Advantages/Disavantages of HFSS and Q3D extractor for EM simulation

[RFaddict]

What are the advantages and Disadvantages of those 2 Tools developped by Ansys (HFSS and Q3D extractor)?


Thanks!

 

[reidintransit]

Compared to what? The adavantage is they are much faster than calculating with an abacus. The disadvantage is that they cost more than an abacus. They are tools. It's like asking what are the advantages and disadvatages of shovels and hammers. Are you asking about when to use one or the other?

 

[RFaddict]

Compared to what? The adavantage is they are much faster than calculating with an abacus. The disadvantage is that they cost more than an abacus. They are tools. It's like asking what are the advantages and disadvatages of shovels and hammers. Are you asking about when to use one or the other?

Yeah i have to say that my question was not enough clear. Yes i mean when to use one or the other one? The answer can be a kind of comparison of those both tools for a specific application like "micrometers TLines EM simulations.."

 

[SI_Product_Manager]

HFSS is full-wave 3D which means it solves all of Maxwell's Equations. The limit on accuracy is basically dependent on the mesh size. HFSS is very good at extracting s-parameters and fields.

Q3D is quasi-static 3D which means it dosesn't solve the time dependence piece of Maxwell's equations. Essentially you want the EM wave to have a wavelength of about 8-10x larger than the structure you are simulating. This essentially will ensure the quasi-static approximation holds true. Q3D is very good at extracting R, L, C, and G parasitics.

Both HFSS and Q3D use automated adaptive meshing to converge to the desired accuracy of the solution. Speed varies depending on the structure and the usage of compute resources. For instance HFSS can distribute frequency points over a cluster which will dramatically speed it up (say 10-20x) where as Q3D can distribute the CG, ACRL, and DCRL solvers over a cluster and then use MPI for CG frequency points. In general different q3d applications can see speedups ranging from 10x to 100x on a large cluster.

In the end the best tool depends on your application. You should answer the following: Do you want s-parameters or do you want RCLG parameters?

 

[RFaddict]

HFSS is full-wave 3D which means it solves all of Maxwell's Equations. The limit on accuracy is basically dependent on the mesh size. HFSS is very good at extracting s-parameters and fields.

Q3D is quasi-static 3D which means it dosesn't solve the time dependence piece of Maxwell's equations. Essentially you want the EM wave to have a wavelength of about 8-10x larger than the structure you are simulating. This essentially will ensure the quasi-static approximation holds true. Q3D is very good at extracting R, L, C, and G parasitics.

Both HFSS and Q3D use automated adaptive meshing to converge to the desired accuracy of the solution. Speed varies depending on the structure and the usage of compute resources. For instance HFSS can distribute frequency points over a cluster which will dramatically speed it up (say 10-20x) where as Q3D can distribute the CG, ACRL, and DCRL solvers over a cluster and then use MPI for CG frequency points. In general different q3d applications can see speedups ranging from 10x to 100x on a large cluster.

In the end the best tool depends on your application. You should answer the following: Do you want s-parameters or do you want RCLG parameters?

Thank you a lot ! Your post is very Helpfull.
By the way, I want to simulate a 3D capacitor.. I think Q3D extractor will be the best tool because I want to have the equivalent circuit " ESR - C- ESL" of the capacitor

 

[SI_Product_Manager]

Your welcome. Both HFSS and Q3D are used for capacitor design. I think you should be good to go. Don't forget if you decide HFSS is the way to go you can determine the C, L, and ESR by looking at the s-parameters in the end.

Simply you can fit the linear portions of the Touchstone file using the simple equation for inductive and capacitive reactances. Xl=2*pi*f*L and Xc=1/(2*pi*f*C).

 

[RFaddict]

Thank you a lot again.. I think that Ansys did not provide sufficient Tutorial/examples for Q3D extractor beginners. Do you have any example/tutorial of how to model a 3D system with Q3D? (system with a PCB layer, package and on chip interconnect...). ALso do you have the procedure steps for importing GDS file in q"d extractor?

 

[wavenology]

This question is huge, and will easily cause argument. I'd say, you have to focus on a specific problem. For example, a waveguide design or input impedance matching of a RF device or antenna. Unstructured solver is much better such as HFSS, IE3D which uses tetrahedron. In such problems, devices is usually less then 10 wavelengths and finite solvers such as FEM or FDTD is very good, if problem is electrically large, such as scattering of a battleship which geometry size is >> 10 wavelengths, then a integral solver or ray-tracing solver is better. For EMC/EMI problem, most of the time it is a system-level simulation, all these tools suffer from system-level simulation since there are too many things to model. And the computation cost is unrealistic.

So in summary, if you ask in general what is the advantage of a commercial solver, most of the professionals will directly ask you what is the problem you encountering? We can not simply compare apple with orange and say which one tastes better.

 

[SI_Product_Manager]

ANSYS has examples and tutorials. Please contact their technical support as I'm sure this would be ANSYS IP.