Passive Cooling Assembly for Flat Panel Displays with Integrated High Power Components

Transcription

1 Paive Cooling Aembly for Flat Panel Diplay with Integrated High Power Component Ilker Tari Abtract Paive cooling of flat panel diplay deign with integrated high power component i invetigated with the help of recently available emi-emprical and CFD baed heat tranfer correlation. A heat-preader-heat-ink aembly i propoed for effective external natural convection cooling of the diplay panel. A flat vertical urface and plate finned heat ink with variou fin height are conidered a heat ink in the aembly. Heat diipation limit for both type of heat ink are determined for variou panel dimenion. It i hown that for large panel, it i feaible to ue paive cooling even when integrated computer component are ued in panel for demanding application uch a video game, high definition video proceing and 2-D to 3-D converion 1. Index Term Paive cooling, flat panel diplay, radiative heat tranfer. I. INTRODUCTION Flat panel diplay are improving very rapidly in term of technology and functionality. Conventionally eparate component uch a game conole, atellite receiver are being integrated into TV panel together with proceing unit (CPU and GPU) and data torage media. Common flat panel TV diplay a well a the one with baic extra functionality can be cooled by letting ambient air pa through the grill on the back cover of the panel where the electronic are placed. Air entering the back cover from the bottom of the panel through the grill carrie heat from the component via natural convection and exit from the top grill. Thi olution provide ufficient cooling due to the very large available heat tranfer area and due to the fact that there are no localized high flux heat ource. However, if one want to integrate high power component uch a Central Proceing Unit (CPU) and Graphic Proceing Unit (GPU), thi internal cooling ytem may not be adequate due to the localized high flux heat ource. In that cae, there i a need for fan or blower to circulate air inide the panel. Fan and blower generate noie, ue electricity, require active temperature control, and reduce reliability due to their mechanical nature. Integration of uch high heat flux proceor with diplay panel tarted to appear or i propoed for the future. A new integration application: 3-D TV, PC TV, TV integrated game conole, HD PVR, HD content delivery and torage can be 1 I. Tari i with the Middle Eat Technical Univerity, Ankara, 6531 Turkey ( itari@metu.edu.tr). cited. For thee application, powerful proceor uch a the H.264 decoder propoed in [1] are needed. Alo ome eparate unit uch a the DVDR and PVR ytem propoed in [2] and [3] can be integrated into flat panel TV diplay. In thi tudy, we propoe and analyze a paive cooling aembly for flat panel diplay with integrated high power component. II. PROPOSED PASSIVE COOING ASSEMBY Flat panel diplay are uually cooled by internal natural convection with the help of the grill at the top and bottom of the back cover of the panel. In thi tudy, we propoe an alternative paive cooling aembly to achieve effective cooling by external natural convection and radiation. The propoed aembly i hown in Fig. 1. Heat generated by the integrated component i rejected with the help of the heat preader flat heat pipe and the heat ink. The heat ink alo function a the back urface of the panel. The heat ink hown in Fig. 1 i a vertical plate finned heat ink. t Back D I S P A Y Front H Finned heat ink Heat preader flat heat pipe Integrated component (GPU, CPU etc.) Fig. 1. Top view of the paive cooling heat ink aembly. Heat preader flat heat pipe i in contact with the integrated component. Plate finned heat ink i attached to the heat preader. Heat i diipated from the heat ink with natural convection and radiation. Two different heat ink geometrie are conidered for the back ide of the flat panel. Firt one i a flat vertical urface that can only provide enough heat tranfer area for low heat diipation rate. Second one i a finned heat ink a hown in Fig. 1 and Fig. 2 with plate fin of height H (.5 or.1 m) and thickne t=.1 m. The fin dimenion are elected a dimenion that are ergonomically uitable from the deign point of view (not harp to touch), and that give an overall urface efficiency that i very cloe to 1 o that entire extended urface i effective in heat tranfer.

