Real life span of real LED fixtures

[drmaxx]

I just looked at a few high powered light bulb here in a store. The 3 x 3W LED bulb was on display. It was quite hot - still able to touch the casing of the bulb though. So, I am guessing that the dies get substantially hotter...

Now, I am wondering: How long do these lightbulbs actually last. Let's say till 80% light output. Are there any real tests that could confirm the claimed several 10'000 h life-span. Does anybody of you did some testing?

 

[JohnR66]

No way to give you an exact answer. It depends on the quality of the LEDs in use, the current drive and temperature.

With 5mm LEDs, I found some show signs of fading in 48 hours of use while better quality ones have not faded after 200 hours.

Power LEDs tend to last longer, but I'd imagine cheap Chinese versions may suffer short lifespans as well. Cree rates their power LEDs for 50K hours at 70% light output lifetime. The junction temperature has to be 80 Deg C or lower. If the lamp has a good thermal path from the die to the heatsinking I would expect the lamp to get quite hot yet maintain the expected lifetime of the LED (other components may still be in question).

 

[Lynx_Arc]

if it is getting uncomfortably hot I wouldn't say it would last more than 6 months to a year without fading to 50% IMO. personally unless I had something that really needed an LED light bulb I would not buy them at this time they are way too expensive to be cost effective at this time... perhaps in a few years if the environmentalists get things fouled up where electricity prices go way up then the market will take off.

 

[jtr1962]

Now, I am wondering: How long do these lightbulbs actually last. Let's say till 80% light output. Are there any real tests that could confirm the claimed several 10'000 h life-span. Does anybody of you did some testing?

I'm in the process of both lumens testing and long-term testing LED lamps for someone. When I get the OK from him, I'll start a thread. The 2 ones I long-term tested which used 5mm LEDs faded to under 70% in 300 hours. The rest using 5mm I deemed not even worthy of long-term testing. Most had outputs not much more than night-light bulbs.

I'm testing 6 lamps with power LEDs long term right now. So far 3 of the 6 are actually above initial output at 300 hours, one is at 98%, the other two are around 95% and 87%. They hit the 1000 hour mark on November 29. Results should be interesting.

I just looked at a few high powered light bulb here in a store. The 3 x 3W LED bulb was on display. It was quite hot - still able to touch the casing of the bulb though. So, I am guessing that the dies get substantially hotter...

That's the problem right there. Based on my testing, the light bulb form factor is incapable of dissipating more than about 5 watts using passive cooling, period. In fact, I'm really more comfortable keeping the dissipation below 3 watts. Anyone who makes a lamp which dissipates more than that without active cooling is making garbage which won't last. Unfortunately, this simple little fact is going to limit LED screw-in lamp replacements until LEDs become somewhat more efficient. Either they won't last, or they won't have much output, or they'll need active cooling (which will add to the price).

if it is getting uncomfortably hot I wouldn't say it would last more than 6 months to a year without fading to 50% IMO. personally unless I had something that really needed an LED light bulb I would not buy them at this time they are way too expensive to be cost effective at this time... perhaps in a few years if the environmentalists get things fouled up where electricity prices go way up then the market will take off.

No, decent LED bulbs aren't going to get that much less expensive than they are right now. I'm figuring retail about $40 to $50 for a decent lamp with good electronics, 100K hour life, and active cooling (which is needed to reach outputs rivaling 100 to 200 watt incandescents). These prices may drop by perhaps $10 with mass production, perhaps even more if emitters get so efficient that no active cooling is needed, but you'll never see decent LED bulbs for the same prices as today's CFLs. If you do, then it means it's junk which won't last, same as the problematic $2 or $3 CFLs we already have. So you can have cheap or good, but not both. We've seen the same thing with CFLs. The $2 ones, or even $6 ones, are often a crap-shoot with regards to quality. Only once you hit the $15 or $20 price point are you fairly sure of getting a decent lamp with a guarantee the manufacturer will back.

Honestly, for what it might cost to put a few decent LED bulbs in an existing fixture, we can more cheaply engineer a brand new LED-specific fixture which will cost less. This is why I think long-term LED bulbs will be a passing fad. Consumers will balk at the price compared to purpose-built LED fixtures, and will just go with the latter.

