White LED lumen testing

[jtr1962]

Post 313 updated with results of high temperature testing (i.e. constant 60°C base plate) for the K2 cool white, K2 neutral white, and LEDEngin LZ4-40NW10 10 watt neutral white star. I used a thermoelectric cold plate in heating mode for these tests. Summary of the results (60°C base plate testing in red):

K2 Cool White -220



K2 Neutral White -180



LEDEngin LZ4-40NW10



The results are pretty much what I thought they would be except for the K2 neutral white. The K2 cool white lumens output more or less tracked the output at room temperature across the entire range, except scaled by a factor of roughly 92% to 94%. Same thing for the LEDEngin except the scale factor was closer to 90% over most of the range (although it did rise to 96% by 1300 mA which I thought was strange).

The K2 neutral white started out like the other two, holding at roughly 94% to 95% of room temperature output until maybe 1.7 amps. Above that the output dropped like a rock as seen in the chart. Yet at room temperature output continued to rise until about 2.6 amps (although this was still less than the 3.1 amps managed by the K2 cool white). Only thing I can think of here, since the dice are likely the same, is the phosphor. Perhaps the neutral white phosphor is a lot more temperature sensitive. That would explain my results here, and also to a lesser degree the results at room temperature where the neutral white maxed out at a lower current than the cool white.

Just for kicks after I finished these tests I put the thermoelectric back into cooling mode. I managed to get 463.7 lumens out of the K2 cool white at a plate temperature of 0° C and a drive current of 3100 mA. I was even able to get 328.6 lumens at the maximum rated current of 1500 mA by cooling the base plate down to -12.5°C ( 9.5°F).

[HarryN]

Hi JTR1962,

Thank you very much for running those tests on the K2 at 60 C. That is particularly interesting information, as Lumileds is going to add another spec to their LEDs - the ratio of output at 60 C / performance under standard test conditions.

It seems that the K2 will do very well with this spec.

Any idea how the Cree or SSL parts will come out?

I can't believe that no one has sent an MC-E to you for testing yet, or at least paid for the parts like I did for some of the tests. Just amazing that this kind of resouce is out there for free and people cannot even buy some LEDs for you.

Thanks

HarryN

[saabluster]

Originally Posted by HarryN

Hi JTR1962,



I can't believe that no one has sent an MC-E to you for testing yet,

I had intended on sending him an M bin MC-E a while back but my resources have been low recently. I am going to send him a K bin 4A and an SST90 today.

[jtr1962]

Originally Posted by saabluster

I had intended on sending him an M bin MC-E a while back but my resources have been low recently. I am going to send him a K bin 4A and an SST90 today.

Received them today! Thanks! I should have time within a few days to run the tests.

Originally Posted by HarryN

Thank you very much for running those tests on the K2 at 60 C. That is particularly interesting information, as Lumileds is going to add another spec to their LEDs - the ratio of output at 60 C / performance under standard test conditions.

It seems that the K2 will do very well with this spec.

Any idea how the Cree or SSL parts will come out?

I'll see if I can do these tests on a Cree or SSC in the near future. The experiment itself doesn't take long to do, but setting it up does.

[HarryN]

Hi - that sounds great. I am just curious as heck to see how the SSC and Cree chips handle the heat under the more "real life - like" conditions at 60C.

Lumileds claims to have some strong technology in that area, and it will be very interesting to compare single and multi die LEDs of different brands together at 60C to see which ones really put out the most lumens at 5, 10, and 15 watts. I suspect that people are going to be surprised when you really run the test - which setup has more light output when you have 10 - 20 watts to play with:
- A pair of K2s
- A pair of single die Cree parts
- MC-E (4 die)
- P7 (4 die)

IMHO, this will really start to define future light builds.

