Static Current Conduction question

Numerical methods and mathematical models of Elmer
Post Reply
sqrtmo
Posts: 2
Joined: 20 May 2024, 11:55
Antispam: Yes

Static Current Conduction question

Post by sqrtmo »

Hello,

This is my very first try with Elmer and this kind of simulations at all. My try is based on this tutorial.
https://www.youtube.com/watch?v=hNHTwpegFBw

My setup consists of V shaped PCB trace with a source pad on the sharp edge and two sink pads at the ends of the two branches.
And it aims to visualise the potential drop, volume current and joule heating.

At the source pad I did applied Dirichlet boundary condition with a potential value of 5V.
My sink pads has equal arias of .895mm^2. I applied Neumann BC's.
So at pad A I applied 1A current equal to current density value of -1117318 and to pad B 4A equal to CD of -4469273

Everything seem to work, but my output shows too much of a voltage drop. Pad A potential value is 1.2 and pad B is -8.

It should be a stupid mistake that I am making somewhere, can anyone give an advice?
I am also not abled to find any examples on Static Current Conduction model except some documentation here https://www.nic.funet.fi/index/elmer/do ... Manual.pdf which I don’t fully understand.
Is there a good place to start learning that?

I am attaching my project.
https://drive.google.com/file/d/1z1YkMB ... share_link

Thanks in advance!
Attachments
Screenshot 2024-05-21 at 14.11.42.png
(90.98 KiB) Not downloaded yet
kevinarden
Posts: 2418
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Static Current Conduction question

Post by kevinarden »

The default values for materials and constants in Elmer are SI units.
The SI unit for distance is meters. You could work in a different unit system but you would have to change al of the constants and values supplied by Elmer.

So Elmer thinks you pad area is 0.895M^2 and that you board is 10 M wide

Adding
Coordinate Scaling = 0.001
to the simulation section of the sif scales your model from mm to M
potential.png
potential.png (29.43 KiB) Viewed 425 times
Tapegoji
Posts: 28
Joined: 12 May 2024, 00:22
Antispam: Yes

Re: Static Current Conduction question

Post by Tapegoji »

I just ran your model with Kevin's suggestion. I ran just fine for me as well.

However, I noticed it took a bit longer that I expected for a dc-conduction simulation. I wonder if you mesh the pcb using hexahedral, the simulation will converge faster. The reason, I am asking is that, when you run the full pcb it may takes a lot of time to solve it.

What do you think?
kevinarden
Posts: 2418
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Static Current Conduction question

Post by kevinarden »

I just change it from iterative solver to direct and it ran much quicker, obviously

Linear System Solver = Direct !Iterative
Linear System Direct Method = Mumps

Windows version may not have Mumps, but it would have banded or umfpack.

Direct may run out of memory, in which case you are stuck with the iterative.

I have found that a hexahedral meesh can be made by meshing the back face in quads and then extruding each area to the desired thickness.
Tapegoji
Posts: 28
Joined: 12 May 2024, 00:22
Antispam: Yes

Re: Static Current Conduction question

Post by Tapegoji »

Thank you Kevin,

I have one more question. In this case the pad and the copper have the same nodes (conformal meshing).

what if the pad and the pcb are different volumes that are touching each other but don't have conformal meshing? like the picture attached.

It seems that there is no contact between the two even though they are touching.
Attachments
Screenshot from 2024-05-22 14-54-21.png
Screenshot from 2024-05-22 14-54-21.png (36.28 KiB) Viewed 395 times
kevinarden
Posts: 2418
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Static Current Conduction question

Post by kevinarden »

Then data will not be shared since Elmer shares data through the nodes. A conforming mesh isn't absolutely required but you can use mortar conditions to join the two boundaries. If there are coincident nodes causing the break the the nodes can be merged in Salome or ElmerGrid

ElmerGrid 8 2 my.unv -autoclean -merge 1.e-5

or if the mesh is already created

ElmerGrid 2 2 mesh -o mergedmesh - merge 1.e-5
sqrtmo
Posts: 2
Joined: 20 May 2024, 11:55
Antispam: Yes

Re: Static Current Conduction question

Post by sqrtmo »

Hello, thanks a lot for all the replays and the followup!

@kevinarden: I did a stupid mistake using google aria converter. So my scaled aria should be .895e-3m^2 instead of .895e-6m^2 : (. Thanks a lot!
@Tapegoji: Thanks for pointing that out!

I am working on Mac and I installed Elmer through macports, it seems that I do not have MUMPS installed as well. I am planning to put effort in compiling it ASAP. However Direct>Umfpack works faster, I will try simulating the whole PCB and see what will happen with the RAM.
I will research the hexahedral meshing options, now I am using gmsh trough FreeCAD, I don't thing that there are different meshing options there.
It seems like SALOME is not available for MAC. If you guys have any proposals will be great!

Thanks again!
Tapegoji
Posts: 28
Joined: 12 May 2024, 00:22
Antispam: Yes

Re: Static Current Conduction question

Post by Tapegoji »

I suggest using the Gmsh GUI as it provides a lot of options. I use the Python API of Gmsh frequently. You can do pretty much anything with it. I believe FreeCAD's Gmsh integration is limited and not fully up to date.
Post Reply