Jeff Duntemann's Contrapositive Diary Rotating Header Image

Let There Be (Long-Lived) Light!

A recent story on the 113-year old light bulb reminded me that I needed to say something about light bulbs here, as they’ve been a long-running low-level project of mine that’s been so low-level that I keep forgetting to post a report. Money quote: LED bulbs are (finally) ready for prime time. It took awhile, but we’re there. Furthermore, there’s upside in LED technology that should make all things LED-ish even better in five or ten years.

Like a lot of people, I stocked up on incandescents when the Feds outlawed them. I did so because my experience with alternative lightning technologies has been hideous. I was curious about CFLs, and I tried them once they became commonplace. If there are light bulbs in Hell, man, they will be CFLs. Their light quality can only be described as sepulchral. They are never as bright as the package says they are. They don’t reach peak brightness immediately, and sometimes take several minutes to get there. (Good luck trying to pee in the middle of a cold night in a one-CFL powder room.) They have mercury in them (granted, not much) which is released into the environment when they break. Oh, and they remind me of spirochetes or intestinal worms.

Fortunately, they die quickly. I recently replaced a couple that were less than a year old. Some have died in a matter of months. I have incandescents in this house that were installed during construction in 2003 and are still in service. Why some bulbs last so much longer than others has always puzzled me. The Phoebus Cartel was real, and it’s not beyond imagination that keeping the tungsten thin for ostensible cost reasons could cover for deliberately limiting the bulb’s life. Still, this doesn’t explain why I have 11-year-old bulbs in some places, and bulbs that repeatedly die in a couple of months in others.

I have theories. One is that some sockets have center contacts that aren’t quite close enough to the bulb to make a firm connection when the bulb is screwed in. Nothing kills a bulb faster than rattly intermittents in the fixture, especially if thermal expansion and contraction of parts in the fixture cause the intermittents. (This is why bulbs shouldn’t be installed with the power on. The moment when the bulb touches the center contact during screw-in is not one moment, but several.) I have also observed that the bulbs that die quickly tend to be mounted either horizontally or at some odd angle, as in my great room ceiling fixtures that are fifteen feet off the floor. The long-lifers are nearly all mounted vertically, bulb-down. I can see how that might work: The filaments of vertical bulbs experience the same gravity load no matter how far they screw into the fixture. Horizontal or angled bulbs will place their filaments in different gravity load situations depending on the angular position of the bulb in the socket, which in turn depends on the manufacturing details of both the socket and the bulb.

Those atrocious CFLs made me cautious. I bought my first LED bulb only about six months ago, having watched them converge on incandescents in terms of spectral signature for some time. That first one was kind of blue, and it’s now in the pantry ceiling fixture where color doesn’t much matter. We’ve been buying Cree TW (True White) bulbs for a couple of months, and they are so close to 60W incandescents that I’ll be ready to install them in critical places (like the master bathroom over-the-sink fixtures) once the incandescents are gone. The Cree bulbs evidently use neodymium-doped glass to add a notch filter to get the spectral signature closer to incandescents. The sweet spot for color seems to be 2700K, and if you want a swap-in for those evil outlawed cheap light bulbs, 2700K is the number to look for. The lerss expensive Feit LED bulbs are bluer, even at the same Kelvin rating, and serve well in places like the laundry room ceiling.

I’ve just started replacing those angled 65W ceiling floods in our great room vault with 650-lumen Duracell Procell BR30s. They’re just a hair brighter, and at 3000K a hair whiter, than the generic incandescent floods we’ve used for ten years now. Replacing angled bulbs fifteen feet off the floor with a sucker pole is a royal nuisance, so even though the Duracells are $20 each, they use a fraction of the power and supposedly live forever.


The clock’s ticking. I’m skeptical. Yes, Phoebus was real. But in the meantime, you really can get 65 watts’ worth of instant-on light with 10 watts’ worth of electricity, in a color that doesn’t resemble a zombie’s complexion. If any of them die on me, you’ll definitely hear about it.


  1. Tony says:

    The problem with all these new bulbs is for the visually impaired. Also the LED bulbs I’ve seen don’t fit in the fixtures for our ceiling fans which we just bought and installed in 2005 (late, or early 2006).

    My wife needs LOTS of light. CFLs did not work for that. Have not tried a lot of LED lights but if it means having to wrestle with the ceiling fans in 5 rooms… wait the one in the basement needs replacing so 6 ceiling fans I’ll wait until I run out of bulbs.

