Ok, this is the first official release of an LED test from my setup. I have a few others already complete that I will be releasing as I get the time to make the charts all pretty and ready for upload.
Cliff notes version
- My sphere and setup is finally calibrated to a level that I feel good about releasing numbers from
- I figured out how to turn a PVC sphere into a true integrating sphere that reads the lumen value directly on the lux meter so no math is involved
- The lumen readings line up perfectly with the Cree data sheet confirming that the readings are indeed accurate
- The XP-L2 is a tank of an LED, handling up to 10A without any damage and over 15A with damage but still working!
- Over 2200 lumens (2450 peak from cold start) from a 3V single die LED in a 3535 footprint, that is over 40w of power through a single LED! It handled 70W at peak while still working!
First off thanks to djburkes for donating the LED’s! If anyone else has a new LED that you think would be worth putting through the ringer, let me know.
So I will start off by giving a quick rundown of my setup. I have 2 spheres actually, the Josh BLF “standard” sphere and a PVC sphere I made using the stuff I learned from Joshes sphere. Long story short, I have been dialing it in for the last few months to get it as accurate as possible, I am quite happy with the current setup. It is FAR FAR better then the plain jane PVC setup I started out with.
If people would like to see what I did to my sphere, let me know and I might be talked into making a thread on it, the biggest difference is that it is a true integrating sphere and doesn’t allow any light to escape. It also reads the lumen readings out directly on my lux meter, no calculator needed!
Now that it is setup to my satisfaction I can start releasing numbers from it as I will not be changing it any more. Thus my numbers will remain comparable. They are also pretty accurate if I do say so myself, this test for example lines up perfectly with the Cree data sheet.
For the test itself I have the LED’s all reflowed to DTP copper stars unless otherwise noted, they are then mounted on a PC heat sink with fan that I setup to hold said stars. The fan is running the whole time.
I then start the test and work my way up in current using my DPS5015 to power it. I stop every .25A to take a voltage, current and lumen reading.
The voltage drop over the wires is calculated out in the corrected voltage column, thus giving the actual forward voltage (although remember in a flashlight that resistance will not be subtracted).
Now to the test itself!
Ok, so djburkes sent me 2 XP-L2’s to test. I have to say I was not expecting much from them to start but I am VERY impressed with the results! These things are TANKS! I took this guy all the way up to 15 amps and it still worked! Although it was hurt pretty bad at that point.
The test really went great up to around 10A (A whopping 40W), the LED was not even hurt up to this point as I retested it at 1A and it had only lost 1 lumen!
After that it started getting stressed, by 12A it had been hurt a bit and after 13A the tint started turning a bit blue. Somewhere around 14A it started burning the dome, it literally cooked the dome off but even at 15A it was still kicking! I left it there for a little while as well. An astounding 70W of power was passing through it!
The spikes in the graph are when I retested from a cold start to show what would happen in a well cooled flashlight over the first 30 seconds or so. I generally try to time this around peak output.
I plan to install the second LED in a Convoy S2+ tomorrow and see what kind of results it gets there but I was too excited and had to post this test now.
So enough with the talk, time for data!
Click images for larger versions that are easier to read! Or download and zoom in all you want.
I am quoting a post from later in the thread where I put the second LED in a S2+ to see what it did in the real world:
LED’s preform just as good in a light as they do on the bench! If I can now just find a neutral tint version preferably in 90+ CRI with cheap shipping.Holy cow, theseI just built an S2+ with the second XP-L2 that was sent to me (I like to get 2 of any LED’s I test so that I have a spare in case something goes wrong or if the first test results are inconsistent for some reason, plus I like to install one in a light so I can see what it looks like in the real world).
Wow, this is easily the brightest single emitter tube like I have!
The really surprising thing is how much more efficient it is. With the TA17 drivers I have a regulated 350ma mode that generally nets me around 120-150 lumens depending on the LED, the highest I have seen was 180 with an XP-L V6 HD IIRC.
With the XP-L2 I am getting 200 lumens @ 350ma!
I then have a 2.45A regulated mode that normally nets me around 850-900 lumens, with the XP-L2 I am getting 1150 lumens!
Then we come to turbo, here are the results with different cells, these are all from a cold start, this is why I retest the peak output on the bench from a cold start to see what it would do in a flashlight.
Cheap laptop pull – ~6A – 1950 lumens
GA – ~7A – 2100 lumens
30Q – ~8.5A – 2300 lumensNow of course things get hot fast and the output drops fast as well but I didn’t feel like testing each at 30 seconds. Still, 2300 lumens @ turn on from an EDC S2+ is seriously impressive!
It is also nice to see that the cold start test on the bench is indeed a good indication of what happens in a real light. It also explains why the numbers people report from flashlights are sometimes higher then what djozz’s tests says they would be.
