Thermal Simulation For Metal Core Board Design

A metal core board is chosen when the heat a circuit produces cannot be moved by copper alone. The aluminium base spreads heat far better than a laminate, but the path from the device to that base passes through a thin insulating layer that dominates the thermal resistance of the whole stack. Predicting the temperature of the device therefore means modelling the path rather than applying a rule of thumb.

This article describes how heat flows through a metal core board, how to set up a useful thermal simulation, and where the model has to give way to measurement.

The structure is simple enough to describe in a sentence: a circuit layer of copper, an insulating layer, and a metal base. The behaviour is not, because the current and the heat are rarely uniform.

The Heat Path Through The Stack

Heat leaves a device through its thermal pad into the copper beneath it, crosses the dielectric layer, and spreads into the aluminium base, from which it reaches a heatsink, an enclosure or the surrounding air. Each step has a thermal resistance, and they add in series, so the largest one dominates the total.

In most metal core designs the dielectric layer is that step. It is thin, but its thermal conductivity is far lower than that of copper or aluminium, and it is the deliberate compromise that makes the substrate possible: the layer has to insulate electrically while conducting heat, and those two requirements pull in opposite directions. Everything else in the design, including the copper area and the base thickness, exists to feed heat into that layer efficiently and to remove it from the other side.

The Dielectric Layer Sets The Limit

Thicknesses for the insulating layer are typically in the range of fifty to one hundred and fifty micrometres. Thinner means a shorter path and lower thermal resistance, which is why the highest performance boards use the thinnest layer the voltage allows. But a thinner layer also means a lower breakdown voltage and a more demanding process, since the coating has to be uniform over a large area without pinholes.

The material matters as much as the thickness. Ceramic filled epoxy and silicone based dielectrics are the common choices, and their thermal conductivity covers a wide range, with the better materials several times more conductive than the basic ones. A design that needs a specific junction temperature has to start by selecting the layer for its conductivity and its dielectric strength, because no amount of copper spreading can compensate for a layer that does not conduct.

Copper Area, Spreading And Vias

Heat enters the substrate over the area of the device pad, and it has to spread laterally to use the whole base. A copper area larger than the component, connected under it and spreading outward, reduces the local flux density and the temperature of the layer immediately beneath the device. That is why metal core layouts use generous copper areas in the high power region rather than minimum geometry.

Where the board is a multilayer metal core construction, thermal vias connect the surface copper to an inner layer, and their number and arrangement determine how effectively the heat is shared. The design rules for thermal vias are like the electrical ones in principle, but the objective is different: the aim is the shortest practical path from the source into the spreading layer, with enough vias that no single one is carrying a disproportionate share. In a metal core board the base itself usually provides the spreading, so the vias matter most at the interface.

Setting Up The Simulation

A thermal simulation of a metal core board only needs to be as detailed as the decision it supports. The pieces that have to be right are the power dissipated by each device and its location, the layer stack with the correct thicknesses and conductivities, the copper geometry in the high power region, and the boundary condition at the back of the base, which is the interface to the heatsink or the air.

The boundary condition is the element most often got wrong. A model that assumes a perfect heatsink will predict a device temperature that the product can never achieve, and one that ignores the interface material between the base and the heatsink will be optimistic by the resistance of that layer. The air side needs a realistic convection coefficient for the enclosure, not for an open bench. Where the current in a trace is what generates the heat, the trace geometry has to be modelled too, and sizing a trace for current gives the input for that calculation.

Reading The Results

The useful output is not a single temperature but a map. The device junction temperature is what the reliability calculation needs, and the peak copper temperature is what tells you whether the dielectric layer is inside its rating. A map also shows whether heat is spreading as intended or whether it is concentrating in a narrow path, which is usually a sign that the copper area is smaller than it should be.

Comparing two versions of the same design is more reliable than trusting one absolute result, because the errors in the model largely cancel when only the geometry changes. Running the model with a larger copper area, or with a thicker base, and comparing the result is a good way to decide whether the change is worth the cost. The mechanical outline and mounting also belong in the model, since the route from the base to the environment is often decided there.

Where Measurement Takes Over

The model gives a prediction, and the first article gives the measurement. Thermocouples or infrared imaging on a board carrying its rated load will show whether the prediction was close, and a large disagreement usually means the boundary condition was wrong rather than the material properties.

Measuring at several points is worth the effort: at the device, on the copper beside it, on the base close to the source and at the far edge. The gradient between them shows whether the bottleneck is where the model assumed it was, and where the measured temperature is lower than predicted, the design has margin that can be traded for cost. gopcb builds metal core boards and reviews the thermal design with the customer, including the layer thickness and the copper area that the simulation assumes, so that the built board matches the model.

Process Control and Verification

On a design of this kind, copper area is the item that decides how the rest of the board is arranged. Running a first article through the same checks as the production panel confirms that the two agree, and that comparison is the cheapest form of process control available at prototype stage. Keeping a sample from the panel turns a dispute into a measurement, because the same coupon can be re-examined by both parties without rebuilding the batch.

FAQ

Does a metal core board always run cooler than an FR-4 board? It spreads heat far better, but the dielectric layer is a thermal resistance of its own. The benefit depends on how much area is available for spreading.

How accurate is a thermal simulation? It is accurate enough to compare design options and to identify hot spots. Predicting an absolute temperature within a few degrees requires the boundary conditions to be characterised.

Should the dielectric layer be as thin as possible? Thinner is better thermally and worse for dielectric strength. The choice is a voltage decision first and a thermal one second.