Enclosure electronic cooling : CPU temperature is way too high then predicted

[sbingDU6EK]

Hi,

I'm working on an enclosure design for an embedded PC assuming a 7.68W for the CPU and about 4W for the rest of the board. All the others bodies are set to heat sink. The board is in a close aluminium enclosure, an heat transfert block and transfert plate are attached to CPU and enclosure body. Enclosure is to be wall mounted, so gravity has been set accordingly. I can't get realistic results of out my simulations, CPU temperature approach 700°C, while the enclosure body are nearly at room temperature...

So to check my design, I've tried other simulation software, SimScale, and I got more realistic results.

I would really like to know how to get Fusion 360 results more realistic so I can work all the design in Fusion 360. I'm a bit crossed, I've lost many simulation credits for bad results...

I've remarked that in SimScale, I need to imprint parts that are partially in contact to create full surfaces of contact. Is this the same in Fusion 360 ?

Also, what type of material should I use for the CPU, I've use discret component... I've create a Tin body between CPU and aluminium block to simulate thermal paste, but maybe I'm wrong. In SimScale I could define a thin layer of thermal paste of 8W/(m.K)...

I must say it is my first go at Fusion 360 Electronic Cooling...

Any help would be welcome. Regards !

Samuel

[heath.houghton · Accepted solution]

I have a couple of thoughts on what can cause higher than expected temperatures for e-cooling simulations.  One thing to try is to increase the resolution under settings.  Since your main mode of lowering temperature is actually by spreading heat on the board, we want to capture that thickness accurately.  Under Manage>Settings, drag the accuracy slider all the way to the right towards accurate. 

Another thing and likely the most impactful is that both you and the software are making some assumptions about material. For the CPU, I'd change the selection to something more conductive than FR4.  There is a material called "discrete component" which gives a 20x higher thermal conductivity than pure FR4 meant to approximate the conglomerate thermal conductivity of an IC.  This by itself will act to normalize the temperature of the CPU, thus lowering the maximum temperature. 
The board in this model would act as a heat spreader in real life.  The board can have quite a bit of copper in it.  Currently, the default assumption in e-cooling is to assume pure FR4 for the board.  You can create a material that approximates the thermal conductivity of the combined copper/FR4 content of the actual board to get a more realistic spreading effect.  This will lower the maximum temperature as well.

I think the assumptions (very conservative) on material properties are your main culprits contributing to the elevated temperatures.  

I hope this helps, and if so, please post the results.  We would love to see you got a good outcome.

[sbingDU6EK]

Hi Heath,

I do have assigned high conductivity material to PCU : I have used "Tin" as material. Parts name could be confusing, like CPU PCB that is in fact a small PCB board on witch the CPU is mounted...

I have assigned heat sink to all non power emitting parts, I'll try assigning heat sink only to the aluminium block in contact with the CPU.

I have try to create a custom material for the PCB, but I can't assign custom material from inside simulation. I need to do it from the design work place. Since my parts were simplified, I did not wanted to go trough the process again. I guess I'll have to start from scratch to test all the solutions.

I didn't know I could raise the simulation precision. It was set in between high and low for my previous simulations. I'll try at high precision.

Thanks for the tips, I'll post back the results.

Here are a picture of simulation out of SimScale which gave more realistic results with same materials, ambiant temperature, etc. :

[sbingDU6EK]

So I've redo my simulation with those parameters :

• Simulation accuracy et to maximum "Accurate" ;
• Heat sink applies only to the aluminium block touching the CPU ;
• CPU material set to "Tin" for high conductivity ;
• PCBs material set to custom material "FR4 + 10% copper" which is more heat conductive ;

Results are good and seems to be realistic. I think it have more to do with accuracy and/or heatsink only applied to the aluminium block touching the CPU. I will not try to only one of this because I don't want to spend more credits for now. Maybe in another simulation...

Thank you for your help !