Text-Alternative Version: LED Color Stability Webinar

Michael Royer: All right, welcome ladies and gentleman. I'm Michael Royer of Pacific Northwest National Laboratory, and I'd like to welcome you to today's webinar on LED Color Stability, Ten Important Questions, brought to you by the U.S. Department of Energy Solid-State Lighting Program.

We're happy to have as our speakers today Chad Stalker of Philips Lumileds Lighting Company and Ralph Tuttle of Cree Hong Kong Limited. And we'll go through this sort of sequentially so you'll hear a bit from each of us. But I'll give some intro bios right now.

Chad Stalker is the Regional Marketing Manager, Americas at Philips Lumileds Lighting Company. As the regional marketing manager for Philips Lumileds Lighting Company, Mr. Stalker is responsible for working across the lighting industry with fixture manufacturer, lighting designers and other market influencers to support the adoption of solid-state lighting technology. He has been working in the LED lighting industry for over 10 years with such companies as Luminous Devices, Osram, and Color Kinetics.

And following him, you'll hear from Ralph Tuttle, an Engineering Manager at Cree Hong Kong, Limited. He is currently living in Hong Kong where he is managing Cree's Asian-Pacific applications engineering group. Ralph has been with Cree for over 18 years. Prior to his current Hong Kong posting, he held various process engineering, product development, and development management positions in the materials, optoelectronics, and components group at Cree's headquarters.

He managed Cree's LED wafer fab, and was then involved in the development of Cree's original explant product family. Ralph is active within NEMA, IES, ANSI, and is a U.S. representative to the IECC. Ralph was coauthor of IES LM-80 and was the technical coordinator of the IES TM-21 Working Group. He has worked in the semiconductor industry for 20 years.

All right, now that I can be done reading, we'll get on with the webinar. So as I mentioned earlier, the title is LED Color Stability: Ten Important Questions. And so those questions I will be going through today are listed here. Just briefly, so you know what we're going to get to, so you don't have to jump ahead and be asking these questions that we might get to later; why and where is color stability important? What metrics are used to describe color shift and color stability? When does color shift become noticeable? Are there established tolerances for color shift? What standards apply to the measurement of color shift? What types of products do color stability measurement standards apply? What causes color shift? Are there methods or standards for projecting color shift in the future? Are there warranties available that cover color shift? And how should end users or specifiers monitor color over time?

So, these are all important questions, and I'm sure all of you have many more that you'll come up with during the presentation. So, we'll go through may be 30, 35 minutes, and then we'll have a significant amount of time to go through some of your questions.

So number one, and you get to listen to me a little bit more first: Why and where is color stability important? So, first let's sort of frame the question here. And color shift is really a comparison of spectral power distributions over time. And there's three aspects of this that are all important, but we're really talking about one of them today.

So, a lot of types of lamps emit light differently, depending on the operating condition. So, if the ambient air temperature changes, for example, you know, their output may change and their color may change. A lot of times, those are recoverable situations, so it's not really as concerning, and it's sort of just an inherent problem that can be remedied somewhat, but is not the major concern today. Really, what we're talking about now is, for some lamps and luminaires, the materials or construction may change over time, resulting in changes to the spectral output.

And then the sort of third aspect of this is that color shift is sort of an independent thing from lumen depreciation, although they can be related. So, if lumen output is changing, that can be because of color shift, or they could be happening simultaneously.

Now, another important thing is that what we're talking about is separate from color consistency. Where color consistency is the comparison of the initial output and the initial spectral power distribution for a group of matching lamps or luminaires. So, if I'm making an installation and there are 30 lamps aimed at a wall, you know, how consistent is the color of each one of those, that's really a separate issue from the color stability and the over-time aspect that we're going to be talking about today.

So, why and where are color stability important? Well, to meet lifetime claims, performance expectations must be met at all times. You know, LEDs are often sold as a long-term solution, but if they aren't meeting those expectations the whole time, they really didn't perform as expected.

So color shift can be a problem anywhere appearance is important, which is a lot of situations we're talking about with lighting, especially in museums, retail, residences. We're doing wall washing and we have multiple lamps or luminaires lined up next to each other, and if they're changing at different rates, we get sort of a rainbow effect on the wall. The same with facade lighting, coves, and direct view applications.

There are also some situations where it's less important, say roadways, parking lots, utility spaces where the specifier or purchaser might not want to pay a premium for a product that's going to have as tight of tolerances as something that would be used in a museum, for example.

In this slide, you see two pictures. The one on the top right is actually an early LED installation and a museum type application. And really, that's a major problem there where the color has shifted, and that's what LED not to be known as, something that ends up like this. And then in the bottom slide, we see the long-term results of a metal halide installation where I don't know exactly where this was taken — it looks like an airport, but I don't think they were going for some sort of national pride with the colors. It's more just an unintended consequence of those lamps shifting over time.