After I got done being impressed with it on the sphere I then started looking at the beam.
Edit, just built a 5700k 90+ CRI XP-G3 for direct comparison. Updated the rest of the post accordingly.
First thing I noticed is that it is much like the XG-G3 with the tint shift as others have mentioned around the edge of the beam. The tint shift is basically the same as the G3, the tint plays a larger factor then the type of LED I think.
Compared to the 5700k G3 I would say the tint is close in the hot spot although the L2 has more red in the beam giving it a slight purple hue when white wall hunting. When used in the wild it is actually not that bad of a tint, but still too cool for my tastes.
I am thinking I would want 4000k to get a neutral tint, this one is far too cool for my liking, CRI is not about average for a 70CRI.
The beam size is naturally larger then the G3 but not a ton, maybe 25% larger.
Overall except for the cold tint I really like this LED, the tint shift is not ideal but it is livable for the performance you get. A good TIR lens that mixes the beam would really improve things, the TIR I am using helps but doesn’t fix the tint shift.
If these can be found in neutral tints and 90+ CRI I can easily see these becoming the go to LED’s for myself.
In other news I was comparing the graphs from the XP-L2 and XHP35 (test not released yet) and I am pretty sure that we are reaching the limits of the 3535 footprint at around 40W. Both LED’s have the same general curve shape when looking at wattage, I am guessing that 40W is just the limit for how fast the heat can be moved out of the LED no matter what star you have it mounted on. Honestly that is seriously impressive for that tiny footprint.
This does make me look forward to an XM-L3 to be released with this latest die tech, it is possible it could reach even higher numbers due to the larger footprint.
EDIT 1-16-17
I tested an XP-L V5 as a comparison and the results were interesting. Keep in mind the lower Vf of the XP-L2. When compared by wattage the XP-L2 is slightly ahead of the XP-L the entire time. AKA, the XP-L2 is more efficient overall.
Also in the real world the lower Vf means we get much better numbers from the XP-L2 then the XP-L since we can pull much higher currents.
My LED Test series - XP-L2 V5 - Nichia 219C 90+ CRI - Latticebright "XM-L" - XHP35 & PWM efficiency - XHP50 - XP-L V5 - XM-L2 U2 - XP-G3 S5 - XP-L HI V2 - Oslon Square & direct comparison to Djozz tests - Nichia 319A - Nichia 219B 9080 CRI - Nichia 219C D320 - Nichia 229AT - XHP70.2 P2 - XHP50.2 J4 - Samsung LH351D
If anyone has any comments on the layout of the charts, the data in them or anything else please do speak up. I would like to standardize my test layout now before I release anymore so that they are all easily comparable.
Texas Avenger Driver Series
My LED Test series - XP-L2 V5 - Nichia 219C 90+ CRI - Latticebright "XM-L" - XHP35 & PWM efficiency - XHP50 - XP-L V5 - XM-L2 U2 - XP-G3 S5 - XP-L HI V2 - Oslon Square & direct comparison to Djozz tests - Nichia 319A - Nichia 219B 9080 CRI - Nichia 219C D320 - Nichia 229AT - XHP70.2 P2 - XHP50.2 J4 - Samsung LH351D
Easy comparison tool for all my LED tests
Holy crap
What if you liquid cooled them when doing testing to keep temps below 50C?
Do you think you would get more amps through it before the output starts decreasing?
The OPTOFIRE - 4.63Mcd aspheric LED flashlight The SYNIOSBEAM - 10Mcd recoil LED flashlight List of the farthest throwing flashlights
Thanks for testing. Interesting info, but what’s more interesting is your pvc sphere! It sounds like a winner. How much did it all cost to build? And would it still the same amount of time to get another sphere up and calibrated since you have experience now?
Do you have a set of standard XP-L data to overlay with?
@1theDeals
If you mean like liquid nitrogen then yes, it would do a lot better
If you mean like the normal PC style setup liquid cooling, it would improve things a little but much like PC’s the cooling is still limited to the ambient air and as such liquid cooling would not drop temps all that much once the internal fluid was heat saturated.
I could actually improve cooling on this setup more then it is now but I did not was the results to be too far off from what we can expect in a flashlight for at least the first 30 seconds. It is already cooled way better then a flashlight, more would simply further remove it from what we would see in the real world.
In the end with improved cooling I think it could make another ~100 lumens but we are reaching the limits of how fast the 3535 footprint and copper star can transfer the heat away from the die more then the heat sink itself.
If this was in a 5050 footprint I am guessing it would make a reasonable amount more lumens all else being the same.
