White LED lumen testing

[Ron Schroeder]

Hi jtr,

Can you get reliable measurements from a very wide angle LED that only puts out about 5 lumens? I have some of the JKL 5mm LED "condoms" that convert a blue LED to diffused white. I am very interested in comparing that technology to the phosphor on die technology.

[Gunner12]

This is amazing, 500 lumen from a single die commercially available LED. I'm surprised it didn't die at above 2 something amp.

From what you posted, I'm assuming the LED didn't loose much of it's performance even after this abuse?

A well cooled MC-E might even be able to break 2000 lumen, but that would be a lot of cooling.

Great job jtr!

[jtr1962]

Thank you for this fascinating test. I am truly amazed by the thoughness of these small dies.

Me too. I thought for sure I would hear a pop by 2.5 amps but the emitter just kept on cooking!

This is amazing, 500 lumen from a single die commercially available LED. I'm surprised it didn't die at above 2 something amp.

From what you posted, I'm assuming the LED didn't loose much of it's performance even after this abuse?

A well cooled MC-E might even be able to break 2000 lumen, but that would be a lot of cooling.

Great job jtr!

When I remeasured the lux at 350 mA after the tests it was exactly the same as before. The LED went through all this abuse yet came back in pristine condition. While I have little doubt it wouldn't have survived long running continuously at those currents, brief excursions seem to be harmless.

Yes, it's entirely possible an MC-E may break 2000 lumens. I think the limiting factor there and even in the case of the XR-E is the thermal attachment resistance. For the XR-E it's listed as 8°C/W. That means when I was running the LED at 2.7 amps the temperature rise of the die was in excess of 80°C (accounting for the percentage of input power which exited the die as light rather than heat). An 80°C rise will degrade output by about 15%. In other words, with the fan-cooled tests had there been zero thermal impedance between the LED junction and the heat sink I probably would have obtained perhaps 530 lumens from the LED, and even more with below ambient cooling. I suspect the thermal attachment impedance will affect the MC-E even more. I believe it's listed as 3°C/W, which translates into 12°C/W for each die.

Hi jtr,

Can you get reliable measurements from a very wide angle LED that only puts out about 5 lumens? I have some of the JKL 5mm LED "condoms" that convert a blue LED to diffused white. I am very interested in comparing that technology to the phosphor on die technology.

Yes, I've measured many low-output, wide-angle LEDs as you can see here:

[jtr1962]

Wow! 502 lumens! Excellent work JTR. I really appreciate your doing this for me. I don't think I did a very good job getting the led really close to the heatpipe but it still seems to make a difference. I have an even more extreme way of mounting the LED that should give it even more of a boost. It will be awhile before I can get it made but it should be incredible when I do.

I'll be happy to see if I can better my results with your new mounting method. I may also repeat the below ambient cooling part of the tests in the winter. The amount of humidity in the air was causing condensation which had to continually be wiped off the LED. In the winter these problems would be far less severe. Perhaps I could even try a liquid-cooled thermoelectric setup allowing me to bring the thermal pad well below 0°F. That might make breaking 600 lumens possible.

So are you going to try to unmount the LED from the heatpipe to use in one of your projects? I wonder how difficult that would be.

I'll leave it as is. Unmounting it would probably destroy the LED. I'm thinking of doing a long-term 350 mA lumen degradation test with it similar to the ongoing one I started with a Q bin Luxeon 4.5 years ago (yes, it's still going strong).

One other thing. What program do you use to make those graphs. I would like to use it to display run times and such. Thanks again for your hard work.

I use Microsoft Excel '97. I believe the free Open Office has a spreadsheet as part of its suite which would work equally well here.

[saabluster]

JTR could you make a graph showing the original R2 tests you did with these new tests(including the active cooled) all together on one? I think it would help us visualize the differences better. Thanks!

[jtr1962]

JTR could you make a graph showing the original R2 tests you did with these new tests(including the active cooled) all together on one? I think it would help us visualize the differences better. Thanks!

Is this OK?

Yellow line is with my stock setup, purple line is your setup but passively cooled, and the dark blue line is your setup actively (fan) cooled.

[saabluster]

Is this OK?

