Edgar J. Kaiser
Certified Consultant
Please login with a confirmed email address before reporting spam
Posted:
7 months ago
Aug 22, 2024, 12:25 p.m. EDT
Hi Luke,
that is indeed a little mysterious.
I added an Ampere's Law node that overrides all the Free Space assignments and set the conductivity for air and the dielectric to 1 S/m for solver stability.
Now the result looks realistic.
I think the Free Space thing is relatively new and I never used it. I always use Ampere's law with suitable material assignments. Free space is essentially equivalent to air. The Free Space node should actually just add the stabilization conductivity that is needed to get the solver to converge. But maybe I am missing something.
I would be curious about comments from other people.
Cheers
Edgar
-------------------
Edgar J. Kaiser
emPhys Physical Technology
www.emphys.com
Hi Luke,
that is indeed a little mysterious.
I added an Ampere's Law node that overrides all the Free Space assignments and set the conductivity for air and the dielectric to 1 S/m for solver stability.
Now the result looks realistic.
I think the Free Space thing is relatively new and I never used it. I always use Ampere's law with suitable material assignments. Free space is essentially equivalent to air. The Free Space node should actually just add the stabilization conductivity that is needed to get the solver to converge. But maybe I am missing something.
I would be curious about comments from other people.
Cheers
Edgar
Durk de Vries
COMSOL Employee
Please login with a confirmed email address before reporting spam
Posted:
7 months ago
Aug 23, 2024, 3:47 a.m. EDT
Updated:
7 months ago
Aug 23, 2024, 4:02 a.m. EDT
Hi Luke,
Thanks for reaching out!
The reason for the lack of material contrast, is that everything is free space. Adding materials to the model gives COMSOL access to the necessary material data, but it does not specify how to treat the domains. In the physics, the default node is "Free Space" (vacuum/air), this node serves as a canvas. You can add an "Ampère's Law in Solids" node to indicate that your metals are solid (basically; what Edgar says -- see attached).
The reason for this design is that for electromagnetical phenomena (in particular related to anisotropic material properties and eddy currents) the behavior of liquids and solids is conceptually different. Since making everything liquid or solid by default would result in models that are wrong in one way or another, we decided to start with empty space and encourage the end user to make a decision.
On top of this, the free space node comes with built-in stabilization. That is the reason why you got some current in your initial model (although a very small amount). On the forum you may have seen the general advice to use "1 S/m in the air". The free space feature determines an appropriate amount of stabilization conductivity automatically, by looking at the local skin depth.
We've been experimenting with this design for about one release now and your feedback shows that for some use cases this can lead to confusion. We will look into this and improve the usability in this regard.
Thanks for reporting!
Kind regards,
Durk
Hi Luke,
Thanks for reaching out!
The reason for the lack of material contrast, is that everything is free space. Adding materials to the model gives COMSOL access to the necessary material data, but it does not specify how to treat the domains. In the physics, the default node is "Free Space" (vacuum/air), this node serves as a canvas. You can add an "Ampère's Law in Solids" node to indicate that your metals are solid (basically; what Edgar says -- see attached).
The reason for this design is that for electromagnetical phenomena (in particular related to anisotropic material properties and eddy currents) the behavior of liquids and solids is conceptually different. Since making everything liquid or solid by default would result in models that are wrong in one way or another, we decided to start with empty space and encourage the end user to make a decision.
On top of this, the free space node comes with built-in stabilization. That is the reason why you got some current in your initial model (although a very small amount). On the forum you may have seen the general advice to use "1 S/m in the air". The free space feature determines an appropriate amount of stabilization conductivity automatically, by looking at the local skin depth.
We've been experimenting with this design for about one release now and your feedback shows that for some use cases this can lead to confusion. We will look into this and improve the usability in this regard.
Thanks for reporting!
Kind regards,
Durk
Magnus Olsson
COMSOL Employee
Please login with a confirmed email address before reporting spam
Posted:
7 months ago
Aug 23, 2024, 4:15 a.m. EDT
Updated:
7 months ago
Aug 23, 2024, 4:17 a.m. EDT
Hi Luke,
The Free Space is the default domain feature in the Magnetic Fields interface since version 6.2. It is there to automate the modeling of free space (non-conducting media with relative permittivity and relative permeability of unity). Depending on the frequency, some artificial conductivity is needed in such domains and the Free Space feature is there to automate that.
The Free Space feature must be overridden by the user in domains having a (non-free space) material by adding Ampère's Law in Solids or Ampère's Law in Fluids - otherwise the entire simulation domain will be treated as free space.
