H F S S vs C S T

Goto page 1, 2, 3, 4, 5 Next

EDAboard.com Forum Index -> Electromagnetic Design and Simulation -> H F S S vs C S T

Author

Message

mwpro

Joined: 20 Apr 2001
Posts: 83

29 Dec 2002 14:25 pba cst

Does anyone can give a comment on this topic?
Which is one is better. I've experience in H F S S and found that it is pretty slow and memory hungury. For a circuit with two plannar inductors, it takes on more than 1GB memory. Does CST offer anything better?

Gakusei

Joined: 02 Jul 2001
Posts: 43
Location: United Kingdom

29 Dec 2002 16:53 quotation price of hfss

I have found CST to be easy to use, accurate and fast.

I use it to design coaxial components and antennas, if I'm carefull I normaly achieve about 1 percent maximum error between the built prototype and CST model.

The optimisation system is easy to use and saves me lots of time in developement projects.

I tried all the others, but after using CST I haven't looked back.

I don't work for them.

sigmaxyz

Joined: 27 Jul 2001
Posts: 17

30 Dec 2002 13:48 hfss radiation pml site:www.edaboard.com

I stand by your point.

eirp

Joined: 11 Dec 2001
Posts: 639
Helped: 9
Location: Phase center

30 Dec 2002 14:02 hfss phase center 확인

In my opinion, MWS is one of the best EM simulators. I've tried more, but now I'm using MWS and IE3D. MWS is unique due to PBA meshing which save a lot of time and gives accuracy. Another fact, MWS is very user friendly! I can recommend you it.
Eirp

sigmaxyz

Joined: 27 Jul 2001
Posts: 17

30 Dec 2002 14:46 cst radiator cap problems

eirp's point is absolutely correct. I used hfss and mws designing a cell phone antenna. Hfss took 4 times in time than mws. Some complex problems may widen the gap. Like eirp said, to use more friendly is what an enigneer needs especially to new guy in terms of RF aspect.

plasma

Joined: 20 Dec 2001
Posts: 577
Helped: 40

30 Dec 2002 18:24 hfss vs cst

Hi

All above correct but if you need a printed
antenna you can draw it in IE3D. run it for
optimum results, then import by MWS and
run the last one for acureccy.

RGZ ZXC

Gakusei

Joined: 02 Jul 2001
Posts: 43
Location: United Kingdom

31 Dec 2002 13:16 numerical base of hfss

Maybe we should start a C/S/T usergroup?

It would be good to share experience, macros and more.

What do yo think?

eirp

Joined: 11 Dec 2001
Posts: 639
Helped: 9
Location: Phase center

31 Dec 2002 13:22 advantage hfss to cst

It's a good idea (maybe it can cover not only C/S/T but EM simulators as general). I think there's a lot of people using it, with good knowledge and experiences.. They helped me to solved some of problems in my projects time ago Laughing

If we speak about C/S/T, I'd like to say that it's becoming a standart in companies.
Eirp

Last edited by eirp on 31 Dec 2002 13:33; edited 1 time in total

mwpro

Joined: 20 Apr 2001
Posts: 83

31 Dec 2002 13:31 hfss vs mws

This thread demonstrates a positive side of elektroda!

Gakusei

Joined: 02 Jul 2001
Posts: 43
Location: United Kingdom

31 Dec 2002 15:29 cst hfss comparison and applications

Is there a negative side?

cheng

Joined: 01 Jan 1970
Posts: 111
Helped: 2

31 Dec 2002 15:43 cst mws vs. ansys

Well, let me throw my 20 cents inhere. To my opinion both SW are complementing each other. I still prefer the HFSS for they implement many physics and microwave features and not only the GUY. Quite frankly, CST GUY is not even theirs - they have this licensed to them by Dassault spin-off that every one can get today. Interested folk can pm me if they want to know the URL. Next to it: if one reads carefully the CST manual it says that the 'wideband' accuracy is in fact not so high and for truly large badwidth one must resort to splitting the band in parts. Third - CST folks is really smart in a way - they saw the HUGE deficiency the time-tomain solver exhibits and veeeeery quickly implemented the frequency domain solver - feature that most of the engineers highly praise for both speed and accuracy and one that resembles the HFSS one:) Smile

