Journal of Mechanical Science and Technology 25 (8) (2011) 2003~2012 www.springerlink.com/content/1738-494x DOI 10.1007/s12206-011-0531-8 Experimental and CFD analysis of heat sinks with base plate for CPU cooling R. Mohan 1,* and P. Govindarajan 2 1 Faculty of Mechanical Engineering, Sona College of Technology, Salem, Tamilnadu, India 2 Principal, Sona College of Technology, Salem, Tamilnadu, India (Manuscript Received October 1, 2010; Revised April 18, 2011; Accepted May 19, 2011) ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Abstract Experimental and theoretical investigations of the thermal performance of a variety of heat sinks have been made. The heat sinks investigated were: straight finned, elliptical finned, small pin finned, circular disc finned, elliptical disc finned, frustum finned and double base straight finned. Realistic, manufacturable geometries are considered for minimizing thermal resistance at low velocity. The experimental results of several of the simple geometry heat sinks have been compared to those predicted by a commercially available computational fluid dynamics code fluent. The parameters such as fin geometry, fin pitch and fin height are optimized primarily in this paper and a second task is carried out to optimize base plate thicknesses, base plate materials and modify design of heat sink for improving the thermal performance in the next generation. Although the performance of heat sink is good, the temperature of heat sink at center is high. In this research work, the best heat sink geometry is selected and modified in order to reduce maximum temperature distribution and hot spots of heat sink at center by changing the geometry design and adding one more base. It is observed that flow obstructions in the chassis and the resulting air recirculation affect the heat sink temperature distribution. Keywords: Computational fluid dynamics; Disc plate heat sink with core; Double base heat sink; Forced cooling of electronic devices ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- 1. Introduction With the rapid development of electronic technology, electronic appliances and devices now are always ever-present in our daily lives. However, as the component size shrinks, the heat flux per unit area increases dramatically. The working temperature of the electronic components may exceed the desired temperature level. Thus, promoting the heat transfer rate and maintaining the die at the desired operating temperature have played an important role in insuring a reliable operation of electronic components. There are a number of methods in electronics cooling, such as jet impingement cooling [1, 2] and heat pipe [3-5]. Conventional electronics cooling normally used forced air cooling with heat sink showing superiority in terms of unit price, weight and reliability. To design a practical heat sink, some criteria such as a large heat transfer rate, a low pressure drop, and a simpler structure should be considered. Porous-channel heat sinks or heat sinks combined with porous structures have been also suggested to improve the thermal performance of heat sinks [6-8]. Among various types of heat sinks, plate-fin and pin-fin heat sinks are widely used owing to their own advantages. The plate-fin heat sink This paper was recommended for publication in revised form by Associate Editor Dae Hee Lee * Corresponding author. Tel.: +91427 4099706, Fax.: +91427 4099888 E-mail address: rmohan12@gmail.com KSME & Springer 2011 has the advantages of a small pressure drop, a simple design and easy fabrication. On the other hand, the pin-fin heat sink has the advantages of a high heat transfer rate due to the redeveloping regions and an even thermal performance independent of the direction of the fluid flow [9, 10]. Recently, Kim et al. [11] revealed that the effective heat sink type between plate-fin and pin-fin heat sinks could be determined depending on the pumping power and heat sink length. In addition to above research activities, there has been a new attempt to combine the advantages of plate-fin and pin-fin heat sinks [12]. Usually, these types of heat sinks have several cuts, termed cross-cuts in the present study, perpendicular to the direction of the fluid flow. Xu et al. [13] demonstrated a new silicon micro channel heat sink composed of longitudinal micro channels and several transverse micro channels that divide the entire flow path into several independent zones. Experiments for strip fin heat sinks were performed and empirical correlations were proposed to predict the Nusselt number and pressure drop [14, 15]. Noda et al. [16] performed numerical simulations of heat sinks with a cut-fin shape to clarify the effect of a transverse cut on the thermal performance of the heat sinks. Amon et al. [17, 18] studied the fluid flow and heat transfer characteristics in communicating channels that have geometry similar to that of cross-cut heat sinks. Lee and Mahalingam [19] used Flotherm code to simulate detailed flow and temperature fields within a computer chassis with two fans. Simi-
