ISTP-16, 005, PRAGUE 16 TH INTERNATIONAL SYMPOSIUM ON TRANSPORT PHENOMENA A Novel Approach for the Thermal Management of PC Power Supply Fu-Sheng Chuang*, Sheam-Chyun Lin, Jon-Ting Kuo, Chien-An Chou Department of Mechanical Engineering, National Taiwan University of Science and Technology, Taipei, 106, Taiwan, ROC Yu-Shan Luo, Yung-Tai Chou Department of Mechanical Engineering, Tung-Nan Institute of Technology, Taipei, Taiwan, R.O.C. *Corresponding author: E-mail: fschuang@mail.ntust.edu.tw, Tel: +886 776445 Fax: +886 776460 Keywords: Axial-flow Fan, Centrifugal Fan, Heat Sink, Power Supply Abstract This study intends to improve the efficiency of PC power supply by using the numerical analysis and experimental method to raise the heat dissipation of the power supply and obtain high performance and low noise effect of the device. Most power supply used today utilizes 8-cm axial-flow fan. In order to overcome the high system impedance, the fan usually has to provide high pressure and thus generate high noise. And because of the limitation of the flow direction, the fan must be placed sideway to the power supply. This position will cause the exit flow to be mostly at low temperature. In this study, we plan to put a centrifugal fan on top of the power supply. This centrifugal fan sucks in the high temperature air inside power supply and expels it from the opening at the side plate. This deployment greatly improves the heat dissipation performance. Furthermore, the cross-sectional area of the upper plate is 15 14 cm, we can use a 14-cm centrifugal fan. Thus the heat dissipation is enhanced and noise is decreased. The new developed cooling fan uses radiating blades with curled flow channel. The parametric studies of the entrance and shrouded effect were conducted. Best geometry with entrance area of 90 cm was found. At the same operating speed, the new fan s maximum flow rate was 5.% better than the referenced fan, and the maximum static pressure increased 48.9%. The shrouded effect was also better. Flotherm software was then used to analyze the heat dissipation of power supply. Various parameters such as fin length, width, height, casing s opening, the electronic part s positions were studied. Temperature distribution was found. The results show that the increasing of the fin s volume and variation of the position of the electronic part can reduce the temperature decisively. 1 Introduction Computer system s stability and durability is closely related to the power supply. In recent years, the development of power supply is heading toward cheaper price, low noise, high 1
Fu-Sheng Chuang, Sheam-Chyun Lin, Jon-Ting Kuo, Chien-An Chou, Yu-Shan Luo, Yung-Tai Chou efficiency, and better heat dissipation. Among them, the requirement of low noise is the most urgent. Owing to the high heat production in the computer, power supply also plays a role in helping dissipate heat to the outside. In current market, most of the power supply is 15 14 8.5 cm (Fig.1) in dimension, including transformer, capacitance, fin, inductance etc. This study intends to replace the axial-flow fan traditionally used with centrifugal fan, and places the fan on the upper side of the casing to help dissipate the heat. Because of the rising trend of hot air, the fan can thus suck in lots of hot air. Also we can use the top surface of the power supply casing, which is 15 14 cm in area to mount the fan. This will yield a larger area than the original 15 8.5 cm area used by 8-cm axial-flow fan. With the larger-sized fan used, the operating speed can thus be reduced to meet the lower noise requirement. And the horizontally placed new fan can reduce the vertical loading of its own weight on the bearing. The lifetime of the fan will be longer. In this study, a newly designed fan blade was developed and placed in the housing of the traditional centrifugal fan. The advantages of the centrifugal fan, such as high-static pressure, low noise and high flow rate were preserved. Regarding the fan study, in 1996 and 1997, Horng and Lu [1, ] had done some works on the noise of forward aligned and backward aligned centrifugal fan. They found that with increasing tongue angle, the maximum flow rate decreased, the power input was lower, static