2.5d simulator accuracy - some info

[loucy]

Can we characterize the effect of random surface by equivalent impedance? Where does that 100% error come from?

 

[yingyang]

2.5d accuracy

I have tried to simulate vivaldi antenna in IE3D, MWO, momentum, FEKO, HFSS, Empire, Ensemble and other EM softwares. All except HFSS are no good for finite substrate.

 

[filterman]

Re: 2.5d accuracy

I remmember that most of those software companies took part in a benchmarking of EM suites, the test was the vivaldi - most of them claimed victory including EMPIRE .
But as a matter of knowledge there are some classes of EM simulation tools and not all are the same:
1. Lateraly open box method -Mommentum,Ensemble and original IE3D - THOSE ARE SOFTWARE OPTIMIZED FOR printed circuits, excellent for printed antennas, multilayer. Has triangulat cells. so curved metal can be approximated very well , geting both good accuracy and speed.
2. Closed box mom method -Sonnet, MWO : very good for printed circuits in closed box. But usually good for very thin metal layer and the via hole trick to account for thickness is in certain uses a clumpsy approximation. Usually you have only rectungular meshing.
3. New IE3D has both types of algorithms in one software, including a special algorithm that allows triangular cells, but pays for it be a slower simulation.
All 3 are optimized for printed circuits work , some are 3D in metal(IE3D, SOME OF aNSOFT CODE).
But they are not FULL 3D ( which implies finite dielectric substrate or resonators).
When you need finite dielectric, then you will have to use FEM(HFSS) or
FTDT(CST,QW3D,EMPIRE,Fidelity,etc.).
Some antennas are developed , using one of the first 3 types (uaually IE3D or Ensemble) and finishing the project using FTDT or for small antennas(HFSS).

 

[loucy]

YingYang,
What is the size of your substrate? How is it compared with the size of the top metal?

 

[yingyang]

finite substrate

I am working on an aperture area of 265 mm^2 and have to pack as many vivaldi antenna. All the 2.5D simulator fails to see the 377 ohm as they assumed infinite substrate.

Even with HFSS, the result is also not very satisfactory but compared to the others, it is better. The HFSS fails to allow good description on the radiation pattern.

 

[sugm]

for antenna simulation, it's very important to define a right boundary.

 

[Element_115]

I don't know alot about all these software's, but I can say that
AnsoftHFSS is better(faster) that HPHFSS, and at my old job (2002)
the IC designers used AnsoftHFSS to create S-parameter files
to use in Cadence for their design. They said it worked well.
I designed 4 couplers w/ADSMomentum and only one had the
expected performance. Also a coworker designed a coupler using
SONNET (it performed well in the lab) I took his structure and imported
it to ADS2002C (momentum) and the simulation was off! The center Freq
was off and the output magnitudes were off. It seems to me that geometry
is a big factor on the software's ability to properly model your circuit.

Cheers

 

[yingyang]

benchmark

You may like to try multilayer patch to compare the various software. I must say IE3D outperform HFSS.

 

[personal]

antenna comparison

I seen great differences in slot antenna design....

 

[edgerate2222]

i thought hfss can do 3d not just 2.5d

 

[sviodo]

and you were tolly right

 

[toonafishy]

Well, I finally struggled through the *omentum tutorial and then put my design into it. Compared the results from HF$$ and M/W/O. All results were very similar except near resonances. There was some divergence there but not much. One thing I did notice was that HF$$ solution took about 1/2 hour, M/W/O about 1 hour but *omentum about 4 hours. The *omentum mesh was not all that thick so don't quite know why it took so long

 

[mwpro]

Momentum is slow. Interface is poor. And is completely incapable of simulation with thick metal.

 

[g86]

Re: 2.5d simulator accuracy

I think no one takes care of the feed probe properly. Especially if a SMA or N-type connector is fede from the edge side. Almost all simulators try to use ideal feeds. Some simulator uses probe feed but they are also not suitable for feed from the edge side. So to get a better result one should model upto the end of connector too. And many other things like metal surface smoothness are absent in many simulators. The metal thickness also cannot be modelled in some of them. And who cares about those solders dumps? Ensemble is not suitable for your metal box and Sonnet or MWo are not suitable for open box. My suggestion not to use 3D solvers for those ideal ports for approx results or use them with a proper modelling i.e considering all practical aspects.

Remember your pactical design will never the same as the modelled one.

:!: :idea: :?:

I just finished a microstrip balun design and simulated it in M/W/O/(2.5D) and H/F/S/S (3D). The results were similar but not exactly the same. which got me thinking. In you experience, which 2.5D simulator results most closely matched your lab measurements of the final circuit. I'm thinking of M/W/O, S/o/n/n/e/t, Mo/m/e/n/t/u/m, E/n/s/e/m/b/l/e, etc. When I build it in the lab, I'll let you know how closely the results matched the sims.

 

[pewang]

Recently,I have one design which simulated in CST MWS4.0,test result is very similar with simulate result.I think for the most microwave designs,when you do simulaton you must make your model as trueth as possible.and when do comparison,must make sense on instrument erorr and remember to get the refference ports the same.

 

[dragonlea]

i prefer I.E.3.D

 

[JoeSchmo]

I did a Q-Band patch design in a standard 30deg WG simulator a few years back (actually it was the HP HFSS w/ Acad13 interface) and the measured data was extremely consistant with the simulation...

 

[JoeSchmo]

Re: finite substrate

Even with HFSS, the result is also not very satisfactory but compared to the others, it is better. The HFSS fails to allow good description on the radiation pattern.

Actually, if you are using periodic boundaries, HFSS is very accurate. The only problem is that for a full radiation pattern you need to do a solve for every scan point and frequency. This method, though tedious, was verified by measurement for a notch element in both Co and X-pol patterns.

 

[totokevin]

Re: finite substrate

Actually, if you are using periodic boundaries, HFSS is very accurate. The only problem is that for a full radiation pattern you need to do a solve for every scan point and frequency. This method, though tedious, was verified by measurement for a notch element in both Co and X-pol patterns.

If you are modeling only 2.5D structure, you may try the Ansoft Designer - Planar EM, it has new functions specially for periodic structures, and also frequency sweep. That is a great feature for designing FSS/PBG. I have simulated some FSS structures, the results are perfect.

 

[IICCEE]

what is 2.5D mean?