I was hoping you also had the data points with it on the cold plate as well. If not I understand and thank you for posting that last graph.

[PhantomPhoton]

Ah, LED overclocking... brings back memories of me and my Athlons a few years ago.

I'm surprised at the apparent robustness of those Crees. When will we see the first production of lights by some crazy modder attaching a heatpipe to the Emitter and running the heat down to the body (heatsink) of the light while thermally isolating the head? Running a Cree @ 2A+ in a compact handheld light would be pretty sweet.

[jtr1962]

I was hoping you also had the data points with it on the cold plate as well. If not I understand and thank you for posting that last graph.

With the constant condensation of moisture on the LED due to the humidity I wasn't really able to do a controlled test on the cold plate like I would have wanted to. My only data point is 2.5 amps, 502 lumens. At lower currents the LED dome froze over due to the lack of light energy to melt the ice, making measurements impossible. I'll probably try again when conditions for this sort of test are optimal. This is usually January or February when my workroom temps are in the mid 50s.

[saabluster]

With the constant condensation of moisture on the LED due to the humidity I wasn't really able to do a controlled test on the cold plate like I would have wanted to.

I kind of thought that to be the the case. Thanks.

[curry__muncha]

Excellent database!!

very useful!

really appreciate the effort that has been put into this, as many others have said.

=]

keep up the good stuff! ;]

[MattK]

Awesome work as always gents!

Is anyone planning to do some P7 testing?

[ViReN]

Awesome work as always gents!

Is anyone planning to do some P7 testing?

WOW Matt , you have already sent the P7 to jtr1972 for testing....

we all look forward for amazing test results....

[MattK]

LOL I haven't but I'd happily do so - or fund his testing if he wants to source it himself.

[zzonbi]

Marvelous, on par with Osram 500lm world record

http://www.osram-os.com/osram_os/EN/...hip-record.jsp

This OC reminds me of those engine dyno charts. So basicaly we have max hp(450lm) at 2700rpm(mA) without phase-change cooling (normal fuel or natural aspiration) from 1l capacity (1mm^2).

I wonder if the other main chip (Philips die of the Luxeon K2 with TFFC) maxes out around the same 'rpm'.

Even if lm/W is limited, there is no clear limit for emittance (lm/m2).

[Brlux]

Do you know what the flux and tint bin was for your Nichia NSPW500GS-K1 led's?

[Cemoi]

Thanks and congratulations jtr for this excellent and impressive work.

Could you please kindly make your table available as a spreadsheet, which would allow us to sort it to our liking (e.g. by Vf, output, etc.)?

On the other hand, are you still interested in testing 5mm red LEDs? I have purchased some 25 cd 20 degree red LEDs from BestHongKong, so I could send you one for testing.

[redlaw55]

The table listing measurements of LEDs has me thinking of upgrading the 5mm leds in a Nite-Ize module for AA maglites. I wonder how the Nite-Ize leds would do against those listed.

[jtr1962]

Seoul Semiconductor P7 Bin C (tested October 2008)

I obtained a Seoul P7 of the highest brightness bin this week. The C bin is rated at between 740 and 900 lumens. For this test I epoxied the P7 to a 60mm heat sink. I tested the output with and without a fan cooling the heat sink. Here are the results:











At the nominal current of 2800 mA the fan makes only a small difference. Output is 731 lumens without the fan, and 742 lumens with fan cooling. This is barely but still within ratings. Efficiency at nominal current is 70.7 lm/W, good but certainly not in the same league as the latest single-die Cree or Seoul LEDs running at a comparable current of 700 mA. However, that's to be expected since the thermal path per die for the P7 isn't as good as for the Seoul P4 or Cree XR-E. If you underdrive the P7 then you can get significantly greater efficiency. For example, at 1000 mA the efficiency is 97.9 lm/W, way better than even an R2 bin XR-E at that current, and output is 329 lumens, about 20% greater than the Cree at that current. At 350 mA the P7 can manage 115 lm/W and 125 lumens, both figures somewhat greater than the best single die LEDs. Efficiency peaks in the 125 lm/W area at currents under 100 mA. Output versus current actually remains linear to well under 10 mA, but with the slight drop in Vf you're only gaining a few percent efficiency compared to 100 mA.

