How do LED's go out?

Solar Powered Pickle

Newly Enlightened
Joined
Feb 11, 2008
Messages
55
I know most of these LED's are good for at least 50,000 hours but how do they go out? Do they just get dimmer and dimmer as they reach 50,000 hours of use or do they just simply do not turn on anymore?
 
The number on most of the high power LEDs is the time it takes to reach 70% output. So it will still be lit after 50,000 hours but at a less efficient level.

I've also read that LEDs don't degrade linearly, they degrade faster and faster as they are used.

After a long enough time, the LED would not light anymore. But you would probably have changed lights by then.
 
While the LED itself might not go out, to be practical, you need to consider rest of the whole package that you are using the LED with.

The wires/connections/circuits/switches/batteries might fail. When there is a failure it may either destroy the LED or costs more to fix the replacing the entire unit.
Especially with high powered flashlights, just about every one has some circuit or other part that is probably more prone to failure then the LED itself.
And even if the parts are good, one bad battery could leak strange chemicals everywhere.
 
Food for thought ...

L1 Cree ... on a fresh battery lasts 1.6 hours on high.

50,000 hour LED life / 1.6 hours per battery = 31,250 batteries ...

If you spent $1 per battery, you'd have spent $31,250 on batteries by the time it goes out.
 
Hello Solar Powered Pickle,

I have had a couple of LED's just go out like an incandescent lamp goes out. I think this is less common than the circuit going out, but it can happen.

Tom
 
Hello Solar Powered Pickle,

I have had a couple of LED's just go out like an incandescent lamp goes out. I think this is less common than the circuit going out, but it can happen.

Tom

Doesn't that happen when you seriously overdrive the LED.
 
Ive had LEDs flicker before dying but these were ones that had been overheated. likewise the tint can shift (before they die) when overheated.
 
So how do we prolong the useful life of our LED lights? I know that overdriving an LED is a big no-no due to overheating.

For instance, if my Cree P60 drop-in is recommended to run between 3.7v-9.0v, would running it at 9.0v cause the LED to shorten its life and dim significantly faster than compared to running it at say 3.7v???

What part does the regulation/driver play in maintaining the life on an LED??
 
I know most of these LED's are good for at least 50,000 hours but how do they go out? Do they just get dimmer and dimmer as they reach 50,000 hours of use or do they just simply do not turn on anymore?

I'm not certain what the damage mechanisms are. I do know that if one of these LED doesn't suffer from SID (sudden infant death) they live a long time.

The 50,000 hour figure (at least for the CREE parts) is to 70% of the rated output. The 'exponential' decay of the LED means (I think) that most of degradation occurs right away - as time goes on it fades more slowly.

(see this thread 'LED's Fade over time: true or false..'):
http://www.candlepowerforums.com/vb/showthread.php?t=146371

Assuming the LED is properly heatsinked at not driven overspec the part should essentially last a lifetime, and that's with a lot of use. There's also some evidence the Scintillator/phosphor degrades faster than the LED itself. The scintillator takes the blue light from the LED and creates the green & red needed to make the white light - nearly all commercial white LEDs are a combination of a blue LED with a phosphor. The phosphor fades over time (like a burned in old school crt TV) and this can effect the color balance and brightness, especially used a high drive levels.

Anyway the newer LED offerings should last much longer than the older 5 mm LEDs.

I've been using an L2D CE as computer desk lamp for over a year, and it probably has 1-2 thousand hours of use. I can see no change in the light output.
 
here is a good read, hope the links work

Failure modes
The most common way for LEDs (and diode lasers) to fail is the gradual lowering of light output and loss of efficiency. However, sudden failures can occur as well.

The mechanism of degradation of the active region, where the radiative recombination occurs, involves nucleation and growth of dislocations; this requires a presence of an existing defect in the crystal and is accelerated by heat, high current density, and emitted light. Gallium arsenide and aluminium gallium arsenide are more susceptible to this mechanism than gallium arsenide phosphide and indium phosphide. Due to different properties of the active regions, gallium nitride and indium gallium nitride are virtually insensitive to this kind of defect; however, high current density can cause electromigration of atoms out of the active regions, leading to emergence of dislocations and point defects, acting as nonradiative recombination centers and producing heat instead of light. Ionizing radiation can lead to the creation of such defects as well, which leads to issues with radiation hardening of circuits containing LEDs (e.g., in optoisolators). Early red LEDs were notable for their short lifetime.

White LEDs often use one or more phosphors. The phosphors tend to degrade with heat and age, losing efficiency and causing changes in the produced light color. Pink LEDs often use an organic phosphor formulation which may degrade after just a few hours of operation causing a major shift in output color.

High electrical currents at elevated temperatures can cause diffusion of metal atoms from the electrodes into the active region. Some materials, notably indium tin oxide and silver, are subject to electromigration. In some cases, especially with GaN/InGaN diodes, a barrier metal layer is used to hinder the electromigration effects. Mechanical stresses, high currents, and corrosive environment can lead to formation of whiskers, causing short circuits.

