COMPUTATIONAL METHODOLOGIES FOR THERMO-FLUID DYNAMICS PROBLEMS IN ELECTRONICS
COMPUTATIONAL METHODOLOGIESFOR THERMO-FLUID DYNAMICSPROBLEMS IN ELECTRONICS
KITE GROUP STRUCTUREKite Group in Italy: TORINO & CHIVASSO (HQ)Kite Group in Adriatic & Balcan: ZAGREBKITE GroupMAIN OFFICE - ITALYKITE GroupADRIATIC & BALCAN DIVISION•Virtual prototyping•Engineering intelligence•Software distributionD e lta _ P & ND ifferen z a tra sp esa reale e fo rfaitd a ily a v g-4 0 ,0-3 0 ,0-2 0 ,0-1 0 ,00 ,01 0 ,02 0 ,03 0 ,04 0 ,0R o c c a s e c c a D a n ie leR ic c io G iu s e p p eP u n zo C iroP o m p ili Ma s s im oN a ta le Fa b rizioC e ra V itto rioC a s ta ld o A le s s a n d roD r iv e r sG iu g n oMa g g ioA p rileMa rzoFe b b ra ioG e n n a ioMe s e
CHALLENGES IN THERMAL DESIGN INELECTRONICS (1)• During the electronic system design steps, electrical and thermo-mechanicalissues are strongly coupled,• It is important to get accurate and predictive chips operating temperatureestimations during the operations.• Three different methods to analyze the problem:1. Analytically: substantial underestimation of temperatures2. Experimentally: it drives costs in terms of resources and time3. Numerically: early stage of the design, full range of operational in time andwith low costs• Heat Transfer is an Empirical Science,• Computational multiphysics: ability to solve fluid dynamic, conductive,convective and radiative problems by means of simulations
CHALLENGES IN THERMAL DESIGN INELECTRONICS (2)• Computer simulation helps in facing thermo fluid-dynamics problems• Assumptions very key to result quality (Experience vs. CFD)• Accurate Thermal Prediction within an electronic system is a difficult task:1. Complex Geometries, Bends, Obstructions, etc...2. Dozens of active and passive electronic components3. Circuit boards as laminates of composite sheets4. Complex phenomenology of the flow ruled by low Reynolds number.• Select the right modeling tool is critical,• Test validations whenever possible, actually before set-up the virtual model• 5% - 7% accuracy of simulation to test results is acceptable.
THERMAL SIMULATION– TEST CASE / INITIAL CONDITIONS• The problem here presented is an Info-telematic system: hi-fi system with 40W, satellite navigation, bluetooth and wifi access point,cluster radio• Operations capability at 65 °C.• Some external surfaces always below 100 °C.• This is good case to test the computational specification of a commercial solver due tothe strong thermal interaction among components.
THERMAL SIMULATION– EXPERIMENTAL SET-UP• six monitoring points were considered: these are the red andblue dots on the boards and aluminum heat sink
THERMAL SIMULATION– SCSTREAM• KITE GROUP is official reseller for CRADLE (SC/Tetra, STREAM,Heat Designer) software. On these products, KITE GROUPgives to its customers Technical Support and Training.
THERMAL SIMULATION– MODELING• The numerical model:o actual geometrical dimensions,o metal housing of 195 x 165 x 53 [mm] containing all the electronics,o a wooden box of dimensions 250 x 245 x 100 [mm] is including theelectronic system mentioned above.• Some computational methodologies in the context of CFD are introduced inorder to decrease the computing resources both in terms of hardware andtime:o modeling the stratification of the circuit boards,o accurate modeling of integrated circuits,o choice of the best turbulence model characterizing the flowo methods to reduce drastically the number of degrees of freedom of thesystem , by means of MultiBlock
• The circuit boards are composite elements: conductive sheets of copper alternatewith layers of epoxy resin, also known as FR4.• Several ways can be implemented to model the PCB:1. importing within scSTREAM Gerber files of the electric circuit2. equivalent material having anisotropic thermal conductivity• It is needed to know:1. The number of insulating layers in FR4,2. number of sheets of copper and3. the position of the thermal vias…THERMAL SIMULATION– MODELING ( PRINTED CIRCUIT BOARDS )
• Generally the PCB thermal resistance of a PCB can be lowered by verticalinterconnections of copper (in thickness, known as thermal vias).THERMAL SIMULATION– MODELING ( THERMAL VIAS)Array of viasa) area with a filling ofcopperb) equivalent materialvias and PCBTop viewSide view
• To model an integrated circuit (IC) it is important:1. to define the main dimensions,2. to identify the materials used,3. to create the dissipative inner and4. to modeling the welds by using a TIM• information from the chip vendor (appropriate resistivethermal model) scSTREAM is suitable to handle thosefollowing the directives of the project DELPHITHERMAL SIMULATION– MODELING ( INTEGRATE CIRCUITS)
• The fan was modeled not as a movable part (very expensive approach fromthe computational effort) but for its equivalent effect, so by inserting in themodel the pressure vs flow rate curve.THERMAL SIMULATION– MODELING ( FAN )
• To mesh the electronic system was used multi-block method: prevents an unwantedincrease of the discretization elements even in areas far from the place of interestTHERMAL SIMULATION– MODELING ( MESH )
• Because of the absence of a characteristic dimension useful todescribe the flow regime, the problem has been solved eitherin laminar that in turbulent regimes.• A turbulence model at low Reynolds number has been used:specifically the Abe-Nagano-Kondoh (AKN) model• It is also reported the simulation without and with radiationheat exchange mode (numerical results will be shown in thelast slide)THERMAL SIMULATION– SET-UP DEFINITION
SIMULATION RESULTS vs EXPERIMENTAL DATA% DISCREPANCYTHERMAL SIMULATION– RESULTS (4)Type of analysis DDR2 nVIDIA PCB HeatSpreader Heatsink FAN outlet air MB PCBLaminar flow & no-radiation 1,77 1,36 3,01 3,94 5,74 5,14AKN model & radiation 1,13 1,65 0,09 0,09 2,49 1,92
CONCLUSIONS• It has been highlighted the opportunity to exploit a numerical approach in orderto simulate thermal and fluid-dynamical behavior of a complex electronic system.• The goal mainly consists in disposing of a predictive and flexible tool forcharacterising equipment during several functioning conditions.• Using the numerical models defined, prediction accuracy is assessed againstcorresponding experimental measurements of component surface temperature.• Very good correlation is found between numerical values and experimental data(mainly less than 3% of error).• scSTREAM from SOFTWARE CRADLE is oriented towards Electronic Cooling problems• By means of scSTREAM is very simple to create a CFD model starting from a verycomplex geometry