A method for calculating the heat distribution of a double-sided heat-dissipating SiC power module

By constructing a numerical algorithm based on the Fourier formula for heat conduction and the Laplace equation, a method for calculating the heat distribution of a double-sided heat-dissipating SiC power module is developed, which solves the problems of large computational load and underestimated junction temperature in the existing technology, and achieves fast and accurate temperature assessment.

CN116796537BActive Publication Date: 2026-11-17FUDAN UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310752080.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-06-25
Publication Date
2026-11-17
Estimated Expiration
2043-06-25

AI Technical Summary

Technical Problem

Existing technologies require a large amount of computation and are time-consuming when calculating the heat distribution of double-sided heat-dissipating SiC power modules. Furthermore, traditional methods cannot accurately calculate the lateral heat transfer of the chip, resulting in an underestimation of the junction temperature and failing to meet the needs of rapid evaluation.

Method used

A numerical algorithm based on the Fourier formula for thermal conduction is used to construct a method for calculating the heat distribution of a double-sided heat sink-level fan-out (FOPLP) package structure. By simplifying the model into a regular cuboid stacked structure, and combining the Laplace equation and the method of separation of variables, an accurate representation and rapid solution of the heat diffusion phenomenon can be achieved.

Benefits of technology

It enables rapid and accurate temperature assessment of double-sided heat-dissipating SiC power modules. The calculation speed is fast, making it suitable for rapid assessment in the early stages of design, and it can also be used for temperature assessment of other double-sided heat-dissipating modules.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116796537B_ABST
    Figure CN116796537B_ABST
Patent Text Reader

Abstract

This invention belongs to the field of power semiconductor technology, specifically a method for calculating the heat distribution of a double-sided heat-dissipating SiC power module. This invention provides a fast numerical solution method for FOPLP (Flexible Open-Plate Power Module). The numerical algorithm is based on the Fourier transform equation for heat conduction, which can calculate the lateral heat transfer during the packaging process. In the FOPLP, the chip is located in the center of the structure, and the heat dissipation paths are the upper rearrangement layer, the upper heat sink, and the lower substrate. During the numerical solution process, the model is simplified to a cuboid block structure as the basic model for double-sided heat dissipation. The chip in the module acts as a heat source, mainly transferring heat in the upward and downward directions. This invention constructs the numerical temperature distribution of the FOPLP structure by calculating the solid-state heat transfer and convective heat transfer laws within the power module. The numerical solution starts from the steady-state Laplace equation for heat conduction, uses the method of separation of variables, and combines boundary conditions to determine the analytical solution for the double-sided heat dissipation of the FOPLP structure, achieving a rapid evaluation of the module's temperature distribution.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] Technology Neighborhood

[0002] This invention belongs to the field of power semiconductor technology, specifically relating to a method for calculating the heat distribution of a double-sided heat-dissipating SiC power module. Background Technology

[0003] Power semiconductor modules, as crucial energy conversion components in power electronic devices, generate significant heat due to losses during chip operation. Temperature affects power devices primarily in two ways: firstly, the electrical performance is influenced by junction temperature; secondly, thermal stress arises from the mismatch in thermal expansion coefficients between different structures. These are potential heat-induced failure issues. Finite element analysis is generally used to accurately predict the thermal characteristics of power modules; however, this method is computationally intensive and time-consuming.

