What Happened To The 100,000

Early adopters of LED lighting will remember 50,000 hour or even 100,000 hour lifetime ratings printed on the box. But during a recent trip to the hardware store the longest advertised lifetime I found was 25,000 hours. Others claimed only 7,500 or 15,000 hours. And yes, these are brand-name bulbs from Cree and GE.

So, what happened to those 100,000 hour residential LED bulbs? Were the initial estimates just over-optimistic? Was it all marketing hype? Or, did we not know enough about LED aging to predict the true useful life of a bulb?

I put these questions to the test. Join me after the break for some background on the light bulb cartel from the days of incandescent bulbs (not a joke, a cartel controlled the life of your bulbs), and for the destruction of some modern LED bulbs to see why the lifetimes are clocking in a lot lower than the original wave of LED replacements.

Any discussion of light bulb lifetime would be incomplete without mention of the Phoebus cartel, an international organization formed in 1924 by the world’s leading light bulb manufacturers to manipulate the bulb market. As discussed by Markus Krajewski in “The Great Lightbulb Conspiracy”, the cartel assigned territories to member companies, limited production, and dictated a shortened 1,000 hour bulb life. Previous bulbs had burned for a much longer 1,500 – 2,500 hours. Purportedly imposed to increase quality, efficiency, and light output, the new 1,000 hour limit also resulted in many more bulb sales. Archived documents show that significant research was expended to devise bulbs that lasted their 1,000 appointed hours and no more. It wasn’t only household lighting that took a hit: flashlight bulbs originally lasting for three sets of batteries were reduced to two, with a proposal to limit their lifetime to a single set. Again, brightness increases were touted as the reason. However, that last step, halving bulb lifetime, would increase brightness only between 11%-16%, while doubling sales. This was about selling more bulbs and making more money.

The cartel enforced production quotas and bulb lifetimes with a system of monetary fines, backed by the power of GE’s patent portfolio. Bulbs from each producer were tested, and penalties imposed for bulbs lasting significantly shorter or longer than 1,000 hours. Phoebus continued to exert influence on the market until World War II ended its reign. The cartel is often cited as one of the first instances of planned obsolescence: designing products with an artificially shortened lifespan. A 2010 documentary, “The Light Bulb Conspiracy,” explores the history of the cartel along with some more recent instances of planned obsolescence. I wonder what the conspirators would have thought of bulbs that supposedly last 100,000 hours? Or even 7,500?

Tucked into a lower shelf in the lighting isle at the hardware store, a few lonely incandescent bulbs waited for some Luddite consumer. Picking up a box, I read the rated lifetime: 1,000 hours.

What exactly does the box mean with this 1,000 hour lifetime? This is the bulb’s Average Rated Life (ARL) — it’s the length of time for 50% of an initial sample of bulbs to fail (abbreviated B50). What “failure” means depends on the type of bulb; we’ll explore this in more depth later on. The definition of B50 reveals a common misinterpretation, namely that a bulb will last for its rated lifetime. In reality, only half of them last that long, although this rating doesn’t tell you anything about the distribution of failures around the median lifetime.

Manufacturers use these ARL values to forecast how many years a bulb will last based on using the bulb a specified number of hours per day (typically 3). LED bulbs suffer less wear-out through power cycling than incandescents, so the conversion is just a division: years of service = ARL/(3*365). For example, half of a set of 100,000-hour bulbs would still be in service after 91 years according to this calculation. But this simple metric doesn’t tell the whole story. LED bulb failure mechanisms are complex and fundamentally different from the well-known incandescents. To understand more, we need to shed some light on the inner workings of a bulb.

Before leaving the store, I threw a few bulbs in my cart so I could see firsthand what was inside.

There’s more to an LED bulb than just the LEDs. Outlets in our homes are actually fairly dirty sources of AC power. LEDs want clean, constant-current DC sources, so circuits inside the bulbs must rectify and filter the incoming AC, then limit current to the LED packages. To see how this is done, I dissected three different A19 style bulbs: one each from the GE “Basic” and “Classic” lines (7,500 and 15,000 hours), and a Cree model offering a 25,000 hour life.

This GE bulb has a plastic dome covering a circular aluminum PCB which carries eight LED packages and the driver electronics. The driver consists of an MB10F bridge rectifier, an electrolytic capacitor rated for 105 °C, and an SM2082D linear constant-current driver. There are three resistors on the PCB: one bleeds charge from the capacitor when the bulb is off, and two others set the SM2082D current to 54 mA. In fact, the circuit looks like it was taken directly from the SM2082D datasheet.