pressure increases slightly. The efficiency had no significant improvement. But noise was reduced. They concluded that if the flow rate or pressure was the major concerns of the fan, tongue angle should be an important parameter for consideration. In 001, Shu and Yu [, 4] used fan-blade theory to design a radial and flat radiating blade and put it in a centrifugal fan casing. They changed various geometric parameters, such as blade number, blade angle etc. and had done a complete study on the fan performance and noise. In the aspect of system simulation, in 1999, Wang [5] used Flotherm software to analyze notebook computer with various types of heat sink inside. In this study, we used the centrifugal fan in place of traditional axial-flow fan, and horizontally placed it on top of the power supply casing. The consideration is due to: (1) Usually, axial-flow fan is placed vertically in the casing (Fig. 1). The heat released by power supply causes the hot air to rise. Hence the fan s lower part can only suck in cold air. The cooling efficiency is thus smaller. But the centrifugal fan can be horizontally placed on top of power supply (Fig. ). It has a direct contact with hot air, and gets higher efficiency. () The components in the power supply cause large system impedance. This requires high-static pressure fan. The centrifugal fan s high-pressure characteristics suit this need. () Vertical-aligned fan produces side-loading on the bearing due to its weight, which causes the wearing of the bearing and reduces the operating life. On the contrary, horizontally aligned fan will not have this bad effect. The lifetime of the bearing can thus be longer. (4) Traditional power supply has a large opening. The components inside can be seen through
A NOVEL APPROACH FOR THE THERMAL MANAGEMENT OF PC POWER SUPPLY this opening. Dust and insects can also easily get inside. Small opening for a centrifugal fan prevent the occurrence of this situation. The new cooling fan uses NACA-441 blade shape with 85 angle and 9 blades. First, the mock-up was made by using CNC. Second, various parametric studies were performed, such as varying fin s length, height, thickness, and position etc. The objectives were to achieve low noise and low heat resistance. Next, by combining the heat sink with fan, the relative position of the electronic components, the inlet and exit position and area, the whole system was analyzed to obtain its velocity and temperature fields. From this result, optimal modification of the system was expected to be achieved. Experimental Apparatus The fan performance test in this study uses Air Movement and Control Association (AMCA) standard AMCA 10-99 [6] and the noise test were using Chinese National Standard CNS-895 [7], with the usage of semi-anechoic chamber. This section describes some details of fan test, noise measurement and the operation of loading machine..1 Fan Performance Measurement The body of the equipment of the fan performance measurement is set up according to AMCA 10-99 regulations. It includes main body, rectifier plate, multi-nozzle, and auxiliary fan. This will provide a good flow field for measurement. For the sake of good measurement, good measuring devices, data acquisition system, and other assisting devices are also needed.. Noise Measurement Sound level meter connected with Frequency Analyzer was used in this study. Narrow band and board band noise can he obtained. (I) Sound pressure system-noise measurement system comprises of multifunction sound pressure meter and double-band FFT analyzer. (II) Measuring Environment semi-anechoic chamber was used to conduct noise measurements. The positions of various equipments were shown on Fig... Loading Machine The loading experiment of the power supply was accomplished by using FA-88 ATE loading machine by Hwa-Da Company. The experiment can be automatically executed through computer control. The experiment includes: loading current test, on-time test, off-time test, short circuit protection test, voltage adjustment test, Rise time, power efficiency and short-circuit efficiency. And the temperature measurement at some positions in power supply under full load had also been made.. Experimental Program The whole experiment can