Simulating THz in the RF module

Please login with a confirmed email address before reporting spam

Hi All,

I am currently trying to simulate a feed horn at terahertz frequencies and eventually encorporate the feedhorn design in to a cassegrain antenna. My current issue is that I will run out of memory either at the meshing stage or during the computational stage. Since my design is not cylindrically simetrical, I am restricted to a 3D environment. The design is 4-fold symetric so I am able to take advantage of applying PEC and PMC boundary conditions, but the load is still to large for the memory available. Is there something else I could try to overcome this issue?


4 Replies Last Post Apr 12, 2024, 9:04 a.m. EDT
Robert Koslover Certified Consultant

Please login with a confirmed email address before reporting spam

Posted: 2 months ago Mar 8, 2024, 10:38 a.m. EST
Updated: 2 months ago Mar 8, 2024, 10:34 a.m. EST

Well, if you haven't already: 1. Carefully adjust your mesh to include no more detail than necessary. 2. Restrict your computational space to no larger than necessary. 3. Use linear discretization. 4. Use the GMRES iterative solver with SOR vector preconditioner. (Slow, but memory efficient and generally gets the job done, eventually.) However, none of the above is likely to be adequate if your dish antenna is sufficiently large (e.g., 40 dB or more of gain), especially if you don't have hundreds of GB of RAM and up to days of computational time available, or need to run many frequencies. In that case, you may need to employ an alternative or hybrid computational approach. If you can share more about your problem, you may be able to get more specific advice here.

-------------------
Scientific Applications & Research Associates (SARA) Inc.
www.comsol.com/partners-consultants/certified-consultants/sara
Well, if you haven't already: 1. Carefully adjust your mesh to include no more detail than necessary. 2. Restrict your computational space to no larger than necessary. 3. Use linear discretization. 4. Use the GMRES iterative solver with SOR vector preconditioner. (Slow, but memory efficient and generally gets the job done, eventually.) However, none of the above is likely to be adequate if your dish antenna is sufficiently large (e.g., 40 dB or more of gain), especially if you don't have hundreds of GB of RAM and up to days of computational time available, or need to run many frequencies. In that case, you may need to employ an alternative or hybrid computational approach. If you can share more about your problem, you may be able to get more specific advice here.

Please login with a confirmed email address before reporting spam

Posted: 2 months ago Mar 11, 2024, 2:56 p.m. EDT
Updated: 2 months ago Mar 11, 2024, 2:56 p.m. EDT

Hi Robert,

Attached is my current design of a diagonal horn (no dish included in the simulation).

Best, Daniel Heligman

Hi Robert, Attached is my current design of a diagonal horn (no dish included in the simulation). Best, Daniel Heligman


Please login with a confirmed email address before reporting spam

Posted: 2 months ago Mar 11, 2024, 3:15 p.m. EDT
Updated: 2 months ago Mar 11, 2024, 3:10 p.m. EDT

I also, do not seem to have the GMRES solver as an option. Only MUMPS, PARDISO, SPOOLES, and Dense matrix.

Best, Daniel Heligman

I also, do not seem to have the GMRES solver as an option. Only MUMPS, PARDISO, SPOOLES, and Dense matrix. Best, Daniel Heligman

Robert Koslover Certified Consultant

Please login with a confirmed email address before reporting spam

Posted: 2 weeks ago Apr 12, 2024, 9:04 a.m. EDT
Updated: 2 weeks ago Apr 12, 2024, 9:04 a.m. EDT

Use a manually-configured mesh. In the Model Builder, right-click Mesh, then Free Tetrahedral. Right click Free Tetrahedral 1, then size. In the Settings pane, choose specific geometric entities and specify sizes appropriate to the calculation. You can set sizes in domains, on boundaries, edges, and points, with separate size entries. You will also probably want to define a swept mesh for your PML layer. The simple Port 1 setting of TE10 isn't going to work with your PMC boundary. Instead, set that port so as to provide fields appropriate to 1/4th of the (full, virtual) physical waveguide there. Unfortunately, setting the PMC and PEC boundaries doesn't do this automatically. There are other ways to configure ports, so I encourage you to think about it. To set the solver to GMRES (if needed), in the Model Builder, expand Study --> Solver Configurations --> Solution 1 --> Stationary Solver 1, then right click Stationary Solver 1 and choose Iterative. Then you can choose among various solvers in the Settings pane.

-------------------
Scientific Applications & Research Associates (SARA) Inc.
www.comsol.com/partners-consultants/certified-consultants/sara
Use a manually-configured mesh. In the Model Builder, right-click Mesh, then Free Tetrahedral. Right click Free Tetrahedral 1, then size. In the Settings pane, choose specific geometric entities and specify sizes appropriate to the calculation. You can set sizes in domains, on boundaries, edges, and points, with separate size entries. You will also probably want to define a *swept* mesh for your PML layer. The simple Port 1 setting of TE10 isn't going to work with your PMC boundary. Instead, set that port so as to provide fields appropriate to 1/4th of the (full, virtual) physical waveguide there. Unfortunately, setting the PMC and PEC boundaries doesn't do this automatically. There are other ways to configure ports, so I encourage you to think about it. To set the solver to GMRES (if needed), in the Model Builder, expand Study --> Solver Configurations --> Solution 1 --> Stationary Solver 1, then right click Stationary Solver 1 and choose Iterative. Then you can choose among various solvers in the Settings pane.

Reply

Please read the discussion forum rules before posting.

Please log in to post a reply.

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.