It seems weird that Thomas Pynchon's name is nowhere to be found on this page or any other context about the source of this story (its from Gravity's Rainbow). There are also spelling errors.
As it's not directly explained on the link, I'll ask here... How/Why is it still burning?
I have light bulbs in my house that blow after only a few months use, and normally only blow when I turn them on, if this bulb was power cycled, would it blow?
Modern lightbulbs use thinner filaments that are substantially more efficient than the thick one used in this bulb, but they don't last as long. Further, this particular bulb never gets turn off and back on again which I'm led to believe causes wear.
If it were turned off and back on it would likely be fine.
Always a fun segue into the Phoebus Cartel [1] which pushed for shorter lifespans for bulbs - including by mandating thinner filaments. They did it to keep people on the treadmill but it was actually a substantial improvement in energy efficiency.
I also do think that the filament thickness has a lot to say for longevity. I have a "decor type" large warm glowing light bulb with a quite thick and long filament that's been on for hours daily and turned on and off multiple times daily for 10 years now and counting. And yes, it is a real incandescent bulb, not LED :)
Incandescent light bulbs have a trade-off between brightness and longevity. This bulb specifically appears to be running at a fraction of its original designed brightness.
The carbon filament has gotten thin enough that it runs at a lower temperature and doesn't draw much current (or produce as many lumens), so it doesn't wear out as quickly.
The bulb was accidentally off for nine hours in May 2013 due to a faulty UPS - originally installed to ensure the bulb would not go out in the case of a power failure - but fortunately it did not blow when power was restored.
Besides the other great answers, there's also survivorship bias. This bulb wasn't necessarily designed to last a hundred years. It's just an outlier that has lasted that long. We wouldn't give a shit about it had it burned out fifty years ago. We don't talk about other bulbs (today) that lasted fifty years even though that's an extremely impressive number itself. If the Internet was around fifty years ago we might have half a dozen webcams pointing to the "Half Century Bulbs" and being fascinated with them lasting fifty years and wondering if any will get to a hundred years.
"Survivorship bias" is a favourite go-to for contrarians, but it's not a strong effect and when you see something surprising way out at the tail end of a distribution, attributing it to survivorship bias is a lost opportunity to learn something important. When something survives 2x longer than expected, it might be survivorship bias. When it survives 10x, 100x, or 1000x longer than expected, you should probably look closer and see what material or operational differences might be responsible, because there's going to be something worth learning.
Even if what you learn is just that any incandescent lightbulb can have a greatly extended life if you run it at below-spec brightness.
When you have an outlier like this bulb you can't make any good statistical judgements about it unless it was part of a controlled experiment.
This bulb is a sample size of one. We might roughly know some details about its original construction and early operational life but not enough to really draw conclusions from. The only way to get more details about it will be a postmortem examination of its physical qualities.
Due to its age we have little knowledge of its specific construction details. Were the factory specs followed exactly (or even recorded) or was some of the fabrication machinery tuned by the operator? Did this bulb end up with thicker than normal glass or filament? What was the mortality rate of other bulbs in its batch? Does the fire station have better (or worse) than typical wiring? Did it have bad wiring early in the bulb's life to unintentionally under-load it?
Sometimes extreme outliers are just extreme outliers. A postmortem examination of the bulb might tell you it had a manufacturing defect giving it thicker than typical filament that combined with babying it for the past 50 years extended its life.
Sorry it's not just being contrarian. Outliers are outliers. Just because something is an outlier in a distribution doesn't automatically mean there's some hidden truth of the universe to be revealed.
> When something survives 2x longer than expected, it might be survivorship bias. When it survives 10x, 100x, or 1000x longer than expected, you should probably look closer
You're implicitly assuming a normal distribution or something like it, but it's easy for a lognormal distribution to span multiple orders of magnitude like that, and the Weibull distribution commonly used in reliability modeling for this kind of thing can also be heavy-tailed if k > 1.
A simple physical example of how this could arise would be if the filament diameter of a batch of bulbs was normally distributed spanning a factor of 2, resulting in filament resistances spanning a factor of 4, resulting in filament powers spanning that same factor of 4, resulting in filament powers per unit area of filament surface spanning a factor of 2 (since the thickest filaments have the lowest resistance and thus the highest power at a fixed voltage; it'd be 8x instead on a constant-current source), resulting in filament Stefan-Boltzmann temperatures varying by about 19% (1.19 is the fourth root of 2), which works out to a temperature difference of about 200 K for filament temperatures of about 1000 K. That's probably (handwaving here) enough temperature difference for about an order of magnitude difference in vapor pressure, which (handwaving wildly at this point) might mean an order of magnitude difference in filament evaporation rate.
