NOTE - This is an archive/historical copy of this page as of just before I pared out older matter on 5/11/2008. This note was added to this page on 7/22/2008 and is the only editing of the "5/11/2008 version" of this page. Other than this note, the following update date is valid.
Latest update 5/11/2008.
Back up to my LED main page, ledx.html.
News and links! UPDATE 12/22/2006
General Notes and Disclaimers
Red LEDs - 46
lumens/watt
Red-Orange LEDs - 53
lumens/watt
Truly Orange LEDs - 18-22
lumens/watt
Yellow and "amber" LEDs -
34-35 lumens/watt, with a way for 40-60 for less-amberish yellow.
Yellow-Green LEDs - 8
lumens/watt, but there is a way to get a way to get 80.
Green LEDs - 58-79
lumens/watt.
Nichia green LEDs get 90-120 lumens/watt at 3 milliamps
Blue-Green LEDs - 50-75
lumens/watt
Blue LEDs - 17 - 26 to
optimistically 30 lumens/watt
White LEDs - 60-100
lumens/watt.
Increased Efficiency of InGaN LEDs at lower currents
Important Notes on Gallium Nitride and InGaN LEDs!
LEDs with high candela "beam candlepower" figures
LEDs with high total lumen output
Notes on Where To Get Bright LEDs
Converting / comparing lumens, candelas, millicandelas
UPDATE 2/1/2007 - Lumileds has issued this 1/23/2007 press release announcing a prototype single chip LED at 350 mA, roughly 1.2 watts achieving 115 lumens/watt.
UPDATE 12/22/2006 - Nichia has issued on 12/20/2006 this press release announcing achievement of a laboratory prototype white LED with an overall luminous efficacy of 150 lumens per watt.
UPDATE 6/20/2006 - Cree has announced in this press release that they have demonstrated a white LED achieving an overall luminous efficacy of 131 lumens per watt.
UPDATE 11/4/2004 - Cree has announced on 10/12/2004 blue LED dice (chips) that produce 27 milliwatts of blue light having a dominant wavelength of 460 nm at 20 mA with a typical forward voltage drop of 3.2 volts. This works out to a conversion efficiency or "wallplug efficiency" of 42 percent! This is an announcement of sampling of chips that efficient, and commercial production of ones that produce 24 milliwatts at 20 mA (37.5% efficiency).
Relevant press release: HERE.
Relevant datasheet: HERE.
UPDATE 11/3/2004 - A link to check at Sandia labs for LED news!
UPDATE 11/2/2004 - WITH THANKS TO FANS OF THIS WEBSITE:
In August 2004, Osram announced a prototype of an orange-red LED that achieved 108 lumens/watt, producing over 2 lumens at 10 mA with a voltage drop presumably of about 2 volts.
Look in This Osram Press Release.
UPDATE 12/1/99 Agilent (now Avago) has improved models working in the lab. Odd
shaped LED dice such as a truncated inverted pyramid let more light out than
rectangular ones by reducing the amount of light trapped by "total internal
reflection". Such dice are now used in some Lumileds "Luxeon" red, orange and
yellow LEDs - namely those with the "Lambertian" radiation pattern, which are
their "High Dome" ones.
Look in
http://www.photonicsonline.com/read/nl19991130/39722
for details including claims of overall luminous efficacy as high as 102
lumens/watt for orange and 68 lumens/watt for orange-amber.
Compound Semiconductor magazine (the May/June 2000 issue) has mentioned surface texturing to increase efficiency. This seems to be another approach to reduce total internal reflection trapping of photons. Total internal reflection is a serious issue in many LEDs due to the extremely high refractive index of some LED materials.
You definitely want to look at Craig Johnson's LED Site.
IMPORTANT NOTE Many laboratory prototypes mentioned above are not yet in production even if years old! Barriers apparently include but may not be limited to patents held by competitors! Also, gallium nitride based LED chips just may not be cheap (substrate material is generally about 9 on the Moh's hardness scale where diamond is 10). Production of chips of size favoring a more optimum current density may price such things out of the market more than improve efficiency.
Some brands mentioned here as achieving efficiencies less than those of the best mentioned here but without recent updates mentioned may have since improved.
Luminous efficacies were largely determined by one or more of several methods,
mostly almost laughably crude. I give NO warranty as to accuracy. I have at
times slightly changed my estimates for some models. Your mileage may vary.
Unless otherwise noted, luminous efficacy is stated for a 25 degree Celsius
(77 degree F.) ambient with a current of 20 milliamps in an environment
that does not build up heat around the LED. Red, orange, yellow, and
yellow-green models (other than high power / heatsinkable ones) mentioned
here generally have maximum efficiency at currents around 20-25 milliamps.
Efficiency of these is typically less at low currents of a few milliamps or
less. The decrease is not as bad with GaAlAsP and "T.S." AlGaAs and Avago's
most efficient InGaAsP red ones as it is with most other InGaAsP (orangish
red through yellow-green).
Indium gallium nitride types (most ultrabright blue, blue-green,
non-yellowish green, and white types) usually have efficiency that
increases with decreasing current, although the increase will reverse as
current decreases below something like 2 mA or a fraction of rated current.
Please beware that performance of many LED models at currents below a few tenths
of a milliamp or a percent or so of maximum rated current may be unreliable.
There may be significant tolerance in light output. Your mileage may vary.
Some of these LEDs radiate light of slightly different colors in different directions. This is most apparent with white models and narrow-beam yellow models, as well as the most efficient yellow-green models.
PLEASE NOTE that the lumen per watt figures in this document are lumens of light produced per watt of electricity delivered to the LED. Some LED manufacturers state much higher figures indicating the lumens per watt of radiated light. The latter, high figure is typically in the ballpark of 65-100 for red, 130-180 for orange-red, 220-265 for red-orange, 440-500 for yellow, 400-plus to near 660 for green (656 for Avago HLMP-C115), and around 50-90 for blue. The main purpose of this "high" figure is for conversion between radiometric and photometric units for the emitted light. If the conversion efficiency and the "high" lumen/watt figures are both known, multiply them to get the lumens out per watt in.
Beam brightness and width figures below are mostly ones claimed by manufacturers/distributors and I have NOT confirmed most of these.