So, moving on to questions number two: What metrics are used to describe color shift or color stability? Something you might all be familiar with, but it's often really misunderstood, is MacAdam ellipses. So, just some brief background on those. They were experimentally derived using essentially one highly trained observer who is looking at a very small field of view and matching colors to a targeted chromaticity that was at the center of each of those ellipses.

And so, you might also hear them referred to as a standard deviation of color matching, because really, the actual ellipse came from measuring the standard deviation of that observer's matches. And then we can multiply that standard deviation by different steps, say three different steps to enlarge or decrease those ellipses. And probably one of the most important things about them is that they do not convey the direction of the color shift.

Some more on MacAdam ellipses. As I mentioned, they can be multiplied to get a one step, two step, or three step. Here, you'll see sort of an enlarging of that ellipse. And one of the important things that people often mistake is that the step difference is always relative to the center of the ellipse shown in the red circle on the screen.

And so, A and B, even though they're on the boundaries of a theoretical one step ellipse are not a one step difference as defined by this original experiment with the original highly trained observer. And the same with C and D. So, they're on the boundaries of a three step ellipse. They're actually more than a three step difference. They're actually double that. And so, that's one we see if we move into you know, ANSI tolerances, where we have a seven step ellipse. We're really at the opposite end of that ellipse. You're at a 14 step difference, so it's almost doubling the difference in the tolerance you might expect.

Now, MacAdam ellipses were developed in the 1931 chromaticity diagram, and as you saw previously, they were all different sizes. And so, this led a lot of smart people to modify these chromaticity diagrams, and without going through that whole history, we ended up in the 1976 chromaticity diagram, which is much more uniform in terms of its color difference throughout the whole diagram.

So, one thing we can do is use that diagram in a metric called Δuˈvˈ. And this is different from a metric you might have heard of called Duv, where Duv is just measuring the distance above or below the black body locus, and it's sort of related to correlated color temperature, or used in combination with correlated color temperature, Δuˈvˈ is something very different, and it's used for describing the color difference or color shift between two points. So, it's a simple, mathematical or Euclidean distance. You all think back to your geometry class in high school. You know, you can just take the square root of the squares of the differences in the coordinates and get your Δuˈvˈ.

So, one thing, again, like MacAdam ellipses, it does not convey the direction of the color shift. So, if I shift in one direction here or in the other direction, that would have the same number, but it would result in a very different appearing final light source appearance.

So, one thing that Δuˈvˈ does do is combine the effects of a shift in Duv and a shift in CCT, whereas either one of those alone wouldn't be very effective in conveying the absolute difference of the magnitude of the color shift. You know, Δuˈvˈ does a better job.

So, sort of relating these two back together, tying it in to perhaps some more familiar metrics, when we look at that chromaticity diagram, and as I mentioned, it was designed using the MacAdam ellipses to make them essentially into circles, or at least around the area of the black body locus. So, that allows for the conversion that a one step MacAdam ellipse is approximately equal to a Δuˈvˈ difference of .001. So, we can sort of translate one step, 001. Sort of an easy little trick to remember.

Now, as I mentioned, seven step MacAdam ellipses for CFLs were sort of modified into the quadrangles of ANSI for chromaticity bins for nominally white light. Again, seven steps. But really, the boundaries of this quadrangle are actually more like 20 steps. So, a very, very large tolerance there, and it's actually a lot of color difference.

And so, actually, leading into the next question, when does color shift become noticeable, and this is something I get asked all the time by all different kinds of people, and there really is not a good answer to it, because it depends. And what does it depend on? Well, it depends on the exact year. We all have slightly different visual tolerances, although that's a minor factor.

Your field of view. If I'm looking at a small two degree field of view, like in MacAdams experiments, I would have slightly different results than if I have a full field of view based on the distribution of photo receptors in our eyes. It depends on the surface characteristics. If I'm looking at a white wall, it would probably be more noticeable than if I'm looking at a very multi-chromatic wall, art work or something along those lines.

Proximity is a very important one, where if I have two light sources shining on a wall right next to each other, something — a relatively small difference might be noticeable. But if those two light sources are instead downlights at the opposite end of a large auditorium. I might have much greater tolerances for what's noticeable to a typical observer.

Now, in MacAdams experimental setup, a just noticeable difference was determined to be three times the standard deviation of color matching, essentially a three step ellipse for that given observer. For each observer, it varied a little bit, but generally, three steps was just noticeable for any observer if we redid the experiment.

So, that's sort of a general tolerance. Again, that applies to that very specific setup, which again, was more like the situation where you'd have two lights shining on a wall. But this has sort of generally made its way into our language, and a lot of people referred to, you know, in or outside that three step ellipse as being noticeable or not.

Again, though, I point out that it really depends on the application. In a museum, that three step ellipse might not be enough. In a general office environment, something larger, like a seven step difference might not get people too riled up about the difference.

All right, and the last little question you'll get from me. Are there established tolerances for color shift? Really, because of this application dependency, there's not a lot of very well defined tolerances. ENERGY STAR, which essentially is geared around the residential application, says that the lamp changing chromaticity from zero hour measurement at any measurement during the first 6,000 hours shall be within a seven step difference, .007, using Δuˈvˈ.