High-power LEDs are susceptible to current crowding, nonhomogenous distribution of the current density over the junction. This may lead to creation of localized hot spots, which poses risk of thermal runaway. Nonhomogenities in the substrate, causing localized loss of thermal conductivity, aggravate the situation; most common ones are voids caused by incomplete soldering, or by electromigration effects and Kirkendall voiding. Thermal runaway is a common cause of LED failures.

Laser diodes may be subject to catastrophic optical damage, when the light output exceeds a critical level and causes melting of the facet.

Some materials of the plastic package tend to yellow when subjected to heat, causing partial absorption (and therefore loss of efficiency) of the affected wavelengths.

Sudden failures are most often caused by thermal stresses. When the epoxy resin used in packaging reaches its glass transition temperature, it starts rapidly expanding, causing mechanical stresses on the semiconductor and the bonded contact, weakening it or even tearing it off. Conversely, very low temperatures can cause cracking of the packaging.

Electrostatic discharge (ESD) may cause immediate failure of the semiconductor junction, a permanent shift of its parameters, or latent damage causing increased rate of degradation. LEDs and lasers grown on sapphire substrate are more susceptible to ESD damage
 
The 50,000 hour figure (at least for the CREE parts) is to 70% of the rated output. The 'exponential' decay of the LED means (I think) that most of degradation occurs right away - as time goes on it fades more slowly.

Brightness declines slowly at first, then faster. This post has a good chart of accelerated decay.
 
For instance, if my Cree P60 drop-in is recommended to run between 3.7v-9.0v, would running it at 9.0v cause the LED to shorten its life and dim significantly faster than compared to running it at say 3.7v???
A light that gives you a range of voltages like that is running what is called a buck converter -- that is, it takes in a DC voltage (in this case anywhere from 3.7v - 9.0v) then lowers it to the Vf of the LED, which is usually around 3.5. Boost converters, such as the L0D-CE, taken in a lower voltage (1.2V from a AAA) and increase that to the Vf of the LED. Buck-boost converters can go both ways, for example, my EDC, a Liteflux LF2, can run on either a AAA (boost) or a 10440 LiIon (buck).

With any good DC converter, as long as the battery input voltage is within the specified range, the LED should be receiving the same amount of power no matter what. In your case, if your battery voltage drops below 3.7V, the light will direct-drive, and gradually lose output. If you load a battery over 9V, you will probably burn out the converter board.
 
Last edited:
Ive had LEDs flicker before dying but these were ones that had been overheated. likewise the tint can shift (before they die) when overheated.


Explain the tint shifting before the LED goes out. If the LED has a white tint and starts to go out does the ting shift to a warmer color? or even a cooler color like blue or purple?
 
Brightness declines slowly at first, then faster. This post has a good chart of accelerated decay.

lol, ah I'm the one that originally posted that.

At first I thought as you did, but I was wrong. Most of the decay occurs early on, then it slows down. Look at the horizontal part of the graph, it's logarithmic.

From the original thread:
http://www.candlepowerforums.com/vb/showthread.php?t=146371&page=2

"IMSabbel"
Logarithmic graphs can be VERY decieving. This one especially makes the first 1000 hours cover the biggest part of the graph, making it seem much more stable.
Take a look at the lowest of those lines:

After 1000 hours its down to 72%. (losing 28% of original intensity)
Going to 2000 hours (only tiny bit further right than the 1000h mark) only dropy to 56%, a drop of 18% of the original intensity.
"wasBlinded"
The curves on your graphs are just like IMSabbel suggests. The decay is greatest in absolute terms early. Put those data curves on a linear time base graph and it would be obvious.
 
Explain the tint shifting before the LED goes out. If the LED has a white tint and starts to go out does the ting shift to a warmer color? or even a cooler color like blue or purple?

It's because of how the 'white' LED is constructed. Actually there are no true 'white' LEDs. LEDs typically produce light in a narrow part of the spectrum, that is they appear as a single color (red, green, etc.). To get white light from a LED, a blue LED is combined a phosphor. Some of the blue light is converted by the phosphor to other colors, which when combined with the blue LED create a good approximation of white light.

The phosphor is placed on top of the LED.. if you look at a cree LED (when it's off!) you'll see it appears a yellow color. This is caused by a dome of phosphor goo sitting on top of the blue LED.

Over time the phosphor degrades and as it does the tint of the LED will change. I'm not certain how the tint will shift, this would depend on the composition of the phosphor and the type of LED. Note,, the tint shifting occurs as the LED ages (over many years likely)... not as the battery dies (assuming the light is regulated). {edit} however, the tiny of white LEDs does vary depending on drive current. The white they are rated as is only at a given drive current. Lower drive currents tend to make the output warmer I believe.

It's interesting to look at the tint of LEDs, because often it will vary depending on where you look in the beam. For most crees the light near the edge of the spot will be warmer then light in the center, which has a cooler cast. This likely is because light on the edges of the beam travels thru more phosphor. I've noticed Luxeon LEDs seem to have less of a tint variation across the beam.
 
Last edited:
Top