The document discusses various topics related to LED performance metrics and reliability testing, including: 1) Hot binning and how it does not eliminate binning but creates a new framework. 2) Hot/cold factor is a real issue but should not be the only parameter focused on. 3) LM-80 testing provides lumen maintenance data over time but does not indicate lifetime. TM-21 allows projecting lifetime from LM-80 data. 4) LED lifetime is irrelevant - system lifetime that provides value depends on factors like heat sinks and drivers. Proper design leads to long lifetimes. 5) Chemical compatibility is important - VOCs can contaminate LEDs over time, especially with heat and light

1. “Hot Binning”, Droop, TM-21 and other ‘Fun’ LED Topics June 14, 2011

4. Supplier B: 7%

5. Supplier C: 11.5%“H/C factor” is real and could be an important consideration at very high temperatures; it’s within measurement uncertainties in normal operating ranges and there are other critical issues to keep in mind… (there’s more…)

7. Color Performance and Leveraging Binning to Save Money LED CCx, CCy Yield to Bin Start with the right target, expectation – what traditional lighting technology are you trying to match? “Perfect” color performance is achievable with LED – sometimes at a price…

10. “Holes” are great for making demo fixtures for the exhibition floor at a trade show. Try ordering a million “holes” and watch the marketing guys run for cover (and/or you for your wallet)…New!

11. Some Smart Ways To Specify LEDs Buy only the distribution you need (e.g., CFL quality) Specify multi-chip LED arrays that do the color mixing for you – inside the lamp Make your own “hole”: Buy the full distribution to get the best supply/ lowest cost, but do the mixing yourself – in the fixture

12. Tools for Doing It Works well for downlights, most bulb types, most diffused light applications More challenging for linear, wall-washers, etc. http://www.cree.com/binonator/

13. 50,000 hours is: 137 Years at 1 hour/day 68.5 Years at 2 hours/day 34.2 Years at 4 hours/day 22.8 Years at 6 hours/day 17.1 Years at 8 hours/day 11.4 Years at 12 hours/day 5.7 Years at 24 hours/day …A WAG when it comes to LED lifetime…

14. Semiconductor Reliability Testing Reliability test methods and acceptance criteria for semiconductor components have been standardized (JEDEC, EIAJ, others…) and practiced for decades Think: processors, regulators, microcontrollers, etc.. If you’ve ever flown in an airplane, driven in a car, or talked on a cell phone, you’ve depended on this body of scientific work and testing…

15. LED Reliability Testing LEDs are semiconductor components that happen to emit light… Most LED manufacturers conduct standardized semiconductor component reliability testing – the same tests Intel tests their microprocessors with – on their LED lamps The Illumination Engineering Society of North America published IES LM-80 in 2008 to characterize the Lumen Maintenance aspect of LED semiconductor components Note: Lumen Maintenance ≠ LED Lifetime

16. LEDs Last Forever!! [under ideal conditions] Well-designed systems with Lighting-class LEDs at low TA, TJ will run a very, very long time…

17. Typical LM-80 Lumen Maintenance Behavior 100% TSP = TA = 85˚C; IF = 1500mA 94.1% 90% Lumen Maintenance (%) 80% 70% 2,000 1,000 3,000 5,000 4,000 6,000 Time (hours) LEDs do not normally fail catastrophically; gradually lose light output over very long time periods Small “hump” is frequently observed between 0 and 500 hours Lower drive currents and lower temperatures yield higher Lumen Maintenance curves

18. Everyone Asks for an “LM-80 Report” Here is what one looks like (too detailed, no interpretation, just data…):

20. Projection limited to 6x the available LM-80 data set

21. Projection algorithm: least squares fit to the data set

22. L70, L80, L90, Lp projections easily possible

23. Nomenclature: Lp(Yk) where p is the Lumen Maintenance percentage and Y is the length of the LM-80 data set in thousands of hoursExample: L90(12k)

24. Typical LM-80 Test Behavior and TM-21 Lumen Maintenance Projection (6k) 100% 94.1% 6 x 6,000 = 36,000 hours (max) 90% Lumen Maintenance (%) 80% Projected L70(6k) = 35,000 hours Reported L70(6k) = 35,000 hours 70% 10,000 20,000 30,000 40,000 50,000 Time (hours) First 1k hours is ignored for TM-21 projection purposes Upper reporting bound set by 6x available data (6 x 6k = 36k hrs) Exponential extrapolation to least squares mathematical fit between 1k and 6k hours Reported and projected L70 may or may not be the same number

26. Upper reporting bound set by 6x data (6 x 10k = 60k hrs)

27. Exponential extrapolation to least squares mathematical fit between 5k and 10k hours

29. Upper reporting bound set by 6x data (6 x 20k = 120k hours)

30. Exponential extrapolation to least squares mathematical fit between 10k and 20k hours

33. Adhesive outgases VOCs, concentrates under optic

34. VOC penetrates silicone lens

35. Heat, photonic energy turn the VOC brown

37. Chemical Compatibility Experiment Vented Optic@ 336 hrs Standard LED (Control) SealedSecondary Optic C

38. Happens On All Types of LEDs, All Suppliers Multi-chip LED (sealed) Applications Notes… Multi-chip LED (vented) Cree Philips Lumileds

39. Chemical Incompatibility – Prevention Cree

40. Chemical Incompatibility – Prevention, cont. Philips Lumileds

42. Simple boards

43. Instructions

46. Small chips sometimes appear to have higher efficacy since they are customarily binned higher up the droop curve

48. LED Capacity Example LED Capacity Light Output, Efficacy LPW efficacy Un-utilized capacity Light Output = $$$ Max Drive Current (1500mA) Binning Current (350mA) Input Current (If, mA) At the binning current only 23% of the capacity of this LED is utilized;77% of the LED goes UNUSED! There are, of course, practical limitations to this…

51. What About LED Cost? The Semiconductor Industry is a solution looking for a problem Articulating the problem in a way a chip company can understand is often the challenge A problem – once understood – gets solved in a very predictable way…

52. Moore’s Law for Transistor Cost -36% CAGR

53. Driving Lumen Affordability with Technology Working on both numerator and denominator!! 93% improvement in 5 years Cool White (6000K) Normalized $/klm * At maximum drive current

54. How Would You Like That Lumen…? U.S. DOE Multi-year R&D Program, March 2010, p.28 Cool? Warm? High CRI? Efficient? Long life? Stable Color? Uniform? Optically-controlled? At what drive current? Energy Star? Simplistic comparisons like this do not work – everything matters with LED…

55. By That Metric, LEDs are Already Cheap! Raw $/klm Cost of One Commercially-available LED XLamp XM-L $/klm $/klm @ Binning Current DOE Values $/klm @ Max Drive Current  LED Drive Current  Raw LED cost is already close to parity with most incumbent technologies – if we must be simple about it, but… How would you like that lumen…? Raw LED cost is only part of the story (driver, optic, etc.) Must factor in application efficacy, energy savings, maintenance avoidance, environmental impact, etc., etc.to get the real answer on LED VALUE (different than cost!)

57. 650 lm

59. 575 lm

60. 10.5WCree LR6 Cree CR6 >$100 Commercial Wholesale $50 Retail