LEDs waste 75% as heat

[MikeAusC]

"the theoretical ηL limit of white LEDs (260–300 lmW−1)."

"White light emitting diodes with super-high luminous efficacy"
Nitride Semiconductor Research Laboratory, Nichia Corporation, 491 Oka, Kaminaka, Anan, Tokushima 774-8601, Japan

http://iopscience.iop.org/0022-3727/43/35/354002/

So if an LED was producing 300 lm/Watt of electrical input, it would be working at the maximum theoretical limit of white LEDs !!!!

[uk_caver]

So if an LED was producing 300 lm/Watt of electrical input, it would be working at the maximum theoretical limit of white LEDs !!!!

I suppose it's worth clarifying whether 'white LED' covers any white LED, or just blue+phosphor ones.

And also even if the theoretical maximum efficacy of blue+phosphor was ~300lm/W, that could presumably square with a potential ~350lm/W LER.

[Kinnza]

Hi, Stephen! Thank you for pointing me to that discrepancy.

I checked it, and your values are correct, it is the official Photopic function of CIE, V(λ), approved on 1931 (built in 1924). It is the one used to calibrate all photo detectors, so it is the adequate to calculate LER (as lm emission are calculated according to it).

I realized I was using the Vm(λ) CIE function, approved in 1988 (built in 1978). Actually, it represent more accurately the human response, which is higher in the blue region, below 460nm, than the original function. But I have to change it in my spreadsheets, as it overestimates LER of short wavelengths (from 460nm and up, they are identical). Due all the standards were built for the older, CIE refused to change the standard and just endorsed the new function as more "physiologically accurate" for research purposes.

But your values, still being the official ones, have a problem. They are the values for the peak wavelength, not for the range until the next peak. LER value for 450nm is just for 450nm, while 451, 452, 453,454 have higher values. As you get SPDs usually in wavebands, if you use a 5nm wavebands, as 450-455nm, 455-460nm, etc, results are less accurate. For balanced spectrums, probably difference will be little, as in the blue underestimate lm and in the red, overestimate them.

This is the table with 1nm figures. If you use 5nm wavebands, it is better to average them in groups of 5. But anyway, for LEDs is good to perform calculations on 1nm. They are just the coefficient for each nm, to obtain LER you need to multiply it by 683.002 (I had V(λ) in text format, now I have to import it to Excel)

Wavelength V(λ)