[jtr1962]

Here are the results of my most recent round of testing:

Seoul Semiconductor N42182L bin S2SL0H warm white high CRI star (acquired September 2009)

This is the first high-CRI LED I've tested. The S2 flux bin is speced at 60 to 70 lumens at 350 mA, and the SL0 tint bin is between 3250K and 3500K. Here are the results:





The N42182L slightly missed the S2 bin at 58.5 lumens at 350 mA. However, this is only marginally low, and likely within the margin of error of my tests. Vf is a low 3.08 volts at 350 mA, rising to 3.35 volts at 1000 mA. Efficiency at 350 mA is a pretty decent 54.2 lm/W given the nature of this LED (lower CCTs and high CRI both significantly reduce luminous flux). CCT did indeed appear to be in the low-middle 3000s as per the spec. Color rendering, especially with warmer colors, was much better than that of cooler LEDs. The N42182L didn't do particularly well at higher currents. Output peaked at 111.3 lumens at 1000 mA, only 1.9 times the value at 350 mA. This seems to be part of a general trend I'm noticing where warmer-tinted LEDs don't scale as well with current as their cooler cousins. It may have to do with the phosphor used perhaps being more sensitive to heat. There are certainly greater Stokes losses, which in turn translates into heat, in a phosphor outputting more longer wavelengths.


Lumileds Rebel neutral white -100 (acquired September 2009)

This is the highest available bin of Lumiled's neutral white Rebel presently available. Here are the results:





Color temperature looks to be around 4500K. This is no surprise. Due to nature of LED manufacture at present cooler bins tend to emit more output, and therefore the cooler binned neutral whites will be most likely to hit the -100 spec. The -100 neutral white actually falls a little short of the spec, coming in at 98.6 lumens at 350 mA. However, the margin of error in my testing likely exceeds the amount that the spec was missed. Vf at 350 mA is a very low 2.97 volts, and rises to a mere 3.30 volts at 2000 mA. Output with current scales even better than the cool white Rebel -100 I tested last year, and maxed out at 288.3 lumens at 2000 mA. Despite marginally missing the spec, efficiency at 350 mA is 94.9 lm/W (compared to 98.2 for the Rebel -100 cool white) due to the very low Vf. Overall the Rebel is slowly but steadily improving.


Cree MC-E bin K tint 4A (acquired September 2009)

I finally received an MC-E to test courtesy of saabluster! The K flux bin is 370 to 430 lumens at 350 mA per die, and the 4A tint bin is 4500K to 4750K. Here are the results:





I connected the dies in series as it was the easiest way for me given my power supply. As can be seen, the MC-E falls right in the middle of the K bin with 403.7 lumens at 350 mA. Efficiency at that current is an excellent 93.6 lm/W. Vf at 350 mA is a low 12.32 volts, or 3.08 volts per die. Output peaks at 915.2 lumens at 1200 mA. Vf at 1200 mA is 13.37 volts ( 3.3425 volts per die), and efficiency is a pretty decent 57 lm/W. CCT does indeed appear to fall into the 4500K to 4750K range. Overall, this is a very nice LED which handily meets specs, and provides an output sufficient for general lighting.


Luminus Phlatlight SST-90 (acquired September 2009)

Along with the MC-E, saabluster sent me a Phlatlight SST-90 mounted on a copper slug for testing. The SST-90 is a single die LED using a huge die of roughly 3mm x 3mm. Multiple bond wires and photonic lattice technology enable even surface current distribution. The die is bonded to the thermal pad with a resistance of only 0.64°C/W. The SST-90 is speced at 3.2 amps. After using my usual test apparatus to plot the radiation pattern I mounted the SST-90 on a fan-cooled Pentium 4 heatsink. It was obvious given the specs of this LED that the smaller extruded heat sink on my test apparatus would not be up to the task. I had installed a temperature sensor in the P4 heatsink for testing thermoelectric modules. This would come in handy and allow me to determine the LED's slug temperature. The thermal resistance of this heat sink is about 0.24°C/W. Therefore, the total thermal resistance between the die and ambient should be around 0.24 + 0.64, or 0.88°C/W. Without any further ado, here are the results:






As can readily be seen, the SST-90 is impressive. Output at the rated current of 3200 mA is 925.4 lumens, Vf is 3.48 volts, and efficiency is 83.1 lm/W. To put things into perspective, the closest competitor to the SST-90 which I tested was the Seoul Semiconductor P7. At the same current the P7 only managed 842.6 lumens and 72.3 lm/W. As current rose, the SST-90 only increased its lead. The P7 ran out of steam at 7 amps and 1195.8 lumens. The SST-90 managed 1782.9 lumens, or 49% more, at the same current. By that time, I was muttering Will Smith's line from "Men in Black" when he fired off the big laser gun: "Now that's what I'm talking about". On the other end of the current scale, behavoir at low currents was very interesting. The die doesn't even emit light until drive current reaches 13 mA. Efficiency starts out low, and doesn't peak until 500 mA. It remains above 100 lm/W in the 350 to 700 mA range. And it falls very slowly, remaining above 60 lm/W even at the maximum rated current of 9 amps.

