Learning Forum

[AndyJP]

The most regrettable limitation in HFSS Transient is inability to simulate ferrites. It is kind of inconsistency, because HFSS is considered the main tool in designing ferrite RF components.

 

Will ANSYS ever try implementing ferrite transient simulation, or this topic is banned forever because of dusty 1989 creator's decision?

 

...well, I know that magnetization anisotropy is normally treated in frequency domain. But there's also a split-step method which perfectly models not just anisotropy, but a nonlinearity as well. It may be heavily ineffective, but should be easily parallelizable using GPGPU/Xeon Phy. There should be other approaches developed in 60 years. I've seen papers on colculating gyrotropy with FIT, Galerkin, and FDTD methods.

[Charlotte Blair]

 There are ways to analyze ferrites with HFSS and depending on your needs there may be several options for you. You can always couple Maxwell with HFSS or create a named expression for the permeability or import the permeability with a dataset and of course write a script that models the anisotrophy for use in HFSS.

[AndyJP]

You are probably talking about different HFSS, not the one, sold by Ansys in Electronics Desktop 2020 bundle. In Ansys HFSS Transient it is not possible. Maxwell is a static solver not relying on anisotropic Polder's tensor and not giving out any high frequency solution. Nvidia along with Ansys released several Fraudlent presentations and papers where ferrite circulator design example coexist with claims of many times GPGPU acceleration. It is clearly a fraud built up for boosting sales of gpgpus. (Few months ago nVidia was sued by investors for another fraud in annual analytics with similar false claims) In fact ferrite rf solution is not accelerated by gpgpu in driven modal, the only solver by ansys capable of treating errite anisotropy. Gpgpu acceleration was implemented in the transient solver about 5 years ago. But it appears, it is not capable of solving magnetic anisotropy problems. Which is itself quite amazing - more than 30 years an billions of dollars has passed, but general algorithms were not improved a bit. Just the code was optimized for newer processors, thats all.

 I am kind of angry, because I followed these publications and promos and purchased these nVidia Quadro cards, which are useless in fact!  Extra licenses would be more beneficial.