Giles Humpston, Applications Manager
LEDs and tungsten filament bulbs have two things in common. First, they convert electricity into light. While neither technology is particularly good at this task, LEDs do at least manage to achieve around 40 percent electrical-to-optical conversion efficiency compared with about 5 percent for a traditional Edison light. In both cases the energy wasted appears as heat. Filament bulbs shed the heat by radiation and convection, while for LEDs the heat is removed by conduction through to a heat sink.
Key to the success of manufacturing a reliable LED light is minimizing the thermal resistance between the LED and the heat sink. The LED cannot be placed directly on the heat sink since it also needs electrical connections. Thus the LED is mounted on a PCB, which in turn is attached to the heat sink. Owing to this arrangement choice of PCB is critical. For high-power LEDs, the thermal conductivity of traditional organic PCBs, like FR4, is orders of magnitude too low.
Fortunately there exists a special class of PCBs engineered to provide exceptional through-thickness thermal conductivity. They are known by a variety of names, including “insulated metal substrate” (IMS) and metal clad PCB (MCPCB). These PCBs are generally of a tri-layer construction comprising a metal plate, a dielectric layer, and a copper tracking layer. The metal plate does the majority of the ‘heavy lifting’ in terms of heat transport, while the dielectric provides electrical isolation between the copper tracking and the metal plate.
Very few dielectric materials are good thermal conductors. Ceramics like beryllium oxide and aluminium nitride superficially look attractive but must be applied as particle-loaded polymers. At face value the thermal conductivity of these composite materials is quite good. However the packing density of particles means the minimum layer thickness is tens of microns, so the thermal resistance of the dielectric ends up being quite high. Often too high for the latest generation of high brightness and short wavelength LEDs.
Some dielectric materials can be applied directly to metal. Spray coatings are generally too thick and the surface finish too rough for an LED PCB. At least one company is depositing thin layers of diamond as the dielectric, but the cost/benefit ratio might be unsuitable for some LED products. Another approach is to convert the surface of the metal to its oxide. Metal oxides are generally insulators. Aluminium metal is an excellent thermal conductor and the surface can be converted to alumina by anodizing, electrochemical oxidation and other processes. Alumina does not have particularly good thermal conductivity, around 25W/mK in sintered form. But, it has good dielectric properties, around 20V/um. Given LEDs are low voltage devices only a very thin layer of alumina dielectric is required so the thermal resistance (= thermal conductivity divided by thickness) of the dielectric can be very small.
Nano materials are defined as having an average grain diameter less than 100nm. This threshold is a little arbitrary since a true nano material has properties that are different from the bulk form. Effectively they are entirely new materials. For example, nano silver is being evaluated as a die attach material for LEDs to replace hard solder (gold-tin eutectic). Nanograin alumina has a dielectric strength around 60V/um. This makes Nanoceramic coated aluminium an interesting candidate in the challenge to develop new PCBs for the LED market.
The second and a lesser known fact about LEDs and tungsten filament bulbs is that both fail in a near identical manner. This can be easily verified by connecting either type of device to an inappropriate power supply, having first taken the precaution of replacing any inconvenient fuses with a short length of welding rod. The first stage of failure occurs when the LED or filament bulb partly or wholly vaporizes, emitting a brief but brilliant flash of light. This will be followed moments later by a cacophony from the exploding power supply and conclude with smoke pouring from the building utility transformer. Quod erat demonstrandum.
About the Author
Dr. Giles Humpston is a metallurgist by profession and has a doctorate in alloy phase equilibria. He is a cited inventor on more than 250 patents and has co-authored over 150 papers as well as several text books. Dr Humpston currently works as the Field Applications Manager for Cambridge Nanotherm on thermal substrate technologies.