New 2nd Generation Cree XP-G2 LED!

[Colonel Sanders]

Why? Because lighting nerds such as myself have short attention spans and will stop throwing money at Cree if they don't give new toys on a regular basis.

How? Unicorn tears and panda sperm.

+1. Yep! You nailed it. They know what they're doin'.

 

[SemiMan]

Ok, time for people to guess why and how Cree did it this time. Go!

Why? .... to stay ahead of the competition and give their end customers, many of whom are using the XPG an upgrade path without changing vendors.

How? .... packaging/die technology to address bonding wire issues to bring cost down and textured die surface to improve light extraction in order to improve efficiency

It's called progress.

Semiman

 

[degarb]

I am pretty sure this number is correct. I can say that it might be even more, it is about 7300-7400.

And here we are, just check some more numbers, that i didn't manage to upload them yesterday:
Fenix PD32 R5: was 323 lumen and 5175 lux, now 382 lumen and 7300 lux!
Armytek PRedator R5: was 345 lumen and ~17,000 lux, now 411 lumen and 27.150 lux!
Quark mini123 ti S3: was 293 lumen and ~3450 lux, now 310 lumen and 4470 lux!

All readings are @~30sec after activation, i have put a lot of effort to measure them accurately, but still, to be sure for all expect a +-3% difference.

But with lower forward voltage, simply swapping in a xp-g2 into an xp-g without changing the driver/resistor, naturally you would pull more current, have greater wattage and output, with lower runtime. Am I wrong?

 

[degarb]

I would disregard the last numbers I posted. It was a quick test like I said. I just bought a new power supply and I trusted too much what the readout was telling me. Not liking this new PS very much. grr. So I used a meter to see exactly how much current the LED was seeing and reran the test. This time with it attached straight to copper. These numbers are much more accurate.

I want to emphasize that the higher numbers are not achievable in a flashlight unless active cooling is used to keep the junction temps down. The lower you go in current the more representative these figures are going to be to real world flashlight usage. Still it is amazing to see this relatively tiny LED putting out over 1000 lumens. 1033 at it's peak. I chilled the LED to about 45F and got a reading of 1100 lumens even @4.91A.

.02 9
.08 36
.17 74
.2 87
.28 122
.38 163
.46 194
.57 233
.67 268
.75 294
.85 327
.93 353
1.03 383
1.13 412
1.21 436
1.32 465
1.4 487
1.5 513
1.6 537
1.68 557
1.76 575
1.87 600
1.97 624
2.07 646
2.15 664
2.2 673
2.33 699
2.43 718
2.53 738
2.62 755
2.72 773
2.8 786
2.9 800
3.0 816
3.09 829
3.15 839
3.2 847
3.27 858
3.38 876
3.48 890
3.57 903
3.66 914
3.75 925
3.85 937
3.95 949
4.03 960
4.13 970
4.22 979
4.32 988
4.43 999
4.5 1006
4.61 1016
4.69 1021
4.8 1028
4.91 1033
4.99 1032

Another dumb question: It looks like the sweet spot is near 2 AA range. Would this not make two 2AA the most efficient setup. Now a 4 AA with killer runtime-two 2 AA packs in parallel? Making the need for entirely new controllers to maximize this?

 

[ShineOnYouCrazyDiamond]

But with lower forward voltage, simply swapping in a xp-g2 into an xp-g without changing the driver/resistor, naturally you would pull more current, have greater wattage and output, with lower runtime. Am I wrong?

That would only apply if you were dealing with a voltage controlled power source. LEDs are driven by a current controlled power source - that means that the current is held constant by modifying the voltage. So regardless of forward voltage (don't forget that changes as LEDs heat up and cool down) the current should be held constant.

Unless the light uses PWM in which case it may change a bit. But I know that at least Fenix and 4Sevens use current controlled drivers and no PWM.

 

[bose301s]

Why? .... to stay ahead of the competition and give their end customers, many of whom are using the XPG an upgrade path without changing vendors.

How? .... packaging/die technology to address bonding wire issues to bring cost down and textured die surface to improve light extraction in order to improve efficiency

It's called progress.