The LRC ASSIST program released some documentation when they were developing the ANSI chromaticity standards. Two step or four step ellipse for LED binning, so this is very different from necessarily color tolerances, were an application, but that was, again, sort of depending on how demanding the application is.

So, in general, to sort of give you an idea on what might be necessary, and again, on a case by case basis, we'll have to establish these specific tolerances for a project.

So, that's my first bit there. I'm going to turn it over to Chad now who is going to go through a couple more questions for you. Give it a second to get this transitioned.

Chad Stalker: Thank you, Michael. Good afternoon, good morning to everyone. So, one of the things that comes up pretty typically is people are asking what standards apply to the measurement of color shift. And in the industry now, probably the most common standard that you see is at the device level with LM-80. And some people are kind of surprised when we talk about this, because in their mind, they think of that as just being the lumen maintenance report that you get that has those typical graphs that you're familiar with, with the curves.

But if you were to read back through the reports, you'll find typically towards the end, a table that outlines some of the color differences that happen at each time. So, at a test level, what happens is, not only are you testing the lumen output and lumen maintenance of the devices, but they actually look at the total spectral flux every thousand hours, which is kind of the minimum time duration that you test an LED at before you take a data point.

So, at this time, understand that the LEDs are subjected to some level of drive current, some level of temperature for that thousand hour period. Then, they're removed and tested. Now, the testing is typically done at ambient temperature, so it's not something that's done in situ. It's actually taken out and done on a specific test bench. And that total chromaticity shift is tracked over the period of time for the testing, and it's actually included in the report.

Now, I kind of have an eye chart here for everyone. You can see here an example of what the tables typically look like in the report. And here, you'll see it's a Δuˈvˈ measurement at whatever drive current the device is being run at. So, you have this for the different drive currents and different temperatures.

To kind of give you a summary here, this is for a product, Luxeon TX. We're looking at what the performance is at 4000 Kelvin, 85 CRI. Drive current is 700 milliamps, and there's four different test points. You can see here, it's up to 10,000 hours of test, and over that period of time, I took basically what the differences are across at 10,000 hours is anywhere from .0006, which is basically 6/10 of a one step ellipse, up to .006, or saying basically 6 MacAdam ellipse. So, you can see how it changes typically like the higher temperature devices will show the largest amount of color deviation. Ralph's going to get into more of the details of how all that is done and what impacts that, but at a reporting level, at a standards level, this is the most common methodology for tracking and communicating this into the industry.

Now, this typically brings up the next question, which is, what about the system and luminaire level? And people typically assume that it would be in the IES LM-79 report, which is part of the testing of the overall solid-state lighting products. But at that test level, it's really only done at a fixed ambient temperature for the system and at a single point in time. So, that testing does not take into consideration tracking the actual color shift. All it does is report on — it states the color quantities, both chromaticity, CCT, and CRI, basically validating the output of the system at a certain ambient temperature. So, it doesn't continue on to look at overall color shift over time.

So that presents some questions that people start to look at. Well, what about what is being done in that case? And Ralph's going to talk to some of the things that are happening in the industry to look at system level color shift tracking and how to characterize that.

So, the next question we typically get, though, as people are looking at this, they understand the information is in the marketplace. So the question is what products does this apply to? And the simple answer, kind of going back to the previous question, right now, it's only being tracked and measured at the device level. So, we typically get kind of the obvious next question, is, can you take that information and apply it to a lamp in luminaire?

And the rationale where that comes from is people are used to being able to look at the lumen maintenance data out of an LM-80 report, utilize TM-21 methodology to characterize what the expected lumen maintenance of the system is over time, and then, that's typically captured as part of the LM-80 reporting, too. But in this case, that methodology, those characteristics have not been done yet where it's into a standard. But it still begs the question, you still have to think about it, because in some cases, people need to understand this.

So, I took an approach to look at it as where does it make sense, and how might you approach it for both an LED lamp and an LED luminaire? Thinking back, can we do this? Does it make sense to try to do it? So on an LED lamp side compared to an LED luminaire, I looked at, first thing is, what are these systems designed for and what's the criteria?

LED lamps are typically designed for lowest price, a reasonable amount of life, 10,000, 20,000 hours, and comparable photometric performance kind of versus the traditional lamp. I always describe LED lamps as similar to real estate. Real estate, it's basically location, location, location, while for LED lamps, it's price, performance and price again. So, that presents some challenges when you start looking at the system wide, which we'll talk about in a second.

But on the LED luminaire side, those are typically designed for more moderate to high priced, depending on the system, even comparable to existing technologies in the market. Robust life, 25,000 hours, 50,000 hours of life is expected. And I'll even challenge and say it's typically a high photometric performance versus traditional sources.

LED systems are typically expected to perform at equal or better performance levels when you look at things like uniformity, like we already talked about. Color consistency. It's an expectation at the luminaire level that it's actually a higher level of performance. So, the system design is typically more robust.