The SST-90 bought me into uncharted territory in terms of output and drive current. I continued to ramp up the current, expecting that eventually it would run out of steam and output would level off, but output kept increasing. I started telling my power supply in Star Trek fashion "Can you give us any more?" (anyone remember that line from the first Star Trek pilot?). By the time drive current hit 11.75 amps the answer was no, but the SST-90 would gladly have taken more if the power supply could have given it. Vf at 11.75 amps was still only 3.95 volts while output was 2530 lumens! And as can be seen in the lumens versus current chart, output wasn't even close to leveling off. I think the SST-90 could have exceeded 3000 lumens at perhaps 15-16 amps, assuming the bond wires were up to the task. However, since this SST-90 was a loaner, I wasn't about to try to parallel supplies to get more current into it. 2500+ lumens was impressive enough, as was 2136.3 lumens at the maximum rated current of 9 amps! And for good measure, intensity at 1 meter exceed 1000 lux. I dread to think about the intensity the SST-90 can put out with a suitable aspheric.

[ICUDoc]

Yet another terrific post contributing to our knowledge- thanks mate. I note that at 16W the MC-E puts out 915 lumens but the Phlatlight is at 1246! That looks very impressive. What colour temp was it? And I wonder how much difference the excellent heatsinking made to the relative efficiencies?
Thanks again.

[jtr1962]

Originally Posted by ICUDoc

Yet another terrific post contributing to our knowledge- thanks mate. I note that at 16W the MC-E puts out 915 lumens but the Phlatlight is at 1246! That looks very impressive. What colour temp was it? And I wonder how much difference the excellent heatsinking made to the relative efficiencies?

Thanks!

Color temp of the SST-90 looked to be in the 6000s, probably around 6500K. And the relative heat sinking didn't make all that much difference at the currents the MC-E was running at. I had a fan blowing on the heat sink when I was testing the MC-E. The heat sink only reached perhaps 10°C over room temperature. I tested the SST-90 on the same heat sink as the MC-E up to 2 amps, then moved it to the other one, and retested. At 2 amps there was only a 2% difference in the flux, although this was at a 6.6W power level. So the MC-E might have done perhaps 5% better on the other heat sink, but the SST-90 still would have handily beaten it.

[saabluster]

OMG! Those are better results than I expected. I was a little sceptical when Luminus released their specs on these LEDs but I have to hand it to them. Luminus is much to be feared. I wonder how much of this performance if any comes from their strategic relationship with Nichia.

Excellent job as usual. BTW it's saabluster not blaster. Although thats not all that far off either. Thank you Thank you Thank you!

[jtr1962]

Originally Posted by saabluster

OMG! Those are better results than I expected. I was a little sceptical when Luminus released their specs on these LEDs but I have to hand it to them. Luminus is much to be feared. I wonder how much of this performance if any comes from their strategic relationship with Nichia.

What really impresses me is how the SST-90 performs at high currents. It just takes want you give it and asks for more!

Excellent job as usual. BTW it's saabluster not blaster. Although thats not all that far off either. Thank you Thank you Thank you!

You're quite welcome! And sorry about the mispelling. I just fixed it. That's what happens when I post on a few hours sleep.

Almost forget to ask but do you know what flux bin this SST-90 was? It falls into the WM flux bin ( 850 to 1000 lumens @ 3.2A ) by my tests.

[saabluster]

Originally Posted by jtr1962

Almost forget to ask but do you know what flux bin this SST-90 was? It falls into the WM flux bin ( 850 to 1000 lumens @ 3.2A ) by my tests.

There was no flux bin provided as it was a pre-production sample.

[milkyspit]

Originally Posted by ICUDoc

Yet another terrific post contributing to our knowledge- thanks mate. I note that at 16W the MC-E puts out 915 lumens but the Phlatlight is at 1246! That looks very impressive. What colour temp was it? And I wonder how much difference the excellent heatsinking made to the relative efficiencies?
Thanks again.

In fairness to the MC-E, the sample you're using for comparison is a markedly warmer tint using less efficient phosphors... a better comparison would involve a cool white MC-E of M-flux and similar color temp... my guess is that plus identical heatsinks would give the Cree at least a 20% better result, and more importantly, it's a far more valid comparison since test conditions would be more consistent in both cases.

Still, the Phlatlight did deliver a great showing, there's no disputing that!

[jtr1962]

Originally Posted by milkyspit

In fairness to the MC-E, the sample you're using for comparison is a markedly warmer tint using less efficient phosphors... a better comparison would involve a cool white MC-E of M-flux and similar color temp... my guess is that plus identical heatsinks would give the Cree at least a 20% better result, and more importantly, it's a far more valid comparison since test conditions would be more consistent in both cases.

I agree 100% here. Looking at my earlier tests with the cooler binned P7s I was obtaining roughly 62-63 lm/W when driven at the 16 watt level, or roughly 1000 lumens. Add another 5% or so for the better heat sink, and that would be about 1050 lumens. An M-binned MC-E would probably slightly outperform the P7 based on the relative performance of the XR-Es and P4s, so I'd guess 1100 lumens at 16 watts on the same heat sink as the SST-90. OK, the SST-90 still beats it, but at least it's not the LED equivalent of a shut out!

Still, the Phlatlight did deliver a great showing, there's no disputing that!

You have a definite talent for understatement. My jaw was dropping to the ground by the time the SST-90 hit 10 amps, and begged for more. If this is the way LEDs are heading, then it's going be .

[saabluster]

Originally Posted by jtr1962

My jaw was dropping to the ground by the time the SST-90 hit 10 amps, and begged for more. If this is the way LEDs are heading, then it's going be .

BTW I managed to buy the last two CST-90s that are confirmed to be the top bin. These use the same die as the one I sent you but the die is directly mounted to a large piece of solid copper! They are rated up to 13A but we both know it can handle a bit more than that. Question is can you power it up that high?

[jtr1962]

Originally Posted by saabluster

BTW I managed to buy the last two CST-90s that are confirmed to be the top bin. These use the same die as the one I sent you but the die is directly mounted to a large piece of solid copper! They are rated up to 13A but we both know it can handle a bit more than that. Question is can you power it up that high?

Cool! Is your question "Do I have a power supply capable of going past 13 amps?" The answer is I do, although nothing electronically regulated like the one I usually use when testing LEDs. However, all this is giving me incentive to make some sort of current regulated switching supply which I can attach to my main supply. The power supply I was using is a linear supply I built. It can put nearly 12 amps and about 15 volts into a load. It was made mainly for testing thermoelectrics. As you can imagine, the heat sink for the MOSFETs got pretty warm powering a low voltage, high-current load like the SST-90. I think my idea of making a switching current regulator powered by the main supply makes more sense. I'm figuring I could probably design it to go past 25 amps. If huge, single-die LEDs are part of the future, then I need a better way of powering them than I do now.

Hopefully after I finish a few days of work, I'll get started. I'll obviously be using a chip which uses external MOSFETs and probably synchronous rectification. It might be a bit of a pain to design, although I suppose I'll find a ready market here if I'm able and willing to mass produce them. Anyway, were you thinking of sending me one of the CST-90s to test? I looked at the data sheet for those. They look like 16 or 17 amps should be no problem.

[HarryN]

Wow - very nice work. I would love to see how the rebel, MC-E and Luminus parts work at 60C as well so we can estimate real world output. Is there any way to fit that info into your test plans?

[jtr1962]

Originally Posted by HarryN

Wow - very nice work. I would love to see how the rebel, MC-E and Luminus parts work at 60C as well so we can estimate real world output. Is there any way to fit that info into your test plans?

I'll try to do that as soon as I finish up this latest round of work I need to do (building 100 regulator boards for a customer). Hopefully in a few days.

The SST-90 especially should be easy to do. I just need cut the fan speed enough to keep the heat sink at 60°C. The other two parts will need active heating over most of their operating range (although the MC-E at higher currents should produce enough heat to keep the heat sink at 60°C).

[saabluster]

Originally Posted by jtr1962

Cool! Is your question "Do I have a power supply capable of going past 13 amps?" The answer is I do, although nothing electronically regulated like the one I usually use when testing LEDs. However, all this is giving me incentive to make some sort of current regulated switching supply which I can attach to my main supply. The power supply I was using is a linear supply I built. It can put nearly 12 amps and about 15 volts into a load. It was made mainly for testing thermoelectrics. As you can imagine, the heat sink for the MOSFETs got pretty warm powering a low voltage, high-current load like the SST-90. I think my idea of making a switching current regulator powered by the main supply makes more sense. I'm figuring I could probably design it to go past 25 amps. If huge, single-die LEDs are part of the future, then I need a better way of powering them than I do now.

Hopefully after I finish a few days of work, I'll get started. I'll obviously be using a chip which uses external MOSFETs and probably synchronous rectification. It might be a bit of a pain to design, although I suppose I'll find a ready market here if I'm able and willing to mass produce them. Anyway, were you thinking of sending me one of the CST-90s to test? I looked at the data sheet for those. They look like 16 or 17 amps should be no problem.

Judging by your earlier post I was not sure if you were able to test at the higher range safely. I am not the electronic wizard you are so most of what you said went over my head. I am perfectly willing to send you the CST-90 although unfortunately like the other one I will need it back as I have plans for it. I also have an MC-E M bin for you to test. I have a special little test for it as well which I will PM you about. Thanks!

[txg]

this is such a GREAT thread, keep it going jtr1962. thank you very much, very informative.

are you planning to test one of the new cree xp-g r4 leds?

[jtr1962]

Originally Posted by txg

this is such a GREAT thread, keep it going jtr1962. thank you very much, very informative.

are you planning to test one of the new cree xp-g r4 leds?

Thanks for the compliments! The thread will keep going so long as I'm alive and there are LEDs to test. As for the XP-G, I have some R4s on the way from Cutter for a bike light project, and I'll certainly be testing one of them!

[NIMA1966]

Hi JTR1962This link might help http://www.national.com/an/AN/AN-1937.pdf

[JoeF]

Thanks very much for your testing of white LEDs. I have some Light of Victory 1 Watt 10mm, and 0.5 Watt 8mm (140 degree viewing angle) white LEDs. Can I send a few to you? I tried to send a private message, but was denied access for some reason.

[jtr1962]

Originally Posted by NIMA1966

Hi JTR1962This link might help http://www.national.com/an/AN/AN-1937.pdf

Thanks for the link. Interesting IC. Not sure if I'd want to use it however due to the weird input voltage requirements.

Originally Posted by JoeF

Thanks very much for your testing of white LEDs. I have some Light of Victory 1 Watt 10mm, and 0.5 Watt 8mm (140 degree viewing angle) white LEDs. Can I send a few to you? I tried to send a private message, but was denied access for some reason.

Thanks for the compliments. I'm glad you've enjoyed the thread!

PM privileges are only given after the first few posts, which is why you were unable to PM me. I might be interested in testing the 0.5 watt 8mm. I've already tested a few narrow beam 10 mm's however, so there is probably not much to be gained by testing another. Besides that, the very narrow beam angle emitters are a major pain to test.

