07-27-2006<br />
<br />

Many have asked, some actual data I ran across (FYI, these are Nichia 5mm LEDs):<br />
<br />
<img src="http://www.molalla.net/~leeper/5mmlife.jpg" border="0" alt="" />
<br />
<br />
You can draw a line from 100 to 80% at 1000 hours for 20mA drive.  If a fella overdrives them, they degrade faster and faster, the more they are overdriven...  Primary factor is heat.  My Nichia fella said to use 296 to 350 C/W for the thermal resistance from die to lead, and then one needs to also add in the C/W of the lead to ambient.  From my own testing, the C/W from lead to ambient is also a major factor, as heat is the #1 degradation mechanism- which has been proven over and over...phosphor degradation is #2 (in 5mm LEDs).  Though, die attach adhesive can be the biggest culprit of for some of the LEDs that are not made well-heat and wavelength affect the die attach adhesives.  You'll find CREE and LumiLEDs soldering their die instead of utilizing adhesives...<br />
<br />
Right, I have a graphic here:<br />
<br />
<img src="http://www.molalla.net/~leeper/ledlife.jpg" border="0" alt="" /></div>
<br />
The ones shown above are the 510BS LEDs.  It really isn't going to make a whole lot of difference (as compared to the CS).  You still have the thermal resistance from the die to the cup, then down the lead that runs out of the 5mm package.  It is in the 300-350 C/W range.<br />
<br />
When you over drive these at all, they are going to degrade *much* more rapidly.<br />
<br />
The lifetime of white LEDs has never been a secret.  It has been published widely since 2002 that they degrade fast, and the degradation rate accellerates *extremely* fast in overdrive.  Absolutely nothing at all new about this.  <br />
<br />
Now there have been alot of folks running around, ill-informed, making up stuff about 100,000 hours.  I asked the major players about this, and it was big news to them.  When pushed, one of the manufacturers provided this official factory testing information:<br />
<br />
<img src="http://www.molalla.net/~leeper/nichlife.jpg" border="0" alt="" /><br />
<br />
You don't remember when folks used to keep asking Gransee about what he did to his new ARC AA/AAA, as they kept saying that it is so much brighter than my old one!!!  It wasn't, it was the LED in the old one had degraded so rapidly.<br />

<br />
No, it is not the high power LEDs that have the rapid degradation.  They can degrade fast, if you over drive them, or have poor heatsinking, or both (which is really bad).  There is a reason for specs, and if one doesn't understand them, they are bound to have issues.  It is the 5mm white LEDs that have the very rapid degradation.<br />
<br />
However, if you underdrive the 5mm LEDs (1-5mA), I have data that shows their lifetime can be *greatly* extended.  Additionally, if the leads are soldered to copper slabs, right where they come out of the package (not 3mm later...), and heatsinks are added, and you use 30cfm of air on the heatsink, their life can be additionally increased on top of the underdrive, rather significantly yet again.<br />
<br />
Arrays of like three 5mm LEDs, that are closely spaced (say at 1 cm spacing) are a very bad idea, as the heat by the nearby LEDs makes matters even worse, and things compound alot.  Its kinda like riding a pissed off bull that is being stung by hornets heading for the safety of a barn that is on fire while an earthquake is in progress.  It gets much worse for arrays of more than three.<br />
<br />
If you chop off the leads, when you solder them flush on a pcb (which is usually FR-4/G-10), things get even worse.  The copper layer is typically 0.000707&quot; thick, and as such, provides really piss poor heatsinking.<br />
<br />
Oh, yeah, and their lm/W numbers go to hell in a hen basket when over driving, even just a little bit.</div>
<br />
<br />

<b><i>From a comment by SemiMan</b></div><br />
For 5mm white LEDS, the main source of light degradation IS NOT heat. Heat certainly plays a major impact when you start talking about LEDs life being reduced into the 5-10K hour life, but white 5mm LEDs have significant degredation at as little as a couple hundred hours into the low thousands of hours.<br />
<br />