UPDATE 10/30-2005 - I recently purchased some Avago HLMP-ED33-SVOOO LEDs from Digi-Key (catalog number 516-1390-ND). These achieved a good 28 lumens per watt, possibly 30. With this efficiency advancing so slowly in recent years and tight competition between at least some brands in achieving this efficiency in recent years, I do expect some LEDs of some other brands to achieve similar efficiency.
Avago HLMP-E*** types mostly come in 2-3 brightness-sorted subtypes.
UPDATE 10/9/2003 - I just received some samples of ETG's ETG-5TS630-30. Results: Around 28 lumens/watt.
UPDATE 7/4/2003 - best I have seen so far for high power red LEDs: *Test Results* Lumileds LXHL-MD1D Luxeon Stars (5 pieces tested) test as averaging 35 and give or take a few lumens/watt at 325 mA. The figure I am using for lumens per radiated watt may be low for these. The average for this product may indeed be more like 42 lumens/watt.
UPDATE 7/5/2001 - Lumileds datasheets for their "lambertian"/"high dome" red Luxeons indicate that they produce 30.6 lumens minimum, 44 lumens typical at 350 mA with a 2.95 volt typical voltage drop. This is 29.6 lumens/watt minimum (assuming the typical voltage drop), 42.6 lumens/watt typical. I expect overall luminous efficacy to be even a little higher at somewhat lower currents around 200 mA. This model uses those truncated inverted pyramid dice mentioned above.
The "Low Dome"/"Batwing" is not as efficient and produces less light. The high dome but not the low dome ones use truncated inverted pyramid dice.
In my experience, all red LEDs achieving near or over 25 lumens/watt have a dominant wavelength (color specification) close to 625 nm and the peak wavelength is probably typically close to 635 nm.
Other bright red LEDs:
UPDATE 1/4/2001 - a few various Toshiba orangish red LED lamps with dominant wavelength around 626 nm seem to get around 14 to maybe sometimes 15 lumens/watt. This is for some samples sent to me by Craig Johnson. Models of samples that did about this good include TLRH180P and TLRMH151P.
Toshiba TLRME20T, Hosfelt 25-405 $1.35 (2003 catalog), 12.5-14 lumens/watt, supposedly 8 candela. Not as orange as many other Toshibas.
Toshiba TLSE20T, Hosfelt 25-406 $1.35 (2003 catalog), not tested, interpolated 14-15 lumens/watt. Very orangish red, supposedly 9 candela.
Toshiba's TLSH180P, Hosfelt 25-359, US$ .79 (2003 catalog), 8 candela 6-8 degree beam VERY orangish red more like red-orange - 15-16 lumens/watt! Similar but untested is AND's AND180HSP (Newark 52F6144).
Other red LED lamps:
Avago HLMP-8103, 9 lumens/watt, 3 candela, 7-8 degree beam
Avago HLMP-C116, 9 lumens/watt, 2 candela, 14 degree beam.
Avago HLMP-C124, 9 lumens/watt, 24 degree beam.
Radio Shack's 276-086A 8 lumens/watt, 10 mm. jumbo lamp size, 5 candela,
4 degree beam, peak wavelength 665 nm, dominant wavelength 648-650 nm deep
red, conversion efficiency 12-13%
All other Radio Shack red LEDs available in or prior to early 2002 had
overall luminous efficacy averaging 8.5 lumens/watt or less. I have not
seen any improvements since as of 2007.
Chicago Miniature CMD53SRC/E, approx. 8-8.5 lumens/watt. (October 1997)
Chicago Miniature CMD53SRD/G, approx. 8.5-9 lumens/watt. Claimed 1000 mcd
60 degree beam but I found the beam to be narrower. Consistency was
notably and unusually good. (October 1997)
Red LEDs with GaAlAsP and Avago's similar AlGaAs in my experience, when
doing well, tend to achieve 8-9 lumens per watt.
AND's AND180HRP 8-8.5 lumens per watt. Claimed peak wavelength 644 nm, approx.
dominant wavelength 632 nm. Newark Electronics catalog number 92F2646.
Toshiba TLRH190P (untested) presumed similar in color and efficiency. 10
mm jumbo, claimed brightness 15 candela, Hosfelt catalog number of 25-339,
price US$ 3.49 according to their 2000-A catalog.
Others with luminous efficacy around 8 lumens/watt or less:
Panasonic LN261CAL(UR)/Digi-Key P408ND
Toshiba TLRA191P / Hosfelt Electronics cat. no. 25-246.
Toshiba TLRA180AP / Hosfelt Electronics with cat. no. 25-249.
Electronic Goldmine G9601, approx. 5 lumens/watt.
UPDATE 7/5/2001 - The Lumileds datasheets for their "lambertian"/"high dome" red-orange Luxeons indicate that they produce 39.8 lumens minimum, 55 lumens typical at 350 mA with a 2.95 volt typical voltage drop. This is 38.5 lumens/watt minimum (assuming the typical voltage drop), 53 (corrected 5/25/2003) lumens/watt typical. I expect overall luminous efficacy to be even a little higher at somewhat lower currents around 200 mA. These use those truncated inverted pyramid dice mentioned above.
The "Low Dome"/"Batwing" is not as efficient (typically 25 lumens/watt at 350 mA, probably more at 200 mA) and produces less light. The high dome but not the low dome use truncated inverted pyramid dice.
UPDATE 9/4/2000 - Lumileds HPWT-PH00, formerly a Hewlett Packard part number. The datasheet says 3.75 lumens typical at 70 mA with a 2.5 volt voltage drop which works out to 21.4 lumens/watt. I expect the same for HPWT-BH00, HPWT-DH00, HPWT-MH00, and HPWT-RH00 (whether Lumileds or Avago).
The datasheet has a set of curves that indicates light output at 25 mA is about 44 percent of that at 70 mA. I think these curves are a bit off in the low current end (I know efficiency actually decreases as current is further decreased to just a few mA) but it seems reasonable to expect 40 percent of the 70 mA light output at 25 mA. The voltage drop at 25 mA is 2.1 instead of 2.5 volts, so you get 40 percent of full light at 30 percent of full power. 4/3 of 21.4 is 28.5 lumens/watt.