Now, let's talk about what's in the systems comparing the two. LED lamp, from a systems standpoint, you have your LEDs. Typically, the simplest optics possible. They want to keep managing the price. The lowest cost driver you can get and honestly, minimal thermal management. You're not looking at something where — because again, we mentioned about the lifetime being relatively shorter versus a system, but 10,000 hours, 20,000 hours is typically considered typical or even high.

On an LED luminaire side, though, things are different. You still have the LEDs. You have comparable optics, meaning you're looking for optics that will give the same performance or better than the basically traditional system. There is a driver that is designed to meet the overall lifetime, and there's also thermal management that's designed for the specific application. Indoor applications. They look at temperature differences in an environment. It can be 20, 30 degrees. Outdoor can be significantly larger. So, the thermal system is designed to handle that.

So, when you think of those scenarios of an LED lamp versus an LED luminaire, and you ask the question again about, can an LED package data be applied? When you think about it on an LED lamp side — so take that data and apply at the lamp, it's fairly risky. Those systems are typically designed to or close to the limits of most of the components. So, you're in a situation where any variations in the environment, any variations in manufacturing can typically put that performance at risk. It's designed for a specific life, a specific price, a specific performance level.

But take the situation of an LED luminaire. In this case, to take the performance of the LED luminaire, look at the system, you have a lot more flexibility. So, if you approach it as a system and you look at the elements that impact color, uniformity, and color stability, you actually can start to characterize it.

With the LEDs, you have the LM-80 characterization. You know what the performance is, temperature and drive current. You have that information in it. You can then look at the optics, and there's work being done with optics and material suppliers, whether it's at a lens material standpoint, the materials are going into a mixing chamber, the solder mask on your LED boards — all of those elements and all of those suppliers have been helping get more information on how light interacts and impacts those materials. And there's a level of characterization that's happening there.

So, you're able to take that, combined with LEDs, and looking at the thermal management system and be able to come up at a system level with a determination on what the overall color consistency would be over time with the product. And I think that, when you look at it for products in more the LED luminaire, where you can break down the system and understand what the system requirements are and performance are, you are able to look at what the color stability would be of that system over time.

So, I think the point being here is with LED lamps, you're running some risk, but with LED luminaires, looking at it from a system standpoint, you're able to put in some rationale and characterization to support that. With that, I think the next person — I'll turn this over to Ralph, and Ralph will continue on with the questions.

Ralph Tuttle: Okay. Thanks very much. Hopefully, everyone can see my screen at this point in time. Good afternoon, good morning, good evening — whatever time it is where you all are. It's currently 1:30 in the morning here in Hong Kong. It's a pleasure to be able to talk to you all today. I want to thank the Department of Energy for organizing this webinar.

Now, I'm going to be talking today about two items that are kind of near and dear to my heart. One is, what causes color shift, and also, can we actually predict color shift in solid-state lighting products. These are two subjects that I've been studying closely for the last year and a half or so.

What causes color shift? Okay, well, we're talking about three different situations here, three different causes. One is the LED packages themselves. We know that the materials, the manufacturing methods that are used in the construction of LED packages can contribute to color point stability over time. And with LED packages, the primary factor that contributes to LED color shift over time is actually operating temperature.

There are other factors, also. Primarily, there's operating temperature. And I'll go into some details on that in a moment. As far as luminaires are concerned, as with the packages themselves, the materials and the methods used to assemble the luminaires will affect the color shift of the product. And once again, temperature can be a very important factor in the shift in luminaires.

And lastly, the actual application of the luminaire. What is the environment that that luminaire is placed in to operate? And we have seen several instances where the environment can significantly affect the long point color control, color stability of the actual product.

So, let's talk about what causes color shift. And we're going to start out with LED packages, and in this case, plastic low and mid power types of packages. This is a cross section of a mid power LED package. You can see that there are LED chips inside the package. They are normally placed onto a reflective lead frame. The package itself is a white reflective material. And when we make this type of LED, we typically will mix the phosphor with a silicone material, and we simply pour it into the cavity of the package and thus, we have a plastic package mid power part.

One interesting thing about these plastic materials that are used to make most of these low and mid power packages is that the material is called polyphthalamide, or it's abbreviated PPA. In the industry, we know that this PPA material will discolor at high temperatures, and that discoloration will reduce the light being emitted from the LED package.

The PPA will also discolor when the blue photons, the high energy low wave different photons actually hit the side or hit the surface of the plastic. That will discolor the material, as well. So I've got this LED, and when I'm operating this LED, there is one component of light that is essentially emitted straight out of the LEDs, passes through the phosphorus silicone mixture and is emitted from the package. Also, there is light that is emitted from the LEDs and reflects off the side walls of the package itself, and then, is emitted from the package.

The color point of this LED is actually an average of all of the different light, all of the different photons that are emitted. There are going to be some photons which will be cooler than others. The cooler photons, the distance that they travel through the phosphor is relatively short, so the odds that they will impinge on a phosphor particle are low.