This quick degradation occurs due to the yellowing of the epoxy encapsulation used in 5mm LEDS. The phosphor essentially &quot;traps&quot; the blue light near the phosphor causing rapid degredation of the epoxy. Almost all high powered LEDS and some mid-power surface mount LEDS use silicon encapsulation. This was one of the big advances that Lumileds brought to the market. This eliminated the rapid degredation of white LEDS. To get the 50K+ hours at 70% light output, of course lots of other things are needed such as good thermal management of the die, robust die structures that do not cause any &quot;hot spots&quot; due to current crowding, etc. It is difficult to even get reputable LED suppliers like Cree and Nichia to give you a very clear statement of how long their LEDs will last at a given die temperature. For the no name guys, they just say 100K hours on their data sheets but with little or no description under what conditions this will occur.<br />
<br />
The Luxeon and Luxeon3 dies were attached with solder, but I think the new K2 uses a different die attach method.<br />
<br />
Semiman</i><br />
<br />
<br />
Humm, don't you realize that epoxy degrades on exposure to heat?<br />
<br />
You should read a few of Mr. Narendran's papers where he studied the degradation mechanism, and himself, said it was heat, amongst other things.  It is the heat that affects the epoxy, combined with the short wavelength light.<br />
<br />
Even more so, it is good to recognize that it has been shown repeatedly that blue LEDs actually degrade slower than White LEDs.  This is counter-intuitive, if the major mechanism is due to the short wavelengths involved...as the White LED's phosphor actually convert a large portion the blue to longer wavelengths.  Visit the link below, under "Journal of Crystal Growth 268 (3-4): 449-456".<br />
<br />
I have just the critter under test right now, where I've *very* agressively addressed the heat issue, and they are running out past 20,000 hours now, and it will be a very long time before they cannot make the required light level.  My testing has actually been in process since 2001.<br />

<br />
IEEE/OSA Journal of Display Technology, Vol. 1, No. 1, September 2005<br />
<b>Life of LED-Based White Light Sources</b><br />
Nadarajah Narendran and Yimin Gu<br />
<i>"Even though light-emitting diodes (LEDs) may have a very long life, poorly designed LED lighting systems can experience a short life.  <b>Because heat at the p-n juncion is one of the main factors that affect the life of the LED</b>, by knowing the relationship between the life and heat, LED system manufacturers can design and buld long-lasting systems."</i><br />
<a href="http://www.lrc.rpi.edu/programs/solidstate/pdf/narendran-gu-JDT2005.pdf" target="_blank">http://www.lrc.rpi.edu/programs/sol...-gu-JDT2005.pdf</a><br />
<br />
<br />
Journal of Crystal growth, Vol 268, No. 3-4, pp. 449-456, August 2004<br />

Solid-State lighting failure analysis of White LEDs.<br />
Nadarajah Narendran and Yimin Gu and L. Deng<br />
<i>"Therefore, based on past studies, <b>the primary reason for the degradation of indicator-style white LED packages is the yellowing of the epoxy that is caused by excessive heat</b> at the p-n junction of the LED."</i><br />
<br />
<br />
Society of Photo-Optical Instrumentation Engineers<br />
Third International Conference on Solid State Lighting, Proceedings of SPIE 5187: 107-114<br />

Nadarajah Narendran and Yimin Gu<br />
<i>"Therefore, based on past studies, the <b>primary reason in the degradation of 5mm white LEDs is the yellowing of the epoxy due to thermal effects.</b>"</i><br />
<a href="http://www.lrc.rpi.edu/programs/solidstate/pdf/NonContactMethod.pdf" target="_blank">http://www.lrc.rpi.edu/programs/sol...ntactMethod.pdf</a><br />
<br />
<br />
International Society of Optical Engineers<br />
Fourth International Conference on Solid State Lighting, Proceedings of SPIE 5530<br />
White LED Performance<br />
Yimin Gu, Nadarajah Narendran, and Jean Paul Freyssinier<br />