    1. One reason we chose 3000K instead of 2700K floods for the great room is that Carol reads in there a lot, and brushes dogs, which requires a fair amount of light to do correctly. 2700K might have made the place feel a little more quaint and comfy. It would also have made the room less useful.

  2. zeph says:

    I must say, my experience with CFLs has been much different. Only the very oldest ones take any time to start up. I had some newer ones that have only lasted four or five years, which was disappointing, as most of mine have been around for decades.

    However, I replaced the CFLs that died with LED lights this year. They provide much better lighting at the same minimal power cost, with smaller bulbs. How long they will last is yet to be determined.

  3. Tom Roderick says:

    Newer CFL’s are available in a wide range of color temperatures which has helped some, but I have often replaced a 60 watt incandescent with a 75 watt equivalent CFL to get enough light.

    My biggest two gripes are that they regular phase control dimmers don’t work with them and some of the cheap ones emit that really foul and persistent aroma of burning electronics when they fail. I have learned that you CAN use CFL’s on timers that have a Triac switch instead of a relay if you have an incandescent in parallel — it is best to have two since if one fails the CFL is on its on.

    One thing I want to hear more about regarding LED replacements is some of the issues of RF noise that I have been hearing about. Maybe Jeff can report if he has had any problems.

  4. Lee Hart says:

    Light is a tricky customer. The physics are complex, and the response of the human eye significantly distorts our perception of it. Most people do not understand it, which leaves marketeers free to invent all sorts of claims. “Lying with statistics” is really going into high gear with all the new light sources coming on the market.

    Color temperature is just the color; not the brightness or illumination of a light source. It used to be easy; the apparent “color” of an object heated to that temperature. It doesn’t really apply to objects like CFLs or LEDs that use phosphors or have a spectrum with sharp peaks and holes.

    Illumination (in Lumens) is a measure of the total amount of light produced, in all directions. If you want a lot of light to illuminate a room, buy lumens.

    Brightness (in Candela) measures the peak intensity at a single infinitesimal point. Brightness matters for light sources that you look directly at; car tail lights, LED indicators, etc. But brightness has nothing to do with illumination. Stars are bright; but provide negligible illumination (you can’t read by starlight).

    Lamp life has a *lot* to do with quality. Lights are mass produced consumer items, and inclined to be manufactured as cheaply as possible. I too have had many CFLs with early failures. Cutting them open revealed appallingly cheap parts and bad construction.

    An LED can last 100k hours at rated power and temperature — but an offshore producer will happily run it hot at 10x rated current to get the same light with cheaper parts. Its life expectancy may now be only be 1000 hours; but they know most consumers won’t notice, or won’t return them under warranty.

    Tungsten lamp life is drastically affected by voltage, current, vibration, and quality control.

    1. 10% overvoltage shortens life about 3:1. If your AC line voltage tends to run high, there goes your lamp life.

    2. A cold filament draws 5-10 times its normal current (if there isn’t any series resistance to prevent it). Each turn-on seriously shortens the lamp’s life, so adding a little series resistance or slowly turning up a dimmer improves life dramatically. As Jeff suggests, screwing a lamp in or out with the power on could indeed seriously reduce its life from this effect, as it could easily see ten “turn-on” surges.

    3. Vibration from slamming doors etc. also causes a big reduction in life. Use “rough service” lamps if you notice the filament vibrating.

    4. It is said that a 1% tolerance in manufacturing has a 25% effect on lamp life. Like most things, quality control problems can thus cause big differences (one lamp lasts twice as long as an identical lamp used exactly the same way).

    1. Alex Dillard says:

      The candela isn’t a unit of brightness, although it can correspond to how intense a light source appears to be. As far as I am aware there is no unit of perceived brightness since it is a subjective property of how something looks to an individual observer in their specific viewing conditions (which makes it hard to quantify). Candelas specify luminance, or more specifically luminous intensity (luminance itself is candelas per square meter). Lumens refers to luminous flux, which as Lee points out is related to a light bulb’s ability to effectively “light” a room.

      Note that both lumens and candelas are photometric units, meaning they are normalized to the sensitivity of a normal human eye. There is a complete set of corresponding quantities called radiometric units which are not normalized to any particular spectral response profile. The units of light measurement are a particularly confusing category of physics units. I’ve always felt that spectral radiance (watts per steradian per square meter per hertz) takes-the-cake in this regard.

