Development of Short Time Period Evaluation Technique for Determining Quality of LED Package
The degradation mechanism of the LED package is still unclear, although several degradation models have been proposed. Many authors have pointed out that heat is the major factor for the degradation of the LED package. The general evaluation technique for determining the quality and reliability of LED packages measures the light intensity during continuous emissions under the rating current condition. Because of the long life and complex behavior of LED light emissions, over 10 k hours is needed to test them to make sure they are sufficiently capable. In Figure 2, the difference between LED A and LED B can be distinguished after 8 k hours. However, it is difficult to ensure their quality and reliability within a short time period.
We propose a new evaluation technique that can distinguish the quality and reliability of LED packages within a short time frame. This technique has the ability to measure the change in function of LED packages taken from different noises. In this technique, the LED packages are exposed to five different types of complex stresses for a maximum of 500 hrs. After that, the relation between the current and light intensity of these LED packages is measured. The degradation level sensitivity of a LED can be improved by measuring the change in function (the relation between the current and light intensity) of a LED. The given noises are listed as follows.
(1) Light stress: deteriorating the resin and package sealing the LED device.
(2) Heat stress: deteriorating the resin, package, and LED device.
(3) Moisture stress: deteriorating the resin, package, and Ag mirror film.
(4) Heat cycle stress: deteriorating the resin and package.
(5) Corroding the gas stress: changes color of Ag mirror film.
Figure 3 shows the relations between the current and light intensity of LEDs A and B, respectively. The relations are measured initially and after being exposed to the five complex stresses for three LED packages. These figures show that by measuring their relations, we can judge the level of quality of each LED package within a short time period (within 1000 hrs). The level of quality of LEDpackages is quantified by using the S/N ratio proposed by G. Taguchi in the field of Quality Engineering. The S/N ratio is defined using the following formula .
S/N ratio (db)=10 log (β2/σ2) 
In this formula, β is the regression coefficient of the relationship between the current and intensity, and σ2 is the regression error variance. An increase in the S/N ratio means that the variability of the LED package is decreased and/or the emitting efficiency of the LED package is improved. Then, a LED package with a large S/N ratio is determined to be more stable and/or has a higher level of efficiency than one with a smaller one. Table 1 itemizes the S/N ratios of LEDs A, B, and C. In the table, the quality of LED A is higher than that of LEDs B and C. In addition, this LED is the lowest one. This result corresponds with the results obtained by using the general evaluation technique over a long time period, as shown in Figure 2.