LED life versus current (at and above rated) in white light bulbs

Is the bulb rated for enclosed use? If so, it shouldn't be. It looks like a thermal nightmare, which kills electrolytic caps and LEDs.

I bought some Great Value LED bulbs from Walmart recently. I like them. They're rated for enclosed use, but they were running so hot in open air I don't see how they'd survive long. I added heat sinks, lowered their temp rise. by ~30C, for longer life.

Heat sinks described here:

They're close to 100lm/watt in 2,700K; over, in the 5,000K version, non- dimmable (for best efficiency).

The 5W/450lm 2,700K bulbs are just $2.44 now.

Cheers, James Arthur

The Philips LED bulbs are great, but I think they are getting out of the business. It's the usual Chinese race to the bottom.

The Chinese are building passenger jets now!

Are they over-spec'd for 25,000-hour illumination too?

Grins, James

Paul, here's a tear-down video of the 5W Walmart LED I mentioned:

It's hard to beat $2.44.

A different tear-down, very detailed, of a related unit (efficiency has improved since then):

They seem pretty serviceable, except for running so hot.

Cheers, James

Recently (Dec 2014) I bought some 9W LED light bulbs from

Then, a couple weeks ago, the new lamp started flickering, and also failed. Similar mode of failure. Several chips showed low resistance and did not light, while the others lit using the DMM diode check. So the lifetimes of both lamps were similar, about 2000 hours, and actually less than 1500 since they are used only about 12-18 hours a day.

I replaced the bad LED chips with some much smaller 1206 white LEDs, rated about 30mA or 100mW. I figure the current in the LED strings should be about

My thought is that the original LEDs were over-rated and failed because they were driven by too much current. But I could not find any definitive information on the effect of (over)current on lifetime. The sources I found only said that LEDs fail mostly because of excess temperature, and the lifetime is determined by measuring the slight drop in luminous output after continued operation at nominal current after several thousand hours, and then extrapolated to EOL defined as something like 50%-70% of initial output.

It seems that there should be some studies on the relationship of current and lifetime, but I could not find anything. It seems like the 5x data point is about 5 minutes or 0.1 hour, and perhaps it is an exponential function with 1x=1,000 hours, 2x=>100, 3x=>10, 4x=>1, 5x=>0.1. Obviously not quite that simple, but perhaps similar.

Of course, this early failure may be typical for cheap Chinese LEDs and bulbs, but I have also heard that some of the newer Cree lamps also have much shorter lives than expected. I may order some higher quality Osram chips from Mouser for about $0.20 each and replace the ones in this lamp and see if it lasts a lot longer. Not that it's really worthwhile to "fix" it, but to see if it is the quality of the LED chips or something else. It is possible that the two strings may have unequal current causing one to fail and the others to draw even more, for a cascading failure, but there were bad LEDs in both strings.

For more info and pictures, see my thread on:

Paul ============================================================================

Here is a document from Cree where they discuss the lumen output versus time for their leds, and the test methods they use:

Haven't read this version yet, but several years ago their data showed that at full output the lifespan would be 5-10,000 hours, and dropping 3-6x in output power would extend that to 50,000+ hours. Of course, everyone called the leds "50,000 hour" leds, and gave the max rated power, and no one mentioned that the two specs couldn't be met at the same time :-).

----- Regards, Carl Ijames

So did the Soviets. I wouldn't fly in one of those, either. "If it ain't Boeing, I ain't going!"

I flew Aeroflot once. It was interesting. They treated foreigners like royalty and let the Russians stand outside in the cold rain, waiting to board. I thought that was weird.

That is useful, but doesn't answer the question of reduced life above rated current. It may be purely academic, as there is no good reason to drive LEDs that hard. Apparently the actual output luminosity does not increase very much at the top end, so the only plausible reasons are design error, misrepresentation of actual specs, or deliberate reduction of life for planned obsolescence.

