Key takeaways
- Thermal resistance is a chain, and the largest term is often the interface rather than the sink.
- A datasheet figure is only valid at the airflow, source size and mounting condition it was measured in.
- Spreading resistance in a thin base can exceed the fin resistance on a small, high power device.
- A test should report ambient, power, reference temperature and the point of measurement, or the result cannot be compared.
The resistance chain
Heat leaves a semiconductor junction and passes through several materials before it reaches the air. Each layer has a thermal resistance, and the resistances add in series. The junction to case resistance belongs to the device package. The case to sink resistance belongs to the interface and the mounting pressure. The sink to ambient resistance belongs to the fin geometry and the airflow.
The simple model is a series of resistors. Junction temperature equals ambient temperature plus the total resistance multiplied by the power dissipated. That model is useful for steady state and for comparing designs, but it hides two effects. The first is spreading: heat enters the base over a small area and has to spread sideways. The second is that the sink to ambient resistance is not a property of the sink alone, because the enclosure and the airflow change it.
A good calculation therefore has four terms, not two: junction to case, case to sink, spreading in the base, and sink to ambient. The first three are the ones most often assumed.
Reading a datasheet without being misled
A heat sink datasheet usually quotes a thermal resistance at a defined air velocity, input power and source size. The source size matters because a small source spreads less heat into the base. The velocity matters because the convection coefficient rises with airflow. The mounting matters because a clamped sink with a thin pad is not the same as a sink resting on a bench.
A figure measured with a large, uniform heat source flatters the sink. A real device may contact only a fraction of the base, so the spreading resistance is higher. A figure measured at 3 m/s does not apply to a sealed enclosure with no fan, and a figure measured with the fins vertical does not apply under a shelf.
The practical response is to ask three questions before using a catalogue number: what was the source area, what was the air velocity and how was the sink mounted? If the supplier cannot answer, use the number for a first estimate and test the assembly.
A worked calculation
Take an illustrative device that dissipates 60 W, with a maximum junction temperature of 125 degrees Celsius and a maximum ambient of 40 degrees Celsius inside the enclosure. The allowable total resistance is the temperature difference divided by the power, which is 85 divided by 60, or about 1.42 K/W.
Now add the chain. The device datasheet gives a junction to case resistance of 0.35 K/W. A thermal pad at the mounting pressure gives 0.10 K/W. Spreading from a small source into a 6 mm base adds about 0.15 K/W. The sink, at the available airflow, is rated at 0.55 K/W. The total is 1.15 K/W, so the junction sits at 40 plus 60 times 1.15, or 109 degrees Celsius.
The 16 K margin is useful but not generous. A dirty fan, a blocked filter or a hotter day can remove it. The sink is less than half of the total, and the interface and spreading together are about 22 per cent, which is why a better sink alone may not fix a hot unit.
| Term | Value | Share of total |
|---|---|---|
| Junction to case | 0.35 K/W | 30 per cent |
| Case to sink, through the pad | 0.10 K/W | 9 per cent |
| Spreading in the base | 0.15 K/W | 13 per cent |
| Sink to ambient | 0.55 K/W | 48 per cent |
| Total | 1.15 K/W | 100 per cent |
Values are illustrative and chosen to show the method. Use the device datasheet and a tested sink figure for a real design.
Spreading resistance and base thickness
Spreading resistance appears when heat enters a base over an area smaller than the base itself. The heat travels sideways through the base before it can reach the outer fins. A thin base has little sideways conductance, so the fins above the device do most of the work and the outer fins run cool.
The effect is largest when the device is small relative to the sink and the base is thin. A 10 mm square device on a 100 mm wide sink with a 3 mm base is a spreading problem. The same device on a 6 mm base is much better. Copper spreaders or a vapour chamber are used where the spreading resistance dominates and the weight budget allows.
The calculation is not a single formula because spreading depends on the source size, base thickness, base area and the boundary condition at the fins. Finite element analysis or a dedicated correlation is the usual route for a critical design. For an early estimate, treat the base as a constriction resistance.
