Peter,
You keep the junction below 60C?
How do you do this? Here are my thoughts below.
Junction = the LED die.
Case = Slug on the emitter.
Or does the chassis of the ARC4 stays below 60C?
I highly doubt you keep the junction below 60C. Why I'd be led to believe this:
Level 1 is ~ 3.6V and 700mA for 2.52 W.
Thermal resistance of LED die (junction) to slug is ~15 C/W
15 C/W * 2.52 W = 37.8 C rise junction to slug.
Normal room temperature ARC4 housing is 25C.
Assumming a perfect transfer from slug to air (which it is far from this) you'd be at a junction temperature of 62.8 C right when you turn it on.
When running it in human hand, human hand is 37 C. If you had perfect thermal transfer from human hand (the ARC4 body was at human hand temperature and it had a perfect transfer to the slug, again far from it) 37.8 C + 37 C = 74.8 C junction (die) temperature.
What you probably meant to say is you hold the slug at 60 C or the ARC4 body at 60C?
If the actual slug was held at 60 C, at level 1, 700mA and 3.6V, you have 2.52W. 15 C/W * 2.52 = 37.8 C rise from slug to junction, so we take the slug temperature of 60C and add the rise due to the thermal resistance between the slug and junction, and we get 60C + 37.8C = 97.8C junction temperature.
See Application Note: AB05 located at
http://www.lumileds.com/pdfs/protected/AB05.PDF
*****
Also in your comment, "This projection is based on constant current 350 mA " (which would be a 1W emitter at 350 mA, not 2.52W (700mA for level one))
****
Luxeon Application Brief AB07,
http://www.lumileds.com/pdfs/protected/AB07.PDF
"Figure 4 shows the average lumen maintenance at 4000 hours for white Luxeon Power Light Sources driven at 0.35A at 25°C case temperature (TJ @ 45°C). Note that virtually nolumen depreciation occurred after 4000 hours. Figure 5 shows the average lumen maintenance at 2000 hours for white Luxeon Power Light Sources driven at 0.35A at 85°C case
temperature (TJ @ 105°C). Note that the lumen
depreciation is somewhat higher at elevated
temperatures."
To me, it appears that there is a 10% drop in light output after only 2000 hours, according to chart 5.
Does the ARC4 actually embed a temperature sensor into the luxeon case (slug)? Or is the sensor clear back on the board, and really measure the ARC4 housing temperature, which is lower than the slug temperature?