[0004] Currently, there are usually two methods for thermal modeling of power modules. One method uses a thermoelectric network model, which compares heat conduction to current conduction and solves the problem using a circuit model, namely the traditional Foster and Cauer models. This method uses thermal resistance and thermal capacitance to form an RC thermal equivalent circuit to establish a thermal impedance equation, which equates the power module to a one-dimensional heat transfer path. It has the significant advantages of simple expression and fast calculation speed [1]. However, this method cannot calculate the lateral heat transfer of the chip and underestimates the junction temperature. In order to achieve an accurate description of thermal diffusion coupling under different operating conditions, reference [2] proposed a 3D fourth-order Foster thermal network that takes into account the thermal coupling effect between IGBT and diode, references [3 and 4] proposed a 3D third-order Cauer thermal network that takes into account the thermal coupling effect between different bridge arms, and reference [5] established a hybrid fourth-order Foster and Cauer thermal network. However, the RC parameters in the network are obtained through finite element dynamic simulation, which takes a long time to solve. At the same time, the number of RC parameters in the accurate model is huge, the identification complexity is high, and the calculation efficiency is not high. Another method is numerical analysis, which is a solution method for obtaining the temperature distribution by directly solving the heat conduction equation. This method takes into account the lateral heat conduction process during the solution process. Reference [6] proposes a general analytical solution for the thermal diffusion performance of an eccentric heat source on a rectangular substrate, and also proposes a solution for the steady-state temperature distribution of a rectangular double-layer structure. Reference [7] extends the numerical model to a multi-layer structure, and the flexibility of the application scenario increases the difficulty of solving the Fourier coefficients. In order to solve this problem, Reference [8] proposes a steady-state thermodynamic model, the uppermost layer of which contains any number of rectangular heat sources. The difference between the model simplification process and the actual thermal performance is also discussed. The distribution of multi-chip thermal coupling under various layouts is fully analyzed and discussed in [9]. However, the above studies are all numerical modeling of traditional single-sided cooling power modules. At present, almost no scholars have studied the numerical heat dissipation solution of double-sided heat dissipation modules. In order to evaluate the temperature characteristics of double-sided heat dissipation more quickly and accurately, it is necessary to establish a set of numerical heat dissipation solution models for double-sided heat dissipation.

[0005] Using thermal resistance and thermal capacitance to construct an RC thermal equivalent circuit to establish a thermal impedance equation, the power module is equivalent to a one-dimensional heat transfer path. It has the significant advantages of simple expression and fast calculation speed, but it cannot calculate the lateral heat transfer of the chip and underestimates the junction temperature.

[0006] To accurately predict the thermal characteristics of power modules, the finite element method (FEM) is generally used for evaluation. However, this method is computationally intensive and time-consuming, requiring significant computational resources, involving numerous meshes, and has a long solution cycle, making it unsuitable for rapid evaluation in the early stages of design. Currently, most numerical heat conduction solutions focus on numerical modeling of traditional single-sided cooling power modules, with limited research on numerical heat dissipation solutions for double-sided cooling modules. This invention provides a fast numerical solution method for FOPLP double-sided cooling power modules, offering high computation speed and accuracy, and is applicable to temperature evaluation of other double-sided cooling modules. Summary of the Invention

[0007] The purpose of this invention is to provide a method for calculating the heat distribution of a double-sided heat-dissipating SiC power module that is fast and accurate.

[0008] The thermal distribution calculation method for double-sided heat-dissipating SiC power modules provided by this invention is a fast numerical solution method for double-sided heat-dissipating power modules with double-sided heat sink level fan-out (FOPLP) packaging structure. The numerical algorithm used is based on the Fourier formula for thermal conduction, which can calculate the lateral transfer of heat during the packaging process and has a good representation of the thermal diffusion phenomenon.

[0009] The double-sided heatsink level fan-out (FOPLP) package structure is described in [reference needed]. Figure 1 The diagram shows a high-power module composed of a single chip and its interconnect materials. The chip is located in the center of the structure, and the heat dissipation paths are the upper rearrangement layer (RDL), the upper heat sink, the lower solder layer, and the bottom substrate. In the numerical solution process, the model needs to be simplified to a stacked structure of regular cuboid blocks with the same cross-sectional area, i.e., a basic model for double-sided heat dissipation, as shown below. Figure 2 The heat sink, molding compound, solder, and substrate will all be included in this model according to the actual thickness of the structure. The RDL layer contains both copper material for electrical connections in vias and molding compound for insulation and mechanical support; therefore, the material parameters for this part need to be reassessed.

[0010] The basic model for dual-sided heat dissipation, such as Figure 2 As shown, the overall structure of the power module is a cuboid; its internal heat source is a rectangular region. The chip acts as the heat source within the power module, primarily transferring heat in the upward and downward directions. Therefore, the model simplifies the chip into a thin-film structure with dimensions a*b, and its area is denoted as A. s The longitudinal heat transfer region has a length and width of L. x *L y The region has an area of ​​A. c The chip provides power loss of Q, and the origin of the z-axis is located in the plane where the chip is situated. The coordinates of the chip's center are (x...). c ,y cThe lengths of the upper and lower conduction paths are t respectively. p , t n The air convection heat transfer coefficient is h.