360 0.0000039170000 361 0.0000043935810 362 0.0000049296040 363 0.0000055321360 364 0.0000062082450 365 0.0000069650000 366 0.0000078132190 367 0.0000087673360 368 0.0000098398440 369 0.0000110432300 370 0.0000123900000 371 0.0000138864100 372 0.0000155572800 373 0.0000174429600 374 0.0000195837500 375 0.0000220200000 376 0.0000248396500 377 0.0000280412600 378 0.0000315310400 379 0.0000352152100 380 0.0000390000000 381 0.0000428264000 382 0.0000469146000 383 0.0000515896000 384 0.0000571764000 385 0.0000640000000 386 0.0000723442100 387 0.0000822122400 388 0.0000935081600 389 0.0001061361000 390 0.0001200000000 391 0.0001349840000 392 0.0001514920000 393 0.0001702080000 394 0.0001918160000 395 0.0002170000000 396 0.0002469067000 397 0.0002812400000 398 0.0003185200000 399 0.0003572667000 400 0.0003960000000 401 0.0004337147000 402 0.0004730240000 403 0.0005178760000 404 0.0005722187000 405 0.0006400000000 406 0.0007245600000 407 0.0008255000000 408 0.0009411600000 409 0.0010698800000 410 0.0012100000000 411 0.0013620910000 412 0.0015307520000 413 0.0017203680000 414 0.0019353230000 415 0.0021800000000 416 0.0024548000000 417 0.0027640000000 418 0.0031178000000 419 0.0035264000000 420 0.0040000000000 421 0.0045462400000 422 0.0051593200000 423 0.0058292800000 424 0.0065461600000 425 0.0073000000000 426 0.0080865070000 427 0.0089087200000 428 0.0097676800000 429 0.0106644300000 430 0.0116000000000 431 0.0125731700000 432 0.0135827200000 433 0.0146296800000 434 0.0157150900000 435 0.0168400000000 436 0.0180073600000 437 0.0192144800000 438 0.0204539200000 439 0.0217182400000 440 0.0230000000000 441 0.0242946100000 442 0.0256102400000 443 0.0269585700000 444 0.0283512500000 445 0.0298000000000 446 0.0313108300000 447 0.0328836800000 448 0.0345211200000 449 0.0362257100000 450 0.0380000000000 451 0.0398466700000 452 0.0417680000000 453 0.0437660000000 454 0.0458426700000 455 0.0480000000000 456 0.0502436800000 457 0.0525730400000 458 0.0549805600000 459 0.0574587200000 460 0.0600000000000 461 0.0626019700000 462 0.0652775200000 463 0.0680420800000 464 0.0709110900000 465 0.0739000000000 466 0.0770160000000 467 0.0802664000000 468 0.0836668000000 469 0.0872328000000 470 0.0909800000000 471 0.0949175500000 472 0.0990458400000 473 0.1033674000000 474 0.1078846000000 475 0.1126000000000 476 0.1175320000000 477 0.1226744000000 478 0.1279928000000 479 0.1334528000000 480 0.1390200000000 481 0.1446764000000 482 0.1504693000000 483 0.1564619000000 484 0.1627177000000 485 0.1693000000000 486 0.1762431000000 487 0.1835581000000 488 0.1912735000000 489 0.1994180000000 490 0.2080200000000 491 0.2171199000000 492 0.2267345000000 493 0.2368571000000 494 0.2474812000000 495 0.2586000000000 496 0.2701849000000 497 0.2822939000000 498 0.2950505000000 499 0.3085780000000 500 0.3230000000000 501 0.3384021000000 502 0.3546858000000 503 0.3716986000000 504 0.3892875000000 505 0.4073000000000 506 0.4256299000000 507 0.4443096000000 508 0.4633944000000 509 0.4829395000000 510 0.5030000000000 511 0.5235693000000 512 0.5445120000000 513 0.5656900000000 514 0.5869653000000 515 0.6082000000000 516 0.6293456000000 517 0.6503068000000 518 0.6708752000000 519 0.6908424000000 520 0.7100000000000 521 0.7281852000000 522 0.7454636000000 523 0.7619694000000 524 0.7778368000000 525 0.7932000000000 526 0.8081104000000 527 0.8224962000000 528 0.8363068000000 529 0.8494916000000 530 0.8620000000000 531 0.8738108000000 532 0.8849624000000 533 0.8954936000000 534 0.9054432000000 535 0.9148501000000 536 0.9237348000000 537 0.9320924000000 538 0.9399226000000 539 0.9472252000000 540 0.9540000000000 541 0.9602561000000 542 0.9660074000000 543 0.9712606000000 544 0.9760225000000 545 0.9803000000000 546 0.9840924000000 547 0.9874182000000 548 0.9903128000000 549 0.9928116000000 550 0.9949501000000 551 0.9967108000000 552 0.9980983000000 553 0.9991120000000 554 0.9997482000000 555 1.0000000000000 556 0.9998567000000 557 0.9993046000000 558 0.9983255000000 559 0.9968987000000 560 0.9950000000000 561 0.9926005000000 562 0.9897426000000 563 0.9864444000000 564 0.9827241000000 565 0.9786000000000 566 0.9740837000000 567 0.9691712000000 568 0.9638568000000 569 0.9581349000000 570 0.9520000000000 571 0.9454504000000 572 0.9384992000000 573 0.9311628000000 574 0.9234576000000 575 0.9154000000000 576 0.9070064000000 577 0.8982772000000 578 0.8892048000000 579 0.8797816000000 580 0.8700000000000 581 0.8598613000000 582 0.8493920000000 583 0.8386220000000 584 0.8275813000000 585 0.8163000000000 586 0.8047947000000 587 0.7930820000000 588 0.7811920000000 589 0.7691547000000 590 0.7570000000000 591 0.7447541000000 592 0.7324224000000 593 0.7200036000000 594 0.7074965000000 595 