<br />
<i>"Tests of 5mm white LED arrays showed that junction temperature increases produced by dirve current had a greater effect on the rate of light output degradation than junction temperature increases from ambient heat...The dissimilarity in temperature effect among 5mm and high-flux LEDs is likely caused by packaging differences between the two device types."</i><br />
<a href="http://www.lrc.rpi.edu/programs/solidstate/pdf/guSPIE.pdf" target="_blank">http://www.lrc.rpi.edu/programs/sol.../pdf/guSPIE.pdf</a><br />
<br />
<br />
Journal of Crystal Growth 268 (3-4): 449-456<br />
N. Narendran, Y. Gu, J.P. Freyssinier, H. Yu, and L. Deng<br />
<i>"Failure Analysis of White LEDs<br />
Experimental results showed that the degradation rate depends on both the junction temperature and the amplitude of short-wavelength radiation.  However, <b>the temperature effect was much greater than the short-wavelength amplitude effect.</b>"</i><br />
<a href="http://www.lrc.rpi.edu/programs/solidstate/pdf/narendran2004_jcrystalgrowth.pdf" target="_blank">http://www.lrc.rpi.edu/programs/sol...ystalgrowth.pdf</a></div>

<br />
<br />

Just so folks don't get confused, 5mm LEDs and SuperFlux LEDs are quite different creatures.  The SuperFlux LEDs have a very large metal pad inside, with multiple leads, which greatly enhances their thermal properties and allows you to drive them up at 50-70mA without severely degrading the LED.  They are also known as Spider LEDs, but not all are created equal.<br />
<br />
<img src="http://www.molalla.net/~leeper/supflux.jpg" border="0" alt="" /><br />
<a href="http://www.lumileds.com/products/line.cfm?lineId=4" target="_blank">http://www.lumileds.com/products/line.cfm?lineId=4</a><br />
<br />
I have seen these in white, but I don't remember who made a white version in this package.<br />
<br />
<br />
There is also the SnapLED, which is designed for 70-150mA, but is only available in the reds:<br />

<a href="http://www.lumileds.com/pdfs/DS08.PDF" target="_blank">http://www.lumileds.com/pdfs/DS08.PDF</a></div><br />
<br />
<br />
To further explain things, the Spider has much better thermal transfer than a 5mm LED.  This allows them to drive the part up at 70-150mA, and if you look at the picture above, you can see the large copper surface area which helps to spread out the heat, getting it away from the die, and helping to get rid of it, along with the two additional leads.<br />
<br />
The Luxeon was a later creation, which was even further improved over the 5mm and Spider LEDs.  In this case, they started out early on, with a large aluminum slug, which was designed to lower the thermal resistance even further.  Additionally, this slug was brought out of the bottom of the package, to allow attachment to a heatsink, in order to further get rid of the heat generated.  They also improved the die attach method, to lower the thermal resistance even further.  On top of all that, the Luxeon switched from an epoxy lens/case of the 5mm/Spider LEDs, which degrades with light exposure and heat.  The material they switched to is a silicone gel, which is much more robust in both cases.  Over this, they use a thin acrylic dome, that seals it.  This greatly reduces the lumen depreciation.  The die size was increased, in order to reduce the current density in the die (efficiency increase), and to allow for better thermal transfer.  There are also other improvements which also help.  After a little while, LumiLEDs changed the Aluminum slug over to a Copper slug, which further reduced the thermal resistance even more.  <br />
<br />
So, the Luxeon incorporates a great number of advances over the older 5mm and Spider LEDs.<br />
<br />
As such, the price that the market can bear for the Spider LEDs has dropped to well under a dollar.  They are not depressed as bad as the 5mm white LEDs, which go for under 25 cents each, even the premium 5mm Nichia white CS.  There is quite alot of competition in the small white LED area, with many companies competing directly with Nichia and Toyoda-Gosei (who used to be the two largest small white LED suppliers).  White SMT and 5mm LEDs made by other manufacturers besides the top five, can now be had in the 9 cent range.<br />
<br />

A company known as Citizen, has liscensed technology from Nichia, and purchases their die, to make new parts that are made from arrays of small LED die.  A few of these parts have small arrays of 24 die, the type that are used for 5mm LEDs, as the die are even lower cost yet.  These die are  mounted on metal substrates which can be bolted to metal heatsinks, to keep the efficiency and lifetime of the LED, up.<br />
<br />
There is another company that is known as Lamina Ceramics which does a similar thing to Citizen, with a different substrate make up, but their parts have lower efficienies-but are designed for higher power drive.  But if you underdrive them, you can gain alot of efficiency.</div>
<br />
<br />