Other photons have a much longer path to get through the phosphor. The odds that they will impinge on a phosphor particle are quite high. The component of light that comes right out of the LED package will be much cooler than the light that is reflected off the interior of the package. So, I'm going to run this LED for some period of time, and as I run it, the package itself will begin to discolor.

The photons that are emitted from the top of the package, right out of the LED, they're going to continue to be emitted. The photons that are reflected off the sidewalls of the package are, in fact, going to be absorbed, because that package is discolored. So, the warm component of the light is going to be reduced, and as a result, we will tend to see on this kind of a product a blue shift in the color point.

This is data from some mid power LEDs that we tested, and in this case, we tested them for about 17,000 hours. These graphs are actually from a paper that I had published not too long ago in LEDs Magazine. And what we can see is that after 6,000 hours, the lumen maintenance is actually not too bad on this LED. The lumen output is still, on average, about 95 percent. And at about 6,000 hours, the shift in color point has only been about four steps — three to four steps on average.

However, after 17,000 hours, we can see that the lumen output has decreased significantly to where it's only at about 75 percent of the original output, and the color shift is up in the range of about a 20 step MacAdam ellipse. Clearly, not a good LED for general lighting applications.

Okay, how about ceramic high power LEDs? This example is for an LED where we take the blue chip, we coat it with phosphor, and generally, when we do that, the phosphor that is applied to the chip surface is mixed with some type of a binder. Some people use different silicone materials. There are some people who are trying to use some glass materials. But overall, the industry as a whole typically uses a binder of some type to mix the phosphor with, and apply it to the top of the chip. Then, the chip and the phosphor coating are encapsulated with a silicone lens.

Now, during the operation of this LED, over time, this phosphor that is applied to the top surface of the LED chip can crack. It can delaminate from the surface of the chip. It is all temperature related and it's a due distress. Now, if we're starting with this LED with a color point — let's say it's right on the black body curve at about 3000 Kelvin, this cracking and delamination can cause some real issues, because if the cracking and delamination is severe enough, the color will start to shift warm up above the black body curve.

Now, this color shift is due to a scattering of the light that occurs due to the cracking and delamination here on the surface of the LED, and essentially, what happens is the path that the blue photons are taking through the phosphor is increased with the additional scattering. And as a result, we get additional down conversion of those blue photons to the longer wavelengths, and we see this shift to the warmer end of the spectrum.

Now, depending on the type of binder that phosphors are mixed with, or the type of silicone that is used to encapsulate the LED, this effect may be greater or less with certain LEDs. I can say that we have tested many, many thousands of LEDs, not just our own LEDs, but LEDs from around the world, and we find that there is a significant variation from one LED type to the next as far as the color point stability when it comes to these high power ceramic based LEDs.

This color shift is definitely temperature dependent. It is directly correlated to the solder point temperature of the LEDs. A higher solder point temperature gives you a faster shift. Now, another thing that we have found is that LEDs where the junction of the LED chip is close to where the phosphor is located will result in a higher shift. So, the higher the junction temperature, the faster the shift. In addition, the higher the drive current, that correlates as we all know to a higher junction temperature and the faster shift.

So, for a given solder point temperature, if I have two LEDs that are running at the same solder point temperature, I'm able to maintain that. Yet, one is running at a higher drive current than the other. The LED with the higher drive current will shift faster in time than will the LED operating at a low drive current.

As the blue photons go through the phosphor, they hit the phosphor particles and they heat those phosphor particles up. The temperature of those phosphor particles is generally thought to be in the range of 35 to 45 degree C above the junction temperature of the LED itself. Although I actually gave a paper at an IEEE conference in Belgium last week, and there was a speaker there from an LED company located in Germany whose data indicated that the phosphor temperatures could be over a hundred degrees higher than the junction temperatures. And so, we're kind of reviewing his data to determine where he came up with those numbers. But in general, people believe those phosphor temperatures to be in the range of 35 to 45 degrees above the junction temperature.

The efficiency of that phosphor is also going to have an effect on how quickly the color shifts in the package. A phosphor that is very efficient, say it's 75 percent efficient, 25 percent of the photons are lost — of the energy is lost naturally due to heat from — if you get an LED that's less efficient than that, or the phosphor is less efficient, then the phosphor is going to get hotter, and it will shift faster than the more efficient phosphor will shift.

Okay, how about luminaire construction? We talked about packages. Well, we know that the luminaire construction itself can have a significant impact on long-term color point stability. The reflective surfaces that are inside the product can, over time, oxidize during the operation of the product, and you can get both a loss of light and a color shift there.

Lenses themselves, whether they be acrylic or polycarbonate, may begin to degrade over time. The spectral distribution of light that is coming from the product, that can also affect color point. Diffusers. You might start with a diffuser that is relatively transparent, and over time, it will discolor, turn dark. It can also affect the color point.