2 Heat tranfer from the panel component to the heat ink require very good thermal contact and a heat preader for uniform ditribution of heat. Heat pipe and thermoyphon that are elf-powered have been improving rapidly for the lat decade and they can ditribute heat from localized ource to large urface area. Epecially flat heat pipe are very uitable for thi tak. In thi tudy, heat pipe heat preader combined with a large heat ink urface i propoed a the paive cooling aembly that doe not require any power. The aembly i thermally analyzed and cooling limit (achievable heat tranfer rate) for variou panel ize are determined. Heat ink form the back urface of the panel (flat urface or finned-heat-ink Integrated component (CPU, GPU etc.) are placed behind the panel between the lid component and the ambient air i hown in Fig. 3. Since the anodized layer thickne (t ox ) i very mall and thermal conductivity of Aluminum (k Al ) i very high, the temperature difference between the heat diipating component and the urface of the heat ink lid i dominated by the thermal contact reitance between the component and the heat ink a well a between the Aluminum and the anodized layer. The combination of the thermal contact reitance together with the reitance of the heat preader and heat pipe (R t,c ) can only be experimentally determined after the complete deign of the aembly. If good thermal contact can be maintained, the urface temperature (T ) of 5 C hould be enough to keep the component urface below 6 C. That i le than the deign maximum temperature (T max ) for common integrated circuit and data torage. Therefore, heat ink urface temperature, T i taken a uniform 5 C with the aumption that the heat preader flat heat pipe i capable of preading heat uniformly. Thi temperature i alo the effective temperature everywhere on the fin. Thee aumed temperature are roughly baed on the experimental tudy uing a imilar integrated heat pipe with heat preader arrangement by Take and ebb [4] in which they report C and C a hot-cold ide temperature for two different heat preader. T component T h,bottom T anodized T T q R t,c t Al/k Al t ox/k ox 1/(h conv+h rad) Fig. 3. One-dimenional thermal reitance network between the urface of the component and the ambient air. The temperature node from left to right are fo component, heat ink bottom urface, anodized layer, heat ink urface and ambient air. The reitance value can only be determined after the complete deign of the heat preader heat ink aembly. Fig. 2. Plate finned vertical heat ink geometry. III. ACHIEVABE HEAT TRANSFER RATES The heat tranfer calculation for the propoed aembly are done by conidering both natural convection and radiation a heat tranfer mode. To be able to approach the heat tranfer problem analytically, ome implifying aumption are made that are dicued in Section III-A. A. Aumption The ambient air temperature i taken a 25 C. All thermophyical propertie are obtained at that temperature. Aluminum i elected a the heat ink material for it high thermal conductivity, and lower denity and cot compared to imilar high conductivity metal. There i a thin anodized layer on the urface of the heat ink for durability, electrical inulation and high radiative emiivity. By electrically inulating the heat ink, it i poible to ue the heat ink a the back urface of the panel. A imple one dimenional thermal reitance network B. Natural Convection Cooling For the firt heat ink option i.e. the flat heat ink, the convective heat tranfer coefficient i obtained from Churchill and Chu correlation for laminar natural convection over vertical plate. Churchill and Chu correlation a preented in [5] for average Nuelt number, Nu baed on the vertical length of the vertical flat plate i: 1/6 h.387ra Nu = = k [1 + (.492 / Pr) ] 9/16 8/27 where h i average convection heat tranfer coefficient, k and Pr are the thermal conductivity and Prandtl number of air and Rayleigh number baed on, Ra i defined a 3 gβ ΔT Ra = (2) να Here, g i the gravitational acceleration (9.87 m/ 2 ); ΔT i the temperature difference between the urface and the ambient 2 (1)