be divided into three parts. First part is to improve the fan s performance. A new prototype14 14 0 mm fan with curled flow channel was produced by using CNC. Another A-brand centrifugal fan was chosen to compare with. Then the new fan was studied for various inlet areas to investigate the influence on its performance and noise production. In addition, for testing the capability of overcoming high system impedance, a plate of 1 1 m was placed in front of the inlet of the new fan and a B-brand axial-flow fan (Fig.4). The distance between fan and plate was adjusted. The second part is to find the way of reducing power supply heat production. Flotherm was
Fu-Sheng Chuang, Sheam-Chyun Lin, Jon-Ting Kuo, Chien-An Chou, Yu-Shan Luo, Yung-Tai Chou used to do some parametric studies. The temperature distribution in the power supply was obtained. Then the influence of the opening of the casing was studied. The third part is to conduct the actual measurement of the temperature of the electronic parts in the power supply under full load, the measurement of the air temperature at the exit plane, and the noise. At the same noise condition, heat dissipation capability of the new centrifugal fan was compared with the traditionally axial-flow fan..1 The Design and Experiment of a New Fan (1) Rotor design NACA cross-section was chosen as the - D base, and then extended out to form the -D blade. Prototype of the blade was made for further performance testing. () Variation of the inlet area The excess flow through the tip of the blade will greatly influence the performance of the fan. Thus the inlet area was reduced to study its effect on the performance and noise production. () Shroud effect To test the capability of overcoming high system impedance, an 8-cm axial flow fan and a 14-cm new fan (Fig.5) were used to study the performance. A plate of 1 1 m was placed in front of the inlet of both fans. The distance was adjusted and the performance was tested.. Numerical Simulation of Power Supply (1) Change of fan Flotherm was used to calculate the temperature distribution for the power supply using traditional fan and the new fan. The temperature at seven hot spots in the power supply (Fig. 6) was compared. The result of the simulation was used for further modification. () The increase of fin s number and volume The volume of fin (Fig. 7) and number of fin (Fig. 8) was changed for the heat sink used in the power supply. Its effect has been studied.. The Experiment for Power Supply The comparison of maximum temperature and noise for different power supply was conducted. The new fan was combined with the power supply. Under the full loading, the maximum temperature inside the power supply and the noise generation of this assembly was measured and compared with those of the power supply with traditional fan. 4. Numerical Simulation The physical model of the numerical simulation should be close to the real model in order to obtain the realistic and meaningful result. The model is as shown in Fig.9. 4.1 Code and Grid Verifications In this study the numerical result was compared with the experimental results at seven measuring points of a full-loaded power supply. This can be used to verify the validity of the program. In the grid verification. From the result shown in Table 1, 1.6-million grid system was chosen for all the later studies. In addition, a 1.5-million grid system was used to run the program with 00, 500, 700, 1000 iterations. When the iteration number reaches 700 (Fig.10), the temperature at all the monitored points has stabilized. Thus, 1.6 million grid point and 700 iterations are selected 4. Experimental and Numerical Comparison Seven monitored points were chosen in a full-load power supply for measurement. These experimental results were compared with those of 4
