XP-L vs XM-L2 OTF Tests

I don't know either why the results differ, bare emitter or in a flashlight, differences between leds should come out equally well. I tested two XP-L V6 2A leds and they tested nearly the same.

Last week I did a test twice on XP-L U5 7A3 80CRI leds (from intl-outdoor, unfortunately I killed them both) and they put out 1650lm at 7.5A, while the 2A V6 puts out 2120lm at 7.5A . These results seem to match the bin-difference between the two XP-L's, I have no idea why the XP-L V6 tests similar to a XM-L2 T6 (2100lm@7.5A).

BTW, my XM-L2 bare led result is a bit higher than match's test 1.5 years ago (match:1740lm@6A, djozz:1920lm@6A), assuming that match used a T6-bin, he did not mention the bin. (our spheres use the same luxmeter (Ceto 1330B) so that can not be much of a source of error between us)

Later in the thread, match said U2 bin for his tests. He should have updated the OP - didn't though.

If you look at the binning and labeling of XM-L/XM-L2 you will see that for making a "good choice" with not tons of tints you got group E1 or E2 for cool whites, basically U2 E1 or U2 E2 so as far as CREE is concerned they sell you U2 brightness and for them E1 (1B, 1C, 1A, 1D), E2 (2B, 2C, 2A, 2D) or other tints (2T, 2R, 0C) fits the U2 brightness bin.
As we know the min. max. difference is about 6% from the least bright T6 to the most bright T6 (as an example). A much higher difference than that should be put on the lightmeter's accuracy levels, various temperature factors, etc.

I’ve been looking at the Cree spec sheets and come to some conclusions,while drinking 6 cans of Pelforth blonde(x% error :beer: )For the XP-L and XM-L2 emitters,the binning is the same,all at 85°c,T4 is 240-259 lm,T5 is 260 to 279lm,T6 is 280 to 299lm,U2 is 300 to319lm,U3 is 320 to 339lm…V6 is 480 to 499lm _but_the XM-L2’s are measured at 700mA and the XP-L’s at 1050mA,50% higher current.

The Cree data sheet for XM-L2’s give min lumen flux at 1,1.5 and 2 amps,the XP-L’s for 1.5,2 and 3 amps.Extrapolating the XP-L V4 from 440 lm min at 1050mA gives 420 lm min at 1000mA.The XM-L2 U2,at 1000mA, has 412 lm min.So XP-L V4=XM-L2 U2.The V5 has 460 lm min at 1050mA,438 lm at 1000mA,XM-L2 U3 has 439 lm min at 1000mA,data not there for U3 so assume same efficiency drop with current rise as U2.V5=U3.So what is a XP-L V6 equivalent?XM-L2 U4 or XM-L3 I want a V6 :bigsmile:

Billy, maybe 6 more cans of Pelforth blonde will enlighten you and then you can share with us the final results of your query…

But first…. answer this riddle

If you spell coast C*O*A*S*T
and you spell boast B*O*A*S*T
What do you put in the toaster?

I figured it out! I figured it out! I’m so proud of myself… :bigsmile:
If a blue house has blue bricks, and a red house has red bricks, what color bricks does a green house have? :wink:

I suppose almost any colored brick could be used to smash the glass of a greenhouse…

:bigsmile:

…or the Cree measurement error in binning. One the datasheet they state “Cree maintains a tolerance of ±7% on flux and power measurements.” That’s an error in binning basically the size of a one bin step. Even before any errors produced in testing by an end user that nominal T6 could be putting out low T5 or high U2 output when it shipped.

"Cree maintains a tolerance of ±7% on flux and power measurements." - this is so CYA all the way. If you were CREE and dominant a market as they do, you can get away with pretty much anything. This reminds me of our touch screen vendor at work. Get the things out in the field for a few months, and customers have to touch the displays harder and harder. Send back a bunch of returned pieces to the vendor for testing, and they say all are within spec -- duh, the spec's tolerance is so big, it's useless! So, field service tells the customer to roll your finger - make contact with the finger nail and it works fine... This is not a solution, so aggravating...

Lumens XM-L/XP-L
1 amp 1.5 amp 2 amp
T5 357 502 631
T6 385 541 679
V3 404 566 708
U2 412 580 728
V4 423 593 742
U3 439 617 777
V5 443 620 776
V6 462 645 809

All data recalculated from Cree XM-L2 and XP-L data sheets.Minimum lumens from CW with 85°c junction temp and Cree 7% + or - tolerance.T and U bin are XM-L2,V bin are XP-L.