0.6949000000000 596 0.6822192000000 597 0.6694716000000 598 0.6566744000000 599 0.6438448000000 600 0.6310000000000 601 0.6181555000000 602 0.6053144000000 603 0.5924756000000 604 0.5796379000000 605 0.5668000000000 606 0.5539611000000 607 0.5411372000000 608 0.5283528000000 609 0.5156323000000 610 0.5030000000000 611 0.4904688000000 612 0.4780304000000 613 0.4656776000000 614 0.4534032000000 615 0.4412000000000 616 0.4290800000000 617 0.4170360000000 618 0.4050320000000 619 0.3930320000000 620 0.3810000000000 621 0.3689184000000 622 0.3568272000000 623 0.3447768000000 624 0.3328176000000 625 0.3210000000000 626 0.3093381000000 627 0.2978504000000 628 0.2865936000000 629 0.2756245000000 630 0.2650000000000 631 0.2547632000000 632 0.2448896000000 633 0.2353344000000 634 0.2260528000000 635 0.2170000000000 636 0.2081616000000 637 0.1995488000000 638 0.1911552000000 639 0.1829744000000 640 0.1750000000000 641 0.1672235000000 642 0.1596464000000 643 0.1522776000000 644 0.1451259000000 645 0.1382000000000 646 0.1315003000000 647 0.1250248000000 648 0.1187792000000 649 0.1127691000000 650 0.1070000000000 651 0.1014762000000 652 0.0961886400000 653 0.0911229600000 654 0.0862648500000 655 0.0816000000000 656 0.0771206400000 657 0.0728255200000 658 0.0687100800000 659 0.0647697600000 660 0.0610000000000 661 0.0573962100000 662 0.0539550400000 663 0.0506737600000 664 0.0475496500000 665 0.0445800000000 666 0.0417587200000 667 0.0390849600000 668 0.0365638400000 669 0.0342004800000 670 0.0320000000000 671 0.0299626100000 672 0.0280766400000 673 0.0263293600000 674 0.0247080500000 675 0.0232000000000 676 0.0218007700000 677 0.0205011200000 678 0.0192810800000 679 0.0181206900000 680 0.0170000000000 681 0.0159037900000 682 0.0148371800000 683 0.0138106800000 684 0.0128347800000 685 0.0119200000000 686 0.0110683100000 687 0.0102733900000 688 0.0095333110000 689 0.0088461570000 690 0.0082100000000 691 0.0076237810000 692 0.0070854240000 693 0.0065914760000 694 0.0061384850000 695 0.0057230000000 696 0.0053430590000 697 0.0049957960000 698 0.0046764040000 699 0.0043800750000 700 0.0041020000000 701 0.0038384530000 702 0.0035890990000 703 0.0033542190000 704 0.0031340930000 705 0.0029290000000 706 0.0027381390000 707 0.0025598760000 708 0.0023932440000 709 0.0022372750000 710 0.0020910000000 711 0.0019535870000 712 0.0018245800000 713 0.0017035800000 714 0.0015901870000 715 0.0014840000000 716 0.0013844960000 717 0.0012912680000 718 0.0012040920000 719 0.0011227440000 720 0.0010470000000 721 0.0009765896000 722 0.0009111088000 723 0.0008501332000 724 0.0007932384000 725 0.0007400000000 726 0.0006900827000 727 0.0006433100000 728 0.0005994960000 729 0.0005584547000 730 0.0005200000000 731 0.0004839136000 732 0.0004500528000 733 0.0004183452000 734 0.0003887184000 735 0.0003611000000 736 0.0003353835000 737 0.0003114404000 738 0.0002891656000 739 0.0002684539000 740 0.0002492000000 741 0.0002313019000 742 0.0002146856000 743 0.0001992884000 744 0.0001850475000 745 0.0001719000000 746 0.0001597781000 747 0.0001486044000 748 0.0001383016000 749 0.0001287925000 750 0.0001200000000 751 0.0001118595000 752 0.0001043224000 753 0.0000973356000 754 0.0000908458700 755 0.0000848000000 756 0.0000791466700 757 0.0000738580000 758 0.0000689160000 759 0.0000643026700 760 0.0000600000000 761 0.0000559818700 762 0.0000522256000 763 0.0000487184000 764 0.0000454474700 765 0.0000424000000 766 0.0000395610400 767 0.0000369151200 768 0.0000344486800 769 0.0000321481600 770 0.0000300000000 771 0.0000279912500 772 0.0000261135600 773 0.0000243602400 774 0.0000227246100 775 0.0000212000000 776 0.0000197785500 777 0.0000184528500 778 0.0000172168700 779 0.0000160645900 780 0.0000149900000 781 0.0000139872800 782 0.0000130515500 783 0.0000121781800 784 0.0000113625400 785 0.0000106000000 786 0.0000098858770 787 0.0000092173040 788 0.0000085923620 789 0.0000080091330 790 0.0000074657000 791 0.0000069595670 792 0.0000064879950 793 0.0000060486990 794 0.0000056393960 795 0.0000052578000 796 0.0000049017710 797 0.0000045697200 798 0.0000042601940 799 0.0000039717390 800 0.0000037029000 801 0.0000034521630 802 0.0000032183020 803 0.0000030003000 804 0.0000027971390 805 0.0000026078000 806 0.0000024312200 807 0.0000022665310 808 0.0000021130130 809 0.0000019699430 810 0.0000018366000 811 0.0000017122300 812 0.0000015962280 813 0.0000014880900 814 0.0000013873140 815 0.0000012934000 816 0.0000012058200 817 0.0000011241430 818 0.0000010480090 819 0.0000009770578 820 0.0000009109300 821 0.0000008492513 822 0.0000007917212 823 0.0000007380904 824 0.0000006881098 825 0.0000006415300 826 0.0000005980895 827 0.0000005575746 828 0.0000005198080 829 0.0000004846123 830 0.0000004518100