 

[Lynx_Arc]

No, decent LED bulbs aren't going to get that much less expensive than they are right now. I'm figuring retail about $40 to $50 for a decent lamp with good electronics, 100K hour life, and active cooling (which is needed to reach outputs rivaling 100 to 200 watt incandescents). These prices may drop by perhaps $10 with mass production, perhaps even more if emitters get so efficient that no active cooling is needed, but you'll never see decent LED bulbs for the same prices as today's CFLs. If you do, then it means it's junk which won't last, same as the problematic $2 or $3 CFLs we already have. So you can have cheap or good, but not both. We've seen the same thing with CFLs. The $2 ones, or even $6 ones, are often a crap-shoot with regards to quality. Only once you hit the $15 or $20 price point are you fairly sure of getting a decent lamp with a guarantee the manufacturer will back.

Honestly, for what it might cost to put a few decent LED bulbs in an existing fixture, we can more cheaply engineer a brand new LED-specific fixture which will cost less. This is why I think long-term LED bulbs will be a passing fad. Consumers will balk at the price compared to purpose-built LED fixtures, and will just go with the latter.

I agree that unless LEDs get more efficient prices under $25 for anything approaching 75 watts incan equivelent isn't promising. I also agree that the screw in base design will be supplanted by a new design that would allow both LED and fluorescent to be interchangeably used in fixtures. Perhaps they will make a kit that you can replace standard sockets in lights in many fixtures but unless fixtures use multiple bulbs a single LED bulb solution for lighting isn't a good design. I could see LED chandeliers in the future in mansions with perhaps 20 single high power modules in place of a small 16 watt CFL etc. LEDs in flashlights and lanterns have required the same redesign that fluorescent lanterns and fixtures went through before the first CFL screw in bulb replacement surfaced costing around $20-$25 then.
I have bought the cheap CFLs..... bad color, flakey etc but you can get semi decent utility CFLs for upwards of $3 that do ok. With the exception of heatsinks I don't see the electronics in LED fixtures costing more than CFL electronics do.

 

[jtr1962]

I also agree that the screw in base design will be supplanted by a new design that would allow both LED and fluorescent to be interchangeably used in fixtures.

I actually think the linear tube format lends itself fairly well to LEDs. Plenty of surface area to dissipate heat, plus room to mount an external ballast. The problem with compact form factors like the screw-in lamp, or even the new GU-24 format, besides lack of room for heatsinking, is trying to cram the ballast electronics into a small space right next to a lamp which gets hot.

I could see LED chandeliers in the future in mansions with perhaps 20 single high power modules in place of a small 16 watt CFL etc.

A conventional chandelier running a bunch of 1 watt or so emitters, and with a ballast hidden in the main body of the chandelier driving those emitters, would probably work fairly well. Plenty of open space to dissipate heat, and the ballast needn't be near the heat-generating lamps. XP-Gs running at 350 mA will generate only about 0.6 watts of heat each. A very small heat sink can cope with that.

With the exception of heatsinks I don't see the electronics in LED fixtures costing more than CFL electronics do.

That's true, although any decent CFL driver electronics will probably run around $5 in quantity, same as decent LED drivers. The ones we see in the cheap CFLs right now are horrible for the most part. I've taken a few apart. They use the crappiest components imaginable. The only reason they sometimes get away with it is the fact that the ballast is throw-away. It only has to last as long as the tube, which is perhaps 6000 to 8000 hours. Electronics have reached such a point in reliability that even garbage can be made to last that long.

With LEDs you no longer have this luxury. The main selling point of LEDs right now is ultra long life. They can also save energy over CFLs, but for now those savings are marginal. Ultra long life means the driver must also last 50K or 100K hours. Easy to do with a $5 ballast, impossible to do with a 75 cent ballast as currently is used in many CFLs. Also, for what these lamps cost you'll want them dimmable on a standard triac dimmer. Easy to do with a decent ballast, impossible to do with a cheap one. My main point here is that if you want an LED bulb to live up to the long-life claim, then you can't get away with skimping on the ballast as you currently can with CFLs. So look for $40-$50 retail for now. As the costs of emitters comes down, and perhaps we no longer need active cooling, you may see decent LED bulbs in the $25 to $30 price range. Still a bargain for what it is, which is basically a screw it in and forget about it for the next few decades solution.