There are even greater red-orange LEDs at Lumileds - the HPWS-FH00 and the HPWS-TH00. According to the datasheet, the standard grade produces 6 lumens minimum and the highest projected-available grade (HPWS-FH00-N4000 and HPWS-TH00-N4000, supposedly available 2002) supposedly will produce 10 lumens minimum at 150 mA with a typical forward voltage drop of 2.55 volts. The high grade figures work out to 26.1 lumens/watt minimum and the 2000-available standard grade works out to 15.7 lumens/watt minimum. I expect significantly higher efficiency at currents below the 150 mA maximum, such as 60-110 mA.
As for a T1-3/4 / 5 mm. model Avago model? This is the HLMP-DH16 and related models. My tests indicate an overall luminous efficacy ranging from 17.5 to 22 lumens/watt. The HLMP-DH16 supposedly has a 4000 mcd 15 degree beam.
Note that many of these LED lamps are sorted into brightness grades with different suffix portions of the part number.
Other orange LEDs worth mentioning:
Hosfelt Electronics 25-359, Toshiba TLSH180P tests as 15-16 lumens/watt. Hosfelt calls it red but I found it to be red-orange. Beam is 6-8 degrees and 8 candela at 20 mA.
Toshiba's TLOA190P(WX), a 10 mm diameter jumbo orange LED with an overall luminous efficacy of roughly 14 to maybe 15 lumens per watt. Hosfelt Electronics 25-276, US$ 3.49 each according to their 2003 catalog. They have a very narrow main beam that is roughly 2.4 degrees square with a dim spot in the center (my finding). The beam brightness at 20 mA seems to meet the claim of 36 candela.
Radio Shack's 276-206 10 mm jumbo has an overall luminous efficacy of roughly 12-13 lumens per watt. It has a somewhat irregular main beam that is roughly a 2.5 by 3.7 degree rectangle with a claimed brightness of 12 candela at 20 mA.
Toshiba TLOA180AP, 10-11 lumens/watt, with a claimed 3 candela, 18 degree beam, Hosfelt Electronics Cat. 25-277, US$ .75 according to their 2001-A catalog.
Note that the above orange LEDs have a red-orange color, with dominant wavelengths generally around 610 to 617 nm. Some of these sometimes look more red than orange.
Avago has the HLMP-DJ** LEDs, which are truly orange. The dominant wavelength is supposedly 605 nm, which is about that of an NE-2H neon lamp, and less red than the color of most other orange LEDs. The HLMP-DJ08 supposedly has a 9.5 candela 6 degree main beam.
UPDATE 5/27/2000 - HLMP-DJ24, 18-21 lumens/watt! The color was extremely slightly more red than that of an NE-2H neon lamp and much less red than that of other orange LEDs. The HLMP-DJ24 is a 23 degree model with a typical beam intensity of 2 candela.
Also note that these brightest HLMP-D*** part numbers are obsoleted and the current ones have HLMP-E*** part numbers.
The loss is greater if filtering to an orangish or "amber" shade of yellow suitable for traffic signals. Therefore, such orangish yellow LED lamps should be made with amber/yellow LEDs rather than by filtering white ones, especially when longer life or better aging characteristics are required.
UPDATE 7/5/2001 - Lumileds datasheets for their "lambertian"/"high dome" amber Luxeons indicate that they produce 23.5 lumens minimum, 36 lumens typical at 350 mA with a 2.95 volt typical voltage drop. This is 22.7 lumens/watt minimum (assuming the typical voltage drop), 34.8 lumens/watt typical. I expect overall luminous efficacy to be even a little higher at somewhat lower currents around 200 mA. This model uses those truncated inverted pyramid dice mentioned above.
The "Low Dome"/"Batwing" is not as efficient (typically 20 lumens/watt) and produces less light. The high dome but not the low dome use truncated inverted pyramid dice.
UPDATE 5/27/2001 - TESTED - Toshiba TLYE20T, Hosfelt Electronics 25-408, $1.45 each in their 2003 catalog, 2 pieces tested 14 and 15.5 lumens/watt. Claimed brightness 9.5 candela.
UPDATE 1/4/2001 - I just did tests on a Marktech 54E38xx (apparently a Toshiba model) sent to me by Craig Johnson. My tests indicate a good 16 lumens per watt overall luminous efficacy by that sample. This seems to have about a 15 degree beam probably around 10 candela at 20 mA.
UPDATE 9/4/2000: Avago's HPWT-ML00-00000 and related types may well typically get 14.5 lumens/watt at 25 mA as well as 11 lumens/watt at their maximum current of 70 mA. Typical light output is 2 lumens at 70 mA and probably .8 lumen at 25 mA. Typical voltage drop is 2.6 volts at 70 mA and 2.2 volts at 25 mA. The datasheet shows curves for light output as a function of current. They show 44 percent of the 70 mA value at 25 mA but I think they are a bit optimistic on low current and I think more like 40 percent at 70 mA. These LEDs have efficiency actually decrease when current is reduced further to a few mA.
Then-Agilent's "amber" HLMP-DL16 has an overall luminous efficacy of about 13
lumens per watt accodring to my tests. This is a 5 mm. model with a
supposedly 4.5 candela 15 degree beam.
Note that these are now by Avago and have HLMP-EL** part numbers and at least
some of these come in 2-3 brightness sorted subtypes with different suffix
portions of the part numbers.
There are amber Avago "autoLEDs", HPWT-DL00 and HPWT-ML00. The data sheet works out 11 lumens per watt for a typical overall luminous efficacy. The typical light output is supposedly 2 lumens at 70 mA and the typical voltage drop at 70 mA is supposedly 2.6 volts. Overall luminous efficacy is probably maximized closer to 13 lumens/watt at lower currents of 20 to 40 mA.
Other notable yellow LEDs:
Toshiba's TLYH180P and the suspiciously similar AND AND180HYP both have luminous efficacy around 11 lumens/watt and 8 candela 6-8 degree beams.
Agilent HLMP-DL08 and HLMP-DL10, 9500 mcd 5 mm. lamp size, probably around
11 lumens/watt. (Now Avago HLMP-EL* part numbers with suffix-noted
brightness-ranked subtypes.)