2004 R. Mohan and P. Govindarajan / Journal of Mechanical Science and Technology 25 (8) (2011) 2003~2012 lar work was also done by Wong and Lee [20]. Subramanyam and Crowe [21] described evaluating designs of electronic cooling heat sinks by using thermal finite element analysis (FEA) and computational fluid dynamics (CFD). Linton and Agonafer [22] simulated an entire desktop PC with one fan using Phoenics code. Yu and Webb [23] analyzed the flow and heat transfer inside a computer cabinet for the high power conditions expected in desktop computers. In this research, CFD (Icepak) has been used to identify a cooling solution for a desktop computer. The 40W PCI card, different case fan size and different ducting positions have been studied. Chang et al. [24] report the results of CFD analysis to cool the 30W socketed CPU of a desktop computer with minimum air flow rate and minimum heat sink size. In the paper the methodology of CFD analysis for the heat sink, and duct design has been described as well as the experimental procedures to validate the predictions. Lober [25] discussed some thermal management considerations involved in choosing an enclosure and demonstrated the use of CFD thermal modeling software which optimally integrated a computer system into an existing enclosure and reduced the design cycle time. In this research CFD (Icepak) has been used to identify a cooling solution for a desktop computer. In the paper the methodology of CFD analysis for the heat sink, and duct design has been described as well as the experimental procedures to validate the predictions. In future desktop computer systems, the CPU will dissipate in the range 100-140 W. AGP card, memory, chipset and peripherals such as hard disk drive will dissipate more power and also air moving devices such as fans and vents must be located appropriately to deliver air to critical components. In addition to the higher CPU power dissipation, the CPU die will be smaller so it results in a significant increase of the interface thermal resistance between the CPU die and the heat sink. The thermal resistance of the computer packages increases, air flow rate and size of the system decreases and the performance of the CPU is decreased accordingly. All of the changes expected in future desktop factors will make air cooling of CPU more difficult. The heat sinks with base plate are placed that will minimize the thermal resistance and increase the performance of the system. In this paper the base plate thickness, fin thickness, base plate material are optimized and to analyze the flow and heat transfer inside a computer for the high power conditions. The purpose of the present work is two-fold. First, obtain experimental data for the thermal characteristics of some individual heat sinks to enable a generalized comparison of the differences between the heat sinks. Second, develop computational models for the heat sinks with the simplest geometries and compare the results with the experimental results to verify the fidelity of the models. 2. Experimental setup and procedure This test setup is not the whole computer chassis system, but a smaller domain in order to simplify the experiments. Fig. 1. Schematic arrangement of experimental setup. An experimental setup is arranged with an electrical heater of size 25 x 25 mm as a heat sources to imitate a processor and it is supplied by direct current power supply. A heater strip was mounted on a piece of circuit board which was, in turn, mounted on a piece of insulation. The insulation was machined so that the top surface of the heater lies just below the top surface of the insulation. The insulation was drilled or milled to allow routing of power and thermocouple wires to minimize free stream interference in the vicinity of the heat sink. The heater is fixed with heat sink using anabond compound with thermal conductivity of 0.437 W/m C and wakefield thermal compound with thermal conductivity of 0.735 W/m C is applied to the contact surface of the heat sink to make proper surface contact between heat sink and the heater. The heat sink rejects the heat into air which is enhanced by placing the blower. The area covered by the resistance coils (on the heat strip) is less than the area of the base of the heat sink, so that the bottom surface of the heat sink completely covers the active heater section. This ensures that the entire power output of the heater is dissipated through the heat sink. Fourteen J-type thermocouples were used for temperature measurement. To measure the maximum temperature of the heat sink, seven thermocouples were mounted through 5mm deep holes at the base plate of the heat sink. They are 7 mm apart. The first thermocouple is positioned 8mm from the leading edge of the heat sink and the last one is positioned 8 mm from the end of the heat sink. Fig. 1 shows the schematic representation of the arrangement for the experimentation. The 25 x 25 mm as a heat sources to imitate a processor is attached to different heat sinks models and heated with heat loads 80W. Since the test setup is an open domain, the atmospheric temperature is the temperature of the air blown to the heat sink. The atmospheric air is passed around the heat sink which is heated then is exhausted by blower and also the pressure drop has been noted. In this setup other heat sources are not considered for simplifying the experiments. In this experimental setup the desired volume flow rate of the air was generated by a suction type blower in steadily of CPU fan and exhaust fan. The steady state was assumed that the change in the maximum temperature of the heat sink was smaller than ± 1 C for a period of 3 min. The temperatures are recorded and used to calculate the thermal resistance of the heat sink.