A NOVEL APPROACH FOR THE THERMAL MANAGEMENT OF PC POWER SUPPLY the numerical simulation. From Fig.11, we can see the operating point of the fan is 11 cfm for the experiment, and 1 cfm for the numerical result. The difference is 1cfm. The main reason for this is in the numerical simulation, some small components in the power supply were neglected. This will cause the system impedance to be smaller. The error is still in the acceptable range. Table shows the result for temperature. The error is all below.5%. 5 Results and Discussions The presentation of the result of this study will be divided into three parts. The first part is about fan performance. The second part is on the numerical study of the power supply. And the third part is on the experiment of power supply. 5.1 Fan Performance The fan performance of the traditional axial-flow fan and the new centrifugal fan will be compared for various inlet areas. Shrouded effect will also be studied. 5.1.1 Variation of the Inlet Area A. Axial-flow Fan Table shows the result for the referenced axial-flow fan. The point for maximum flow rate has an optimal value as the inlet area decreased when the inlet diameter is 96 mm; the flow rate has the largest increment, which is 14.97%. The maximum static pressure is increased as the inlet area decreased. The most is when the inlet diameter is 86 mm, static pressure increases 68.77%. The reason for these is when the fluid flow through the blade channel, there is separation and some excessive flow is recirculating, which hinder the flow into the blade rotor and thus reduces the flow rate. At the same time, such disturbing flow will generate noise. From Table 4, varying the inlet diameter has little effect on the total noise level. The noise at the inlet (monitored point 1) is about dba. At the exit (points and ), the lowest noise is 7 dba, when the inlet diameter is 106mm. In summary, the referenced axial-flow fan has an better flow rate, an increasing maximum static pressure as the inlet diameter is decreased. However, there is no significant acoustic change. B. New Fan For the new fan, experiment was conducted for inlet diameter of 10, 96, 90mm respectively. The result is as shown in Table 5. The max flow rate decreases as inlet get smaller. When inlet diameter is 10mm, largest max flow rate was obtained. The result for inlet diameter of 90mm decreased 4.5%. The maximum static pressure increases as the inlet area decreases. The 90mm diameter has the highest increment of 5.7% On the maximum flow rare, the result shows new fan changes very little as the inlet area changes. The reason is that the recirculation effect between the blade tip and casing is small, thus changing of inlet area would not have much influence on max flow rate. On static pressure, due to the decrease of inlet area, the incoming flow velocity is increased. This produces higher-pressure gradient in the flow channel and hence maximum static pressure increases. Table 6 shows the result of noise measurement. At the inlet, inlet diameter of 90 has the min. noise level of 8.5 dba. At the outlet, diameter of 10mm has the minimum noise of 7.0dBA. Because when the inlet diameter is smaller the volume flow rate will also be smaller. Less 5
Fu-Sheng Chuang, Sheam-Chyun Lin, Jon-Ting Kuo, Chien-An Chou, Yu-Shan Luo, Yung-Tai Chou excessive flow will strike the upper casing. The noise level will be lower. 5.1. Comparison of Fan Performance Considering the result of the previous measurement, inlet diameter of 90mm was chosen for the new fan (Base 1). Other two choices are the original referenced fan (Sample) and original fan with inlet diameter of 96mm (Base ). The fans are operated at 1600 rpm. The results are listed in Table 7. The maximum flow rate and static-pressure produced by Base 1 are of value 1. cfm and 11.1 mm-aq, which are 5% and 48% higher than those of sample fan. Base also has higher maximum flow rate and maximum static pressure than sample fan, which are 14.56% and 9.18% higher, respectively. Hence the performance of sample fan can be improved by simplily using a smaller inlet diameter. The final choice for this study is to choose inlet diameter of 90 mm for the new fan, which has highest maximum flow rate and static pressure. The new fan will be combined with the power supply for further study. 