The reason for having two flavors of the Ampère's Law feature is to be able to correctly handle models with moving parts/mesh as then solids and fluids must be treated differently.
When adding Ampère's Law in Solids to your non-air domains (above the Coil as the Coil also overrides Free Space and is overridden by Ampère's Law in Solids), the solver warned for some missing material properties in the CFRP_homogenised material (permittivity and conductivity) so I added, hopefully intelligent, guesses for those see the attached version.
I hope this helps.
-------------------
Magnus
Hi Luke,
The Free Space is the default domain feature in the Magnetic Fields interface since version 6.2. It is there to automate the modeling of free space (non-conducting media with relative permittivity and relative permeability of unity). Depending on the frequency, some artificial conductivity is needed in such domains and the Free Space feature is there to automate that.
The Free Space feature must be overridden by the user in domains having a (non-free space) material by adding Ampère's Law in Solids or Ampère's Law in Fluids - otherwise the entire simulation domain will be treated as free space.
The reason for having two flavors of the Ampère's Law feature is to be able to correctly handle models with moving parts/mesh as then solids and fluids must be treated differently.
When adding Ampère's Law in Solids to your non-air domains (above the Coil as the Coil also overrides Free Space and is overridden by Ampère's Law in Solids), the solver warned for some missing material properties in the CFRP_homogenised material (permittivity and conductivity) so I added, hopefully intelligent, guesses for those see the attached version.
I hope this helps.
Edgar J. Kaiser
Certified Consultant
Please login with a confirmed email address before reporting spam
Posted:
7 months ago
Aug 23, 2024, 5:23 a.m. EDT
Durk and Magnus,
thanks for your comments. In earlier versions the default node was Ampere's Law and all domains were treated according to their material settings. Now we need additional Ampere's law nodes for everything that is not free space. That is fine for me.
Is the automatic stabilization conductivity in the Free Space node available in the model variables? I didn't find it in the equation view.
And I guess I need to do some reading about the differences of solid and fluid Ampere's law nodes.
Cheers
Edgar
-------------------
Edgar J. Kaiser
emPhys Physical Technology
www.emphys.com
Durk and Magnus,
thanks for your comments. In earlier versions the default node was Ampere's Law and all domains were treated according to their material settings. Now we need additional Ampere's law nodes for everything that is not free space. That is fine for me.
Is the automatic stabilization conductivity in the Free Space node available in the model variables? I didn't find it in the equation view.
And I guess I need to do some reading about the differences of solid and fluid Ampere's law nodes.
Cheers
Edgar
Magnus Olsson
COMSOL Employee
Please login with a confirmed email address before reporting spam
Posted:
7 months ago
Aug 29, 2024, 3:52 a.m. EDT
Hi Edgar,
It shows up in equation view in the regular tensor conductivity variables, e.g: mf.sigmaxx
The equivalent skin depth that is aimed for in the frequency domain when using "Automatic" is 100 times the linear size of the geometry.
From the manual:
"A suitable rule of thumb for choosing the stabilization conductivity is given by the skin
depth, as compared to the overall geometry size. The assumption is that if the skin
depth in Free Space is around one hundred times the device size, the resulting loss and
the impact on lumped device properties are insignificant. At the same time, the model
will still be sufficiently stable. What is considered a good value will differ per model,
though. It is therefore recommended to double-check the free space loss as compared
to other loss terms in the model. If the stabilization conductivity has a negative impact
on the overall model accuracy, consider lowering it."
Best regards,
-------------------
Magnus
Hi Edgar,
It shows up in equation view in the regular tensor conductivity variables, e.g: mf.sigmaxx
The equivalent skin depth that is aimed for in the frequency domain when using "Automatic" is 100 times the linear size of the geometry.
From the manual:
"A suitable rule of thumb for choosing the stabilization conductivity is given by the skin
depth, as compared to the overall geometry size. The assumption is that if the skin
depth in Free Space is around one hundred times the device size, the resulting loss and
the impact on lumped device properties are insignificant. At the same time, the model
will still be sufficiently stable. What is considered a good value will differ per model,
though. It is therefore recommended to double-check the free space loss as compared
to other loss terms in the model. If the stabilization conductivity has a negative impact
on the overall model accuracy, consider lowering it."
Best regards,
Please login with a confirmed email address before reporting spam
Posted:
7 months ago
Sep 2, 2024, 4:31 a.m. EDT
Thank you all so much for your help and explanations - makes perfect sense now that I know what the problem was.
Thank you all so much for your help and explanations - makes perfect sense now that I know what the problem was.