R. Mohan and P. Govindarajan / Journal of Mechanical Science and Technology 25 (8) (2011) 2003~2012 2005 Table 1. Interior conditions. Object name Material Heat dissipation CPU SILICON 80 CPU Heat sink AL CU - CD AL 15 DVD AL 15 HDD AL 20 Power supply POROUS 75 Miscellaneous card FR4 20 Table 2. Fan conditions. Fig. 2. Computer chassis model. Name of the fan Pressure rise Heat flow rate CPU Heat sink fan 30 Pa 30CFM Case fan 40 Pa 40CFM The wind tunnel is an open circuit type consisting of a variable speed 1/2 horse power direct current motor powering a squirrel cage blower. Air velocities were measured from ports located on the sides of the test section tube. Velocity measurements were made with a Dwyer model 471-2 digital thermo-anemometer. Temperatures were measured with thermocouples and temperature data was collected at the locations upstream of the heat sink, and at the base of the heat sink. 3. CFD simulation approach The CPU heat sink with base plate is attached to the CPU together with a fan. The mainboard and all the other components are enclosed in a chassis. There are many other heat sources in addition to the CPU. Some of them are on the mainboard (e.g., northbridge chip), some of them are attached to the mainboard (e.g., memory modules) and some of them are in the chassis volume (e.g., DVD). The CFD 3D chassis model is shown in Fig. 2. The chassis is modeled using standard dimensions of a common ATX chassis by hollow blocks, and internal components are represented as lumped objects. During modeling, all the components inside the chassis are standard sized components and exact dimensions are obtained by measurement. The CPU is modeled as a 2D area which dissipates 80W. The 25mm x 25mm cross sectional area of CPU is taken, which is commercially available AMD CPU. For simplicity, the mother board, chipset card are modeled as zero thickness with heat generated uniformly. The CPU fan is modeled as a lumped parameter model and does not have blades. Ram cards are fixed on the motherboard. They are also heat sources and accurate dimensioning of space between ram cards is difficult. Therefore, these things are not considered for study. Power supply is a very complex geometry which includes many electric components, wiring and heat sinks. It is assumed as a lumped media which exerts a resistance on the cooling air flow streams. SMPS (switch mode power supply) and few miscellaneous cards are modeled and many small electronic Fig. 3. Surface grid on one of the CPU heat sink. components on these cards are not modeled. The computer cases have small holes which are used to allow inlet air for cooling and discharge hot air through outlet. The modeling of these holes in accurate dimensioning is difficult and computationally expensive. Therefore, it is modeled as a zero thickness flow resistance. The HDD (hard disk drive), DVD, CD are modeled as solid blocks generating a specified amount of heat uniformly inside the volume. The scope of this study is investigation of temperature distributions on CPU heat sinks. The thermal boundary conditions for the objects inside the chassis are listed in Tables 1 and 2. A total of 225W of heat is dissipated. The fans inside the domain are modeled as circular surfaces which add momentum source to the flow. The added momentum source is given as the pressure rise across the fan versus the flow rate curve. The relationship between the pressure and the flow rate is taken linearly. The boundary condition for the power supply is different. The power supply is geometrically very complicated. Therefore, it is modeled by simplifications. The power supply is a rectangular box which is a resistance to flow. The
2006 R. Mohan and P. Govindarajan / Journal of Mechanical Science and Technology 25 (8) (2011) 2003~2012 resistance is different in the y-direction. A closer view of the surface grid on one of the CPU heat sinks is shown in Fig. 3. 3.1 Governing equations and boundary conditions Time-independent flow equations with turbulence are solved. The viscous dissipation term is omitted. Therefore, the governing equations for the fluid flow and heat transfer are the following form of the incompressible continuity equations, Navier Stokes equations x-y and z direction momentum, and energy equations together with the equation of state. The continuity equation: ρ v = 0. The X, Y, Z momentum equations: p τ τ xx yx τzx ρuv = + + + + Bx, x x y z p τxy τ yy τzy ρvv = + + + + By, y x y z p τ τ xz yz τzz ρwv = + + + + Bz. z x y z The energy equation: ρhv p = v+ k T +Φ+ S h Equation of state: P = ρrt ( ). where ρ is the density, u, v and w are velocity components, v is the velocity vector, P is the pressure, B terms are the body forces, h is the total enthalpy and τ terms are the viscous stress components. Reynolds averaging is employed to handle the turbulence effects. In the Reynolds averaging, the solution variables are decomposed into mean and fluctuating components. For the velocity components u = u + u, where u and u are the mean and fluctuating velocity components for x direction. Likewise, ' for the pressure and the other scalar quantities Φ=Φ mean +Φ where Φ is a scalar such as pressure