5.1. Shroud Effect In power supply, there are many electronic parts and heat sink, which will produce high system impedance. The effective use of such a narrow space to reach the high fan performance is important. Thus, and 140 140 0 80 80 5mm axial-flow fan mm new fan are selected to execute the shroud effect experiment. Fig. 1 shows the P-Q relation for the new fan as the distance between fan and stop plate changes form 1 to 8mm. For the flow rate, Fig. 1 shows the flow rate increases as the distance increases. When the distance becomes 8 mm, the Maximum flow rate has increased to 95% of that of no stop-plate case. On static pressure, Fig 14 shows the static pressure increases as distance increase. As the distance becomes 8mm, the maximum static pressure becomes 99% of that of no plate case. Thus for the new fan, within 1~8mm distance, the flow rate and max static-pressure are significantly influenced. The result is very suitable for the operation in the narrow space of a power supply. As for the axial-flow fan, the distance of stop plate has to become 70mm to reach 99% of the max flow rate of no plate case. At 75mm distance, the max static pressure is 78%. Hence in the power supply, the centrifugal fan with the characteristics of high static pressure has better performance than axial flow fan. 5. Numerical Study of the Power Supply 5..1 Different Fans The cases of the traditional fan and new fan in a power supply were studied by using the Flotherm software. Temperatures at seven monitored points as shown in Fig. 6 were obtained. As listed in Table 8, the temperatures at all the points are lower for the new fan case. The maximum decrease of 8.6 occurs at the inductance. This is because of the higher flow rate and higher static pressure produced by the new fan, which will bring more heat outside and thus lowering the temperature. 5.. Volume and Number of Heat-Sink Fin The transistors in the power supply rely on the heat sink to dissipate the heat. In this section, some changes have been made to the fin of the heat sink. The total volume of the fins was increased 8 mm (Fig. 1) and the number of fin was also varied (Fig. 14). The effect on the heat transfer was studied. Table 9 shows increasing 6
A NOVEL APPROACH FOR THE THERMAL MANAGEMENT OF PC POWER SUPPLY the volume of the fins appreciably decrease the temp of all the monitored points. The maximum temp of point 7 has changed 10.9 and point 5 changed 1.. This is because the fins that have its volume changed are all located on top of the 5 major heat-generating transistors. The heat generated by the transistors can thus be transferred quickly through the fins and thus yield lower temperature. For increasing the number of the fins, the lowering-temp, effect is not as pronounced as that of the increasing volume case. This may due to the total volume increment is smaller in this case than the increasing volume case. 5. The Experiment Power Supply In this part, the new fan was combined with the A-brand power supply, at the full loading of 00W, the max temp and the noise level was measured and compared with that of B-brand power supply. The temp of the transistor in the power becomes stable after 0min of the operation. The temp in A-brand is about 4 lower than brand B. This would verify that new fan in brand A power supply achieves better heat dissipation the brand B. Considering the noise, the noise level at the exit, inlet and side of the casing for brand A are.6, 1., 1.4 dba, respectively lower than brand B. In real operation, the exit of the power supply will face the user. Hence to a customer, noise at the exit is more important. Brand A has a lower noise level at the exit than brand B as indicated. 6 Conclusion This research has done a series of investigation on power supply regarding the heat dispassion and noise level. A new fan was developed to improve these two effects. New fan of inlet diameter of 96mm has increased the max, flow rate by 5.15% and increased the max static pressure by 48.90%. The new fan also has a better shroud effect than the traditional axial flow fan. In the study of the power supply, using the new fan and increasing the total volume of the fin of the heat sink has a marked effect on improving the heat dissipation of the power supply. The temp of the whole system was lower. References [1] Horng J. Y., Noise Analysis of Backward-Inclined Centrifugal Fan, Master Thesis, National Taiwan University of Science and Technology, June, 1996. [] Lu S. Y., Noise Reduction of Forward-Curved Centrifugal Fan, Master Thesis, National Taiwan University of Science and