Semiman

The die has always been textured, no change there.

 

[ergotelis]

That would only apply if you were dealing with a voltage controlled power source. LEDs are driven by a current controlled power source - that means that the current is held constant by modifying the voltage. So regardless of forward voltage (don't forget that changes as LEDs heat up and cool down) the current should be held constant.

Unless the light uses PWM in which case it may change a bit. But I know that at least Fenix and 4Sevens use current controlled drivers and no PWM.

Exactly. And moreover, this means, due to lower Vf, that we are going to have less energy spent, other than more light output. So, other than the classic lumen increase, in flashlights with good circuits, we are going to see a slight decrease in the current demand from the battery and higher runtimes.

 

[tstartrekdude]

Or longer time in regulation for the ubiquitous AMC7135 driven lights, something I am always paying mind to.

 

[ergotelis]

Or longer time in regulation for the ubiquitous AMC7135 driven lights, something I am always paying mind to.

You will have a bit longer runtime regulation in AMC7135 based drivers, but also a larger energy loss, because this driver is linear.

 

[SemiMan]

The die has always been textured, no change there.

Would you be happy if I said more "pronounced" texturing of the surface or faceting? .... I was trying to keep it simple. In theory you are a Cree employee ... or so you have said. Why not reveal who you are and shed some light ... as opposed to adding nothing really to the conversation?

 

[bbb74]

The die has always been textured, no change there.

Are you sure about that? I thought the texturing SemiMan is talking about is pretty new, the die has all these angular prism type things on it, I saw a picture somewhere but can't find it now...

 

[bshanahan14rulz]

I thought these weren't using that new flip chip faceted die, even though it would seem that Cree has a scaled up version the size of the ez1400 xpg die. I was wondering if anybody was gonna guess that Cree had made some sort of breakthrough in standard plane InGaN quality without having to move to non-polar nitride substrates, or that perhaps Cree has built some secret space station in outer earth orbit where they grow secret, massive, m-plane substrates.

Also, Cree has textured the light extraction layer for a long time. Flipping the chip means more material between the active layer and the surface of the chip, meaning easier to get internal reflections at weird angles, mandating some larger "texture" to reduce the incident angle of these very acute light rays and allow them to be projected more forward. You'll notice that the active layer is now at the bottom of the chip. And Philips made flip chips a while ago. I'm glad Cree has found a way to do so with their (imo) superior dice.

Also, here's a link to a pdf about the xp-c, this teardown was done back in 08, IIRC: http://www.systemplus.fr/plaquettes/MuAnalysis/Cree-XLAMP-LED-Lamp-Teardown-Report-short-version.pdf You can see the texturing on some of the pics at the bottom.

 

[saabluster]

Would you be happy if I said more "pronounced" texturing of the surface or faceting? .... I was trying to keep it simple. In theory you are a Cree employee ... or so you have said. Why not reveal who you are and shed some light ... as opposed to adding nothing really to the conversation?

To be honest it sounded to me like you were referring to the XT-E die which has facets. To my surprise and delight this XP-G2 does not have that. It also has two bondwires for which it seemed you were referring to them having been done away with like in the XT-E.
As an employee I think he would enjoy having some anonymity around here. Maybe he can't tell us what it is but feels more comfortable telling us what it is not. His call. Since he would bear the brunt of any repercussions from talking at all I don't think it kind to censure him for his decisions. Not that he has really ever said anything we don't already know.

 

[AnAppleSnail]

Would you be happy if I said more "pronounced" texturing of the surface or faceting? .... I was trying to keep it simple. In theory you are a Cree employee ... or so you have said. Why not reveal who you are and shed some light ... as opposed to adding nothing really to the conversation?

This company is VICIOUS about employees saying anything or revealing anything about how their products are made. Even describing in detail the geometry gives away knowledge they have developed. That geometry is carefully engineered to be easily manufacturable and effective for light extraction. I think the company wants to avoid any knowledge leaks that would let interested third parties imitate their expensively-made gains in performance.

I've seen dies like this before, but I can't say where.

