Heat Sink Design and Fabrication

Heat Sink Design and Fabrication

Final Report

T.E.A.M. M.

Ryan Condon

Steve Gehlhausen

David Keene

Todd May

Nathan Piccola

Washington State University – Vancouver

Mechanical Engineering Department

Mech 442 – Advanced Thermal Systems

March 31, 2011

i

Abstract

Modern computer chips require additional cooling efforts in order to keep the internal temperature at an operable level and prevent permanent damage. A common technique to accomplish this is to use heat sinks in the form of fin arrays to increase the overall heat transfer from the chip to the ambient surroundings. It has been decided that one of the aging computers in the fluid and thermodynamics lab is going to be used as a PIV processor, which requires more work and power from the hardware. Because significant speed and power increases the amount of heat generated from a chip, it is necessary to replace the stock heat sink with a new design. Many stages of the engineering process were used to solve this problem, including design, simulation, fabrication and testing. Besides designing for optimal performance, other engineering characteristics were taken into consideration including geometric constraints, ease of manufacturing, cost, and environmental impacts. The results of the project showed the thermal resistance of the heat sink to be approximately 0.3 K/W, an efficiency of 92%, and an overall effectiveness of 16.5. Another important result showed that the estimated combined cost for a professional to fabricate the device would be approximately $260. Other non-numerical results were also considered and it was concluded that they all fell successfully within the scope of the project. Overall, the heat sink meets the design goals but it was not cost effective.

i

Table of Contents page

Abstract i

Introduction 1

Problem Description 1

Design Process 1

Methodology 1

Design Description 2

Fabrication 3

Calculations 4

Analytical 4

Simulations 6

Experimentation 7

Results 9

Conclusion 13

References 14

Appendix A 15

Appendix B 19

Hand Calculations 19

List of Figure page

Figure 2. A sample cross sectional temperature profile of entire heat sink 5

Figure 3. ANSYS Icepak setup for simulations. 6

Figure 4. Velocity and temperature profiles generated by ANSYS Icepak. 7

Figure 5. Heat sink mounted in wind tunnel for experimental tests. 8

Figure 6. Temperature at the top of a fin half way between the edge and center 10

Figure 7. Thermal grease after un-mounting shows the heat sink was not mounted flat. 10

Figure 8. The effect of air velocity on temperature. Power = 32.1 Watts. 11

Figure 9. The effect of power on temperature. Air velocity = 6.0 m/s. 12

Figure 10. Drawings for manufacturing the heat sink. 15

Figure 11. Location of holes drilled for mounting thermocouples. 16

List of Tables page

Table 1. Geometric fin dimensions and properties of air. 5

Table 2. Properties of each trial. 9

Table 3. Analytical calculation results for each state compared. 17

Table 4. Computational simulation results for each state compared. 17

Table 5. Experimentally measured temperatures for each trial. 18

i

Introduction

The objective of this project was to design and build a heat sink to dissipate heat from a central processing unit (CPU) in a computer. The central processing unit is to be upgraded, and the old heat sink will no longer be sufficient. Due to the critical task of cooling the processor, the heat sink must be analyzed using multiple methods before being mounted in the functioning system. Once all these methods are completed the results will be compared and the suitability of the heat sink evaluated.

Problem Description

The CPU was modeled as a thin plate 7.09 mm by 11.25 mm that generated up to 100 watts of power. These values were specified by the customer as a reasonable approximation of the actual CPU. The ambient air velocity ranged from 0.25 to 6.1 m/s, and the ambient temperature range was 18.3 to 65.5 degrees Celsius. The aluminum stock provided for manufacturing was a cylinder 2” in height, and 4” in diameter. Material could be removed from the cylinder, but not added. The primary objective of the heat sink was to keep the CPU below 90 degrees Celsius, and the secondary objective was to remove as much heat as possible.

Design Process

Methodology

The first idea considered for maximizing the heat sink performance was to create the greatest surface area possible. Then every other optimization idea generally evolved from there. The easiest way to increase surface area was to use more volume of material, which is what lead to the decision to keep the circular shape of the cylinder instead of cutting it down to a rectangular cube. Another way to increase surface area was by minimizing channel width and fin width. This allowed more fins to be packed on the heat sink, increasing surface area significantly. The minimum fin width for this design was simply determined by the fabrication limitations. There is an actual minimum width for conduction, but this design did not exceed that due to equipment and material capabilities. Minimum fin spacing was determined by the thickness of the slitting saw used, but there were also calculations done to make sure stagnation between the fins was not an issue. Impeded flow occurs when the boundary layers from adjacent fins significantly interact, so the fin spacing should theoretically be at least double the maximum boundary layer thickness. Boundary layer thickness was determined according to Equation 1.

δ=5xRex (1)

This equation shows that the boundary layer is dependent on both the fin length and the velocity of the coolant. Since the middle fins in the heat sink design are longer the spacing between them should have been wider than the outside fins. One reason this was not implemented was that the minimum spacing that could be manufactured was much wider than needed for the outside fins at all air velocities and for the center fins it would not cause problems at the higher velocities. After discovering this, it was determined that the heat sink would be designed for medium to high air velocities to make sure that it would be able to handle the most extreme heat dissipation needs that the CPU could produce. Another reason the fins all maintained the same spacing was to speed up manufacturing with the tools available.

After the initial calculations, some basic simulations were run to verify the calculations. The simulations indicated that using the smallest width of the fins and spacing was going to get the best performance over a range of conditions involving power, air velocity and air temperature.