I still wander what CST folks has to say about it:). Next to it, the default settings of CST are too poor for any decent structure. One has to tune the meshing quite fine and some other advanced FDTD specific techniques. Time domain solvers generally exhibit some stability issues to handle - please pay attention that FDTD DOES NOT use the acclaimed FIT (Finite Integration Technique)in MWS but only in MAFIA. Simply said, the unstructured grid is difficult to handle in terms of stability etc. It is a rather long topic to discuss in here.
HFSS - rock solid core, great meshing with best possible manual meshing flexibility, MANY advanced features like Master-Slave BNDs, Freq-dependent materials, ferrite devices easy to do, superior antenna simulations, phase array is like a pleasure doing, etc. great flexibility in applying boundaries one over the other with clear order. Many more ports and boundaries -
Cons: we all know about the shitty and ugly interface - Ansoft folks is heavily working on it - and my information is that their interface will be far superior. One other thing to consider - for heavy WG structures - CST is desperately inaccurate and slow - especially for resonant structures, but not only. Price to pay with CST is finer mesh and looonger time.
So, my opinion is that the 2 solvers compliment each other for various applications - CST is definitely more user friendly and 'colorful', while HFSS is more 'professional solver' of the old time when every one was working under DOS like (or rather Unix) style. CST has its own merits but if I have to send a structure to the CNC machine I'd rather put in scrutiny with both - HFSS being the last one and compare. In terms of accuracy - I'd rather stick to HFSS, while speedier award is granted to CST.

Hope it helps. I very much praise the idea to have eaither CST MWS or HFSS (or both:)) forums in here.

greetz,
P.S About IE3D - speedy Gonsalez but lacks accuracy - Zeland is claiming accuracy is not their biggest concern (private talks with them) but rather being able to simulate 'larger' structures. It is why no one uses IE3D to design MMICs for instance. Losses in IE3D are incomparably poorer than Sonnet or Momentum - Again, best user interface and other 'cosmetics' in it - for patchy antennas be aware of about 5% freq shift downwards whatever you do in the setup. That's all for now.

Wish everyone Happy New Year.

DanTheMan

Joined: 24 Jan 2002
Posts: 203

31 Dec 2002 19:28 tilt box hfss

Somebody should mention that MWS's eigenmode solver does not allow materials to be specified, whereas HFSS's does.

/Dan

kato01

Joined: 02 Dec 2002
Posts: 84
Helped: 2

31 Dec 2002 20:34 vivaldi .hfss

It would be great if we could take a few geometries and each of us would analyze them on their own machine/soft.
Generating a significant data base for references would increase/decrease the confidence in the software.
I am an antenna guy and I already lost money manufacturing something that was not properly predicted.
I'm really pissed and now I work like 30 years ago: manufacture step by step and TEST!

If your money is at stake, you could be in trouble trusting the 50K MWS or others.

There is a paper (u can find it at
http://www.anteg.net/tuli/paperspdf/conference/20-APS-SALTLAKE-2000.pdf) which makes a timid atempt to just do that, but there are not many details. for a start, we could just take those examples.

If interested pm-me

Needbias

Joined: 03 Mar 2002
Posts: 164

01 Jan 2003 5:15 fdtd cern

Every software has it's merit,So HFSSvsCST is not a simple compare.

Gakusei

Joined: 02 Jul 2001
Posts: 43
Location: United Kingdom

01 Jan 2003 7:12 hfss limits

I have already done a lot of C/S/T/realworld comparisons and found the accuracy to be good for return/insertion loss in coaxial components and antenna radiation patterns.

However, I can't really give any of you drawings of the items compared for legal reasons (and someone could work out who I am).

I'll have a look around our workshops over the next couple of weeks and see if I can find a few simple items to compare. I can then post both sets of results and some simple drawings onto this forum.

If someone could then run the same models on H/F/S/S we can do a comparison.

I think a resonator, coaxial transformer and a dual dipole array should be good enough.

