How can they possibly know lifespan of LEDs?

peacefuljeffrey · May 25, 2006

If white LED technology is as recent as I think it is, how is it that anyone (especially when I first bought a small, cheap Brinkmann in 2000 or so) can possibly know that they will last 100,000 hours (11 years, give or take)?

Did they sit on some prototypes for 10 years or more before releasing them to the public, just to make sure they would last as long as that? Let's say 2000 was the first year for white LEDs; would that mean they were developed in 1989 or so?

What is the test that they use to be able to say, "This ought to last for 100,000 hours"?

-Jeffrey

PhotonWrangler · May 25, 2006

They can measure the decrease in light output over time and make projections based on that.

martytoo · May 25, 2006

From what I've heard they take a few beer cans and some old garbage of vaious lineage, throw it in the DeLorean and then check the light every year or so until it is found to be extinguished.

http://www.bttf.com/delorean.htm

:grin2:

Redjam · May 25, 2006

I thought they ran them at 50 amps for 4 nanoseconds and extrapolated from that.

PhotonWrangler · May 25, 2006

That could run the risk of turning them into NEDs (noise emitting diodes).

FirstDsent · May 25, 2006

You have to hold your mouth just right.

Bernie

peacefuljeffrey · May 26, 2006

Hey hey hey hey HEY!!!!

Around here, when 90% of the time people are talking techno-speak that I can't begin to fathom, why is it that when I ask a serious question I get five replies that are nothing but jokes?! GRRR! :rant:

Seriously, I was asking for real -- so where are all the helpful flashaholic tech-geeks to come out and explain it to me?

How could measuring the dimming over time be useful for determining the lifespan of an LED? Would they start by calling the light output on day 1 "100%" and then wait til one year had gone by (if they even waited that long) and if it had dimished in brightness by 10% they'd say, "Oh, well, then, 10% per year, it'll be dead in 10 years"? Wouldn't that mean you'd be expecting 10% light output for that final year?

No, I don't suspect that they measure the dimming and then extrapolate. Doesn't make sense. Besides, I have no way to know whether LEDs even are known to experience gradual dimming as a function of age/use, and whether that would occur on a predictable gradient reliable enough to be used as a lifespan projection.

I mean, how do we know that LEDs don't function function function until one day they just die? I'm sure some do. But a proper, non-defective one, they claim to know that it will work for 50,000 or 100,000 hours, generally. HOW can they KNOW that?

-Jeffrey

FirstDsent · May 26, 2006

They may as well guess. It's obviously rounded even if they have careful projections. They may as well say 1 million hours. If one of the luxeons in my trafficl lights dims out 15 years from now what am I going to do? Beat-up Lumileds for a warranty replacement?

"Dear Lumileds,
One out of 672 of the Luxeon I emitters in one of our city's 2,280 traffic lights only lasted 79,126 hours. You'll have to take my word for it because it is impossible to document that It was on for 79,126 hours. I assure you that it was well heat sinked, driven within limits, and sufferd no voltage spikes in the 10 years it operated at the intersection of Peachtree Rd, and Peachtree Dr. near Peachtree Plaza in the Peachtree district of Atlanta. It never came in contact with moisture, or excess heat from direct exposure to sunlight to the metal enclosure. On behalf of the City of Atlanta, I demand full compensation! Please send me a new $3.45 emitter".

Soooooo,
Who cares about their silly claim?

For flashlight use, its a worthless number anyway. NOBODY will ever experience a "natural death" of a Luxeon if operated within it's recommended drive limits. If I were building traffic lights I would only give a crap so I can repeat the lie to my customers. They won't come back on me if it only lasts 79,126 hours either. For flashligts, I would buy one if it only lasted 500 hours. If I ever have to put that kind of mileage on a light, it will be after the apocalypse, and some mutant will probably kill me for it first.

Damn mutants!
Bernie

carrot · May 26, 2006

I'm under the impression that Marketing pulls these numbers out of their rear ends. Aside from that, degradation of a LED {is|should be} fairly predictable, so they can come up with numbers that are fair estimates.