3 air, T -T ; β, ν, and α are the volumetric thermal expanion coefficient, the kinematic vicoity and the thermal diffuivity of air, repectively. Uing (1) for the entire range of Ra, the convective heat tranfer rate that can be diipated from a vertical flat plate, q conv,flat can be calculated uing: Nu k q = A( T T ) (3) conv, flat where flat urface area, A i equal to the product of the length, and width,. For the econd heat ink option with plate fin (ee Fig. 2), the eparation ditance of the plate that i denoted with i important. There i an experimental tudy that i performed with imilar plate fin heat ink by Yazicioğlu and Yüncü [6] in which they uggeted a correlation for optimum eparation between the plate, opt that i taken a the eparation ditance in the preent tudy: 1/4 = = 3.53 Ra (4) opt They alo uggeted a correlation for the convection heat tranfer rate, q conv for a heat ink with optimum plate eparation ditance by conidering the enhancement over q conv,flat : 1/2 q = q Ra khδ T (5) conv conv, flat In general, (5) give conervative reult, therefore the convection heat tranfer rate that are calculated from (5) can be taken a minimum poible value, q conv,finned,min. Thu q = q (6) conv, finned,min The eparation ditance, value calculated from (4) are cloe to the eparate plate limit, thu the plate fin do not affect each other much. There will be ome diturbance where the fin meet with the bae that reduce the local heat tranfer coefficient due to the increae in the boundary layer thickne at thoe location. The upper limit of convective heat tranfer rate, although not achievable, can be obtained by taking the entire extended urface a a virtual flat plate with the total area of the finned heat ink. Thu, the ame average Nu that i obtained from (1) can be ued for calculating the maximum convection heat tranfer rate from: Nu k qconv, finned,max = ηo Atotal ( T T ) (7) where η i the overall urface efficiency of the heat ink that o can be taken a 1. Here, the total heat ink area, A total i calculated uing from (4) and by lightly extending the heat ink width to form a heat ink between two plate fin (end plate). The number of fin, N f i obtained a a function of heat ink width, fin eparation and fin thickne t from: N f = ceiling (8) + t where ceiling() i rounding-up operation. Therefore, the total heat tranfer area for the conidered fin height H can be obtained from: conv A = N (2 H + t) + ( N 1) (9) total f f Here, the firt term on the right hand ide i the finned area and the econd term i the unfinned area for the heat ink. For optimum fin pacing, opt and convective heat tranfer rate, q conv, there are two other recent le conervative et of correlation. One of the et of correlation uggeted by Yazicioğlu and Yüncü [7] by re-evaluating available tudie in literature i the following: and 1/4 = 3.15 Ra (1) opt = + Δ (11) 1/2 q q.2116 Ra kh T conv conv, flat The other et of correlation i uggeted by Cakar [8] uing Computational Fluid Dynamic (CFD) imulation: and.236 = Ra (12) opt = + Δ (13).51 q q.1898 Ra kh T conv conv, flat Thee two et of correlation are alo ued in the thermal analye to obtain more reliable convective heat tranfer rate etimate than the conervative one from (5). C. Radiative Cooling The contribution to heat diipation by radiative heat tranfer i coniderable for both flat plate heat ink and plate finned heat ink geometrie. In radiative heat tranfer rate calculation, the temperature of the urrounding i taken a equal to the ambient air temperature. Thi approximation i perfectly valid for well inulated indoor pace without any nearby high/low temperature radiant heat ource/ink. For the flat plate geometry, radiative heat tranfer rate i given with: 4 4 