LED, semiconductor, and package designers use complex thermal analysis software to analyze their products, but the analysis is only as good as the data provided. Reliable clarification of heat data upstream and downstream is critical. The process of obtaining this clarification traditionally has been cumbersome, requiring manual intervention. But now, for the first time, an automated process combines accurate thermal measurement with precise analysis, giving companies the ability to design better LEDs and power semiconductor device packages, create accurate simulation models, and verify systems for proper heat management.
The Importance of Accurate Thermal Analysis
Unlike incandescent light bulbs, light emitting diodes (LEDs) are semiconductors and thus must be kept cool. As an LED’s temperature increases, the light output decreases, the light changes color, and the lifetime of the LED is reduced. Temperature adversely affects both the functional performance of the LED and its longevity, so thermal management is a predominant issue in solid state lighting (SSL) design.
A long-established division of responsibility in the electronics thermal supply chain is that vendors should provide information that characterizes their part independent of any application environment in which they are used, and customers are responsible for building products that provide an application environment in which the part can operate within specification.
When designing SSL, thermal performance is a key consideration for creating a commercially successful product. Thermal performance of an electronic part is usually reported as thermal resistance, which is an indication of how difficult it is for heat to flow out of the package. Thermal resistance is calculated as the temperature rise of the junction divided by the power heating that junction. These metrics must be accurate because they are the primary inputs to thermal models used for design.
Metrics can be used for early design calculations; however, their use becomes more problematic the more complex the product. Placed close together, LEDs interact, each heating its neighbors as well as itself. The LEDs’ thermal, optical, and electrical operations are interdependent. For example, the amount of electrical power that produces heat varies with temperature. Thermal metrics should be based on the actual heat dissipated in the LED rather than the supplied electrical power, which requires accurate measurement of the light output as a function of temperature and applied elec trical current. If only the supplied electrical power is used and the energy emitted in the form of light is omitted, the calculated thermal resistance is far too low. The more efficient the LED, the more the error increases.
Published metrics are part of the commercial agreement between vendors and customers. If parts don’t perform as advertised, the financial effects on profits can be severe, with vendors facing huge product recall and warranty costs, as well as negative publicity. On the other hand, publishing correct metric values and providing accurate thermal models enhances LED sales, increases prestige, and maintains customer loyalty.
A Methodology-Based Standard
Fortunately, JEDEC (Joint Electron Devices Engineering Council) established a new standard, JESD51-141, for junction-to-case thermal resistance measurement based on transient measurement techniques. This metric has been used for decades, but measurement procedures according to older standards were not accurate enough and did not allow a sufficiently high level of repeatability of the measurement results. The junction-to-case resistance is the most appropriate metric also for packaged LEDs because it characterizes the heat flow path from where the heat is generated at the PN-junction down to the bottom of the case—exactly how LED packages are designed to be cooled.
The new transient method uses a “dual interface” approach in which the part is measured against a cold plate with and without thermal grease. The junction-to-case resistance is determined by examining where the two measurements differ. Very high measurement repeatability is required because the thermal impedance curves for the two measurements must be identical to the point where the heat starts to leave the package and enter the thermal interface between the package and the cold plate. This ensures that the point where the curves deviate is clear.
The Mentor Graphics transient thermal tester T3Ster uses a “smart” implementation of the JEDEC static test method to provide accurate thermal transient testing. Combining T3Ster with TERALED, a CIE 127:2007–compliant total flux measurement system with temperature control, creates a comprehensive LED testing station that enables self-consistent thermal and radiometric/ photometric characterization of LEDs. The system is fully automated, allowing an LED to be characterized at approximately 50 operating points (forward current and temperature combinations), in an hour.
LED vendors often report thermal metrics at only one temperature – for example, a junction temperature of 25°C, which is far from the temperature at which LEDs normally operate. Their data is supplemented by diagrams showing the relative light output as a function of junction temperature, but no standardized method has been used for obtaining these curves.
The T3Ster post-processing software allows the temperature versus time curve obtained during the transient measurement of a single LED to be recast as “structure functions”2. These graphs help identify magnitude of the thermal capacitances and the partial thermal resistances of the most important structural elements along the junction-to-ambient heat flow path.
The graphs in Figure 1 show the magnitude of all the thermal resistances in the heat flow path from the die through the structure to the cold plate used in the test setup. Deviation of the cumulative structure functions obtained for the two measurements (with two different qualities of the thermal interface at the package’s case surface), forms part of the JESD51-14 methodology.
The structure function can be represented by a piecewise linear fit, separating the curve into a number of discrete thermal resistance and thermal capacitance steps. The resistance and capacitance values within the package provide a measurement-based thermal model that is computationally efficient and accurately captures the heat flow path. These dynamic compact thermal models (DCTMs) can capture the thermal response of the LED as a function of time. When generated with T3Ster and TERALED, the models are consistent with the measured light output properties.