Actual LED lumens maintenance. Not 100,000 hours.

[liveforphysics]

Thank you so much for all the excellent info!

It seems you folks are telling me much of what I had also thought about the performance relationship between T8/T5 vs LED. For our specific application, a big part of the appeal to LED tech was the labor involved to change bulbs. This enviroment is very risky to change bulbs in, due to the potential for bumping a ladder or man-lift into some part of the mostly overhead data infrastructure.

saabluster- You also recall seeing some large output drops in the infancy period of LED lifespans from tests that may have been from Newbie? This is exactly the sort of thing which causes me concern. I'm fairly certian you would not see evidence of the early phosphor failure by reading the SSC manufacture's spec sheets. I can't help but question the manufacture spec sheets with 10's of Kilohour output durability estimates for products with a development to production time measured in months. The thermal management consists of mounting 16 x Q4's on a little aluminum strip that mounts into the existing fixture, and replacing the ballast with a 700mA current supply. The product really seems like the sort of thing that you could DIY in your garage in a few hours for about 40$ over the cost of the LEDs, yet they each sell for a small fortune. They carry 10 year warrenty on manufacture defects ONLY. I get the impression that if they all become dim in 2-3 years, we are entirely on our own.

In our application, we always have techs handling server issues on the datacenter floors 24/7, so our lights are never switched off. When you have 400,000 servers, it requires continous babysitting to keep things running smoothly.

I personally think CRI is important, but I seem to be alone on this in the datacenter lighting field.

There are really only two factors that seem to most strongly come into play. Reduction of maintenance time, and hence risk to incidents which effect server infrastructure, and power usage.

Does anyone have any non-manufacture supplied data on lumens maintenance for CREE products?


Thank you for being such an amazing group of people!

-Luke

 

[Juctuc]

About lumens: Most of the ledproduct-manufacturers are telling the lumens of the LED, at tj.25c. Those are not the lumens when the light is on. Manufacturers like to use also the highest lumens, and not typical lumens. In that kind of investment what we are talking here, i think that manufacturer should have some information about fixturelumens as well.Even better if there is a IES files for it.

And if the lumens are ok, then we can start to talk about the lifetime..Most of my personal testing is been on the lowcurrent leds..results are not very good, at least on for the 5mm led with white colour. I have one E27 with semileds(from U.S.A) inside.5x1w. In 5600hrs my luxmeter is showing 6% less than it was at the start..now im testing CREE XRE 3X1W MR16 warmwhite.But i just started...im waiting about 10% fading in 10 000hrs. But we will see after one year=)

 

[liveforphysics]

Juctuc- That is exactly the sort of info I'm interested in hearing. Tests performed out of the manufacture's lab, and in a real working enviroment.

6% over 10,000hrs for a 5mm white? I'm just taking a wild guess that the spec sheet for that LED lists something like 100,000hr to 80-90% lifespan?

I love CREE products, and I'm wild about solid state lighting. I also don't feel like they are working to mislead us, but I do think perhaps the testing procedure is not quite perfect.


This is part of a sample kit given to us by the company. Its a small desklamp. It has a 12VAC trasformer that powers it.

I don't know anything about these emmiters, but they oddly appear to be multi-die perhaps? I took the unit apart and took some photos.

Again, the lighting for the server rooms is a different product, and CREE Q4 based, this is just a novelty sample they sent out. Any idea's why it appears to have 6 dark areas under the phosphor in each emmiter?

Thanks!
-Luke

 

[Juctuc]

That 6% is with so called powerled, not 5mm led. 5mm LED will last about 5000hrs, and then it has lost about 50% from the original lightoutput..i think.

I dont know what leds those are in your picture. In the website in this company, who is making these, there is no information what led they are using. Most of the manufacturers will tell if they are using CREE. But maybe this company is starting to use them just now.

 

[liveforphysics]

Wow! 6% from a powerled that was not being overdriven? These experiences are very different than what most LED info would lead you think.



These are the sample T8 replacement LED strips they sent us about 6 months ago. They definately use CREE leds. I don't have one of the newer Q4 based units laying around, or I would photograph it for you. They don't look much different though. Just emmiters on an alumium stick, an LED ballast, and a crazy price tag.

They seem to do the job just fine for the 6 months I've had them in my office. My concern is fear of dropping 3M dollars on lighting that becomes unsafely dim for this enviroment after a few years.

 

[Lunal_Tic]

Could you not just write lumen maintenance/output into the contract? Something like OSHA's illumination requirements for your application: guarantee x number of foot-candles for a given number of years with the option to replace units that fall below that. I don't really know all the ins and outs but IIRC they have ratings for everything from washrooms to warehouses.

