Why Metal Core Thermal Conductivity Measurements Differ

Two suppliers quote the same dielectric thermal conductivity for their metal core laminates and the finished boards perform differently. The discrepancy is not necessarily dishonesty; it is usually a difference in what was measured and how. Thermal conductivity of a thin bonded layer is not a simple property of a material, and the number in a datasheet depends on the method, the sample and the direction of the heat flow.

This article explains where the differences come from, what a meaningful specification looks like, and how to compare suppliers fairly.

The subject matters because the thermal performance of a metal core board is almost entirely determined by that thin dielectric layer, and the rest of the stack is straightforward to model.

Why A Thin Layer Is Hard To Measure

Thermal conductivity relates a heat flux to a temperature gradient. Measuring it requires establishing a known flux through a known thickness and measuring the temperature difference across it. With a layer a hundred micrometres thick bonded between copper and aluminium, both of which conduct far better than the dielectric, the temperature difference across the layer is small and the measurement uncertainty is large.

The practical consequence is that the thermal resistance of the whole stack is often easier to measure than the conductivity of the layer, and it is the more useful number. A specification that gives the thermal resistance of the construction at a defined copper area tells the designer what will happen, while a conductivity figure requires an assumption about the interface quality that may not hold.

The Measurement Methods

Several methods are used, and they do not agree. A steady state method applies a known heat flux and measures the temperature difference after equilibrium; it is accurate but slow and it requires careful control of the heat losses. A transient method applies a pulse and fits the response to a model, which is faster but depends on the model being correct.

Laser flash measures the thermal diffusivity of a sample and derives conductivity from it with a value for the specific heat and the density. The method works well on homogeneous samples and less well on a bonded stack where the interface is part of the structure being measured. Each method reports a value with a defined meaning, and the values are not interchangeable. The general properties of aluminium based metal core boards are affected by how the number was obtained.

The Interface Is Part Of The Answer

Heat crossing from the copper into the dielectric and from the dielectric into the aluminium passes through two interfaces. Each interface has a thermal resistance of its own, arising from incomplete contact and from the chemistry of the bond, and it can be a significant fraction of the total.

A measurement made on a free film of the dielectric will not include those interface resistances, and a measurement on a bonded stack will. That is one reason two datasheets can quote the same material and different performance. The way the interfaces behave in the finished board is discussed alongside the insulation and thermal path of the construction.

Thickness And Its Effect

Thermal resistance is proportional to thickness, so a dielectric layer that is thicker than nominal has a proportionally higher resistance. The thickness of a bonded layer is also difficult to measure directly, because it varies across the panel and it is affected by the copper pattern beneath it.

The pattern matters because the dielectric flows during lamination and fills the spaces between copper features. Under a large copper area the layer may be thinner than under an isolated trace, and the thermal resistance accordingly lower. A specification should therefore state the thickness and the copper coverage with which it was measured, and the manufacturing tolerances on that thickness should be part of the qualification.

Comparing Suppliers

A fair comparison uses the same quantity. The most useful is the thermal resistance of the board from the copper pad to the base, expressed in degrees per watt for a defined pad area, measured with the same method by both suppliers. Where such a figure is available, it is directly usable in the thermal calculation.

Where only a conductivity figure is offered, the comparison should be made at the same thickness and with a statement of the method. Converting between a film measurement and a bonded stack measurement is not possible without knowing the interface resistance, so a table that mixes the two is misleading even when every individual number is correct.

Temperature is a variable in the measurement as well as in the application. Thermal conductivity of most dielectric materials falls as temperature rises, so a figure measured at room temperature overstates the performance of a board that will operate hot. Where the application runs at a high temperature, the comparison should be made at that temperature, and the supplier should be asked for the temperature dependence rather than the room temperature value.

Modelling And Its Limits

A thermal model of a metal core board is straightforward once the layer resistance is known: the copper spreads the heat, the dielectric resists it, and the base conducts it away. The uncertainty in the model is concentrated in the dielectric and in the spreading resistance of the copper.

That is why measurement on a real assembly is worth the effort. A thermocouple bonded to the base and a measurement of the component’s case temperature under a known load give an effective thermal resistance for the assembly, including the solder, the copper and the interface. The number is less elegant than a model and considerably more reliable.

The copper area and its distribution change the effective performance in a way that a single conductivity figure cannot express. A wide copper area spreads the heat before it enters the dielectric, which reduces the flux density and therefore the temperature drop. For a small component on a large pad, the spreading dominates the calculation, and the dielectric resistance is only part of the answer.

Practical Consequences

Three habits follow from the above. Specify the thermal resistance of the construction rather than only the conductivity of the material. Ask for the thickness and the copper coverage that the figure was measured with. And qualify a new supplier by measuring a test coupon from a real production panel rather than by comparing datasheets.

Where the application is thermally critical, the qualification should include an accelerated life test with thermal cycling, because a dielectric whose thermal resistance is acceptable when new may degrade as the bond ages. The interface is the part of the construction that is most likely to change.

It is worth recording the measurement conditions whenever a number is used in a design. The method, the temperature, the sample thickness and the copper coverage should travel with the figure, because a design decision made from a number whose context has been lost is a decision that cannot be defended when the product runs hot.

FAQ

Is a higher thermal conductivity always better? It is better thermally, but the materials with the highest conductivity sometimes have lower dielectric strength or poorer adhesion, so the choice is a compromise between the thermal and the electrical requirements.

Can the metal core be modelled as an ideal heat sink? It is a good approximation for the base, because aluminium conducts far better than the dielectric. The limitation is the interface between the base and whatever it is mounted to, which is a separate thermal resistance.

How much does the copper thickness matter? It spreads the heat laterally, so a thicker copper layer reduces the temperature directly under a small component. Beyond a certain area the spreading is no longer the limiting factor and the dielectric dominates.