Temperature limits both the performance and the life of electronics, and the temperature of a component depends on the whole path from the junction to the ambient air: die, package, interface material, heat sink, board and airflow. Conjugate heat transfer (CHT) simulation solves that path as one coupled problem. This guide covers what the method involves, the modelling decisions that drive accuracy, and what a useful study should report.
What CHT couples, and why a prescribed coefficient falls short
A CHT model solves the flow and energy equations in the fluid and the heat-conduction equation in the solids together. At every fluid–solid interface it enforces two conditions: the temperature is continuous, and the heat flux leaving one side equals the flux entering the other. Nothing about the surface heat transfer is assumed in advance.
The simpler alternative is to solve conduction in the solids alone and apply a heat-transfer coefficient h on the wetted surfaces. This has three weaknesses in electronics work.
- h is not a property. It varies over each surface with the local flow, and correlations exist only for idealised geometries.
- h is defined against a reference temperature. Inside an enclosure the air warms as it passes each component, so downstream parts see hotter air than upstream ones.
- Conduction in the board and heat sink redistributes heat, which changes where it enters the air, which in turn changes the flow in buoyancy-driven cases.
Treating the fluid–solid interface
A conformal mesh, with matching faces on both sides of the interface, is the cleanest option. Non-conformal interfaces with interpolation are acceptable where cell sizes on the two sides are comparable. Thin sheets such as enclosure walls or shields can be represented as shell-conduction layers without a volume mesh. Always confirm that the heat leaving the solids equals the heat entering the fluid in the converged solution.
Natural or forced convection, and when radiation matters
In fan-driven systems convection usually dominates, and buoyancy and radiation are secondary. In sealed or fanless products the balance is different. Buoyancy drives the flow, so gravity direction and product orientation become inputs, and density must vary with temperature: the Boussinesq approximation for small temperature differences, an ideal-gas density otherwise.
Radiation is easy to forget and often significant. Linearised, its equivalent coefficient is about 4εσT3 (T in kelvin), which for a high-emissivity surface a few tens of degrees above room temperature is roughly 6 W/m2K. That is the same order as natural convection in air, which is typically a few W/m2K.
In natural-convection cooling, include surface-to-surface radiation by default and use realistic emissivities. Bare aluminium has a low emissivity; anodised or painted surfaces and most plastics are high.
Turbulence and transition in enclosure flows
Reynolds numbers inside enclosures are often low. Flow between fins and boards may be laminar, flow behind a fan is turbulent and swirling, and much of the rest is transitional. A high-Reynolds-number k-ε model with wall functions tends to over-predict mixing and heat transfer where the real flow is laminar, which gives optimistic temperatures.
Reasonable practice is a low-Reynolds-number formulation such as k-ω SST on a wall-resolved mesh, or a laminar solution where the Reynolds or Rayleigh number supports it. Where the regime is uncertain, run both: the difference shows how much the answer depends on that assumption. Our article on choosing a turbulence model covers the options.
Compact or detailed component models
A detailed package model, with die, die attach, substrate and mould compound, is justified for the few components that limit the design. For the rest, compact models keep the problem tractable:
- Two-resistor models use junction-to-case and junction-to-board resistances from the datasheet.
- Multi-resistor network models of the DELPHI type are intended to remain valid across different cooling environments.
- Printed circuit boards are commonly modelled as orthotropic solids, with in-plane conductivity far higher than through-plane because of the copper layers. Copper planes and thermal vias under critical parts may need explicit representation.
A datasheet junction-to-ambient resistance is measured in a standard test environment. It is not a substitute for a system-level model.
Thermal interface materials and contact resistance
The interface between a package and its heat sink is thin, but it often carries a large share of the total temperature rise. Model it as a thermal resistance on the interface, not as a meshed layer. The resistance combines the bulk term, thickness divided by conductivity, with contact resistance at both faces. The installed bond-line thickness depends on clamping pressure and surface flatness, so treat the value as an uncertain input and run a sensitivity case.
Mesh and time-scale disparity
The fluid needs boundary layers resolved on heated surfaces. For heat transfer that means a wall-resolved mesh, as discussed in y+ and near-wall meshing. The solids need resolution where conduction gradients are steep: thin fins, spreaders and around heat sources.
Time scales differ even more. Air passes through an enclosure in seconds, while the solids take minutes to hours to warm up. For transients such as power cycling, marching both regions at the fluid time step is wasteful. In forced convection the flow field can often be frozen while the solid temperatures evolve. In natural convection the flow depends on the temperatures, so it must be updated periodically.
What to report
- Junction and case temperatures for each critical component, against their limits and at the stated ambient.
- Location of hot spots on boards and enclosure surfaces.
- The heat-flow breakdown: how much leaves by each path, including radiation.
- System pressure drop plotted against the fan curve. The operating point is where the two intersect, and it is always below the fan's free-air flow rate.
- Sensitivity to the uncertain inputs, and evidence of mesh convergence for the reported temperatures.
How CFD Pro can help
CFD Pro provides thermal and CFD simulation services and supports clients from problem definition through to verified results and reporting. To discuss an enclosure, heat sink or board-level cooling problem, send us a project brief.
