Thermal simulation of PCB

The graphical user interface of Elmer
Post Reply
jssidi
Posts: 3
Joined: 18 Jul 2017, 11:25
Antispam: Yes

Thermal simulation of PCB

Post by jssidi »

Hi all. I'm starting to work with Elmer and I've encountered some issues.

I'm trying to simulate the thermal behaviour of a PCB board with a CPU that generates the heat. I set the case as I think it's correct but the results are not satisfactory (The CPU reaches 10^5 K and the rest of the board remains at 0).

I'm using the default heat equation, with a volume source of heat applied to the body of the CPU. For its value I assume that it dissipates ~0.4125mW/mm^3 = 146.22W/Kg, so I set the value to be 146.22 (Is this correct?). Furthermore, the initial condition and some boundaries are at room temperature. I have checked the dimension of the model with FreeCAD and it's right.

The .sif file is attached but the mesh is to big to upload. Anyone knows what is wrong?

Thanks in advance.
Attachments
case.sif
(3.11 KiB) Downloaded 405 times
mzenker
Posts: 1999
Joined: 07 Dec 2009, 11:49
Location: Germany

Re: Thermal simulation of PCB

Post by mzenker »

Hi,

without checking in more detail, the most frequent error sources in similar cases are the following:

* Dimensions: The coordinates in CAD and mesh are in mm, whereas Elmer assumes m
* Nonconformal mesh: between the CPU and the board, there may be a duplicated mesh with different nodes on both sides. In that case, the heat cannot pass from CPU to board. Which mesher did you use?

HTH,

Matthias
jssidi
Posts: 3
Joined: 18 Jul 2017, 11:25
Antispam: Yes

Re: Thermal simulation of PCB

Post by jssidi »

Hi mzenker, thanks for your help.

For the mseh generation I used Salome, stacking all the layers together and adding a box as CPU. I then made a partition including everything and made groups. Finnally Salome generates the mesh with 1D2D3D Netgen algorithm and I create groups within the mesh and export it to a .unv file. Elmer reads the file correctlly and all the different bodies are in there.

For the units, I have made sure that the geometry model is correct measuring it in FreeCAD and it gives me 34x44mm. Are you saying that Elmer is taking it as a 34x44m board?

Sorry if this are stupid questions, but I'm new with Elmer. I have been reading for over a week and this is what I have undertood of it.
mzenker
Posts: 1999
Joined: 07 Dec 2009, 11:49
Location: Germany

Re: Thermal simulation of PCB

Post by mzenker »

Hi,

if you proceeded like you wrote for the meshing, the mesh should normally be conformal. But it's strange that there seems to be no heat flow from CPU to PCB.

You can check the model dimensions in ElmerGUI by examining the output of Model->Summary. The values given there are in m. So if Elmer sees a 34mx44m board, you need to write

Coordinate Scaling = 0.001

in the free text field under Model->Setup, in the Simulation section.

HTH,

Matthias
jssidi
Posts: 3
Joined: 18 Jul 2017, 11:25
Antispam: Yes

Re: Thermal simulation of PCB

Post by jssidi »

Hello again Matthias.

Your help has been trully great, by scaling the coordinates now all the results are in reasonable ranges.
As a final question, I'm supossing it only modifies the coordinates and therefore I don't need to change the value of the rest of the constants in the problem (Heat conductivity, capacity, etc...), right?

Thanks again for your help.
mzenker
Posts: 1999
Joined: 07 Dec 2009, 11:49
Location: Germany

Re: Thermal simulation of PCB

Post by mzenker »

Hi,

glad to hear that it works now for you.
Yes, all the constants are in MKSA, so if your mesh dimensions are in meter (or converted to be by Cooordinate Scaling), you shouldn't need to change anything else.

Matthias
Post Reply