Also, typically there are hotspots in a filament where it's thinner than the rest, which evaporate more rapidly because they're hotter, which makes them get thinner faster, making them even hotter. So the dominating factor in filament lifetime (at a given voltage) might not be how thick it is but how smooth it is.
So, you might find out something interesting. Or you might just find that this bulb had an unusually thick or smooth filament.
> You're implicitly assuming a normal distribution
I'm not really making that assumption. Instead, I'm pointing out that learning this particular bulb survived due to "unusually think or smooth filament compared with other bulbs" is indeed something interesting you can learn by inspecting the survivor, and that might be something you can apply for process quality improvement.
Basically, I'm advocating the point of view that if you are trying to improve lightbulb, or say, ball bearing quality (rather than just model it), then it's useful to look beyond the fact that the Weibull distribution is long-tailed. A wide distribution is a reflection of ignorance and thus a learning opportunity; the long tails are still caused by something. Yes, from a given batch of ball bearings, some might last 10x or 100x longer than others under the same loading conditions. But if you put those super-long-lasting bearings under a microscope, you might observe real, important differences compared with others from the same batch. The balls might all be rounder or more identically-sized compared with their cousins. (You can do the inspection before the lifetime test if you want to be extra fair).
But it's even more important when you don't expect a long-tailed distribution to begin with; for example, if you have a linear damage model. In this case, if you control for filament diameter, resistance, or brightness, and still see that one bulb is a mysterious outlier, you might want to inspect the joints or run the gas in the bulb through an analyzer to see if there are any other surprises.
The flip-side of it being the world's longest lasting light-bulb is that every other light-bulb made alongside it has burnt out by now.
Its longevity is most likely due to a unique confluence of factors, not simply because of a robust design produced before the concept of planned obsolesce, as is often alluded to whenever the subject is brought up. I also suspect its longevity is helped by all the delicate care and attention it has been given since people realized that it has been running for so long. It has a dedicated power supply, it is under-powered, and great care has been taken not to damage it. Normal lightbulbs never get such a luxury.
It is certainly possible a newer lightbulb has already been produced that will burn for twice as long as the Centennial Light. We just won't live long enough to see it.
As I understand it the issue with ordinary incandescents is that the alternating heat and cold from being used eventually causes the glue attaching the glass to the metal part to crack, which lets air into the bulb (ordinarily a vacuum/inert gas). The next time you turn it on, the air causes the filament to burn, causing the characteristic flare-up and scorch.
Er, that glue only holds the glass bulb in the socket. The atmosphere contained within is completely glass sealed.
What happens is the filament sputters away (like evaporation) and becomes thinner over time. It eventually becomes so thin it breaks from thermal shock when turned on. The dark spot on dead bulbs in the evaporated tungsten.
Fun fact: In theater a blown bulb can ruin a production so some of the more critical lamps are dimmed instead of power cycled to avoid blowing a lamp.
Really? I was under the impression that the glass isn't a fully enclosed "bulb" so to speak, but is cut off at one end to accommodate the rest of the necessary hardware, and glue completes the seal. Perhaps you're right. I can't seem to find a diagram or anything that shows this, though.
If you examine a clear incandescent bulb, you will see that the support for the filament is hollow and has a tube inside extending down. This little "straw" is the point of connection for evacuation where the air is removed and then back-filled with argon. The argon coaxes the tungsten back onto the filament preventing rapid sputtering which would happen in a pure vacuum. After the argon the little tube is heated until it hits melting point and crimped off for a near perfect hermetic seal. The argon pressure is still well below atmosphere so the bulb is still considered under "vacuum" compared to atmosphere.
The leads are inserted through a glob of hot glass, along with a glass tube. The bulb's open base is heated to make it soft, the blob containing the leads with attached filament are inserted, then the base is welded to the lead blob. Last the tube is used to evacuate the bulb and/or inject inert gasses. While under this vacuum the tube is melted and smashed closed, permanently sealing the bulb. Attachment of the base comes later.
The simple reason is that it's not a light bulb, it's a heater.
They are running it at much lower than specified voltage, so are generating mostly heat, and almost no light, so the filament is not very hot. It's glowing a dull red, not white.
There is a quadratic relationship between temperature; and efficiency plus bulb life.
You may have heard of the phoebus cartel - they specified a specific bulb life, which directly translates into a particular energy efficiency. There's nothing about the manufacture of the bulb that they casued to be worse, rather if the bulb lasts too long, you just up the temperature and make a more efficient bulb (uses less electricity).
You can buy stage bulbs that last barely 10 hours, and run extremely hot, and also efficient, and for film that is the tradeoff they prefer (I guess they need a lot of light, and they don't really keep bulbs for long).
Every time you increase temperature by 10 degree, it shorten the lifespan of electric device by 50%. If you run device that was meant to run at 4600°C at 600°C, well you do the math...
