Reliability Prediction

Can you Google, "Reliability prediction"?

A statistician could do you the math. All you need are real data quanity of defect and corresponding hours(RIP-rest in peace). By mere guessing, you could form a normal distribution right there.

Probably, you could use chi.

I don't know, well its a pretty guess, I took stastics about 8 years ago.

You should be able to get more guidance from the LED manufacturer. If the LED you are using doesn't have supporting data from the manufacturer, then I suggest you switch to an LED that does. For example, the LEDs made by Philips-Lumileds are backed by a great deal of supporting documentation such as Application Briefs, Lumens Maintenance guides and Reliability guides. I'm an optics engineer and 've used Lumileds 'Luxeon Rebel' LEDs quite a bit in the past and was able to get, or derive, the info I needed from their data sheets. Their personnel are very helpful, too.

Here's a link:

http://www.philipslumileds.com/support/documentation

The LED fixture life is not dependednt on the life of the LED but the life of the Capacitors on the drive. They fail before the LEDs get a chance to take revenge on you.

Cheap capacitors, lesser half-life, faster degradation, higher sensitivity...

You should try to find the life of the capacitors based on your environmental constraints. The life of capacitors and performance is not linear.

Never compromise on the Capacitors rating for the LED driver... ;)

Take care..

Vishal...

The capacitors are the weakest link but what tests this weakest link is the rest of the LED driver circuitry, the LED themselves, the power grid, and the ambient environmental conditions. A realistic MTBF and cost benefit analysis of a LED system can be easily done but a good analysis is not a trivial task. Then again there's also the tin whisker problem....

Life testing of semiconductors is a mature science and usually uses a number of shorter term tests at high temperatures to extrapolate to 100,000 hours at ?? temperature. We used to rate motors for various classes of service this way.

life test search

how long would they last 5 years? 10 years?

That is totally dependent on how difficult they are to access.

If they are installed in a position that is within easy reach they will last 100 years.

If they are installed in an auditorium with 12metre high ceilings they will fail immediately after the scaffolding has been returned to the hire shop.

They will not last very long if they are both easy to access and available to the public.

I guess u didn't recognise the sarcasm

So very true.

This has already been done.

Review IESNA's LM-79, LM-80, and TM21 documents as well as numerous articles in EDN and other magazines and journals so you are up to date on this subject.

Even better join IESNA and help contribute something of value to the lighting industry, preferably something that has not already been done. An added benefit is that you get a copy of IESNA's Lighting Handbook so you can get yourself up to date on lighting.

I don't mean to sound sarcastic but...

I can do it, but I need to know all the parts that are in the lamp and their specs. Supply voltage and if is stable. That being offered, MTBF (mean time between failures) is just a number. I have seen many times where the numbers and real life are not the same.

The ultimate test is and endurance test where you but 10 or more lamps into the chamber run them until they break and then do a failure analysis to make sure the failure was not a manufacturing problem.

Running the lamps without a guideline will not yield any good data. You would need to set up something similar to the test outlined in Rensselaer Polytech's ACTV Stress Test. Granted the ACTV test is designed for GU24 based CFL's but this is a good starting point for putting together a closer to real world test. A real world test would need to include power cycling and cooldown stages as well as endurance.

Check out the ACTV Stress Test guideline and Final Report:

http://www.lrc.rpi.edu/programs/lightingtransformation/actv/index.asp

Hi MSEELADY,

Thanks!

Well for a start, I have a sample size of 10 units of LED assembly units that have been running with failures up to 1k hour in T&H chamber at 60/90 with bias at 12vDC, So far we have none functional failures, I need to extrapolate what would be the life time be, based on the 1K hours passing. If possible, please let me know the methodology of doing it.

Thanks

accelerated aging uses a heated chamber, and they start at 100C and run until they fail, then repeat at 90C, 80C, 70C and chart this on logarithmic paper.

Different things start at different temperatures.

Sigh. Are you actually trying to reproduce the MTBF measurement of just a DC biased LED that the manufacturer already performs. I recommend that you instead refer to the manufacturer's application notes on how they test their product.

I think he wants a real life test he controls. The Chinese tests are suspect, IMHO.

The reputable makers, like Cree, provide good data - coupled with far higher prices.

But, you get what you pay for or test for.

The only way to screen the Chinese makers is to test samples of each batch forever

Yes u are absolutely right!, but i am not testing chinese led products, Some or if not most of them are known to be notoriously bad. Actually i am using reputable LEDs and its jut that i cant find the appropriate tools to use to predict their lifetime, perhaps i am ignorant. I saw weibull way of prediciting liftime, but then it says u cant do anything unless u get the shape parameter and the characteristic life.

Thanks.

Well LED data is LED data its got nothing to do with the system data. I believe they cannot be equated. While i have assembled it into a system then it will have a different set of reliablity numbers right?

My point exactly. The least reliable parts in most electrical systems are the capacitors in the voltage converting circuitry, followed by any frequently used mechanical parts. The weakness in the capacitor is typically a slow chemical reaction occurring in the dielectric. As such one can use a variation of the Arrhenius reaction rate equation to identify the anticipated lifetime of the system. Many good capacitor manufacturers provide an applet or application note on their products expected lifetime curves. Illinois Capacitor is just one example. What many people forget in their calculation is that no real capacitor is an ideal capacitor. The self heating by a capacitor's ripple current through the parasitic resistance can raise the actual temperature in the capacitor above the ambient temperature.