or energy. The Reynoldsaveraged Navier Stokes equations are solved together with the Boussinesq approximation. While the Navier Stokes equations are solved inside the domain, no-slip boundary condition is applied to all the walls in the domain. Therefore, at all of the surfaces u = v = w = 0. It is assumed that the system fan does not drive a flow cell around the computer chassis and the heat transfer mechanism at the chassis outer walls is natural convection. Heat transfer coefficients at the outer walls are estimated from the empirical correlations. To use the correlations, the average wall temperature must be prescribed. To do that, a first cut analysis must be run. As the typical values of the natural convection heat transfer coefficient lie between 2 and 25 W/m 2 K, a value of 5 W/m 2 K is selected to be the heat transfer coefficient at the computer chassis walls. The analysis result by taking the ambient temperature as 27 C gives an average temperature of 33 C at the walls, and then heat transfer coefficients are calculated using this value and the available correlations with the uniform surface temperature assumption, and with the definitions of the Rayleigh number and the average Nusselt number as: ( s ) 3, gβ T T L Ra = αυ hl Nu = k where is L the characteristic length, k, h, g, β, α and υ are the fluid thermal conductivity, the convection heat transfer coefficient, the gravitational acceleration, the volumetric thermal expansion coefficient, the thermal diffusivity, and the kinematic viscosity, respectively. T s and T are the surface and the ambient temperatures. Here, Ra is less than 10 9 ; therefore, the flow is laminar. Using the correlations for laminar natural convection on the vertical plate by taking the thermal conductivity of air as k = 0.027 W/mK, the heat transfer coefficient is 4 W/m 2 K. Similarly, for the horizontal top plate the Rayleigh number is calculated as 1.5 x 10 5, where the characteristic length is calculated from L =A/P, where A is the plate surface area, and P is the plate perimeter. The average Nusselt number is correlated to the Rayleigh number with Nu = 0.54 Ra 0.25 which gives 0.04 W/m 2 K. The calculated heat transfer coefficients are applied to all of the exterior walls of the chassis except the bottom horizontal wall which sits on the ground that is considered to be adiabatic. 3.2 Selection of heat sink design In this work, mainly instead of increasing the base height, the different base plates are attached with heat sink to enhance the performance. The straight finned, pin finned and elliptical finned heat sinks of variable fin thickness and base plates are used to cool the CPU. The 54mm x 65mm heat sink base plate size is selected for this work. The different shape of extruded fins and a 3.5 mm thick base plate is finished of aluminum materials. In addition, to enhance the heat transfer 2.5mm to 5 mm thick copper and ccc base plate has been provided as a spreader to conduct heat from CPU processor. For all fin geometries 2.5 mm fin pitch, 40 mm fin height, 54 mm x 65mm heat sink base plate, there is 5mm clearance between the power supply and the fin tips of the heat sink at the flow
R. Mohan and P. Govindarajan / Journal of Mechanical Science and Technology 25 (8) (2011) 2003~2012 2007 Table 3. Heat sink combinations with corresponding response. Experimental No. Fin height No of fins Fin pitch Base height Thermal resistance 1 15 30 4 3.5 0.257923 2 30 30 4 3.5 0.249132 3 50 30 4 3.5 0.285937 4 15 26 4 3.5 0.265322 5 30 26 4 3.5 0.257022 6 50 26 4 3.5 0.184986 7 15 22 4 3.5 0.259425 8 30 22 4 3.5 0.176525 9 50 22 4 3.5 0.201964 10 15 40 2.5 3.5 0.327302 11 30 40 2.5 3.5 0.318745 12 50 40 2.5 3.5 0.249761 13 15 36 2.5 3.5 0.316159 14 30 36 2.5 3.5 0.209891 15 50 36 2.5 3.5 0.221920 16 15 30 2.5 3.5 0.219678 17 30 30 2.5 3.5 0.181234 18 50 30 2.5 3.5 0.228912 19 15 56 1.5 3.5 0.376615 20 30 56 1.5 3.5 0.279812 21 50 56 1.5 3.5 0.354217 22 15 48 1.5 3.5 0.259872 23 30 48 1.5 3.5 0.271234 24 50 48 1.5 3.5 0.321291 25 15 40 1.5 3.5 0.289735 26 30 40 1.5 3.5 0.2762896 27 50 40 1.5 3.5 0.215850 Table 4. Description of heat sinks with base plate materials. Sink A/1 Sink B/2 Sink C/3 Fin material Aluminum Aluminum Aluminum Fin profile Plate fin Plate fin Plate fin Fin dimensions in mm 25 x 40 x 3 25 x 40 x 4 25 x 40 x 5 Base plate thickness in 5/2.5 5/2.5 5/2.5 mm Base plate material / ccc Sink D/4 Sink E/5 Sink F/6 Sink G/7 Aluminum Aluminum Aluminum Aluminum Pin fin Pin fin Elliptical Elliptical 40 x 4 40 x 5 40 x 5.5 x 4 40 x 6.5 x 5 5/2.5 5/2.5 5/2.5 5/2.5 Table 5. Description of modified heat sinks with base plate materials. Sink a Sink b Sink c Fin material Aluminum Aluminum Aluminum Fin profile Core diameter in mm No of discs Circular disc plate fin Circular disc plate fin Circular disc plate fin 25 25 25 9 (vertical direction) 9 (vertical direction) 6 (vertical direction) Slots in disc Without 4 slots with D10 4 slots with D10 Fin dimensions in mm Base plate thickness in mm Base plate material R27 40 x 2 R27 40 x 2 R27 40 x 2 5 5 5 / ccc Sink d Sink e Sink f Sink g Aluminum Aluminum Aluminum Aluminum Elliptical disc plate fin Plate fin Plate fin Frustum pin fin 