500 Lumen at 1.4A

[luschnouar]

From the German Caving forum......

http://gizmodo.com/5027606/osram-push-wh...uper-efficiency

It's an interesting week in the world of LEDs: on the weekend we heard about ultra-cheap ones, and today Osram (yes, the lightbulb people) has news that they've pushed white LEDs to world-record brightness. By optimizing the diode, light converter and the package, their lab test squeezed 500 lumens out of a single LED at 1.4A. That's bright enough for projector tech, and certainly makes the single unit good for car lighting and even interior lights. At a lower, more optimal, current the 1mm-square white LED had an efficiency of 136 lumens/W which makes it about twice as efficient as standard fluorescent lamps and 10 times a normal bulb. Press release below.

 

[phantom23]

http://www.candlepowerforums.com/vb/showthread.php?t=203330

 

[JohnR66]

This is terrible news. Now I'm going to have to throw away all my XR-E based lights to get new ones...:laughing:

 

[Barbarin]

According to the rule:

50% of max current means 70% of max ouput (more or less it works) even a 700 lm SSCP7 should get 490 lm when driven at 1400 mA.

 

[monkeyboy]

According to the rule:

50% of max current means 70% of max ouput (more or less it works) even a 700 lm SSCP7 should get 490 lm when driven at 1400 mA.

A more precise calculation:

For the SSC P7, page 6 of the data sheet (W7240.pdf) shows the relationship between current and flux.

1400mA -> flux = 1.0 (arbitrary units)
2800mA -> flux = 1.75

for C bin emitters, the new binning structure is 700 to 800lm so we can assume that there are no LEDs over 800lm

lumens at 1400mA:

700 x 1/1.75 = 400lm
800 x 1/1.75 = 457lm

Of course this calculation doesn't take Vf into account since lm/W is what's important not lm/A.

The advantage of single die is better beam pattern and much smaller reflector needed for the same throw. Imagine if Osram works this technology into their 4 chip and 6 chip LEDs

 

[Barbarin]

You are right, I must confess I did not take a look deeply at the datasheet.

Anycase, 400 lm at 1400 mA sounds pretty good for a single LED light.

FYI, I have been testing some P7 C Bin samples:

350 mA @ 2,95 V= 1,03 Watt. (¿149 lm, 144 lm/watt?)
700 mA @ 3,20 V= 2,24 Watt (¿261 lm, 116 lm/watt?)
1400 mA @ 3,58 V= 5,012 Watt (assuming 457 lm means 91 lm/watt)
2800 mA @ 4,27 V= 11,956 Watt. (for 800 lm that is 67 lm/watt)

Next week we will test them on the integrating sphere.

 

[phantom23]

So P7 with four R2 dies and 840lm@2,8A doesn't exist? That's why, even though specs says 700-800lm, currently shipping C bins of P7 still have 740-900lm@2,8A.

 

[tebore]

So P7 with four R2 dies and 840lm@2,8A doesn't exist? That's why, even though specs says 700-800lm, currently shipping C bins of P7 still have 740-900lm@2,8A.

Those are just out of date specs. You can't expect a C-bin to have much over 800lm at 2.8A. I doubt you'll even get 800lm out of a C-bin at 2.8A.

Interestingly I did some testing Where I pushed 5.5A at ~7v in to a C-bin and it didn't blow. The LED ran with no noticeable dimming for about 5 minutes. The vf kept dropping and the amperage kept going up after a period of time. I believe thermal runaway was about to occur.

I had to use an Thermalright SLK800U. The measured temp of the base of the heatsink near the LED was about 70degrees, using a thermal couple. There was no active cooling.

 

[LEDninja]

So P7 with four R2 dies and 840lm@2,8A doesn't exist? That's why, even though specs says 700-800lm, currently shipping C bins of P7 still have 740-900lm@2,8A.

You are assuming SSC only use R2 dies when making C bin P7s.
Where does SSC promise to use only R2 dies and nothing but R2 dies in C bin P7s?

Due to the shortage of R2 LEDs most SSC-P7s are probably a mix of Q5 and R2. There may be a SMALL NUMBER that may be all R2 but I do not think all of them are 4*R2.

CPF members who manage to get a good look at the SSC P7 mention 1 die is brighter than the other 3.
Here is a picture:

Why is 1 die brighter than the other 3? If all 4 are R2 they should be equally bright. Here is what I think.