[0011] Let the temperature distribution function of the entire structure be T(x,y,z), and the ambient temperature be T. air Steady-state heat conduction satisfies the Laplace equation, specifically:

[0012]

[0013] The general form of its general solution is:

[0014] θ(x,y,z)=X(x)*Y(y)*Z(z), θ(x,y,z)=T(x,y,z)-T air (2)

[0015] θ(x,y,z) represents the temperature rise caused by the heat from the heat source, and X(x), Y(y), and Z(z) are the coordinate functions when solving the Laplace equation using the separated parameter method.

[0016] For the entire structure, the four walls are insulated and there is no heat exchange with the outside; therefore, the boundary conditions are:

[0017]

[0018]

[0019] Therefore, the general solution to the Laplace equation can be obtained as follows:

[0020]

[0021] Where λ=mπ / L x ,δ=nπ / L y , B i A i Let i be the coefficients to be determined, i = 0, 1, 2, 3;

[0022] Heat transfer occurs between the upper and lower interfaces and the air via convection, with the following boundary conditions:

[0023]

[0024]

[0025] Since the heat transfer directions of the z-axis are opposite, they need to be discussed separately. From z = t p and z = -t n The Fourier coefficients for heat transfer at the convective heat transfer boundary of the interface can be solved separately for the upper and lower heat transfers:

[0026] B i=φ(ξ)A i (5)

[0027] For heat conduction in the positive direction, the Fourier coefficients are:

[0028]

[0029] For heat conduction in the negative direction, the Fourier coefficient is:

[0030]

[0031] In this context, ξ can be replaced by λ, δ, and β, respectively.

[0032] The chip generates heat in both upward and downward directions, but the power conducted in these two directions differs. Therefore, the power distribution of the chip as a heat source needs to be considered, which will be analyzed in detail after the formula derivation is completed. Let the total power generated by the chip be Q = Q1 + Q2, where Q1 and Q2 are the power of the upper and lower surfaces of the chip, respectively.

[0033] Let's first discuss the case where Q1 is input in the positive direction. In this case, the chip layer boundary consists of two parts:

[0034] In A s Areas included:

[0035]

[0036] In A s In areas other than the interface, the total heat conducted is equal to the total heat transferred out in the negative direction, which can be expressed as:

[0037]

[0038]

[0039]

[0040]

[0041] For the upward-conducted power Q1, we have:

[0042]

[0043] For the power Q2 conducted downwards, we have:

[0044]

[0045] At this point, the theoretical numerical model derivation and solution for the cuboid module with embedded heat source and double-sided heat dissipation are basically complete. However, the power distribution problem of heat conduction between the top and bottom of the chip still needs further discussion. The junction temperature is represented by the average surface temperature of the chip, denoted as T. j The average temperature of the upper and lower surfaces of the structure is T. p and T n Thermoelectric networks can clearly describe the temperature transfer process of heat conduction, such as... Figure 3 As shown.

[0046] Since the ambient temperature is constant, the upward and downward power transfer can be considered as parallel current splitting. Specifically, Q1 and Q2 need to be distributed through the cuboid conduction thermal resistance and the convection thermal resistance with air. Therefore, the thermal resistance needs to be calculated first.

[0047] Thermal resistance is a physical quantity determined by the material's structural dimensions and is independent of power. Therefore, when discussing thermal resistance in both positive and negative directions, it can be calculated separately. However, thermal resistance is expressed as the degree of temperature rise, so the thermal resistance is often inferred from the temperature difference in calculations. This involves using the difference between the average temperature of the chip surface and the ambient temperature. This represents the temperature difference between the junction temperature and the ambient temperature.

[0048]

[0049] According to the solution process of the Laplace equation for the steady-state heat of the chip, without considering heat conduction in areas other than the heat source on the chip layer, it is equivalent to the general solution for single-sided heat dissipation. At this time, the steady-state heat solution in the positive direction is:

[0050]

[0051]

[0052]

[0053]

[0054]

[0055]

[0056]

[0057]

[0058] At this point, the difference between the chip junction temperature and the ambient temperature can be expressed as:

[0059]

[0060] Therefore, the thermal resistance in the positive direction can be obtained as follows:

[0061]

[0062] For thermal resistance in the negative direction, the input power direction is negative, and φ needs to be replaced with a negative direction coefficient.