Seven of the 3.5 x 2.8 mm LED packages show around 18 V of forward drop when driven with 50 mA, indicating that they contain six LED dice in series. One LED on the board shows a drop of 9 V, so it has only three LED chips. All the LEDs, totaling 45 dice, are wired in series to drop approximately 135V.

When they say classic, they mean it. This bulb is in a glass envelope just like incandescents, and like those old bulbs, the glass is easily removed with a ball-peen hammer. In place of the tungsten filament is an aluminum PCB folded into a squat obelisk. Sixteen 3.5 x 2.8 mm LED packages are connected in series on the board, with each one showing a forward voltage of around 9 V at 50 mA. So, this version has 48 LED chips vs 45 for the Basic bulb, except they’re in twice as many packages – this is good for keeping the LEDs cool.

Another difference with this longer-lived bulb is that the driver electronics are not thermally coupled to the LEDs; they are hidden on a separate PCB in the screw base. This keeps the rest of the components from heating with the LEDs. On the driver PCB is a bridge rectifier, an electrolytic capacitor again rated for 105°C, and an SOIC-8 IC. Interestingly, this bulb also contains a metal-oxide varistor for transient suppression. Although I couldn’t determine what the house-marked (“BYSACT”) driver IC was, the lack of any inductive components on the PCB indicates this is another linear supply.

The Cree bulb has a diffused plastic dome like the GE Basic model. Inside, a larger aluminum PCB holds (16) 3.5 x 2.8 mm LED packages. Each LEDs drops around 8.5 V at 50 mA, so they contain 3 chips; like the GE Classic bulb, this one uses 48 total LED dice. The LEDs are wired as eight sections of two paralleled LEDs, so the total drop is around 68 V. The LED PCB is coupled to a thick aluminum heat sink with silicone thermal compound.

As with the GE Classic bulb, the power supply electronics are on a separate PCB, thermally decoupled from the LEDs. The driver IC is an SOT23-5 package inscrutably marked with “SaAOC”, but the presence of a transformer and stout Schottky diode reveals that this is a switch-mode power supply. The filter capacitor on the switcher output is an aluminum electrolytic rated for 130 °C.

It’s not much to go on, but what conclusions can we draw from the design of these three bulbs? It helps to consider how they typically fail, and what factors affect their lifetime.

Since the LED bulbs contain a number of parts, it’s natural to ask which ones might be responsible for failures. The US Department of Energy (DoE)’s solid-state lighting program supports research and development of LED technologies, and their website contains volumes of data on LED lighting systems. Their Lifetime and Reliability Fact Sheet contains data on the failure rate of 5,400 outdoor lamps over 34 million hours of operation. Interestingly, the LEDs themselves account for only 10% of the failures; driver circuitry, on the other hand, was responsible almost 60% of the time. The remainder of failures were due to housing problems, which may not be as applicable for bulbs in indoor use. This data shows that at least for catastrophic failures (where the lamp ceases to emit light), extending lifetime means improving the power supplies.

The lifetime of a bulb (or power supply) can be no longer than the lifetime of any of its components. Among the components found inside the bulbs, two stand out as life-limiters: the semiconductors and the electrolytic capacitors. Both of these components suffer from a failure rate that is a strong function of temperature. The typical model for this effect, based on the Arrhenius equation, predicts a doubling of lifetime for each 10 degree Celsius decrease in temperature, at least over a limited range.

The two longer-lived bulbs use twice as many packages to carry approximately the same number of LED dice as the GE Basic lamp, decreasing thermal resistance to their respective heatsinks, and presumably reducing their temperature. These bulbs also both mount the failure-prone driver electronics on separate PCBs from the LEDs to keep them cool. Finally, the 25,000-hour Cree bulb uses an electrolytic capacitor rated for 130 °C as opposed to the 105 °C caps in the other two. For similar operating temperatures, this could multiply the expected life of the capacitor by a factor of five. Each of these measures probably contributes to delaying catastrophic failure of the bulb, resulting in the longer rated lifetimes.

But when it comes to the LEDs themselves, there is more to lifetime estimates than predicting catastrophic failure.

Like the soldiers in Douglas MacArthur’s famous line, old LEDs don’t die, they just fade away. We all know what an incandescent lamp failure looks like: one second it’s burning bright; the next, it’s not (and every once-in-a-while, you hear a pop followed by a faint jingling as the liberated filament richochets inside the bulb). Power supplies aside, LEDs typically don’t fail with so much fanfare. Instead, they gradually lose brightness as they age. In the lighting industry, this is known as lumen depreciation, and is a separate failure mode from the catastrophic failure we usually think about.

