Cree XP-G3, testing a S5 3A emitter

Thanks for the long awaited review djozz.
Do you think that this emitter should be my choice for running small triple/quad lights on 14500 cells? I’ve mostly even using XP-E2 (because I have 50 of them), but I need a change up.

These things are absolutely amazing when fully dedomed. I put one in a reflector light last night, and the lumens looked almost identical to the XPL next to it, but with a pure white hotspot 2/3 the size. I honestly would not trade the dedomed tint for anything I have seen yet dedomed. But 7-8 amps at max output? Holy…crap. YES! :partying_face:

Nice, have you posted your method for dedoming these?

No… He’s teasing us;) but he’s been graciously giving hints along the way.

Hi djozz

Very useful as always! :student:

Many thanks

Have you taken any lux measurements? As we have seen with the 219C, just because it is dedomed does not mean that the process will double the luminance like with the XP-G2.

So, de-domed properly, these should out-throw the old style XP-G2 S4 2B. Is that what you guys are saying? I’ve got a couple of the XP-G3 from the VOB group buy from Cutter. I would really like to know how to de-dome one of these. I’ve never had a XP-G2 S4 2B light. I’d really like to see what it’s like. :innocent:



You might try the method commonly used to dedome other LED’s with embedded phosphor: by using a razor.

Based on Djozz’s excellent data, in theory the G3 would be more intense than a G2 S4 until the die approached 1.7mm² die size. That’s assuming the G2 is 1.5mm² die size.

I haven’t scraped a G3 to measure the actual die size. Can anyone confirm what the G3 die actually is who has scraped the phosphor off to measure?

Just looking at it, I would say there’s no way it’s 1.7*1.7mm.

I am putting these dedomed emitters in M3XS-UT Javelots, and I will be posting the throw photos up by morning now that I have a good mix of LEDs for comparison all in the same hosts. It should be interesting to see the results against the G2, which will be compared in the beamshots.

EasyB, luminance may not double, no, but a dome is a magnification optic. Any optic which magnifies also decreases luminance. I will be writing more on the details later about these LEDs. So far, I am definitely impressed with it being dedomed in a reflector light, though.

Instead of a razor could one use a cnc machine and a very sharp bit to shave the dome down to a very exact height? Would this give the flat dome a nice and even finish? just a wild guess…

I’m excited about this new LED. Its performance results above are indeed promising for it to be a good thrower.

MEM, I look forward to your measurements.
My understanding of luminance (cd/mm^2) is that simple optics cannot change it. Most of my understanding on this subject is based on Dr. Jones thread: Flashlight Optics - Dome, Dedoming and Throw
In it I learned it’s not simply the removal of the magnifying dome which increases the throw. Removal of all of the silicone material results in a sort of light-recycling, which takes some of the light that would have left the LED at shallow angles and puts it back on the LED surface, increasing the luminance. In fact Dr. Jones compares it to a hemispherical reflecting aperture.
I’m a little fuzzy on the details. The test performed in this thread helped convince me.

I doubt this would work, dome is not that firm material, it’s silicone, even if you crank the speed up and set a really slow feed, it might just rip it off at some point, also, in case it works, I doubt you would get clean surface of the remaining silicone dome, no milling bit is sharper than razor + when you use razor idea is to do the job in one slice in order to achieve as flat surface as you can!


I think they are saying that XP-G3 emitter will out throw classic XP-G2 but only if we leave dome on.
And since 80% of us here do de doming, and expecting better performance with newer types of emitters it seems such performance upgrade will not happen with G3.

Wait until someone confirms kcd numbers in well known hosts. I know I would not be impressed with anything that will throw less than old production XP-G2 S4 2B de domed no matter how nice tint is.

Djozz sliced XP-G3 and did put it in well known and consistent performance aspheric host(B158) and results are really not good.

So when it will be pure clean de dome only 10-20% gain could be achieved(in my experience) and not even that will be enough to dethrone old de domed xpg2s42b

Djozz wrote:
How about dedoming this led that is build quite different than the XP-G2 ? Some first attempts by other BLF-members at hot dedoming or in gas were not promising. My own (limited skills !) first attempt at a sliced dome produced a working led but the throw was disappointing. See here. So for now, the ’king of throw’ is still the ’old type’ XP-G2 S4 2B, which unfortunately is not available anymore. Cree changed production process for XP-G2 emitters! - #221 by djozz

I believe the tests from Djozz are not that helpful in determining the g3s ability as it is not dedomed. It still has silicone covering the phosphor.

Luminarium has a tendency to form out-of-context passages from other authors into tidy, neat statements saying exactly what he wants to hear the first time he reads them. :stuck_out_tongue: :smiley:

Totally kidding, Luminarium! :laughing:

No but honestly, these fully dedomed G3s do something quite odd or even possibly incredible for some people I would imagine. After dedoming, the hotspot becomes about 5800-6000k. Reverse tint shift! It’s such a neat dual tone hotspot that I’ve begun to like it, a lot.

The die has more phosphor at edges, less in the center, like a dip when dedomed. I have macro photos I will post after I get back from the store of what that looks like.

My video of XPG3 full dedome using a diluted fuel mixture to slow the process down, is here: MEM Dedomes Cree XP-G3 S5 3A on Vimeo

Why would you say that we are saying “with the dome on”, when I’m talking about a domeless LED? Once again you’re modifying what has been stated by others. :stuck_out_tongue:

A good experiment is not necessarily putting the LED in a host you are particularly fond of and getting one throw measurement from a sliced-dome LED. To be fair, a sliced LED is going to vary person to person, method to method, slice to slice.

A proper experiment would be using any host, with a “control LED” installed at the start. Then, other LEDs would be placed into the same host, directly comparing dedomed XPG2s to dedomed XPG3s. It makes no sense to compare a “sliced” G3 to a completely dedomed XPG2. If you want max-to-max comparison, then they should use the LED in its maximum intensity form (at maximum output power), not guesswork of theoretical gains it could have in other scenarios.


Can you report in this or any other thread throw or lux values of de domed XP-G3 emitter in direct comparison with old XP-G2 S4 2B…? I say old type and that is the key… You can’t compare them with new production G2’s cause although they share same name they ain’t same emitters. Since you are the only person who knows how to de dome it it would be much appreciated in community.

What I know (and probably Djozz you and other guys here) that with “slicing the led method” vs “de dome method” difference can be very little from 5-20% depending on how much silicone did you sliced…

So If Djozz got 200kcd with sliced XP-G3 in well known B158 host (and it really looks like good slice to me):

and he make that test in direct comparison with old type de domed xp-g2 s4 2b that gives 320 kcd that should tell a lot, cause even your properly de domed G3 if we assume that it will have maximum gain of 20% over Djozz sliced one it will still not beat old XP-G2 S4 2B. It will have about 240 kcd in same host with more current sag in dd light. So nicer tint? OK I respect nicer tint but I always choose better performing emitter.

Don’t forgot you claimed that new production process XP-G2 is good and that you can’t spot differences between new and old type? Although imho you are terribly wrong on that and I (and lot of other guys from that thread) have more than one arguments for that(and still waiting for your reply in that thread).