Continuation of 'Domain decomposition in HFSS'

  • Last Post 29 August 2018
asamaiyar posted this 08 August 2018

Thank you. That was very helpful.

But this leads to my second problem. I have a feed and reflector antenna system. And because the model is large I am putting feed and reflector in different FE-BI regions. In earlier versions of HFSS I was able to use Direct Solver (which I always prefer, based on my experience) to run this kind of model. But in v16+ the direct solver option is inactive and it is forcing me to use Domain Decomposition Method (which, in some cases, is not converging even after 200 iterations). 

Is there a way to run Direct Solver with FE-BI regions in HFSS?

Order By: Standard | Newest | Votes
haoxuanzhang posted this 17 August 2018

Thank you for your answer! And  I have some questions,what kind of model is suit for domain decomposition solver in HFSS? And HFSS's domain decomposition method refer to which paper?

dsurendr posted this 29 August 2018

In reference to this thread:

asamaiyar - Whenever you're running a hybrid solution (combination of FEM and MOM in your case), the domain solver will be used. To use Direct solver only, you must use only one solver time (all FEM for example --- in your case, change FEBI boundary to Radiation or PML).

dsurendr posted this 29 August 2018

In reference to this thread:  

haoxuanzhang - Information (including reference papers) for HFSS's domain decomposition method can be found in the Help documentation. Domain decomposition is used by default in all hybrid solve models. It's also used in our periodic boundary solutions when you Create Array from unit cell designs. It can also be used to distribute a model that can't be solved on one machine (due to RAM limitations).

asamaiyar posted this 29 August 2018

Hi dsurendr,

But in HFSS v15, Ansys allowed Direct Solver with FE-BI hybrid boundary. Is there a specific reason that this feature was discontinued?

The reason I am asking this is because I have a large model. And the domain decomposition solver is failing to converge after 200 iterations. I was recommended to use Direct Solver here. But I want to use FE-BI boundary because Radiation boundary is giving some reflections. If I use PML over the full model, the problem gets extremely large and not feasible to solve. Is it possible to bring back FEBI with direct solver (just like HFSS v15)?



slouie posted this 29 August 2018

Domain solver uses the direct solver method, but there are iterations between domains (FEM domain and IE domain) so it's not purely a direct solve. This has always been the case. Failure to converge after 200 iterations is usually indicative of a poor setup. Please refer to Antenna training to make sure you're following best practices in setting up your model: