underlying technology making LEDs better?

[wa5ngp]

Can someone comment on what is the underlying technology improvement that is making LEDs brighter and more efficient?

I understand how line widths of CMOS processes keep going smaller to yield more transistors for processors etc. I think we are at 14nm and going down at the moment. I assume that smaller line widths is not really what is driving the technology improvements of LEDs. So my question is what is the main area of technology that is making LEDs better?

Heatsinking?
improvements in doping of the raw LEDs for increased luminescence? ie secret sauce additives
what else?

don

 

[IMSabbel]

Can someone comment on what is the underlying technology improvement that is making LEDs brighter and more efficient?

I understand how line widths of CMOS processes keep going smaller to yield more transistors for processors etc. I think we are at 14nm and going down at the moment. I assume that smaller line widths is not really what is driving the technology improvements of LEDs. So my question is what is the main area of technology that is making LEDs better?

Heatsinking?
improvements in doping of the raw LEDs for increased luminescence? ie secret sauce additives
what else?

don

Well, it basically boiles down to improving the quality of mass producing the semiconductors.

Cree has published a paper in phys rev b in which a hand picked LED had >80% wall plug efficiency at >mA currents. That is so close to perfect it is crazy. Now, the challange is making 100million/year of them and keeping the same quality. I.e. have all the processing of the LEDs perfected, as well as the nuances of doping, etc.

In that regard, now that the groundbreaking stuff has been handled (i.e. finding the right semiconductors, etc), LED devolepment is following more the normal, evolutionary improvement process of industrial goods compared to the revolutionary of semiconductors (Where we STILL expect 50% density gains every 3 years or so).

 

[saabluster]

There is more to the increases being seen in the industry than merely the improvement in the semiconductor. The main areas in my mind are as follows.

epitaxial
optical
thermal
current density reduction


Epitaxial- Closely guarded secrets that I don't really understand or have insight into but involve the construction of the active light creating layers themselves.

Optical- Improvements in this area range from reducing/removing the current spreaders from the front optical path. Improvements in encapsulation from both a materials perspective but also basic design such as increased dome size and progressive indices.

Thermal- improving the extraction of waste heat allows for higher input current, longer lifetimes, and increased lumen output due to reduction of thermal droop. Improvements in this area can still be made but are happening at a slower pace now than in years past as they have reached somewhat of an impasse in what they are willing to do for cost reasons.

Current density reduction- The more power you put into a given space with white(blue) LEDs the less efficient the LED becomes. This has led to the "band-aid" approach of simply making the LED dies bigger and bigger. It has lead to greater performance when measured at a singular point such as the somewhat industry standard point of comparison of 350mA but has not made a fundamental change in the performance of LEDs.

 

[bose301s]

There is more to the increases being seen in the industry than merely the improvement in the semiconductor. The main areas in my mind are as follows.

epitaxial
optical
thermal
current density reduction


Epitaxial- Closely guarded secrets that I don't really understand or have insight into but involve the construction of the active light creating layers themselves.

Optical- Improvements in this area range from reducing/removing the current spreaders from the front optical path. Improvements in encapsulation from both a materials perspective but also basic design such as increased dome size and progressive indices.

Thermal- improving the extraction of waste heat allows for higher input current, longer lifetimes, and increased lumen output due to reduction of thermal droop. Improvements in this area can still be made but are happening at a slower pace now than in years past as they have reached somewhat of an impasse in what they are willing to do for cost reasons.

Current density reduction- The more power you put into a given space with white(blue) LEDs the less efficient the LED becomes. This has led to the "band-aid" approach of simply making the LED dies bigger and bigger. It has lead to greater performance when measured at a singular point such as the somewhat industry standard point of comparison of 350mA but has not made a fundamental change in the performance of LEDs.

Epitaxial, that's a big one :) ;)

 

[bshanahan14rulz]

The GaN crystal grown for LED substrates must have a very precise crystalline structure for the LED to be grown onto. I think the active layer is grown on the GaN, and the crystalline structure of the GaN determines how the active layer's constituents arrange themselves. Through some sort of mathmagic, they probably have a molecular structure that they are trying to produce. Or maybe they just get to play with fun stuff in the lab until a certain blend works better? Although Cree always talks about its G*SiC technology, so that must be part of it too. I think I remember SiC having nice thermal attributes, that can't hurt either.

I hear their air circulation system at the factory has installed new chip-weevil countermeasures, so that may very well come into play as well

 

[wa5ngp]

Things get really wierd in FABs. I heard a story about a FAB where yield went south. After all the engineers gave up on figuring out why, the mgmnt got employees together and brainstormed about what has CHANGED? One comment was that they had started putting FRITOS in the vending machine. Sure enough, they then traced the problem back to increased sodium contamination. That's why the new 300mm processor FABs are all sealed up and have no humans inside. Seems like magic until you get to the bottom of things.

I see now that LEDs are definitely on a different trend line than logic and CMOS. Does anyone know what is the largest wafer currently used?

best regards.
Don