Use of Superposition to Describe Heat Transfer from Multiple Electronic Components Gerald Recktenwald Portland State University Department of Mechanical Engineering [email protected] Convection from PCBs These slides are a supplement to the lectures in ME 449/549 Thermal Management Measurements c 2006, Gerald W. Recktenwald, all rights reserved. The material is provided to enhance and are the learning of students in the course, and should only be used for educational purposes. The material in these slides is subject to change without notice. The PDF version of these slides may be downloaded or stored or printed only for noncommercial, educational use. The repackaging or sale of these slides in any form, without written consent of the author, is prohibited. The latest version of this PDF file, along with other supplemental material for the class, can be found at www.me.pdx.edu/~gerry/class/ME449. Note that the location (URL) for this web site may change. Version 0.81 Convection from PCBs May 30, 2006 page 1 Overview • Overview of the Physics • Experimental Data • Superposition and the adiabatic heat transfer coefficient • Sample Calculation Convection from PCBs page 2 Heat Transfer Modes Vin, Tin radiation convection conduction in the board • • • • • conduction within devices and attached heat sinks conduction in the multilayer, composite PCB forced and natural convection from devices and heat sinks radiation between devices and adjacent boards radiation between the fins of a heat sink Convection from PCBs page 3 Geometrical Complexity H B2 B1 L1 S12 L2 B3 S23 L3 • multiple length scales: large boxes and small components • irregularly shaped flow passages with blockages • three-dimensional flow patterns around heat sinks and in the wake of discrete components • internal board configurations may change in the field Convection from PCBs page 4 Fully-Developed Flow Hydrodynamically fully-developed flow: • velocity field is independent of the flow direction dp • = constant dx Thermally fully-developed flow: • flow is hydrodynamically fully-developed • heat transfer coefficient is independent of the flow direction Flow over arrays of blocks in a channel exhibits fully-developed behavior after the third or fourth row of blocks Convection from PCBs page 5 Laminar, Transitional, and Turbulent Flow Industrial equipment tends to be turbulent flow • little or no noise constraint ⇒ high flow velocities • high power consumption equipment Office equipment tends to have transitional flow • equipment must be relatively quiet ⇒ lower flow velocities Convection from PCBs page 6 Natural Convection Applications Some equipment uses natural convection only • low power devices ⇒ battery power makes fan use “expensive” • portable test equipment • optimize internal heat conduction paths conduct heat to external case use of heat pipes in lap-top computers Convection from PCBs page 7 Mixed Convection Buoyancy effects can be present in a forced convection flow Convection from PCBs page 8 Recirculation in Plan View device with heat sink Convection from PCBs page 9 Recirculation in Elevation View Experiments by Sparrow, Niethammer and Chaboki [3] H b Nu = Nufd 1.00 1.46 Re = 3700 Convection from PCBs 1.49 1.30 1.21 1.15 t b–t = 1 H 5 page 10 Thermal Wakes (1) Thermal wake for a flush heat source Convection from PCBs page 11 Thermal Wakes (2) Three-dimensional representation of a wake, T (x, y) Convection from PCBs page 12 Unmixed Temperature Profile Flow tends to organize into • By-pass flow above the devices • Array flow around the devices bypass flow, above blocks array flow, between blocks Convection from PCBs page 13 By-pass and Array Flow (1) bypass flow, above blocks array flow, between blocks By-pass flow • Higher velocity than array flow • Streamlines are topologically simple • Relatively higher turbulent fluctuation at interface between by-pass flow and top of blocks. Flow may still be considered unsteady laminar for many applications. • Gross flow features may be predicted with CFD. Convection from PCBs page 14 By-pass and Array Flow (2) bypass flow, above blocks array flow, between blocks Array flow • Lower velocities than by-pass flow • Streamlines are topologically complex: many