From
http://www.candlepowerforums.com/vb/showpost.php?p=2170040&postcount=141
3*Q5@700mA each=3*184.89=554.94 lumens
From
http://www.candlepowerforums.com/vb/showpost.php?p=2412997&postcount=158
1*R2@700mA=210.57 lumens
554.94+210.57=765.51 lumens total.

Remember SSC only promised 740+ for the old C bin. 3Q5+1R2 meet that spec.

SSC does NOT promise 840+.
phantom23 you should not assume 4*R2 unless you have proof SSC only uses R2 for P7 production.

-----

Hard numbers from actual test results should help us decide if we should use the old C bin datasheet or the new one. Please remember to post.

Next week we will test them on the integrating sphere.

 

[VanIsleDSM]

You can't assume the flux bin by how bright it is at a very low current.. differences in Vf, since they are run in parallel, could easily make a P3 look brighter than an R2 if they were in the same P7 package.

 

[matt0]

Might I ask why "Q5" and "R2" are being used in reference to a P7? I know SSC P4 is supposed to be the same die as the CREE X-RE but aren't the "Q5" and "R2" bins assigned after production of the X-RE has finished?

Is there something I've missed?

 

[phantom23]

SSC uses EZ1000 dies in both P4 and P7. Minimum brightness level of P7 C bin is higher than 4xQ5 so they have to use higher binned dies like R2. Maybe that's why it's still limited?
http://www.cpfmarketplace.com/mp/showthread.php?t=180106
'LEDninja' go ahead and tell PhotonFanatic or Litemania that they're wrong and their P7s aren't 740-900lm but 700-800lm only and you don't care that manufacturer told them that nothing changed except specs on paper.
Dies can be brighter, dimmer but that's because of different Vf not bin! Cree (unlike SSC) doesn't control Vf of their products. Lower VF=brighter die.

 

[monkeyboy]

Agreed; the differences shown in that picture are probably due to Vf differences not flux. e.g. The difference between R2 and Q5 is not that big. I'm guessing that picture was taken was taken at a very low drive level? Perhaps the brightness evens up when driven at higher levels?

The fact that they've changed their binning structure to C = 700-800lm and D = 800-900lm and no one has seen D bins yet clearly indicates that none of the previous C bins were over 800lm.

 

[LEDninja]

My estimate show 765 lumens which meets both old C bin specs 740-900 lumens and new C bin specs 700-800 lumens.

I have no objection to PhotonFanatic listing 740-900 lumens.
My objection is you insisting they are 840 lumens.

You can not assume all P7 are 840-900 lumens. SSC never promised 840-900 lumens. That is a 'phantom23 spec' not an SSC one.

Even with 4*Q5 dies 4*184.89=739.56 which is effectively the same as 740 lumens. So there is no need for SSC to use only R2 equivalent dies.

matt0 is correct in pointing out Cree will ship dies without pre-sorting them so SSC will get a range covering current production equivalent to Q5 through R2 bins. I just cannot see SSC throwing out 1/2 the dies after they sorted them (that would cost too much, the CFO will fire the factory manager who try to lose money by doing that). So what to do? By mixing the top quality dies with the bottom quality ones they can save the low end dies from the garbage bin. And result in most of the SSC-P7 LEDs being average, not necessarily the best they can make.

-----

We are using Q5 & R2 as jtr1962 has tested those LEDs giving us numbers to estimate with. The Q5 numbers is the low end and the R2 numbers is the high end of current EZ1000 production.
So a SSC-P7 will be somewhere between 4*Q5 (~740 lumens) and 4*R2 (~840 lumens) in output. I think current P7 production covers the whole range. phantom23 thinks they only use R2 equivalent dies.

Might I ask why "Q5" and "R2" are being used in reference to a P7? I know SSC P4 is supposed to be the same die as the CREE X-RE but aren't the "Q5" and "R2" bins assigned after production of the X-RE has finished?

Is there something I've missed?

-----

To avoid unbalanced dies prematurely killing the LED, SSC would be grouping them by Vf, so I do not think differences in Vf would be significant within each P7.

You can't assume the flux bin by how bright it is at a very low current.. differences in Vf, since they are run in parallel, could easily make a P3 look brighter than an R2 if they were in the same P7 package.