Canuke · May 26, 2006

peacefuljeffrey said:
Hey hey hey hey HEY!!!!

Around here, when 90% of the time people are talking techno-speak that I can't begin to fathom, why is it that when I ask a serious question I get five replies that are nothing but jokes?! GRRR!

Seriously, I was asking for real -- so where are all the helpful flashaholic tech-geeks to come out and explain it to me?

I thought Photon Wrangler's answer was straighforward. Are we being a bit tetchy because we didn't get what we wanted right this instant?

How could measuring the dimming over time be useful for determining the lifespan of an LED? Would they start by calling the light output on day 1 "100%" and then wait til one year had gone by (if they even waited that long) and if it had dimished in brightness by 10% they'd say, "Oh, well, then, 10% per year, it'll be dead in 10 years"? Wouldn't that mean you'd be expecting 10% light output for that final year?

Hardly. They could measure over a year, observe a 5% drop, which equates to 50% at 10 years, and mark that as the end point.

They may test at some multiple of spec current, and test thousands of units, count the failures, and then derive a meantime between failure (MTBF) figure, rather like the hard drive manufacturers. Those guys are quoting what, now 300,000 hours MTBF?

I'm inclined to thing that the failure rate of LED's is rather high, but that the ones that fail tend to do so right away -- units that last past a certain length of time tend to go on for a loooooong time.

Anyhow, the answer lies less in the testing they do than in the sum result of statistical number crunching and marketing BS.

So I think the answer boils down to applied statistics.

brightnorm · May 26, 2006

What about those notorious "portable" 5 watters that had a 500 or 1000hr life to 50% even when run within spec? A heavy daily user could put on that many hours relatively quickly. And many mods push the heck out of LED's, the LionCub for example.

Personally, I'd like to see figures to 90-95%, not 50%.The jury is still out on actual run times but most knowledgable modders don't seem to be too concerned about current Lux III's.

Brightnorm

peacefuljeffrey · May 26, 2006

I was not complaining about what I would do if my LED module in a given flashlight failed. I was just C-U-R-I-O-U-S about how they ARRIVED at the figure!

I didn't expect to be taunted about the curiousity; I actually did really expect some of the people who really do know their <stuff> here to come and tell me how the developers of the technology have arrived at their life-expectancy figures.

So sue me!

-Jeffrey

peacefuljeffrey · May 26, 2006

Canuke said:
I thought Photon Wrangler's answer was straighforward. Are we being a bit tetchy because we didn't get what we wanted right this instant?

It was hardly authoritative, and even a layman like me could see the potential flaws in figuring that this was the manner in which they test the technology. Not nearly scientific enough -- more like a random, moderately educated guess. It wasn't what I wanted. I wanted someone with actual knowledge of the tests to come forward. Thanks.

-Jeffrey

smknman · May 26, 2006

Before testing, LED is operated for 1000 hours for an initial seasoning. this is required because of rapid changes and LED experiences initialy, they then settle into a gradual decline. this decline is exponential in nature so can be expressed as L=Lo * e^-at where L is current output, Lo is intial output (normalized to 1) and t is time. a is a weight used to more closly match the decline to the exponential model. using this model we can estimate the life of the LED to whatever output % we wish.

Hope that clears it up

FirstDsent · May 26, 2006

PJ,

Please never mistake one of my rants for criticism of the original post. Your post did not have to be provocative to set me off. I'm nothing, if not opinionated. It just prompted me to think critically about Lumileds' claim. This kind of thing is my prime pet peeve! Some people don't believe that the truth sells, so they make up something instead. Take the Democratic party propagandists for instance...

Lumileds' claim sounds fairly arbitrary to me, and I wouldn't be surprised to learn that it is an educated guess. Whatever the case, Lumileds would clearly only give a "lab condition" estimate because real world variables can't be controlled for very accurately in a HUGE projection like that.