-LT

 

[jtr1962]

The heatsinking for those strips really doesn't seem adequate to me. I really think power LEDs can exceed 100,000 hours or more, but only if underdriven, and only if run very cool. Neither seems to be the case with that product. You really need a purpose-built LED fixture. I have little doubt the LEDs will still be working ten years from now as the company said, but they will probably be quite dim by then.

Regarding tests, I actually have a Q3J Luxeon which has been running continuously since January 2004 at 350 mA. Although I didn't have a light meter when the test started, it's measured the same for the last 3 years. With over 41,000 hours on it, I'm confident 50,000 hours with >70% initial lumens is easily possible. Indeed, it seems my LED is doing significantly better than that.

Since these fixtures are on 24/7, you might want to look into these. They will run only $5 each in quantity, and have improved color rendering (CRI = 91) plus 34,000 hour life. Note that lifetime for fluorescent tubes is specified with 3 hours on, 30 minutes off. When run continuously most fluorescent tubes exceed rated lifetime by about 40%. This would mean you would get about 48,000 hours in your continuous-duty application. This is over 5 years.

 

[Rogerg]

Reflectors / mirrors overhead and lights lower, out of the way directed upward and accessible. No clue if practical or possible.

 

[matrixshaman]

I'll bet the high price tag has a lot to do with a salesmans' commission. How about trying some of the ones available from various sources in a smaller area and hiring or using your own people to install them? Perhaps do a gradual switchover and over a couple years you will likely be installing more efficient ones and if the older ones degrade than upgrade those also.

Or hire CPF people to build them for you ongoing for a gradual switchover

 

[saabluster]

WOW! I could design something better in my sleep. I agree with jtr1962. There just seems to be little thought given in how to dissipate the heat. I would stay away from this. What I would suggest is sending one of these to jtr(if he is so inclined) to test and see what how hot the LEDs are getting. This would give us a better idea of the true life span of these.

Just emmiters on an alumium stick, an LED ballast, and a crazy price tag.

They seem to do the job just fine for the 6 months I've had them in my office. My concern is fear of dropping 3M dollars on lighting that becomes unsafely dim for this enviroment after a few years.

 

[VanIsleDSM]

As someone else who dabbles in machining their own aluminum LED fixtures, I'll third the fact that the fixture pictured looks nowhere near adequate to remove the heat.. You'd at least want some vertical fins on either side of the "LED tube" to utilize convection currents to cool it down. Since LEDs become less efficient as they heat up, you're generally looking at 10% less output that claimed by the LED manufacturer at 25C, that's with a good heatsink design with a die temp around 60-70C. With a poor design you'll lose out on another 10-20% output from the LED due to heat, and it will degrade much faster.

I'd test to see what the heatsink temp is as close to one of the LEDs as possible, and then determine the die temp through the typical thermal resistance of a CREE LED of 8C/W to get a real idea of how well these fixtures are removing the heat. 3 Million is a large investment, I wonder how much they are charging for each fixture+driver (16 CREEs in large quantity placed on aluminum extusions).

 

[liveforphysics]

All these LED nuts, and nobody can ID that emmiter? LED museum guy?

Does anyone know why is appears to have 6 dice?

Thanks!
-Luke

 

[LukeA]

All these LED nuts, and nobody can ID that emmiter? LED museum guy?

Does anyone know why is appears to have 6 dice?

Thanks!
-Luke

That's a 6-die 1W Nichia device, possibly one of the 083 high-CRI emitters.

 

[snarfer]

Noticed the markings on the PCB so I looked up the company. Supposedly they have "patent-pending" technology for thermal management that conducts the heat from the LED strip to the housing.

Google patent search did not produce anything relevant name of the company or the CEO/founder in it. So I think that it must be some sort of secret "provisional patent" that allows them to say that it is patent-pending. Or maybe they are licensing it from another unnamed company.

Only problem with conducting heat from strip to housing is that of course the housing is made of steel, which has much less thermal conductivity than aluminum or copper, by about a factor of 10.

They have listed some photometrics in which they play games with "fluorescent equivalent lumens." What the hell are those?

[rant]I am pretty sick of all these companies that think they are getting over by making up their own measuring units. Some guys make a panel out of a bunch of crap 5mm LEDs and then say that it is "400 watt equivalent" even though it only draws 35 watts, and a quarter of that is in the resistors. [/rant]

Actually I was doing some tests last week comparing fluorescent lamp from KinoFlo to a project I am working on using multi-emitter LED devices and I was amazed at how inefficient the fluorescent fixtures were.

We were shooting some color charts, and we set up 72 watt LED lamp with no optics or anything (so it had a very wide spread), at same distance as LED Diva 400, which uses 4 x 55 watt PL-L (essentially T-5 u-tubes). The Diva tubes are rated at 2560 lumens each, so total of over 10,000 lumens, while the LEDs are supposedly 60 lumens/watt so a little over 4,000 lumens.

