Re: What really causes (White) LED\'s to fail?
We've theorized that it must be related to the phosphor, since the blue/green 5W devices are rated at a 50,000 hour life by Lumileds. Everything else is the same between the colored and white 5W devices, thus the phosphor must be what's causing the reduced light.
We can see the relative junction temperature difference between the various families of Luxeons to determine how the phosphors are being treated:
(assuming the emitter mounted directly to an infinite 25C heatsink, 1C/W S-A thermal resistance)
1W:
Rj-s = 15C/W
Tj = 25 + (16 * (3.3 * 0.35) ) = 44C
3W (700mA) :
Rj-s = 13C/W
Tj = 25 + (14 * (3.5 * 0.7) ) = 60C
3W (1A):
Rj-s = 13C/W
Tj = 25 + (14 * (3.7 * 1.0) ) = 77C
5W:
Rj-s = 8C/W
Tj = 25 + (9 * (6.8 * 0.7) ) = 68C
What is interesting is that the 3W at 1A has a higher Tj than a 5W, yet a 3W at 1A is rated at 20K hours to 50% brightness.
However, we are just given the j-s thermal resistance. The phosphor must be at a higher temperature, since it is farther from the heat sink slug than the junction is. It could be that there is so much heat produced at the phosphor of a 5W (compared to any of the others) that the phosphor is being baked alive, even though the junction temperature is at reasonable levels.
If anyone is willing to sacrifice the above 3 LEDs, and have a very sensitive thermocouple probe, they could drill into the dome of the emitters, then measure the temperature at the top of the phosphor and report back what they found. It might be an interesting experiment.