[Kinnza]

I'm curious as to why it often seems that for a given LED type, LED vendors have the higher lumen bins available at the cooler end of the spectrum.

[edit]
Is it just that even though the LER is higher, that's more than compensated for by other losses, giving a lower overall efficacy.
If so, is that unavoidable in practice, or something that's theoretically alterable?
[/edit]

You are right, the warmer the tones, the lower the efficiency. The loss in efficiency is higher than the gain due higher LER, thus warmer colors have lower efficacies than coolwhites.

As both uses blue chips of similar performance, the problem is with the phosphor.

First because warmer tones means converting photons to longer wavelengths, thus higher Stokes losses.

Second, they need to convert more blue light (check how blue peak in warm tones are way smaller than in cool ones), achieved by using a larger concentration of phosphor. This result on more losses at the phosphors (they are not 100% efficients) and more light scattering that reduces light extraction efficiency (a larger fraction of light emitted is unable to leave the package).

Manufacturers are working hard on these issues, and each day differences in efficacy between cool and warm whites are smaller.

Better phosphors, with higher quantum efficiency, reducing losses at the phosphor layer (actually, now there is little to improve here).

New ways of using the phosphor, reducing scattering and reduction in light extraction efficiency. For example, remote phosphor systems and phosphors embedded in ceramics.

Another trick which is resulting very useful is using red (or red and yellow,as Cree) LEDs along the white ones. This allows to use cooler LEDs and still get a warm tone, without reducing the efficiency of the white LED. But this trick is basically only valid at the whole luminary level. Some manufacturers (Epistar, at least, reported it for no far production) are trying to implement it into a single LED. Most of the top performance warm white luminaries in the market uses this solution.

