Fem/L@b vs. @nsys
i am interested in experience exchange between two FEM simulators:
fem-L@b <--> @nsys
i am intended to use one of these programms to simulate and investigate the electromagentic interferences on PCB's. which program is easier to handle and serves the best? or should i suggest my boss to buy another, maybe better tool for this kind of experience.
thanks to anybody for help
ciao
Special s/w is for this analysis: Siwave of Ansoft, Speed of Sigritys, signal itegrity modules of PCAD etc.
Ansys could handle a larger problem better than FEMLAB. FEMLAB allows you to customize the equations that are solved. I have not used either for EM simulation. I use ANSYS mainly for thermal-electric and mechanical. I've looked at FEMLAB but found it difficult to learn and become productive.
I'm using FEMLAB for estimation of patch antenna's modes and resonators.
I like it because you can fully control FEMLAB from MATLAB (as .m file).
It's good for people familiar with PDE and it's physical meaning.
As I understand your question, FEMLAB isn't much suitable --> recommend you to aim to software mentioned by JNekas...
Best regards,
Eirp
thanks a lot for your help.
my problem seems to be too difficult for p-C@d, because i want to compute the e.m. radiation at any distance to the board.
the problem is the following:
@/ns/ys and Fe/ml@b fits well for computing the electro-magnetic field, but as i know only @/ns/ys has some electric (which are still primitive) circuit elements which can be coupled with the field equations. so i think the only way to simulate this problem is to record voltages and current of a circuit using a simulator like P/S/P/I/C/E or Sp/e/t/r@, including reflection etc.. using this recorded data can be used for the stimuli for example @/ns/ys to compute the e.m. field. i know that this flow would contain a lot of errors and approximations, but i don't know a better and simpler way for this problem. especially i'm interested in digital high speed systems with signal lenght in teh range of lamda.
thanks
Sorry, but i didn't looked at the Si/Wave product. it seems to be a very powerfull tool. can the spice simulator coupled with any commercial simulator or libaries respectiveley?
this would be great!!! 8O
Eirp and all:
I am using FEMlab to calculate capacitor with
a 2D cross section, or in other sense, capacitor per-unit-length.
The manual of FEMlab does NOT give a clear way to
calculate capacitor. There are two 2D examples given.
One calcuates the capacitor by
"postint(fem,'2*We')"
The other one uses
"postint(fem,'2*pi*r*We')"
Please show me your understading and suggestion.
Thanks
Div
Hi, div!
Yes, it isn't shown clearly. Let now forget that for these capacitors
there are 2 methods used. What is common, is to know stored electric energy WE=integral over domain Omega of (We) d(Omega).
(WE is total energy and We is its density)
MEMS capacitor:
They're using C=2*WE/V0^2
More clearly written
WE=postint(fem,'We','dl',1) (V0 is constant, so it can be removed from the integral) and we are integrating in cartesian coordinates.
C=2*WE/V0^2 (V0=1e-6 in this case)
SPHERICAL capacitor
Another equation for capacity C=Q^2/2*WE
WE=int.over domain Omega of (We) d(Omega)=int(0-2pi)int(0-r) We(r)*r d(phi) d(r). We has only radial dependecy.
As We doesn't depend on agle, can be moved before integral: WE=2*pi*int(0-r) We(r)*r d(r), where r is Jacobian.
So
WE=postint(fem,'2*pi*r*We')=2*pi*postint(fem,'r*We ')
From FEMLAB notation I can't prove that it's really spherical capacitor, integration looks like in cylinder coord(Jacobian) ?!
I'm not sure, if I'm really correct, but the difference seems arising from diferent coordinates used... Only my suggestion...
Best regards,
Eirp
Eirp and all:
Thank you for the very instructive explanation.
I get the idea of using "2*pi*r*We".
But, then, the next question is how FEMLab understand
whether you want to integrate over cartesian or cylinder
coordinate?
For instance, if you put "postint(fem,'We')",
Does FEMLab do [int (We) d(x) d(y)] or
[int (We* r) d(phi) d(r)]?
Another related question is when you do
"export FEM structure as fem", how many variable
can you have access to?
Do you do "fem.equ.var" to get all the variable?
Third question is what if I want to the following:
C=Q/U, I can specify the U as different potential at
different mental, then how can I integrate the
rho--charge density-- to get Q?
Fourth question is how you specify "Magnetic wall"?
Do you use "elekctric insulation"? Are they same?
Thanks
Div
Dear Div!
As I'm not so friendly with using this quasistatic module, I can give you answer to only a part of your questions:
It's given by the module type used in Model Navigator: In-plane quasi static (x,y,z) and Axisymmetric quasi static (r,z,phi)
Do export FEM structure as fem, switch to Matlab and write fem, fem.equ, fem.equ.var,fem.xxx, and you will see what's accessible.
I think yes, because of following:
PMC means H=0 at the boundary and Z=E/H Thus Z-->infinity as H-->0.
Cheers,
Eirp
PS: From the Axisymmetric definition I really see that there are cylindrical coordinates used, not spherical.
--> are wrong with "Spherical Capacitor" name.
I am trying FEMLab with other modes.
Hope it would solve the lossy capacitor problem.
I would very grateful if you compare these solvers from the viewpoint of their performance and large scale problems in electromagnetic field simulation.
Thanks
Can Femlab deal with electromagnetic scattering problem ?
I am interest in Femlab, but can not find any help about solve EM scattering problems.
anyone like to help me? thanks
Div,
For your last question about the magnetic wall:
An electrical insulation has nothing to do with H or B; it means that
Dn1-Dn2=Qs, and Et1-Et2=0; where n=normal and t=tangential, and 1,2-refer to the marterials of the interface and E/D are the electric and displacement fields respectively and Qs is the surface charge density. Boundary conditions for the magnetics are like the opposite of the electrical where a nonmagnetic material results in Bn1=Bn2 and Ht1=Ht2, where H is the magnetic field intensity and B is the magnetic induction field. So I'd be careful how you specify this constraint.