How much better will LEDs get?

[Wacki]

It seems like only a few years ago I was amazed by a 120 lumen LED light. Now they have compact lights running 700 lumens or more. My Fenix TK12 simply isn't that old and it's obsolete!

Is there a Moore's law of LEDs? How much better are these things going to get in both the short and long term.

 

[anuragwap]

Haitz's law. For long term, who knows? Maybe >100% efficient as the MIT folks have shown?

If you think more carefully, you don't need to upgrade everytime the latest/greatest bins are out. Most people will have a hard time noticing a 20-30% flux increase on a blind test.

 

[Yoda4561]

There are theoretical limits to how much white light you can extract from x amount of electricity, and further there are apparently limits to how much you can do via a phosphor white LED. As far as I know the absolute limit for a white phosphor led is around 320 lumens per watt, this is only double what we have now in top bin cree's, however the heat output would be 100% less. This lets you disregard current requirements for heatsinking/cooling from designs, and your output will scale much higher with high power levels. As far as practical terms, I doubt we'll ever get to that level, however Cree does have LED's hitting over 260 lumens per watt in controlled lab testing at lower power levels.

 

[idleprocess]

Suspect we'll see 200 ln/W in white going into production eventually. This could mean ~20 lumens via a commodity 5mm/SMD LED or ~2000 lumens via something like a Luminus 10W package.

 

[Illum]

Eventually though, its going to hit the same thermal ceilings that has thusfar blocked manufacturers from making more powerful processors having the same or smaller dimensions than its predecessors. As a result, many manufacturers have done to the dual/quad core architecture. When CREE came out with the MC-E and SSC came out with the P7, I figured that was where the trend was going. But know, who knows :shrug:

 

[zespectre]

There is apparently still a lot of theoretical room left in "nano-scale perforation" to vastly increase the available surface area (and thus the light emitting capabilities) however on the nanoscale light does some very... odd ... things and as mentioned already there is still a thermal issue to contend with, although the same nano-perforation methods may give us a form of "super-heatsink" as well. The theoretical stuff I've read seems pretty solid but of course the jump from theory to practice is often a large one.

 

[Yoda4561]

Eventually though, its going to hit the same thermal ceilings that has thusfar blocked manufacturers from making more powerful processors having the same or smaller dimensions than its predecessors. As a result, many manufacturers have done to the dual/quad core architecture. When CREE came out with the MC-E and SSC came out with the P7, I figured that was where the trend was going. But know, who knows

It's not really a thermal ceiling, it's a power ceiling, they can only push so much current through the wires and nanoscale circuitry before making it go poof. Making a single core faster causes the power requirements to increase exponentially regardless of how cold you keep it. We'll run into the same thing with LED's at higher drive currents, for most applications I don't think it will end up being a major limiting factor. Battery technology probably won't be able to keep up, household A19 bulbs only need so much light and we can already get it today, and large outdoor/commercial indoor lighting can use arrays to increase efficiency and lower current requirements.

 

[uk_caver]

As far as I know the absolute limit for a white phosphor led is around 320 lumens per watt, this is only double what we have now in top bin cree's, however the heat output would be 100% less.

I know it's being a bit pedantic, but for a 'perfect' phosphor-based LED, wouldn't there still be some need for heatsinking (even if relatively small compared to current needs) to take care of heat from the phosphor losses?

 

[jtr1962]

I know it's being a bit pedantic, but for a 'perfect' phosphor-based LED, wouldn't there still be some need for heatsinking (even if relatively small compared to current needs) to take care of heat from the phosphor losses?

Probably not-see this thread where I discussed that subject. This chart I made for the amount of power and waste heat when replacing a 100 watt incandescent with LED is particularly relevant:

For our flashlights power to the LED is really limited by batteries and runtime. This means once LEDs start getting ~80% efficient or better in many cases an mcpcb might provide sufficient heatsinking. If you're outputting 1000 lumens with an 80% efficient LED (~250 lm/W), you'll be putting 4 watts into the LED and getting out about 0.8 watts of waste heat. This might require heatsinking to the flashlight body. Lesser outputs of a few hundred lumens won't. You could mount the LED on an mcpcb in a plastic flashlight body without heat issues.

 

[Yoda4561]

I know it's being a bit pedantic, but for a 'perfect' phosphor-based LED, wouldn't there still be some need for heatsinking (even if relatively small compared to current needs) to take care of heat from the phosphor losses?

At "reasonable" drive currents with similar applications to current LED's, such as flashlight, household, and automotive lighting probably not. At extreme drive currents like in uber high power flashlights and military searchlights then there will be other problems. Say you take an XML sized LED and since 100% efficiency isn't happening, let's do 95. Now you can drive the LED something like 50x harder than today with the same heatsink requirements. The blue light radiation from the LED is likely to instantly vaporize the phosphor and probably the transparent dome as well. (this would be something like 400 watts of light, lasers that can cut clear materials are around 100 or so) http://www.parallax-tech.com/cutting.htm

The bond wires and leads to the LED will likely suffer a similar fate as they're so small. What will happen is larger LED's or arrays with collimated optics so they can keep the phosphors and stuff happy at the needed outputs. I imagine HID or other future technologies will take over for the extreme high output applications.

 

[Canuke]

Don't forget losses in the optics, and potential degradation thereof; that energy is turning into heat also. I've already seen a few folks ask about IR in LED's because they could feel heat during operation, that was actually the light output heating their skin.

 

[Norm]

Previous threads.

Norm