by Dr. Johannes Adam, Flomerics Ltd. (LpR Issue 04, p 45-48)

Readers of this journal are well aware that the junction temperature TJ of LEDs plays a key-role for their performance; not only for light output (efficacy) but also for wavelength stability. The cooler the LED, the better will be the performance. However, the temperature of the LED is not only a function of its own termal power loss, but also depends on the other heating components on the board and the cooling and heating paths due to ambient conditions, i.e. the mechanical and thermal environment in the final application. These ambient conditions might be quite harsh e.g. in car entertainment systems where high ambient air temperatures can occur, other heat sources are near-by and space for cooling is precarious.

It is not only necessary to find a good solution for the temperature problem but, also to find it as early as possible, so that any necessary heat control measures can be implemented in good time and at a reasonable cost. Modifications to the finished design are always the most expensive solution. Also, one would like to know the answer with reasonable accuracy. Frequently found rules of thumb (for example simple temperature estimates with thermal resistance calculations) only hold true for specific geometrical configurations or physical assumptions and yesterday’s empirical values probably cannot be extrapolated to today’s situation. Experimental mock-ups (heating resistors and “shoeboxes”) are not precise enough. Flomerics advocates numerical simulation as still the most precise way to get realistic early temperature predictions (without hardware prototypes). It is flexible, provides insight into the heat flow paths and it can be used in early phases of the product development. As development proceeds, early estimates for model parameters and geometry can then be increasingly replaced with hard and fast values. .
Thermal calculations
All three software packages mentioned above analyse the system, calculating airflow, infrared radiation, and heat conduction on a 3D calculation mesh. The algorithms are similar to those used to forecast the weather but much of the work is automated by the program, making them much easier than weather forecasting. For example, the user doesn’t have to worry about the heat transfer between the surface of the components and the air, and the mesh generation is fast and intuitive. The degree of detail in the virtual model is controlled by the engineer and the project status. FLO/PCB is focusing on the thermal scene on and around a PCB. It is simple to use and very intuitive: components, layer patches and thermal vias can be defined by drag and drop, by numbers or by extracting from libraries or partially from IDF files or electronics layout software. The calculation process is fully automated and therby suitable for an electrical engineer. FLOTHERM goes beyond the PCB level where the user (usually a mechanical engineer with more background knowledge about electronics cooling) is creating a geometrical model of the device by selecting and placing objects (enclosures, circuit boards, perforated sheeting, fans, heat sinks etc.) and assigns the physical parameters (power dissipation, materials etc.) to the parts. The calculation and meshing can be controlled in various ways. PCB geometry from FLO/PCB can be imported in a FLOTHERM model. While FLOTHERM can be used stand alone and is strongest, most effective and very fast with rectangular cartesian geometries (one might say in prototyping a thermal model), EFD is embedded – really embedded - in CAD Systems such as Solid Works, ProEnginneer or Catia V5. According to the Flomerics philosophy, the burden of complicated meshing or numerical analysis expert knowledge is taken away from the user. A volume model of a CAD geometry can be turned quickly into a calculation model. EFD is adressed to mechanical designers who have to use pre-defined CAD parts from a company wide library and whos geometry is far from beeing rectangular. All in common to all simulation tools is that only good input gives good output. This holds especially for thermal power loss, spatial dimensions (area and thickness) and material properties (thermal conductivity). Geometrical shape is a form factor of 2nd-level importance - more or less - mostly.In the following sections we will describe some examples to illustrate the simulation approach for LED applications.

Light-emitting diodes
In terms of its inner structure the LED is still one of the simplest electronic components but it is still neither possible nor desirable to include all the details in a model. You need a good (!) data sheet with clearly-defined conditions for the thermal resistance between the chip and the circuit board, which effectively translates the inner structure into the thermal resistance between the chip and the heat spreader. For an example we use an Osram TOPLED LA E67F with a thermal resistance junction to board of Rth JB=130 K/W given in the data sheet and in the application note. The simulation model of the LED consists of no more than a 2-resistor network of size 3 x 3.4 x 2.1 mm and a FR 4 substrate mounted on an aluminium base plate with adhesive.

...

Full article and access to all LpR back issues - Subscribe HERE