Thermal Characterization and LED Lamp Design

LED lighting poses a unique design challenge because heat buildup can reduce an LED's light output and cause a color shift while shortening the component's useful life. A luminaire's mechanical design can be evaluated with virtual models and CFD tools, but thermal resistance measurements on the LED and luminaire are an essential part of the overall design process.

Thermal resistance from the LED's P-N junction to the ambient must be characterized to ensure a safe, reliable design and satisfactory performance. Often there are multiple thermal interfaces such as glue layers in the heat flow path, and their thickness and resistance can be difficult to control in manufacturing.

Like most other products, lighting systems are designed to balance cost and performance. The power budget — and therefore the need for cooling — is largely determined by the LED's radiometric output expressed in lumens per watt of input energy. A thermal management solution that provides effective cooling delivers more useful lumens of a more consistent color in a given application.

As explained earlier, heat dispersion begins at the package level, which is the province of the LED vendor.

Although standardization efforts are underway, LED vendors may still define their products' thermal resistance and other temperature-related characteristics in differing ways. If, for example, they do not include optical power as a conditioning variable when determining an LED's thermal resistance, the resulting resistance values will be much lower than “the actuals.” If the real thermal resistance is higher, the junction temperatures also will be higher and the luminous flux from the luminaire will be insufficient. Knowing the real thermal data of the LED is essential.

Measurements: Heat Before Light

It is a given that temperature in an LED is proportional to the forward voltage drop across the device. Therefore, by observing the converse (changes in voltage) it is possible to accurately infer temperature changes. To perform such tests it is necessary to leave the software domain of modeling and simulation and instead deploy hardware systems designed for the purpose. Commercially produced automated systems such as the MicRed T3Ster (“Trister”) from Mentor Graphics exemplify the toolset required for the job.

The third illustration in the slideshow below depicts the measurement setup itself in simplified form, not to scale. The measurement begins with the determination of the forward voltage drop at a specified current. Next, a large current is applied to the LED, causing it to heat up. Then the large current is turned off and a much smaller current is forced into the device. Lastly, the forward voltage measurement is performed before the junction has time to cool down. Due to the LED's fast thermal response, the measurement hardware must be able to capture the temperature (voltage) change within a few microseconds of the device being powered off. As the illustration shows, the measured device resides within an enclosure — a JEDEC standard still-air chamber that provides a controlled environment.

The complete Thermal Transient Measurement Process (shown in the fourth illustration in the slideshow below) consists of many repetitions of “Step 5,” yielding a plot known as the “Zth curve” which displays temperature change vs. time. As in electronics, the term “Z” represents impedance — in this case thermal impedance. In the Zth curve the plotted values express the temperature difference divided by the heating power. Thus, the Zth curve (illustration five in the slideshow below) depicts the temperature change for a heating power of 1W.

Standing alone, the Zth curve is rather smooth and lacking in the detail engineers need to interpret what is going on within the device. But it is made up of a very large number of closely spaced data points, so the underlying information is rich. Advanced mathematical applications bundled within a full-featured thermal measurement system can produce very useful analytical transforms of the Zth vs. time curve.

Data within the Zth curve plot makes it possible to calculate a cumulative plot of thermal capacitance vs. resistance, known as the cumulative structure function. This is the graphical representation of the network model of the thermal impedance of the junction-to-ambient heat-flow path. The shape of a structure function maintains a consistent one-to-one relationship with the properties of the actual junction-to-ambient heat flow path. The device junction is always at the plot's origin. The sixth illustration in the slideshow below illustrates the concept.

In an LED device the heat generated by the semiconductor travels outward from its origin. The junction heats up, and then the heat passes through a number of thermal resistances, heating up objects along the way. The further the heat travels, the more resistance it passes through and the more thermal capacitance (thermal mass) it heats up.

In the Structure Function graph, the sixth illustration in the slideshow below, the junction heats initially and the plot climbs steeply as subsequent capacitances are heated. Here the line is annotated to point out the transitions between the LED/MCPCB, the mounting medium (thermal grease in this case), and the luminaire. But within these boundaries the steps in the line represent smaller transitions through the die attach, the heat slug and even the glue used to secure the copper heat slug to the MCPCB. Note that the graph validates the earlier contention that the LED itself represents about 50 percent of the junction-to-ambient thermal resistance in the system.

Looking back at the measurement setup illustration, note that the voltage measurements span only the LED device. How does the system generate thermal data that spans the entire luminaire? The answer lies in monitoring and observing the cool-down cycle.

When the LED die initially starts to cool down, the decrease in temperature is slowed by the only thing connected directly to it in the thermal sense: the thermal resistance immediately beyond the die. The amount of time the die takes to cool down depends in turn on the thermal capacitance that is storing heat on the opposite side of that resistance. The measurement system sees this discharge as a perceptible increment and successive resistance/capacitance nodes, exhibiting similar behavior, are similarly observed. Each step further from the LED die increases the sensitivity required from the measurement system.

Models Spin off From Measurements

Structure functions help designers evaluate the discrete sections of the complete thermal path. Importantly, they can reveal design problems that might affect the device's manufacturability or reliability.

The cumulative structure function can be further refined into a compact model, that is, an equivalent network of purely resistive and purely capacitive components that embodies all the values and transitions expressed in the structure function plot. The last schematic in the slideshow below depicts a generic model. In an actual model the R and C components would have specific values.

R/C network models developed with the aid of transient thermal measurements are directly usable in thermal design tools, where they simulate the behavior of the LED system. Answering the market's demand for more data about their products, some semiconductor vendors have begun to use transient models to report the thermal performance of their LED devices.

Conclusion

The thermal resistance characterization procedure has confirmed that the LED can perform safely in its intended environment. But does it provide the intended amount of light? Photometric measurements made concurrently with the thermal measurements just described can validate this aspect of the design. Additional fixturing and instrumentation is required for this task.