It also now has an efficiency of 0.05 lm/watt, 200 times less than a standard "planned obsolescence" tungsten filament light bulb and 3000 times less than a modern LED bulb.
Planned obsolescence is a thing, but it is often a trade-off, not just companies being evil. By making things less durable, you can also make them cheaper, lighter, more efficient,... You also have to take technological progress into account. Your old dishwasher may last for 50+ years but it is heavy, loud, wastes water and electricity... So you may prefer a cheaper dishwasher that only lasts 10 years because you may want to change it later anyways because the energy savings of a new one can make keeping the old one not worth it.
There are cases of planned obsolescence that are, in my opinion, inexcusable (looking at you smartphone manufacturers) but lightbulb longevity is a worthwhile trade-off.
The Livermore bulb only has a small part, but this reminded me of the story 17776: What Football Will Look Like in the Future [0]. It's a fun science fiction story that is also a great example of web storytelling.
17776 is the most important speculative fiction that I've read in the past decade. Don't be turned off by the eye-melting lo-fi hypermedia aesthetic, or the fact that the framing narrative is ostensibly about football, hosted on a sports website, and written by a sportswriter (all of whose work is brilliant, BTW, even if you don't follow sports). As absurdly silly as the premise is, it's quietly a thought-provoking treatment on how humans adapt to post-scarcity and utopia, and changed how I view the human relationship to modern video games.
I have no idea what I just experienced. My PC fan spooled up to max speed when I clicked on the link and everything went all 2001 Entering The Monolith on me. I panicked, but I stuck with it.
Wow !! This is amazing. My last comment on HN was 5 days ago when I threw a bottle to the sea for a website I saw years ago : a story told in a calendar.
And there it is !!! I find it again here thanks to you and your comment. (I was not much interested in this story so I was about to leave and I just had a quick look at the first comment : yours) This is an incredible luck ! Thank you very much !
36 comments
[ 4.6 ms ] story [ 87.6 ms ] threadI have light bulbs in my house that blow after only a few months use, and normally only blow when I turn them on, if this bulb was power cycled, would it blow?
If it were turned off and back on it would likely be fine.
Always a fun segue into the Phoebus Cartel [1] which pushed for shorter lifespans for bulbs - including by mandating thinner filaments. They did it to keep people on the treadmill but it was actually a substantial improvement in energy efficiency.
[1] https://en.wikipedia.org/wiki/Phoebus_cartel
https://www.mercurynews.com/2011/02/03/tests-shine-light-on-...
Even if what you learn is just that any incandescent lightbulb can have a greatly extended life if you run it at below-spec brightness.
This bulb is a sample size of one. We might roughly know some details about its original construction and early operational life but not enough to really draw conclusions from. The only way to get more details about it will be a postmortem examination of its physical qualities.
Due to its age we have little knowledge of its specific construction details. Were the factory specs followed exactly (or even recorded) or was some of the fabrication machinery tuned by the operator? Did this bulb end up with thicker than normal glass or filament? What was the mortality rate of other bulbs in its batch? Does the fire station have better (or worse) than typical wiring? Did it have bad wiring early in the bulb's life to unintentionally under-load it?
Sometimes extreme outliers are just extreme outliers. A postmortem examination of the bulb might tell you it had a manufacturing defect giving it thicker than typical filament that combined with babying it for the past 50 years extended its life.
Sorry it's not just being contrarian. Outliers are outliers. Just because something is an outlier in a distribution doesn't automatically mean there's some hidden truth of the universe to be revealed.
You're implicitly assuming a normal distribution or something like it, but it's easy for a lognormal distribution to span multiple orders of magnitude like that, and the Weibull distribution commonly used in reliability modeling for this kind of thing can also be heavy-tailed if k > 1.
A simple physical example of how this could arise would be if the filament diameter of a batch of bulbs was normally distributed spanning a factor of 2, resulting in filament resistances spanning a factor of 4, resulting in filament powers spanning that same factor of 4, resulting in filament powers per unit area of filament surface spanning a factor of 2 (since the thickest filaments have the lowest resistance and thus the highest power at a fixed voltage; it'd be 8x instead on a constant-current source), resulting in filament Stefan-Boltzmann temperatures varying by about 19% (1.19 is the fourth root of 2), which works out to a temperature difference of about 200 K for filament temperatures of about 1000 K. That's probably (handwaving here) enough temperature difference for about an order of magnitude difference in vapor pressure, which (handwaving wildly at this point) might mean an order of magnitude difference in filament evaporation rate.
Also, typically there are hotspots in a filament where it's thinner than the rest, which evaporate more rapidly because they're hotter, which makes them get thinner faster, making them even hotter. So the dominating factor in filament lifetime (at a given voltage) might not be how thick it is but how smooth it is.
So, you might find out something interesting. Or you might just find that this bulb had an unusually thick or smooth filament.