25 Corner base 25 x 30 Center base 65 x 54 x2 2 base 9 (vertical 4 corner base Bottom and direction) center 4 slots with D10 major, minor axis 32.5 x 27 40 x 2 25 x 40 x 5 25 x 40 x 5 40 x 5 x 4 5 5 5 5 rate of 30CFM. The flow of air is parallel to the heat sinks and vertical flow of air is pulled upward by a fan mounted at the top of the heat sink with clearance. Through the optimization process, the thermal performances of three types of heat sinks, straight finned heat sink, pin finned heat sink and elliptical heat sink are analyzed and the viable heat sink geometry design is selected. Even though the performance of the heat sink is good, the temperature of the heat sink at center is high. Disc plate heat sink with cylindrical core, elliptical plate heat sink with cylindrical core, frustum heat sink and double base heat sinks are implemented in order to reduce the temperature of heat sink at center. The dimensions and characteristics of the heat sinks are contained in Tables 3, 4 and 5. 4. Results and discussion The chassis model with different combination heat sinks was analyzed by CFD simulations. The results were obtained by varying the heat sink model and keeping the entire computational domain the same. The simulation results are compared with experimental results. In this work, 27 combinations of heat sinks are selected and analyzed. The thermal resistance curve is shown in Fig. 4. These analyses are taken for opti-
2008 R. Mohan and P. Govindarajan / Journal of Mechanical Science and Technology 25 (8) (2011) 2003~2012 Fig. 4. comparison of experimental results with simulation results. Fig. 6. Thermal resistance of fin vs number of fins for various fin heights with 2.5 mm fin pitch. Fig. 5. Thermal resistance of fin vs number of fins for various fin heights with 4mm fin pitch. Fig. 7. Thermal resistance of fin vs number of fins for various fin heights with 1.5 mm fin pitch. mizing the fin height, fin pitch and fin thickness. It is seen that densely stacked fins do not allow much air to cool the hottest center parts of the heat sink. However, due to better flow path for the air, the performance of the heat sink is increased. The effect of changing the fin height and fin pitch is shown in different curves. From this comparison, it is assumed that fin pitch varies from 1.5 mm to 4mm. Varying the number of fins implies that the fin spacing varies, so that the overall flow width is fixed. Note that as the number of fins increases, the space between the fins decreases. Due to smaller fin spacing increases the pressure drop and decreases the total volume flow rate supplied by the fan. From Figs. 5, 6 and 7 the thermal resistance is high for 15mm fin height. For increasing fin height the thermal resistance is decreased moderately and then increased when fin height is nearly 50mm. Comparing the performance curves, there are physical limitations to producing a viable heat sink. The thermal resistance of a heat sink is strongly affected by the number of fins. The performance of the heat sink decreases due to small heat transfer area. For viable heat sink fin height is optimized as 40mm and also fin pitch is 2.5mm. As shown in Figs. 6 and 7, at the high end of the range of number of fins, the volume flow rate through the small spaces is so low that air heating outweighs the heat transfer coefficient advantage of the smaller spacing, and dominates the thermal resistance. At the low end of the range of number of fins, there is simply not enough fin area to achieve high performance. There are a few conclusions to note at this point. Most significant is that the sensitivity to number of fins decreases as the fin height increases. The larger flow space has lower pressure drop. Making the fin space narrower by adding more fins produces less increase in pressure drop for a tall fin space than for a short one. Keep in mind that the fan responds to higher pressure drop by delivering less volume flow rate. The lower flow rate translates to a higher air temperature rise contribution to the thermal resistance. Secondarily, the addition of fin height brings less and less performance return as the fin height increases. This effect stems from the decrease in air speed (and thus heat transfer coefficient) and from the decrease in fin efficiency. It is worth noting that experience shows that the thermal resistance is typically not quite as sensitive to the number of short fins. For all of the heat sinks, it is viewed that their centers are the hot spots since the heat source corresponds to the closeness of the base center. The fans installed on the heat sinks are identical with dimensions. The fans have hubs where air cannot pass through and it makes the center parts hotter. In the current simulations, the swirl of the fan is not modeled since
