CFD simulation to improve thermal management

”Keep cool“ is not only important in the hot summer months or during crucial project phases. Did you know that the life expectancy of electronic components (i.e. semiconductors) is highly dependent on operating temperature?

With an increase of ten degrees, life expectancy is halved (logarithmic relationship) and the overall performance decreases drastically. Moreover, overheating can lead to fire, resulting in catastrophic failure modes, and needs to be considered in order to comply with regulatory standards. Therefore, thermal management is a key aspect in the development of any electronic product in order to obtain a robust, safe, durable, and high-performance device.

At IMT, state-of-the-art simulation tools are used to address such requirements at a very early stage of the development phase to achieve an optimal product layout and design. However, not only are colorful marketing pictures produced…we also challenge the computational models and verify them with experimental measurements.

Problem

Within the project concept phase, IMT was commissioned to investigate possible options to increase the cooling efficiency of a product. The need for this optimization arose from direct market feedback, as long-term observations showed that the devices should be more robust in terms of thermal management.

Restrictions were imposed by the connector layout for proper use and the resulting internal wiring. Another important consideration was the maintenance requirements. Every component of the device should be easily accessible and exchangeable.

Due to the tight schedule, multiple iterations on prototypes were not an option.

Method

Computational fluid dynamics (CFD) is a branch of fluid mechanics that uses numerical analysis and data structures to analyze and solve problems that involve fluid flows. Using this method, product designs can be investigated in terms of their aerodynamic behavior including thermal effects, without the need for physical prototypes. Thus, design changes and their impacts can be studied in a very early development phase.

Numerical Model

After some research on enclosure layout, a basic concept for the device was worked out taking into account the restrictions imposed (i.e. PCB layouts, cable length, etc.). A numerical model to perform CFD simulations was set up in the Solidworks® flow simulation software to obtain a high-performance prototype at the first attempt. This is intended to save time and costs and is the only way to comply with the schedule laid down. For efficient iterations, the concept of the enclosure was abstracted in a block model approach in order to be able to analyze the device on the system level (Figure 1).

The idea is to increase cooling efficiency for critical components by separating and guiding the airflow at the inlet of the main unit into a primary (blue) and secondary (orange) stream (Figure 2). This allows better tuning of the spatial distribution of the mass flow. Additionally, fresh inlet air can be directly guided to critical areas with high power dissipation to benefit from a higher temperature gradient.

The concept of the enclosure design splits the device into two sections by a “false bottom”, where the fan is located. The mass flow distribution inside the upper part can be adjusted by changing the position and size of the opening at the inlet and outlet of the main device respectively.

Components with higher heat loss such as the power board and power supply are encapsulated and located in the upper part. This not only allows more control over heat accumulation but also supports electrical insulation to comply with standards such as EN 60601-1, EN 61010-1, IEC62368-1, etc. Additionally, the component orientation and outlet position are optimized to take full advantage of natural convection leading to a well-formed flow over the mainboard.

Figure 2: Intended airflow separation/guidance

Pre-Studies

The fan was implemented on the basis of a mathematical description to reduce the required calculation time. Therefore, the fan curve, provided by the manufacturer, was implemented by means of a lookup table. The implementation and mesh dependency were investigated in a preliminary study (Figure 3).

Figure 3: Mathematical model of fan

In order to obtain meaningful results from the simulation, a Mesh Sensitivity Study (MSS) was performed to determine the required spatial discretization of the numerical model. Particularly for thermal loads, such studies are essential in order to achieve the required resolution for the boundary layer where the heat transfer takes place (Figure 4).

Figure 4: Mesh sensitivity study (MSS) on convection

The correct definition of the thermal loads turned out to be the crux of the model. Generally, the heat loss can be approximated by the power dissipation.

For distinct hot spots such as integrated circuits (IC), the calculation is straightforward as long as the required data is available. However, for extended heat sources, such as the power board in this case, another solution had to be found. Using a reverse engineering approach, the block model of the PCB was calibrated. Using thermal imaging, the steady-state temperature during operation in a free convection setup was measured (Figure 5).

Temperature (°C)

Figure 5: Thermal imaging of the power board

The resulting temperature could be used in a corresponding CFD simulation as the boundary condition to determine heat flux, which could then be used in the block model of the device (Figure 6).

Temperature (°C)

With these pre-studies, the physical phenomena involved were investigated and their implementation in the simulation was verified. Now the block model of the device could be set up.

Simulation

First, the initial layout was analyzed in order to understand if the intended principle of airflow separation works and to identify problematic locations.

Velocity (m/s)

Figure 7: Airflow for the initial design

According to the simulation, the general idea has proven successful (Figure 7). However, the detailed investigation revealed a problematic vortex that needed to be resolved. Additionally, a support plate had to be included to increase structural stability. Within hours, the design change was simulated and the effect could be assessed. The integrated support plate leads to an improvement in the vortex and overall heat discharge (Figure 8).

Temperature (°C)

Figure 8: Critical vortices leading to heat accumulation

Compared to the “free convection” measurements, the ventilation of the enclosure was shown to work efficiently with both designs and led to a significant reduction of the temperature of electrical components at the different probe locations (Figure 9).

Satisfied with the results predicted by simulation, the prototype could be released.

At IMT, we not only put trust in our tools, we question them and our work, always striving for evidence rather than belief. Once the prototype was assembled, it was fitted with thermocouples at probe locations from the numerical model and put through a stress test (Figure 10).

Figure 10: Prototype fitted with thermocouples

The simulation results were compared with the experimental measurements showing a discrepancy of around 15 – 25% (Figure 11). This sounds like a lot, but given the high level of abstraction, the results met the expectations.

During development, it was noticed that the concept device would allow the implementation of a larger fan, which could further optimize thermal management. Without further ado, the prototype was adapted to examine the effect of a larger fan. Since the fan control is designed to operate the fan in low and high RPM mode, this parameter was investigated as well (Figure 12).

Most important was the comparison with the current device, which was used as a benchmark. Comparing the internal sensors, which are equivalent in both devices, a tremendous temperature reduction could be achieved for all sensor positions (Figure 13).

Interestingly, the concept device with a large fan achieves a performance equivalent to the configuration with a small fan operating in high RPM mode. This is essential for customers regarding noise emission, as the large fan option was shown to be even more quiet while providing a margin for extreme conditions.

Simulations allow product designs to be investigated in a very early development phase and reduce the need for prototyping. Additionally, physical effects can be visualized, which could not be detected or only with great effort using experimental methods.

This leads to very high efficiency in the development of new products or identification of weaknesses in existing designs while reducing valuable resources such as time and money. Moreover, it facilitates communication between different departments or stakeholders to better understand requirements. This simplifies decision-making in crucial questions…true to the motto: “Keep cool”!