Doug S
Flashlight Enthusiast
Forgive me if this horse has been flogged in another thread somewhere.
I have noticed that some here haven't noticed that Nichia and Lumileds use very different test conditions for the "typical" output figures provided on their datasheets. Nichia test data is for an ambient temperature of 25C while Lumileds is for a junction temperature of 25C. While not specificed on their datasheet, per a phone call to a Nichia rep years ago, the Nichia test condition is mounted an a PCB with a rather minimal copper footprint. The Nichia test condition is therefore much closer to the "real world" conditions under which the LED might be operated. Those here that expect to get the "typical" outputs out of the Lumileds are mistaken with the possible exception of those who are using their lights under arctic conditions. It may be underappreciated by some just how hot the junctions will be running without very careful heatsinking. I performed a test with a /C version of the "1W" device [25mm X25mm metalcore PCB with no optics]. I drilled a tiny hole in the aluminum heatsink just under the LED and inserted a low mass thermocouple. With the heatsink mounted vertically in free air to maximize convective cooling, I measured a junction to ambient thermal resistance of 60C/watt. Now consider the case of running a "1W" device at 350mA. Using a typical Vf of 3.42V this is a power of 1.2W for a temperature rise over ambient of 72C. For an ambient of 25C this yields a junction temp of 97C. Consulting the datasheet for output vs temp one sees that the output is reduced to only 75% of the 25C Tj value. In many cases the reality may be worse than above. For a device enclosed within a housing the junction to ambient thermal resistance will likely be greater than the 60C/W that I measured. It may also be presumed that the devices with optics will have somewhat higher thermal resistances as the holder for the collimator will obstruct airflow on most of one side of the metalcore PCB. A small sampling of the posts on this board reveals that there are many who like to drive LEDs to the limits and beyond. It is therefore probable that junction temps of 120C [the rated absolute maximum] and beyond are being obtained. At 120C the output is reduced to 69% of the 25C Tj typical. Now consider the issue of permanent degradation. For the white devices, the output degradation vs time graph shows that after 1000 hours of operation the output is reduced by 12%. This does not look too bad until you. consider that the data is for a Tj of 70C. In the absense of specific knowledge of the degradation mechanisms, a widely accepted "rule of thumb" is that thermally influenced degradation rates double for each 10C rise in temperature. Using this, it only takes 31 hours to reach the [permanant] 12% degradation of output at a Tj of 120C.
I have noticed that some here haven't noticed that Nichia and Lumileds use very different test conditions for the "typical" output figures provided on their datasheets. Nichia test data is for an ambient temperature of 25C while Lumileds is for a junction temperature of 25C. While not specificed on their datasheet, per a phone call to a Nichia rep years ago, the Nichia test condition is mounted an a PCB with a rather minimal copper footprint. The Nichia test condition is therefore much closer to the "real world" conditions under which the LED might be operated. Those here that expect to get the "typical" outputs out of the Lumileds are mistaken with the possible exception of those who are using their lights under arctic conditions. It may be underappreciated by some just how hot the junctions will be running without very careful heatsinking. I performed a test with a /C version of the "1W" device [25mm X25mm metalcore PCB with no optics]. I drilled a tiny hole in the aluminum heatsink just under the LED and inserted a low mass thermocouple. With the heatsink mounted vertically in free air to maximize convective cooling, I measured a junction to ambient thermal resistance of 60C/watt. Now consider the case of running a "1W" device at 350mA. Using a typical Vf of 3.42V this is a power of 1.2W for a temperature rise over ambient of 72C. For an ambient of 25C this yields a junction temp of 97C. Consulting the datasheet for output vs temp one sees that the output is reduced to only 75% of the 25C Tj value. In many cases the reality may be worse than above. For a device enclosed within a housing the junction to ambient thermal resistance will likely be greater than the 60C/W that I measured. It may also be presumed that the devices with optics will have somewhat higher thermal resistances as the holder for the collimator will obstruct airflow on most of one side of the metalcore PCB. A small sampling of the posts on this board reveals that there are many who like to drive LEDs to the limits and beyond. It is therefore probable that junction temps of 120C [the rated absolute maximum] and beyond are being obtained. At 120C the output is reduced to 69% of the 25C Tj typical. Now consider the issue of permanent degradation. For the white devices, the output degradation vs time graph shows that after 1000 hours of operation the output is reduced by 12%. This does not look too bad until you. consider that the data is for a Tj of 70C. In the absense of specific knowledge of the degradation mechanisms, a widely accepted "rule of thumb" is that thermally influenced degradation rates double for each 10C rise in temperature. Using this, it only takes 31 hours to reach the [permanant] 12% degradation of output at a Tj of 120C.