R. Mohan and P. Govindarajan / Journal of Mechanical Science and Technology 25 (8) (2011) 2003~2012 2009 the fans are lumped parameter models. For real cases, the center would not be as hot as the present simulations predict, due to the swirl. It is observed that the upper right and left part of the heat sink have lower temperature when compared to the center part of the heat sink. This is due to more air flow circulation in the sides of the heat sink, and also the exhaust fan sucks the hot air which is nearer to side of the heat sink. The cooling becomes less efficient at other sides of the heat sink. It is observed that the conduction rate is high in ccc base plate rather than copper base plate and also it is enhanced by increasing the number of fins. Since, for same heat source CPU the bottom heat sink temperature is decreased for increasing number of fins. In this work, base plate thickness is increased from 2.5 to 5mm and base plate material is changed while keeping the fin lengths constant. The temperature difference of various heat sinks as is shown in Figs. 8, 9 and 10. The difference between the maximum and the minimum temperatures on the ccc base plate heat sink is around 9 C, whereas the copper base plate heat sink is around 10 C. For small thickness of ccc base plate heat sinks performed well rather than larger thickness of copper base plate. The performance of heat sinks is enhanced, the weight reduction, space limitation is an advantage of using ccc base plate heat sinks. Ccc base plate heat sink performs well when compared to copper base plate heat sink. By increasing the base plate thickness and changing the material of base plate, the performance of the heat sink is enhanced. It is also observed that by adding the base plate the heat conduction rate is enhanced in place of increasing the fin height. As shown in Fig. 4 the thermal resistance is high for 15mm fin height. For increasing fin height the thermal resistance is moderately decreased and it is increased when fin height is more than 50 mm. The performances of 2.5 mm ccc base plate heat sinks are obviously comparable to 5mm copper base plate heat sinks. The straight finned heat sinks give better performance, but it shows that still their centers are hot spots. The temperature distributions of heat sink e are presented as shown in Fig. 11. Emre Ozturk and Tari used four copper fins at the center of the heat sink for reducing hot spots. The thermal performance of the heat sink is not affected, indicating that these fins do not contribute much to heat transfer. This result is expected from Ozturk s analyses where removing the four center fins did not change the temperature distribution. For reducing the hot spots, the heat sink (a to d) has 25 mm diameter core with round and elliptical disc plate with four slots are used. Disc type fins wrapped around the cylindrical core which allow heat conduction in the radial and vertical directions. The disc diameter is 2 times diameter of core. However, the thermal performance of the heat sink (a to d) decreases due to more heat transfer area. Even if heat sink (a to d) core is provided it gives more hot spots when compared to heat sinks 1 to 3. The cylindrical core offers more thermal resistance to reduce the performance. Frustum fin heat sink f performs better than pin finned elliptical finned heat sinks. By reducing hot spots heat sink e and f are used. Heat Fig. 8. Heat sink combination vs temperature difference. Fig. 9. Heat sink combination vs temperature difference. Fig. 10. Modified heat sink combination vs temperature difference. sinks f show better thermal performance by 41-57% compared to heat sinks a to d and g. In heat sink e four corner base straight finned heat sink is attached to the bottom base which is designed for reducing the temperature of the heat sink at center. It is observed that power is equally distributed to the four corner base heat sink from the bottom base. As shown in Fig. 10, heat sink e performs better by approximately 61% compared to other heat sinks. In heat sink f the straight fins are extruded from the bottom base by 19 mm and then center base is attached. From the center base straight
2010 R. Mohan and P. Govindarajan / Journal of Mechanical Science and Technology 25 (8) (2011) 2003~2012 Fig. 11. Temperature distributions on heat sink e with 5mm CCC base. Fig. 12. path lines and temperature distribution for heat sink g. fins are extruded by 19 mm. By providing center base in addition to bottom base the thermal performance is increased and center temperature of the heat sink is also reduced. This performance result is nearly closer to the results in earlier heat sinks. In heat sink g fin thickness is varied by 1.5mm. First and last 15 mm height of heat sink g fin thickness is 5 mm and intermediate thickness is 6.5 mm. The heat sink g performs well, which is closely equal to performance of heat sink e and f. It is observed that even if the performances are same, but heat sink g has more hot spots compared to heat sinks e and f. The heat sinks e and f nearly reduced the center temperature up to 1.5 to 2.5 C. Although the heat sink dimensions are similar, ccc base plate heat sink enables higher conduction rates, and heat is conducted to the whole heat sink in a more efficient way. When the computer chassis is investigated, it is observed that only the upper right part of the heat sink has a free path for the air flow. Therefore, air driven by the CPU fan can travel to that side and the effect of which can also be seen in the temperature