[Kinnza]

So if an LED was producing 300 lm/Watt of electrical input, it would be working at the maximum theoretical limit of white LEDs !!!!

Great! You finally got it.

If a LED has a LER of 300lm/W, meaning that emits an spectrum able to produce 300lm per optical watt, at 100% efficiency it will emit 300lm/W.

Im glad you understand it, now we can move along and maybe look for alternative efficiency measurement systems as the one you tried. Always is good to have alternative ways of measuring things.

[slebans]

.

This is the table with 1nm figures. If you use 5nm wavebands, it is better to average them in groups of 5. But anyway, for LEDs is good to perform calculations on 1nm. They are just the coefficient for each nm, to obtain LER you need to multiply it by 683.002 (I had V(λ) in text format, now I have to import it to Excel)

Thank you very much for posting the table, Kinnza. When you get it imported into Excel would you mind Emailing me a copy?

Finally, I was using the value of 683 so with your correction I should be 100% accurate.

Stephen Lebans

[bbb74]

Wow I think I'm getting it, that didn't hurt at all Actually there's probably a few things I don't quite get...

Lets say I invented some new type of device, that magically converts 100% of its input electrical energy into pure 555nm light. Would it produce 683lm per watt of input electricity? And its LER would be ... also 683lm/W? If the device was only 50% efficient, it would produce 683/2 lm/W but the LER would still be 683lm/W?

Lets then say I invented some new type of device, that magically converts 100% of its input electrical energy into "white" light. Would it produce almost ~400lm per watt of input electricity? I know it depends on the cct but what would its LER be roughly in this case? The same? I get stuck here...

The maximum theoretical efficacy of leds of 300lm/W I see about - what is that based on? The fact that LEDs are designed in a certain way (eg blue+phosphor) and there are certain known losses that must occur (eg Stokes) that simply cannot be avoided by improved technology?

Last question ... Kinzza you seem to be able to do some calculations regarding the junction temperature. I was wondering if you could guestimate (or guess) the junction temperature in real world led torches if you could measure the temperature of the body? Eg. Tk41, XM-L at 10 watts. Is the junction temp going to be "nice" down around say 80 degrees, or is it going to be more likely cooking up around the 150 degree mark?

Based on what you've been saying, a moderately driven XM-L has about 30% of input current wasted as heat, but at 3A it will be more like about 50% wasted as heat, right?

[slebans]

I suppose it's worth clarifying whether 'white LED' covers any white LED, or just blue+phosphor ones.

And also even if the theoretical maximum efficacy of blue+phosphor was ~300lm/W, that could presumably square with a potential ~350lm/W LER.

It is not accurate to state that the theoretical maximum efficacy of a white LED is 300lm/W. The statement must be qualified concerning the specific attributes of the LED and its spectral output.

With regard to the document Mike quoted, please note the research was performed in 2009. There are also several issues with the document that are probably related to the translation of the source Japanese document and the desire to protect the company's IP. I am not questioning the science behind the document -those scientists have forgotten more than I know about LEDS.

Stephen Lebans

[uk_caver]

Lets say I invented some new type of device, that magically converts 100% of its input electrical energy into pure 555nm light. Would it produce 683lm per watt of input electricity? And its LER would be ... also 683lm/W? If the device was only 50% efficient, it would produce 683/2 lm/W but the LER would still be 683lm/W?

Yes - the LER is effectively a 'subjective colour efficiency' which doesn't depend on how good the device is at turning electricity into radiation.

Lets then say I invented some new type of device, that magically converts 100% of its input electrical energy into "white" light. Would it produce almost ~400lm per watt of input electricity? I know it depends on the cct but what would its LER be roughly in this case?

The LER in that case would depend only on how 'efficient' the spectrum it produced was - if you like, on the precise kind of white the device produced.

For comparison, it was stated earlier that 'kind' of white light typically produced by current cool white phosphor LEDs has an LER of ~300lm and that by warmer LEDs a bit more, and that doesn't depend on the basic electricity-light power conversion efficiency of the devices.
Taking those figures as correct for the sake of argument, they'd mean that if you could produce a new device that was 100% efficient at converting electricity to light *and* it produced an output spectrum the same as a cool white LED, you'd have an overall luminous efficacy (electricity-to-lumens) of ~300lm/W, or higher if your device mimicked the spectrum of current warmer white LEDs.

