State of the art white LED efficacy?

[splin]

The most efficient LEDs available at Digikey are 229 lumens/Watt form Everlight though Seoul, Osram and Samsung are close behind. Cree used to regularly boast about setting new efficacy records, but the last was in 2014:

https://www.cree.com/news-media/news/article/cree-first-to-break-300-lumens-per-watt-barrier

303 lumens/Watt (in the lab). So are we likely to see 300+ lumens/Watt in the foreseable future or has development hit a plateau?

Most domestic lighting LED bulbs seem to top out at around 117 lumens/Watt - how much of the shortfall from the best available LEDs is likely to be down to:

1) Using cost effective, lower efficacy LEDs
2) Driver efficiency loss - 10% perhaps?
3) Losses due to light absorbtion by the encapsulation/diffusers etc?
4) Use of lower colour temperature (2700K), lower efficacy LEDs because of customer demand for 'warmer' lighting?

Given that LED bulbs can last a very long time if properly designed, with decent quality drivers, surely there must be some demand for state of the art efficiency bulbs albeit at a somewhat higher price? A decent 1521 lumen, 13W bulb can retail at £10 in the UK; 1521 lumens equates to 40 of these 3500K, 80 CRI LEDs:

https://www.digikey.co.uk/product-detail/en/seoul-semiconductor-inc/S1W0-3030358003-0000003S-00001/897-1400-2-ND/9997401

They would cost £2.12 at Digikey prices. So with markups and taxes that might add £5 to £10 to the cost (with a high efficiency driver included) at retail. I'd be happy to pay £15 for a bulb that is 30% more effiicient, for locations where the lights are on for a long time. 30% CO2 reduction would be welcome and the extra cost could be recovered in 12 to 18 months; clearly you need to expect a decent lifetime from an expensive bulb, but the higher efficiency would help by reducing the thermal stresses.

The 30% more efficient comes from 213 lumens/Watt less 10% driver loss and 15% optical losses = 164 lm/W compared to the 117 lm/W of a 1521 lm, 13W bulb.

Do such bulbs exist?

[mzzj]

add to your list
5. operating temperature: specified at 25C junction temp, real-life situation is closer to 100C
Based on Cree's PCT tool modern medium-power white leds drop lose about 10% in efficiency if junction temperature rises 60C from theoretical marketing value.

E27 light bulb just sucks as a led form factor.
High efficiency led lightning is easy to find in fluorescent tube replacements. Philips has 4000K tubes with 172Lm/W.

[splin]

add to your list
5. operating temperature: specified at 25C junction temp, real-life situation is closer to 100C
Based on Cree's PCT tool modern medium-power white leds drop lose about 10% in efficiency if junction temperature rises 60C from theoretical marketing value.

Good point, but witrh higher efficiency, temperatures will be lower. Samsung LM301B LEDs lose 6% flux at 85C, but forward voltage drops by 3% so efficiency loss should only be 3% from 25 to 85C provided the driver is suitably designed.

E27 light bulb just sucks as a led form factor.
High efficiency led lightning is easy to find in fluorescent tube replacements. Philips has 4000K tubes with 172Lm/W.

That's much better but it can't all be down to the form factor. Perhaps it's because tubes are generally bought by companies who value the electricity cost savings (and maybe additional air conditioning power savings) rather more than domestic customers.

[OwO]

Are we talking about bulbs? as in those that screw into incandescent bulb sockets? If so there is no way you can dissipate 10W in that form factor without heating to near 100 degrees C. Ditch the bulb form factor and you can do far better.

Here in new homes you rarely see bulb based fixtures anymore. It's almost always  big ceiling mounted lights using LED strips internally, with area big enough to keep the LEDs cool.

[DaJMasta]

Don't know if the bulbs exist as what market would they be for?  There's virtually no residential or commercial application where that level of efficiency would be a significant benefit over baseline performance that would use a standard bulb.  The cost would be substantially more because of the price of the LEDs themselves, so with the exception of products emphasizing tiny form factor over almost all else, it's going to be hard to justify the expense.