recirculation zones • Very hard to accurately predict the details because of small scale flow features. Convection from PCBs page 15 Hierarchy of Analysis Strategies In order of increasing effort: • • • • hand calculation of energy balance use of heat transfer correlations for board-level analysis resitive network of entire enclosure Conduction modeling in the board: fluid flow is treated only as a convective boundary coefficient. • PCBCAT layer-based models • Full 3-D CFD models of conjugate heat transfer Convection from PCBs page 16 Example: Fan-Cooled Enclosure (1) disk drive Convection from PCBs power supply page 17 Fan-Cooled Enclosure (2) ΣQ1 ΣQ2 ΣQ4 X m1 ΣQ3 m2 m3 Q1 = ṁ1cp (To,1 − Ti,1) P Q1 ṁ1cp P Q2 = Ti,2 + ṁ2cp P Q3 = Ti,3 + ṁ3cp ⇒ To,1 = Ti,1 + To,2 To,3 Convection from PCBs page 18 Fan-Cooled Enclosure (3) What contributes to P Qi? • Power dissipation of devices • Heat loss directly through the cabinet to the ambient • Heat gain/loss through the PCB to an adjacent channel containing other board Perhaps individual control volumes should be connected into a thermal network. Convection from PCBs page 19 Board-Level Energy Balance Tm(x) Tout Tin x • min Q1 A B Q2 C Q3 D E • 3D effects fan wake non-uniform inlet blockage by obstacles including heat sinks • Channel by-pass and unmixed temperature profile Convection from PCBs page 20 ——————- Convection from PCBs page 21 Heat Transfer Correlations for Board-Level Analysis m• 1 h, Ta Convection from PCBs page 22 Heat Transfer Correlations for Board-Level Analysis • Energy balance only gives the air temperature. • We need values for thermal resistances to estimate junction temperatures. • Thermal resistance of heat sink comes from heat sink manufacturer. (But does that test data apply to your configuration?) • Other convective resistances are estimated from heat transfer coefficients. • General correlations for heat transfer coefficients from arbitrary devices on a PCB do not exist. Convection from PCBs R values for a high performance CPU: Rsa ∼ 0.4 W/C Heat Sink Case (cover) Rim ∼ 0.1 W/C Rjc ∼ 0.3 W/C Die Substrate Q Rjb Rba page 23 Resistive Network Models SINDA: http://www.webcom.com/~crtech/sinda.html http://www.indirect.com/user/sinda/ See also Thermal Computations for Electronic Equipment, by Gordon Ellison [2] Convection from PCBs page 24 Conduction Modeling (1) Internal resistance can be obtained from finite-element analysis of conduction heat transfer inside the device. This data is usually supplied by the device manufacturer, because only they know the details of the internal construction. m• 1 h, Ta Convection from PCBs page 25 Conduction Modeling (2) • • • • Need heat transfer coefficient at all fluid-solid interfaces. Analysis is a standard procedure with most FEM packages. Practical limit to the geometric detail Analysis time is short compared to model building time. Convection from PCBs page 26 CFD Modeling (1) m• 1 Convection from PCBs page 27 CFD Modeling (2) • Significant investment in model development =⇒ CFD model run time is often short compared to model building time. • Detailed solution still requires significant computing requirements • Momentum equations are nonlinear • Turbulence models • Inlet vents and fans need to be modeled. • Practical limit to the geometric detail • CFD packages for electronic cooling FlothermTM http://www.flomerics.com/ IcePackTM http://www.fluent.com/ Convection from PCBs page 28 Experimental Data • • • • flush mounted heaters Ribs Arrays of blocks arrays of “heater” devices Convection from PCBs page 29 Correlations Flow over a flat plate a Nu = C Re Pr b Which length scales to use? Proper application requires • geometric similarity • dynamic similarity • thermal similarity Convection from PCBs page 30 Heat Transfer Coefficients Vin, Tin s Experimental Procedure 1. Adjust flow rate 2. Set power level of each block 3. Wait for thermal equilibrium 4. Measure temperature of each block 5. Compute heat transfer coefficient Convection from PCBs page 31 Which heat transfer coefficient? Based on inlet temperature: hin,i = Qconv,i/Ai Tb,i − Tin Based on local, mean fluid temperature: hm,i = Qi/Ai Tb,i − Tm,i Pi Tm,i = Tin + j=1 Qj ṁcp Based on adiabatic wall