 

[SemiMan]

Seoul buys dies from Cree.... that is it. They put the phosphor on and they package them. Those two items contribute to the overall light output. I suspect that Seoul has an efficient phosphor, but I question whether its reliability is as good as others out there based on tests we have seen from Newbie and others where there was quick degradation.

Semiman

 

[monkeyboy]

Seoul buys dies from Cree.... that is it. They put the phosphor on and they package them. Those two items contribute to the overall light output. I suspect that Seoul has an efficient phosphor, but I question whether its reliability is as good as others out there based on tests we have seen from Newbie and others where there was quick degradation.

Semiman

That's interesting. I wonder if the upcoming Cree MC-E will be able to handle more overdrive if:

1) the phosphor is more reliable.
2) connecting in series will ensure equal current to each die.

 

[phantom23]

So a SSC-P7 will be somewhere between 4*Q5 (~740 lumens) and 4*R2 (~840 lumens) in output. I think current P7 production covers the whole range. phantom23 thinks they only use R2 equivalent dies.

That's what I'm talking about!!! C bin can be between 740-840lm. That's why current shipping P7s are rated at 740-900lm - because it covers 740-840lm range! Maybe when R2 will be in mass production, and R3 or R4 as a limited run we'll see 700-800lm C bin and 800-900lm D bin...

 

[TexLite]

Seoul buys dies from Cree.... that is it. They put the phosphor on and they package them. Those two items contribute to the overall light output. I suspect that Seoul has an efficient phosphor, but I question whether its reliability is as good as others out there based on tests we have seen from Newbie and others where there was quick degradation.

Semiman

Semiman,
If my memory serves me correctly,Newbie tested the P4's when they first became available.These were the ones that didnt take to any current over 1000ma(sometimes not that) before the infamous "angry blue" tint shift.The P7's are assembled differently,something about the way the dice are bonded to the substrate.This is supposed to allow for higher current capacity,and I think it (the new process) was applied to the P4's as well.I know they certainly seem much more resistant to tint shift at high currents.

CPF members who manage to get a good look at the SSC P7 mention 1 die is brighter than the other 3.
Here is a picture:

LEDninja,
I saw that thread when it was first started and never posted,but it did cause me to test mine.I dont remember what the specific voltage was,but it was well below 1v.My results werent anything like that,mine looked very well matched.I could not discern a difference between the dice.

You can not assume all P7 are 840-900 lumens.

Right,I talked to a sales manager at SSC who confirmed all that is currently available are the C/700-800lm.I asked about the first run emitters I had gotten,whether they were 700-800lm or the 740-960lm.He said it was hard to tell.Up until that point I was doubtful that anyone had gotten any emitters with luminous flux over 800,now I'm not so shure.

On an interesting note,I did see,believe it or not,D bin P7's for sale at an electronics dealer.They were sold out,but I've been checking back to see if they get another reel.

Thanks,
Michael

 

[VanIsleDSM]

To avoid unbalanced dies prematurely killing the LED, SSC would be grouping them by Vf, so I do not think differences in Vf would be significant within each P7.

That's purely speculation, we have no information to support this from SSC. Lumileds didn't do it with the LuxV. Is SSC? only time will tell I suppose.. either way I maintain that the main differences you're seeing at low current levels is due to Vf, not bin.

Do an expirment to see for yourself if you like. Take 4 XR-Es, or SSC P4s, with a close Vf, and run them in parallel at low current, see the differences in the dies.. then run them in series at low current.. the differences between the dies will be much less. Doesn't take much of a difference in Vf to to send a lot more current through.. hence constant current drivers in the first place.

The only reason why I haven't bought a P7 is because of the parallel dies.. bad form SSC. I knew CREE would do it right.. Can't wait for that MC-E, already almost have the host and driver all ready for it, all from scratch.

 

[TexLite]

If we were seeing,or begin to see a high failure rate with the P7's I would see the cause for concern.But the only instance of failure I have seen was way over-current and insufficient heatsinking.Many of us here have pushed the P7's way over spec,some with CC power supplies,without a die failure.In my opinion its worrying/speculating a over minor issue.

I'm not saying they're all are perfectly matched,or that parallel is always the always the best configuration,I'm just saying I haven't seen a real cause for concern yet,but as you said,time will tell.

I have tested the P7's at very low current,and I could see absolutely no variation between the dice,not anything like the pictures above.

The MC-E will be very nice,one very nice aspect of the design will be the ability to measure the the Vf of each die independently.

Thanks,
Michael