      More info:

  5. Jack Smith says:

    A few random thoughts on LED/CFL/Incandescent lamps

    1) Base up / base down has significant thermal differences, particularly for LED and CFL lamps. With base down, the heat generated by the light emitting part of the device convects away from the electronics end of the lamp. Base up increases the temperature of the electronics, which can significantly shorten the life of components, particularly electrolytic capacitors. Even with an incandescent lamp, base up/down changes the temperature of the seal and also of the socket. Base up sockets should be ceramic, base down can be a suitable thermoplastic.

    2) If you disassemble the electronics package of a cheap CFL, you will find marginal components; electrolytic capacitors of 85 deg C rating, not 105 or greater, and operated with little to no peak voltage de-rating.

    3) As I understand it – and this is not an area in which I have a great deal of first hand experience – the tungsten filament can have hot spots and the tungsten sublimes preferentially from the hot spots. Over time, the hot spots thin out, increasing their resistance and hence become hotter under normal operating conditions, leading to increased sublimination, etc. Filling lamps with nitrogen instead of operating them as vacuum tubes reduces sublimination, but the real fix for that is found in the halogen lamps.

    4) Line voltage has increased over time; the earliest central power plants built by Edison distributed power (DC, of course) at 100V, plus a 10% loss factor, thus leading to the 110V “standard.” When AC replaced DC, the 110V standard was maintained as 110V RMS. Since outside plant distribution networks operate more efficiently at higher voltages, the “standard” voltage has increased, being 110V, 115V, 117.5V and 120V. Look at the rating nameplate on old electrical equipment and you will see this progression.

    5) It is possible to purchase 130V incandescent lamps – readily available on the Internet or from your local industrial supply house. These will last much longer than a 120V lamp due to the 4th power relationship mentioned earlier. There are also “long life” incandescent lamps. When I worked in broadcasting, we used 130V lamps (also known as traffic signal lamps) in tower lighting obstruction beacons because it was so expensive to get a steeplejack out to change a lamp. (Code beacons have two lamps for this reason as well.) A 100W “standard” incandescent lamp has a typical operating life of a few thousand hours, usually quoted as 0.9 years or the like. A 130V or “long life” lamp will usually have a rated lifetime of 12 years or so.

    6) For those of us that are amateur radio operators, or short wave listeners, CFL and LED lamps are terrible – they have a switching power supply running at about 50 to 60 KHz and radiate harmonics well up into the lower shortwave and amateur radio bands. They also radiate enough energy to make using them on a test bench difficult; I have a couple of medium price Philips LED spotlights on one of my work benches and if I am measuring low signal levels with an oscilloscope or spectrum analyzer, it’s necessary to turn them off or else the desired signal is contaminated with induced signals from the LEDs. In some cases I have to turn off the T-8 electronic ballasted fluorescent lamps as well.

    7. Yes, there are FCC and CE limits on radiated emission from electronic lamps, but if you disassemble a lamp you will often find the PCB is screened for chokes and bypass capacitors to control radiated emission, but the parts are not installed. Yet, the lamps still carry the FCC or CE compliance mark.

    8. I’ve written a bit about this subject at — probably needs updating but I believe the general information there remains correct.

  6. RH in CT says:

    Creating an incandescent bulb that lasts a long, long time is not difficult, mostly being a tradeoff between life and efficiency.

    The early CFLs were horrible, and the next generation were poor, but if you never used good ones you gave up too soon. When millions are sold every year they usually get better.

    That said, I have been waiting for the LEDs to fulfill there promise and they are finally starting to get there. Until there is a 100 watt replacement at a reasonable price (not $20) we won’t be there. The problem is that the actual manufacturing cost of a brighter LED is considerable, since it requires more of the most expensive component – the LEDs – rather than a trivial change in the filament or a longer twisty tube. But they are getting there. Here in CT the Cree 60w equivalent (with state incentives) is under $5, while the 100w is up around $20. I suspect that the 100w is simply awaiting some sort of approval by the state, after which the price will drop by $5 to $8. In the meanwhile I try not to buy anything that can’t take multiple bulbs. Decorative bulbs are going to be a problem.

    I have heard that LED bulbs used in a garage door opener can cut the effective distance of the remote severely. Cree is said to be looking into it.

Leave a Reply

Your email address will not be published. Required fields are marked *