The interface is part of the sink
The case to sink resistance is set by the interface material, the bond line thickness, the contact pressure and the flatness of both surfaces. A thermal pad is easy to assemble but has a bulk resistance and two contact resistances. Grease fills the gaps but can pump out under cycling. A phase change material is dry at room temperature and flows when hot.
Pressure matters. A pad needs enough pressure to conform to the surfaces, and a clip that looks tight may deliver much less pressure than a screw at the specified torque. The pressure also has to be spread across the device, not concentrated at one corner. A warped base or a bowed device contacts at the high points and leaves the rest of the interface thick.
The interface is also where a design review usually finds the cheapest improvement. A thin, higher conductivity pad can remove more resistance than a larger sink. The gain depends on the pressure and the flatness, so it should be measured.
Testing the assembly
A thermal test applies a known power to the device, holds the assembly in a defined environment, and measures the temperature rise at steady state. The result is a resistance in kelvin per watt. The test is simple in principle and easy to get wrong in practice, because the ambient, the power and the measurement point are often not reported.
Measure the ambient at a point that is not heated by the sink, and shield the sensor from radiation. Measure the reference temperature at a defined point on the device or the sink base, not on a fin. Apply the power with a stable load, and record voltage and current so the power can be recalculated. Wait until the temperature changes by less than a small amount over a defined interval.
Report the result with the airflow, orientation, interface material and mounting pressure. A resistance without those conditions cannot be compared with another test. For a production acceptance, define the sample size, the pass limit and the action on a failure. A one-off laboratory result is useful for design, not for shipping.
Measure the air, not the room
A sensor near the enclosure wall can read several degrees below the air entering the fins. That error goes straight into the resistance and makes the sink look better than it is.
When the calculation and the test disagree
A test that runs hotter than the calculation usually means one of four things. The airflow is lower than assumed, the interface is thicker or poorer than assumed, the spreading resistance was underestimated, or the inlet air is hotter than the room. Each has a different fix.
Start with the airflow. Measure the velocity at the fin entry, not at the fan outlet. Then compare the sink base temperature under the device with the case temperature. A large difference points to the interface. A small difference with a hot base and cool fins points to spreading. A hot sink with cool air points to the fin geometry or the airflow path.
A test that runs cooler than the calculation deserves the same scrutiny. The ambient may be measured too close to the sink, the power may be lower than recorded, or the device may be switching in a way that reduces the average loss. A design with no margin will fail in the field, even if the first test passes.
Frequently asked questions
What is a good thermal resistance for a heat sink?
There is no single good value. The sink resistance has to fit the allowable junction temperature, the ambient and the other terms in the chain. A 0.5 K/W sink suits a small natural convection box and is ordinary for a large forced air unit.
Why is my measured thermal resistance higher than the datasheet?
The common causes are lower airflow than the datasheet condition, a poorer interface, a smaller heat source than the test used, and a hotter inlet than the room. Measure the air at the fin entry and the base temperature under the device.
Does a thicker base always lower thermal resistance?
It lowers spreading resistance, but it adds weight and cost and does little for the fin to air resistance. The benefit is largest when the device is small relative to the sink. Beyond a certain thickness, a copper spreader may be better.
How long should a thermal test run?
Long enough for steady state. A practical rule is to wait until the change is less than about 0.5 K over 15 minutes, then record for a further period. The time depends on the thermal mass of the sink and the enclosure.
Sources and further reading
- JESD51-1, Integrated circuit thermal measurement method. Electrical test method
- JESD51-2A, Integrated circuits thermal test method environmental conditions. Natural convection (still air)
- JESD51-6, Integrated circuits thermal test method environmental conditions. Forced air convection
Qumek Engineering Desk
Engineering team
The engineering team behind Qumek drawings, DFM notes, tolerances and inspection standards. Every guide is reviewed against the drawings and processes used on live programmes.
Have a thermal calculation that needs a test behind it?
Send the loss, the ambient and the proposed stack. The engineering desk will mark the terms that dominate and the measurements that will settle the question.
Send the thermal case