Actual LED lumens maintenance. Not 100,000 hours.

liveforphysics

Enlightened
Joined
Dec 4, 2006
Messages
343
I'm an engineer in a datacenter. We have a company which ensures that they can save us money by swapping all of our T8 fixtures to LED fixtures.

They feature a cool white bin of a CREE Q4.

The basis of the cost savings mainly comes from labor time savings due to the complex process to change the bulbs in our fixtures. We have over 2,000 4 bulb fixtures placed above rows of servers and thousands of miles of cables.
They drive each LED at 700mA.

They are claiming we would still have 90% lumens maintenance after 10 years of operation.

For some reason, I have this memory floating around in my head of some respected members of CPF doing lumens maintenance testing of LEDs, and finding results which were much different than the 100,000hr claims given by manufactures.

I'm trying to avoid making a 3million dollar mistake if I go forward with this, and they are dim after a couple years.

I've searched Google and CPF, and CPF searches with google, and I can't seem to find anything but manufactures data on LED lifespan specs.

Was I just dreaming that I read people performing tests that indicated 10-20% output drops after a few months of operation?

Any input from folks who have insight on LED lumens maintenance testing?

Thanks!
-Luke
 

LukeA

Flashlight Enthusiast
Joined
Jun 3, 2007
Messages
4,399
Location
near Pittsburgh
Cree specs 50k hours to 70%. If they're on 8 hours a day, then they'll last 17.1 years. Twelve hours a day is 11.4 years, 24 hours a day will be 5.7 years.
 

liveforphysics

Enlightened
Joined
Dec 4, 2006
Messages
343
Thank you LukeA!

50,000hrs to 70% when driven at 700mA, or 350mA? This is certianly quite a difference from the 10% at 10years they are claiming with their retro-fit product.

Those numbers seems closer to reasonable, but I seem to recall seeing some non-manufacture data done by a CPF member, perhaps McGizmo or someone similarly well respected who recorded big losses over a much shorter time span.

-Luke
 

jirik_cz

Flashlight Enthusiast
Joined
Jul 29, 2007
Messages
1,605
Location
europe
I recommend you to look on DOE Caliper test reports. I think that with good heatsinking 50000 hours is realistic claim. But some manufacturers make SSL retrofits with poor heatsinking and then the lifetime may be very short.
 

broadgage

Enlightened
Joined
Nov 23, 2007
Messages
445
Location
Somerset UK
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.
 

jtr1962

Flashaholic
Joined
Nov 22, 2003
Messages
7,506
Location
Flushing, NY
With good heatsinking you can probably get 50,000 hours or more at 700 mA. While it's true that there are some fluorescent tubes which last this long, they're not that common, and probably not exactly cheap. With inexpensive commodity tubes figure you'll get about 20,000 to 25,000 hours.

A better design decision might be to drive the LEDs at 350 mA, or perhaps even lower. Although data for underdriving is sparse, if power LEDs are anything like 5mm LEDs then life will increase exponentially. In other words, it might not be totally unrealistic to expect 1 million hours to 70% brightness via severe underdriving (i.e. 100 mA). The downside of course is that you'll need 2 to 5 times the number of LEDs. I think this is where streetlight design is headed. The desire is to get the light emitter to last as long as the fixture, which by most accounts is around 200 years ( that's about 1 million hours if on 13 hours per day).
 

saabluster

Flashlight Enthusiast
Joined
Oct 31, 2006
Messages
3,736
Location
Garland Tx
Those numbers seems closer to reasonable, but I seem to recall seeing some non-manufacture data done by a CPF member, perhaps McGizmo or someone similarly well respected who recorded big losses over a much shorter time span.

-Luke
I think it may have been Newbie. Not sure though. I think what you are remembering are tests that were done on the SSC P4. It showed a large loss of lumens in a very short time with it then leveling off some. It was speculated that although SSC's phosphor was very efficient, it was not to stable, and that was causing the reduction in output as well as a shift to blue. This is all just off the top of my head though. I don't remember ever seeing anything saying there was a problem with Cree's lumen maintenance.
 

SteveDavis

Newly Enlightened
Joined
Nov 20, 2007
Messages
66
I think it's best to consider this from the application standpoint. My understanding of datacenters is that they are well cooled and occupied very sparingly. That means that your flourescent fixtures are likely less efficient than normal from the cold, and lower life than normal from being turned on for short periods of time (look up the lifetime data on your tubes. I bet there's a big difference between 12hr starts and 3hr starts). That puts the application on the bad end for T8's, and on the good end for LEDs.

Cree rates mean lumen maintenance at 70% over 50,000 hrs for a junction temperature of 80 deg C. I believe this data is also at 350mA. Decreasing current or junction temperature will increase life, increasing these will decrease life. If the lighting company has excellent thermal management and the lights aren't on all that much, I don't think 10 years at 90% is unreasonable. I'd take a little time to figure out the average time that your lamps are on each on period, and how many hours a year that translates to. That will give you a good estimate of how long your tubes last, and how long the LED replacements will.

For more info on Cree lifetime, check here:

http://www.cree.com/products/pdf/XLamp_Reliability.pdf
 

monkeyboy

Flashlight Enthusiast
Joined
Mar 7, 2006
Messages
2,327
Location
UK
It's all just speculation. No one has actually done a real world test of the lifetime of these particular LEDs due to the fact that they haven't been available for long enough. So you don't know if you're going to get 200,000h or 5000h. Anyone can maintain a T8 fluorescent system but this LED system ties you into one company and leaves you at their mercy. They can then charge what they like for maintenance or disappear off the face of the earth if it all goes horribly wrong.

There are factors other than lifetime also to consider;

1) Fitting cost. To produce the amount of light required for an office will require a LOT of Cree emitters. 3 million dollars is a large amount to spend in one go for dubious long term savings.

2) Colour rendering. Cree Q4 LEDs can't match the CRI of tri-phosphor "daylight spectrum" fluorescent tubes. So maybe that would be a better upgrade in producing a healthier working environment?

3) In 5 years time, Cree Q4 will be obsolete. There will probably be silicon based LEDs with better efficiency and better CRI for one tenth of the cost.

IMO, It's not worthwhile switching to indoor LED lighting for at least another 5 or 10 years. Definitely not on that scale anyway. The first people to adopt a new technology are always the ones to get ripped off.
 
Last edited:

Oznog

Enlightened
Joined
Dec 2, 2006
Messages
595
The lifetime depends very very strongly on how low the die temp is maintained within the device. Unfortunately, heatsinking is the most expensive and difficult part of the design and it's easy to skimp on. It's unlikely they'd be able to use a fan. The traditional enclosures in the ceiling don't have a good way for air to flow through at all.

People won't notice anything wrong right away and the sale goes through.

Note that lumen maintenance affects efficiency too. What was 100lumens/watt initially becomes 70 lumens/watt by the end. That changes the bottom-line for the efficiency goals.
 

jtr1962

Flashaholic
Joined
Nov 22, 2003
Messages
7,506
Location
Flushing, NY
IMO, It's not worthwhile switching to indoor LED lighting for at least another 5 or 10 years. Definitely not on that scale anyway. The first people to adopt a new technology are always the ones to get ripped off.
I tend to agree with that assessment. LEDs at this point in time make quite viable replacements for incandescent-based light. However, I think it will be at least a decade before LED becomes cost-effective to replace T8 fluorescent systems. Right now efficiency is about the same, color rendering is the same or worse, and cost is much higher, even factoring in the longer life of LEDs. LED needs to be markedly better in at least two out of those three criteria before wholesale T8 replacement makes sense.
 

asdalton

Flashlight Enthusiast
Joined
Dec 12, 2002
Messages
1,722
Location
Northeast Oklahoma
For some reason, I have this memory floating around in my head of some respected members of CPF doing lumens maintenance testing of LEDs, and finding results which were much different than the 100,000hr claims given by manufactures.

The largest discrepancies were for 5-mm white LEDs, which are not designed to dissipate heat. They show degradation of output after thousands of hours--shortened to hundreds or even tens of hours if overdriven (which in practice they often are).

The original "100,000 hour" claim may have been carelessly transferred from the performance of underdriven monochromatic LEDs. Phosphor degradation is a weak point for white LEDs, and colored LEDs have no phosphor.

Properly driven, properly heatsinked power LEDs (Luxeon, Cree, Seoul, etc.) should last many thousands of hours easily. But I still wouldn't believe any "100,000 hour" claim without data.
 

JohnR66

Flashlight Enthusiast
Joined
Aug 1, 2007
Messages
1,052
Location
SW Ohio
Don't do it! BAD idea! A T8 system can give you 90 to 100 lumens per watt with good electronic ballasts. Q4s can match that, but as I understand the die temp can't be kept cool enough in normal operation to keep the LED that bright. The system is likely very expensive and you probably will give up some light to save money.

One 32 watt 48" T8 tube can put out nearly 3000+ lumens and the bulb costs a couble bucks. A single die LED puts out 200 lumens under ideal conditions and is more expensive. It would take 15 of them to equal the output of the fluoro tube.

I like LEDs a lot, but for general lighting, they don't make the grade. In some directional, color and low power uses, LEDs can be perfect, but metal halide and Flouro are still the kings for efficient lighting at high lumens now.
 
Last edited:

pril

Newly Enlightened
Joined
Jan 30, 2007
Messages
1
My .02$

Led is future without doubt but your question is all about money.

If you can sign contract (with good standing company and its life expectation for next 10+ years and with business insurance) with guaranty for power consumption, light durability, quality etc then this is ok, otherwise you should be very careful.

In this second case you have to do some cost analysis. You probably have exact data (power consumption and maintenance costs etc) for current system. You can easy compare your current costs with oferred costs for new system. Maybe costs analyze shows small difference for next two years - then is better to wait.

Maybe LED technology is not mature yet so you can try with small testing area for this. You can also compare other benefits from each technology (light quality, resistance to frequent switching etc.).

I understand your dilemma: wrong decision can cost you thousands of dollars. For this money is reasonable to order independent report (created by professionals for light systems) that will compare both systems. Maybe you can find someone right here …

Primoz

(Sorry, English is my second language)
 
Last edited:

blasterman

Flashlight Enthusiast
Joined
Jul 17, 2008
Messages
1,802
I'll agree with most of the LED posts above in that we need more data, and that solid state retrofit kits are for the most part bad news.

However, I'll happily 'kick the dog' and offer the advice that you should be dumping T8's and moving to T5's anyways. T5s *are* the state of the art in fluorescent design and in some cases displacing metal halide. 4100K T5's with a CRI greater than 90 are exceptional for data centers and offices.
 

jtr1962

Flashaholic
Joined
Nov 22, 2003
Messages
7,506
Location
Flushing, NY
Just going through some rough numbers:

Typical four-foot T8 tube = 2850 lumens
Four tube instant start ballast typical ballast factor: 0.89
Four tube instant start ballast typical power input: 110 watts
Output of 4-tube fixture = 4 x 0.89 x 2850 = 10146 lumens
Typical fixture efficiency: 80%
Actual light output of 4-tube T8 fixture = 0.80 x 10146 = 8117 lumens
Overall system efficiency = 8117/110 = 73.8 lm/W

Now let's do the numbers for Cree Q4's driven at 700 mA

Typical output: 170 lumens
Typical Vf: 3.5 V
Number of LEDs per fixture needed to match output of 4-tube T8 fixture: 48
Power input to LEDs = 48 x 3.5 x 0.7 = 117.6 watts
Power input from AC line (assuming 95% efficient LED ballast) = 123.8 watts
Overall system efficiency = 65.9 lm/W

Going through the same thing but driving the LEDs at 350 mA instead requires about 82 LEDs and has an overall system efficiency of around 82 lm/W.