Let me know if you think I should do more.

kato01

Joined: 02 Dec 2002
Posts: 84
Helped: 2

01 Jan 2003 15:27 triangular lattice dispersion hfss

Radiation patterns should not be a problem in simulations.
The impedance IS.
I am working on microstrip arrays, where the knowledge of the input impedance is critical before the design of the feed network.
This is important especially in the design of low sidelobe arrays.
Even more critical for the design of serial fed arrays.

DanTheMan

Joined: 24 Jan 2002
Posts: 203

01 Jan 2003 15:31

Gakusei I will run the HFSS simulations if you post the required models. I've been using this program extensivelly in the last while, and only recently have been thinking of moving to MWS. However my first experience with MWS has been quite bad (I was probably doing something wrong).

Anyway, we definitely should get some sets of comparisons of the two EM simulators -- the sooner the better.

Regards,

/Dan

Gakusei

Joined: 02 Jul 2001
Posts: 43
Location: United Kingdom

02 Jan 2003 3:07

kato01

Actualy the problem of antenna self/mutual impedance and distribution system design is a circular one which I have lost much sleep over in past years. Just changing the distribution cable lengths in an array can have a profound effect on the radiation pattern and impedance if the mutual coupling is high.

I think the best thing to do would be to model a dual dipole array which is serialy fed (I presume this is what I call a traveling wave feed system) and include the feed system in the model.

Google
AdSense
Google Adsense

02 Jan 2003 3:07 Ads

eirp

Joined: 11 Dec 2001
Posts: 639
Helped: 9
Location: Phase center

02 Jan 2003 10:13

Guys, let's start the match, please!!

I'm waiting for your results too, it can be very interesting for us!
I prefer to make some simulations which is measured in literature also, so we can compare it absolutively with measurement.
What about a Yagi antenna? It's very similar to what Gakusei noted about dipole arrays and we can found measured results both for impedance and radiation pattern.

Best regards,
Eirp

PS: You can read this article too:
http://www.mwee.com/magazine/2000/CAD_nov.htm
There is discussed Vivaldi antenna simulated by H/F/S/S,M/W/S and E/M/P/I/R/E

kato01

Joined: 02 Dec 2002
Posts: 84
Helped: 2

02 Jan 2003 12:24 comparison between simulators

Gakusei,

Yes, you are right about cables and stuff, but I was refering to microstrip antennas where the rasdiating elements and the feeding distribution network are printed on the same surface.

EIRP,

The article in Microwave Engineering just shows how little these programs can be trusted.

For simulation (since we already have results), I suggest the structures described in
http://www.anteg.net/tuli/paperspdf/conference/20-APS-SALTLAKE-2000.pdf

any takers?

I can do IE3D. I'd love to do MOMENTUM but I donm't have it

eirp

Joined: 11 Dec 2001
Posts: 639
Helped: 9
Location: Phase center

02 Jan 2003 13:49 Re: comparison between simulators

Quote:

For simulation (since we already have results), I suggest the structures described in
http://www.anteg.net/tuli/paperspdf/conference/20-APS-SALTLAKE-2000.pdf

But there aren't measured results...

kato01: Yes, I know about low reliableness of such papers as in MWEE. I hope we will find some good structures with measured results.
Maybe I can give up measured data of some patch antennas we've built.

