Online proceedings - EDA Publishing Association

24-26 September 2008, Rome, Italymust be the junction temperature if there could beunambiguous ways to identify the junction temperature ofthe LED in practice.board/heat sink and finally the ambient.• The resistances are defined locally, in other words,the thermal resistance from die to heat slug is onlydependent on local parameters.• Consequently, the individual resistances areindependent of each other. For example, thethermal resistance from die to heat slug is notdependent on the board thermal conductivity, noron the heat transfer coefficient 1 .In order to check the validity of these assumptions, aLuxeon Rebel as depicted in Fig. 7 was modelled on a boardusing Flotherm in conduction-only mode.Fig. 5: Current and temperature dependence of theWPE of a red Dragon LED (measured by MicReD)The series thermal resistance approachAs an example of the problems that are associated with aseries resistance approach let us have a look at the datasheets of one of the major LED manufacturers.In Fig. 6 we present a sketch taken from a typicalApplication Brief.Fig. 7: Top: sketch of Luxeon Rebel,Bottom: half symmetry model of a Rebel on a PCBFig. 6: From a Lumileds Application BriefThe assumptions underlying the series resistance approachare the following:• The power generated in the die is converted to heatand follows the path sketched via the thermal pad,The procedure to check was as follows. Variousparameters such as board thermal conductivity and heattransfer coefficient were varied over a wide but still1 This may sound trivial, but it is possible to define a seriesresistance network of which the resistors are dependent on eachother. This principle is underlying the well-known heatspreading approach proposed by e.g. Lee [15].©EDA Publishing/THERMINIC 2008 216ISBN: 978-2-35500-008-9