Toshiba TLYH190P, supposedly 23 candela beam which is about 2.4 degrees
roughly square. Hosfelt Electronics catalog no. 25-342, $3.49 according
to the 2003 catalog. Overall luminous efficacy only about 6 lumens/watt.
Radio Shack wide angle yellow LED lamp 276-302, approx. 10 lumens/watt. 500 mcd 40 degree beam. Very orangish amber-yellow, dominant wavelength approx. 594 nm.
Radio Shack 276-301, roughly 6 degree square beam with a dim spot in the center (my findings), a few thousand mcd, approx. 7 lumens/watt. Very orangish amber-yellow, dominant wavelength approx. 594 nm.
Radio Shack 276-350 widebeam yellow 5 mm lamp - preliminary findings of 5.5-6 lumens per watt overall luminous efficacy.
Radio Shack 276-205 10 mm jumbo, about 4-5 lumens/watt. The beam is irregular and roughly a 2.5 by 3.7 degree rectangle and supposedly 6 candela.
Some Toshiba LEDs and AND ones that look like Toshibas have an overall luminous efficacy around 3-4 lumens per watt. These are very yellowish green LEDs with a dominant wavelength around 570-573 nm:
Toshiba TLGA183P, Hosfelt Electronics catalog number of 25-341, US$ .99 according to their 99-B catalog, beam approx. 7-8 degrees, claimed brightness 3 candela.
AND's AND183HGP, Newark Electronics catalog number 92f2644.
Toshiba TLGE185EP, Hosfelt Electronics 25-366 has a better defined beam of supposedly 3.5 candela (I think 2-2.5) and maybe about 15 degrees but no higher efficiency. $.99 from Hosfelt according to their 2003 catalog.
LEDs of similar color but with overall luminous efficacy similar to, maybe slightly higher than that of white ones can be made by suitably adding a suitable fluorescent dye to white or blue ones. That means about 20-80 lumens/watt! If you put a white or blue LED into a piece of green-fluorescing yellow acrylic this will probably work! Go Here for LED Hacking for Fun and Danger.
UPDATE 11/2/2004 - Cree XL7090 green "XLamps" typically produce 34 lumens at 350 mA with a typical voltage drop (according to the XB900 chip datasheet) of 3.4 volts. This works out to 28.6 lumens/watt. Other green LEDs with Cree chips have already achieved 33 lumens/watt in my tests.
Now for "low power" green LEDs:
UPDATE 6/14/2004 - Cree has announced on 5/4/2004 "XThin" green LED dice
(chips) that produce 9 mW at 20 mA with a typical forward voltage drop of 3.2
volts. I expect the overall luminous efficacy from this to be 65-70 lumens per
watt.
The relevant Cree press release is here.
UPDATE 1/1/2008 - I tested several units of Nichia NSPG510AS, NSPG520AS and NSPGF50AS. My tests indicate at 20 mA average forward voltage drop of 3.05 volts and light output ranging from 3.7 to 4.4 lumens. The datasheet says typical forward voltage drop is 3.2 volts. This indicates 58 to 72 lumens per watt overall luminous efficacy.
UPDATE 4/30/2001 TESTED 10 pieces of Nichia NSPG-500S of brightness rank T (supposedly typically 14 candela) - 37.5 lumens/watt average!
Optical output ranged from 5.05 to 5.7 mW, averaging 5.4 mW at 20 mA. Voltage
drop at 20 mA was 3.4 volts - conversion efficiency was 7.4 to almost 8.4
percent, averaging 7.9 percent. These LEDs had a slightly yellower shade of
green than average for Nichia color rank G green (525 nm dominant wavelength)
- estimated luminous efficacy of the emitted light 475 lumens/watt (corrected
from 500 1/1/2007). This works out to an overall luminous efficacy of 35
to 40 lumens/watt, averaging about 37.5.
YOUR MILEAGE MAY VARY both in conversion efficiency and in the luminous
efficacy of the emitted light.
UPDATE 1/15/2001 on apparently Nichia NSPG-500S pieces purchased elsewhere:
HOSFELT ELECTRONICS (1-800-524-6464) 25-377, total of 14 pieces from three different orders at various times in 2000 ranged from 25 to 40-41 lumens/watt, averaging 31 lumens/watt.
BG MICRO (1-800-276-2206) LED-1050 - 58 pieces from three batches ordered at various times from early July to late December 2000 had overall luminous efficacy ranging from 27 to 39 lumens/watt, except for two "duds" which got around 24-25 lumens/watt.
The better ones are probably best obtained from Nichia's sales office.
UPDATE 5/27/2001 - Toyoda Gosei has improved versions in the works. If this is their only improvement since they made the ones below, then their greens should soon get about 21 lumens/watt. But I think this is an underestimate since they probably already had improvements since 1999.
UPDATE 12/25/2007 - I just got a batch of NSPG510AS units that average 72 lumens/watt at 20 mA. At 3 mA, they had a forward voltage drop averaging 2.6 volts and produced on average .94 lumen, for an overall luminous efficacy of 120 lumens/watt. Other similar vintage green Nichia LEDs had efficiency about 3/4 that of these good ones in my testing.
Now for "low power" LEDs:
UPDATE 7/30/2000 - TESTED BG MIcro LED-1052 narrowbeam bluish green Nichia NSPE590S - preliminary figure 24-26 lumens/watt on the single sample ordered. This model has an irregular beam with a small central bright pattern (roughly 7 by 4 degrees) with a claimed typical brightness of 20,800 mcd at 20 mA.
TESTED Hosfelt Electronics 25-376, apparently a Nichia model. Tentative figure for overall luminous efficacy 24-27 lumens/watt. Optical power output 7 mW at 20 ma with a 3.75 volt voltage drop - 9.3 percent conversion efficiency! Expect even better efficiency in the future since the voltage drop will usually be 3.5-3.6 volts.
Cree and Avago also make bluish green LEDs suitable for traffic signals.
UPDATE 7/17/2007 - Cree XRE blue LEDs in their flux rank of J0 produce minimum 23.5 maximum 42 lumens at 350 mA with a voltage drop of typically 3.3 maximum 3.9 volts. This works out to minimum 17.2, highside 30.7 lumens per watt at 25 C junction temperature.