distributions. On the other sides of the CPU, air returns to the proximity of the heat sink by hitting the wall, the fan sucks the returning relatively hot air and the cooling becomes less efficient at these sides of the heat sink, as can be observed in Fig. 12. It is also observed that these results to model not only the CPU-heat sink assembly but the whole chassis are important for predicting heat sink performance. To investigate this issue further, everything inside the chassis is fixed and the heat sink model is changed. The mesh is kept the same to able to compare the results with the detailed chassis model. The model with CPU heat dissipation values also resembles the experimental setup. The air can bounce off the chassis walls and recirculate in the chassis, but the temperature distribution is much more symmetric compared to the detailed whole chassis model. It is viewed that the 5 mm thick base plate heat sinks performed well when compared to 2.5 mm thick base plate heat sinks. Also, the ccc base plate heat sinks performed well when compared to copper thick base plate heat sinks. It is noticed that the performance of the heat sink is increased by increasing the thickness of fins instead of increasing the number of fins. However, the air flow path is better and heat transfer rate does not change; and if heat transfer area is decreased, then the thermal resistance is not significantly changed. The reduction in heat sink material and weight creates worth for the manufacturer. The heat transfer rate is enhanced by increasing fin thickness. The heat transfer rate is enhanced by increasing fin thickness up to 25% in 5mm base plate heat sink models and up to 10% in 2.5mm base plate thickness heat sink models. It is observed that the performances of elliptical fin, pin fin heat sinks are identical for 2.5mm and 5 mm copper base plate heat sinks and also 5 mm ccc base plate heat sinks. However, compared to pin fin heat sinks with 2.5 mm ccc base plate, the elliptical fin heat sinks performed well. 5. Conclusions The number of fins and their distribution, fin material and base plate thickness are investigated for enhancing the heat dissipation rate from CPU. Improvements on heat sink designs are possible by the use of CFD. It is possible to design a new heat sink with suitable base plate which has better thermal performance and uses less material using CFD simulations. The heat sink base thickness is also a parameter for increasing the performance of the heat sink. If base plate material is selected to be ccc rather than copper or aluminum, then the thermal resistance of the heat sink is decreased. When the base plate thickness was varied from 2.5mm to 5 mm, the heat sink performed better. Due to space limitations of a heat sink in a computer, it is not possible to increase the height of the heat sink. Therefore, the base plate is attached with heat sink to enhance the performance rather than increasing the height of heat sink. In the current study, it is observed that stacking too many fins is not a solution for decreasing the hot spots on the
R. Mohan and P. Govindarajan / Journal of Mechanical Science and Technology 25 (8) (2011) 2003~2012 2011 heat sink since they may prevent the passage of air coming from the fan to the hottest center parts of the heat sink. In this paper, three thicknesses of heat sinks with base plate are selected and analyzed, from which the optimal design of the heat sink is selected which gives more heat transfer rate. It is observed that the velocity field around the heat sink is affected by the presence of the other components inside the chassis as well as the chassis walls which redirect the hot air back into CPU heat sink. If the heat sink is plate fin type, plate fins can reduce the recirculation. It is also possible to increase intake of cooler air by using a heat sink fan. The CPU is considered as the largest heat source, which is cooled directly drawing cooler ambient air from outside the chassis to the CPU fan with a duct is a viable option that has been recently implemented by many chassis manufacturers. The present study together with Ozturk outlines the details of CFD simulation steps for a computer chassis thermal management solution by concentrating on CPU cooling. The results and conclusions obtained in this present work are found to be in good agreement with the conclusion obtained by Ozturk. In this study, it is suggested that heat sinks e, f and 1 to 3, will benefit the design engineers involved in electronic cooling. Chassis models with different combination heat sinks are analyzed by CFD simulations. Nomenclature------------------------------------------------------------------------ Ra : Raleigh number Nu : Nusselt number h : Enthalpy P : Pressure T : Temperature or celsius temperature scale A : Area h : Heat transfer coefficient k : Thermal conductivity k : Turbulence kinetic energy L : Length P : Perimeter Pr : Prandtl length R : Gas constant u,v,w : Velocity components υ : Kinematic viscosity S : Modulus of the rate of strain tensor T s : Surface temperature : Surrounding temperature T References [1] Y. Chung and K. Luo, Unsteady heat transfer analysis of an