[0063]

[0064] The power distribution here can be broken down by the parallel relationship of thermal resistance:

[0065]

[0066] After calculating the power transfer in both directions, due to the linear superposition of temperatures, the temperature distribution of the double-sided heat dissipation base structure can be expressed as:

[0067] T(x,y,z)=θ Q1 (x,y,z)+θ Q2 (x,y,z)+T air (27)

[0068] This invention constructs a numerical temperature distribution model for the FOPLP structure by calculating the solid-state heat transfer and convective heat transfer laws within the power module. The numerical solution starts with the steady-state Laplace equation for heat conduction, employs the method of separation of variables, and, combined with the boundary conditions in practical applications, determines the analytical solution for the double-sided heat dissipation of the FOPLP structure. This enables rapid evaluation of the temperature distribution of the module. Attached Figure Description

[0069] Figure 1 This is a schematic diagram of the FOPLP (Fan-Out Power Module) structure.

[0070] Figure 2 This is the basic model for a double-sided heat dissipation power module.

[0071] Figure 3 This is a schematic diagram of the power distribution between the upper and lower parts of the heat conduction system for double-sided heat dissipation.

[0072] Figure 4 The temperature distribution of the chip is shown in Figure 1. (a) shows the temperature distribution of the chip layer; (b) shows the temperature distribution of the upper surface; and (c) shows the temperature distribution of the lower surface. Detailed Implementation

[0073] The FOPLP package is equivalent to a six-layer structure, with four layers on top of the chip and two layers on the bottom. Substituting the package parameters into the numerical model established above yields the result.

[0074] Assuming the chip's total output power is 50W, the temperature distribution of the chip layer is as follows. Figure 4As shown, the maximum junction temperature is 90.88℃. The temperature distribution on the upper surface is as follows. Figure 4 As shown in (b), the highest temperature is 51.54℃. The temperature distribution on the lower surface is as follows. Figure 4 As shown in (c), the highest temperature is 89.67℃.

[0075] References:

[0076] [1]Wang Z., Qiao Wei.A physics-based improved Cauer-type thermalequivalent circuit for IGBT modules[J]. IEEE Transactions on PowerElectronics, 2016, 31(10): 6781-6786.

[0077] [2]Iachello M, De Luca V, Petrone G, et al. Lumped parameter modeling forthermal characterization of high-power modules[J]. IEEE Transactions on Components, Packaging and Manufacturing Technology, 2014, 4(10): 1613-1623.

[0078] [3]Li Jianfeng, Castellazzi A, Eleffendi MA, et al.A physical RCnetwork model for electrothermal analysis of a multichip SiC power module[J]. IEEE Transactions on Power Electronics, 2018, 33(3): 2494-2508.

[0079] [4] Ma Mingyao, Guo Weisheng, Yan Xuesong, et al. Compact thermal network model for thermal analysis of power modules in electric vehicles [J]. Proceedings of the CSEE, 2020, 40(18): 5796-5804.

[0080] [5]Bahman A S,Ma Ke,Ghimire P,et al.A 3-D-lumped thermal networkmodel for long-term load profiles analysis in high-power IGBT modules[J].IEEEJournal of Emerging and Selected Topics in Power Electronics,2016,4(3):1050-1063.

[0081] [6]Y.S.Muzychka,J.R.Culham,and M.M.Yovanovich,“Thermal spreadingresistance of eccentric heat sources on rectangular flux channels,”ASME.J.Electron.Packag.,vol.125,no.2,pp.178–185,Jun.2003.

[0082] [7]Geer,J.,Desai,A.,and Sammakia,B.(April 7,2007)."Heat Conduction inMultilayered Rectangular Domains."ASME.J.Electron.Packag.December 2007;129(4):440–451.

[0083] [8]K.R.Choudhury and D.J.Rogers,"Steady-State Thermal Modeling of aPower Module:An N-Layer Fourier Approach,"in IEEE Transactions on PowerElectronics,vol.34,no.2,pp.1500-1508,Feb.2019.

[0084] [9]H.-Y.Tang,H.-Y.Ye,X.-P.Chen,C.Qian,X.-J.Fan and G.-Q.Zhang,"Numerical Thermal Analysis and Optimization of Multi-Chip LED Module UsingResponse Surface Methodology and Genetic Algorithm,"in IEEE Access,vol.5,pp.16459-16468,2017。

