My first freeCAD Tutorial

[makkemal]

@easy-fc
"Enable backlight color" in your preferences to get VTK to display correctly.

vtk.png (113.59 KiB) Viewed 2513 times

[easyw-fc]

@easy-fc
"Enable backlight color" in your preferences to get VTK to display correctly.

vtk.png

thanks a lot!
here some simulation I did
viewtopic.php?f=3&t=17068&p=135728#p135728

[makkemal]

@easy-fc
thanks a lot!
here some simulation I did
viewtopic.php?f=3&t=17068&p=135728#p135728

Glad you got how it works !
Just a few notes I followed an example on the internet of which the input values are debateable.
For accuarcy I think the convection coeff of the PC board to air value of 19W/M^2K is a bit a high I would reccomend 10W/M^2K.
With the chips so high I would also add convection of 10W/M^2K to the side of all chips
Regards
Michael

[easyw-fc]

@easy-fc
thanks a lot!
here some simulation I did
viewtopic.php?f=3&t=17068&p=135728#p135728

Glad you got how it works !
Just a few notes I followed an example on the internet of which the input values are debateable.
For accuarcy I think the convection coeff of the PC board to air value of 19W/M^2K is a bit a high I would reccomend 10W/M^2K.
With the chips so high I would also add convection of 10W/M^2K to the side of all chips
Regards
Michael

thanks again! I will update my simulation according to that...
EDIT may be with some ground planes that I normally have in my boards, the convection could be increased to the previous used value... do you agree?

PS for kicad users, I posted the FEM results also on kicad forum
https://forum.kicad.info/t/kicad-stepup ... lysis/3828

[makkemal]

@easy-fc
EDIT may be with some ground planes that I normally have in my boards, the convection could be increased to the previous used value... do you agree?

It will depend on the area of the ground planes but genrally i dont think this would make a big diffrence.
The lower your convection coeff the more conservative your answer i.e. the higher the tempearture will be on the board

[easyw-fc]

@easy-fc
EDIT may be with some ground planes that I normally have in my boards, the convection could be increased to the previous used value... do you agree?

It will depend on the area of the ground planes but genrally i dont think this would make a big diffrence.
The lower your convection coeff the more conservative your answer i.e. the higher the tempearture will be on the board

Hi Michael
I noticed that in my simulation
download/file.php?id=26127
the temperature values are from 50°C and 70°C, but I inserted 25°C as in your tutorial...
I attached a simplified version of my board saved in FC 0.17 8318 win10 64b

fem-simplyfied-v1.FCStd

(843.57 KiB) Downloaded 32 times

here, my FemConstraints are 25K as in your tutorial, but the result is from 102 and 136 degrees ...
(I edited .inp file changing the second instance of '*BOUNDARY' to '*CFLUX' and removing '11,' before running the simulation)...

would you mind to see if I missed something or if my simulation is wrong because I'm on windows?
Ranges are wrong, but the color map seems fine anyway..

thank you again
Maurice

[makkemal]

@easy-fc
I just realised that *CFLUX applies that load to every node selected.(It needs to be distributed or only applied to single node)
Thus the finer the mesh or the larger the chip the more energy you end up putting into the system.
On my example i did not notice since I used a very course mesh.

So if you only apply the *CFLUX to a vertex it gives the energy input.
On the attached example the chip temperatures are wrong but the PCB will be close.
If you split chip to create a vertex in the middle of and apply *CFLUX to that vertex, it would be a good approximation.

Regards
Michael

[easyw-fc]

@easy-fc
I just realised that *CFLUX applies that load to every node selected.(It needs to be distributed or only applied to single node)
Thus the finer the mesh or the larger the chip the more energy you end up putting into the system.
On my example i did not notice since I used a very course mesh.

So if you only apply the *CFLUX to a vertex it gives the energy input.
On the attached example the chip temperatures are wrong but the PCB will be close.
If you split chip to create a vertex in the middle of and apply *CFLUX to that vertex, it would be a good approximation.

Regards
Michael

Hi Michael
I tried two simulation with a vertex at the center of the chip and with a small area at the center of the chip
here are the two FC files and the simulation results

fem-simplyfied-v6-c.FCStd

(861.62 KiB) Downloaded 60 times

central-vertex.png (272.06 KiB) Viewed 2407 times

fem-simplyfied-v7-c.FCStd

(881.06 KiB) Downloaded 34 times

central-area.png (299.45 KiB) Viewed 2407 times

Which one would you would consider the best to be used for a simulation?
I tweaked some values, but feel free to trim the simulation as you prefer...

At the moment my target is just to have an idea of the temperature on the board, starting from the maximum temperature measured on the top of the chip, just to verify if the simulation is close to what I can test...

Thanks again
Maurice

[makkemal]

@easy-fc
You are not putting the same energy into the two examples.
In the first one you are putting in 110W
In the second one you put in 9x5 =45W it should be 9x12.222=110W to compare (Check amount of nodes in input file)
If you do that you will see the board temperatures are exactly the same although the chip peak temps differ.

If you are interested in the temperature of the PC board the first one (vertex only) is good enough
Since the chip is very conductive it spreads the energy through the Chip quickly the PCB that has a lower conductivity and spreads heat slower
You can make the chip temperature look more constant by just setting the maximum in colorbar (double click on color bar)

Uniform_temp .png (240.11 KiB) Viewed 2390 times

As far as the heat transfer coeff goes it looks like for electronic chips 10 is still a high value
http://www.electronics-cooling.com/2001 ... lat-plate/
It should be more in the line of 5 W/m^2
In the example when they apply 24W/m^2 it is because the have a heatsink on to of the chip

Regards
Michael

[easyw-fc]

@easy-fc
You are not putting the same energy into the two examples.
In the first one you are putting in 110W
In the second one you put in 9x5 =45W it should be 9x12.222=110W to compare (Check amount of nodes in input file)
If you do that you will see the board temperatures are exactly the same although the chip peak temps differ.

If you are interested in the temperature of the PC board the first one (vertex only) is good enough
Since the chip is very conductive it spreads the energy through the Chip quickly the PCB that has a lower conductivity and spreads heat slower
You can make the chip temperature look more constant by just setting the maximum in colorbar (double click on color bar)

Uniform_temp .png

As far as the heat transfer coeff goes it looks like for electronic chips 10 is still a high value
http://www.electronics-cooling.com/2001 ... lat-plate/
It should be more in the line of 5 W/m^2
In the example when they apply 24W/m^2 it is because the have a heatsink on to of the chip

Regards
Michael

thank you very much for your valuable advises

Maurice