New HFSS FE-BI as compared to PML and ABC, Thoughts

[dafioram]

Even though HFSS is on version 14 as of version 13 they introduced another boundary condition in addition to ABC and PML for normal simulations. The new boundary type is called Hybrid Finite Element Boundary Integral Method. An easy way to implement it is to take a model with an ABC boundary and go to the options for the radiation boundary and check the "Model exterior as HFSS-IE domain".

This will allow the solver to solve the currents along the boundary of the absorbing surface, which in effect lowers computation time, similar to PML FE-BI method is best placed lambda/10 away from the structure, see the attached document from ansoft for uses.

My question is has anyone tried this? The results I got were encouraging it also converged in a fairly decent amount of time. The results I get are slightly different from what I get for ABC, which is discerning since most of model are ABC, but I've become doubtful of ABC and PML because of the problems raised in the paper that advertises FE-BI (attached).

 

[kae_jolie]

hmmm...i was just reading about HFSS-IE and FE-BI and ran into your post. So, all we have to do is check the "Model exterior as HFSS-IE domain" and set boundary box to lambda/10 and we should have a Hybrid FEM/IE solution? Is there anything else we need to do?

Also, how about just simulating using HFSS-IE alone. I am looking for a way to move my design, reports, excitation source, etc from HFSS to HFSS-IE and run the simulation in HFSS-IE. Do you know how we can do this? If so, please be as detailed as possible. I read about a tool called data link, but I can't find much resources on it.

Thank you.

 

[dafioram]

As long as the boundary has been created, you just right click and select "Model exterior as HFSS-IE domain", and it only works in HFSS (not in HFSS-IE). You don't have to set it lambda/10 away, but what i understand that is the minimum distance.

When I first started using HFSS I used IE only because I was limited in computer resources. There are some situations where you can drag all HFSS objects into an HFSS-IE project, but that isn't always the case. For example not all the same port types exist in both models (among other things), like there is no wave port in IE. You might like to know that if you would like to control the excitation of the ports then you can only control the total value on the line (like the total voltage), but in HFSS you can control both the incident amount or the total amount on the line, you can get around this limitation by writing an algorithm that uses the s parameters of your ports to determine what are the necessary total excitation values on the ports to achieve a certain incident value, but this can be trouble some. Additionally HFSS-IE is similar to HFSS-terminal in some ways in that the ports are excited with voltages, unlike in HFSS-modal where the ports are specified as powers. If you are trying to compare the "accuracy" between HFSS-IE and HFSS you can't just compare the delta s, since they solve the models completely different. In this instance I suggest observe the quantity you are focused on as a function of the pass number.

The way data link works is if you have a HFSS-IE project whose fields are already solved and you would like to have that same object in an HFSS model you can just import it into the HFSS model. There is a datalink example out there that solves a horns antenna pattern in IE and then imports it into a HFSS model that has a parabolic reflector (so as to have a parabolic antenna with a horn feed).

 

[drkirkby]

If I recall correctly, if you display your model, then select all items, then you can go Edit -> Copy (or control C), then insert an HFSS-IE design, and use Edit -> Paste (or Control V) in the empty HFSS-IE model. (Copy and Paste the window to the right - I forget what its called). You will then need to sort out the issues that remain - ports, and whatever else. I was shown by an Ansys engineer how to do this.

Dave