[jtr1962]

Cree XP-G bin R5 (acquired October 2009)

A package containing my XP-Gs arrived from Cutter today. I had ordered R4 bins but Cutter substituted R5s. Naturally, the first thing I did after opening the package was to set up my test jig. The R5 bin is specified as 139 to 148 lumens. Color temperature of my sample appeared to be roughly 6500K. The XP-G was rather difficult to set up for testing due to its form factor. I mounted it on a PCB I had made for Rebels. It was necessary to modify the board a bit due to the different pad layout. I then thermal epoxied this on to a brass tab which was bolted onto my test jig. I'll admit the thermal path could have been a little better, but it didn't appear to affect test results very much. Here are those results:







Beam angle is 125.4°, Vf at 350 mA was only 3.01 volts, output was 145.4 lumens (well within the R5 bin), and efficiency was an amazing 137.8 lm/W! Owing to the larger die size, output and Vf scaled very well. Vf at 700 mA was 3.17 volts, output was 265.3 lumens. The corresponding numbers at 1000 mA were 3.26 volts and 351.1 lumens. Output at 1000 mA relative to 350 mA was 2.415, a bit short of the roughly 2.48 in the spec sheet. However, I'll attribute this small difference to my fairly lousy thermal path. Despite this, output continued to rise with current well past 2 amps, peaking at 546.6 lumens at 2500 mA! My previous highest result for a single die normal-sized emitter was 436.7 lumens, also at 2500 mA, for a Cree XR-E R2 mounted on a heat pipe and copper block. I've little doubt the XP-G could break 600 lumens on a similar setup. Another amazing thing was that efficiency remained above 100 lm/W until 1200 mA. Even at 2000 mA it was 77 lm/W.

It has been mentioned that the XP-G's superior performance can be attributed solely to a larger die size, as opposed to a better die. My test results also indicate a superior die. This is evidenced by the higher peak efficiency of the XP-G, as opposed to the best-binned XR-Es. My best result for an XR-E was 148.3 lm/W at 20 mA. The XP-G peaked at 157.6 lm/W between 60 and 80 mA. The chart below is further evidence of this. The red line is a plot of lumens versus current for the XP-G. The white line is a plot of lumens versus current for two XR-E R2s in parallel. Two XR-Es in parallel roughly simulates the die size of the XP-G. Note that the XR-E R2s I tested mostly likely have the larger die size Cree was using, prior to switching over to a slightly smaller size, due to the fact that I tested them in June 2008. This makes the comparison valid. Note how the XP-G outperforms two XR-E R2s in parallel up to roughly 1700 mA. Above that the XR-Es have an edge owing to their superior thermal path (heat pipe and copper block as opposed to brass tab and thermal epoxy).



Also interesting to note is that the XP-G outperforms 4-die emitters such as the MC-E up to roughly 1500 mA. Even at 2000 mA the XP-G managed 528.7 lumens, while a K bin MC-E I tested only managed slightly more, 538.5 lumens, at 500 mA per die. Granted, an M-bin MC-E would do somewhat better, but even there the difference wouldn't be huge.

Overall, the XP-G is another quantum leap in performance from Cree.

[rav]

jtr1962, big thanks for the measurements. XP-G looks sensational.

[Solo Rider]

Jtr1962.

Yeah, good post, nice work, thanks.



Solo.

[NIMA1966]

Try to drive LEDs with pwm the output will be lower but you can eliminate the thermal loss.
At 1% duty cycle the output will be 100 times lower @ low currents but WILL peak FAR farther.
And finally you will calculate the lumens without the thermal effect

[HarryN]

Nice work as usual Joe. I would really like to see the comparable and more realistic 60 C results for an MC-E and the XP-G when you have a chance. Droop can be a big factor in LED applications, and it would be interesting to see if Cree has really worked its way past this.

Thanks

Harry

[ICUDoc]

Wow that efficiency at 1000mA is amazing! Now to find the right reflector...

And thanks very much again!

[blasterman]

Another amazing thing was that efficiency remained above 100 lm/W until 1200 mA. Even at 2000 mA it was 77 lm/W.

Yeah...that's pretty startling. I'm hoping we see pure blue versions of these devils because they are likely to be a quantum leap over blue P4s, and we've already discussed a significant application for those. I'll let you know in a week or two how my initial experiment is going.

Otherwise, as somebody mentioned above, we need to hold some applause for the performance of these killer emitters and awknowledge that 6500k -vs- 4500k is like benchmarking dual core processors vs single core. Hyper efficient cool-white emitters is a given, but when you move to the neutral/warm-white phosphor category the efficiency hit is quite severe. This is really what I'm waiting to see with the XP-G.

Good testing regardless.