DOE OLED Planning Meeting Trovato Mfg, Rochester, NY, Oct 1, 2013 Common Manufacturing Platforms and Testing Mike Lu, Director OLED Lighting Design Center Acuity Brands Lighting, Inc. OLED Lighting Design Center, Berkeley, CA
Overview OLED standards in progress Notable items OLED product standardization Acuity product specification Color Quality Life Test Reliability and Early Failures
OLED Standards UL UL 1598 Standard for Luminaires Governs all luminaires UL, ETL, CSA all test to UL standards and generally accepted by customers UL 8752 Standard for Safety Organic Light Emitting Diode (OLED) Panels Concerned with issues of safety: sharpness of edges and corners, security of wiring connections, flammability of materials, over current/voltage etc. At the component level, UL listing is necessary Michael S. Shulman is the engineer responsible for drafting this standard. In the field, we have referred vendors to Walter Das for obtaining UL listing. I m on the UL technical committee and gets to vote/give input on proposed changes.
OLED Standards IEC I m serving as a subject matter expert on US TAG34 and is part of the US delegation in this working group. Working on two standards: safety and performance Safety standard is concerned with proper marking, flammability, etc. Deals with hot spots associated with shorting, intentionally provoking a short, then run at rated current for 30 min: C.1 Method for an OLED panel with glass substrates An internal short circuit shall be provoked by applying a small solder tip (preheated to minimum 430 C) to the glass surface or by laser treatment. The solder tip shall be applied as short as necessary and shall be removed immediately when a short has been created. The location of the internal short circuit provoked intentionally shall be close to the edge of light output area at around 2 mm distance. In case that the final product of an OLED panel has an external outcoupling system (e.g. outcoupling foil) or a plastic film for safety, slightly peel off the external layer just enough to expose the glass for provoking treatment of the internal short circuit and then reattach the peeled off layer to the glass after the treatment. C.2 Method for an OLED panel with flexible plastic substrates In the case of an OLED panel equipped with a flexible plastic substrate, the internal short circuit shall be provoked by applying vertical pressure on the surface of the panel with round tip of steel for about 1 second. The temperature of the tip shall be 25 C ± 5 C and the radius of curvature of the round tip shall be 0.5 ± 0.05 mm. The force applied on the tip is desirable to be about 100 N. Draft (IEC 62868) at the committee draft stage
OLED Standards IEC Performance standard is concerned with standardize measurement method and terminology still at the committee stage Example: luminance uniformity Chromaticity angular dependence Li L U 1 Max dij j L d av max
OLED Standards IESNA IES has a technical committee formulating OLED lighting standards Chaired by Jeremy Yon of Lite Control, I m a member. Recent discussion centered around panel orientation during measurement, camera vs. spot spectrophotometer, etc. Other standards work: CIE Notable recent paper Tokihisa Kawabata and Yoshi Ohno, Optical measurements of OLED panels for lighting applications, J. Modern Optics (2013).
Overview OLED standards in progress Notable items OLED product standardization Acuity product specification Color Quality Life Test Reliability and Early Failures
OLED Panel Standardization There are several dozen commercial white OLED panels on the market today. All with different form factors and current-voltage characteristics. LED standardization came at a later stage of development and only at the module level so far. It s too soon to talk about OLED panel standardization.
Acuity Brands OLED Lighting Panel Specification White OLED panels for general illumination Outer dimension: 100 x 100 mm, 50 x 200 mm, and 53 x 55 mm Emitting area: 90 x 90 mm, 40 x 190 mm, 46 x 46 mm, respectively Efficacy: minimum of 55 60 lm/w @ CRI, Ra 85 90, R9 20 50 (higher CRI is preferable) Efficacy goals expected to increase over time Luminance measured by 2 integrating sphere L70 > 15,000 hrs Soon to be 18,000 hrs and will increase over time CCT: 3000, 3500, and 4000K; Duv < 0.002 Luminance uniformity > 85% intra panel, within +/ 10% inter panel Chromaticity uniformity: +/ 2 MacAdam steps (tighter is better) Chromaticity angular dependence: u v < 0.004 Chromaticity change at end of life: u v < 0.004 Reliability: <1:1000 failure during 4 wk burn in, to increase to 1:10000.
Overview OLED standards in progress Notable items OLED product standardization Acuity product specification Color Quality Life Test Reliability and Early Failures
COLOR IS KEY!!! Color Quality 7 sdcm is WAY to wide!