So, people say, well, you know, why don't we just use remote phosphors for all of our products? That way, the phosphor itself is away from the chip surface, so the delamination that you might see will not occur. If you've got a blue chip with the phosphor suspended above it, no problems. However, again, that phosphor will heat up due to the photonic energy from those blue photons, and the plastic materials themselves that the phosphors are coated onto can, in fact, discolor and can affect the color point of the product.

Application conditions. It's another issue that can definitely affect color point of a product over its time. We know in the industry that there are contaminants that can come into contact with the luminaire and with the LEDs that are in the luminaire, and it will accelerate and color shift. The most common is sulfur, I think. Everybody in the industry now knows that sulfur, which is present in the environment all around us, can actually get to the LEDs, and it can adversely affect the reflective silver surfaces in the product.

There are other corrosive materials that we know of that can attack plastics, that can attack the lens materials. You just need to be sensitive to that with our products and where we're placing those, what the application actually is.

So, how do we determine whether or not an LED package is going to shift color or an LED luminaire is going to shift in color? Chad had mentioned LM-80 testing, and the fact that during LM-80 testing, we, of course, measure not only lumen output of the product, but also, the chromaticity of the product.

Well, LM-80 recommends testing LEDs at 55, 85 degrees, and a third temperature that's generally left up to the discretion of the LED package manufacturer. Most manufacturers now also include either a hundred or a hundred and five degrees as a third test temperature. We at Cree and a few other people have been now testing LED packages at temperatures from 120 to 125 degrees C or higher. And I will tell you unequivocally, that there are many LEDs that do not like 125 C operating temperatures.

So, the question is, can we use those high temperatures to accelerate, to color shift and then predict what is going to happen over time? So, this is some data that I have from some LEDs that we tested, and we can see that at 125 degrees, here's the change in u-prime v-prime. At 125 degrees C, the change happens rather quickly. At 105 degrees, it happens, but it's pushed out. At 85 degrees, it's even slower. And for these LEDs, at the point in time when I made these measurements, the 55 degree data had not shifted at all. And these LEDs, of course, we still have on tests. We continue to test them.

So, the question is, can we take data of this type to develop an algorithm or algorithms that will then allow us to project the color shift over time the same way that we projected lumen maintenance a la TM-21. IES has approved a project initiation form, a PIF, to develop this type of a method, and the same way that I was the coordinator for TM-21, I will be the coordinator for this technical memorandum.

And I have been encouraging LED manufacturers, and I do so today in a public forum — I have been encouraging LED manufacturers to provide test data to help us all develop accurate production models. And we did that with TM-21. The major LED manufacturers pooled their data and we looked at it very closely and used that to develop TM-21. I'm hoping that we can do the same thing for a method to project color point stability of LED packages.

How about luminaires? Lynn Davis at RTI and his team have been doing a lot of very good work for the Department of Energy on developing on what they're calling a hammer test to evaluate luminaire reliability. And their testing actually involves multiple steps. You can see I've got — you know, there's a wet high temp operating life test sequence, and there's thermal shock. There's another wet high temp operating life sequence, and then there's a high temp operating life sequence.

And they have actually been able to cause color point shift in luminaires during their testing. So, the question now that I have is, can we, in fact, use this in the future to predict luminaire performance? And that's something that we're going to have to see as we move forward. So, that's the end of my section, and I'll turn it back over to Mike now.

Michael Royer: Okay, thanks, Ralph. Wrapping up these last couple of questions that we had previously identified here, and then we'll try and get over to your questions. Are warranties available that cover color shift? Device level warranties are often available, so we're talking LED packages, one to three years performance related to the component spec covering workmanship performance.

You know, extrapolating that over to end products, as sort of Chad alluded to earlier, can be a lot more difficult task, because you have a lot of multiple components going into one end product, all performing differently. Not one manufacturer has control over all of those components.

The most common warranties are based on lumen maintenance. So, L70 at 25,000 hours, L70 for five years, for example. There are some warranties that offer a little bit broader coverage, including color stability along with lumen maintenance. Sometimes, you'll see those relative to the balance of the installation or even some might be based on the owner's perception.

But really, it's not always straightforward here. What exactly is allowable? Is it a warranty relative to the starting point? If so, what is that starting point? Is it the specification for the product? Is it an actual measurement for one specific product that's going to be monitored over time? Or is the warranty relative to other lamps with the same hours of use? Is that essentially saying it's okay if they shift, as long as they all shift together and they're not shifting away from each other?

How much time should be covered? You know, we don't have a standardized production method yet. As Ralph mentioned, there's lots of work going on in that area. We do have sort of 6000 hour data, if a product has been tested according to LM-80. But as Ralph showed in some of his slides, it's often after that point where this color shift is becoming a problem.

And then, the final sort of big question is what are the procedures for documenting color shift exceeding a tolerance? And we have these other variables where the local ambient temperature is affecting the color output of a product. So, we have to take those products down and shift them and have them LM-79 tested to confirm some type of color shift. You know, it doesn't become an easy question to answer, and there's a lot of complications.