Technology, June, 1997. [] Hsu H. C., Experimental Study for a Cooling Fan Applied on Pentium 4 Notebook Computers. Master Thesis, National Taiwan University of Science and Technology, 001. [4] Yu Z. J., Numerical Study for a Cooling Fan Applied on Pentium 4 Notebook Computers. Master Thesis, National Taiwan University of Science and Technology, July, 001. [5] Wang F. M., Numerical Simulation of Notebooks. Master Thesis, National Taiwan University of Science and Technology, 001. [6] "Publication 10, Laboratory Method of Testing Fans for Rating," Air Movement and Control Association, Inc., 1999. [7] CNS 875, Determination of Sound Power Level of Noises for Fan, Blower, and Compressors, Chinese National Standard. 7
Fu-Sheng Chuang, Sheam-Chyun Lin, Jon-Ting Kuo, Chien-An Chou, Yu-Shan Luo, Yung-Tai Chou Table 1 Grid verification Exp. result 0.70 M grids 1.06 M grids 1.6 M grids 1.46 M grids 1.76 M grids Point 1 7.0 69.1 71.0 7.78 7.89 7.90 Point 81.0 85. 84.4 8.1 8.1 8.11 Point 8.4 86.91 85.0 8.0 8. 8.5 Point 4 8.0 86.74 85.69 8.6 8.54 8.5 Point 5 94. 88.1 90.8 9.01 9.0 9.07 Point 6 81.0 86.4 84.67 8.79 8.64 8.60 Point 7 95.0 10.1 100. 98.1 98.7 98.6 Unit C Table Comparison of experimental and numerical results Exp. Temp CFD Temp Deviation (%) Point 1 7.0 7.78 0. Point 81.0 8.1 1.5 Point 8.4 8.0 0. Point 4 8.0 8.6 0.8 Point 5 94. 9.01.0 Point 6 81.0 8.79. Point 7 95.0 98.1.5 Table Comparison for various inlet areas (A axial-flow fan at 1600RPM) Inlet diameter, mm 117 11 106 96 86 dddiameter(mm) Qmax(CFM) 1.89 14.08 14.6 15.97 15.9 Increase (%) 1.6.8 14.97 10.08 Pmax(mm -Aq).17.4.8 4.44 5.5 Increase (%) 7.89 0.50 40.06 68.77 Table 4 Noises for various inlet areas (axial-flow fan) Inlet diameter, mm Point 1 Point Point 117. 8.1 8.4 11.7 8.5 8.8 106.4 7. 7.4 196.8 8. 8. 86.8 8.6 8.8 Unit dba Table 5 Performance for various inlet areas (New centrifugal fan at 1600RPM) Inlet diameter, mm 10 96 90 Qmax(CFM).1 1.69 1.1 Increase (%) -.5-4.50 Pmax(mm -Aq) 8.0 9.96 11.1 Increase (%) 1.6 5.7 Table 6 Noises for various inlet areas (centrifugal fan) Inlet diameter, mm Point 1 Point Point 10 9.6 7.0 7. 96 9. 7. 7. 90 8.5 7.1 7. Table 7 Comparison of fan performance Unit dba Sample Base 1 Base Qmax(CFM) 1.94 1.1 15.97 Increase (%) 5.15 14.56 Pmax(mm -Aq).19 11.1 4.44 Increase (%) 48.9 9. Table 8 Temperatures in power supply for different fans Old Fan New Fan Difference Point 1 7.7 68.5 4. Point 8. 75. 7.0 Point 8. 76.5 6.7 Point 4 8.6 74. 8. Point 5 9.0 84. 7.7 Point 6 8.7 74. 8.5 Point 7 98. 89.7 8.6 Unit: C Table 9 Temperature in the power supply for changing the volume and number of fin Original Fin Increasing Increasing Fin Volume Fin Number Point 1 67.9 6.9 64.9 Point 75.7 7. 7.8 Point 76. 7.7 7.9 Point 4 76.7 7.7 77.4 Point 5 8.0 81.8 85.7 Point 6 76.5 74.7 78.8 Point 7 86. 75. 85.8 Unit: C Fig. 1 Traditional fan position in power supply 8
A NOVEL APPROACH FOR THE THERMAL MANAGEMENT OF PC POWER SUPPLY Fig. New centrifugal fan position in power supply Fig. 6 Positions of the measured points in power supply Point 1 1 m Test Fan Inlet Outlet 1 m o 45 o 45 1 m Point Point Fig. Measuring positions of microphones inside a semi-anechoic chamber Fig. 7 Increase of heat-sink fin s volume Fig. 4 Schematics for shroud-effect experiment Fig. 8 Increase of heat-sink fin s number Fig. 9 Physical model of the numerical simulation Fig. 5 Schematic diagram of new centrifugal fan 9
Fu-Sheng Chuang, Sheam-Chyun Lin, Jon-Ting Kuo, Chien-An Chou, Yu-Shan Luo, Yung-Tai Chou.5.5 1mm mm mm 4mm 5mm 6mm 7mm 8mm free Ps(mm-Aq) 1.5 1 Fig.10 Convergence after 700 iterations 0.5 0 0 4 6 8 10 1 Q(CFM) 0.14 0.1 FAN SYSTEM Fig. 1 The P-Q curves of new fan under shroud effect Ps(in-HO) 0.1 0.08 0.06 110 100 90 80 0.04 0.0 0 0 4 6 8 10 1 14 16 Q(CFM) (a) The experimental operating point AirFlowRate(%) 70 60 50 40 0 0 10 CentrifugalFlowFan AxialFlow Fan 0 0 5 50 75 100 Distance(mm) Fig. 1 Flow rate of fans for various stop-plate distance 110 100 90 80 (b) The numerical operating point Fig. 11 Comparison on the operating points from experiment and simulation for 00W power supply Staticpressure(%) 70 60 50 40 0 0 10 CentrifugalFan AxialFlowFan 0 0 5 50 75 100 Distance(mm) Fig. 14 Fan static pressure for various stop-plate distance 10