The XP-L V3 bin sits between XM-L2 T6 and U2.The U3 and V5 bin are effectively the same.

The V6 bin is ahead on it's own,one bin higher than the highest,and still rare,XM-L2 U2 bin.

What's not to shout about,it's the new highest bin emitter from Cree.

I'm still practicing formats,sorry it is not easier on the eye.

Billy X, I already posted the images from PCT.CREE.COM for comparing the same current on XP-L V6 vs XM-L2 U3, XP-L V5 vs XM-L2 U2, XP-L V4 vs XM-L2 T6 all at 25C, check the previous pages on this thread.

Hikelite,it was your post,along with the OP graph that got me trying to match up XM-L2 and XP-L bins.

Hikelite said

Is this the first test showing that CREE was not exaggerating their ratings for the XP-L LED?

They have never done that, so that is why my question was, why would they do it now?

Nice test pflexpro!

There it is.

I got a little motivated by Selfbuilt’s excellent analysis of the L3 Illuminations L10 on CPF so I brushed off some old and very basic statistics.

The graph we have only shows means for the two types of emitters tested. Without the standard deviation for the XM-L2 and XP-L a t-test isn’t possible. I decided to make up SD for both to redresh myself on a t-test and now I am sharing. My absolutely wild as guess for SD was based on 1.5% of the mean for each tested emitter. I took the lowest SD from selfbuilt’s analysis as a percentage of it’s mean (XP-G2 on high) and just divided it by two. I believe it’s a very conservative assumption. Still I want to say this is a giant assumption given the lack of raw data to compute the variance.

I used the mean for each emitter type at 30 seconds, the assumptions for SD, and samples sizes of 2 and did a simple t-test computation (alpha=.05, 2 degrees of freedom, two tail test). More information on what I did is in selfbuilt’s CPF review and all around the net. I leave the rest to the reader who wants to torture themselves. What I found…

There is not a statistically significant difference between the output of the XM-L2 U2 1D and the XP-L V5 2A in the data from these tests (again given my SWAG assumption about variance.) Standard deviation would need to be even lower than I assumed (given that all the other data need is known) to find a likely difference.

Long way around saying “I don’t know.”

For three words I’d go “no proven difference” as a summary.

No proven difference may be flattering, but when one must take “Wild as guesses” and “giant assumptions” then the truth is, you just don’t know. And that, my friend, is usually a good time to stay silent.

A lot of people read these threads, and a great many of those are trying to learn. So wild guesses and assumptions to base a theory of standard deviation is just blatantly wrong and misleading. No offense intended, but without a series of tests to establish facts it’s truly just guesswork.

Ever since the XP-L came out, people have been trying to prove that it’s better (or worse) than the XM-L2. It’s different. And the same. It does pretty much the same thing, with the ability to fit more dies in the same area, thereby achieving Cree’s target goal for array’s and such use. For us? (flashaholics) This small footprint allows us to put the big die where it previously wouldn’t fit. And in that instance, it’s very much appreciated.

The XP-L V5 2A has enabled 3129 lumens in my new Sinner Ti XC with the use of one 18650. And 2960 lumens from a small all copper Sinner Cypreus, from a single 18350 cell. They’ve enabled a AA MiniMag to make 1663 lumens with 3.7A from a Qlite. 921 lumens from a single 10440 cell in a 3” custom Ti light. As far as I’m concerned, this is a win all the way around. Not one of these lights can use the bigger substrate of the XM-L2. Not in the configuration that yields these results. Standard Deviation be danged.

Edit: The Ti XC is using a Carclo 10511 TIR Optics, polished. The Cypreus a 10507. The AA MiniMag is using a quad. All these utilized de-domed XP-L emitters.

The output of statistics more often than not is the same as what we thought already, but it really is a valid way to be more certain about uncertainties. The problem is that people are notoriously bad in dealing with uncertainties, and therefore with the outcome of statistics.

That said, if statistics is applied to unsufficient data, it does not magically make it good data. it can only make it look good :bigsmile:

Thanks for the review, seems consistent with one bin higher.

Where can I get some XM-L SMO reflectors? The ones I bought from Solarforce don’t fit their XM-L or XM-L2 drop ins.