(Carbon has 1000x higher vapor pressure than tungsten, though, which is relevant here: https://www.powerstream.com/vapor-pressure.htm.)
I'm not really making that assumption. Instead, I'm pointing out that learning this particular bulb survived due to "unusually think or smooth filament compared with other bulbs" is indeed something interesting you can learn by inspecting the survivor, and that might be something you can apply for process quality improvement.
Basically, I'm advocating the point of view that if you are trying to improve lightbulb, or say, ball bearing quality (rather than just model it), then it's useful to look beyond the fact that the Weibull distribution is long-tailed. A wide distribution is a reflection of ignorance and thus a learning opportunity; the long tails are still caused by something. Yes, from a given batch of ball bearings, some might last 10x or 100x longer than others under the same loading conditions. But if you put those super-long-lasting bearings under a microscope, you might observe real, important differences compared with others from the same batch. The balls might all be rounder or more identically-sized compared with their cousins. (You can do the inspection before the lifetime test if you want to be extra fair).
But it's even more important when you don't expect a long-tailed distribution to begin with; for example, if you have a linear damage model. In this case, if you control for filament diameter, resistance, or brightness, and still see that one bulb is a mysterious outlier, you might want to inspect the joints or run the gas in the bulb through an analyzer to see if there are any other surprises.
Its longevity is most likely due to a unique confluence of factors, not simply because of a robust design produced before the concept of planned obsolesce, as is often alluded to whenever the subject is brought up. I also suspect its longevity is helped by all the delicate care and attention it has been given since people realized that it has been running for so long. It has a dedicated power supply, it is under-powered, and great care has been taken not to damage it. Normal lightbulbs never get such a luxury.
It is certainly possible a newer lightbulb has already been produced that will burn for twice as long as the Centennial Light. We just won't live long enough to see it.
What happens is the filament sputters away (like evaporation) and becomes thinner over time. It eventually becomes so thin it breaks from thermal shock when turned on. The dark spot on dead bulbs in the evaporated tungsten.
Fun fact: In theater a blown bulb can ruin a production so some of the more critical lamps are dimmed instead of power cycled to avoid blowing a lamp.
They are running it at much lower than specified voltage, so are generating mostly heat, and almost no light, so the filament is not very hot. It's glowing a dull red, not white.
There is a quadratic relationship between temperature; and efficiency plus bulb life.
You may have heard of the phoebus cartel - they specified a specific bulb life, which directly translates into a particular energy efficiency. There's nothing about the manufacture of the bulb that they casued to be worse, rather if the bulb lasts too long, you just up the temperature and make a more efficient bulb (uses less electricity).
You can buy stage bulbs that last barely 10 hours, and run extremely hot, and also efficient, and for film that is the tradeoff they prefer (I guess they need a lot of light, and they don't really keep bulbs for long).
https://en.wikipedia.org/wiki/Oxford_Electric_Bell
I used to work in the same building as it. You can't actually hear it ringing, it's extremely quiet anyway and it's behind two layers of glass.
I made a video of it ringing way back in 2011: https://youtu.be/1Dx1-f8xQio
Planned obsolescence is a thing, but it is often a trade-off, not just companies being evil. By making things less durable, you can also make them cheaper, lighter, more efficient,... You also have to take technological progress into account. Your old dishwasher may last for 50+ years but it is heavy, loud, wastes water and electricity... So you may prefer a cheaper dishwasher that only lasts 10 years because you may want to change it later anyways because the energy savings of a new one can make keeping the old one not worth it.
There are cases of planned obsolescence that are, in my opinion, inexcusable (looking at you smartphone manufacturers) but lightbulb longevity is a worthwhile trade-off.
[0] https://www.sbnation.com/a/17776-football
There's also (the first half of) a sequel, 20020: https://www.sbnation.com/secret-base/21410129/20020/
I have no idea what I just experienced. My PC fan spooled up to max speed when I clicked on the link and everything went all 2001 Entering The Monolith on me. I panicked, but I stuck with it.
https://news.ycombinator.com/item?id=29009014
And there it is !!! I find it again here thanks to you and your comment. (I was not much interested in this story so I was about to leave and I just had a quick look at the first comment : yours) This is an incredible luck ! Thank you very much !
Centennial Light - https://news.ycombinator.com/item?id=20427081 - July 2019 (20 comments)
This Light Bulb has been burning for 110 Years straight - https://news.ycombinator.com/item?id=8749576 - Dec 2014 (1 comment)
The Centennial Light, a bulb that's reportedly been burning for 113 years - https://news.ycombinator.com/item?id=8353200 - Sept 2014 (29 comments)
The lightbulb that's been burning for almost 110 years - https://news.ycombinator.com/item?id=2642483 - June 2011 (9 comments)