impinging jet, J. Heat Transfer, 124 (2002) 1039-48. [2] K. Nishino et al., Turbulence statistics in the stagnation region of an axisymmetric impinging jet flow, Int. J. Heat Fluid Flow, 17 (1996) 193-201. [3] K. Kim et al., Heat pipe cooling technology for desktop PC CPU, Appl. Therm. Eng., 23 (2003) 1137-44. [4] Y. Wang and K. Vafai, An experimental investigation of the thermal performance of an asymmetrical flat plate heat pipe, Int. J. Heat Mass Transfer, 43 (2000) 2657-2668. [5] Z. Zhao and CT. Avedisian, enhancing forced air convection heat transfer from an array of parallel plate fins using a heat pipe, Int. J. Heat Mass Transfer, 40 (13) (1997) 3135-47. [6] G. P. Peterson and C. S. Chang, Heat transfer analysis and evaluation for two-phase flow in porous-channel heat sinks, Numer. Heat Transfer A, 31 (2) (1997) 113-130. [7] S. Y. Kim et al, Effect of anisotropy in permeability and effective thermal conductivity on thermal performance of an aluminum foam heat sink, Numer. Heat Transfer A 40 (1) (2001) 21-36. [8] T.-G. Maxime and G. Louis, Thermal resistance minimization of a fin-and-porous medium heat sink with evolutionary algorithms, Numer. Heat Transfer A, 54 (4) (2008) 349-366. [9] S. Y. Kim and R. L. Webb, Thermal performance analysis of fan-heat sinks for CPU cooling, Proceedings of the IMECE 03, IMECE2003-42172 (2003). [10] D. Kim et al, Thermal optimization of microchannel heat sink with pin fin structures, Proceedings of the IMECE 03, IMECE2003-42180 (2003). [11] S. J. Kim et al, Comparison of fluid flow and thermal characteristics of plate-fin and pin-fin heat sinks subject to a parallel flow, Heat Transfer Eng., 29 (2) (2008) 169-177. [12] Y.-T. Yang and H. S. Peng, Numerical study of thermal and hydraulic performance of compound heat sink, Numer. Heat Transfer A, 55 (5) (2009) 432-447. [13] J. L. Xu et al., Microscale heat transfer enhancement using thermal boundary layer developing concept, Int. J. Heat Mass Transfer, 48 (9) (2005) 1662-1674. [14] H. Jonsson and B. Moshfegh, Modeling of the thermal and hydraulic performance of plate fin, strip fin, and pin fin heat sinks-influence of flow bypass, IEEE Trans. Compon. Packag. Technol., 24 (2000) 117-123. [15] I. E. Idelchik, Flow resistance, A Design Guide for Engineers, Hemisphere, New York (1989). [16] K. Noda et al, Effect of cut-fin shape on heat transfer and pressure loss performance, Natl. Heat Transfer Sympos. Jpn., 3 (2004) 799-800. [17] C. H. Amon and B. B. Mikic, Numerical prediction of convective heat transfer in self-sustained oscillatory flows, J. Thermophys, 4 (2) (1990) 239-246. [18] C. H. Amon et al, Numerical and experimental studies of self-sustained oscillatory flows in communicating channels, Int. J. Heat Mass Transfer, 35 (11) (1992) 3115-3129. [19] T. Y. Lee and M. Mahalingam, Application of a CFD tool for system level thermal simulation, IEEE transactions on components, Packaging, and manufacturing Technology- Part A, 17 (1994) 564-571. [20] H. Wong and T. Y. Lee, Thermal evaluation of a PowerPC 620 Microprocessor computer, IEEE transactions on components, Packaging, and manufacturing Technology-Part A, 19 (1996) 469-477. [21] S. Savithri and E. C. Keith, Rapid design of heat sinks for
2012 R. Mohan and P. Govindarajan / Journal of Mechanical Science and Technology 25 (8) (2011) 2003~2012 electronic cooling computational and experimental tools, IEEE Symposium (2000) 243-251. [22] R. L. Linton and D. Agonafer, Thermal model of a PC, ASME Journal of Electronic Packaging, 116 (1994)134-137. [23] C. W. Yu and R. L.Webb, Thermal design of a desktop computer system using CFD analysis, Seventeenth IEEE SEMI- THERM SYMPOSIUM (2001) 18-26. [24] J. Y. Chang et al., Identification of minimum air flow design for a Desktop computer using CFD modeling, 7 th intersociety conference on thermal and thermomechanical Phenomena in Electronic systems, 1 (2000) 330-338. [25] D. Lober, Optimizing the integration of an electronics system into an existing enclosure using CFD modeling techniques, International journal of microcircuits and electronic packaging, 22 (1999) 146-151. [26] E. Ozturk, CFD modeling of forced cooling of computer Chassis, Engineering applications of computational fluid mechanics (1) 304-313. [27] E. Ozturk and I. Tari, Forced air cooling of CPUs with heat sinks, IEEE Transactions on components and packaging Technology, 31 (2008) 650-660. P. Govindarajan received the B.Sc (Maths) degree from Madras University in 1966. The B.E and M.Sc (Heat Power Engineering) degrees from Madras university, India in 1969 and 1972 and Ph.D in Mechanical Engineering from I.I.T Kharagpur in 1983. He is principal at Sona college of Technology, Tamil Nadu, India. He was a staff member in mechanical engineering at Government college of Engineering, Tamil Nadu, India. His research interests are gas turbine combustion, heat transfer applications, solar Energy, I.C Engines. R. Mohan received B.E and M.E degrees from Madras and, respectively, Anna University in 2001 and 2005. He is currently doing doctoral works in Mechanical Engineering at Anna University. His research interests are CFD and heat transfer application in Electronics industries.