> >> and the cost? > > > > I don't think they mentioned cost :-). > > Though you seem to think that your solution wouldn't be cost effective > compared to theirs (I'd have suggested it otherwise!) I know that my LEDs are not cost competitive Watt per Watt with XRE's or any modern high Wattage white LED as mine are very low wattage and cost does not, alas, scale down linearly with Wattage. So I'd certainly hope that anything new was more cost effective than using 3+ year old Nichia 100 m,W LEDs. I' d need 100 x NSPWR70CSS-K1 at say around 50 cents each or $50 of LED's. An XR-E in true production volumes is probably about $1.50 so you'd hope a new 10W source of 1 to 3 LEDs will be say $5 and falling when it arrives as a real product. > >> =A0Personally I'm impressed we're still seeing significant advances ..= .. > > That was only a quantum leap wrt prior offerings they had. My 3+ year .= ... > It was a quantum leap wrt any comparable high power LED made by > anybody. How do your Nichias compare to a derated XR-E (ISTR those > become far more efficient when run at very low power)? Short: Surprisingly well given their age. Lonnnnng: There are two data-sheet factors that govern l/W efficiency. These are not primary factors in their own right - just visible indicators of the deep underlying magic. - One is a relatively slight kink in the close to straight line lumen per mA curve. This is close to the form of y =3D kX with a very slight kink towards Y =3D k.X^(1-1/N) where N is largish (flu on Y axis, current on axis). For most purposes you read this as "if I double he current I don't quite get double the lumens" but it works backwards when candlepower is the aim. - The greater factor is the forward voltage with current curve which is a classic exponential curve. As you increase current through say 10% to 100% of rate value you get modest increases in Vf - say from 2.9V to 3.3V, corresponding at constant lumen per mA to a loss of efficiency by a factor of 2.9/3.3 =3D~ 12% in this example. Of interest is the shape of the curve and the absolute vaklue of Vf at rated current. With improving LED technology VF operating is dropping. Most LEDs have a substantial spread of production Vf values leading to a wide binning range of l/W values. My Nichia Raijins cluster immensely impressively around Vf =3D 2.95 V at whatever reference current they used. (I have a Nichia report based on large volume sampling from production. Results slowly get better with time). Few modern white LEDs are getting down to that Vf to start, so far. Manufacturers don't usually expect you to run an LED at 5% or 10% of it's rated Wattage. eg an about 5 Watt capable XR-E at 500 mW or less. So hard data is not usually offered so you need to either ask (and may be told) or blow up datasheet graphs immensely and try to plot data from them in a manner not probably intended. At lower levels these are liable to not be intended as definitive and are also only typical values. BUT I have in fact done this for the XR-E a while ago and it looks good but not stunning wrt the Raijin. Raijin gives 160+ l/W at 30 mA. It was originally a 30 m rated part but Nichia appear to have decided that it was so good that they rated it at 50 mA (it's in a P4 7mm 4-lead through hole package) and provided decreased lifetime figure to match. From memory it's 14,000 hours to 70% at 50 mA. (Actual measured as noted is +1% to +2% increase in output at 2000 hours at about 45 mA :-). XR-E gives > 160 l/W at say 10% rated power but not so vastly better than Raijin as to make your eyes water. (Do not stare into beam with ....). For some reason, 3+ years ago Nichia made an utterly stunning small LED, packaged it strangely (P4 only, no SMD), did not match it with an SMD equivalent, did not come close to matching it with a higher power version, and then they and everyone else waited several years to finally catch it up. They have now done so, but not stunningly so. There will shortlyish be new products around which compete in a similar application area but so far at the lower power levels it's still the best. Russell --=20 http://www.piclist.com PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist .