LED inefficiencies: 82% of lighting energy lost as heat

Actually Dave, I've measured parts at a steady state condition that are doing 130 lm/W. One needs to pay careful attention to addressing thermal issues. And they particular ones were not Philips LumiLEDs devices. Recently the bar has been raised and competition has resulting in a couple of LED makers who have exceeded the 100 lm/W number. I did try to order through Future, some of the new Philips LumiLEDs parts which are specified to be greater than 100 lm/W, but they are still not shipping and no idea when they will have them, short some unofficial engineering samples. And it definitely was not Nichia, one of their power LEDs did manage to do 105 lm/W at a steady state die temperature of 41.2ºC and 0.30002 Amps when under powering it a little. But yes, some of their SMT devices will hit 130 lm/W if pulsed, and then it drops quickly due to the packaging thermal resistance of the small edgelight and many keypad LEDs. Many of their package range from 200 to 300 for their ºC/W numbers. However, they do have a couple of ceramic "boat" parts (instead of plastic) that have a lower thermal resistance. I believe it was CREE that was pushing to use the NIST method of steady state measurements, and there were some papers on the same, and CREE was submitting their parts to NIST for independent third party verification. CREE has also been using ceramic (which is nice as they have electrically isolated thermal pads) which has their ceramic package XP-G ~3.45mm device that produces 132 to 139 lumens per watt at 350mA. For lab results, they have hit 161 lm/W. It will be interesting to see that part get on the market. They have a number of parts that exceed 100 lm/W that are available now. OSRAM has been catching up, and certain new parts like OSLON SX are claimed at 73 lm/W, or their new Diamond Dragon which is claimed at 53 lm/W, and their new Golden Dragon Plus which is claimed at 93 lm/W. Yes, it was CREE that was using NIST steady state measurements: compoundsemiconductor.net/cws/article/news/31137 The was done with a 5 minute warm up to steady state, and it had a tiny heatsink, so there wasn't much thermal mass to heat up. They used to put the NIST reports up on their website, and there were pictures in the pdf. I just looked on DigiKey, and even DigiKey has over 4,000 of the older CREE parts that are over 100 lm/W, XPEWHT-L1-R250-00E01, minimum 114lm @ 350mA I see Future has Philips LumiLEDs Rebel devices, LXML-PWC1-0100, which are specified at 100 lm/W minimum, which are probably in your 95lm/W range. Sorry Dave, the best continuous lm/W is above your 70 lm/W numbers, including Nichia- which doesn't have the highest efficiency.

herbster, the efficient incandescent lamp you mention did make it to market is and called an an IRC Halogen bulb (OSRAM make them for example). Such bulbs generate about 25lm/W.

There''s absolutely no argument that LED is more efficient at producing visible light than lowly incandescent bulb -- just what is needed in California. But in certain circumstances the huge infrared radiation emitted by the incandescent lamp is very beneficial -- such as melting snow/ice forming on a car headlamp during a storm, or evenly heating a cold room, being a mini heatlamp. That''s something an LED conduction fixture''s heat dissipated in the ceiling cannot do since the dissipation isn''t radiated -- but I guess you can add a fan into the LED assembly (it''s done). I haven''t seen an LED heatlamp yet... What is best depends on the application. Personally I''m waiting for LEDs to replace compact fluorescents once smart power meters used by utilities start dinging consumer wallets due to CF''s poor power factor...

SteveP''s comments on visible emissions are totally correct but depending on the context may be misguiding the public into demanding LEDs. An LED automotive taillamp assembly (not a resistor + LED) is a complex device due to a design requirement to limit LED temperature over a broad operating temperature range -- lots of LED heatsinking too. It is misleading to think that LEDs do not generate heat. An incandescent taillamp assembly is simple despite the hot bulb. As for being ''green'', in Quebec the power is hydroelectric and the utility has admitted that using incandescent lamps in the winter (the IR warms the entire room like a heatlamp) is cheaper and less polluting than heating with expensive oil or natural gas (it may be different elsewhere where you need airconditioning) and using fancy lamps. So LEDs (3 times more efficient for visible light than incandescent, but they heat the ceiling fixture only) aren''t always better.

An interesting comparison on LED versus incandescent efficiency is found by GOOGLEing the terms "Comparison of Power Conversion of White Light Sources" and "energy" to find a USgovt DeptOfEnergy website LED has 15% radiant emissions versus 80% radiant emissions for incandescent lamp -- the opposite of what you'd expect. An LED creates useless heat that needs to be dissipated by conduction, whereas an incandescent lamp's huge radiant emissions are largely infrared, and not visible light. So an LED wins efficiency for visible emissions, whereas an incandescent lamp wins for total radiant emissions.

In all three of these devices, you are taking electrical energy (moving electrons) and transforming the energy in that movment into electromagentic waves (pure energy). This step in the process has always been the lossy one, and there are physics limits to how much it can be improved. I think, but I can't prove, that the losses in the driver electronics, especially for LEDs, can be reduced to a negligible level compared to the losses in the LED. CFL's will be more of a challenge, as they are essentially high voltage devices. Incandescent lighting really doesn't need electronics, but it probably has been improved about as much as it can be at this time. (2.8 percent was quoted above, but I have seen 1 percent more commonly given for conversion efficiency.) One thing to keep in mind, especially for incandescents and maybe LEDs-- heat is also electromagentic energy. Maybe there is a way to convert this into light with a phosphor, etc.

LED 18% CFL 9-11% Tungsten 2.6% see wikipedia - luminous efficiency

Or, if the source you quote is giving the efficiencies of the individual, cascaded, subsystems then the total efficiency is the product, and the loss is 60% rather than 82%. Can you clarify? The idea that LED's are "cold" is a futurist myth.

Perhaps a more meaningful comparison would be lumens/watt. A typical incandescent bulb is about 14-16 lumens/watt. A CFL is about 60-72 lumens/watt. The newest HP LED''s are running at about 95 lumens/watt. Perhaps another measurement tool would look at the total life cycle cost to include the cost of the luminary.

The way I understand this is if LEDs were 99% efficient, the 1% of the energy lost would be broken down into these percentages. Fairly misleading..what is the absolute, not relative number?

I have wondered for some time now if anyone anywhere in the last 50 years has worked on finding a more efficient emitter for an incandescent bulb

The first three writers asked exactly the first questions that came to my mind as well. I hope EDN will address these definitively soon.

And are LED drivers going to be as heat-sensitive as CFLs? Most of my lighting applications are either enclose or high-power 3-way (50/200/250 W) - CFLs don't last in the former, and don't exist for the latter :^(

For comparison, what is the efficiency of tungsten-halogen or CFL lamps??

How do these losses compare to incandescent? How do these losses compare to CFL?