HFSS simulation out of memory

EDAboard.com Forum Index -> RF, Microwave, Antennas and Optics -> HFSS simulation out of memory

Author

Message

ptshao

Joined: 30 Jan 2007
Posts: 19

18 Sep 2008 21:10 hfss 11.1.1

-------------------------------------------------------------------------------
Hello, guys:
I got some problems in my hfss simulation. My case is a very thin film antenna, around 1um in thickness. I know the large ratio of length/width to thickness is bad for meshing and simulation, but I have to. Every run ends with errors as followed. Now I attach my project here. Could you please try to run it and comment?
Thank you very much in advance!

MeanderLineAntenna1
HFSSDesign1 (DrivenModal)
[error] Out of memory (5:41 PM Sep 18, 2008)
[error] Simulation completed with execution error on server: Local Machine. (5:41 PM Sep 18, 2008)

Sorry, but you need login in to view this attachment

toonafishy

Joined: 18 Jun 2002
Posts: 461
Helped: 25
Location: Earth

19 Sep 2008 1:04 hfss airbox restriction

It converges on the 5th pass and simulates on my computer in 23 minutes. 58459 tetrahedra, 1.95GB ram used.
Maybe you don't have enough RAM or your page file is set too small.

jyfx

Joined: 07 Jun 2004
Posts: 26
Helped: 1

19 Sep 2008 4:56 simulation out of memory

let me see

Google
AdSense
Google Adsense

19 Sep 2008 4:56 Ads

ptshao

Joined: 30 Jan 2007
Posts: 19

19 Sep 2008 10:44 how to choose size of pml in hfss

Dear toonafishy：
Firstly thanks for your warmly help. I got a different result. It runs 40 minutes, and 165858 tetrahetra and then end with out of memory. You can refer to the profile I exported from HFSS. The passes and convergence plot is availbe as well in the attached. I have no idea why there are a difference between yours and mine. Welcome professionals to comment.
Thanks to all of you!
BTW: 4 Gb RAM is in my computer, but only 3.24GHz can be seen in property of my computer.

Sorry, but you need login in to view this attachment

Ivan_

Joined: 16 Nov 2007
Posts: 111
Helped: 21
Location: ITALY

19 Sep 2008 15:01 hfss radiation pml site:www.edaboard.com

Hi,
I got ur email and I think u may try to reduce required resources just by halving the structure and simulate one half. Due to the image theorem, u may try to use symmetry, since ur structure is perfectly symmetrical. In order to do that, select all objects (ctrl+A) and the right-click on the antenna-->boolean-->split-->yz-->keep negative fragment (or positive, it's the same). Now u have halved the structure. Then, apply radiation conditions to all the faces of the airbox except the one on the symmetry plane and apply radiation condition. On the face lying on the symmetry plane, u must apply a symmetry condition. The choice of the proper symmetry depends on the electric current direction and versus at the lumped port.
In such a case, the choice is E symmetry (see image theorem for more details). The 3rd thing to do is to set the impedance multiplier. In such a case, with only E-type symmetries, choose a imped. multipl. equal to 2. In this way, you reduce the resources required by almost one half.
Try the attached file.
The second possibility could be the reduction of simulation time by changing the metal (which is a solid) with a sheet. Good conductors imply high meshing. Then, apply a finite conductivity boundary on that sheet and choose the desired (emulated) thickness and conductivity. This reduces mesh in good conductors.

Let me know.

I.

Sorry, but you need login in to view this attachment

toonafishy

Joined: 18 Jun 2002
Posts: 461
Helped: 25
Location: Earth

20 Sep 2008 2:56 decreasing number of tetrahedra in hfss

You would expect the same result. I am using HFSS 11.1.1. What version are you using? It looks like if you moved your target to a little bigger than 0.02, it would converge sooner and maybe avoid the memory problem. I know the higher the number of passes to converge, the more tetrahedra and the more memory. Your computer has the same amount of memory as mine. On my project I am doing, if I make the air box smaller, less memory is needed. You can try that. I'm just learning HFSS so I can't give expert opinion.

akedar

Joined: 27 Jul 2004
Posts: 295
Helped: 15

20 Sep 2008 4:52 site:www.edaboard.com hfss memory error

use PML boundaries. Will reduce ur problem size

ptshao

Joined: 30 Jan 2007
Posts: 19

20 Sep 2008 7:07 Re: HFSS simulation out of memory

Hello, guys:
I would like to say thanks all of you, specially to Ivan_ and toonafishy. I have solved the problem with the help of Ivan_. And I think toonafishy's suggestions are also correct. Now I think the conclusions to reduce the memory required include:
1. Use the symmetrical plane boundary if your structure is sysmmetrical.
2, Config thin filmconductor by the altenative sheet with finite conductivity boundary.
3, Choose a litter bigger convergence restriction.
4, Using PML to substitute radiation boundary.
I have not poven the 4th approach. But I think if you have the similar question, you can try one or all of these methods.
For reducing airbox to decrease the memory consumed, I can't say. I choose the minimun required distance from the radiation component is a quarter wavelength. I have no idear if I keep slander the size of the air box. Hope someone can comment this point.
Thanks again for all of you!