The results when cranking up the current are interesting. Without the fan cooling the heat sink, the output levels off at 6 amps and 998 lumens. With the fan I was able to go to 7 amps and 1114 lumens, although the increase in output was flat from 6.8 to 7.0 amps. The increase from 6.5 to 6.8 amps barely registered on my light meter. Therefore, even with very good forced-air cooling there is no point taking the P7 past about 6.5 amps, and only then if you want to squeeze out every last lumen. In fact, if you consider that the eye can't detect brightness differences of 10%, then the absolute maximum current of the P7 should be no more than around 4.5 amps, even less with the heat sinking typically available in flashlight bodies. At 4.5 amps you will get about 1000 lumens with very good cooling. Note however that 1000 lumens is only 35% more than what the P7 will give you at its nominal current of 2.8 amps. The lumen increases from overdriving the P7 just aren't as dramatic as with single die LEDs. In a typical situation with less than stellar heat-sinking I doubt it even pays to go above the nominal 2.8 amp current.

[jtr1962]

Do you know what the flux and tint bin was for your Nichia NSPW500GS-K1 led's?

It looks to be rank W (the highest) based on my results.

Thanks and congratulations jtr for this excellent and impressive work.

Could you please kindly make your table available as a spreadsheet, which would allow us to sort it to our liking (e.g. by Vf, output, etc.)?

On the other hand, are you still interested in testing 5mm red LEDs? I have purchased some 25 cd 20 degree red LEDs from BestHongKong, so I could send you one for testing.

The spreadsheet for the table is available in the zip file containing all my spreadsheets: Link (look for the file LED Comparison Chart.xls )

Regarding the BHK 5mm reds, I don't test colored LEDs because the results would be meaningless. When I participated in the CPF light meter testing I found that my meter was way off with colored LEDs.

[Gryloc]

jtr1962,

I just wanted to check. Could you give us the exact bin of SSC P7 emitter that you used? I remember that you said in a PM that the emitter that you tested was actually a D-bin. If you cannot be 100% certain there, then could you try to give us the forward voltage bin? I was curious if it was a J or I bin. According to their documentation, I believe that Seoul bins their Vf when the I = 2800mA. The I bin means a Vf of 3.25V - 3.50V, and the J bin means a Vf of 3.50V - 3.75V. Seeing your data, it seems to be a J bin. However, on my own Vf vs I tests on P7 emitters, the actual data seemed different than what the bin supposed to be (actual Vf is lower than what the bin states). Maybe the Vf drops a bit over time. Thanks!

-Tony

[Holzleim]

Hi jtr1962,

any plans testing CREE Mc-E?

I made a quick test with my MC-E (WG, K-bin) soldered to solid copper giving the feeling that output still increases till 2A. Between 2A and 2,1A output nearly stagnated so I stopped measuring at 2A in order not to destory my only MC-E.

But I have no equipment to measure real lumen output, the values shown below are just Lux-meter reading of the same optical system while increasing current. Temperature of the copper was nearly static.

Current Lux x100
0,1 88
0,3 188
0,7 349
1,0 455
1,5 600
1,8 670
2,0 708


Thanks for your superior work,
Regrads,
Holzleim

[jtr1962]

I just wanted to check. Could you give us the exact bin of SSC P7 emitter that you used? I remember that you said in a PM that the emitter that you tested was actually a D-bin. If you cannot be 100% certain there, then could you try to give us the forward voltage bin? I was curious if it was a J or I bin. According to their documentation, I believe that Seoul bins their Vf when the I = 2800mA. The I bin means a Vf of 3.25V - 3.50V, and the J bin means a Vf of 3.50V - 3.75V. Seeing your data, it seems to be a J bin. However, on my own Vf vs I tests on P7 emitters, the actual data seemed different than what the bin supposed to be (actual Vf is lower than what the bin states). Maybe the Vf drops a bit over time. Thanks!

