A SIMPLIFIED CASE STUDY

In this case study, we will investigate how CFD simulations (Computational Fluid Dynamics) can improve electronics cooling for a simplified case study of a cabinet with one PCB (Printed Circuit Board).

What is the issue at hand?

The PCB contains 1 GPU and 8 memory chips. They all produce heat and have a maximum allowable junction temperature of 70°C. The junction temperature is the highest temperature in the chip. To remove some of the heat, a fan is installed at the cabinet inlet.

We might have a gut feeling, that this single fan is not good enough. But how bad is it? And what is needed to achieve the required temperature?

How do we solve it?

To evaluate the temperatures and improve the cooling, we used CFD simulations to simulate different scenarios that could have a positive impact. This process involved the following steps:

1. Model creation: We designed a 3D model of the electronic cabinet and its components.
2. Mesh Generation: This step involves dividing the geometry into small elements called meshes. The mesh size and quality have a significant impact on the accuracy and efficiency of the simulation. Expert knowledge and software play an important role at this stage in the process.
3. Physics and solver selection: In this step the relevant physical phenomena and their governing equations is specified. In this case turbulent flow and heat transfer for both air and solids are important physics. The solver is responsible for solving the governing equations. The appropriate solver is selected based on the problem complexity and computational efficiency.
4. Model Setup: Here the physical properties of the air and components are defined, such as density, viscosity, and thermal conductivity. The simulation parameters such as boundary conditions and convergence criteria are also set.
5. Simulation: The simulation is run to obtain results of the temperature distribution and air flow velocity in the system. The results are then analyzed and compared with the desired outcomes.
6. Iterate: Based on the results, we iterated and tried different cooling scenarios until we found the solution, which matched our specific needs.
   

Iterations

The initial simulation with one fan as the single cooling solution, showed a junction temperature of more than 300°C, which is way above the allowable limit as seen on the above picture.

To lower the temperature of 300°C, we added a heat sink to the PCB, which reduced the temperature to around 90°C. As this temperature was still too high, the next iteration added a fan at the outlet as well, which brought the temperature down to below 80°C, and close to the limit.

In the final iteration and attempt to go below 70°C, the PCB was placed in the jet from the inlet fan. This resulted in a junction temperature of 69°C and finally, we had reached an acceptable temperature level inside the cabinet, as seen on the below picture.

What did we learn?

The CFD simulations showed us that adding a heat sink and a fan at the outlet helped reduce the temperature, but it was not enough to keep the temperature below the allowable limit. Only by using a fan at both the inlet and outlet combined with an optimized PCB position, it was possible to lower the junction temperature to an acceptable level.

This simplified case study highlights the importance of using CFD simulations to optimize thermal management solutions for electronics. It also shows the iterative process involved in finding the most effective solution, that allows the designer to validate the efficiency of their product.

Do you find CFD simulations interesting, you can read more about Computational Fluid Dynamics here and if you are interesting in exploring the possible value of CFD within your company, then don't hesitate to reach out to our CTO Jacob: