LED efficiency

[Srbel]

How many lm/W do you have to have in order to have 100% LED efficiency?

[sleemanj]

Lumens and human vision are a very tricky business, but if we handwave away details, then this random person on he internet says that there are effectively 680 lumen per watt and gives a reference for that number;

http://www.onlineconversion.com/forum/forum_1108372773.htm

Therefore if your led was 100% efficient it would produce by definition 680 lumens per watt.

[Srbel]

Thank you.

[ConKbot]

Note, thats only for 555 nanometer green light.  As soon as you add other frequencies, or a broadband source (I.e. white light) the lm/w going to fall off significantly.

Food for thought though, the daylight sun is 1kw/m^2 maximum, and ~120k lux maximum, or 120k lumens/m^2 or about 120 lumens/w. Though I'm not sure how far into infrared and UV the 1kw/m^2 figure goes, which will really mess up the lm/w figure, but that gives you an idea how the lumens/w relation can change with color temerature.

[ajb]

If you have the emission spectrum for the source and its nominal output in lumens you can calculate its radiated power in watts.  You can then compare that to its power consumption to determine its exact radiometric efficiency.  However, because lumens are defined based on the sensitivity of the human eye which changes with ambient lighting conditions, you must also know what curve was used when the output in lumens was measured.  The principle curves are meant to model photopic vision, which is when the eye is adapted to bright conditions where color vision receptors (cones) are dominant, scotopic vision, which is when the eye is adapted to dark conditions and rods dominate, or mesopic vision where the eye is adapted to intermediate conditions and the responses of both types of receptors are combined in some way.  There are standardized curves available, such as CIE's 'standard luminosity function', although if you're looking at an LED datasheet, I'm not sure if there's a commonly accepted standard for luminosity measurement.  I would think it would be the CIE standard, but I wouldn't be surprised if some manufacturers pick a more favorable curve to give better datasheet numbers.  A quick look at a CREE datasheet didn't find any indication of what curve they use. 

I'm not sure how easy it is to find tables of the standard luminosity functions (you might have to purchase the relevant CIE documents to get them), but if you can tabulate the emission spectrum of the LED and the approximate luminosity function you want to compare against, it's a quick bit of work in Excel to get at least a ballpark conversion between lumens and watts, and once you have the tables, it's easy enough to run the conversion against a different curve.

[Srbel]

I asked because I was wondering about the efficiency of the latest available high power (few Watts) white (blue) LEDs. Efficiency or at least efficacy, in %.
Nobody seems to give such information in the datasheet. They only give lm, and lm/W, and so on. Although, I once saw stated 57,4% efficacy of some Luxeon high-power white LED, but now when I look at the same datasheet (and it was page 5), there is no such data anymore. WTF?

So I thought, if I knew at least rule of thumb value of lm for 1W, I could know the efficiency in %.

[mikerj]

It's wavelength dependant.  This table might be of use to you.

[ajb]

I asked because I was wondering about the efficiency of the latest available high power (few Watts) white (blue) LEDs. Efficiency or at least efficacy, in %.
Nobody seems to give such information in the datasheet. They only give lm, and lm/W, and so on. Although, I once saw stated 57,4% efficacy of some Luxeon high-power white LED, but now when I look at the same datasheet (and it was page 5), there is no such data anymore. WTF?

So I thought, if I knew at least rule of thumb value of lm for 1W, I could know the efficiency in %.

Out of curiosity, why are you so interested in watts/watts for your LEDs?

For the most part, nobody really cares about the watt per watt efficiency of LEDs because total radiated power (which is to say watts) is not a useful way to measure illumination.  Note that illumination is distinct from irradiation because in the former we're generally talking about throwing light at something so that we can see it, which means we need to account for the spectral response of the eye, whereas with the latter we're talking about throwing energy in the form of light at something for some other reason.  Lumens account for the human eye response, so that's why we measure illumination in lumens/area and irradiation in watts/area.  If you were to look at the datasheet for a UV LED, for example, you'd probably see the output listed in W or mW, because generally with those you're throwing energy at some process (killing bacteria, curing adhesives, or whatever) that responds to energy at that particular wavelength (and you don't care about the human eye response because you shouldn't be looking at the thing anyway).

Furthermore, if you're using LEDs for illumination, the efficiency of the emitter itself is only one part of the total efficacy of the finished system.  You also have to account for how much light makes it through your optics (optical efficiency) and then how much useful light makes it to the subject (total efficacy).  Note here the distinction between efficiency and efficacy.  Efficiency deals with getting the same type of thing out as you put in, such as putting in watts of electricity and getting out watts of light, and efficacy deals with getting a different thing out as you put in, such as putting in watts of electricity and getting out lumens.  By definition, efficacy isn't going to be a percentage, because if you have lumens/watts the units don't cancel out like they do if you had watts/watts.  To bring this all back to the beginning, if the ultimate receptor of the light is the human eye, then you need to account for its spectral response, and since that's what lumens do, lumens are the useful thing that you get out of the LED, and watts are a completely meaningless unit for light*.

If you still really want to know the radiometric efficiency of your LED, you can use the table mikerj linked to at least get in the ballpark.  For a single-color LED, you can simply pick a point on the table (or interpolate between two points--note that a few nm in wavelength makes a big difference, especially at the ends of the visible spectrum).  For white LEDs, you'll need to tabulate the emission spectrum from the LED's datasheet, normalize the resulting table as well as the table that mikerj linked, and then multiply the two and sum the result. **

* Unless you have a LOT of light (or a lot less coherent light) and are calculating ocular hazard, in which case watts and/or joules are very important.  While the luminosity function is crucial to determining perceived brightness, energy is all that it takes to cause retinal injury.

** If you have multiple single-color LEDs, things get tricky in the other direction, as the response of the eye to multiple discontinuous wavelengths does not necessarily add up the way that the luminosity function would suggest.  That means that you can't really put the output of RGB LEDs in lumens with much confidence.  There's some interesting work out of CIE on this subject, but last I looked the consensus was basically "it's really complicated".

[Srbel]

All I wanted to know is how many % of electrical power is transformed into optical power (light), by the most efficient high power white LEDs commercially available.

[SteveLy]

All I wanted to know is how many % of electrical power is transformed into optical power (light), by the most efficient high power white LEDs commercially available.

I've read up on this a while ago. So IIRC, it's around 200Lm/W, perhaps a bit more (but <300). But that's state of the art stuff, very expensive. Practical stuff is in the low 100 + few 10s (and even that's pricey). Sorry I have no references. But info is readily accessible; search and you'll find.

[coppice]

All I wanted to know is how many % of electrical power is transformed into optical power (light), by the most efficient high power white LEDs commercially available.

Which white? Warm white? Daylight? Algerian Tropical Daylight (that one really exists as a flourescent lamp colour). The different colours vary quite a bit in visual efficiency, both because the phosphors vary in efficiency, and because their colour differences put more or less energy in the bands where your eyes are most sensitive.

[IanB]

Regardless of all the arguments about how and why lumens are used to measure illumination, it is still interesting to know (from a physics standpoint) what the efficiency of a typical white LED is in terms of radiated power vs input power (ballpark, even). One reason for knowing this is that whatever power is not radiated is presumably dissipated as heat, and this will be relevant to your heat sink design.

[Srbel]

Just a ballpark value is what I want.

I know that incandescent light bulbs go up to 7% or so, fluorescent up to 14% or so, but what about LED?

I thought that the most efficient ones go up to 50%.

[SteveLy]

Just a ballpark value is what I want.

I know that incandescent light bulbs go up to 7% or so, fluorescent up to 14% or so, but what about LED?

I thought that the most efficient ones go up to 50%.

Lumen is a difficult unit because it's calibrated to the human vision. Theoretical maximum for a lightsource is 680 Lm/W. But that's with all the energy going into green light, where the eye is most sensitive.

For white light the figure is around 300 Lm/W. So in the sense you're asking, a super-efficient white LED that puts out 150 Lm/W, would be about 150/300 = 50% efficient. Most commercial LEDs are more like 70Lm/W for cheapies to ~120+Lm/W for relatively (but not insanely) expensive brand names, so you're looking at a range of 70/300 = 23% to 120+/300 = 40+%.

The rest of the energy goes into mostly heat, but some of it also into a bit of IR and UV light.

[Srbel]

Hmm, I am not sure. Look at this: http://www.cree.com/News-and-Events/Cree-News/Press-Releases/2014/March/300LPW-LED-barrier

303 lm/W. Surely, they are not 100% efficient...

[Kjelt]

White is not white, you have cold white 6500K-5000K and warm white 2700K-3000K the first range having much higher efficiency.
The reason is that the led in almost all white leds is blue and that a small phospor plate transforms the blue light into yellow light. The mixture of the yellow and the blue light is the white light, and the colder white uses more blue light so less loss in the transition. These colder white leds are often used in flashlights and bikelights where they show off the amount of lumens but most users rather have a 4000K white light to do work and in the livingroom it is often 3000K and in the kitchen even 2700K (yellowish white).

For 3000K I thought the 100% eff was indeed around 200lm/W and we are not far off now
Indeed the led is only one part of the fixture, the PSU, lenses, diffusors and all drop the efficiency hard.
So there will be no giant leaps anymore in white led efficiency.
That is why prices are dropping and competition is fierce with led makers. It is all about production efficiency now, making as many dies for the $ in the shortest amount of time is where they are doing their research.

[IanB]

According to this paper http://physics.ucsd.edu/~tmurphy/papers/lumens-per-watt.pdf the luminosity of ideal white light (equivalent to the visible light from the sun) is about 250 lm/W of radiated power. This compares to 683 lm/W of radiated power for monochromatic green light at 555 nm where the human eye is most sensitive.

This means that any claim of luminous efficacy greater than 250 lm/W must be for light that is not "white" in the sense of sunlight (and presumably therefore shifted towards longer wavelengths making it redder).

Sunlight has a color temperature of 5800 K and the Cree LED referenced has a stated color temperature of 5150 K, so it is indeed redder than sunlight.

Where I remain puzzled is about the extra input power dissipated as heat leading to the requirement for big heat sinks. I suspect there may be a degree of bullshit taking place with how these claims are being stated. This once more is why the radiant efficiency is important to know. Unless we also know the ratio of radiated power to input power we cannot know the whole story.

(E.g. if an LED emitted 250 lm/W of input power at a color temperature of 5800 K, then it would be 100% efficient and it would not get warm at all. It would need no heat sinking. I highly doubt such "magic" LEDs exist.)

[coppice]

Sunlight has a color temperature of 5800 K and the Cree LED referenced has a stated color temperature of 5150 K, so it is indeed redder than sunlight.

The impressive Cree claim is for a LED running cool and at very low current. When running at high currents their efficiency drops a lot. That warms them up, and the resulting high temperature reduces their efficiency a good deal further. The difference between the impressive claims of what seems like near 100% efficiency, and the more modest efficiencies of lamps in the shops using those LEDs, is big. Trying to maintain high efficiency at high currents is a major area of research right now, and there seem to be some interesting advances happening.

[Kjelt]

Sunlight has a color temperature of 5800 K

Uhhhh , a black body at the suns surface temperature radiates at 5800K.
The color temperature of sunlight depends on the time of day and the atmospheric conditions and can range anywhere from 2000K to 8000K (see picture).

Video (monitors) often use 6500K as their standard white.
Humans prefer to work in environments with a lot of light from 4000K to 6000K.
Humans like to have warmer white in their homes (because usually that is at the end of the day and they then like to have "sunlight" in the color of the end of the day from 2700K to 3000K).

[SteveLy]

Hmm, I am not sure. Look at this: http://www.cree.com/News-and-Events/Cree-News/Press-Releases/2014/March/300LPW-LED-barrier

303 lm/W. Surely, they are not 100% efficient...

They've fudged the figures. Definition of "whiteness", and colour temperature and the watts-to-lumens formulas leave a lot of wiggle room as to the characteristics of a light source - as you may have gathered from the contributions of other posters on this thread.

When you have a team of marketing, engineering and data analysis gurus at your disposal, you can make claims like Cree does and no one can touch you. I bet you those 300Lm/W LEDs would be horrible to live with using them as indoor lighting. They'd distort colours and you'd miss out on some colours altogether. (Normal fluoros and LEDs already do that but I'd expect much at worse for Cree's 300Lm/W.) But they are probably legit on paper and will have useful applications.

If you want as close to natural light as possible ~250Lm/W is the limit at 100% power-to-light conversion efficiency. A poster above linked a PDF about why that is re sunlight being limited to visible range. It's a good one. If you're willing to compromise a bit and go more "green", then 300Lm/W will pass for white light. Any more than that and you'd need to be colour blind to be happy with it for ordinary lighting.

[Srbel]

OK.

But what is the average efficiency (ballpark) of "standard" white 3W LEDs that are readily available? If, for example, I want to know how much Watts will they dissipate, so I can choose an adequate heat sink.

[GNU_Ninja]

How many lm/W do you have to have in order to have 100% LED efficiency?

This may shine some light on your question http://www.hi-led.eu/wp-content/themes/hiled/pdf/led_energy_efficiency.pdf 

[Srbel]

But those are not answers to my question.

[Kjelt]

But what is the average efficiency (ballpark) of "standard" white 3W LEDs that are readily available? If, for example, I want to know how much Watts will they dissipate, so I can choose an adequate heat sink. 

Only correct answer: You have to look that up in the specific datasheet of the specific LED.

I do can tell you that power leds (>1W) have a lower efficiency than medium power leds (<= 0,5W) and that a lot of modern esp. industrial luminaries have a lot of medium powered leds (100mA/150mA @ 3V) in them in a couple of parallel strings of 10 to 12 pieces (so for instance multiple pcbs , each pcb totalling 36+ leds a board running at 36V 250-400mA).
Power leds are for lighting a bit obsolete.
What you do see however is that manufacturers build power leds from clusters of leds inside the led. I call that a led package. So if you look at power ledpackages >3W they consist of tens or even more of extremely small medium power leds in a matrix under a phosphor layer.
All these ledpackages have specific heat transfer characteristics since they are build upon a metal or ceramic carrier.
The only place to find out how much heat should be transferred and what size of heat sink you need is on the datasheet of that specific module.

[GNU_Ninja]

But those are not answers to my question.

There isn't a specific answer to your question. The 'answer' is that it depends. You will have to measure your particular LED for whatever particular set of conditions with reference to its datasheet and find out.

Why bother with a mere 100% check out what the MIT optical group have been doing http://physicscentral.com/explore/action/led.cfm fascinating stuff