 

[Lynx_Arc]

I actually think the linear tube format lends itself fairly well to LEDs. Plenty of surface area to dissipate heat, plus room to mount an external ballast. The problem with compact form factors like the screw-in lamp, or even the new GU-24 format, besides lack of room for heatsinking, is trying to cram the ballast electronics into a small space right next to a lamp which gets hot.

the problem though with exchanging LEDs for linear tubes is most of the newer tubes are already more efficient than LEDS so unless electricity prices go up considerably the expenditure vs the energy savings is less attractive.

A conventional chandelier running a bunch of 1 watt or so emitters, and with a ballast hidden in the main body of the chandelier driving those emitters, would probably work fairly well. Plenty of open space to dissipate heat, and the ballast needn't be near the heat-generating lamps. XP-Gs running at 350 mA will generate only about 0.6 watts of heat each. A very small heat sink can cope with that.

it would be easy to mass manufacture LEDs on heatsinks (like stars are) and you could hide the electronics in the ceiling above the chandelier which IMO is optimal if done properly it would be easy to repair/replace.

That's true, although any decent CFL driver electronics will probably run around $5 in quantity, same as decent LED drivers. The ones we see in the cheap CFLs right now are horrible for the most part. I've taken a few apart. They use the crappiest components imaginable. The only reason they sometimes get away with it is the fact that the ballast is throw-away. It only has to last as long as the tube, which is perhaps 6000 to 8000 hours. Electronics have reached such a point in reliability that even garbage can be made to last that long.

I would say in quantity CFL drivers around $2 would be decent, if made in china the parts probably cost 75 cents or so and labor is dirt cheap and the stores selling the good ones for less than $10 probably have a markup of $4-$5 on them. IMO if a $5 CFL lasts 5 years is ok to me but a $3 one that lasts 6 months is unnacceptable

With LEDs you no longer have this luxury. The main selling point of LEDs right now is ultra long life. They can also save energy over CFLs, but for now those savings are marginal. Ultra long life means the driver must also last 50K or 100K hours. Easy to do with a $5 ballast, impossible to do with a 75 cent ballast as currently is used in many CFLs. Also, for what these lamps cost you'll want them dimmable on a standard triac dimmer. Easy to do with a decent ballast, impossible to do with a cheap one. My main point here is that if you want an LED bulb to live up to the long-life claim, then you can't get away with skimping on the ballast as you currently can with CFLs. So look for $40-$50 retail for now. As the costs of emitters comes down, and perhaps we no longer need active cooling, you may see decent LED bulbs in the $25 to $30 price range. Still a bargain for what it is, which is basically a screw it in and forget about it for the next few decades solution.

If you sell an LED bulb for $25 that means the stores have to buy it for about $12-$15. I believe a decent cree emitter could be had in huge quantities at about $2.5-$4 each, add $5 electronics that leaves $3-$7 for a package which could be done. I think right now the biggest obstacle is the same one that CFLs had when they were over $20, people would rather buy a $1 incan bulb. It is quite possible in the future energy prices will soar and cities/states/countries will start subsidizing LED technology in standard house lighting and we will see dramatic price drops including emitters handling 10 watts for less than $1 in quantities of 100 million pieces a year

 

[drmaxx]

I'm in the process of both lumens testing and long-term testing LED lamps for someone. When I get the OK from him, I'll start a thread. The 2 ones I long-term tested which used 5mm LEDs faded to under 70% in 300 hours. The rest using 5mm I deemed not even worthy of long-term testing. Most had outputs not much more than night-light bulbs.

I'm testing 6 lamps with power LEDs long term right now. So far 3 of the 6 are actually above initial output at 300 hours, one is at 98%, the other two are around 95% and 87%. They hit the 1000 hour mark on November 29. Results should be interesting.

Very interested in seeing the results. Hope you get the permission to publish your results.

That's the problem right there. Based on my testing, the light bulb form factor is incapable of dissipating more than about 5 watts using passive cooling, period. In fact, I'm really more comfortable keeping the dissipation below 3 watts. Anyone who makes a lamp which dissipates more than that without active cooling is making garbage which won't last. Unfortunately, this simple little fact is going to limit LED screw-in lamp replacements until LEDs become somewhat more efficient. Either they won't last, or they won't have much output, or they'll need active cooling (which will add to the price).

So what you're saying is, that buying a 3 W 'LED-bulb' is about the max. that makes reasonably sense. Good to know and it confirms my suspicion that they won't replace the cfl at home anytime soon.

As special fixtures the high powered LEDs pop up more and more. E.g. in the new public transportation (trains) they use single die LEDs for additional illumination beside the CFLs. The 'bulbs' used to illuminate a hallway at my work are 9 LEDs arranged in an approx. 4 inch circle.

