Discussion Forum

I haven't looked at that particular model, but I can tell you that when you are changing the details of how you define the ports, you are changing the RF behavior there (that is, the detailed field distributions, at the ports). Boundary conditions matter. The accuracy of your model depends upon you specifying the boundary conditions in a manner that represents reality. Lumped ports are circuit-like, and thus are only circuit-like approximations to real-world RF field distributions. Some lumped-port approximations are better than others. If you are not confident that your lumped port representation is a good one, consider specifying your port fields via a more accurate method.

Hello Robert,

Thank you so much for your replay!

Since my strucutre is a little bit complex so It would be better for me to calculate the S21 with some well developed comercial software, and I noticed that COMSOL has been used to do lots of S parameters' calcualtions, and I also researched that the multielement uniform ports type is the most suitable one for my strucutre. However, I always get positive S21 (The S21 is a positive value at low frequency and gradually decreases to be negative), I have attached a results plot.

Also I read the S parameter calcualtion section in the COMSOL RF User manual (see attached PDF). The S paramter calcuation equations in the manual look strange to me, it is not the typical definition of the S parameters.

Do you have any comments?

Thanks again!

Best

Ke

The page you attach is not a definition of the S parameters, but it does describe how Comsol calculates them.

Briefly: Comsol launches a wave of the type you select (say, TE10) and assumes a reflected wave of the same type with an unknown amplitude and phase (that's S11). Same thing on the output port except there is only an outgoing wave. Then Comsol obtains a solution for the electric field everywhere that satisfies the boundary conditions AND has incoming and outgoing TE10 waves. When it obtains that solution it gets the "best" values of S11 and S21.

This works fine UNLESS there is mode conversion and the true, physically correct solution has outgoing waves of a different mode. When that is the case your solution is not accurate.

To solve correctly, you must add ports for all the potentially significant outgoing modes. (Yes, you can have more than one port at the same physical location, with different mode types).

Additional modes can be significant even if your input waveguide is cut off for those modes, because cutoff only means that the mode decays exponentially, not that is has zero value everywhere. This can, by the way, lead to a dependence of S11,S12 on the length of the input and output waveguides. Now if you make the length long enough, then the magnitude of the higher mode will approach zero at the ports, and the "extra" modes will be unimportant.

Thank you so much! Dave,

The strucutre I am simulating is a simple two ports coplannar waveguide (quite similar as the example "Coplanar Waveguide Bandpass Filter" see https://www.comsol.com/model/coplanar-waveguide-bandpass-filter-12099). RF Wave is only excited in port #1. As you can see my S21 (in dB) data it is positive at samll drive frequencies. Do you know what could be the reason?

S21 (in dB) should always be negative according to its definition.

Also I cannot find how to set the modes type of the port in the COMSOL manual, do you have any instructional document ahout the setting?

Best

Ke

If you are using lumped ports then the page you linked is not relevant. With lumped ports Comsol will calculate S parameters from the port voltages and currents.

I suggest either (1) simulating a length of uniform transmission line, to be sure you get the expected results (S11 = small, S21 =1) and/or (2) posting your model.

Many postings here result in no useful advice because without a model to look at it is only possible to guess.

Hi Dave,

Thank you for your suggestion. Unfortinately, I am not allowed to share the simulation because of the confidentiality.

Best

Ke