temperature: had,i = Convection from PCBs Qconv,i/Ai Tb,i − Tad,i page 32 Superposition Principle (1) Consider flow in a tube with an arbitrary axial variation in heat input. ξ x ∆x u(r) R r hydrodynamically fully-developed flow qw''(x) Energy Equation ∂T k ∂ ρcpu(r) = ∂x r ∂r „ « ∂T r ∂r Boundary conditions ˛ ∂T ˛˛ = 0 (symmetry) ˛ ∂r r=0 Convection from PCBs ˛ ∂T ˛˛ 00 k = q w (x) ˛ ∂r r=R page 33 Superposition Principle (2) General solution is + Tw,ad(x ) − Tin R = k Z x+ + 00 g(x − ξ) qw (ξ) dx 0 where g(x+) is the superposition kernel function + g(x ) = 4 + ` ´ 2 X exp −γm x+ m 2 A γm m For a single heated patch this reduces to + Tw,ad(x ) − Tin = Convection from PCBs Q + g(x − ξ) 4ṁcp page 34 Interpretation of Kernel Function (1) ∆x r . m y Q Tw,ad(x) Tm Tin ∆Tm ∆Tm ∆Tm y Tin Convection from PCBs T(y) Tin Tm Tw,ad Tin Tw,ad page 35 Interpretation of Kernel Function (2) Energy balance gives increase in mean fluid temperature ∆Tm = Q ṁcp Solve equation defining Tw,ad for g(x+) Tw,ad(x+) − Tin g(x − ξ) = Q/(4ṁcp) + Tw,ad(x+) − Tin = 4 ∆Tm Convection from PCBs page 36 Application to PCB Heat Transfer (1) 3 2 m=1 n=1 2 3 4 The adiabatic temperature of a block is the temperature it attains when it is has zero internal heat generation. Note that if no blocks are heated, then Tad,i = Tin. Remember that “adiabatic” in this context means unheated, not insulated. Convection from PCBs page 37 Application to PCB Heat Transfer (2) The temperature difference between block i and the inlet air can be decomposed as Tb,i − Tin = (Tb,i − Tad,i) + (Tad,i − Tin) Convection from PCBs Tb,i = average surface temperature of heated block i. Tb,i − Tad,i = temperature rise due to selfheating Tad,i − Tin = temperature rise due to heat inputs from other heated elements (1) page 38 Application to PCB Heat Transfer (3) The adiabatic temperature rise of block i due to heat input from all blocks is Tb,i n X Qj ∗ gi,j = Tin + ṁcp j=1 (2) Interpret as sum of two major contributions n X Tb,i − Tin = j=1, | Qconv,j ∗ gi,j ṁc p j6=i {z } + upstream contribution Qconv,i ∗ g ṁcp i,i | {z } (3) self-heating Temperature rise due to self-heating is rise due to self-heating alone is Tb,i − Tad,i = Convection from PCBs Qconv,i had,iAi (4) page 39 Application to PCB Heat Transfer (4) Equating the right hand side of Equation (4) with the second term on the right hand side of Equation (3) gives Qconv,i Qconv,i ∗ = g (5) had,iAi ṁcp i,i Thus, ṁcp ∗ gi,i = (6) had,iAi ∗ Equation (6) shows that gi,i and had,i are intrinsically related. This is no accident since ∗ both gi,i and had,i are derived from measurements in which only block i is heated. Convection from PCBs page 40 Application to PCB Heat Transfer (5) Substituting Equation (5) into Equation (3) gives Tb,i − Tin = n X j=1, Qconv,j ∗ Qconv,i gi,j + ṁc had,iAi p j6=i (7) ∗ With measured values of gi,j and had,i, Equation (7) uses superposition to compute the effect of any power distribution on the temperature of each block in the domain. All that ∗ remains is a procedure for determining gi,j from the experimental data. Convection from PCBs page 41 Measuring had for a 3 Block Experiment (1) Measure had,i for i = 1 and Tad,i for i = 2, 3: 1. adjust flow rate 2. turn heat on for block 1 3. turn off heat for block 2 and block 3 4. wait for thermal equilibrium 5. measure temperatures of all three blocks Convection from PCBs page 42 Measuring had for a 3 Block Experiment (2) Write out Equation (2) for i = 2, j = 1: Tb,2 = Tin + Q1 ∗ Q2 ∗ Q3 ∗ g2,1 + g2,2 + g ṁcp ṁcp ṁcp 2,3 (8) Since Q2 = Q3 = 0 in this experiment, the preceding equation reduces to Tb,2 Q1 ∗ = Tin + g ṁcp 2,1 (9) ∗ Solving for g2,1 gives ∗ g2,1 = Tb,2 − Tin Q1/(ṁcp) only block 1 is heated (10) Because only block 1 is heated, Tb,2 − Tin is the temperature rise of block 2 due to heat input at block 1. Convection from PCBs page 43 Measuring had for a 3 Block Experiment (3) Define Twake,i,j = temperature of block i when only block j is heated. The term “wake” is suggestive of the mechanism of heating: Twake,i,j > Tin because block i is downstream of block j . Thus, when only block 1 is heated, the value of Tb,2 is Twake,1,2, and Equation (10) is ∗ g2,1 = Twake,2,1 − Tin Q1/(ṁcp) (11) Remember that