3D simulatoring

I quite don't get the yuyu argument about HFSS being not accurate on radiation problems. It is just absolutely contrary - YuYu I've tested hfss on almost any kind antennas (but reflector) and the one that matches nearly everything was hfss. How can you say it is not good for antennas while the prime strength of hfss is namely in the precise antenna simulations. I really don't get it. And its post processing maybe lacking the polish of MWS but is nontheless rock solid and offers better functionality. In antenna problems, what does matter is the proper sizing of B.C but to say that hfss is no good for antennas means to twist truth.
Next - I disagree that none of the SW but MWS on the above list is 3D. Simply it isn't true for some real cases 3D task in stratified media can be tons faster run with "2.5D" tool. Let's not get into useless discussion on what is a real 3D tool. Let's make it opposite - what is really 3D task to solve - which is much more correct. Else we are wasting efforts and time.

cheers,

Can you imagine a big dishy antenna being simulated with any of the above mentioned tools? - No, of course, that's where Ticra and the like come into being. So, there are different applications that call for different tools and not the other way around.

Hi Cheng,

The problem is not only the accuracy!
HFSS is not suitable to solve radiation problem; the reason is simple. HFSS is a FEM code and then it is necessary to have a lot of memory and time to solve an antenna problem. In fact the numerical effort (memory and time) raises not linearly with problem size (instead of FDTD code).
Very often you can't simulate a radiation problem; only for simple radiation structure it is possible to have some results.
Further, FEM codes have some instability that generates spurios frequencies and the fast frequency sweep (a HFSS feature) causes other spurios frequencies. FEM was born in civil engineering and then in electrical engineering; in antenna engineering it shows his limits.
Regarding 3D SW, IE3D is not a true 3D simulator; that is the truth.

Regards

Yuyu,

Bit disappointed for you are talking things "in general". Let us be more specific:
The truth of the matter is that indeed FEM is proportional to N^2 in terms of memory but it matrices are all sparce and thus easy to handle. A lot of algorithms have been developed to handle Band matrices.
Contrary - all MOM are having rather ill-conditioned matrices, full, and difficult to precondition to say the least.
With FDTD we have some error by definition for the Yee algorithm is a leap-frog one thus having E- and H- fields misaligned both space and time

Next, the PML in HFSS is non-isotropic, thus the user can approach the structure from very short distance - much less for instance the one that MWS is using. (Again, I am not against MWS in here). Thus I do not see what increase in the simulation voluma we are talking in here if I can punch my PML at 1/8 lambda afar?! So, to say that HFSS is no ggod for antennas only is like I said - totally misleading. It does not really matter (it does, but a bit) if you simulate close structure or non-homogeneous one or "slightly bigger" antenna box. Indeed the solver must besh the entire volume, plus the PML but it is not too difficult. So, I would not separate antenna from other "similar in electrical size" structures.
Next to it - Your argument about the civil engineering is totally baseless. - FEM is a general Numerical method to solve PDE and as such is applicable to various problems all over. To the question about spurios solutions - I will grant 1000 bucks to anyone that ever had a spurios modal eigenmode calculation. The 2D port solver of HFSS is very robust and I know not a single case where the eigenvalues are mistaken by "spurious" solutons. The rest is pretty classical - vector elements and matching. To scracth a method for it is used in Mechanics is a nonsense (excuse my language). What is the damn difference between FDTD and FEM - FEM is variational formulation due to the second order decoupled Maxwell equations and FDTD is just FD of 2 coupled first order space-time PDE equations. Simple as that.
Again, a lot can be said about antenna modelling with HFSS that is beyond that post but to me there is "real" 3D "suitable" problem and not strictly 3D suitable software. Else we would not have a handful of excellent EM tools on the market.
To give you a token - why would one use the "full" 3D MWS or HFSS for MEMS switche modelling while I can do it at the fraction of time with EM3DS, which some folks can classify as 2.5D?! Isn't MEMS switch a "real" 3D device?

cheers

In terms of simulating the radiation condition, FEM is at least as good as the FDTD method. FDTD obviously scales bettter then FEM. I think the treatement of irregular shape is still the main difficulty in FDTD.

Who can tell us more about the Perfect Boundary Approximation in MWStudio?

Hi,

The ANSYS software is the best tools based on FEM simulation.
and can solve your problems.

Mems123

Hi Cheng,

you wrote

"The truth of the matter is that indeed FEM is proportional to N^2 in terms of memory but it matrices are all sparce and thus easy to handle".
The problem is that; FEM is proportionale to N^2 and often for many radiation models it means that you can't simulate your system!
Probably you are not an antenna designer and you don't know well this kind of problem.
To the question about spurios solutions the problem is related to 3D solution not 2d eigenmode!

3D or 2.5D software is not useful discussion.

Regards.

Hi Cheng,

you wrote: "Your argument about the civil engineering is totally baseless. - FEM is a general Numerical method to solve PDE and as such is applicable to various problems all over".

Since FEM is a general numerical method to solve PDE then is applicable to various problems; but, when you try to solve a radiation problem, for example an array antenna with dimension of several lambda, you can see that this is only an academic argument.
When the problem dimension is several lambda FEM is not applicable; in fact, if you see the ansoft web site, you can download an example on antenna arrays and you can see that the model was solved with a SUN with 8 processors and 6 Gbytes of RAM!
The model was solved by ansoft application engineers not yuyu.