I do think you can get somewhat better than 90% efficiency on the power supply conversion, though, especially when a string of LEDs can be driven at a fairly high voltage (close to mains).  You can get drivers better than 95% efficient, so if you were looking for power savings in bulbs, it may actually be more economical to optimize the driver and then get the slightly better luminous efficiency LEDs.  The combination of the two could probably shave off a few percent of the total losses from a standard bulb for maybe twice the price, whereas going for the ultimate in LED tech and driver optimization could maybe get a few more percent for several times the price.

Hard to say.  Looks like what's commercially available is to about 230lm/W, but the price for LEDs seems reasonable (~10 cents each in a few thousand quantity).  I wonder if they really just jump on the bargain basement LEDs to make consumer grade LED bulbs.

[T3sl4co1l]

I'm not sure if there are any physical limits to the quantum efficiency, i.e., one photon emitted for every electron crossing the junction.  I know the high index of refraction of the semiconductor is a big problem (total internal reflection for most randomly oriented photons), and they're using patterned surfaces to help that (and probably antireflective layers?).

Perfect lime green is 660 lm/W (or whatever the exact figure is), so white can't possibly be anywhere near that.  A flatter (solar?) spectrum is supposed to be more like 230?  So, anyone claiming anything that high, with phosphors, is full of something.

I wonder if they've developed bi-level or wavelength-doubling phosphors?

Tim

[Siwastaja]

Also note that while the datasheet front page might specify a large lm/W number, they are binned as per efficacy, and of course, the best bins are [nearly] unobtanium.

Cree has a track record of giving press releases of total vaporware, they always have almost x2 ratio between lab and real products. Who knows whether the lab results are true, or total lies? You can't buy them even 5 years after the press releases.

For real products actually available, under realistic operating conditions, we seem to be somewhere around 140-150 lm/W, which, after integrated into a bulb with a mains supply and diffusor, becomes around 100-120 lm/W max. This is still remarkable, compared to the real-world of around 40 lm/W for fluorescent lamps.

[T3sl4co1l]

Yup, and they're nothing to sneeze at, at all.

I'm quite impressed by the newest generation LEDs I bought, to replace the aging LEDs on my hand made fixtures.  The new ones:
- are smaller,
- have better color (warm/neutral white, versus what was labeled as "warm white" but ended up bluish, or purplish after several years' use),
- drop less voltage (about 2.8 vs. 3.6V),
- are much brighter,
- and run cooler (lower Tc).

Of course, doesn't help that the old ones were Chinese specials, like < $0.50/ea (ca. 2013 vintage), while the new ones are close to $1/ea.  Even at Digi-Key prices, they'll pay themselves back over their lifetime.

And this is with a CC supply, so it's not like they're brighter just because Vf is lower.  Actually, I had to adjust my desk lamp to run at lower currents anyway, because it's just too damn bright to use at full power in most room lighting conditions.  (Namely, I had to increase the UC3842's aux supply, so it doesn't brown out and hiccup at low load.  Yeah, it was kinda marginal to begin with.)  I can probably disable the cooling fan on it now, or make it temp activated.  (It's currently temp _limited_, i.e. the output is throttled down at high temperature.)

Tim

[splin]

Are we talking about bulbs? as in those that screw into incandescent bulb sockets? If so there is no way you can dissipate 10W in that form factor without heating to near 100 degrees C. Ditch the bulb form factor and you can do far better.

Nonsense, unless by 'far better' you mean 5% or so at 100C compared to 25C? Except that the better form factor lights still won't be running the LEDs anywhere near 25C - they will use fewer LEDs and drive them harder to cut costs.

More importantly a 100W equivalent LED bulb doesn't have to dissipate 10W - as the efficacy of the LED increases the waste heat drops dramatically. See jtr1962's table in reply#7 here:

https://www.candlepowerforums.com/vb/showthread.php?381202-CREE-about-to-break-300-lumens-per-watt-in-R-amp-D

For a 1700 lumen light: Efficacy, waste heat
100lm/W, 11.8w
125lm/W, 8.4W
150lm/W, 6.2W
175lm/W, 4.6W
200lm/W, 3.3W
225lm/W, 2.4W

Heat is pretty much a non-issue for high efficiency LEDs.