Bottom line is that LEDs probably need to improve to at least 150 lm/W at higher currents such as 1 amp before replacing T8 tubes will begin to make much sense, at least from an efficiency standpoint. There may be mitigating factors where difficulty of changing lamps, low temperature, or frequent starts make LED a better choice to replace T8 tubes right now, but for most general linear fluorescent lighting, I think we're at least 5 years away. LEDs will probably be viable for replacing CFLs in 2 or 3 years however.
 

jtr1962

Flashaholic
Joined
Nov 22, 2003
Messages
7,506
Location
Flushing, NY
4100K T5's with a CRI greater than 90 are exceptional for data centers and offices.
All of the common T5s I've seen have a CRI of 86 regardless of color temp. There just don't seem to be lower or higher grades. T8s generally come in 2 commodity grades-one with CRI in the high 70s and the other in the mid 80s, with >90 available for a premium, and generally only for CCTs around 5000K-5500K.
 

blasterman

Flashlight Enthusiast
Joined
Jul 17, 2008
Messages
1,802
Typical four-foot T8 tube = 2850 lumens

That's higly optimistic in a non VHO tube - trust me. The main problem with T8's (and spiral CFLs to be fair) is strikeback absorption, which can suck a huge percentage of the light's emission. You need a very efficient and optimized reflector, one that basically you can see yourself shaving in to hit that efficiency. The big advantage with T5, and linear CFL is thinner tubes and less strikeback. Hence, it's easier to optimize a reflector and actually use all the photons you generate.

Regardless, T5 is the king right now for commercial fluorescent fixture design and efficiency, and T8 is only being sold for legacy fixtures. T5 bays are even displacing HID right now, and doing so at an industrial scale.

I had a site that was selling 90+ CRI T5's, but it isn't working now, so 85CRI / 4100k is your best bet in any tube size. It's a good working color temp.
 

2xTrinity

Flashlight Enthusiast
Joined
Dec 10, 2006
Messages
2,386
Location
California
About the only advantage that could be offered by LEDs is if they are on motion detector control. LEDs are much "happier" being short-cycled than any kind of fluorescents, and if the area that needs to be lit is only occasionally occupied, running these on motion detectors woudl both save power (due to the light being completely OFF much of the time) also, LEDs have the potential to be more optimally focused, and save money by ONLY throwing light where it's absolutely needed. However, probably no commercial LED products on the market today live up to these hypothetical potentials.

I still believe T5 is probably a better way to go. Many of these fixtues have motion detectors as well, and the best way is probably to set this but with a somewhat longer delay before turnoff -- that way oy're not cycling them every few minutes, but they will NEVER be on for hours at a time with nobody around. You may also consider high-output fixtures. Efficiency is slighly lower, but since each tube has double the brightness, and the tubes themselves are physically smaller, you can light the datacenter to acceptable illuminances with fewer total tubes, leading to less money spent during maintenance.
 

jtr1962

Flashaholic
Joined
Nov 22, 2003
Messages
7,506
Location
Flushing, NY
That's higly optimistic in a non VHO tube - trust me. The main problem with T8's (and spiral CFLs to be fair) is strikeback absorption, which can suck a huge percentage of the light's emission.
That's raw tube lumens. I accounted for fixture efficiency later in my calculation (I think 80% is reasonable for an average T8 fixture). I do know that T5 fixtures using highly reflective mirrors can exceed 95% fixture efficiency. For the LED fixture I assumed 100% since Crees emit all of their light over a half sphere (and hence no need for optics of any kind). Diffusers would obviously hurt all fixtures equally, so no need to account for that here.
 
Top