Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | SubscribeqMqM | Next PageqqM qMMQmagsTHE WORLD’S NEWSSTAND ®standards | TM-21Next, the WG analyzed LED lumen maintenancebehavior using over 40 sets of LM-80-08test data (20 sets with 10,000 hours or more)collected from four major LED manufacturers.The working group also examined theaccuracy of proposed projections using variousproposed models and LM-80 data thatextended up to 15,000 hours.Insight from LED manufacturers’ dataThis analysis showed that the LED lumendepreciation trends often change after 6000hours in one way or another and there is noreliable and consistent approach to predictsuch trends from 6000-hour data points.Ts,1, ( 0 C) 55Ts,1, (K) 328.15a11.684E-06B1 0.9639Ts,2, ( 0 C) 85Ts,2, (K) 358.15a23.354E-06B2 0.9525Ea/kB 2699A6.283E-03B09.582E-01Ts,i, ( 0 C) 70Ts,i, (K) 343.15ai2.413E-06Projected L70(Dk) 130,131Reported L70(Dk) >60,000TABLE 1. Parameters of interpolationusing 10,000 hours of LM-80 data for insitucase temperature Ts,i of 70˚C.Even a 10,000 hour stream of data is oftennot sufficient to provide rigid statistical confidencein an extrapolation of a decay curveout to large numbers such as 35,000 hours.In addition, the LED industry understandsthat degradation associated witheffects other than the diode itself tends tomanifest earlier in life rather than later. TheWG concluded that the most reasonableapproach to extrapolation of lumen degradationwas to avoid using any initial variabledata. This initial data variability is commonlyassociated with an early rise in outputfollowed by the eventual decline, identifiedas a hump in the data. Instead, by utilizinga fit to the later data streams (Figs. 2 and 3),a more accurate representation of the product’slumen degradation curve is obtained.What TM-21 is and is not!First, TM-21 does not determine a traditionallife or time-to-failure like we are used to forother lighting sources.What TM-21 does provide is a projectionof the lumen maintenance of an LED source(package/array/module) based on data collectedaccording to LM-80. This projectioninformation can then be used to project theexpected lumen degradation of the lightsource as part of a complete system (fixture).It is important to note, however, that TM-21is not a complete lifetime estimation tool.TM-21 does make use of all data providedin an LM-80 test format. LM-80 requires6000 hours of data at 1000-hour increments,but the TM-21 WG determined that longerstreams of data and data taken at shorterintervals (less than 1000 hours) will providefor a better estimate. The methodology firstnormalizes the provided data to 1 (100%) at0 hours and averages each point for all samplesof the device for each test condition(temperature).TM-21 also provides a suggested samplesize of 20 LED packages, arrays or modules.Based on an evaluation of the uncertaintyof measurement at various samplesizes, a larger sample size (30) does notsignificantly increase the uncertaintyand smaller size (10) would significantlyreduce the uncertainty of the degradationestimates.How it worksAs discussed previously, later data tends toexhibit the more characteristic decay curvethat is of interest. So for 6000 hours of data(LM-80 minimum) and up to 10,000 hours,TM-21 applies curve fitting to the last 5000hours of collected data. For collected datagreater than 10,000 hours, TM-21 uses thelast half of the data.The method then applies an exponentialleast-squares-curve fit to the data:Φ(t) = B exp(-αt)Where:t = operating time in hours;Φ(t) = averaged normalized luminousflux output at time t;Relative lumen maintenance1.101.081.061.041.021.000.980.960.946k hour fitTest data10k hour fit0.920.900 5000 10,000Operated hoursRelative lumen maintenance1.051.000.950.900.85Test data6k hour fit10k hour fit15,000FIG. 2. Later TM-80 data points providea better estimate of long-term lumenmaintenance.0 5000 10,000 15,000Operated hoursFIG. 3. As also shown in Fig. 2, a fi t toLM-80 data collected later provides thebest estimate of lumen maintenance forLEDs.B = projected initial constant derivedby the least-squares-curve fit;α = decay rate constant derived by theleast-squares-curve fit.And a projected lumen maintenance valueis derived using:Lp = ln(100∙B/p)/αWhere:Lp= lumen maintenance life in hourswhere p is the maintained percentage of initiallumen outputThis process can accommodate any useridentifiedlevel of light output (i.e. L70 or L50).If the desired light output level is reachedduring the course of LM-80 testing, then thattime value is reported.During the development of the method, it38 NOVEMBER/DECEMBER 2011 LEDsmagazine.comPrevious Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | SubscribeqMqM qMM MQmags| Next Page q qTHE WORLD’S NEWSSTAND ®
Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | SubscribeqMqM | Next PageqqM qMMQmagsTHE WORLD’S NEWSSTAND ®standards | TM-21became clear that limited 6000-hour data canproduce mathematical results that are notnecessarily rational or believable. For example,flat or increasing delay curves would beunreasonable for LED products. Clearly, reasonablelimits on predictions are needed.The WG analyzed 40 sets of data to determinemodel-fit uncertainty and its relationshipto prediction limits. The recommendedextrapolation limits are:• For a sample size of 20 units, the maximumprojection = 6× the test duration• For a sample size of only 10 units, themaximum projection = 5.5× the testduration.TM-21 also provides for interpolatingbetween test data curves to determine theactual degradation characteristics the productwill exhibit when installed in a fixturewhere heat conditions will change. When thein-fixture measured temperature is withinthe LM-80 test temperatures (55°C, 85°C, andmanufacturer chosen), then TM-21 prescribesthe use of the Arrhenius equation to interpolatebetween test temperatures. The Arrheniusrelationship accounts for the temperatureeffect on decay rate constants, which iscritical for temperature-dependent LEDs.For instance, Table 1 presents an exampleof an extrapolation with an in-fixturetemperature of 70˚C with the graphicLumen maintenance(%)100908055°Cdata85°C55°Cdata 70°C85°C700 10,000 20,000 30,000 40,000 50,000 60,000Time (h)FIG. 4. Degradation curves from measured data and fromin-fi xture interpolation at 70˚C.results in Figure 4. Extrapolation outsideof LM-80 test temperatures is not recommendedor supported.The TM-21 document provides a nomenclaturefor results to provide uniform reportingbetween products. The general form is:Lp (Dk)Where:p = the percentage of initial lumen outputthat is maintainedD = the total duration of the test in hoursdivided by 1000 and roundedFor example:L70(6k) = 34,000 hours represents resultsat 6000 hours of test dataL70(10k) = 51,000 hours represents resultsat 10,000 hours of test dataL70(6k) > 36,000 hours represents resultswith the 6 times rule appliedL70(4k) = 4400 hours represents results ata value reached during testing.The methodology is provided in theTM-21 document as formulas and procedurebut can be developed into spreadsheetsor other software-tool formats. Themethod can also be misapplied in softwareprograms, so a set of examples is includedfor users to check the results.Where does TM-21 fit in the industry’scharacterization of LED life?TM-21 is the result of extensive analysis andit presents a realistic method of projectinglumen maintenance of LED packages,arrays and modules in actual operating conditions.However, diode lumen degradationover time is only one of many possible LEDproduct-degradationmechanisms. The industryis seeking to betterunderstand componentreliability metrics suchas driver life, and theeffect on lighting productperformance.The TM-21 documentmay be applied in variousways. It can be usedby laboratories, manufacturers,or othersas defined by the programor organizationrequiring its application.TM-21 data can berequested as part of LM-80 testing or appliedby the manufacturer of an LED product.TM-21 is expected to achieve widespread useby organizations and program certifications(e.g. Energy Star) as part of their required documentationof performance.____________________LEDsmagazine.com NOVEMBER/DECEMBER 2011 39Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | SubscribeqMqM qMM MQmags| Next Page q qTHE WORLD’S NEWSSTAND ®