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.
2D boundary mode analysis takes ages
Posted 28 juin 2021, 05:13 UTC−4 Wave Optics Version 5.5 2 Replies
Please login with a confirmed email address before reporting spam
Hi All,
I'm doing simulations on so called Echelle gratings using frequency domain analysis with the wave optics module on Comsol 5.5. These things are optical (de)multiplexers, rather large and therefore I can just handle them in 2D simulations. Also these structures have some ports and to run the simulations, a boundary mode analysis is performed for each port. The ports are defined as numerical ports. and I'm just looking for the very first mode. Now I noticed that the boundary mode analysis takes the longer the larger the structure is. The whole structure, the ports always have the same shape! On my current structure (Okay, it's really large) Comsol is calculating for more than a day on a single port (Number of degrees of freedom to solve for: 97). And the structure has 65 ports!. I thought that the boundary mode analysis only looks at, well, the boundary. Why does it take so long? And why does it take longer for larger structures, even if the port is the same and rather small?
Please login with a confirmed email address before reporting spam
Hi MarcS,
The boundary mode analysis should only consider the boundaries, so the behavior you describe is strange. Are you sure that it is the boundary mode analysis that takes time and not the meshing part, that precedes boundary mode analysis, that takes so long time?
If possible, I would recommend that you contact our support with your model. Then they will be able to give you more specific advice about the problem you experience.
Best regards,
Ulf
Please login with a confirmed email address before reporting spam
Yes, I'm sure. After meshing the whole model with the intended final mesh size, it took extremely long for even a single port. Now I circumvented (hopefully) the problem by meshing the unbimportant parts of the model (for the boundary mode analysis) very very coarse, do the BMA and then use the finer mesh for doing the final simulation in the frequency domain. I hope that the boundary modes are copied to the final mesh with an appropriate accuracy... By the way, meshing took 17 hours now, but calculating one BMA five minutes. Still quite a long time, but well...