It is easier to look at color consistency over time, because we have products relative to one another in the same application. In certain situations, that might not be enough, though. If a certain group of LED lamps is installed along with maybe, you know, some fluorescent lamps or another type of LED lamp, if all of those are shifting together, they're all good relative to each other, not necessarily the other type of lamp in the same room.

So, oftentimes, it sort of comes down to a case by case basis, which isn't really an answer a lot of people want to hear, because it's more work for everyone. There's not sort of an industry standard. But in applications where it's really important, that might be the best solution.

So, sort of the last bit, if you're an end user or specifier, what should you be doing to monitor color shift over time? First, it's just establishing it's important. If it's not important, you know, maybe it's not worth all of this extra effort. And looking at manufacturer data, if there's LM-80 test data that's available going out to 20,000 hours and it shows minimal color shift, it's perhaps possibly more confident in that product than with something thought's only been tested to 6,000 hours.

If there is some warranty as an option, if you've discussed that with the manufacturer, what action could be taken if color shift is detected? Is it possible to correct that using some type of driver change or filtering or something? Is replacement the only option? Is replacing even going to be an option in the future? And then, if you are going to be monitoring, it's important to be proactive and establish a baseline.

So, if you get four years down the road, and you say, I think these lamps have shifted, well, if you don't have anything to go back to, what was the original state — pretty hard to establish that that color shift actually occurred. So, possible options, you know, doing a lot of LM-79 testing. I'm sure the testing labs love that, but end users, that's probably not an option for very many installations.

Is storing some extra samples of new products enough? Well, it would be enough if they were all perfectly color consistent, but that has to be evaluated, as well. And then, creating a plan up front for continued monitoring. Obviously, the most simple is just a visual analysis. A little bit more complex would be using a type of handheld meter to monitor things over time, again, remembering that there's some situations where the local ambient conditions are affecting color. Repeated LM-79 photometry takes out those other variables, but it's very expensive. And again, the big question is, what is the repeatability of all of these measurements?

So, a couple of examples here, and Chad provided these for me the other day, and I'll ask him to step in and just sort of explain these. These are just a couple of examples of what he's seen in the industry recently.

Chad Stalker: Thanks, Mike. Yeah, there's two examples that I've seen happen in more than one case. And this is typically very high profile installations — either it's significant investment on the owner's side, or visually, it's in a place that has an impact on the area. But the owner of the building will look for having the lighting designer, ESCO, or someone take as part of their project — take responsibility for monitoring, and they'll do validation testing.

In the case to where they want to hold to a level of warranty, they've had manufacturers put funds in bond that are then released back to the manufacturer as you hit certain milestones. And sometimes, this is done all the way just to really maintain a level of control on the installation. Sometimes, it's a term that's put in the agreement to really kind of ferret out maybe the secondary manufacturers who can't support this from a capital standpoint and really make sure that they get kind of a leading product in place.

The other example is, in a situation where an installation went in, it was very important that the installation maintain consistency, a lot of different sizes, shapes, colors going into play. And what they did is, they ran product in parallel off-site for two reasons. One is to monitor the performance of the off-site systems as a reference. Also, to kind of keep those as replacement products, should product fail at the installation.

Both approaches are pretty aggressive, but in situations where people want to maintain a level of performance and quality, it's probably less costly than having the installation go down or having problems with the installation and having patrons or people involved seeing that area compromised.

I think more and more, I hear from the design community and support community not atypical to have people at least come in around the end of the warranty to do a baseline, to check it and see where it is against the warranty, should they decide they want to pursue it. And so I hear this more and more as I talk to the community. Mike?

Michael Royer: Okay. We're at the conclusions. I'm not going to go through these individually, because we're getting short on time. We can go a few minutes over to answer questions, and I do have a whole bunch of questions, probably 20 or 30 of them. I'm just going to switch over here real quick to pull that up.

A few questions asked, how does color shift compare to T8 and T5 fluorescent, or is this one technology better than another? I don't know if either one of you, Chad or Ralph, want to take a stab at that first?

Chad Stalker: I'll take a first stab at it. I guess my experience has been with LED versus traditional. Sometimes, I've seen it where it actually manifests the same way as the picture that you showed. And I'm thinking of the one that showed the airport, where all of a sudden, the installation was an LED installation at a similar level of failure.

Sometimes, the failure modes were different, though. I'm familiar with the metal halide, where just the whole lot had hit extreme life and started to basically go out of spec. Where I saw it happen on the LED side wasn't an issue with going out of spec, but the installation was put into an environment that was outside kind of the design's ambient, and the system was under a lot of high temperature situation, and you had the failure. So, I've seen it manifest the same, but the root cause being completely different.

Michael Royer: I also want to add in, one important thing to remember is that LEDs and other technologies, too, are not really a monolithic technology. We've measured LED products that have extremely minimal color shift over up to 25,000 hours of actual measurement of end product. We've also seen some — and you've seen some examples where the color shift isn't good.