qrad, flat = σε AT ( T ) (14) where σ i Stefan-Boltzmann contant, ε i the emiivity of the urface, A i the urface area and T and T are the temperature of the urface and the urrounding. For the finned heat ink geometry, a view factor for the compoite of the fin and unfinned area can be calculated by uing view factor equation from the view factor catalogue in [5]. Specifically, Perpendicular rectangle with a common edge i ued to find the view factor between the unfinned bae and the fin ide urface, F b-. Alligned parallel rectangle with a eparation ditance i ued for finding the view factor between the fin ide urface, F -. The view factor between the tip of the fin and the urrounding i 1. By uing thee view factor, with the help of ummation rule and reciprocity, the view factor between the compoite heat ink urface and the urrounding, F h-urr can be obtained from:

4 F h urr A bae 2A 1 F F + A (1 2 F ) + A Aide = 2A + A + A ide b bae b tip ( ) ide bae tip (15) 2H 1 F + (1 4 Fb ) + t = 2H + + t where A ide, A bae and A tip are the fin ide, unfinned bae and fin tip area of the heat ink. Equation (15) i ued for the inner part of the heat ink. The view factor between the outide facing ide of the end plate of the heat ink and the urrounding i 1. Therefore, the radiative heat tranfer rate from the heat ink to the urrounding, q rad,finned can be obtained from: 4 4 qrad, finned = ( Nf 1)(2 Aide + Abae + Atip ) σε Fh urr ( Tb T ) (16) σε (2 A + A )( T T ) ide tip b D. Heat Tranfer Rate Calculation The total heat tranfer rate for both bare flat and finned heat ink geometrie can be calculated by umming the convection and radiation heat tranfer rate. A the apect ratio, 1: 2 i conidered for being the typical apect ratio for flat panel diplay. The heat tranfer calculation reult for the flat bare heat ink geometry are preented in TABE I. For each conidered vertical length, Ra and Nu are obtained from (2) and (1), repectively. The convective heat tranfer rate, q conv,flat are calculated uing (3). The total heat tranfer rate, q flat value are obtained by adding radiative tranfer rate, q rad,flat that i calculated from (14) to q conv,flat. For the lowet value, the contribution of radiative tranfer to total heat tranfer rate i about 5% and it increae with increaing due to the decreaing convection heat tranfer coefficient with the increaing boundary layer thickne with. It i oberved that the total heat tranfer rate that can be obtained with the flat bare heat ink are only ufficient for integrated component uch a low voltage mobile CPU and low power graphic chipet. For more powerful integrated component, finned heat ink are required. TABE I NATURA CONVECTION HEAT TRANSFER RATES ITH OR ITHOUT RADIATION FOR THE FAT HEAT SINK GEOMETRY (m) (m) Ra q conv,flat () q flat () E E E E E E E E E E E E E E E E E If the urface emiivity i low and the radiative tranfer i negligible, the contribution of the fin i eential for ufficient heat diipation. The convection heat tranfer rate reult for the plate finned heat ink geometry for the ame vertical length value are preented in Fig. 4 and Fig. 5 for fin height of.1 m and.5 m, repectively. Here, value are obtained from (4), (1) and (12). The minimum (hown a Ref [6] ) and the maximum ( Flat limit ) convection heat tranfer rate are calculated by uing (5) and (7), repectively. The other two et of reult, Ref [7] and Ref [8] are obtained from (11) and (13) Ref [6] Convection, H=.1m Ref [8] Convection, H=.1m Vertical length, (m) Ref [7] Convection, H=.1m Flat limit Convection H=.1m Fig. 4. Natural convection heat tranfer rate without radiative tranfer for the fin height, H=.1 m.