<i>Comment by-jar3ds</strong></div>
				<div style="font-style:italic">newbie like always thanks for digging up and sharing it for all of us... very helpful info!</div></i>
			
<br />
<br />

<br />
A great example, that I've been testing utilizes a so to speak "overdriven" 5mm LED.  This is a practical example I happened to run across, that definitely applies.  This device was put on a regulated 2.8V, and put in a low cost flashlight made by Garrity, and the light output vs. time is shown.  You will notice instead of taking 875 hours to drop by 30% like the first example that was driven at 40mA, it only takes 20.5 hours to loose 30% of it's output:<br />
<br />
<img src="http://www.molalla.net/~leeper/smjledl1.png" border="0" alt="" /></div>
<br />
<br />
I have an after action report on a 5mm that degraded during testing here:<br />

<a href="http://www.candlepowerforums.com/vb/showpost.php?p=1597726&amp;postcount=163" target="_blank">http://www.candlepowerforums.com/vb...6&amp;postcount=163</a></div>
<br />
<br />
<br />
Here is a photograph of a cross-sectioned overdriven 5mm LED die, showing how the epoxy has "burned", and turned black, which is why it's output dropped down by 95% in only 240 hours (yes, 5% output):<br />
<br />
<img src="http://www.molalla.net/~leeper/smjledd1.jpg" border="0" alt="" /><br />
<br />
<br />
There are a bunch more photos like this, with the LED actually lit up while it is cross-sectioned like this here, post 163 and post 172:<br />
<br />
<a href="http://www.candlepowerforums.com/vb/showthread.php?p=1600819" target="_blank">http://www.candlepowerforums.com/vb...d.php?p=1600819</a><br />
<br />
<br />


Here is what a good ArcMania SMJLED2 PR SS (5mm LED) looks like before it has burned itself up:<br />
<img src="http://www.molalla.net/~leeper/amsmj1.jpg" border="0" alt="" /><br />
<br />
A good amount of additional photos of the 5mm LED before it burns itself up from being overdriven can be found here:<br />
<a href="http://www.candlepowerforums.com/vb/showthread.php?p=1622346#post1622346" target="_blank">http://www.candlepowerforums.com/vb...346#post1622346</a></div><br />
<br />
<a href="http://www.candlepowerforums.com/vb/showthread.php?p=1597726#post1597726" target="_blank">More pictures of internal LED failure mode here</a></div><br /><br />
<br />
<br />
<i>Comment-evan9162<br />

				<div style="font-style:italic">The epoxy isn't turning dark due to heat, its turning dark due to exposure to high power blue light.  There are plenty of high brightness green and red 5mm LEDs that don't suffer lifespan issues the same way white and blue LEDs do, and they still use a regular epoxy encapsulent.</i></div>
<br />
<br />
<br />
I will refer you back to an earlier post, where that common fallacy was dismissed:<br />
<br />
<br />
IEEE/OSA Journal of Display Technology, Vol. 1, No. 1, September 2005<br />

Life of LED-Based White Light Sources<br />
Nadarajah Narendran and Yimin Gu<br />
<i>"Even though light-emitting diodes (LEDs) may have a very long life, poorly designed LED lighting systems can experience a short life. Because heat at the p-n juncion is one of the main factors that affect the life of the LED, by knowing the relationship between the life and heat, LED system manufacturers can design and buld long-lasting systems."</i><br />
<a href="http://www.lrc.rpi.edu/programs/sol...-gu-JDT2005.pdf" target="_blank">http://www.lrc.rpi.edu/programs/sol...-gu-JDT2005.pdf</a><br />
<br />
<br />
Journal of Crystal growth, Vol 268, No. 3-4, pp. 449-456, August 2004<br />
Solid-State lighting&quot; failure analysis of White LEDs.<br />
Nadarajah Narendran and Yimin Gu and L. Deng<br />