If you could make a device that produced an 'efficient white' spectrum rather than mimicking the spectrum of phosphor devices, potentially you could get 400lm/W.

The maximum theoretical efficacy of leds of 300lm/W I see about - what is that based on? The fact that LEDs are designed in a certain way (eg blue+phosphor) and there are certain known losses that must occur (eg Stokes) that simply cannot be avoided by improved technology?

If someone's stating an maximum theoretical efficacy of 300lm/W for a particular device, that will be based on the unavoidable losses in that kind of device and the LER of the light such devices produce.

[Kinnza]

Any given spectrum produces more lumens (higher LER) as larger is the percentage of green light related to the full visible spectrum. On the opposite, it has lower LER as higher is the emissions in blue and red, and especially if such emissions are deeper on both ranges (shorter wavelengths for blue, longer for red).

You can check in the photopic coefficients above how dramatic is the drop the conversion of radiant energy to lm:

While 1W of 555nm (green) produces 683lm, near it, at 600nm (yellow-orange) produces 431lm and at 510nm (cyan) 344lm. And once we enter in blue and red, it drops a lot more yet. For red, 260lm (620nm), 148lm (635nm, typical peak of red LEDs) and 42lm (660nm, deep red). Worse yet for blue, 62lm (470nm), 26lm (450nm, most used peak for white LEDs) and 16lm (440nm) (figures are rounded)

So generally if an spectrum has a broad and large peak around green (plus cyan, yellow, some orange) and little on blue and red (and uses near bands of them), it may obtain "white" light (with an ugly tint, of course) at very high LER (over 450lm/W).

Current LEDs using blue chips and phosphors are somewhat limited about the wavelength range that can be used. Coupling with most phosphors is not good over 460nm, so manufacturers need to use 445-455nm peak (give or take) blue leds (longer ones often are used for warmer tones), and this limits the maximum LER achievable by this technology. But I believe is excessive to put the theoretical limit of this technology at 260-300lm/W of efficacy. But for sure that the authors of that article knows their stuff, so probably I'm unaware of more things that limits it (not related to LER, but to efficiency, probably limits on the external quantum efficiency)

But for sure that statement refers to phosphor based white LEDs. I have read papers designing 4 color bands LEDs surpassing 400lm/W of LER and CRI close to 80. Not using phosphor allows to use longer wavelength blue LEDs and shorter wavelength red LEDs. But this way of obtaining white, apart of the problem of different behavior with changing temperature (color stability problems) for the moment is limited due the low efficiency of green and yellow LEDs. Recently Ive seen proposals of building such 4 band LEDs but using a blue and red one, and two blues with phosphor fully emitting in green and yellow (actually, there is some models of yellow-amber LEDs in the market that are blue chips with phosphor, as the Rebel). But anyway, 4 bands technologies need large improvements on the efficacy of green and yellow LEDs to compete with the simpler blue+phosphor scheme (or the blue with phosphor+red, currently the solution with higher efficacy).

PS: Stephen, I'll do that Excel next week most probably. As soon as I have it completed, Ill send you it. Im thinking on improving it adding the color matching and color coordinates function, so the spreadsheet calculates x,y,z coordinates, u' and v', CCT, CRI and maybe some other color indexes.

[slebans]

PS: Stephen, I'll do that Excel next week most probably. As soon as I have it completed, Ill send you it. Im thinking on improving it adding the color matching and color coordinates function, so the spreadsheet calculates x,y,z coordinates, u' and v', CCT, CRI and maybe some other color indexes.

That would be perfect, Kinnza. I worked through the math last night to convert the color coords to CCT but then got stuck trying to convert to (u,v) and onto to CRI. A couple of hours of math and color theory - my head started to ache!

Stephen Lebans