How to build scripted element of MMI?

Hi, I have some questions about compact model of mmi. Could you help to have a look at it?
For compact model of mode waveguide, the loss, neff, ng can be observed by experimental results, and thus replace part of the result extracted from mode simulation.

  1. For MMI, the S parameter can be used to make circuit model. Are there any part can be replaced by experimental result (like what in waveguide)?

  2. For MMI, is it better to choose S element or script element to build circuit model? Is there any example of scripted element for MMI?

Hi @1683324,

  1. No, there isn’t a built-in parameterized model of MMIs like there is for simple waveguides in Interconnect. If you wanted to, you could define one using Scripted elements if you have a model you want to use. Generally, when modelling devices like MMIs the S-parameters are extracted using a physics solver (like the FDTD or EME solvers, for example) and then these S-parameters are exported to Interconnect. To see how this is done, you can take a look at the MMI coupler example in the Application Gallery.

  2. Usually it is best to use the N port S-parameter element for simple passive devices like MMIs. The Scripted element is only used when more complicated functionality is required, or you have an analytic model for the S-parameters you want to use.

I hope that helps. Let me know if you have any questions.

Hi, kjohnson, thank you for your explanation.

  1. The perforimance of MMI (power imbalance, insertion loss)can be calculated from the result of S matrix. Does it mean the S matrix of MMI in mode solution already contain the performance information of MMI, although the compact model of MMI building by N port S-parameter element doesn’t show the performance of MMI Intuitively?

  2. Put aside these, if I have experimental results (power imbalance and insertion loss) of MMI, which kind of compact model should I choose to make the most of the experimental result?

  3. If I want to combine the simulation data with the experimental results to make a compact model, is it completely impossible? if so, how to do it?

  4. If not, does it mean that I can either choose to simply simulate the MMI by Mode solution and then build a compact model by N port S-parameter element, or choose to combine experimental data with analytical formulas and build them by scripted element?

I hope you could give me some suggesions.

1 - Yes, in a linear circuit model the performance of the components is entirely characterized by their S matrices. You can extract the power imbalance, insertion loss, splitting ratio etc. from the S matrix. For example, in a 1x2 MMI, the transmission into output 2 would be |S_{21}|^2, transmission into port 3 is |S_{31}|^2, reflection back into port 1 would be |S_{11}|^2, insertion loss would be 10\textrm{log}_{10}(|S_{21}|^2 + |S_{31}|^2), etc.

2 - I would still recommend using the N port S-parameter element. You can use the experimental results to obtain the S parameters. This is a bit tricky if your experimental results are based on measuring power (like insertion loss and power imbalance), which lack phase information (unless the results are from some sort of interference experiment, I suppose). In other words, power measurements can give the magnitudes of the S parameters, but not the real and imaginary parts.

You could assume that the S parameters are purely real, but you should keep in mind that this will not include the phase information, which could play a role in any interference or resonance effects. However, if you are using the MMI as a power splitter, for example, this method would be sufficient.

3, 4 - You could use the method I mentioned above and calculate the S parameters based on experimental measurements. You could also obtain the S parameters using simulation and then verify that the simulation results are reasonable by comparing them to the experimental results. The third option you mentioned, using a scripted element and an analytical model, is also possible. It is up to you how exactly you want to model the devices in your circuit.

Let me know if you have any questions.

Hi, kjohnson, thank you for your respond.

About question 1,3,4, I have a clearer comprehension.
There are still some doubts about queation 2.

Suppose there is a 1X2MMI and its power transmission versus frequency has been measured(Its unit is 1 mW). Then, the S parameter don’t incude phase information. Then, how to deal with the phase information in txt file (which are then used inN port S-parameter element) ?
to assign the value of phase to 0( the first column is frequency, the second column is |S| (for example |S21|),the third column is assigned 0 see MMI1by21310-s-params2.txt (7.5 KB) ), or to delete this column(the first column is frequency, the second column is |S| (for example |S21|), the third column is deleted, seeMMI1by21310-s-params1.txt (7.3 KB) , this method shows error when running interconnect )?
Is the first method right or there are other ways to deal with the phase information in txt file?

The first method is correct. If you want all of your S parameters to be real you should set the phase to zero for all of the parameters by putting 0 in the third column.

Hi, kjohnson, thank you for your respond. I got it.

This topic was automatically closed 3 days after the last reply. New replies are no longer allowed.