Of course you deserve a legitimate answer, You'll just have to wait until tomorrow when the daywalkers are out. The kind of geeks you want have early bedtimes. Someone will give you a good answer, just not me. I may be a flashlight geek, but at least I'm not a semiconductor geek. Those guys are wierd! :sssh:

Bernie

mchlwise · May 26, 2006

HHAAAAAAAAHAHAHAHHAHAHAAhahaaaaa!11!!!!!1!1

:hahaha: :crackup:

/dies

FirstDsent said:
They may as well guess. It's obviously rounded even if they have careful projections. They may as well say 1 million hours. If one of the luxeons in my trafficl lights dims out 15 years from now what am I going to do? Beat-up Lumileds for a warranty replacement?

"Dear Lumileds,
One out of 672 of the Luxeon I emitters in one of our city's 2,280 traffic lights only lasted 79,126 hours. You'll have to take my word for it because it is impossible to document that It was on for 79,126 hours. I assure you that it was well heat sinked, driven within limits, and sufferd no voltage spikes in the 10 years it operated at the intersection of Peachtree Rd, and Peachtree Dr. near Peachtree Plaza in the Peachtree district of Atlanta. It never came in contact with moisture, or excess heat from direct exposure to sunlight to the metal enclosure. On behalf of the City of Atlanta, I demand full compensation! Please send me a new $3.45 emitter".

Soooooo,
Who cares about their silly claim?

For flashlight use, its a worthless number anyway. NOBODY will ever experience a "natural death" of a Luxeon if operated within it's recommended drive limits. If I were building traffic lights I would only give a crap so I can repeat the lie to my customers. They won't come back on me if it only lasts 79,126 hours either. For flashligts, I would buy one if it only lasted 500 hours. If I ever have to put that kind of mileage on a light, it will be after the apocalypse, and some mutant will probably kill me for it first.

Damn mutants!
Bernie

Max Brightness · May 26, 2006

I found this real world testing interesting. (Although somewhat off topic)
http://www.bebi-electronics.com/testing.html

flashlife · May 26, 2006

smknman said:
Before testing, LED is operated for 1000 hours for an initial seasoning. this is required because of rapid changes and LED experiences initialy, they then settle into a gradual decline. this decline is exponential in nature so can be expressed as L=Lo * e^-at where L is current output, Lo is intial output (normalized to 1) and t is time. a is a weight used to more closly match the decline to the exponential model. using this model we can estimate the life of the LED to whatever output % we wish.

Hope that clears it up

What SMKNMAN said above...
In my "real" job, I'm a quality Engineer for a company that makes LEDs and VCSELs (Vertical Cavity Surface Emitting Lasers) for the fiber optic communications industry. Since 80% of the internet travels thru one of our products, reliability is a big deal to us and our customers. Here's how it's done:
When a new device is developed, it's burned-in under various conditions, high temp high current, room temp high current, etc. and based on those results, a "reliability model" may be constructed based on the Arrhenius equation, and the "infant mortality rate", as Smknman mentioned above. That model is then tested by long term burn-in at high temp and current, etc to verify the model.
Based on the verified model, the statistical lifetime may be predicted.

jason9987 · May 26, 2006

LEDs life span is rated for 70 or 80% of output I forget which one, you can read it on the lumileds site.

DCFluX · May 26, 2006

Mean Time Between Failures is calculated using a lot configuration. To arrive at a MTBF of 100,000 hours, 1,000 LEDs are ran for 100 hours. If none fail then the MTBF is rated at that mark. The same is done for fans, hard drives, just about anything that has a MTBF certification on it. Problem is it takes real time to wear something out to the point where you discover a design fault, I have 18 hard drives sitting on my desk all made on the same day that quit within 3 weeks of each other.

How can they possibly know lifespan of LEDs?

Newly Enlightened

Flashaholic

Newly Enlightened

Newly Enlightened

Flashaholic

Enlightened

Newly Enlightened

Enlightened

Flashaholic

Enlightened

Flashaholic

Newly Enlightened

Newly Enlightened

Newly Enlightened

Enlightened

Enlightened

Enlightened

Enlightened

Newly Enlightened

Newly Enlightened

Similar threads