And I don't want this presentation to sound too alarming and that we've presented all this bad data where LEDs are shifting all over the place after short hours of use. I think most of that was just meant to illustrate some potential issues. But I think with any type of light source, including LED, there are different, even types within there, and all of those are performing somewhat differently. So, it's important not to generalize one necessarily being better than the other.

So, that begs the question, always, well, how do I tell the difference? It's not always necessarily easier, but if it's an important installation, you know, digging deep, working with that manufacturer to obtain as much data as is available at that point is an important first step.

Ralph Tuttle: And Michael, I think that's a good point to bring up. As I mentioned, at Cree, any time of the day or night, we have between 35 and 40 thousand LEDs that are in LM-80 testing at various temperatures and drive currents. And it's not just Cree LEDs, it's LEDs from manufacturers around the world. And there are, unquestionably, some very, very, very good LEDs. They hold their color point. They hold the lumen output. They last for a long, long time. At the same time, there are some LEDs that are made by some companies that don't fare quite as well. And so, it is very important that the consumers of the products talk to the suppliers and look at the data. Data speaks more than anything else to ensure that the product that they're procuring meets their color point requirements. Michael Royer: Okay. I have like 30 or 40 questions here. A lot of them are good. I'm going to pick another couple. We'll go over it a little bit. Something different here, more technical. It says does decycling in the form, say, of pulse width modulation impact the robustness of phosphor lamination? Ralph Tuttle: Have not seen. That's a good question. Normally, even when you're doing a PDWM operation of the LEDs, that junction temperature doesn't vary very, very fast. So, what we're looking at when it comes to this type of a cracking in delamination phenomena on some high powered LEDs is the stress due to the temperature differentials between the junction and what's happening above in both the phosphor layer and the silicone layer. So, there is no indication from the testing that we've done with pulse width modulation in any way affects color point stability over time.

Michael Royer: Okay. I had a couple related to CRI and color shift. I can do a quick answer to this. Does color shift affect CRI? If you're changing the spectral output, you're going to change the way objects look. One of the issues with CRI is that you're always going to a reference of the same CCT. So if you're changing CCT, you're also changing your reference for your CRI calculation. So, your number from your CRI value, depending on the exact way the shift is occurring, might not actually change much.

But that again does not mean — it's also depending on the magnitude of the shift, that all of your actual colors being illuminated will appear the same. And there's sort of a related question in there, does LED color shift vary as to the function of CRI? Ralph or Chad, do you want to step in on that?

Ralph Tuttle: Sure. I can speak to that. In order to achieve high CRI on warm white products, normally, red phosphors are added to the product in order to of course, get that red component up higher and higher. Red phosphor tends to be notoriously inefficient due to Stokes shifts. And so, I would say that in general, a product that has a high CRI could experience a more significant color shift over time than a product that has a low CRI.

But again, that's with the caveat that the product is built in such a manner that color shift will occur. And I want to once again repeat what I said earlier. There's a lot of good LEDs on the market today, and there's also a lot of bad LEDs on the market today. So, you need to be — you know, when selecting LEDs, it's important to look at the data and understand what's actually happening.

Chad Stalker: So, Ralph, maybe another way to say it is the higher CRI is probably a secondary or tertiary factor versus mechanical design or just overall quality of the LED.

Ralph Tuttle: Yes, you could say that. Yes.

Michael Royer: Okay. I'll get one or two more in here. This one reads, research shows LEDs degrade due to heat without being turned on. How can LEDs be used for exterior locations, and what specs should be used to ensure no failure due to exterior temperature?

Ralph Tuttle: The lumen depreciation of LEDs when exposed to heat is typically due to the breakdown of the silicones that are used to encapsulate that product. Generally, temperatures have to get up very high for that to occur. Most silicones that are used to encapsulate the product are stable to temperatures up above 200 degrees centigrade.

So, if an LED is sitting around or a luminaire is sitting around in an environment that's greater than 200 degrees centigrade, yeah, there could be some problems. But I don't think there's too many solid-state lighting products that are going to be in that type of an environment.

Now, that being said, we do know that at moderate temperatures, say 100 degrees, 125 degrees, there can be some change in the silicone materials that will affect lumen depreciation. But those changes are very minor, and we're now then talking about 2 percent, maybe 3 percent lumen depreciation at those more moderate temperatures. So, it does not really — you know, storage temperature does not significantly impact long-term lumen maintenance of the product.

Michael Royer: Okay. We're about five minutes over now. I'm going to wrap it up there. I know there's a lot of other great questions that didn't get answered. Hopefully, at some point, you can talk to one of us and get those questions answered. Really, this is becoming a more and more important issue, I think, that people are becoming aware of, and I hope the industry continues to have good discussions about this topic as we continue to improve the technology.

So again, this will be available in PDF and WAV form, and thank you all for participating in today's webinar brought to you by the U.S. Department of Energy. You may all disconnect. Thanks. Have a good day.