Now for "low power" blue LEDs:
UPDATE 1/3/2008 - UNTESTED, SIGHT UNSEEN: Nichia has new blue LEDs about to become available or just now/recently becoming available. They mostly have a revision letter 1 beyond that of their long-standing blue ones. An example is NSPB500AS, which is the improvement of NSPB500S and is claimed to have a beam intensity of 11 candela at 20 mA.
According to published millicandela figures and my extrapolation from my testing of related white Nichia LEDs, it appears to me that output is roughly tripled. With a voltage drop of typically 3.2 volts instead of 3.5 volts, efficiency is increased by a factor of *very roughly* 3.3 (and this is an extrapolation of mine, not an actual test result). With NSPB500S achieving probably about 8 lumens per watt, I expect the new models to achieve about 25 lumens per watt. I expect optical output at 20 mA to be about 22 mW, for a conversion efficiency around 34 percent.
UPDATE 11/4/2004 - Cree has announced a claim of actual commercial production of blue LED dice (chips) producing 24 milliwatts of optical output at 20 milliamps with a typical forward voltage drop of 3.2 volts. This works out to an efficiency of 37.5 percent. Even with a very conservative estimate of 55 lumens per radiated watt for this, the overall luminous efficacy works out to 20.6 lumens per watt. Cree claims to be producing even more efficient samples that produce 27 milliwatts at 20 mA, which would achieve 23 lumens per watt.
An LED manufacturer that I am in contact with said around the end of 2004 that yields of dice this good are still low, so do not expect good availability of blue LEDs this good for a little while...
Relevant Cree website items:
http://www.cree.com/ftp/pub/CPR3BV.pdf, the relevant datasheet.
http://www.cree.com/News/news207.asp, the relevant 5/6/2004 press release.
UPDATE 6/29/2002 - tested a sample LED with a Cree C470-XB290-E1000 die. This LED lamp was an early sample of ETG ETG-5XB470-30. Optical output of the single sample received was approx. 6.5 mW at 20 mA, (UPDATE 1/13/2004 - that was with my solar cell that ran low at shorter wavelengths; actual output could have been 7-7.5 mW) with a voltage drop of 3.6 volts. Based on prior experience testing LEDs with Cree dice, I expect an improvement of at least 20% and more likely 25-30-plus percent within a few months. In fact, the relevant Cree datasheet claims a typical optical output of 14 mW at 20 mA for this die/chip.
This early sample had efficiency very nearly usual for Nichia, and the likely prospect is that efficiency will soon exceed that of Nichia. Also consider that ETG appears hellbent on significantly beating Nichia prices.
Also consider that Cree, unlike Nichia, produces reliably available quantities of LED dice with the 460 nm deeper blue wavelength in addition to the usual 470 nm wavelength. 460 is better than 470 for use in LED video displays - 460 nm has greater "blue color impact" useful in an RGB device than 470 nm despite 460 nm having lower photometric performance than 470 nm. Photometric output of an RGB device is normally almost entirely from stimulating green and red receptors in the human eye rather than blue, so blue LEDs with greater "blue impact" and purer shade of blue and producing more radiometric output can easily have inferior photometric specifications despite their superiority. One thing to look for is photometric output after multiplying by the ratio of Z to Y CIE chromaticity figures.
Cree sells dice (chips) and not LED lamps (whole, ready-to-use LEDs). (NOTE - Cree now sells LED lamps that have high power dice.)
UPDATE 1/1/2002 - TESTED - Nichia NSPB-500S pieces of brightness rank T (4.2 candela at 20 mA supposedly typical for NSPB-500S pieces with this rank) and color rank W (usual shade of blue with a rated typical dominant wavelength of 470nm) obtained in April and in early December 2001 consistently produce approx. 7.5 mW at 20 mA (conversion efficiency generally around 10.5-11 percent) percent) and have an overall luminous efficacy of generally at least 7.5 and more likely 8 lumens per watt.
Nichia NSPB-520S, rank WT pieces (45 degree beam as opposed to 15 degree) obtained in early December 2001 tested as producing approx. 6 percent more light than the above NSPB-500S pieces.
UPDATE 5/16 and 5/1 2001 - I have some Nichia NSPB-500S pieces of their rare/obsolete color rank V (deeper blue, dominant wavelength around 460 nm)and the somewhat uncommon color rank X (greenish blue, dominant wavelength in the upper 470's). The V pieces had conversion efficiency averaging at least 10 percent with some around 11 percent. The X pieces got about 11 percent, and they were of the extreme brightness rank of U (typically 6 candela for NSPB-500S) which only the X color rank of Nichia blue achieves (as of then). The XU pieces in my hands got 9.5-10 lumens per watt of overall luminous efficacy. Note that Nichia blue LEDs of color rank other than the usual W have experienced low and spotty production, are not always in stock at Nichia sales offices, and are in danger of being discontinued. Figures here are for use at 20 mA and expect about 30 percent more efficiency and 40 percent more overall luminous efficacy at 3-4 mA.
BRIGHT AND EFFICIENT BLUE LEDs I HAVE TESTED:
Notably, Nichia NSPB-500S pieces of color rank W (the usual one) obtained from BG Micro, catalog number LED-1051. Based on three separate orders in the second half of 2000, I was getting average overall luminous efficacy of at least 6.5 lumens/watt. The worst piece was about 5.5-6 and the best was about 7.5 lumens/watt, possibly approaching 8 lumens/watt.
Avago claims that typical luminous efficacy of the emitted light of their blue LEDs of 470 nm peak wavelength and 472 nm dominant wavelength is 75 lumens per emitted watt. The typical figure for Nichia blue LEDs of their usual color rank W is probably slightly less - I assume 70.
NOTE - Nichia's official figure for typical optical output power of their blue LEDs at 20 mA is 6 mW as of June 2000. With a typical voltage drop of 3.5 volts and luminous efficacy of the emitted light assumed to be typically 70 lumens/watt, the typical overall luminous efficacy would be 6 lumens/watt.
TESTED UPDATE 7/12/2000 - Radio Shack special-order item 900-8005, 2 pieces, apparently Nichia NSPB-500S which contradicts the catalog claim of 450 nm blue, one approx. 6 lumens/watt with the other approx. 7 lumens/watt.