Regards

Yuyu,

Don't wanna be impolite but for 14 years I am designing antennas and do know something on the matter. If one needs only a sw to design antennas he'd better change jobs I reckon:). On your question about the few lambda size radiation problem - you again prove yourself wrong - like I said the problem is not in the radiation or closed boundary, problem is related to the overall size of the problem. FEM scales less favorable than fdtd for instance but the matrix is much easier to handle.
Second, I don't see problem that you need big memory and a number of crunchers to slove your antenna problem - I have never seen a serious antenna designer working on a Pentium III, 256 RAM, have you? The example you took by Ansoft is a very old one, made with version 6 if I remember and today we are talking about version 9, that offers a multifold of improvement in both speed and memory handling. But you somehow miss that point.
Dont'know if you managed to see that MWS did indeed implement an option for a multiple processors - Up to 20 if my memory is fine. Why do you think they'd that?
Have you seen the website of an american company dealing with fractal stuff? They built a cluster of 20 Workstation to deal with their problem and they would not complain, nor did their customers.
I am afraid you are changinga focus of our topic in here - we are not talking whether one needs that machine or others - we are talking if one could use one tool or another to solve problems. Like I said - let's make it better use for ppl that might need that information and not throwing stones for problems of marginal chance to happen or a matter of logistics.

cheers

topic closed

Hi Cheng

I think that is necessary to close this discussion.

Anyway:

1) The radiation models are related often with big dimensions ; I'm talking about complex antennas not antenna for cellular phones!

2) Your consideration about matrices and FDTD are wrong!

3) Seriuos antenna engineer can work also with a pen!

4)I can show two recent examples took by Ansoft website (June 2002, version 8.5, v.9 is not yet available, only beta);

Example 1 - Optical Ring Resonator Coupler, Sun with 8 processors, 5 h and 45 minutes with 8.4 GBytes of RAM!

Example 2 - 12 Arm Power Divider, 7.6 Gbytes of RAM !

Yes, you can have many problems even with closed structures!


5) Further, sometimes HFSS add spurios frequencies and if you have to consider higher order modes that is a serious drawback.


Regards.

Hi all,

I read this thread with very much pleasure ...:)

Thanks you all for sharing your opinion.

Still wait for more post ...
Which one is the best to learn and invest for the future ? ...
Please go on ... and don't stop ...

Can someone out there give me a hint,
which FULLWAVE Software I can use to
design a Relfectarray of 1 square meter, with thousand 37GHz-resonance patches
in its ...

Can Feko do this ?
How about IE3D ?
and CSTW ?

Please, ...I just a newbie ... since
there many EM-CAD softwares ...
learning and testing them all ...:( :(... it
takes too much time ...
I just want to learn 1 or 2 softwares for my coming MSC-project ...
Later when I have time ... sure I will test all of them...

But not at this time .. Student don't much time now ..You know that ..:) :)

Thanks and waiting for your reply,
Thanks in advance ...
Dany

Hi,

I want to add some considerations about CST MWStudio and FDTD method.
MWStudio is based on FIT (Finite Integral Technique) that is applied to Cartesian grids and combined in time domain with the leapfrog scheme; FIT is applied to the integral form of the Maxwell equations and is computationally equivalent to FDTD (FDTD is applied to the differential form).
Thus, FIT shares all of FDTD advantages regarding memory needs. Further, FDTD and FIT, being a fully explicit computation, avoid the diffuculties with linear algebra (matrix solution) that currently limit the size of finit element models.

Regards.

Thanks Yuyu, ...:)

After reading your reply (advice) ...I'm planning
to go for FIT ... This is rather new .. Am I right ?

My further silly questions are :
1) Is there any textbook about FIT...?
I want to learn a little about this theory and technique.

2) Can any Guru in here provide me some tutorial materials
for this methods....

Thanks for anyone who can help ...

TIA and Have a nice day ...

Hi Dderijken,

1) there are papers about FIT by Tomas Weiland (the guru of CST). Anyway, since FIT is FDTD applied to the integral form of Maxwell equations (more suitable to develop conformal version of FDTD), you can try to read a good textbook on FDTD (for example A.Taflove-S.Hagness,
Computational Electrodynamics, the finite difference time domain method, Artech House, 2000). In the textbook there is a section (3.6.8, pp. 93-96) where is discussed an other interpretation of FDTD method starting with a more macroscopic (but still local) combined field description based upon Ampere's Law and Faraday's Law in integral form; in practice, that is FIT.
2) the textbook or 1000 papers on FDTD (try to internet ...)

Regards.

Hi Yuyu, ....

Hehe ...Thanks a bunch ..buddy ..
I will try to find those literature ...

I like the FIT ...
I't seems to me that Integral is better than Differential ...?
(just my 2 cents ...:) ...:) ..)

Have a nice day
Dany,

I think the HFSS is good.

Firstly,The hfss;secondly,The Mwo;I think

MWO is not a 3D field solver!
Regards

Best field solver? My choice goes for SuperSapar roller. It can even dry the ground in the rainy days.:)

Oh IC you are talking about 3D EM fields. Then use Maxwell or Ansys for simulate your transformer or use XFDTD/Fidelity to check the presense of EM inside your head. Check EM inside the klystron or magnetron or waveguide with HFSS. Dont forget to see the near field components of helical antenna using NEC. Actually what I wanted to say is the best choice depends upon the object and also the observer. If you want to see the 3D EM of a power divider inside a metal case you can try Sonnet too with using brick substrate same thing may be possible by MWO but not sure(I used it long back). If you wright your own code, must follow FEMLAB codes then comback to MATLAB use PDE lybraries.

Actually RF is [Really Funny]

cst is a very good 3d em solver, it is based on fit. more information can be found at www.cst.de including a list of publications. this software is developed at the university of darmstadt in germany.

another good 3d solver is wipl-d based on 3d method of moments. more information at www.wipl-d.com

greets

i think HFSS is a good choice.