Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

Impedance boundary condition for EMW eigenfrequency analysis?

Please login with a confirmed email address before reporting spam

I'm trying to calculate the Q factor of a 3D RF resonator, including the effect of skin losses in the metal conductor. When I run an eigenfrequency analysis with impedance boundary conditions, COMSOl 4.3 returns the error below.

The model is attached. Any suggestions would be appreciated.
-Till Rosenband, NIST


Undefined value found.
- Detail: Undefined value found in the stiffness matrix.
.There are 8316 equations giving NaN/Inf in the matrix rows for the variable mod1.E10
. at coordinates: (0.0305,-3.79641e-018,0.000816667), (0.0305,-3.79641e-018,0.000883333), (0.0304731,0.000117369,0.000883333), (0.0304461,0.000234737,0.000816667), (0.0304461,0.000234737,0.00075), ...
There are 4110 equations giving NaN/Inf in the matrix rows for the variable mod1.E20
. at coordinates: (0.0304731,0.000117369,0.000816667), (0.0306791,-5.03007e-005,0.000816667), (0.0306773,-0.000186272,0.000883333), (0.0306773,-0.000185906,0.000816667), (0.0304991,-0.000135634,0.000883333), ...
There are 4110 equations giving NaN/Inf in the matrix rows for the variable mod1.E21
. at coordinates: (0.0304731,0.000117369,0.000816667), (0.0306791,-5.03007e-005,0.000816667), (0.0306773,-0.000186272,0.000883333), (0.0306773,-0.000185906,0.000816667), (0.0304991,-0.000135634,0.000883333), ...
and similarly for the degrees of freedom, NaN/Inf in the matrix columns.




5 Replies Last Post Jun 23, 2012, 5:57 a.m. EDT

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jun 21, 2012, 3:10 p.m. EDT
Can't open your model as it's in 4.3. But try changing impedance BC to Scattering or other type of BC. When solving for eigenmodes, you don't need to have excitation.
Can't open your model as it's in 4.3. But try changing impedance BC to Scattering or other type of BC. When solving for eigenmodes, you don't need to have excitation.

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jun 22, 2012, 5:12 a.m. EDT
Hi

check your materials, and why do you use bulk and boundary copper ?
I can understand as thin layers on the Circuit Board, but not on the copper layer like that

I would also consider to unite the full Cu geometry into one part and remove the interior boundaries, its easier to pick, but it might be trickier to mesh, an idea but not sure it matter that much, apart reducing drastically the number of domains

Then for me you have not defined enough BC to give COMSOl the ability to find an unique solution. Do you not need a GND, a source term etc, are all your dependent variables fully constrained like that ? I believe not, but I do not know all the details of your model either ...

--
Good luck
Ivar
Hi check your materials, and why do you use bulk and boundary copper ? I can understand as thin layers on the Circuit Board, but not on the copper layer like that I would also consider to unite the full Cu geometry into one part and remove the interior boundaries, its easier to pick, but it might be trickier to mesh, an idea but not sure it matter that much, apart reducing drastically the number of domains Then for me you have not defined enough BC to give COMSOl the ability to find an unique solution. Do you not need a GND, a source term etc, are all your dependent variables fully constrained like that ? I believe not, but I do not know all the details of your model either ... -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jun 22, 2012, 8:45 p.m. EDT
Thank you for this suggestion. When I try a scattering boundary condition instead of an impedance boundary, there is a similar error (below). My initial question should have been this:

How can I calculate eigenfrequencies, including the imaginary (loss) term, for a circuit board RF resonator.? Some of the loss comes from the circuit board substrate, and this is easy to include as an option in the wave equation (electric displacement field model). How can the loss fom the conductor resistance (including skin effect) be included? Also, should the circuit board traces be modeled as infinitely thin boundary layers, or should they be made with some thickness so that COMSOL's solution automatically includes skin effects?

An image of the model is attached, including the cylindrical simulation domain, rectangular circuit board substrate, and top and bottom conducting traces.

Thanks again,
-Till

Undefined value found.
- Detail: Undefined value found in the stiffness matrix.
.There are 3552 equations giving NaN/Inf in the matrix rows for the variable mod1.E10
. at coordinates: (0.0305,-3.79641e-018,0.000816667), (0.0305,-3.79641e-018,0.000883333), (0.0303889,-3.76919e-018,0.000816667), (0.0303889,-3.76919e-018,0.00095), (0.0302778,-3.74198e-018,0.000883333), ...
There are 1184 equations giving NaN/Inf in the matrix rows for the variable mod1.E20
. at coordinates: (0.0303889,-3.76919e-018,0.000883333), (0.0304991,-0.000135634,0.000883333), (0.0303871,-0.000135971,0.00095), (0.0302778,-3.74198e-018,0.000816667), (0.0300556,-3.68755e-018,0.000883333), ...
There are 1184 equations giving NaN/Inf in the matrix rows for the variable mod1.E21
. at coordinates: (0.0303889,-3.76919e-018,0.000883333), (0.0304991,-0.000135634,0.000883333), (0.0303871,-0.000135971,0.00095), (0.0302778,-3.74198e-018,0.000816667), (0.0300556,-3.68755e-018,0.000883333), ...
and similarly for the degrees of freedom, NaN/Inf in the matrix columns.