5 Vertical length, (m) Ref [6] Convection, H=.5m Ref [7] Convection, H=.5m Ref [8] Convection, H=.5m Flat limit Convection H=.5m Vertical length, (m) Ref [6] Convection, H=.1m Ref [7] Convection, H=.1m Ref [8] Convection, H=.1m CFD Convection H=.1m Flat limit Convection H=.1m Fig. 5. Natural convection heat tranfer rate without radiative tranfer for the fin height, H=.5 m. In order to how that the actual convection behavior hould be between the maxima ( Flat limit reult) and the minima that are obtained from (5), three CFD reult for H=.1 m cae are preented with black circle at =.21,.25 and.3 m in Fig. 6. Thee imulation reult are obtained uing ANSYS Fluent CFD oftware with the exact geometrie and boundary condition of the heat ink, after defining them on the center of a vertical wall in an air filled cube of 3 m ide, by mehing the domain with approximately 3 million cell that are fine in the vicinity of the heat ink and coare away from it. All of the CFD reult are between the minima and the maxima and they are cloe to Ref [7] and Ref [8] reult. Thu thee two et of reult can be conidered a more realitic. Fig. 6. Natural convection heat tranfer rate without radiative tranfer for the fin height, H=.1 m, comparion of 3 data point obtained from CFD imulation with the reult of correlation. It i oberved form Fig. 4 and Fig. 5 that heat ink with fin of.5 m height perform ufficiently by providing low to moderate diipation rate for the convection only cae, reaching up to 21 for the larget panel ize of 1 m m. Fin of.1 m perform even better, reaching up to 26 for the larget panel ize. The calculation detail for the finned heat ink radiative and total heat tranfer rate that are obtained by uing the correlation from [6] are given in TABE II. Here, N f, F h-urr and q rad,finned value are calculated from (8), (15) and (16), repectively. F h-urr i increaing with increaing due to increaing eparation ditance between the fin. The radiative heat tranfer rate i calculated by auming urface emiivity of.8 for the anodized (oxidized) aluminum. The lat two column are the minimum ( Ref [6] ) and the maximum ( Flat limit ) total heat tranfer rate for each. It i oberved from thee range that with the contribution from radiative tranfer, it i even poible to cool powerful dektop computer component integrated into a flat panel diplay. TABE II RADIATIVE AND TOTA HEAT TRANSFER RATES F h_urr q rad,finned () q finned,min () q finned,max () (m) N f

6 The comparion of total heat tranfer rate for both of the conidered fin height are preented in Fig. 7 and Fig. 8. Each eparate data et i obtained by uing it own optimum fin eparation ditance wherea the Flat limit data et i calculated for opt from (4). Here, the data point are getting farther apart with the increaing due to the increaing contribution from the radiative tranfer Ref [6] with Radiation, H=.1m Ref [8] with Radiation, H=.1m Vertical length, (m) Ref [7] with Radiation, H=.1m Flat limit with Radiation H=.1m Fig. 7. Paive heat tranfer rate (natural convection and radiative tranfer) for the fin height, H=.1 m Ref [6] with Radiation, H=.5m Ref [8] with Radiation, H=.5m Vertical length, (m) Ref [7] with Radiation, H=.5m Flat limit with Radiation H=.5m Fig. 8. Paive heat tranfer rate (natural convection and radiative tranfer) for the fin height, H=.5 m. hen Ref [7] and Ref [8] reult are taken a repreentative of actual behavior in H=.1 m cae, it i oberved that.1 m fin height doe not have much advantage compared to.5 m fin height, due to the higher view factor of.5 m fin (ee TABE II F h-urr reult). For the larget panel ize, the heat ink with.1 m fin can diipate 445 wherea the one with.5 m fin can diipate 41. IV. DISCUSSIONS AND OTHER CONSIDERATIONS The thermal analyi in Section III i valid with the following limitation: The analyi i for teady tate, therefore there hould not be a tranient urpaing the calculated power budget (Table I and III heat tranfer rate). Uniform 5 C heat ink urface temperature i aumed. If heat preader and heat pipe beneath the heat ink are deigned to keep the hot and cold pot temperature difference low, maintaining a 5 C average hould be enough to make the analyi valid.