The guy who gave it to me doesn't even know the Vf bin. Honestly, I don't put much value in Vf binning because I've seen Vf change over time, and it certainly changes with temperature. Binning by Vf ostensibly might be beneficial in that matched emitters should track with temperature, but they still might not track with time. And this isn't theoretical, either. I've carefully matched emitters by Vf when I made a few bike lights using 5mm LEDs only to find some months later some emitters were brighter than others. IMHO matching Vf is a lesson in futility. Now I design so that Vf doesn't matter.

I haven't yet tested the emitters you sent me. Last two weeks I've had to reorganize my work area as it was getting to the point I couldn't function. As soon as that's completed, I'll be resuming my lumens testing.

[jtr1962]

Hi jtr1962,

any plans testing CREE Mc-E?

Tony (Gryloc) mentioned that he'll send me some samples to test as soon as I'm done with the P4s and P7s he sent me, assuming that Cutter sends them to him by then.

[Gryloc]

Tony (Gryloc) mentioned that he'll send me some samples to test as soon as I'm done with the P4s and P7s he sent me, assuming that Cutter sends them to him by then.

We all know that things are a little weird with Cutter. They have not shipped yet. I may order the M-binned MC-E from Kaidomain, but I fear the equally long wait. Is that a bad idea? DX has the 370lm version available, as does LEDsupply and Digikey (which would ship promptly), but I sort of wanted to send jtr1962 the best bin that is available.

To all,

Does anybody have an MC-E currently available to lend to jtr1962 for lumen testing? I offer to pay for shipping charges or what ever as some aid to mail it. Thanks.

-Tony

[saabluster]

I have an M bin that I could send him. I will even mount it to solid copper.

We all know that things are a little weird with Cutter. They have not shipped yet. I may order the M-binned MC-E from Kaidomain, but I fear the equally long wait. Is that a bad idea? DX has the 370lm version available, as does LEDsupply and Digikey (which would ship promptly), but I sort of wanted to send jtr1962 the best bin that is available.

To all,

Does anybody have an MC-E currently available to lend to jtr1962 for lumen testing? I offer to pay for shipping charges or what ever as some aid to mail it. Thanks.

-Tony

[Paul Baldwin]

Hi,
Very interesting info, I'm glad I stumbled across this thread I had wondered if keeping an led properly cool would allow you to push it harder and you've well and truly confirmed that!
I'm now wanting to go and buy a peltier/heatsink and fix an led directly to the cold side and see what can be achieved.
Keep up the great work you are doing

Paul.

[Jarl]

Hi jtr1962,

any plans testing CREE Mc-E?

I made a quick test with my MC-E (WG, K-bin) soldered to solid copper giving the feeling that output still increases till 2A. Between 2A and 2,1A output nearly stagnated so I stopped measuring at 2A in order not to destory my only MC-E.

But I have no equipment to measure real lumen output, the values shown below are just Lux-meter reading of the same optical system while increasing current. Temperature of the copper was nearly static.

Current Lux x100
0,1 88
0,3 188
0,7 349
1,0 455
1,5 600
1,8 670
2,0 708


Thanks for your superior work,
Regrads,
Holzleim

Do you have any idea of what the Vf was doing at such high currents?

Thanks!

[saabluster]

Hi jtr1962,

any plans testing CREE Mc-E?

I made a quick test with my MC-E (WG, K-bin) soldered to solid copper giving the feeling that output still increases till 2A. Between 2A and 2,1A output nearly stagnated so I stopped measuring at 2A in order not to destory my only MC-E.

But I have no equipment to measure real lumen output, the values shown below are just Lux-meter reading of the same optical system while increasing current. Temperature of the copper was nearly static.




Thanks for your superior work,
Regrads,
Holzleim

I missed this post somehow. You have done something really wrong if you can only get it to 2A before lumen output starts to go down. Especially considering the fact that you said the temperature of the copper remained static. That tells you the LED is not able to get the heat out of the package. It should still be gaining in output at twice that drive level.

[thegeek]

I missed this post somehow. You have done something really wrong if you can only get it to 2A before lumen output starts to go down. Especially considering the fact that you said the temperature of the copper remained static. That tells you the LED is not able to get the heat out of the package. It should still be gaining in output at twice that drive level.

Was it wired in series or parallel? Series I can see it peaking at 2A. Maybe it was a BIG piece of copper.