MINOR UPDATE 6/25/2001: Radio Shack 276-316 blue:
Claimed 2600 mcd 30 degree beam. Beam angle is definitely around 18 degrees and not 30, and the beam is more like an 8 degree beam in a 16-18 degree ring than a solid 18 degree beam. Photometric results - around 1-1.5 lumens/watt overall luminous efficacy. The chip is apparently a somewhat-below-average Cree C470-CB290-E1000.
Toyoda Gosei 15 degree models with really impressive beam candela figures have a ring-plus-blotch pattern rather than a more solid beam.
Avago is now making 470 nm blue LEDs. So is Panasonic, and someone told me they are using Toyoda Gosei chips.
Panasonic 470 nm blue LEDs are now available from Digi-Key.
Radio Shack's 276-311 blue LEDs and other 430 nm blue LEDs are nice and actually close to "high brightness", but are not nearly as bright as others mentioned here. Some of these are referred to as 466 nm blue LEDs since Cree, the manufacturer of the chip, specifies the dominant wavelength at 466 nm. (I find it closer to 460 nm eyeball estimate.) The largish spread between peak and dominant wavelengths is a result of spectral distribution being a violet-blue peak on a primary band that is broader than that of most blue LEDs, and having a very broad "tail" on the long wavelength side. I have seen this spectrum with a diffraction grating, and it has appeared to me as a 428-430 nm peak standing out from a very broad band throughout most of the visible spectrum.
The Cree blue chip used in these is their C430-CB290-E1000, or "standard blue".
At 700 mA and 25 degree C thermal pad temperature, the typical light output is 180 lumens and the typical voltage drop is 3.4 volts. This works out to 75.6 lumens per watt typical.
Light output at 700 mA is typically about 1.65 times that at 350 mA. Typical voltage drop at 350 mA is 3.15 volts. This is according to the "DS56" datasheet. 75.6 * .7 / .35 * 3.4 / 3.15 / 1.65 is 98.9 lumens/watt.
UPDATE 1/3/2008 - Lumileds "Luxeon K2 with TFFC" in the top brightness rank available at Future Electronics as of 1/3/2008 is the LXK2-PWC4-0180. It is rated to produce a minimum of 180 lumens at 1 amp with a typical voltage drop of 3.65 volts and a maximum voltage drop of 4.71 volts when the "junction temperature" is 25 degrees C. This works out to 49.3 lumens/watt with typical voltage drop and 38.2 lumens/watt worst case. With thermal resistance of 5.5 degrees C per watt, this means temperature of the heatsinkable surface would be 5 degrees C for a typical unit and -1 degree C for the worst case unit. At heatsinkable surface temperature of 25 degrees C, the typical unit has performance reduced about 3% to ~174.6 lumens minimum and ~47.8 lumens/watt, while the worst case unit has a roughly 4% decrease in performance for ~173 lumens minimum and ~36.7 lumens/watt minimum.
One notable performance figure is typical light output at the absolute maximum current of 1.5 amps: 250 lumens at junction temperature of 25 degrees C. With a heatsinkable surface temperature of 25 degrees C, expect typically about 5% less, or about 237 lumens.
Another notable bit of data: performance at 350 mA. Light output is typically 85 lumens and voltage drop is typically 3.3 volts, for an overall luminous efficacy of typically 73.6 lumens per watt when junction temperature is 25 degrees C. When the heatsinkable surface is 25 degrees C, expect typically ~1.4% less, for 83.8 lumens and 72.6 lumens/watt.
There is a higher brightness rank version expected to become available in the future, LXK2-PWC4-0200. It will produce about 11% more light and have about 11% higher efficiency than the now-available LXK2-PWC4-0180.
UPDATE 1/3/2008 - Cree XRE "cool white" LED has a maximum brightness rank of "Q5" and Cree's website shows three order "kits" available. Light output at 350 mA according to material published by Cree is 107 to 114 lumens at 350 mA for this top rank when junction temperature is 25 degrees C. Forward voltage according to the datasheet is 3.3 volts typical, 3.9 volts maximum at 350 mA. Thermal resistance is 8 degrees C per watt, and the datasheet indicates a roughly 2.6% decrease in light output for a 10 degree C increase in temperature.
All of this indicates that with heatsinkable surface temperature of 25 degrees C, light output at 350 mA is about 104 lumens minimum with typical voltage drop, 103 lumens worst case, and 111 lumens maximum with typical voltage drop. Luminous efficacy with 25 degree C heatsinkable surface temperature works out to 90-95.5 lumens/watt range-of-typical and 75.4 lumens/watt worst case.
These LEDs can take more current, although overall luminous efficacy will be less at higher current. The absolute maximum current is 1 amp, and light output at 1 amp is typically about 2.2 times that at 350 mA for same junction temperature. Typical voltage drop at 1 amp is 3.7 volts according to the datasheet. Temperature rise of the junction above the heatsinkable surface works out to 29.6 degrees C, which appears to me to be good for a roughly 8% decrease in light output compared to that achieved with junction temperature of 25 degrees C. This means roughly 216-231 lumens at 1 amp, and typical overall luminous efficacy of 58-62 lumens per watt at 1 amp if the heatsinkable surface is at 25 degrees C.
UPDATE 5/11/2008 - Nichia NS6W083A with 150 mA current typically produces 46 lumens with a typical voltage drop of 3.2 volts. This works out to typically 96 lumens/watt. This LED is a version with a less-bluish but still "cool" shade of white, with a claimed typical chromaticity that is close to that of 5,000 Kelvin.
The main high power LED page of Nichia is:
http://www.nichia.com/product/led-smd-powerled.html
This page notes as a footnote that some specific current (as of 5/11/2008 150 mA for NS6W083A) is that at which "100 lumens/watt can be achieved".
NOTE - All high power white LED data above is from datasheets and datasheet interpretations and not personally experienced actual test results.
UPDATE 1/3/2007 - Seoul Semiconductor "P4" W42180U, W42182U, (top brightness rank of "U"), available from Mouser Electronics, using dice/"chips" by Cree: 91 lumens minimum, 100 lumens typical at 350 mA at heatsinkable surface temperature of 25 degrees C, and voltage drop of typically 3.25 volts, maximum 3.8 volts. This works out to 68.4 lumens/watt minimum, 87.9 lumens/watt typical.
