Continue to Site

Welcome to

Welcome to our site! is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

EM solver VS. Device solver ?

Not open for further replies.


Member level 3
Jun 17, 2002
Reaction score
Trophy points
Activity points
em solver

Hi all:

One topic in which I am interested is the
comparison between EM solvers and Device solvers.

In order to model a metal on-chip, such as
an interconnect over a multilayered
SiO2 & Si structure, I limit my solution with two
1) Using EM solver (Fullwave/Quasi-TEM),
2) or using device simulator.

The problem with EM solver as far as I know
is how to model a very lossy substrate (highly doped
silicon) and the RLGC elements extracted should
be frequency dependent.

I am not very familiar with device(semiconductor)
solver. But I hope it/they can provide frequency-dependent
RLGC element or other format which can be compared
with EM solvers.

You suggestions are welcome from either or both
side(s) of the solvers.
I list several solvers below but you are
welcome to show your opinion beyond the list.

EM solver
1) ADS momentum
2) CST Microwave studio
3) HFSS (my experience: not accurate for lossy media)

Device(semiconductor) solver
2) Silvaco ATLAS
3) Synopsys MEDICI

If someone could show one comparison, that would be

Thanks for your attention.


sonnet lite + rlgc

I have interested in the same(slightly different) situation which you Said.

Now trying to apply EM simulation result to Microwave Transistor.
However Available Device Simulator(TCAD) for me ,
has only 2d analysis capability... I cannot compare the result
with TCAD regorously...

Do you have fully 3D working TCAD?
Or Can we share the simulation and measurement data for analysis?

:D Have a nice day.


Sure. To share experience and knowledge is the reason we
are here.

But if you are talking about TCAD tool, you'd better
PM me because we need to follow the rules of this

It is good to know at least someone has interest in
my post. :)



hi, are you .I am very insterting about TCAS of silvaco.May I have your email address? so we can communicate freely and happily. my email address is
waiting for your replys! thank you

Hi Div -- I work for Sonnet, Sonnet at least (perhaps the others too) has no problem with lossy/conducting substrates. When loss is added, analysis time increases, but there is absolutely no change in accuracy.

If you network is not too big, you can use the free SonnetLite,

If your network is planar (even including thick metal) should not use a volume meshing code, definately use a surface mesher like Sonnet (shielded analysis) or Momentum (unshielded analysis).

If your circuit has any non-planar aspect, or you want to get a second analysis using a different approach as a check, then I recommend CST. We evaluated CST extensively against all competitors (accuracy, quality of solution, speed, and user interface were the primary criteria) before my company decided it was very much worth it to repersent it in North America.

If you wish to check any EM analysis for accuracy for conducting substrate I suggesting analyzing a parallel plate capacitor. Set the conductivity of the dielectric to a known value. Calculate the resistance you should have. Then see if that is the resistance you get when you analyze it. If conducting (or lossy) substrate is important, you should definately do this test no matter what EM analysis you are using.

Hi rautio and div,

we can use sonnetlite but the problem with it is limited to 16 MB files upon registering because it is free and if your device is frequency dependent and you want to simulate at higher frequency it is impossible to use sonnet lite and it also limits the design to single dielctric layer.

if it is possible for you work on IE3D of Zeland it is efficient.

Hi Saravan -- SonnetLite is limited to three dielectric layers, one layer is usually air. This leave you two layers to play with. This lets you do spiral inductors, which require a cross-over. There are very low cost options to upgrade if you need more memory, see For example $495 gets you double the memory. Much much less expensive than buying a big expensive EM solver that does way more than you need. (Little secret, don't tell anyone: With some experience and if laid out with Sonnet in mind, nearly all practical circuits can be done in SonnetLite...but be sure not to tell anyone!)

Also, you might want to post your circuit on the sonnet users forum (same link, Support-->Sonnet Support-->Forum), there are many ways to reduce memory requirement. There is even an entire chapter in the manual about it. Takes about 10 minutes to read that chapter. And you can save a lot of time and effort. Some of the ideas even work for other EM solvers.

Not open for further replies.

Part and Inventory Search

Welcome to