If the emitted optical power isnít considered, the resulting thermal resistance will be smaller than reality, misleading the designers of the cooling solution. Measurement of the light output properties in combination with thermal tests as suggested, for example, by the JEDEC JESD 51-52 standard, provides useful information about their temperature dependence (Figure 2). Figure 3 shows how the chromaticity of the measured LED changes with forward current and temperature.
The data processing software of the T3Ster system derives structure functions from the transient measurement, which are then converted into models that are accurate in the case of packages that possess one-dimensional heat-flow path such as power LED packages. Such models can be created as ďside productsĒ when the RthJC junction-to-case thermal resistance of the package under test is identified according to the latest JEDEC transient measurement standard (JESD 51-14), based on the so-called transient dual interface method.
Figure 3. Example of a chromaticity diagram for an LED generated by the TeraLED measurement system.
Figure 4. Mutual dependence of LEDs major operating characteristics.
New model for simulating LED thermal resistance and capacitance
The Mentor Graphics FloEFD software includes an LED model that provides the so-called JEDEC 2R thermal resistor model in an extended format. For the 2R model, the necessary information can be found easily in the datasheets. In case of LEDs, the junction-to-bottom resistor of the 2R model is relevant; itís more or less equal to the packageís RthJC junction-to-case thermal resistance.
For the junction-to-top resistance of the 2R model, the junction-to-lens thermal resistance would be needed. This isnít typically provided and it hardly can be correctly tested; usually a sufficiently large value obtained from CFD simulations is provided. The way the standard 2R model is extended in FloEFD is that the junction-to-bottom part of the model is represented as an RC model, which allows for transient simulations because the thermal capacitance is also included in the model. However, the more detailed RC model requires more data from the LED, which often canít be found in datasheets. In this case, FloEFD makes it easier with an interface to the thermal characterization system.
A file can be exported out of T3Ster Master thermal transient data post-processing software that can be read by FloEFD with all the necessary data for the RC model in the form of a Cauer-type ladder model. This file contains the single thermal resistance and capacitance from junction to bottom (Rjb and Cjb) values of the LEDs as a bulk value representing a single thermal time constant for the package. In addition, it represents the heat-flow path structure with details that are appropriate for accurate transient simulation using CFD analysis.
To properly predict the LEDís hot lumens (luminous flux at operating junction temperature), an LED model in FloEFD also contains simple models for the radiant flux and luminous flux for the LEDís constant drive currents. These models use the measured junction temperature sensitivity of these light output properties. Including the temperature sensitivity of these parameters is important to account for the complex, multi-domain operation of LEDs, as depicted in Figure 4.
Measured results from a validation experiment to test this approach made it clear that just relying on datasheet values is insufficient when trying to determine a systemís thermal performance. Other methods provided information that showed a thermal issue with the LED packages under test; however, only the JESD 51-51 compliant transient test method spotted the location of the problem. Besides this failure analysis feature, we could derive an appropriate compact LED model which, when inserted into a simulation tool, can provide better predictions of LED operation.
With the new LED model, designers can obtain more accurate results using real component thermal properties. They also can obtain valuable post-processing data, which in turn could be used in optical simulation tools, for example, to further improve the results because the actual hot lumen is usually not known. In general, this process illustrates that the simulation will provide better results because the component properties used in their application rely on real measured component data, allowing for better judgments on thermal design.