Thank you for this suggestion. When I try a scattering boundary condition instead of an impedance boundary, there is a similar error (below). My initial question should have been this: How can I calculate eigenfrequencies, including the imaginary (loss) term, for a circuit board RF resonator.? Some of the loss comes from the circuit board substrate, and this is easy to include as an option in the wave equation (electric displacement field model). How can the loss fom the conductor resistance (including skin effect) be included? Also, should the circuit board traces be modeled as infinitely thin boundary layers, or should they be made with some thickness so that COMSOL's solution automatically includes skin effects? An image of the model is attached, including the cylindrical simulation domain, rectangular circuit board substrate, and top and bottom conducting traces. Thanks again, -Till Undefined value found. - Detail: Undefined value found in the stiffness matrix. .There are 3552 equations giving NaN/Inf in the matrix rows for the variable mod1.E10 . at coordinates: (0.0305,-3.79641e-018,0.000816667), (0.0305,-3.79641e-018,0.000883333), (0.0303889,-3.76919e-018,0.000816667), (0.0303889,-3.76919e-018,0.00095), (0.0302778,-3.74198e-018,0.000883333), ... There are 1184 equations giving NaN/Inf in the matrix rows for the variable mod1.E20 . at coordinates: (0.0303889,-3.76919e-018,0.000883333), (0.0304991,-0.000135634,0.000883333), (0.0303871,-0.000135971,0.00095), (0.0302778,-3.74198e-018,0.000816667), (0.0300556,-3.68755e-018,0.000883333), ... There are 1184 equations giving NaN/Inf in the matrix rows for the variable mod1.E21 . at coordinates: (0.0303889,-3.76919e-018,0.000883333), (0.0304991,-0.000135634,0.000883333), (0.0303871,-0.000135971,0.00095), (0.0302778,-3.74198e-018,0.000816667), (0.0300556,-3.68755e-018,0.000883333), ... and similarly for the degrees of freedom, NaN/Inf in the matrix columns.


Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jun 22, 2012, 8:51 p.m. EDT
Hi Ivar,

I appreciate your taking a look at this. I used thicker copper layers in an attempt to include the skin effect. How would COMSOL know if my circuit traces are thinner or thicker than the skin depth, unless I make them 3 dimensional?

The bulk copper seemed appropriate, and then COMSOL complained that I should specify materials for the conductor boundaries, too. That is why there is bulk & boundary copper.

I just want to know the eigenmodes. That shouldn't require source terms or ground...

A cleaned up version of the model is attached.

Have a nice weekend,
-Till
Hi Ivar, I appreciate your taking a look at this. I used thicker copper layers in an attempt to include the skin effect. How would COMSOL know if my circuit traces are thinner or thicker than the skin depth, unless I make them 3 dimensional? The bulk copper seemed appropriate, and then COMSOL complained that I should specify materials for the conductor boundaries, too. That is why there is bulk & boundary copper. I just want to know the eigenmodes. That shouldn't require source terms or ground... A cleaned up version of the model is attached. Have a nice weekend, -Till


Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Jun 23, 2012, 5:57 a.m. EDT
Hi

I believe you are doing double up then, as for me: either you define the current flowing in the skin = boundary, and you ignore the domain (bulk), or you solve it fully in the domains (and ignore boundaries as materials) and the EM equations will force the current to flow in the skin depth region of your domain, the latter is certainly heavier to solve but more general.
So I would suggest choose one of the two: boundary or domain approaches.


I'm not sure how to interprete the eigenmodes if you leave V fully unconstrained.

It's like solid and eigenmodes, you do not need to fix your object in space, but if you do not, then the 6 first rigid body modes are all "0" and the others are free-free modes, but they are very different from a fixed-free mode you get when you constrain one boundary to get a fixed reference

I would say you must have a ground "plane" somewhere, but probably it could be on an external boundary

--
Good luck
Ivar
Hi I believe you are doing double up then, as for me: either you define the current flowing in the skin = boundary, and you ignore the domain (bulk), or you solve it fully in the domains (and ignore boundaries as materials) and the EM equations will force the current to flow in the skin depth region of your domain, the latter is certainly heavier to solve but more general. So I would suggest choose one of the two: boundary or domain approaches. I'm not sure how to interprete the eigenmodes if you leave V fully unconstrained. It's like solid and eigenmodes, you do not need to fix your object in space, but if you do not, then the 6 first rigid body modes are all "0" and the others are free-free modes, but they are very different from a fixed-free mode you get when you constrain one boundary to get a fixed reference I would say you must have a ground "plane" somewhere, but probably it could be on an external boundary -- Good luck Ivar

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.