Luminous efficacy improves with excess current??

[Oznog]

I've been wondering about a feature on the Luxeon Rebel spec sheet:
http://www.philipslumileds.com/pdfs/DS56.pdf

On Red, Red-Orange, and Amber they all have the lumens output more than double for a double in current (350mA doubled to 700mA).

Not by much, the Red-40 for example goes from 40 to 85. So 6% more, but the other types (white, etc) have a LOWER lumens/mA ratio when we go above the nominal current.

In fact, the spec listed is for a constant THERMAL PAD temp of 25C. The thermal resistance from die-to-pad means the extra watt of power raises the die temp by about 12C. That factor alone should have causes the device to reduce its output by 5%-10% or more.

I'm curious why this is. Why did it increase, not decrease? Do these 3 colors actually get more efficient at higher currents?

 

[jtr1962]

I noticed that also. It's a typo. The data sheet and my own experimental data suggests otherwise. When I tested the reds their output versus current seemed to be almost linear from a few mA up to about 500 mA. 600 mA was about 16% less than linear. Above that output didn't increase much as the thermal effects on the die dominated. 700 mA output was about 1.55 times 350 mA output. Output actually dropped past 700 mA, and this with very good heatsinking. That's why absolute maximum is speced at 700 mA. No point running these past that. There just isn't more output to be found.

 

[Oznog]

Huh, yeah a typo occurred to me too, but if so it's a typo in 3 types. And I have to wonder because that feature/typo is only present in the AlInGaP types, and all 3 AlInGaP types have it. None of the InGaN types have it. I suppose it could be a systematic error due to some specific difference in they way they handled AlInGaP measurements.

Are you confident you were able to maintain the same pad temp at the two different currents? Allowing the pad temp to rise at higher power is natural, which would reduce the lumen/mA efficiency as expected but wouldn't test the spec's assertion that lumen/mA efficiency goes up with higher currents when a constant pad temp is maintained.

 

[2xTrinity]

Warning: I'm not an expert on the phyiscs involved, so this is speculation:

It's quite possible that lumens/mA does go up with increasing current, though even so it's unlikely it keeps pace with the increase in voltage so that lumens/Watt goes up. That's true of InGaN at extremely low current, eg below about 30mA or so for a 1A max Cree, or below about 1mA for a 5mm LED, because I believe there is some fixed amount of "leakage current" that passes though no matter what, but doesn't effectively produce photons, until you hit some

 

[jtr1962]

Are you confident you were able to maintain the same pad temp at the two different currents? Allowing the pad temp to rise at higher power is natural, which would reduce the lumen/mA efficiency as expected but wouldn't test the spec's assertion that lumen/mA efficiency goes up with higher currents when a constant pad temp is maintained.

Yes, the thermal pad temperature was kept pretty close to ambient. The LED was thermal-epoxied to a 2.5" square microprocessor heatsink. The heatsink didn't exhibit a noticeable temperature rise during testing. Practically all of the temperature rise of the die was due to die to contact pad thermal impedance which is greater for some reason for the AlInGaP parts than for InGaN parts.

 

[evan9162]

I've been wondering about a feature on the Luxeon Rebel spec sheet:
http://www.philipslumileds.com/pdfs/DS56.pdf

On Red, Red-Orange, and Amber they all have the lumens output more than double for a double in current (350mA doubled to 700mA).

Not by much, the Red-40 for example goes from 40 to 85. So 6% more, but the other types (white, etc) have a LOWER lumens/mA ratio when we go above the nominal current.

In fact, the spec listed is for a constant THERMAL PAD temp of 25C. The thermal resistance from die-to-pad means the extra watt of power raises the die temp by about 12C. That factor alone should have causes the device to reduce its output by 5%-10% or more.

I'm curious why this is. Why did it increase, not decrease? Do these 3 colors actually get more efficient at higher currents?

Look at the column headers. The left is "Minimum". The right is "Typical". They aren't referring to the same thing.