CFD analysis helps optimize heat sink designs for desktop computers

Bob Dehoff (Manager, Business Development), David Fast (Thermal Engineer), and Dr. Chong-Sheng Wang ( Thermal Development Engineer), Tyco Electronics

eeProductCenter (01/06/2006 4:53 PM ET)

The accelerated release of faster desktop computers being sold at lower prices into the market place has forced thermal engineers to develop and optimize thermal solutions that not only perform better thermally and acoustically, but deliver more value with less weight than previous generations. The continuing reduction in form factor has also increased the challenge placed on today's thermal management designs.

To meet these challenges, Tyco Electronics utilized Computational Fluid Dynamics (CFD) analysis to optimize both cost and thermal performance of next generation radial folded fin (RFF) heat sinks. This optimized thermal solution approach produced a scalable design capable of meeting the rapidly evolving requirements of the computer market.

Conventional folded fin heat sinks (Figure 1) were very successful solutions for Pentium 3 and Pentium 4 processor-based desktop computers. They were successful because they provided very good thermal performance; were compatible with high volume manufacturing; provided low cost relative to performance, and were relatively easy to design.

Most conventional designs were capable of providing a thermal resistance of less than 0.3°C/W utilizing an axial fan while maintaining all of the required parameters. They started to lose their advantages when motherboard components required increased air flow to cool capacitors and VRMs (voltage regulator modules). This change required engineers to design heat sinks that bifurcated the air to both the heat sink and motherboard. This was not easily accomplished with conventional folded fin or extruded fin heat sinks.

The typical air flow patterns were either across and/or through the fins. These air flow directions did not allow air to cool any components next to the heat sink. Only components at the two air outlets were cooled. To help alleviate this problem, the fins were either perforated or slits were introduced as a secondary means of routing airflow. This had limited success and added cost. To better address these issues, the RFF design was introduced. (Figure 2)

The original RFF designs were not thermally optimized and were heavier than what was allowed by most OEMs. The original designs had the fins mounted to a solid base plate. While these first designs maintained most of the original attributes of the conventional folded fins, they created weight issues and did not provide optimal thermal performance. The design challenge was to take this promising configuration and optimize it to maintain all of the positive points and rectify its deficiencies.

Developing a CFD model

Tyco's thermal engineers relied heavily on CFD analysis to optimize fin material, fin thickness, fin style, number of fins, and center core diameter while maintaining the required specification of thermal performance, weight, volume, and cost. The first step was to develop a CFD model that allowed correlation of model results with actual test results.

The simple geometry of the design allowed the model to be created from within the CFD program instead of importing it from a drawing package. Also, the size of the model was minimized by modeling only a quarter of the heat sink by using the symmetrical nature of the design. This saved modeling time and reduced the computer run time for each of the cases that were evaluated. The model took into account that every fin was at a different angle from the core of the heat sink. (The model configuration is shown in Figure 3.)

The CFD model input parameters were as follows:

• Thermal conductivity: Aluminum 1100 to 220 W/m-K; Copper 110 to 380 W/m-K
• Ambient temperature: 23°C
• Altitude: sea level
• Fan curve: 90 mm x 20 mm AVC Model DA09020B12UP001
• CPU power: 95 watts, 28.7 mm x 28.7 mm Lid

Initial prototypes were fabricated to validate the model’s accuracy. The units were tested using a Cedar Mill thermal test vehicle (TTV) provided by Intel Corporation. The testing protocol followed Intel's recommended test setup and procedure for testing desktop thermal solutions. This procedure can be found at www.intel.com/design/pentiumxe/designex/306830.htm.

Figure 4 shows a unit under test. The units were tested and compared to the CFD modeling results, which were within 5 percent accuracy of the thermal test data from the samples. Table 1 compares the CFD results to measured thermal test data from three different samples.

Note 1: case 1 (33 mm diameter copper core with 104 aluminum fins of 0.4-mm thickness); case 2 (33 mm diameter copper core with 122 copper fins of 0.4-mm thickness); case 3 (33 mm diameter copper core with 104 aluminum fins of 0.4-mm thickness, fins are open at outside edge). Note 2: case 1 heat sink will have ca = 0.226 C/W with Intel Pentium D Processor 830

The original goal was to develop a heat sink that had all of the manufacturing advantages of a folded fin heat sink while taking advantage of the performance benefits of the RFF design. The heat sink constraints are listed below:

• Heat sink Volume: 90 x 90 x 67 mm
• Weight: 550 grams
• CA: 0.30°C/W with Cedar Mill TTV
• Selling price: less than $10.00

With the model validated, several parameters were changed to fully optimize the RFF design. Each parameter was changed while the other parameters were fixed. The first parameter to be determined was the fin material. The model showed that copper had only a 4% improvement over aluminum. Factoring in the added cost and weight of copper, aluminum was chosen as the fin material. The fin thickness was modeled using 0.3 to 0.7-mm thick material.

The results for fin thickness are shown in Table 2, which clearly show that a large improvement is gained by moving from a 0.3-mm thick fin to a 0.5-mm thick fin. It also shows that going above 0.5-mm thickness has only a slight improvement and is not worth the added cost or weight gain. As a result, the fin thickness of 0.4 mm was chosen based on the model results, weight, cost and performance.

The next parameter optimized using the CFD model was the number of fins. The results are shown in Table 3. The model results show an improvement in thermal performance by adding additional fins. The improvement gained starts to decrease at 104 fins. Given the other requirements, 104 or 96 fins were chosen as the number of fins depending on the application.

The center core diameter was the last parameter to be modeled and had the greatest impact on weight and cost of the final solution. The model showed that a 33-mm diameter was required to achieve our overall thermal performance goal while not going over the weight or cost targets. The results are shown in Table 4.

With the design parameters optimized, production-level tooling was built to prove out the design and perform capability studies. The folded fin, fan housing, and attachment bracket tooling was developed and fabricated to the optimized parameters, as were the proper assembly fixtures. The fan, springs, retention clip, and screws are commercially available products.

Various trial runs were made to determine the proper amount of flux/solder required and optimized furnace profile. The initial production units were produced with the developed process. These units were thermally tested using a TTV (thermal test vehicle) to simulate a CPU as described above. The test results showed that a stable manufacturing process was achieved and the product was ready for mass production.

Utilizing CFD analysis is a valuable tool to improve the thermal performance of a radial folded fin heat sink. CFD enabled Tyco Electronics to reduce the design and prototype cycle which reduces development costs and allows Tyco Electronics to introduce a fully-optimized cost-effective heat sink for the desktop market.