting question, so I just played with this a bit. My suggestion is to not even bother with a LUMPED port, in this case. Consider using a "coaxial port." (Boundaries --> Port --> Coaxial). Yes, I know, but it's only a quarter of a coax, right? Well, despite the fact that your port is not the full pair of concentric circles, Comsol Multiphysics seems to understand that it is a legitimate coaxial port (aka, a coaxial RF port) presumably based on the existence of the PMCs and concentric circular arcs. I was actually (and pleasantly) surprised it understood my specification from that perspective, but it did. Anyway, via this approach, it looks as if you can proceed much as if you had started with the full model, and then extract S11, VSWR, etc.

Please login with a confirmed email address before reporting spam

Posted: 5 years ago 02.04.2019, 03:50 GMT-4

Thank you very much for your prompt reply. I used a "coaxial port" instead of "lumped port" but I got this error message 'the slit condition should be applied only on an interior boundary or the port should be placed on an exterior boundary' so I activate the slit condition. I don't get any error message anymore but the results are false. Is there anything else to do when dealing with ports instead of lumped ports ?

Thank you for your help, Emna.

Thank you very much for your prompt reply. I used a "coaxial port" instead of "lumped port" but I got this error message 'the slit condition should be applied only on an interior boundary or the port should be placed on an exterior boundary' so I activate the slit condition. I don't get any error message anymore but the results are false. Is there anything else to do when dealing with ports instead of lumped ports ? Thank you for your help, Emna.

Robert Koslover Certified Consultant

Please login with a confirmed email address before reporting spam

Posted: 5 years ago 02.04.2019, 13:44 GMT-4

You normally shouldn't have to activate the slit condition.
Your port should be on an external boundary. (It is, right?) Make sure that the boundaries (edges) of your port are conductors on the curves and PMCs on the radii. It works for me. See attached example,

-------------------
Scientific Applications & Research Associates (SARA) Inc.
www.comsol.com/partners-consultants/certified-consultants/sara
You normally shouldn't have to activate the slit condition. Your port should be on an external boundary. (It is, right?) Make sure that the boundaries (edges) of your port are conductors on the curves and PMCs on the radii. It works for me. See attached example,


Please login with a confirmed email address before reporting spam

Posted: 5 years ago 05.04.2019, 05:34 GMT-4

Hi Robert,

Thank you so much for your help and for you example ! Yes my port is on an external bounday and normally I defined it as you did in your example but I still have error message. I can't figure out the fault in my model. I have attached my mph-file and my geometry .xt (imported in the mph-file in the hope that you see the error and can help me.

Thank you for your help, Emna.

Hi Robert, Thank you so much for your help and for you example ! Yes my port is on an external bounday and normally I defined it as you did in your example but I still have error message. I can't figure out the fault in my model. I have attached my mph-file and my geometry .xt (imported in the mph-file in the hope that you see the error and can help me. Thank you for your help, Emna.


Robert Koslover Certified Consultant

Please login with a confirmed email address before reporting spam

Posted: 5 years ago 06.04.2019, 16:13 GMT-4

I took a quick look at your file. It seems to me that your ports are NOT external boundaries. They are boundaries of solid dielectrics, but there is air on the other side. An exterrnal boundary is a boundary on the external surface of the model. Air, or any other material domain that gets meshed and has FE fields computed in it, is part of the internal part of the problem.

Although there may be an easier way to proceed, what I normally do if I want to have an "external" port but at the same time I want air or other materials very near/around it, I add (actually, I subtract from the local volume) a small volume that is right up against the port, thus creating a truly "external" surface on the actual port. (This little volume is not to be meshed, and not part of the computational space.)

-------------------
Scientific Applications & Research Associates (SARA) Inc.
www.comsol.com/partners-consultants/certified-consultants/sara
I took a quick look at your file. It seems to me that your ports are NOT external boundaries. They are boundaries of solid dielectrics, but there is air on the other side. An exterrnal boundary is a boundary on the external surface of the model. Air, or any other material domain that gets meshed and has FE fields computed in it, is part of the internal part of the problem. Although there may be an easier way to proceed, what I normally do if I want to have an "external" port but at the same time I want air or other materials very near/around it, I add (actually, I subtract from the local volume) a small volume that is right up against the port, thus creating a truly "external" surface on the actual port. (This little volume is not to be meshed, and not part of the computational space.)