<i>"Therefore, based on past studies, the primary reason for the degradation of indicator-style white LED packages is the yellowing of the epoxy that is caused by excessive heat at the p-n junction of the LED."</i><br />
<br />
<br />
Society of Photo-Optical Instrumentation Engineers<br />
Third International Conference on Solid State Lighting, Proceedings of SPIE 5187: 107-114<br />
Nadarajah Narendran and Yimin Gu<br />
<i>"Therefore, based on past studies, the primary reason in the degradation of 5mm white LEDs is the yellowing of the epoxy due to thermal effects."</i><br />
<a href="http://www.lrc.rpi.edu/programs/sol...ntactMethod.pdf" target="_blank">http://www.lrc.rpi.edu/programs/sol...ntactMethod.pdf</a><br />
<br />
<br />
International Society of Optical Engineers<br />

Fourth International Conference on Solid State Lighting, Proceedings of SPIE 5530<br />
White LED Performance<br />
Yimin Gu, Nadarajah Narendran, and Jean Paul Freyssinier<br />
<br />
<i>"Tests of 5mm white LED arrays showed that junction temperature increases produced by dirve current had a greater effect on the rate of light output degradation than junction temperature increases from ambient heat...The dissimilarity in temperature effect among 5mm and high-flux LEDs is likely caused by packaging differences between the two device types."</i><br />
<a href="http://www.lrc.rpi.edu/programs/sol.../pdf/guSPIE.pdf" target="_blank">http://www.lrc.rpi.edu/programs/sol.../pdf/guSPIE.pdf</a><br />
<br />
<br />
Journal of Crystal Growth 268 (3-4): 449-456<br />
N. Narendran, Y. Gu, J.P. Freyssinier, H. Yu, and L. Deng<br />

Failure Analysis of White LEDs<br />
<i>"Experimental results showed that the degradation rate depends on both the junction temperature and the amplitude of short-wavelength radiation. However, the temperature effect was much greater than the short-wavelength amplitude effect."</i><br />
<a href="http://www.lrc.rpi.edu/programs/sol...ystalgrowth.pdf" target="_blank">http://www.lrc.rpi.edu/programs/sol...ystalgrowth.pdf</a><br />
<br />
Note: the papers have moved to here, so some of the links may have changed:<br />
<a href="http://www.lrc.rpi.edu/programs/solidstate/SSLLRCAuthored.htm" target="_blank">http://www.lrc.rpi.edu/programs/sol...LRCAuthored.htm</a><br />
<br />
<br />
<br />
Pay particular note, where they ran the LEDs at the same current, but at different ambient temperatures, and saw a significant difference in light output over time.  Also how blue LEDs did not degrade as fast.<br />
<br />
<br />

Yes, light causes a part of the depreciation, but heat is also a big factor.  Die attach epoxy, discoloration of the reflector and other items also affect things.<br />
White LEDs degrade even faster than Blue LEDs.  Also, if it was due to phosphor degradation, the LED would shift blue as it ages, but it shifts yellow.<br />
<br />
If it was due to the blue light, much of the light is converted, right out of the die, so the amount of blue drops a lot.  This should cause the Blue LEDs to degrade faster, but the White LEDs degrade faster than the Blue LEDs.<br />
<br />
A good point to start from is here, then you can read the newer papers:<br />
<i>"Although junction heat and amplitude of short-wavelength radiation influenced the yellowing of the epoxy, and hense, the light output degratation rate of these types of white LEDs, the junction temperature had a much greater effect than the short-wavelength amplitude."</i><br />
Page 14<br />
<a href="http://www.lrc.rpi.edu/programs/solidstate/pdf/narendran2004_jcrystalgrowth.pdf" target="_blank">http://www.lrc.rpi.edu/programs/sol...ystalgrowth.pdf</a></div>
		


<br />
<br />

<a href="http://www.molalla.net/~leeper/led.htm">Main site</a><br>
Mirror:<br>
<a href="http://jarhead.hobby-site.com/~jarhead/led.htm">Main site mirror</a><br>

<br>
<br>	
<script src="http://www.google-analytics.com/urchin.js" type="text/javascript">
</script>
<script type="text/javascript">
_uacct = "UA-2114182-1";
urchinTracker();
</script>