7 Both ambient air and urrounding temperature are taken a 25 C giving a bae-to-ambient temperature of 25 C. On a very hot day, it i till poible to keep a component uch a a mobile computer CPU below it T max which i 99 C. The analyi i only for aluminum heat ink with anodized urface. The aumed anodized aluminum emiivity of.8 i an achievable value, however degrading urface condition and dut accumulation may reduce radiative heat tranfer from the urface. Overall, it i oberved that the propoed paive cooling olution are feaible for the thermal management of flat panel diplay with integrated proceor. Heat diipation from the panel component i not uniform. Since the thermal conductivity of aluminum doe not how coniderable directional difference, it i not poible to make the heat ink thick enough for even pread of heat. However, the propoed heat preader flat heat pipe can make the pread uniform [4]. The ize of flat heat pipe are limited due to manufacturing contraint and due to the fact that for large flat heat pipe, the thickne can be prohibitively large, therefore mall heat pipe ection hould be organized to form a complete heat tranfer pathway between the component and the heat ink. According to Yazicioğlu and Yüncü [6] For a given fin length and bae-to-ambient temperature difference, the value for optimum fin pacing do not vary more than an amount of.1 mm i.e. the optimum fin pacing i almot inenitive to the variation in fin height. The dependence of optimum fin pacing on bae-to-ambient temperature difference i not very trong. For a given fin height and fin length, the value for optimum fin pacing do not vary more than an amount of.4 mm. Therefore, the propoed finned heat ink deign can be ued at variou bae-to-ambient temperature difference. Alo, the fin height can be varied with a little effect on the fin eparation. V. CONCUSION Thermal analye of the propoed paive cooling aembly for flat panel diplay are performed conidering a wide range of panel ize with the apect ratio of 1: 2. The paive heat tranfer rate limit are calculated for two different heat ink geometrie, i.e. a flat vertical urface and plate finned heat ink. It i oberved that even thee imple heat ink geometrie provide enough paive heat diipation for a wide range of application requiring integrated high power component. The propoed paive cooling aembly may epecially be ueful in 3D TV application that require powerful proceor to proce 3D data or to convert 2D video to 3D. Early example of uch application, among other, are preented in [9] and [1]. imulation preented in Fig. 6. REFERENCES [1] F. Pecador, G. Maturana, M. J. Garrido, E. Juarez, and C. Sanz, An H.264 video decoder baed on a latet generation DSP, IEEE Tran. Conumer Electron., vol. 55, no. 1, pp , Feb. 29. [2] S. H. ee, J. H. Koo, and Y. Chung, Commercial advance technology with a DVDR ytem, IEEE Tran. Conumer Electron., vol. 53, no. 2, pp , May 27. [3] S- Jung and D-S ee, Modeling and analyi for optimal PVR implementation, IEEE Tran. Conumer Electron., vol. 52, no. 3, pp , Aug. 26. [4] K. Take and R.. ebb, Thermal performance of Integrated Plate Heat Pipe with a heat preader, J. Electron. Packag., vol. 123, no. 3, pp , Sept [5] F. P. Incropera and D. P. Deitt, Fundamental of Heat and Ma Tranfer, 5th ed., New York:John iley & Son, 22, ch. 9 and ch. 13. [6] B. Yazicioğlu and H. Yüncü, Optimum fin pacing of rectangular fin on a vertical bae in free convection heat tranfer, Heat Ma Tranfer, vol. 44, no. 1, pp , Jan. 27. [7] B. Yazicioğlu and H. Yüncü, A correlation for optimum fin pacing of vertically-baed rectangular fin array ubjected to natural convection heat tranfer, J. Therm. Sci. Technol., vol. 24, no. 1, pp , Apr. 29. [8] K. M. Cakar, Numerical Invetigation of Natural Convection from Vertical Plate Finned Heat Sink, Middle Eat Technical Univerity (METU) M.S. Thei, Ankara, Turkey, 29. [9] H-C Shin, Y-J Kim, H. Park, and J-I Park, Fat view ynthei uing GPU for 3D diplay, IEEE Tran. Conumer Electron., vol. 54, no. 4, pp , Nov. 28. [1] H. Takada, S. Suyama, M. Date, and K. Kimura, Front and rear image generation module for depth-fued 3-D diplay, IEEE Tran. Conumer Electron., vol. 52, no. 3, pp , Aug. 26. Ilker Tari received hi M.S. degree in nuclear engineering from the Univerity of Michigan, Ann Arbor, MI in 1991, hi Nuclear Engineer and S.M. degree from Maachuett Intitute of Technology, Cambridge, MA in 1994, and hi Ph.D. degree in mechanical engineering from Northeatern Univerity, Boton, MA in He i Aitant Profeor of Mechanical Engineering at the Middle Eat Technical Univerity (METU), Ankara, Turkey. Prior to joining METU, he wa with the Mechanical Engineering Department at the Univerity of California, Riveride. Hi reearch interet are computational heat tranfer and thermal deign with the emphai on electronic cooling. ACKNOEDGMENT The author thank Ender Ozden for performing the CFD