 

[bbb74]

I thought these weren't using that new flip chip faceted die, even though it would seem that Cree has a scaled up version the size of the ez1400 xpg die. I was wondering if anybody was gonna guess that Cree had made some sort of breakthrough in standard plane InGaN quality without having to move to non-polar nitride substrates, or that perhaps Cree has built some secret space station in outer earth orbit where they grow secret, massive, m-plane substrates.

Also, Cree has textured the light extraction layer for a long time. Flipping the chip means more material between the active layer and the surface of the chip, meaning easier to get internal reflections at weird angles, mandating some larger "texture" to reduce the incident angle of these very acute light rays and allow them to be projected more forward. You'll notice that the active layer is now at the bottom of the chip. And Philips made flip chips a while ago. I'm glad Cree has found a way to do so with their (imo) superior dice.

Also, here's a link to a pdf about the xp-c, this teardown was done back in 08, IIRC: http://www.systemplus.fr/plaquettes/MuAnalysis/Cree-XLAMP-LED-Lamp-Teardown-Report-short-version.pdf You can see the texturing on some of the pics at the bottom.

Thanks for the explanation. I don't understand all of it but was interesting all the same.

 

[SemiMan]

This company is VICIOUS about employees saying anything or revealing anything about how their products are made. Even describing in detail the geometry gives away knowledge they have developed. That geometry is carefully engineered to be easily manufacturable and effective for light extraction. I think the company wants to avoid any knowledge leaks that would let interested third parties imitate their expensively-made gains in performance.

I've seen dies like this before, but I can't say where.

Every competitor has this product stripped down and under optical and X-ray microscopy. Can't hide geometry once it's released. Leaks prior to release are one thing but everything external is public knowledge now.

Semi man

 

[bose301s]

I thought these weren't using that new flip chip faceted die, even though it would seem that Cree has a scaled up version the size of the ez1400 xpg die. I was wondering if anybody was gonna guess that Cree had made some sort of breakthrough in standard plane InGaN quality without having to move to non-polar nitride substrates, or that perhaps Cree has built some secret space station in outer earth orbit where they grow secret, massive, m-plane substrates.

Also, Cree has textured the light extraction layer for a long time. Flipping the chip means more material between the active layer and the surface of the chip, meaning easier to get internal reflections at weird angles, mandating some larger "texture" to reduce the incident angle of these very acute light rays and allow them to be projected more forward. You'll notice that the active layer is now at the bottom of the chip. And Philips made flip chips a while ago. I'm glad Cree has found a way to do so with their (imo) superior dice.

Also, here's a link to a pdf about the xp-c, this teardown was done back in 08, IIRC: http://www.systemplus.fr/plaquettes/MuAnalysis/Cree-XLAMP-LED-Lamp-Teardown-Report-short-version.pdf You can see the texturing on some of the pics at the bottom.

Again, no difference there either. The n and p sides of the die are the same orientation in both XP-G and XP-G2. Basically this is taking some of the stuff from the new DA products and applying it to EZ based products. I was very surprised to see bond wires still since there are no current spreaders on the top of the new chip, still trying to find that one out myself

 

[bose301s]

Would you be happy if I said more "pronounced" texturing of the surface or faceting? .... I was trying to keep it simple. In theory you are a Cree employee ... or so you have said. Why not reveal who you are and shed some light ... as opposed to adding nothing really to the conversation?

You can say more pronounced if you want, but I am pretty sure EZ Gen 2 and EZ Gen 3 use the same exact texturing.

 

[SemiMan]

Again, no difference there either. The n and p sides of the die are the same orientation in both XP-G and XP-G2. Basically this is taking some of the stuff from the new DA products and applying it to EZ based products. I was very surprised to see bond wires still since there are no current spreaders on the top of the new chip, still trying to find that one out myself

I have not dedomed and put under the microscope yet, but I was expecting the bond wires may be to attach the transient protection device.

 

[SemiMan]

The XTE has a very pronounced faceting. Unless it is buried under the phosphor, it does not appear to have been done for the XPG2 on closer inspection. That could have been a tradeoff to ensure the exact same optical pattern in new fixtures without having to requalify the optical design ... and everything associated with that.