Maybe there is a way to detect the temperature of the junction, which probably is the main contributor to shortened life. The temperature is probably proportional to the power dissipation, but some power is in the form of visible light, while the remaining power is converted to heat, and temperature depends on the thermal resistance from the junction to the PCB and its final temperature in the application.

Thanks,

Paul

Ignoring the losses in electronics, that would be 500 mW or 170 mA for each LED.

No wonder.

There are two issues with white LEDs:

2.) Degradation of phosphor output (yellow/red) due to too high blue radiation. To design a LED lamp for 10 000 to 30 000 hours, I wouldn't use more than Imax/3

If a reputable manufacturer runs their LEDs at Imax/3 I would expect them to cost 3x than el cheapo products.

I built my own LED retrofits for twin-tube 48" fluorescent fixtures, using Cree 1 W chip LEDs. These are rated at 102 lm/W at 350 mA, which is what I run them at. One string of 20 of them gives as much light as TWO 48" T12 fluorescents. The first retrofit unit has been running in our kitchen since April, 2013. I did have one LED do something funny, but it was bad soldering, not an actual LED failure. I use a Thomas Research LED power supply from Digi-Key. They are pretty expensive, but seem to work well.

I have a picture and some text here:

Jon

On Mon, 28 Sep 2015 23:41:35 -0500, Jon Elson Gave us:

Sure looks like a lot of wasted circuit board media. I hope you are still putting the difuser back on.

There are huge commercial interest in claiming as much light output [lm] as possible, no matter the product lifetime.

A cool white LED will have well above 100 lm/W at Imax/2 or Imax/3, but drops quite rapidly at Imax not to mention 5 x Imax.

Using such high current and the light output will degrade after a few thousand hours.

Spectography perhaps. as the die gets hot the wavelength drops.

Lots of ways, e.g. briefly dropping the current to 1 mA and measuring its forward voltage.

Cheers

Phil Hobbs

Lumileds has good data too.

This graphic shows the nasty side of running hot:

My experience with an 8.5W LED bulb and ~10in^2 cylindrical heat sink surface exposed to open air was >60C rise, until I added additional heatsinking.

Paul's bulb appears to have a plastic case that relies on vertical convection through some small perforations in the LED board for heat transfer, and he's using it in an enclosed fixture. That's going to run /very/ hot, 80C rise (or worse).

Cheers, James Arthur

It was actually installed in an open fixture (a simple lamp-holder base) on the ceiling. For testing, I used an open ended desk lamp with a partial metal shroud. The translucent cover is part of the bulb, and it would probably be good to have holes or slots around the sides to allow cooler air to come in over the LED chips, through the holes, and out the slots near the base. But such holes might produce light spots around the bulb, although they would hardly be noticeable in an enclosed fixture.

The LEDs are mounted on a metal PCB about 2.5" diameter (4.9 square inches). I found a heatsink with about 1.5 in^2 area with 16C/watt, so the PCB in the lamp might be roughly 5C/watt, or 45C rise with 9 watts. Or for three 16C/W heatsinks with 3 watts each it would be 48C. That's likely a minimum figure in free air. The bulb itself did not seem to get very warm, but the heat sink got hot enough to soften the glue I used to hold it in place. (Shoe Goo, actually!)

There's enough room in the bulb for a small fan. I could probably add a 5 volt zener in series with the string and tap off of it. :)

Paul

Well, that is the heat sink! These LEDs do get quite warm at 1W input, so they need good cooling. It is a bit hard to figure out the temperature. if you put your hand in front of the LEDs, the optical output feels really hot! The back of the PC board material gets to, maybe 40 C or so after a while.

Oh, you HAVE to! These things are blindingly bright, they leave you seeing spots for a while.

Jon

Well, Cree has charts of degradation vs time at specific currents, and they show pretty minimal degradation at 10K hours at the "test current". We've been running them for almost 2 years now, and they still look plenty bright.

Jon