I would have expected a difference of a full stop, but it turned out that with the LED lamp just as much light was hitting the chart. So it goes to show you that fluorescent lamp and reflector combinations may not even be as good as 80% efficient. Maybe 50%. And this is with the newer style tubes. T8s are probably even worse!

For a long time now I've been saying that LEDs just can't compete with fluorescent, but after that test I really started to think differently. Maybe it is possible already.

Hey by the way I have also been experimenting with powering LEDs with fluorescent ballasts, and it actually works pretty well. I did a very crude test with a four dollar T8 ballast, a bridge rectifier, 1 mH inductor and 2.2uF /400v low ESR film capacitor. For some reason the ballast took a long time to turn on, maybe because of such a low voltage, but when it did turn on it provided exactly 350 mA of power to the string at various numbers of LEDs. There was absolutely no flicker. I used only two of the four pins. Had to open up the ballast to figure out which pins.

Something else I just noticed in one of their spec sheets (look here). Actually this looks like a different product from the 12 Crees on aluminum that you posted picture of. But in the T8 troffer retrofit brochure they say they are running 30 or 45 LEDs per bar at either 400 mW/LED, 870 mW/LED, or 1400 mW/LED depending on which option you choose. So it looks like they are doing something that makes sense for long life of LEDs.

 

[jtr1962]

By the way I have also been experimenting with powering LEDs with fluorescent ballasts, and it actually works pretty well. I did a very crude test with a four dollar T8 ballast, a bridge rectifier, 1 mH inductor and 2.2uF /400v low ESR film capacitor. For some reason the ballast took a long time to turn on, maybe because of such a low voltage, but when it did turn on it provided exactly 350 mA of power to the string at various numbers of LEDs. There was absolutely no flicker. I used only two of the four pins. Had to open up the ballast to figure out which pins.

That's interesting. I'd guess it would turn on instantly if you used a string of LEDs equivalent in voltage to the fluoro tube. I'm amazed that it works at all, let alone as well as you say. If this is the case, then maybe someone here can design LED retrofits for the original poster. It should be fairly easy electronically if you use the fluoro balllast, and you also avoid the potential liability issues of designing a 120VAC ballast from scratch.

 

[snarfer]

Yes I'm surprised that more people haven't experimented with using fluorescent ballasts for LED applications. Generally those ballasts are extremely efficient, especially for the European market.

If anyone wants to try it, be sure to use an electronic ballast that switches at high frequency. Also rectification seems to work much better than parallel strings, at least in simulation. I didn't bother actually building the parallel strings design.

If the original poster is interested in developing an alternative solution, I have some designs that could be useful. Sorry I can't be too specific about it in a public forum as it is the intellectual property of my company.

 

[jirik_cz]

I would have expected a difference of a full stop, but it turned out that with the LED lamp just as much light was hitting the chart. So it goes to show you that fluorescent lamp and reflector combinations may not even be as good as 80% efficient. Maybe 50%. And this is with the newer style tubes. T8s are probably even worse!

According to Caliper testing report round 5, T8 in housing with parabolic reflector has 60% efficiency and T12 in "lensed" housing has 72% efficiency.

 

[snarfer]

Thanks. That is an interesting document. I wonder what grade reflector they tested with. Seems like it is possible to order reflector material from 75% all the way up to 99%. Fixture efficiency of 60% would be within the range of what I observed.

If I was considering this purchase I would definitely get out a light meter and do a direct comparison between the LED replacements and the existing fluorescents at this point.

 

[Hooked on Fenix]

50,000 hours is not that impressive, premium brands of flourescent lamp often last several years of continous burning in good conditions.

Phillips produce ultra long life flourescents which are claimed to last 63,000 hours, and also claim 90% lumen maintenance at end of life.

I know of no such actual flourescent lamps that will last that long. I think what you are refering to is induction lighting. I saw this technology at the Electric West conventions years ago. Basically it looks like a fat flourescent bulb. It uses a a phosphor coating on the inside of the bulb like a flourescent. It has no electrodes. Instead of using U.V. light to light up the phosphor, it shoots microwaves through the bulb. If I remember correctly, the bulb and ballast were over $200. The most important thing to know about this bulb is that it can't be used indoors. It causes a lot of radio frequency interference so it is only used for outdoor lighting. It's not an option for office buildings. I think the T5 bulbs would be your best bet. L.e.d.s aren't cheap enough yet for that application. You don't want to be kicking yourself in the butt in a few years when they are cheap enough after spending millions on bulbs that have become way cheaper and better.

 

[KeithInAsia]

I think the key is gaining long LED life is a low operating temperature. The vendor should prove that their design will run the LED at the lowest possible temperature.

Also, they need to demonstrate that the driving electronics can be reliable as well. What is the point of the long life LEDs if the electronics are breaking down and possibily destroying the LED at the same time?

That is my 2 cents.