These LEDs have absolute maximum current of 1 amp. Light output at 1 amp is typically 2.4 times that at 350 mA according to a graph in the datasheet. Typical forward voltage drop at 1 amp is about 3.6 volts according to a graph in the datasheet. This is at heatsinkable surface temperature of 25 degrees C according to the datasheet. This means supposedly 218 lumens low side 240 lumens typical at 1 amp, with luminous efficiency 60.5 lowish side to typically 66.6 lumens per watt at 1 amp and heatsinkable surface temperature of 25 degrees C.
I have actually tested these and on a *preliminary basis* got about 90 lumens at 350 mA. Due to tolerances of my testing methods done so far, I cannot say with any confidence that they were as indicated actually close to barely minimum specified performance, and they could easily have done better.
UPDATE 9/11/2006 - Citizen Electronics AKA "Cecol" has an 8-chip white LED lamp (CL-652 type) rated to typically produce 85 lumens from 1.24 watts. That works out to 68.5 lumens/watt. My tests on samples that I have obtained indicate about 80 lumens from 1.24-1.25 watts (64 lumens/watt).
Update 9/11/2006 - Nichia has an announced "high efficacy" white LED claimed to typically produce 60 lumens from 1.14 watts. This works out to about 52.5 lumens/watt. That model is NS6W083.
Now for "low power" white LEDs:
UPDATE 1/3/2008 - Nichia has a 5 mm white LED lamp, NSPW500DS, that is similar to NSPW500BS except with higher output and a nominally narrower beam - and a beam intensity of typically 27 cendela at 20 mA. Preliminary testing of related units with "D" revision letter (NSPW510DS, NSPWF50DS) indicate 40-55 lumens/watt - they may actually achieve slightly more.
UPDATE 2/19/2002 - Someone at Nichia told me sometime in January that their white LEDs now have a typical overall luminous efficacy of 15-20 lumens per watt. I assume this is at 20 mA.
UPDATE 7/31/2000 and 9/26/2000 TESTED Radio Shack 276-320 two pieces. Preliminary figure 14-16 lumens/watt. Beam is supposedly 100 degrees but looks more like 60. Rated candela is 1.1. The package says made in China but the guts appear to be Nichia. The beam has a blue ring within the broader blurred white beam, worse than usual for white LEDs that are known to be made by Nichia as opposed to maybe someone else using Nichia chips and leadframes. The color overall seemed more blue or violet-blue than usual, with color temperature estimated 7500K for one and 8000K (maybe more) for the other.
UPDATE 10/9/2002 - Nichia white LEDs tested below at 15-16 lumens/watt came in a bag indicating that they were of "R" rank. Seems to me that most Nichia white LEDs now are of "S" rank and should do a little better. I am not expecting a full 40% improvement on that 15-16 lumens/watt just yet, even though the center of the "S" rank is about 40% higher than the center of the "R" rank.
UPDATE 9/26/2000 re-evaluation, original preliminary tests 7/22/2000:
Tested: five white BG Micro LED-1044 pieces, apparently Nichia NSPW-500BS. Preliminary UPDATED results: 11, 16, 16, 18, and 18 lumens per watt at 20 mA. The 11 I consider a dud and the average of the others is 17.
UPDATE 10/31-99 - Nichia NSPW-500BS models ordered mid-october 1999 test as 15-16 lumens/watt at 20 mA, meeting their mid-1999 claim of 15 typical. The then-current NSPW-500BS nominally had a 5.6 candela 20 degree beam.
UPDATE 9/14/99 - Hosfelt Electronics 25-363, which is apparently a Nichia model with a claimed brightness of 4 candela. It tested at 10-11 lumens/watt.
Hosfelt Electronics 25-367 is very nearly as efficient as the 25-363. I am not sure of the manufacturer. This is a 10 mm. lamp with a supposedly 7 candela 10 degree beam. I think that 7 candela is achieved in a noticeable, very bluish and bright ~5 degree central hot spot. Outside this bluish hot spot, the beam is non-bluish white and nowhere near 7 candela. Seems about 9, possibly 10 lumens/watt at 20 mA with good prospects of higher efficiency at a few mA.
The above white LED lamps have blue chips plus a phosphor to convert some of the blue light to yellow light of broadband spectral content from mid-green to mid-red. The color is usually a cool "daylight" slightly bluish or sometimes violetish-bluish white, usually of about 6000-6500 K.
Efficiency continues to increase with decreasing current at least down to 4 mA and is close to peak around 2 mA and remains above the 20 mA efficiency at least down to .5 mA. Efficiency and light output become less predictable below .4 mA.
White LEDs have most of their photometric output from a phosphor, and so the luminous efficacy of emitted light does not improve with decreasing current as much as with colored InGaN LEDs. In fact, spectral shifting of the LED chip's output away from wavelengths best-utilized by the phosphor can impair overall luminous efficacy at low currents. However, photometric output per milliamp seems to usually improve with decreasing current down to about 4 milliamps. No solar cell measurements on white ones at low current yet, but I think output at 5 mA should be 30-32 percent of that at 20 mA for an overall luminous efficacy of 1.25 to 1.35 times that at 20 mA.
Important Notes on Gallium-Nitride LEDs!
1) Blue, blue-green, bright non-yellowish green, and other gallium nitride
and indium gallium nitride type LEDs have a voltage drop around 3-4 volts,
usually 3.4-3.6 volts at 20 mA, higher than the 1.6 to 2.4 volts typical of
other colors.
2) Gallium nitride and indium gallium nitride LEDs are fussy, and do not
like their ratings to be exceeded. Exceeding 5 volts reverse voltage even with
low current is supposedly damaging to them. I suspect that destructive
electrolysis is what goes wrong. Peak forward currents in excess of 100 mA
may also be bad. These LEDs are also supposed to be considered static sensitive.
These are the first well known keychain LED flashlights and may still be the brightest. Available in all colors.
Brightest 5 mm LED lamp - the blue-green Nichia NSPE-590S, typical beam candela at 20 mA 20.8 for ones manufactured around or since late 1999. The bright central portion of the beam is a "dumbbell" of two squares attached at the corners, with overall angular dimensions of roughly 4 by 7 degrees. The two squares are maybe 2.5-3 degrees wide.