What People Perceives as WHITE Black line is what people perceives as WHITE: green shift (above BB line) at CCT > 4000K and pink shift at CCT < 4000K It is very dangerous to make an OLED panel at CCT<3000K, color point at the top of the 7- step ANSI bin, with decent CRI and assume it will be accepted as high color quality. http://www.lrc.rpi.edu/programs/solidstate/assist/whitelight.asp
High Color Quality LEDs Soraa MR16 replacements GaN on GaN Vivid series @3000K Ra 95, R9 95 40.5-43 lm/w Source: Soraa
High Color Quality LEDs Source: Xicato Xicato Artist series modules Remote phosphor design with 2 (?) phosphors Very high CRI, round module efficacy at 47-59 lm/w, rectangular module efficacy at 50-64 lm/w
4 Emitter Panel Candlelight-style organic LEDs: a safe lighting source after dusk Jwo-Huei Jou and Chun-Yu Hsieh http://spie.org/x102843.xml?highlight=x2408&articleid=x102843
Overview OLED standards in progress Notable items OLED product standardization Acuity product specification Color Quality Life Test Reliability and Early Failures
OLED Life Time Models T/T 0 = (L 0 /L) ^ is a phenomenological constant between 1 and 2 Bi exponential model for luminance decay at set current T. Tsujimura et al. World s First All Phosphorescent OLED Product for Lighting Application, IDW 11
LED Life Time References IESNA LM80 08 Measuring Lumen Maintenance of LED Light Sources IES TM21 11 Projecting Long Term Lumen Maintenance of LED Light Sources Testing Protocol Air Temperature to within +/ 5 C, Case Temperature to within +/ 2 C RH less than 65% Minimum 6,000 hours total data, data collected (minimum) every 1,000 hours Data collection at 25 C Constant current, rated voltage (operate panel at rated temperature), prefer using the actual driver in the luminaire Record Lumen Maintenance, Chromaticity, Catastrophic Failures Number of LEDs to test: minimum 20 Use all data. Average all data. Life Time Extrapolation ln( B 0.7) ( t) B exp( t) L70 where: t = operating time in hours, Φ(t) = averaged normalized luminous flux output at time t, B = projected initial constant derived by the least squares curve fit; = decay rate constant derived by the least squares curve fit Data used for extrapolation: for 6000 hrs data, 1000 6000 hrs; for 10,000 hrs data, 5000 10,000 hrs; for >10,000 hrs data, use the last 50%, maximum allowed extrapolation = 6 x available data Recommendation: follow same protocol
Overview OLED standards in progress Notable items OLED product standardization Acuity product specification Color Quality Life Test Reliability and Early Failures
Assessing Total Early Failure Rate from Measurement on a Sub Population Finding a black marble = Find an early failure panel How do we determine the true occurrence rate of black marbles in the jar from measuring a portion of the jar? v
Determining Failure Rate ( ) from a Finite Number of Panel Measurements 90% of the area (cumulative probability) is between the upper and lower bounds. Probability 5% of the area is above the upper bound. Lower Bound Measured Rate Upper Bound Failure Rate When we measure a finite number of panels, there is a measured failure rate for the set. But we don t know if we were particularly lucky or unlucky with that set of panels. What we do have is a probability distribution of the actual failure rate. In the plot, a lognormal distribution is used to represent the probability density function of the true failure rate. This is still to be confirmed but the exact shape does not impact the conclusions drawn here. Lower and upper bounds of the 90% confidence interval are shown. Note the confidence level that the true failure rate is below the upper bound is 95%. What we care about: the upper bound at 95% confidence level as a very conservative estimate of the true failure rate.
Exact Determination of Confidence Interval for a Binomial Distribution Definition of the upper and lower bound Solution Where 1- is the confidence level, n is the number of panels tested, and k is the number of failures observed. p ub and p lb are the upper and lower bounds, respectively. Reference: http://www.sigmazone.com/binomial_confidence_interval.htm Clopper, C. and Pearson, S. Biometrika 26: 404-413, 1934.
Sample Results 1 alpha n k Avg rate x 1000 Pub x 1000 Plb x 1000 Number of panels Number of failures Measured failures per 1000 panels Upper bound of failures per 1000 panels Lower bound of failures per 1000 panels Confidence level actual rate < Pub 0.1 1000 1 1.0 4.7 0.1 95% 0.1 2000 2 1.0 3.1 0.2 95% 0.1 3000 3 1.0 2.6 0.3 95% 0.1 4000 4 1.0 2.3 0.3 95% 0.1 5000 5 1.0 2.1 0.4 95% 0.1 6000 6 1.0 2.0 0.4 95% 0.1 7000 7 1.0 1.9 0.5 95% 0.1 8000 8 1.0 1.8 0.5 95% 0.1 9000 9 1.0 1.7 0.5 95% 0.1 10000 10 1.0 1.7 0.5 95% Assume all measurements have average failure rate of 1:1000 As we increase the total number of panels measured, how does the upper bound evolve? Plotted on next page
Sample Results 1 Implied Failure Rate Upper Bound @ 95% Confidence Level vs. Total Panel Measured @ Constant Measured Failure Rate of 1:1000 5.0 Number of Failures 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 Measured average failure per 1000 Implied upper bound of failure per 1000, 95% confidence 0.5 0.0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 Number of Panels Measured There is a drastic reduction in the upper bound between n = 1000 and 2000, a smaller reduction between n = 2000 and 3000, and levels off beyond n = 3000. n = 2000-3000 seems to offer the best trade-off between certainty of results and the number of panels to test.
Sample Results 2 Having 95% confidence that the implied failure rate upper bound is 3:1000 requires 4 3.5 3 Number of panels measured/1000 Number of observed failures Observed failure rate per 1000 panels 2.5 2 1.5 1 0.5 0 1 2 3 4 5 Number of panels measured/1000 1 1.6 2.1 2.6 3 Number of observed failures 0 1 2 3 4 Observed failure rate per 1000 panels 0.0 0.6 1.0 1.2 1.3 Another way to look at testing requirement is to set a constant upper bound and see what test results are required to reach this upper bound. We also want to keep the measured rate below 1:1000. This plot shows required results to keep the upper bound below 3:1000.
Bath Tub Curve Burn-in must go beyond this point.