Power LEDs should always be mounted on a heat sink. For example, a traffic light using six or seven 1-W LEDs could be mounted alongside the driver electronics on a 6-in. (150 mm) diameter circular PCB. A heat sink could be mounted on the backside of the PCB for removing heat from both the LEDs and the driver electronics. As traffic lights may have to work in high-ambient temperatures, a good thermal conductivity is required.

When designing analog or switching power sources, we must always consider efficiency. This is the ratio power out/power in, and is usually expressed as a percentage. What designers sometimes overlook is that input power minus output power equals power loss in the LED driver circuit (Fig. 14.1). Loss in the driver must be dissipated as heat. A switching LED driver with 90% efficiency, driving a 10-W load will require an input power of: 10 W/0.8 = 11.1 W. This means that 1.1 W is the power loss and must be dissipated in the LED driver circuit, which raises the temperature of the circuit.

Another consideration is the maximum operating temperature of the LED driver circuit. If the ambient temperature around the circuit board is raised too high there could be reliability issues with components. Active components, such as integrated circuits, become less able to dissipate their internal power if the ambient temperature is high. Silicon-based ICs are usually okay up to 125°C ambient, although some are only rated to 85°C. Passive components also need to dissipate heat, so a 0.25 W–rated resistor may have to be derated at high-ambient temperature. Also, as we discussed in Chapter 11, electrolytic capacitor lifetime will be shortened by high-ambient temperature.

Like electrical resistance, thermal resistances can be added when connected in series (Fig. 14.3). Consider a TO-220 package mounted onto an aluminum heat sink. The thermal resistance between the silicon die and the package, added to the thermal resistance of the package to heat sink interface and the thermal resistance of the heat sink to air interface, can all be added together to find the total thermal resistance from the silicon junction to air.

Thermal resistance is given as degrees Kelvin per watt of heat flow and has symbol theta. (Note, a 1K temperature rise = 1°C temperature rise.) The end points of the resistance are given as subscripts: junction (J), case (C), heat sink (H), and ambient (A). So, from junction to case, the thermal resistance is labeled as θ_JC. For example, let θ_JC = 1.2 K/W, θ_CH = 0.1 K/W, and θ_HA = 2.4 K/W, so the case to heat sink resistance is 0.1 K/W. The sum of thermal resistance from junction to ambient is 3.7 K/W; so when the device is dissipating 10 W, the silicon junction temperature will be 37 degrees hotter than the ambient temperature. If the ambient temperature is 25°C, the silicon junction temperature will be raised to 62°C.

Like electrical resistance, having parallel thermal resistance paths reduces the overall resistance (Fig. 14.4). Two paths, each of 2 K/W will create an effect single path of 1 K/W. This makes calculation of the exact temperature more difficult, as thermal paths are not usually as obvious as electrical paths. However, for a first approximation, calculating the temperature drop along obvious thermal paths will give a sufficiently accurate result. Less obvious thermal paths usually have a much higher-thermal resistance and have little effect on the temperature.

As parallel paths reduce the thermal resistance, in general a large-surface area can dissipate heat much better than a small area. Conversely, a small-surface area cannot dissipate a lot of heat. For this reason, a small driver is rarely able to drive a high-power load; remember this when the marketing department asks you to design a smaller driver!

Semiconductor component manufacturers usually specify the minimum and maximum junction operating temperatures for their devices. It is usual for the temperature range to be −40°C to +125°C, but this is not the ambient temperature. Commercial device ambient temperature ratings are 0–70°C, industrial device ratings are −40°C to +85°C. Military and automotive devices are rated for ambient temperatures of −55°C to +125°C, but these require special processing of the silicon material and packaging to achieve a +150°C junction operating temperature and hence are usually more expensive.

Component manufacturers also specify power dissipation (usually based on 25°C ambient temperature). Most manufacturers also provide thermal resistance information in their datasheets and some provide notes on heat sink requirements, which can be very useful to designers.

Surface-mounted power MOSFETs are usually in a D-PAK or D2-PAK housing, which have a tab for dissipating heat. However, this means that the tab must be soldered to a copper surface area on the component side of the PCB, or otherwise a surface-mount heat sink is required (Fig. 14.5). A surface area of one square inch (25 × 25 mm) on a standard FR4 fiberglass board can give a thermal resistance of θ_JA = 30 K/W with a D-PAK device. Surface-mount heat sinks are sometimes made from tinned brass and are soldered to the PCB, either side of the MOSFET body. These reduce the thermal resistance to about θ_JA = 15 K/W.

Through-hole MOSFETs in a TO-220 package can be fitted to various heat sinks with a wide range of sizes. A small heat sink can be fitted, supported by the MOSFET pins or bolted onto the PCB through the TO-220 tab. Larger heat sinks could increase parasitic capacitance and cause an increase in switching losses, but this can be prevented if the heat sink is connected to the ground plane. Grounding the heat sink also prevents undue EMI radiation, but the MOSFET should be electrically isolated from the heat sink using a thermal conducting pad (made from a flexible material to give a large-surface contact). Switching losses will be due to the capacitance between the MOSFET drain (tab) and the heat sink.

Even where electrical isolation between the MOSFET and heat sink is not required, a thermally conductive pad or paste is a very good idea. This is because the surface of the MOSFET tab and the heat sink surface are not smooth. Without the thermally conductive pad or paste, the actual area where good contact can be made is just a small fraction of the surface available. Microcavities between the two surfaces creates air pockets with high-thermal resistance (Fig. 14.6). The thermally conductive pad or paste fills these cavities to create a uniform surface with low-thermal resistance.

When several devices on a circuit board generate the heat, a solution could be to use a fan to blow air across the circuit board. Cooler air from outside the equipment can be blown over warm components to reduce their temperature. Airflow will reduce the effective thermal resistance of the air interface.

Careful placing of cooling fans can make a big difference to the performance. Large objects, such as electrolytic capacitors, will tend to block the flow and may steer the cooling air away from areas of the PCB. If the air flows in the direction of heat sink fins, it will be more effective. Air flowing across the fins will only cool the front and rear fins (Fig. 14.7).

If mounting a fan at the top of equipment, make sure that the air flows upward, with the fan blowing air outward, so that the fan aids the natural buoyancy of the hot air. Fans mounted in the side of equipment are much more effective if two fans are used, one on either side of the enclosure. In a wide enclosure, both fans could be mounted on the rear panel; one fan should blow in and the other should blow out so that air circulates around the components inside.

Fans do have a reliability issue, so consider adding a fail-safe mechanism in case the fan fails to operate. A fail-safe mechanism should monitor the temperature of sensitive components on the circuit board. Driving the LEDs at a lower power or turning them off when the temperature rises too high may be a solution.