Please login with a confirmed email address before reporting spam

Posted: 5 years ago 16.04.2019, 10:34 GMT-4

Hi Robert,

Thank you very much for your response. I understand better now the difference between 'internal' and 'external' port. I checked your method for low frequencies and it works now. Do you have any recommendations on how to speed up Comsol Simulations, apart from defining the PML as "layers" and reducing the degrees of freedom (DOF). I need to run my model on high frequencies but it runs for days and ends with convergence problems. I thought that the simulation will be faster by simulating the quarter of the geometry but it wasn't not enough (I use the FGMRES solver).

Thank you for your help, Emna.

Hi Robert, Thank you very much for your response. I understand better now the difference between 'internal' and 'external' port. I checked your method for low frequencies and it works now. Do you have any recommendations on how to speed up Comsol Simulations, apart from defining the PML as "layers" and reducing the degrees of freedom (DOF). I need to run my model on high frequencies but it runs for days and ends with convergence problems. I thought that the simulation will be faster by simulating the quarter of the geometry but it wasn't not enough (I use the FGMRES solver). Thank you for your help, Emna.

Robert Koslover Certified Consultant

Please login with a confirmed email address before reporting spam

Posted: 5 years ago 17.04.2019, 14:41 GMT-4

Well, here are some free suggestions, with no promises made or implied: Consider using Linear elements. (The default is Quadratic. Look for this under "Discretization.") This can dramatically speed up computations and reduce memory use. But it may also result in reduced accuracy, so be careful. Consider using the PARDISO direct solver instead of the FGMRES indirect solver. If you find that you simply must use an iterative solver due to memory limitations, consider using GMRES with the SOR vector pre-conditioner. It is sometimes slow, but usually converges reliably. Take a look at your mesh quality. If you can remesh or re-arrange your geometry somewhat to achieve or maintain a relatively high mesh quality, then your iterative solvers will converge better/faster and/or in fewer steps. (High mesh quality is especially important if you use the BICGStab iterative solver. In fact, just to get BICGStab to converge at all, you may find that you need to use quadratic-order elements.) You may not need PMLs to give you the accuracy you need, depending on the type of problem you are doing. Consider replacing those PMLs with simple surfaces specified as "scattering boundary" conditions. If you do that, you may find that choosing "second order" works better than the default "first order."

-------------------
Scientific Applications & Research Associates (SARA) Inc.
www.comsol.com/partners-consultants/certified-consultants/sara
Well, here are some free suggestions, with *no promises* made or implied: Consider using *Linear* elements. (The default is Quadratic. Look for this under "Discretization.") This can *dramatically* speed up computations and reduce memory use. But it may also result in reduced accuracy, so be careful. Consider using the PARDISO direct solver instead of the FGMRES indirect solver. If you find that you simply must use an iterative solver due to memory limitations, consider using GMRES with the SOR vector pre-conditioner. It is sometimes slow, but usually converges reliably. Take a look at your mesh quality. If you can remesh or re-arrange your geometry somewhat to achieve or maintain a relatively high mesh quality, then your iterative solvers will converge better/faster and/or in fewer steps. (High mesh quality is especially important if you use the BICGStab iterative solver. In fact, just to get BICGStab to converge at all, you may find that you need to use quadratic-order elements.) You may not need PMLs to give you the accuracy you need, depending on the type of problem you are doing. Consider replacing those PMLs with simple surfaces specified as "scattering boundary" conditions. If you do that, you *may* find that choosing "second order" works better than the default "first order."

Please login with a confirmed email address before reporting spam

Posted: 5 years ago 25.04.2019, 08:50 GMT-4

Hi Robert, Your suggestions were very helpful. Thank you so much for your help. Emna.

Hi Robert, Your suggestions were very helpful. Thank you so much for your help. Emna.

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.