Here in new homes you rarely see bulb based fixtures anymore. It's almost always  big ceiling mounted lights using LED strips internally, with area big enough to keep the LEDs cool.

Bulb fittings are still very common in the UK, but obviously all sorts of light fittings are available, mostly bulb based but a great many use small G8 and G9 bulbs or even integrated non-replaceable LEDs. These are even less efficient but presumably are used to achieve a more aesthetic point source effect.  LED strip based designs are available but not for heat dissipation reasons.  Fashions come and go;  personally I find the strip based lights a bit tacky but what do I know? I could never stand domestic living rooms with several kilowatts of halogen downlighters.   

[Marco]

they will use fewer LEDs and drive them harder to cut costs.

Cheap is good too ... ignoring fixtures for a moment, the flip chip LEDs on large aluminium substrates seem to be able offer a much better balance than bulbs.

[mzzj]

I'm not sure if there are any physical limits to the quantum efficiency, i.e., one photon emitted for every electron crossing the junction.  I know the high index of refraction of the semiconductor is a big problem (total internal reflection for most randomly oriented photons), and they're using patterned surfaces to help that (and probably antireflective layers?).

Perfect lime green is 660 lm/W (or whatever the exact figure is), so white can't possibly be anywhere near that.  A flatter (solar?) spectrum is supposed to be more like 230?  So, anyone claiming anything that high, with phosphors, is full of something.

I wonder if they've developed bi-level or wavelength-doubling phosphors?

Tim

bad luck with led materials or what when eye response and led efficiency vs wavelenght are nearly inverse functions of each.

https://www.nap.edu/openbook/12821/xhtml/images/p2001a18fg70001.jpg

[T3sl4co1l]

Indeed, it's a shame LEDs (direct wavelength) aren't more efficient in that gap.  Which, heh, GaP is indeed your classic green LED material -- it's not very efficient, hence being restricted to indication only.

Indeed, they're so inefficient that modern high-efficiency greens are InGaN blue, either paired with a (much cleaner and more total coverage*) phosphor, or with the phosphor or quantum dots mixed into the junction itself (emitting green directly, but still costing 3V to forward bias).

*Compared to white LEDs, which intentionally use a broad spectrum phosphor (or mix), with partial coverage (some blue is transmitted).

My comment about bi-level or doubling phosphors is (or would be?), rather than discarding the energy difference between blue and yellow (the phosphor is the second hottest component after the chip!), it could generate, say, two red photons, if the energy levels work out correctly.  A deeper (indigo or UV) source could give yellow or green at very high efficiency, and therefore very high efficacy as well.  The remaining gaps (cyan, orange, red) could be filled in, at somewhat lower efficiency, with more conventional emitters and/or phosphors, if maximum CRI is desired.

Tim

[mzzj]

Good point, but witrh higher efficiency, temperatures will be lower. Samsung LM301B LEDs lose 6% flux at 85C, but forward voltage drops by 3% so efficiency loss should only be 3% from 25 to 85C provided the driver is suitably designed.

You know what they say about "typical" specs in datasheets? 
Reading "typical" values off the graph and relying on them to 3% accuracy is bit optimistic to my taste. Could be 30% off as well. 

Even the initial guaranteed values leave some leeway: 
Samsung maintains measurement tolerance of: forward voltage = ±0.1 V, luminous flux = ±5 %

Samsung is also pretty quiet about lumen maintenance (lifetime) of the LM301B.

BTW: Current E27 (A19) replacement bulbs probably run even hotter than 85Cel:
I checked two 10W light bulbs with FLIR E4. Bulbs have been turned on for last 6 hours: One hanging from bare bulb socket measures 101 Cel ! and another brand in large, open, "inverted salad bowl"  luminaire measured at 99 Cel 
Bulbs were rated for 1000...1055 lumens.