 

[jtr1962]

the problem though with exchanging LEDs for linear tubes is most of the newer tubes are already more efficient than LEDS so unless electricity prices go up considerably the expenditure vs the energy savings is less attractive.

No arguments here. I was just saying if you want a form factor which supports both LEDs and fluorescent, then the linear tube form factor is a good one. But I agree 100% it makes zero economic sense at this point to use LEDs in linear fixtures, given how inexpensive and efficient today's fluorescent tubes are.

I would say in quantity CFL drivers around $2 would be decent, if made in china the parts probably cost 75 cents or so and labor is dirt cheap and the stores selling the good ones for less than $10 probably have a markup of $4-$5 on them.

Well, you can probably make a decent driver with OK current regulation for $2. Adding things like dimming or PFC, both of which will become more important with LEDs, might bring it closer to $5. Dimming especially seems to be a desired feature but fluorescent tubes are inherently difficult to dim. LEDs don't have that problem, so hopefully manufacturers will have LED bulbs with dimming drivers.

It is quite possible in the future energy prices will soar and cities/states/countries will start subsidizing LED technology in standard house lighting and we will see dramatic price drops including emitters handling 10 watts for less than $1 in quantities of 100 million pieces a year

Energy prices are going up, no doubt about it. That alone will help LEDs. And I do indeed hope some breakthrough in manufacturing dramatically lowers the cost of emitters. Emitters are really the largest part of the cost of LED lamps.

So what you're saying is, that buying a 3 W 'LED-bulb' is about the max. that makes reasonably sense. Good to know and it confirms my suspicion that they won't replace the cfl at home anytime soon.

LEDs won't replace CFLs anytime soon unless consumers change their buying habits, and start looking at total cost of ownership. You can make an LED bulb to replace a 100 watt or larger incandescent with active cooling which might last 75,000 or 100,000 hours. I'm semi-involved in something like that right now. But it'll probably cost $40 or $50. Commercial customers would likely be all over it because they'll do a cost-benefit analysis. But they will be a hard sell for residential customers. And for smaller wattage incandescent lamp replacements up to maybe 40 watts (even 60 watts using XP-Gs) you can make LED replacements without active cooling, selling for perhaps $35. But again the hurdle is convincing residential customers to buy them. What I fear is that the cheap junk which is price competitive with CFLs will give LEDs a bad rap. I really hope some standards, plus truth in labeling, happen in the near future.

IMO purpose-made LED fixtures are likely what we'll eventually be seeing in homes replacing socket-based lamps. When you go through the numbers, it costs less to make a decent LED fixture instead of putting a few decent LED bulbs in an existing fixture. You have one ballast instead of several, no need for active cooling, plus a cooler running fixture.

Oh, and as soon as I get the OK I'll publicly release my results. The lumen depreciation testing will likely be ongoing until the bulbs dim to 70% or even 50%.

 

[Lynx_Arc]

I remember some of the first CFLs out there.... some were good and some were not. Lights of America puts out lousy fluorescent anything IMO but I have some GE and honeywell CFLs that are pretty good so far the honeywell ones cost $3 each and I used one for 4 years before it died and the GE ones I use about 1/4 as much and still are going. I think the turning point will be LED fixtures at the brick and mortar stores made to replace incans at the same dimensions like can lights. You could make a can light with an aluminum can and optics and LEDs in them and probably get by around the 60 watt or so output. Ceiling fixtures could be made to bolt on and replace existing incan fixtures. I would like to see LED drop in modules that you could swap out like like bulbs to change colors etc. Modules that require no soldering or expoxy but maybe just heat grease. LED celing fixtures with dimmers and other neat features would be kewl.

 

[J_C]

What lighting you see will depend on the amount of time you wait. There will always be early adopters who are willing to pay a price premium but for the most part retailers want to stock what the majority of customers would buy which is a "bulb" that fits in their incan fixture socket.

The orientation of that socket and whether it is recessed, and ambient temperature, can determine the emitter wattage possible while retaining long life. A crude cast metal heatsink housing in a recessed ceiling light, tropical location without air conditioning, would not last running at 3W. A decent extruded heatsink in a standard table lamp in an air-conditioned room could support 6W if not a little more before a fan is required though I agree with prior comments that the driver lifespan is at issue if the bulb is passively cooled.