Heat Sink Mounting and PCB Thermal Design
Adding a heat sink to a hot component is the obvious answer to a thermal problem, and it is often the wrong one. The heat sink is only the last element in a chain that runs from the silicon junction, through the die attach and the package, through the interface material, into the heat sink and finally into the air. The largest resistance in that chain decides the temperature, and it is frequently the interface rather than the heat sink.
The Thermal Path from Junction to Air
thermal resistance is the temperature rise per watt of dissipated power, and each element of the path has its own value. The junction to case resistance is set by the package, the case to heat sink resistance by the interface material and the mounting, and the heat sink to ambient resistance by the heat sink and the airflow around it.
The resistances add, so the total is dominated by the largest term. A package with a junction to case resistance of one degree per watt mounted on a heat sink with a resistance of five degrees per watt is limited by the heat sink, while the same device mounted with a poor interface of four degrees per watt spends most of its budget on the interface.
junction temperature follows directly from the total. A device dissipating ten watts with a total resistance of five degrees per watt in a forty degree ambient reaches ninety degrees at the junction, which may be at the limit of the part. Working backwards from the maximum junction temperature gives the resistance budget that the design has to meet.
Thermal Resistance and Junction Temperature
The maximum junction temperature is the number that matters, and it is usually between 125 and 150 degrees Celsius for a semiconductor and lower for some devices. The margin between the calculated value and the limit is the design margin, and it should be large enough to cover the variation in ambient, in airflow and in the interface material over the life of the product.
The thermal resistance of the heat sink itself depends on the airflow. A natural convection rating applies with the fins vertical and unobstructed, and it can easily double if the fins are horizontal or if the enclosure restricts the air path. A forced convection rating assumes a defined air velocity, and it is meaningless without it.
Derating for altitude is worth remembering in equipment that operates high above sea level. The air is thinner, the convective cooling is less effective and the heat sink resistance rises. Manufacturers often publish a derating curve, and it should be applied rather than assumed to be negligible.
Thermal Interface Materials
thermal interface material fills the microscopic gaps between two surfaces that are not perfectly flat. Without it, the contact area between the device and the heat sink is a small fraction of the apparent area, and the resulting resistance dominates the path. The material is a poor conductor compared with metal, but a thin layer of it is far better than the air it replaces.
Pads, grease and phase change materials differ in how they behave under pressure and over time. A grease gives the lowest resistance when freshly applied but can pump out over thermal cycles, while a pad is more consistent and easier to assemble. A phase change material behaves like a pad at room temperature and like a grease once warm.
The thickness of the layer matters as much as its conductivity. A thin layer of a moderate material can outperform a thick layer of a better one, which is why the mounting pressure and the flatness of the surfaces matter as much as the material specification. Excessive material also increases the resistance rather than decreasing it.
Mounting Methods and Pressure
mounting pressure determines how much of the interface material is squeezed out and how well the surfaces conform. Screws with a defined torque, spring clips and clamps all apply pressure, and each has a different distribution across the surface. A single screw in the middle of a large device bows the package and leaves the edges poorly contacted.
Where the device is mounted through the board, the solder joints carry both the electrical current and the mechanical load of the heat sink. A heavy heat sink mounted directly on a small package can crack the joints during handling or vibration, so a mechanical support that carries the load to the chassis is often required.
Insulating pads are used where the device tab is at a potential that must not reach the heat sink. The pad adds thermal resistance, and the mounting hardware has to be insulated as well if the screw would otherwise short the tab to the heat sink. Both effects have to be included in the thermal budget rather than being discovered when the first unit runs hot.
Heat Sinks on the Board: Copper, Vias and Pads
For a surface mount power device, the copper on the board is the first heat sink. A thermal pad soldered to a copper area spreads the heat laterally, and vias transfer it to the inner and bottom layers where there is more area. A pad connected with a few vias performs far better than the same area connected with one.
The via array should be placed inside the thermal pad area, with a reasonable pitch so that solder does not wick away during reflow. Vias that are not tented or plugged will draw solder through the board, which can leave voids under the device and degrade both the electrical and the thermal connection. The pad geometry rules that govern these features are described in our guide to PCB pad design standards.
Where a heat sink is mounted on the back of the board opposite the device, the thermal path runs through the vias and the laminate, both of which add resistance. Using more vias and a thinner laminate helps, and the copper area on the far side should be at least as large as the device footprint. The mechanical arrangement of that kind of assembly follows the rules in board outline and mounting design.
Mechanical and Assembly Considerations
The heat sink must not obstruct the placement of other components or the access needed for test and rework. A heat sink fitted before the board is tested can prevent probing of the very device it is cooling, so the assembly sequence should be reviewed alongside the layout.
Thermal cycling moves the assembly. The heat sink expands differently from the board, and repeated cycles stress the interface and the solder joints. Where the heat sink is large, a compliant interface material and a mounting scheme that allows some movement reduce the stress. Components near the heat sink also see a higher ambient, and their own temperature ratings have to be checked.
Finally, the heat sink itself has to be treated as a mechanical part with tolerances. A sink with a bowed base will contact only the middle of the device, while one with a rough surface needs more interface material. Both reduce the effective contact area, and both are invisible after assembly unless the base flatness is specified and checked. The fabrication and inspection practices that catch this kind of issue are covered in board quality characteristics.
Verification and Common Faults
Verify the thermal design by measuring, not by calculating alone. Attach a thermocouple to the device case, run the equipment at full load in the intended enclosure, and record the temperature until it stabilises. Comparing the measured rise with the calculated value shows which element of the thermal path is worse than expected.
A device that runs hotter than predicted at low power usually has an interface problem. Checking the mounting pressure and the amount of interface material, and looking for voids under the pad with X-ray, identifies it. A device that runs hotter than predicted only at high power may be saturating its heat sink or its airflow.
Temperature measured on the case is not the junction temperature. The difference is the product of the junction to case resistance and the dissipated power, which can be tens of degrees for a small package. Using the case temperature as the junction temperature is the most common error in thermal verification, and it leads to designs that appear comfortable on the bench and fail in the field.
FAQ
How much interface material should I use? Enough to fill the surface roughness and no more. The layer should be thin and even, and the mounting pressure should be specified so that the layer thickness is repeatable between units.
Is a larger heat sink always better? No. Beyond a certain size the heat sink resistance falls slowly while the weight and the mounting stress rise. Improving the interface usually gives more benefit for less cost.
Do I need thermal vias if the device has a metal tab? If the tab is soldered to the board, yes, because the board is part of the heat path. If the heat sink is mounted directly on the tab, the vias matter less but still help to spread heat from the surrounding area.