Claims

1. A method for calculating the heat distribution of a double-sided heat-dissipating SiC power module, characterized in that, It is a fast numerical solution method for double-sided heat dissipation power modules with double-sided heat sink level fan-out type FOPLP package structure. The numerical algorithm used is based on the Fourier formula of heat conduction, which can calculate the lateral transfer of heat during the package conduction process. The described double-sided heat sink-level fan-out packaging structure is a high-power module composed of a single chip and its interconnect materials. The chip is located in the center of the structure, and the heat dissipation paths are the upper rearrangement layer RDL, the upper heat sink, and the lower solder and bottom substrate. In the numerical solution process, the model is simplified to a stacked structure of regular cuboid blocks with the same cross-sectional area, forming the basic model for double-sided heat dissipation. The chip in the module acts as a heat source, mainly transferring heat in the upward and downward directions. The model simplifies the chip to a thin film structure with length and width of a*b, and the area is denoted as A. s The longitudinal heat transfer region has a length and width of L. x *L y The region has an area of ​​A. c The chip provides power loss of Q, and the origin of the z-axis is located in the plane where the chip is located; the coordinates of the chip center are (x... c ,y c The lengths of the upper and lower conduction paths are t respectively. p , t n The air convection heat transfer coefficient is h. Let the temperature distribution function of the entire structure be T(x,y,z), and the ambient temperature be T. air Steady-state heat conduction satisfies the Laplace equation, specifically: Its general solution takes the form of: θ(x,y,z)=X(x)*Y(y)*Z(z),θ(x,y,z)=T(x,y,z)-T air (2) θ(x,y,z) represents the temperature rise caused by the heat from the heat source, and X(x), Y(y), and Z(z) are the coordinate functions when solving the Laplace equation using the separation of parameters method. For the entire structure, the four walls are insulated and there is no heat exchange with the outside, so the boundary conditions are: Therefore, the general solution to the Laplace equation is: Where λ=mπ / L x ,δ=nπ / L y , B i A i Let i be the coefficient, i = 0, 1, 2, 3; Heat transfer occurs between the upper and lower interfaces and the air via convection, with the following boundary conditions: Since the heat transfer directions of the z-axis are opposite, they need to be discussed separately. From z = t p and z = -t n The Fourier coefficients for heat transfer at the convective heat transfer boundary of the interface can be solved separately for the upper and lower heat transfers: B i =φ(ξ)A i , (5) For heat conduction in the positive direction, the Fourier coefficients are: For heat conduction in the negative direction, the Fourier coefficient is: Where ξ can be replaced by λ, δ, and β respectively; Let the total power generated by the chip be Q = Q1 + Q2, where Q1 and Q2 are the power on the upper surface and the power on the lower surface of the chip, respectively; When Q1 is input in the positive direction, the boundary of the chip layer consists of two parts: In A s The included area: In A s In areas other than the interface, the total heat conducted is equal to the total heat transferred out of the interface in the negative direction, expressed as: For the upward propagation of Q1, we have: For the case of Q2 propagating downwards, we have:

2. The method for calculating the heat distribution of a double-sided heat-dissipating SiC power module according to claim 1, characterized in that, Regarding power distribution for heat conduction on the chip's top and bottom: the junction temperature is represented by the average temperature of the chip's surface, denoted as T. j The average temperature of the upper and lower surfaces of the structure is T. p and T n The temperature transfer of heat conduction is described using a thermoelectric network. Since the ambient temperature is fixed, the upward and downward power transfer can be considered as parallel splitting. Specifically, Q1 and Q2 need to be distributed through the thermal resistance of the cuboid conduction and the convective thermal resistance with the air. Therefore, the thermal resistance is calculated first. In the calculation, the thermal resistance is inversely calculated using the temperature difference, specifically the difference between the average temperature of the chip surface and the ambient temperature. This represents the temperature difference between the junction temperature and the ambient temperature. According to the solution process of the Laplace equation for the steady-state heat of the chip, without considering heat conduction in areas other than the heat source on the chip layer, it is equivalent to the general solution for single-sided heat dissipation. At this time, the steady-state heat solution in the positive direction is: At this point, the difference between the chip junction temperature and the ambient temperature is expressed as: Therefore, the thermal resistance in the positive direction can be obtained as follows: For thermal resistance in the negative direction, the input power direction is negative, and φ needs to be replaced with a negative direction coefficient; The power distribution here is determined by the parallel relationship of thermal resistance: After calculating the power transfer in both directions, due to the linear superposition of temperatures, the temperature distribution of the double-sided heat dissipation base structure can be expressed as follows: The FOPLP package is equivalent to a six-layer structure, with four layers on top of the chip and two layers on the bottom. Substituting the package parameters into the numerical model established above yields the result.