Discussion Forum

Hi all

First of, thanks in advance for reading and hopefully reply to this thread.
The main problem is, that when I change a parameter in a 2D simulation, the results do not alter and that's how it's suppose to be. However, when I change the same parameter in a 3D model, all of a sudden the results jump randomly.

If you're interested in the problem, read the subsequent section. If not, you can just simulate the attached COMSOL file with different "margin_side" values.

I try to simulate the reflection and transmission of a mirror.
Here is how the Simulation works:
- An input and an output port are defined.
- The input port is excited with an electromagnetic wave
- The wave travels within a waveguide towards the mirror
- At the mirror the wave is then partially reflected and partially transmitted
- Thus, some of the incident wave reaches again the input and some the output

Here is how the Reflection and Transmission is calculated:
- The scattering parameters at the input and output port (S11 (reflection) & S21 (transmission)) are a measure of how much of the incident wave reaches the input and the output port
- In order to calculate the reflection of the mirror, the power reflected at the mirror is divided by the power reaching the mirror. This can be expressed in a formula like this:
P_in=P0*exp(2*imag(emw.beta_1)*z0)
P_ref=10^(S11dB/10)*exp(-2*imag(emw.beta_1)*z0)
P_ref / P_in = 10^(S11dB/10)*exp(-2*imag(emw.beta_1)*2*z0)
where P0 is the excitation power which was chosen to be 1W
z0 is the distance between the input port and the mirror. In the simulation z0 = "margin_side"
-The transmission is calculated likewise.

So now let's come to the problem:
In theory, when I change the distance between input port and mirror, neither the reflection nor the transmission should change. When I simulate the problem in 2D, the simulation works like expected (R & T do not alter).
In 3D however, changing z0 gives a drastic change in reflection as well as transmission. Additionally, the change is not monotonic but the R and T jump from value to value in a random manner (see attached pdf file).

Does anybody have a clue where this problem might come from? Has anybody had similar issues?
I don't think that meshing is the problem, because the problem still exists if the mesh is made finer. Furthermore, finding the wrong excitation mode at the input port can also be omitted, as the effective refractive index stays the same for all different distances (see attached pdf file).

Best Regards
Chris