90 lm/W at 72 Watts. I'm impressed!

Directly from the lab... but it points the way at things to come : http://www.ledsmagazine.com/news/6/3/25

Even more interesting is that they didn't use top of the line blue emitters to obtain their results. They used blues which emit about 260 mW at 350 mA. Using the best blue dice (~571 mW @ 350 mA), the result would have been as high as 197 lm/W for the neutral white and 170 lm/W for the warm white.

And these are NIST certified results at a thermal equilibrium of 37C! The only catch is... they didn't tell us how many LED cores were used to produce these numbers, nor the current seen by each core. Still, this looks promising... imagine, almost 200 lum/watt for neutral white!

Don't they say that it was tested at 350 mA.

Calina said:

Don't they say that it was tested at 350 mA.

Click to expand...

Yep but now how much current each core saw if it was multi-core.

We're talking about a NIST report here. Those guys don't have a reputation to play games with numbers. I'm quite confident the test was done at 350mA per die.

The number of dies required to reach the tests wattage is already pretty high at 350mA.

IIRC the die's where never the limiting reactant in the mix. Phosphor coatings are only so efficient. To keep a small footprint the coating must be improved, I imagine though that you could use higher output dies with a larger area of phosphor and get a high output flood.

65535 said:

IIRC the die's where never the limiting reactant in the mix. Phosphor coatings are only so efficient. To keep a small footprint the coating must be improved, I imagine though that you could use higher output dies with a larger area of phosphor and get a high output flood.

Click to expand...

I doubt this is the case. The behavior of the dice on commercial LEDs like the XR-E suggests that the phosphor is pretty much the same efficiency except for currents so high that the die is about to fry. So proximity is not a problem/limiting factor.

For an XR-E run pretty much anywhere between 10mA and 1000mA, there is negligible change in color. Color shifting at extremely low currents happens because the voltage across the LED is insufficient to generate light at the LED's designed wavelength, nothing to do with the phosphor. (This is actually more noticeable with no phosphor, in fact)

If the phosphor were becoming less efficient, one would expect a dramatic shift toward blue as the current through the die ramps up. (this does happen if you crank current up toward ~3A...) The fact that the color stays "white" means the ratio of die light to phosphor light is staying almost exactly the same (our eyes are extremely sensitive to small changes in hue, which would happen for even a minor change in phosphor efficiency)

But that said, I believe that using large-area phosphors <is> a good idea, especially for light fixtuers that are meant to be permanent. The phosphor will tend to degrade faster than the LEDs, and it's a lot easier to replace a plastic panel with phosphor on it, than a permanently installed fixture. A fixture could have blue LEDs built in, then use that to backlight a panel of phosphor (I'm picturing something akin to fluorescent ceiling panels -- but geometry could be made a lot more creative -- and would be easier to make flush-mountable). Different phosphor panels could be used by end users to select different color temp -- rather than having to buy neutral- or warm-white emitters and hard-wire them in. That should simplify mass production as only one version of the hardware will only need to be made.

For that matter, Signage could be made a lot more efficient. If you want, say, a backlit color logo, rather than filtering white light away like most signs, why not make a logo out of various colors of fluorescent dye and backlight it with blue light (blue could just be a clear frosting w/ no phosphor). Something like that could save a lot more than the mere "lm/W" numbers suggest, as instead of filtering away almost all the white liht to backliht a color sign, you're ONLY generating the colors you actaully want, for the most part (some filtering may still be needed of course, but not nearly as much).