the simplification that leads from Equation (8) to Equation (11) is valid because only block 1 is heated. ∗ Similar calculation (from same experiment) gives g3,1 . Convection from PCBs page 44 Measuring had for a 3 Block Experiment (4) Repeat measurements to obtain data for following table Measured Temperatures Heat Inputs Convection from PCBs Block 1 Block 2 Block 3 Q1 0 0 Tself,1 Twake,2,1 Twake,3,1 0 Q2 0 Twake,1,2 Tself,2 Twake,3,2 0 0 Q3 Twake,1,3 Twake,2,3 Tself,3 page 45 Anderson and Moffat Correlation (1) top view Sz side view Lz Sx Lx x row number 8 7 6 5 4 3 2 1 H B z flow direction Anderson and Moffat [1] found • g ∗(x) was related to correlation for had • no interaction between columns • fully-developed flow after third row Convection from PCBs page 46 Anderson and Moffat Correlation (2) For fully-developed region g ∗ = 1 + β1 exp (−α1N ) + β2 exp (−α2N ) For first two rows g1∗ ˘ ¯ ∗ = max 0.8 g , 1 ˘ ¯ ∗ ∗ g2 = max 0.95 g , 1 Convection from PCBs page 47 Anderson and Moffat Correlation (3) Dimension analysis gives a relationship for maximum possible turbulence fluctuations in the channel „ «(1/3) −∆P (H − B) row 0 ũmax = 0.82 Um ρ Lx where Um is the velocity in the bypass region Um = Convection from PCBs VH H−B page 48 Anderson and Moffat Correlation (4) For fully-developed region g ∗ = 1 + β1 exp (−α1N ) + β2 exp (−α2N ) For first two rows g1∗ ˘ ¯ ∗ = max 0.8 g , 1 ˘ ¯ ∗ ∗ g2 = max 0.95 g , 1 0 α1 = 0.31 ũmax + 1.91 0 α2 = 0.098 ũmax + 0.19 „ « 1 ṁcp/A β1 = −1 1.13 32.2 ũ0max + 14.4 β2 = 0.13 β1 Convection from PCBs page 49 Example Calculation (1) parameter H B Lx Sx Lz Sz value 0.0214 m 0.0095 m 0.0375 m 0.0502 m 0.0465 m 0.0592 m Table 1: Geometrical parameters for the example calculations. 3 ρ = 1.185 kg/m cp = 1005 J/(kg K) 2 V = 7.1 m/s − ∆Prow = 7.78 N/m Convection from PCBs page 50 Example Calculation (2) 2 A = 0.00334 m Um = 12.8m/s ṁ = 1.06 × 10 −2 kg/s per row 0 ũmax = 2.44 m/s α1 = 2.6685 α2 = 0.4298 β1 = 29.5387 β2 = 3.8400 Convection from PCBs page 51 Example Calculation (3) row 8 7 6 5 4 3 2 1 Q (W ) 12 18 14 7 2 13 11 15 Table 2: Power dissipated by modules in the example caluclation. Convection from PCBs page 52 Example Calculation (4) The temperature rise in row n due to heat dissipated by the module in row 1 is ” “ Q1 ∗ T e,n − Tin = g1 (n − 1) 1 ṁ cp n g1∗ (n − 1) 8 7 6 5 4 3 2 1 1.000 1.033 1.158 1.351 1.654 2.214 4.438 27.503 “ ” T e ,n − Tin 1 (C) 1.40 1.45 1.62 1.89 2.32 3.10 6.22 38.56 Table 3: Temperature rise due to heat dissipated in row 1. Convection from PCBs page 53 Example Calculation (5) The temperature rise in row n due to heat dissipated by the module in row 2 is “ ” Q2 ∗ T e,n − Tin = g (n − 2) 2 ṁ cp 2 n g2∗ (n − 2) 8 7 6 5 4 3 2 1 1.227 1.375 1.604 1.964 2.629 5.270 32.660 0 “ ” T e ,n − Tin 2 (C) 1.26 1.41 1.65 2.02 2.70 5.42 33.58 0 Table 4: Temperature rise due to heat dissipated in row 2. Convection from PCBs page 54 Example Calculation (6) The temperature rise in row n due to the heat dissipated by row three is “ ” Q3 ∗ T e,n − Tin = g (n − 3) 3 ṁ cp n g ∗ (n − 3) 8 7 6 5 4 3 2 1 1.448 1.689 2.068 2.768 5.547 34.378 0 0 “ ” T e ,n − Tin 3 (C) 1.76 2.05 2.51 3.36 6.74 41.77 0 0 Table 5: Temperature rise due to heat dissipated in row 3. Convection from PCBs page 55 Example Calculation (7) n 8 7 6 5 4 3 2 1 T e ,n − Tin (C) 57.6 72.2 54.9 30.8 18.2 50.3 39.8 38.6 Table 6: Total temperature rise for modules. Convection from PCBs page 56 Example Calculation (8) 100 90 80 Temperature (C) 70 60 50 40 30 20 10 0 1 Convection from PCBs 2 3 4 5 row number 6 7 8 page 57 References [1] A. M. Anderson and R. J. Moffat. The adiabatic heat transfer coefficient and the superposition kernel function: Part 1–data for arrays of flatpacks for different flow conditions. Journal of Electronic Packaging, 114(1):14–21, 1992. [2] Gordon N. Ellison. Thermal Computations for Electronic Equipment. Robert Krieger Publishing Co., Malabar, FL, 1989. [3] E. M. Sparrow, J. E. Niethammer, and A. Chaboki. Heat transfer and pressure drop characteristics of arrays of rectangular modules encountered in electronic equipment. International Journal of Heat and Mass Transfer, 25(7):961–973, 1982. Convection from PCBs page 58
© Copyright 2026 Paperzz