My research has indicated that failure in LED lighting (incandescent replacements) results from failure in the support components rather than the actual LEDS themselves. How can I build and/or modify my final set of lights such that they will fail after I am dead. (I'm 40 and very healthy.)


closed as off-topic by Bence Kaulics, Voltage Spike, Nick Alexeev Nov 16 '16 at 0:25

This question appears to be off-topic. The users who voted to close gave this specific reason:

If this question can be reworded to fit the rules in the help center, please edit the question.

  • 1
    \$\begingroup\$ The LEDs themselves will last 20,000-50,000 hours. At 8 hours per day that will be between ~7 and ~17 years. So you are either planning on not living very long or spending a lot of your life in the dark. \$\endgroup\$ – Andrew Nov 15 '16 at 12:11
  • \$\begingroup\$ As with the LEDs, support device lifetimes will be specified in the relevant datasheets. \$\endgroup\$ – Leon Heller Nov 15 '16 at 12:13
  • 1
    \$\begingroup\$ That upto 50000 hrs Andrew mentions and what is in the datasheet is what the manufacturer estimates as an average lifetime. There is no guarantee that a LED will last that long. To guarantee anything beyond a 10 - 20 years lifespan is almost impossible from an engineering standpoint. You could run the LEDs at a significantly lower power than what they are designed for but even that will bring you no guarantees whatsoever. \$\endgroup\$ – Bimpelrekkie Nov 15 '16 at 12:27
  • 2
    \$\begingroup\$ @Andrew that'll be the lifetime at rated power. What the OP is probably looking for is the art and science of de-rating to increase lifetime. He won't find superior components but he might make a solution through learning this art. It's not very well explored territory and I agree with FakeMoustache that guarantees are next to impossible. But there's a VERY good question in there if it's phrased a bit differently. \$\endgroup\$ – Brian Drummond Nov 15 '16 at 12:27
  • 3
    \$\begingroup\$ Note that LEDs themselves don't fail like bulbs just turning themselves "off". Instead, their luminous intensity fades gradually, and after the specified 20000 or 50000 hours they are said to have reached the end of their "useful" life, which may mean they merely dropped to below 70% of their rated brightness. \$\endgroup\$ – JimmyB Nov 15 '16 at 12:38

What is a resource to use to find superior support components (in terms of lifespan) so that I can build my last set of lights such that they will fail after I am dead

MIL-HDBK-217f (RELIABILITY PREDICTION OF ELECTRONIC EQUIPMENT) is a good industry standard for estimating the lifespan of electronic components versus various operating conditions and stresses. I use it a couple of times a year and find it very useful for estimating overall MTBF of a circuit.

It's a little bit of a learning curve but worth the effort.

I have also found a calculator that uses the same models as MIL-HDBK-217f here. It only does resistors and capacitors but it might be a way of easing yourself into this sort of calculation. And here appears to be a free downloadable software tool that uses the models of MIL-HDBK-217f and many other models.


MTBF in LED specs is usually hypothetical and not reality based. LM70 is a standard test for device hours for Lumen Maintenance hours of life expected when the Lumen Output declines to 70% of initial value. This can be much greater than 50khr if the design is very well-cooled.

In practice there are many assumptions neglected in the arbitrary number assigned with the same value for 1 LED as 1000 LED's which are not redundant and any 1 of 100 LEDs should reduce the MTBF by 1000. But since the failures are more due to lack of product maturity in material quality, design, process, or handling and become obsolete after 10 yrs, for lack of a better number everyone says 50khr but very few can back it up.

Can LED's last 500khr? Sure but no parts have been around that long to prove it but they must run cool at room temp in a benign environment with very little junction temperature rise. I.e. reduced in power output by at least 50% of suggest levels with more than adequate heatsinking.

How is it verified?

Normally life testing consists of testing large quantities and defining confidence factors ( e.g.) for being defect free based on randomness of failures and have few or zero failures. Then if they define a fix for the problem, it is censored as if it never happened.

That's all fine for random failures but what if the failures aren't random but have some flaw in chemistry, environmental stress, process contamination, or stressers in handling by factory or distributor or end user for things like Mechanical or electrical shock (ESD)

So until all the actual potential causes of failure are known and an established reliability is proven for a given product, process and design, you can only predict what an "Ideal " part with no design or process flaws" will do.

Military Reliability is often improved only by inspection during the process but is actually the same part. This includes environmental stress tests on samples, and wafer and pre-cap X-Ray inspection as well as process controls on ultrasonic gold wire bond pull tests and other such critical factors.

No one can guarantee any LED will last a lifetime, even if some may until the whole supply chain has established a failure rate based on field returns. Unfortunately , people don't return LED's they just toss them so data collection is unlikely to exist.

What can you do? Choose the most reliable source recommended by experts with a product designed by high integrity and design standards for high volume and run it as reduced power levels to reduce the temperature rise from 60 to 100'C to 20'C to 30'C rise.

According to the Laws of Chemistry, IF a product is made from flawless materials and has a flawless design with a flawless process, the life expectancy from material degradation from thermionic emissions is defined by the Arrhenius Law which states in paraphrase that the lifetime will reduce 50% for every 10'C rise above 25'C


Not the answer you're looking for? Browse other questions tagged or ask your own question.