Because the people that make them lie.
Not ‘obvious’ lying where lightbulb manufacturers calculate one reliability number and deliberately choose to put another number on their lightbulb packaging. But ‘insidious’ lying where they deliberately choose to measure the wrong thing to get a better number.
Here is a direct quote from a lighting reseller …
"While incandescent light bulbs were built to last around 1,000 hours, the most enduring LED light fixtures have been tested to last as long as 100,000 hours. On average LED light bulbs will not have to be changed for at least 20 years."
Just to clarify – 100,000 hours is over 11 years. And just to clarify more – no LED lightbulb manufacturer has ever tested any of their LED lightbulbs for more than 11 years of continuous use. So that would be another ‘lie.’ And anyone who uses LED lightbulbs knows that none of them ever last 20 years, even if they aren’t used constantly.
So how do they ‘lie’ about LED lightbulb reliability so easily and effortlessly?
Let’s think about human ‘reliability’
Let’s say that one day, we woke up with no death records, no birth records, no recollection of how old people are when they die … and we want to work out human ‘average life expectancy.’ One approach is to run a ‘life test.’ We could get a large number of human test subjects (say 1 million), observe them for one month, and ‘measure’ how many die.
If we randomly select humans, of all ages, of all races, of all genders (et cetera), we might observe 1,094 people die in our one-month test duration (which is around what we expect in reality). If we simply divide the combined life duration of our test (1 million multiplied by 1 month) by the number of deaths … we would get an average life expectancy of 76 years. Which is about right.
But what if we did the test again, and instead of randomly selecting humans of all ages, we chose 1 million young adults who are as fit, healthy and as ‘undamaged’ as humans ever get. And it turns out that 78 of these 1 million fit, healthy and ‘undamaged’ young adults die during our one-month test duration (which is also around what we would expect in real life). This now gives us an average life expectancy of 1,068 years.
The latter approach is (of course) a crazy way to find human life expectancy. And that is because humans wear out. We accumulate damage as we age and become less likely to survive the next day. Our fit, healthy and ‘undamaged’ young adults represent a really short period of the human life experience (unfortunately).
So... what has this got to do with LED Lightbulbs?
Electronic component manufacturers ‘measure’ LED lightbulb reliability in the same, crazy way.
Instead of testing components of all ages and types, they simply test their components when they are brand new, in ‘perfect’ test conditions, and for really short durations.
It is not uncommon for diodes (and other electronic components) to be advertised with a failure rate of ‘100 FITs.’
What is a ‘FIT’? ‘FIT’ is an acronym for ‘failure in time.’ In the electronic component industry, the standard time unit is 1 billion hours. So for example, our diode with an advertised failure rate of ‘100 FITs’ should generate 100 failures if it was used for 1 billion hours. Or have an average life of 1,140 years.
Thermionic (vacuum-tube) diodes were first developed in 1907, which means that diodes (as a concept) have been around for a little over 100 years. Today’s diodes are made in a very different way and have been around for even less time.
It is amazingly crazy to suggest that diodes, on average, can last for around 10 times longer than their entire technological history (the length of time since they were first invented.) And there are some electronic components that are advertised with failure rates as low as 10 ‘FITs’, meaning that manufacturers are telling us to expect them to last for 11,400 years.
But LED lights don’t just have LEDs...
LEDs emit light when they have direct current passing through them. And unfortunately, the mains power we use to power them uses alternating current (changing direction around 50 times per second). So a LED light bulb actually includes circuitry and other components to convert alternating current to direct current. And all these components have their reliability measured in the same, crazy way that the electronic component industry uses to measure reliability.
And as a rule, around 50 percent of failures occur at interfaces. This means connectors, solder joints, contacts and so on. LED lightbulb manufacturers pretend these don’t happen as well.
LEDs wear out... just like humans.
All electronic components do. Failure mechanisms such as diffusion, dendritic growth, tin whiskers and oxidation mean that electronic components accumulate damage as they are used. So an older LED is more likely to fail today than a younger one.
Which is why it is crazy to test them when they are young (before they wear out) and then think that this characterizes their entire life. It’s as crazy as assuming a 90-year-old person is just as likely to die today as a 25-year-old person.
But it gets worse.
LEDs can wear out faster. An increase in temperature, voltage or humidity can increase the rate at which damage accumulates. Mains power in particular can fluctuate above what it is supposed to.
But when our manufacturers test LEDs (and other electronic components), they deliberately choose ‘perfect’ conditions with ‘comfortable’ temperatures and ‘quality’ electricity.
None of this affects traditional, incandescent light bulbs
These light bulbs generate light from passing electricity (doesn’t matter if its direct or alternating current) through a tungsten filament until it glows white hot. There are no electronic components (filaments often aren’t seen as electronics) meaning they are not subjected to the same reliability testing that LED lights are. And for whatever reason, manufacturers don’t pretend that filaments do not wear out.
And because it doesn’t matter if the electricity uses direct or alternating current, there is no circuitry needed to modify it before it reaches the filament. So it is very simple.
And temperature doesn’t really affect it. That’s because the whole idea is to heat the filament to around 4,500 ° F or 2,500 ° C. So any increase in ambient temperature will have no impact on how fast the filaments wear out.
So, LED lightbulbs are trash... right?
NO! They are great. They still use a lot less energy to produce the same amount of light as an incandescent bulb. And crazily enough, some of the very first LED lightbulbs we manufactured in very robust ways with large (ugly) heatsinks, meaning some still work today. But since then, manufacturing quality has decreased, meaning that heat isn’t being dissipated away from the electronic components, solder joints crack, and smaller components wear out faster.
But they are a great example of products where we know an ENTIRE INDUSTRY lies about how reliable they are. And if this happens for relatively simple LED lightbulbs … how do you know that the producer of (for example) a military main battle tank isn’t also testing the wrong thing to come up with better numbers? You don’t … and they do.
So if you are a ‘customer’ that needs your system to be highly reliable, you better not just take your supplier’s word on reliability. You certainly can if you have a long history of working with your supplier in an open and transparent way, building trust and working diligently over the years to jointly resolve problems that may arise.
But if you are browsing products and simply accepting advertised failure rates as gospel, then you are just as crazy as anyone else who believes that they will not have to replace their LED lightbulbs for 20 years.
Crazy.
Comments