Online proceedings - EDA Publishing Association

24-26 September 2008, Rome, Italyone manufacturer quotes reliability in terms of B50, L70 andanother in terms of B10, L50. (By the way, being a nonthermalissue, the question is in which committee this type ofstandardisation should be addressed.)Except for knowledge about the absolute temperature,designers need to have knowledge about temperaturedifferences. To make life more difficult for the designers,LEDs suffer also from current-dependent failuremechanisms, such as electromigration and Joule heatingcausing excessive local temperature rise in current-carryingtracks and wires. For further reading on LED reliabilityissues please consult references [1], [2], [3], [4], [5], [6], [7].In summary, most of the degradation and failuremechanism that rule the lifetime of a LED-based product aretemperature assisted failure mechanisms. Consequently, inorder to estimate lifetime, designers need again reliableinformation about the expected temperature profile overtime.III.SITUATION ANNO 2008: DRAWBACKS OF CURRENTDATASHEETSAbout the important things a designer of an LED-basedproduct wants to know are luminous flux (lumen) andefficacy (lumen/W), not only at zero-hour but also over itsexpected lifetime. The problem is that both key parametersare not linearly related to driving current and temperature.Hence, it is not sufficient to report data only at someoptimistic temperature. It is to be expected that when timepasses only the companies that employ experienceddesigners will survive. As a consequence, it is only a matterof time before these experienced users are going to demandfrom their manufacturers thermal data that they really canuse for prediction purposes. The question is: how bad are thecurrent data sheets? The answer is: pretty bad, from theperspective of the experienced designer (see also Grabner-Meyer [8]), for the following reasons (however, it should benoted that there is a tendency among the leading LEDsuppliers to improve upon this situation):• Data sheets do not reflect real-life operation,especially regarding operating temperature.• They differ strongly in content (T ref , I ref ).• Often a direct comparison is not possible.• Often T junction specified at 25°C misleading,efficacy at max rated power often 50% lower (onthe positive side, leading manufacturers havestarted to quote more realistic values).• While non-thermal but temperature-related:Translation to useful lumens or non-nominal use isnot ‘idiot’ proof, and even for an experienceddesigner not an easy task.• The way of quoting thermal data by manufacturersis the series thermal resistance approach that isquestionable in a number of practical cases forwhich more complex thermal networks arepreferred.A. Examples of problems with current datasheetsThe definition of powerMaybe the biggest problem nowadays hampering a faircomparison is the lack of a standardised way of definingpower. Many manufacturers define their power dissipationby the product of voltage and current, not taking into accountthe efficiency of the conversion from current into light.Let us compare two simple cases to address theconsequences, and let us define R th total as the thermalresistance based on the total power dissipation and R th thermalas the thermal resistance based on the total dissipation minusthe light output.Case 1: efficiency 50%, required light output 1WQ total = 2W, suppose ΔT to be 100 K.Then we have for R th total : 100/2 = 50 K/WFor R th thermal we find: 100/1=100 K/WCase 2: efficiency 80%, same light outputQ total = 1.25W, ΔT = 25K. R th total = 25/1.25 = 20 K/WR th thermal = 25/0.25 = 100 K/W.In our view, the thermal resistance should be related to thephysical properties and the dimensions, and should beindependent of the efficiency, as is clear from the exampleabove. The problem in practice is that the user should knowthe efficiency. Of course the LED community could decidefor a metric based on the total power input, resulting in thesame LEDs differing only in efficiency to get different R th ’s.However, to prevent confusion, we should not call this anR th , but in analogy with IC thermal standardisation maybe aΨ. The advantage is that the user does not need to know theefficiency, but could derive it more or less from comparingtwo quoted values, with the emphasis on ‘more or less’. Onthe other hand, we are of the opinion that a higher efficiencyis an important sales argument with the increasing interest insustainability, and hence it makes sense to decouple thisparameter from the physical R th .As shown in Fig. 5, the overall efficiency (also known aswall plug efficiency or WPE in short, which is P opt /P el )depends on temperature and current. (This of course alsoresults in the current and temperature dependence of theefficacy which is nothing else than the WPE weighted withthe V(λ) visibility function defined by the CIE). That is whyany metric of an LED reported in a data sheet should also bereported together with the current and temperature at whichthe given metric was identified. The value of the forwardcurrent to report is clear but the definition of the temperaturevalue to report needs careful discussion. In an ideal case, this©EDA Publishing/THERMINIC 2008 215ISBN: 978-2-35500-008-9