Avago has bright T1-3/4 (5 mm.) orangish red, red-orange and orange models with 6 degree beams and brightness 8 candela minimum 31 candela maximum at 20 mA (for the brightest bin sort subtypes).
Opto Technology makes 40-chip TO-66 lamps that can take 8-10 watts.
Norlux makes 40-chip heatsinkable lamps.
Lumileds, a joint venture of Agilent and Philips Lighting, makes high power heatsinkable "Luxeon" LEDs, in wattages as high as 5 watts, some of which produce over 200 lumens.
Cree and Nichia are getting into this game with ~1-1.2-2-3 watt LED lamps.
Nichia has a link here.
Osram has their "Golden Dragon" high power LEDs.
United Epitaxy has a heatsinkable 10 watt red LED lamp that produces over 200 lumens.
Lamina Ceramics produces heatsinkable high power LED arrays. Some of their arrays take over 100 watts and are of size 1.06 by 1.25 inches (31.75 by 26.67 mm) - and this obviously needs major heat sinking. Please consider thermal resistance of the device and light output as a function of junction (LED "die" or "chip") temperature!
Some Avago, Chicago Miniature, and most AND (apparently Toshiba) LEDs are available from Newark Electronics. Call 800-4NEWARK or visit http:/www.newark.com. They are in a position to collect sales tax if you live in or order from any U.S. state with a sales tax. Minimum order is US$ 25. In addition, the minimum order for most Chicago Miniature Lamp items is 10 pieces of each type. Shipping is extra and duplicate shipping charges apply if you order more than one type of item not all in stock in the same warehouse. Higher minimums may apply to non-stock merchandise - often 500-1000 pieces of any non-stocked Avago LED lamp and maybe even for some stocked ones, and this applies with some other Avago distributors as well.
Future Electronics is the main distributor of Lumileds products and is a distributor of other LEDs.
Mark Hannah Surplus is a hobbyist-friendly unofficial supplier of Lumileds products.
Hosfelt Electronics sells some bright Toshiba LEDs and some other interesting LEDs.
Phone: 800-524-6464, 888-264-6464, 614-264-6464
FAX: 800-524-5414, 614-264-5414
Web: http://www.hosfelt.com/index2.htm
or http://www.hosfelt.com
Orders shipped to Ohio addresses subject to 6.5 percent sales tax. No minimum
order for prepay or credit card orders, shipping is extra. Prepay orders must
include $6.50 for shipping, any excess will be refunded. Some items mentioned
on this web page may be out of stock or discontinued - check with Hosfelt,
not me for stocking status.
Hosfelt has some typo errors - some blue LEDs have lowish supposed voltage drop figures like 2.6-3 volts, but expect about 3.5 volts at 20 mA. Some strange beam angle figures have appeared in their web site and some of their catalogs - 29.5 and 28.5 degrees. Possibly someone saw the "two-theta-half" symbol which resembles 28 1/2 as the actual number of degrees. These look like typos, maybe on the part of a datasheet from a supplier, and are not attempts on Hosfelt's part to give falsely optimistic specifications.
Nichia LEDs are best obtained from their sales offices. In North America, check out Nichia America.
Toyoda Gosei makes bright green and blue LEDs.
Toyoda Gosei LED info is now on the web at http://www.toyoda-gosei.co.jp/led/e-index.html.
Michael W. Sforza's Custom Model Light Service (e-mail lightforce1@rcn.com) also sells bright LEDs. Stock is limited, selection is limited but growing, and prices seem reasonable. E-mail for prices and inquiries. This business specializes in sales of smaller to medium quantities of LEDs and customized addition of lights to building, railroad, vehicle and other models.
A lumen is defined as the "luminous flux" of 1/683 of a watt of monochromatic
light that has a frequency of 540 terahertz, or a wavelength of approx. 555.5
nm.
One thing worth noting is that a lumen is defined secondarily, in terms of
the candela (which is 1 lumen per steradian), and the candela is defined
primarily (it's the "beam candlepower" of 1/683 watt per steradian of 540 THz
monochromatic light.)
Light of wavelengths other than 555.5 nm have a different amount of lumens
per watt of radiation. The number of lumens in a watt of wavelength other
than 555.5 nm is 683 times the photopic function of
the wavelength in question, divided by the photopic function of 555.5 nm
(which I believe is very close to but not exactly 1).
A "USA-usual" 100 watt, 120 volt, 750 hour "regular" (A19) lightbulb usually produces 1710 lumens.
Lumens per watt is a measure of efficiency in converting electrical energy to light. Multiply this by the watts dissipated in the LED to get lumens. A typical red, orange, or yellow or yellow-green LED has a voltage drop around 2 volts and is getting around .04 watt at the typical "standard" current of 20 milliamps. A blue, white, or non-yellowish-green one typically has a voltage drop of 3.5 volts at 20 mA and gets .07 watt at 20 mA.
A candela is a lumen per steradian, or "beam candlepower". (Actualy, as
mentioned above, the candela is a primarily defined metric unit, while the
lumen is defined in terms of the candela.)
So lumens are candelas times the beam coverage in steradians. Candelas are
lumens divided by the beam coverage in steradians. Ideally, that is -
assuming that all light is within the beam and the "candlepower" is
constant within this beam.
So you may now be wondering what a steradian is. It is 1 / (4 * pi) of a whole sphere or 1 / (2 * pi) of a hemisphere or about 3283 "square degrees", to the extent there is such a thing as a "square degree". To get steradians from the beam angle:
Steradians = 2 * pi * (1 - cos (.5 * (beam angle)))
(NOTE: There are a few other expressions equal to this. Proving that is homework for 12th graders taking trig / "elementary functions".)
So if you determine the steradian beam coverage and multiply that by the candela figure (or 1/1000 of the millicandela figure), you get the lumen light output - very roughly! The beam is not uniform and it does not contain all of the light. Obtaining lumens from beam angle and candela can easily be in the +100 / - 50 percent range. Actual lumens are generally higher than predicted by this formula with smaller beam angles of 8 degrees or less since the nominal beam does not include a secondary "ring-shaped" "beam" that usually surrounds the main one. Also note that some beam angle figures are optimistic and could lead one to expect a lot more lumen light output than actually occurs.
Please read my disclaimer.