cheng

Joined: 01 Jan 1970
Posts: 111
Helped: 2

02 Jan 2003 18:29 HFSS vs. CST

Guys,

I find that topic about comparing the 2 tools a bit obsolete. Reasoning being that much of the quality of any simulation goes to the setup, size of the structure (in terms of wavelengths), solver used and many other factors (PC, cache, CPU, Memory etc) and to me it seems nearly wasted efforts to do in a concerted manner. I'd rather suggest discussing what the 2 tools must improve, share some knowledge and generate good suggestions. In return we may hope that some CST or HFSS folks read that:) (joking over here) and they get serious about making it better. To me the bottom line of our very good discussion was that the 2 tools were good and efficient. It should be a matter of personal preference which one individual may choose - last but not least is the price and support - we haven't touched that sensitive issue and I don't think we can do it inhere. Besides, I've seen some suggestions concerning stripline, or rather - Stratified media simulations being done with HFSS - it is ultimately irrational approach for such kind of structures and no one can beat the 2.5D tools (or planar 3D if you will) for many reasons not subject of thaty post. Last but not least comes the personal skills and knowledge of the particular designer to utilize such tools to accomplish maximum. After all, we are not working for those companies and have no direct interest.
I am sorry to say but the MWEE test (Vivaldi one) and the respective results is very much prejudiced (did anyone witnessed the results or took the effort to chek'em - i doubt it) - if one looks the data, one must choose EMPIRE and not CST - right? But like I said, to me it is just scratching the issue and is emotions laden discussion.

truly,

HxJi

Joined: 07 May 2001
Posts: 16
Location: Just ur side!

03 Jan 2003 3:07

hfss:cst
6.5:3.5
I think...

eirp

Joined: 11 Dec 2001
Posts: 639
Helped: 9
Location: Phase center

03 Jan 2003 3:30

HxJi : very fast comment, can you explain it more?

BTW I have very good experience with MWS.

We've simulated an 50Ohm load for SMA connector - coaxial ended with 50Ohm resistor. Resistor was modeled as ceramic cylinder with resistive layer thin only 7 microns!! I've computed required conductivity to reach 50Ohms (considering the resistor's dimensions) and simulated.
With about <2 mins of running I've get results (0-5GHz) which excellently agree with vector measurement (error is smaller than error of measurement!). Note that mesh has about 80k of nodes which (considering max and min dimensions) isn't reachable without PBA!! Results from another FDTD routines were (after some hours of running) totally wrong.

sviodo

Joined: 02 Jan 2002
Posts: 90

04 Jan 2003 0:23

for further discussion i invite you to EM simulators forum

With Regards,

Gakusei

Joined: 02 Jul 2001
Posts: 43
Location: United Kingdom

04 Jan 2003 2:30

Cheng,

I agree with most of your comments. Its nice to see this subject being debated in such a polite and thoughtfull manner.

However, I do think it would be usefull to compare results between the two packages, if only for peace of mind.

eirp,

Please be patient, I'm a busy man. Your post made me realise that the the structures I was going to model all related to the type of work which I do and may not give usefull information to people working in other fields. If any of you have drawings of structures with measured results you could let me have, I would be greatfull.

sviodo,

Thank you for creating the new forum for us. Following posts will go there.

mwpro

Joined: 20 Apr 2001
Posts: 83

04 Jan 2003 16:05

Dear all,

I feel very impressed by all the discussions. I agree that the best is to try out some benchmark samples to see which one runs better. Would like to thank all members involving in this discussion.

mwpro

yuyu

Joined: 27 Mar 2002
Posts: 187

12 Jan 2003 22:13

In my opinion CST is the best choise especially if you have to solve radiation problems. In fact HFSS is very memory consuming and without good accuracy. Further, since is a FEM based, sometime is possible to have spurios frequencies from the method itself and other spurios frequencies using the fast frequency solver.

eirp

Joined: 11 Dec 2001
Posts: 639
Helped: 9
Location: Phase center

13 Jan 2003 7:56

I've moved this discussion to EM simulators.
Eirp

Kit-the-great

Joined: 30 Mar 2002
Posts: 176
Helped: 1
Location: Russia

13 Jan 2003 12:50

I'm use both programs and I can note the following:

1. HFSS (@gilent) allocates a huge memory, but the calculation accuracy not fine (for example: calculated VSWR=1.05, measured VSWR=1.4). Device (coaxial connector) geometry was not very complex !
2. HFSS (@nsoft) works more accurate, but very slow. We used this program for dielectric resonator calculation and obtain satisfactory results, but this results only satisfactory (not good).
3. CST-MWS (on my opinion) now the best choise. I'm test this program on some structures, which have an accurate (analytical) solution (such as magic-tee etc.) and obtain very good comparision between numerical and analytical solutions without incredible memory and calculation time cost. In general, CST-MWS is a best choise.

With responce,
Kit-the-great