|T O P I C R E V I E W
|Posted - Mar 10 2023 : 12:27:50
I have a simple model composed of a nanoribbon, a plate and dielectric between them. It works when the nanoribbon is far away from the dielectric layer, but when the nanoribbon gets close to the dielectric layer, I get the warning capacitance matrix is not diagonally dominant. Is there any way to troubleshoot what could be going wrong? Here's a link to the input geometries and txt file with conductor and dielectric definitions: ht*ps://w*w.dropbox.com/scl/fo/pw805ugopi29a0wxbcmn3/h?dl=0&rlkey=encf8zyqqei7u3ucovc9acugr
Besides, if I want to calculate the capacitance between two contact conductors, should I do anything to the interface#65311;
Thanks in advance!
|5 L A T E S T R E P L I E S (Newest First)
|Posted - Mar 16 2023 : 07:17:59
I get it! Thank you for your precious help!
|Posted - Mar 15 2023 : 19:07:40
for two conductors in contact, which I don't want to treat as a single conductor
If you have two conductors in contact e.g. two cubes touching along a face, you should actually treat them as a single conductor - they would be equipotential.
If you still want to keep them as separated entities, you should instead *remove both* touching faces as they would be an internal structure completely shielded by the resulting shell; you may then attribute some panels to one conductor and some other panels to another conductor. Up to you how to use the results here (e.g. in a Spice model the two capacitors would actually have a common node as they are electrically in contact, thus shorting the mutual capacitance between the two parts)
|Posted - Mar 14 2023 : 09:22:47
OK, it turns out to be the problem of mesh.
I have another question. You said NO panel should overlap, but for two conductors in contact, which I don't want to treat as a single conductor, I am not sure how to deal with the interface. Should I distribute the interface to one conductor and leave the other conductor an unclosed surface?
|Posted - Mar 14 2023 : 06:17:27
Thank you Enrico for the reply!
In my model, the ribbon is lift to 2855m, while the dielectric layer is 2850 m in height. I think there is no overlap, so I am still confused about the warning. Just as you said, the model uses the actual geometry, and I will try to use a refined mesh.
|Posted - Mar 13 2023 : 15:55:37
your model has some issues; the main one is that the ribbon and the plate (top side of the dielectric slab) have overlappling panels.
As stated in the documentation as well as you can see in the samples, NO panel should overlap, be them conductor, dielectric or both. This means that for your geometry you should cut a slot in the plane at the top of the dielectric corresponding to the base of the ribbon. Also, you should divide the ribbon conductor in two parts, one bottom one that is the conductor in contact with the dielectric with relative permittivity 3.5 and the rest (side and top) that are in contact with air.
Failing to do so creates an invalid geometry and leads to numerical issues in the internal calculation of the integrals.
Also, unless your geometry is actually that one, I would recommend a geometric ratio more consistent with the dimension of the ribbon. As you can see, the dimension of the panels between the ribbon and the dielectric sides is quite different. You will need a decent refinement, that Fastercap can handle automatically, but will lead to a lot of panels.
Hope this helps,