Different simulation results on Windows and Linux

MikeSMikeS Member Posts: 5

Hello,

I am running Lumerical 2021 R1.3 on two computers, one Windows (Windows 10 Pro) and one Linux ( CentOS Linux release 7.9.2009 (Core) ). In both cases I am using the varFDTD solver, but the end result is different on each OS.

The *.lms files were created on the windows computer and the file I am running on Linux is a copy of the original.

Has anyone else noticed something similar? Any ideas what to check so the results converge?

Thanks in advance!

Answers

  • kjohnsonkjohnson Posts: 137Ansys Employee

    Hi @MikeS ,

    Can you please give some more information on the simulation, and how the results differ? Remember that Ansys staff can't download files from the ALF (see this post for more information), so please put any screenshots directly into your post instead of attaching the image files.

  • MikeSMikeS Posts: 15Member

    Hello @kjohnson ,

    Thanks for the reply.


    I am not sure what kind of information would be useful. The parameters of the simulations are identical (geometry, structures, monitors, solver, etc...). The field values recorded by the power monitors are different. Their absolute difference is not important, however the mode profile is. Below are the images of the simulated geometry and the result of the top-view power monitor for Linux (upper images) and Windows (lower images). Since the profiles are different, every calculation I run on them gives different results, so I think they are not needed.


    All images are produced with Lumerical's native plotting capabilities on the same computer running Windows 10.

    I noticed that when I increase the simulation area towards the +y (upper y bound) in Linux, the results agree much better with the windows' outcome. If a more specific piece of information is required, let me know.

    Thanks a lot and cheers!

  • kjohnsonkjohnson Posts: 137Ansys Employee

    Hello @MikeS ,

    Thank you for providing this info. It looks like there is some scattering from the mode source in the Linux simulation that isn't present in the Windows simulation results. I would recommend that you plot the modes used by the mode source in each simulation to make sure that they are the same. It might be a good idea to check the slab mode of the varFDTD solver and the material fits as well before running the simulations again.

  • MikeSMikeS Posts: 15Member

    Thanks for the feedback @kjohnson, I'll check your suggestions.

Sign In or Register to comment.