[SI-LIST] Re: HFSS vs. CST

  • From: "Hassan O. Ali" <hassan@xxxxxxxx>
  • To: mmoeller@xxxxxxxxxxxxxxx
  • Date: Thu, 13 May 2004 00:36:36 -0400

Moeller, Merrick wrote:

>Experts,
>       Are there any benchmarks that compare the performance between 
>HFSS and CST available? 
>

I'm not aware of any public document that does that. However, I'm 
certain many users have made independent evaluation of the two tools for 
specific applications of their interest. The evaluation reports are 
usually not published in the public domain due to confidentiality and 
even legal reasons.

>       In going from a frequency domain solver to a time domain solver
>and vise versa what are the trade offs?
>
Let's be specific here and consider CST Microwave Studio as the time 
domain solver and Ansoft HFSS as the frequency domain solver. (BTW, 
"CST" and "Ansoft" are company names, and "Microwave Studio" (MWS) and 
"High Frequency Structure Simulator" (HFSS) are the names of the 3D 
electromagnetic field solvers).

MWS solves for transient (time-domain) electric field (E) directly 
(natively). That is, it directly solves  for electric field in a 3D 
space as a function of time. It then indirectly, through Maxwell's 
equations, solves for magnetic field (H) as a function of time. Once E 
and H are known, many other related parameters (such as current 
distribution), can be computed (indirectly).  Frequency domain 
parameters are indirectly obtained through Fourier transformation.

HFSS solves natively for  E field as a function of frequency. H field is 
indirectly obtained through Maxwell's equations. Other related 
parameters are indirectly computed just like in MWS. Time-domain 
parameter are indirectly obtained through inverse Fourier transformation.

The keywords you need to observe from my two paragraphs above are 
"direct" and "indirect". Anything computed directly is "potentially" 
more accurate than the one computed indirectly. Therefore, transient 
parameters computed by MWS are "potentially" more accurate than 
transient parameters computed by HFSS. Likewise, frequency-domain 
parameters computed by HFSS are "potentially" more accurate than 
frequency-domain parameters computed by MWS. You have to meet certain 
Fourier transform conditions to obtain accurate results when you change 
parameters from one domain to the other. In most cases the conditions 
are hard to meet thereby limiting the accuracy.

Any one field component, E or H, can be used for S-parameter 
computation. However, the two tools differ slightly in the way they 
compute S-parameters. MWS would need to solve the problem as many times 
as the number of ports to obtain all S-parameters, while HFSS solves the 
problem only once to obtain all the S-parameters. For high port count, 
MWS can potentially be much slower than HFSS due to this reason.

For frequency-domain S-parameters to be valid, MWS requires the field at 
each port to settle to a particular (specified) level. The smaller the 
level the longer the simulation. The field in high Q structures take 
extremely long to die off. Such structures may take very long to solve 
using MWS transient solver.

MWS uses a time stepping algorithm. That is, it solves for E field one 
time step at a time. Once all the time domain E field is computed, the 
fields at all frequencies are computed at once. On the other hand, HFSS 
solves for E field one frequency at a time. If you need time-domain 
results, you will have to solve for many frequencies. It offers at least 
two ways of obtaining fast frequency domain results : Fast Sweep and 
Interpolating Sweep. They are both "interpolating" schemes hence 
introduce their own errors. Generally you cannot really tell which tool 
will be faster in producing broadband results of a particular structure. 
Neither can you tell which one will be more accurate in that case. 
However, for narrowband, you are certainly better off with HFSS.

The other issue is about scaling. Bigger size problem needs longer time 
to settle, which leads to longer MWS simulation time even for narrowband 
solution. Bigger problems also demand more memory and MWS, being a 
Windows software, can't solve problems requiring more than 2GB of 
memory. With HFSS, bigger size problems also means huge matrices to 
solve, and very long simulation time. In memory terms, HFSS is better as 
it can run on UNIX and use up to 8GB of memory (or more).

>       Is accuracy of the HFSS and CST at the same level?
>

This is a very important question! When I evaluate a tool I don't only 
check whether the tool is accurate, rather I check if it is 
*consistently accurate*. I like the tool to be predictable. When I use 
the tool to model a number of known structures I expect it to 
consistently give accurate results - for all the structures, all the time.
Both MWS and HFSS can be setup to yield accurate results but the effort 
needed to achieve that differs considerably between the two tools.

>       Is there any type of problem that will not be accurate with one
>or the other?
>  
>
Frequency-dependent loss is not handled well by MWS. However, you can do 
more kinds of analyses with MWS than HFSS. For example,TDR analysis with 
NRZ waveform.

Good luck.

Hassan.


------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field

or to administer your membership from a web page, go to:
http://www.freelists.org/webpage/si-list

For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field

List FAQ wiki page is located at:
                http://si-list.org/wiki/wiki.pl?Si-List_FAQ

List technical documents are available at:
                http://www.si-list.org

List archives are viewable at:     
                http://www.freelists.org/archives/si-list
or at our remote archives:
                http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
                http://www.qsl.net/wb6tpu
  

Other related posts: