High Power LED Thermal Design

A high power LED converts a large fraction of its input into heat, and that heat leaves through the back of the package. The optical design is what most people notice, and the thermal path is what determines whether the light output and the colour stay constant over the life of the product.

Where the Heat Goes

Roughly half of the electrical power becomes light in a modern white LED at moderate current, and the rest becomes heat in the junction. The light that is not extracted from the package is absorbed and becomes heat as well, so the effective fraction is lower than the internal quantum efficiency suggests.

Almost all of the heat leaves through the thermal pad on the underside of the package, because the plastic lens and the wire bonds conduct very little. The board, the interface material and the heat sink form a chain, and the total resistance is the sum of the parts.

junction temperature is the temperature of the semiconductor, and it is the number that the lifetime and the colour depend on. It is never measured directly in a product; it is calculated from the case temperature and the thermal resistance, or from the forward voltage at a known current.

thermal resistance in the Chain

thermal resistance is quoted in degrees per watt and it is additive along the chain. The junction to case figure is in the datasheet, the thermal pad and the board add their own, the interface material adds another, and the heat sink to ambient contribution completes the sum.

The board contribution is often underestimated. A plastic laminate with a few thermal vias has a resistance many times that of a metal core board, and the difference changes the whole design rather than merely improving it.

The sum is then multiplied by the power to get the rise above ambient. A total resistance of ten degrees per watt and a power of three watts gives a rise of thirty degrees, which on top of a forty degree ambient leaves a junction at seventy degrees.

Thermal Pads and Vias

A thermal pad under the package is the standard way of getting heat into the board, and it is soldered rather than glued. A solder connection has a far lower thermal resistance than an adhesive and it is also the electrical connection in most packages.

Under the pad, an array of thermal vias carries the heat to the other side of the board. The vias should be small and numerous rather than few and large, and they should be filled or capped to prevent solder wicking away from the pad during reflow.

The copper on the far side of the board spreads the heat laterally before it reaches the interface material. A large area of copper with a thin dielectric to the top layer is much more effective than a small pad with thick copper behind it.

Interface Material and Mounting

The interface material between the board and the heat sink fills the microscopic gaps between two surfaces that are never perfectly flat. Its thermal resistance depends on its thickness, which depends on the mounting pressure and on the flatness of the surfaces.

A thick pad of soft material tolerates poor surfaces and has a high resistance, while a thin layer of a hard material has a low resistance and demands flat surfaces and even pressure. The choice follows from the mechanical design as much as from the thermal requirement.

Even pressure is essential. A single screw through the middle of a board bends it away from the heat sink at the edges, and the air gap at the edge is worse than no interface material at all. Several screws or a spring clip that spreads the load is the answer.

Measuring the Junction Temperature

The junction temperature is measured from the forward voltage at a low measuring current, which is the standard method for a packaged device. The measurement is made immediately after the drive is removed, before the junction cools.

An alternative is to measure the case temperature with a thermocouple on the pad and add the junction to case rise from the datasheet. That is quicker and less accurate, and it is adequate for confirming that a design has margin.

Infrared measurement of the package is not reliable, because the emissivity of the lens and the surface varies. A thermal camera is useful for finding hot spots on the board, and not for reading the junction temperature.

Effect on Light Output and Colour

Light output falls as the junction temperature rises, by a few tenths of a percent per degree for a white LED. A design that is thermally marginal loses a noticeable part of its output once it has been on for a few minutes.

The colour also shifts, and for white LEDs the shift is towards the yellow because the phosphor behaves differently from the die as the temperature changes. In a luminaire with several LEDs, a difference in thermal resistance between them produces a visible colour difference across the fixture.

Both effects are reversible in the short term and permanent in the long term. The lumen maintenance curve, which describes the output after thousands of hours, is quoted at a specific junction temperature, and running hotter than that invalidates it.

Board and Heat Sink Design

The heat sink has to be sized from the power and the acceptable rise, and its thermal resistance is quoted for a free convection condition with a defined orientation. A heat sink in a sealed enclosure performs far worse than its catalogue figure.

Openings in the enclosure that let air in below and out above allow a convection current to form. Without them the heat sink reaches the internal air temperature and the enclosure becomes an oven.

For a high power luminaire the metal housing is often the heat sink, and the thermal path from the board to the housing is then the critical interface. The surface finish, the flatness and the interface material at that joint are the parameters to specify, and the general board level practices are in our guide to PCB thermal management design.

Verification and Faults

Verify the design by running the product at full power until the temperatures stabilise, which takes twenty to thirty minutes, and then measuring the case temperature. A design that is checked after five minutes is being measured on its way to the real temperature.

A design that is hotter than calculated by more than twenty percent usually has a problem in the interface rather than in the components. Check the flatness, the mounting pressure and whether the interface material has been compressed as intended.

A single LED in an array that runs hotter than its neighbours has a poor solder joint or a missing via. Comparing the thermal images of the units shows it immediately. The release checks that keep such a light engine consistent are collected in our PCB design release checklist, and the assembly points are listed in judging PCB quality.

FAQ

How do I know my LED junction temperature? Measure the case temperature and add the junction to case rise, or use the forward voltage method immediately after switching off.

Do I need a metal core board? Not always, but a plastic board with a few thermal vias has a much higher thermal resistance and limits the power that can be used.

Why is the light output lower than expected? Check the junction temperature. Output falls with temperature, and a thermally marginal design loses several percent once it is warm.