As it turns out, lumen depreciation happens to incandescent bulbs, too. By the end of their 1,000 hour life, the output has typically dropped 10-15%, but nobody ever notices. With LEDs, the effect is much worse, and the output continues to fall as the device ages. At some point, the LED is no longer producing enough light to fulfill its original purpose, even though it hasn’t “burned out.” Research says that most users won’t notice a gradual 30% drop in light levels; accordingly the industry has defined L70, the time at which the output has dropped to 70% of its initial level, as an endpoint for measuring LED bulb lifetime. Based on how it’s estimated, this measure is typically stated as B50-L70, the point at which 50% of an initial sample of bulbs will retain 70% of their rated output.

Something else happens as phosphor-based white LEDs age: they change color. The US DoE’s report on LED Luminaire Reliability: Impact of Color Shift defines four color-shifts (blue, yellow, red, and green) observed in LED lamps, although the yellow shift dominates in high-power white LEDs. This gradual yellowing of the light output results from phosphor cracking, delamination, and thermal effects, since the phosphor temperature can exceed that of the LED junction by 30 C – 50 °C. Modeling and predicting color shift in LEDs is a difficult task, with all of the mechanisms not yet fully understood. As a result, no standards have yet been established for accelerated testing or projection of color stability over time.

Eventually, these effects can be as detrimental to the function of the bulb as catastrophic failure. Given that lumen depreciation and color shift will in time render the LEDs ineffective, it may not make sense for manufacturers to design bulbs with very long electrical lifetimes. It’s possible that the reduced lifetime ratings we see on current bulbs simply reflect better knowledge about actual performance of existing LED technology over time.

I’ve seen lumen depreciation and color shift first-hand. In June of 2010, I replaced twelve 65W incandescent PAR30 floodlight bulbs in our kitchen with LED equivalents. At the same time, I also replaced three lights in another room with identical LED bulbs. These three bulbs see much less use, so in preparation for this article, I took one bulb from each location and put them side-by-side to see if I could tell the difference in output. The recessed light fixtures in both rooms are identical, so I expect that the bulbs are exposed to similar temperatures when on: any difference should only be due to aging effects. The results were shocking. Since these two bulbs were in different rooms, I never saw them side-by-side, so didn’t notice how bad the lumen depreciation and color shift had become. Sure, I knew they were dimmer and yellower than when I installed them, but had no idea it was this bad.

These bulbs were advertised with a 30,000 hour lifetime. I estimate the total use at 15,000-20,000 hours. During the 8 ½ years these were in service, one failed completely. Instead of replacing it with a newer bulb which would not match the color of the older ones (or replacing them all), I left that socket empty.

In the hardware store, I noticed new 9-watt BR30 LED bulbs for $5 each. The PAR30s I purchased in 2010 were $45 and consume 11 watts. A quick calculation says that the old bulbs paid for themselves more than three times over in electricity savings relative to the incandescents they replaced, and put that much less carbon into the atmosphere. They may well continue to burn for another 15,000 hours, but after weighing the degraded output and the cost to replace them with brighter, more efficient versions, I’m headed back to the store.

I’ve taken a look at some of the technical issues in LED lighting. Of course, there is more to LED bulbs than lifetime — color temperature and color rendering index (CRI) should factor into any purchase decision. There are also a number of larger problems involved, including issues of economics and sustainability. Some of these are addressed in J.B. MacKinnon’s 2016 article, The L.E.D. Quandary: Why There’s No Such Thing as “Built to Last”, in The New Yorker.

Certainly moving away from incandescent bulbs to more efficient lighting makes sense, but maybe we never really needed 100,000 hour bulbs in the first place. The lifetime of even 7,500-hour bulbs is long compared to the rapid pace of advance in lighting technology. Does it makes sense to buy expensive long-lived bulbs today, when better, cheaper, more efficient ones may be available in the near future?

The oldest surviving incandescent light, known as the Centennial Bulb (click to see a webcam of the lamp), is a dim carbon-filament bulb that’s been burning nearly continuously since 1901 — over 1 million hours. In its current state, it throws off as much light as a modern 4-watt incandescent. Would it have made sense to pay a premium for such “million hour bulbs” at the turn of the 20th century if we had any inkling of the advances that would come in the next 117 years?

The new $5 BR30 LED bulbs I just installed in the kitchen are amazingly bright and crisp: tests with a lux meter show the illuminance is more than 60% higher. Plus, they’ll more than pay for themselves in electricity savings compared to the old, inefficient LED bulbs they replaced.