Design Description

The base of the final heat sink design retained the original cylindrical shape with a diameter of 4”. The base is 0.5” thick, and 22 rectangular longitudinal fins extend vertically from it (Figure 1). Each fin is 1.5” in vertical height, and the base length of the fin is determined by the location on the base, i.e. the central fins have the greatest base length and the edge fins have the least. The fins were 0.06” thick and the spacing between them was 0.125”. The design schematic is Figure 10 in Appendix A.

Fabrication

One of the steps in the initial design concept phase was to determine the availability and application of tooling for this project. An inquiry was made to the tool room to determine what the technician had available and what was recommended in terms of manufacturability. The technician indicated that a slitting saw with an 1/8th inch width and a 1.5” depth would be available for use. The slitting saw is a much better tool for high aspect ratio features, like vertical fins, than the typical endmill. An endmill is subjected to higher moment loads and decreased chip flushing (typically referred to as loading) as the aspect ratio increases. Both of these conditions lead to premature tool failure and also decrease tolerance and clarity of features.

In the interest of maximizing surface area, the round stock was used rather than removing material to create a square profile. The round stock had some manufacturing issue to overcome as the vise used on the milling machine is designed for parallel surfaces. However, the only time the fin was clamped on the rounded surface was when the mounting holes and counterbore clearance were added to the material. For the rest of the milling process while using the slitting saw, the vise was clamped onto the bottom and top surfaces, which are parallel. This allowed for a fairly rigid setup that improved manufacturability. Towards the end of the slitting process, as the slitting saw started becoming dull, the stock did start to rotate and slide. This caused a few fins to become distorted. The aluminum was removed from the vise and reoriented. The rest of the manufacturing occurred with slower feed but slightly increased speed to decrease the chip load per tooth.

The heat sink was manufactured over five visits to the machining lab. The total time involved was approximately 13 hours. This included a general approximation of a half hour for setup and a half hour for cleanup per trip. The total redundant time lost for the project was 4 hours. If this lost time was eliminated by starting and finishing the project in one window, the time required for a student with the available tooling would be 9 hours. If this heat sink was manufactured in a commercial machine shop, using gang saws and a proper clamping fixture, this time could probably be reduced to 2 to 3 hours. Using a generalized shop rate of $75 per hour, the labor cost for this fin would be approximately $150 to $225 for a professional machinist. For the student speed (without the lost time) at the same $75 per hour, the cost is $675. The stock cost from McMaster-Carr for the aluminum (6061-T6) is $25.34 plus another $10 for shipping, for a total of $35.34 [1]. This price can be reduced by ordering more stock and cutting it to length for multiple heat sinks rather than the single piece price shown. The estimated combined cost would be $260.34 for a professional and $710.34 for a student.

The environmental impact of this heat sink could have been fairly low. A slight amount of oil was used in the manufacturing and then the final product was cleaned in Simple Green to eliminate the oil residue, which would impede the heat transfer ability of the heat sink. As a material, aluminum is very easy and economical to recycle. The WUSV machine shop does not recycle the scrap from manufacturing so that harms the environment, but it would not have the same impact in a manufacturing facility that does recycle. As a whole, aluminum has a very good life cycle, other than the high initial energy cost used in refining the bauxite. At the end of the life cycle, since the heat sink uses air as the transfer medium, no chemical buildup is created which would make recycling difficult. When compared to a closed loop cycle that uses a fluid medium, there are fewer products to consider for disposal. Since there are no moving parts requiring energy, this heat sink design has no life cycle energy requirements and it is much less likely to wear out and fail.

Calculations

There were three methods used to evaluate the performance of the heat sink design. The first method was analytical. The second method was computational simulations. The third method was to actually build the heat sink and measure its performance experimentally. All three methods are important in the design of a product. Simulations are simple because the problem just has to be setup for a computer to solve. Because simulations can sometimes give answers that do not make sense, it is important to verify the simulation outcomes analytically. Actually building the heat sink and testing comes last since the cost of building is expensive and it is not practical to make many prototypes. It is cheaper to run the simulations first and then build only a couple times. Building and testing is an important step because it can reveal issues not apparent in simulations or computational results.

Analytical

After coming up with a general idea for the heat sink design, analytical calculations were the first step in determining if it was a viable option. Pretty quickly, it was determined that a basic longitudinal heat sink was the best option after evaluating performance and manufacturing capabilities.

More advanced analytical calculations were then setup in a spreadsheet to allow variables to be changed and provide updated calculations instantly. This was a major component for many of the following steps such as deciding fin dimensions, determining the approximate limits of performance, and determining what conditions to run the experiments at. The first part of the analytical calculations was setting up a table with all the substance constants, dimensions, and surrounding conditions (Table 1).

Table 1. Geometric fin dimensions and properties of air.

Fin thickness / 0.001524 / m
Fin height / 0.0381 / m
Fin spacing / 0.003175 / m
Base thickness / 0.0127 / m
Al diameter / 0.1016 / m
Q / 32.1 / Watts
T∞ / 23.3 / C
U / 2.061 / m/s
k al / 167 / W/m*K
ρ / 0.995 / kg/m^3
Μ / 0.00002082 / Pa*s
Cp / 1.009 / kJ/kg*K
k air / 3.00E-02 / W/m*K
Pr / 0.7002460 / -

One of the challenges of keeping the cylindrical shape was that each fin had to be calculated individually since their base lengths were all unique. In the spreadsheet, there was a table with each column representing a fin. Each row was a different property of the heat sink. Eventually, the thermal resistance for each fin was calculated. Using a complex resistor network, all the resistances were added together to obtain the total resistance of the heat sink design. By specifying the amount of heat the CPU would dissipate, the temperature profile for the base and each fin was calculated. Figure 2 represents a sample cross sectional temperature profile for the heat sink. Each column is a fin, and each row is a different height up the fin. Sample calculations for the whole analytical section are included in Appendix B.