Texas Avenger Driver Series
My LED Test series - XP-L2 V5 - Nichia 219C 90+ CRI - Latticebright "XM-L" - XHP35 & PWM efficiency - XHP50 - XP-L V5 - XM-L2 U2 - XP-G3 S5 - XP-L HI V2 - Oslon Square & direct comparison to Djozz tests - Nichia 319A - Nichia 219B 9080 CRI - Nichia 219C D320 - Nichia 229AT - XHP70.2 P2 - XHP50.2 J4 - Samsung LH351D
Easy comparison tool for all my LED tests
I am quite happy with how the sphere turned out. I have not wanted to really post about it till Josh finished with his GB on his spheres. They are nice spheres, just not really suited to the power of flashlights I deal with on a regular basis. It is also many times the size and it was hard to fit it on the desk.
The ideas I got from it through are the only reason I was able to build mine, I blatantly stole the ideas from him. They work.
As far as cost, it is normal PVC P trap sphere with a few 3D printed parts, total cost was around $35 and a lot of time dialing it in, excluding the meter (cheap H1010A that works great, I have 2 of them and they both read exactly the same, purchased 6 months apart from completely different suppliers).
I suppose I could build more of them, they would not be very cheap due to shipping and time involved with calibrating them. Although I suppose I could try an idea I have had to “standardize” the 3D printed parts and it is possible that it would be consistent enough for what we do.
If people really want to see it I will snap some pictures and make a thread now that Joshes GB is done and he is going open source I feel ok doing it.
Texas Avenger Driver Series
My LED Test series - XP-L2 V5 - Nichia 219C 90+ CRI - Latticebright "XM-L" - XHP35 & PWM efficiency - XHP50 - XP-L V5 - XM-L2 U2 - XP-G3 S5 - XP-L HI V2 - Oslon Square & direct comparison to Djozz tests - Nichia 319A - Nichia 219B 9080 CRI - Nichia 219C D320 - Nichia 229AT - XHP70.2 P2 - XHP50.2 J4 - Samsung LH351D
Easy comparison tool for all my LED tests
I don’t have any XP-L’s out of lights with “trustworthy” suppliers with which to compare right now. Actually I might have one from Convoy come to think of it, I will have to check.
I don’t want to test some random XP-L and let an unknown bin skew the results. Sadly I don’t have many LED’s with known bins left in my spare parts.
Although all my “unofficial” tests thus far have been spot on the cree data sheets.
If anyone has some spare known bin emitters that I could use to get some “baseline” results for my setup for comparison sake let me know.
Texas Avenger Driver Series
My LED Test series - XP-L2 V5 - Nichia 219C 90+ CRI - Latticebright "XM-L" - XHP35 & PWM efficiency - XHP50 - XP-L V5 - XM-L2 U2 - XP-G3 S5 - XP-L HI V2 - Oslon Square & direct comparison to Djozz tests - Nichia 319A - Nichia 219B 9080 CRI - Nichia 219C D320 - Nichia 229AT - XHP70.2 P2 - XHP50.2 J4 - Samsung LH351D
Easy comparison tool for all my LED tests
Where can these be purchased?
The ones that were donated to me were originally purchased from Mouser. They are only V5 bin though, there should be V6 and even W2’s floating around before long (although W2 could be awhile).
I am very interested to see how the other one looks in a light.
Texas Avenger Driver Series
My LED Test series - XP-L2 V5 - Nichia 219C 90+ CRI - Latticebright "XM-L" - XHP35 & PWM efficiency - XHP50 - XP-L V5 - XM-L2 U2 - XP-G3 S5 - XP-L HI V2 - Oslon Square & direct comparison to Djozz tests - Nichia 319A - Nichia 219B 9080 CRI - Nichia 219C D320 - Nichia 229AT - XHP70.2 P2 - XHP50.2 J4 - Samsung LH351D
Easy comparison tool for all my LED tests
Nice work! The charts look good to me.
Let's hear more about that sphere.
WOW
I want some !
Nice bit of testing TA
I like the rows with numbers too, in this case showing this XP-L2 will still do 100LM per Watt at 15½ Watts, at just over 3½ Volts !
Very nice work…I was hoping it was something that I was doing wrong because I really didn’t want these to fail. I’m going to say it was the product of a bad reflow on my part…you live and you learn.
Those are some freakin’ awesome LED’s! Nice tests suite as well. Gotta see the sphere as well.
EDC rotation:
FW1A, LH351D 4000k (second favorite)
FW3A, LH351D 3500k
FW3A, SST20 FD2 4000k
FW3A, Nichia 4000k sw40 r9080 (favorite light!)
FW3A, Cree XP-L Hi 5A3
Emisar D4V2, SST20 4000k
S2+, XM-L2 T6 4C
I think it would be awesome if you started another thread for your pvc sphere. I also have Josh’s sphere but I wouldn’t mind having another. Doing calculations and the off wall can be troublesome if your just in a hurry. And 80% of my lights don’t fit the opening and is too powerful too many lumens.
Cool, thanks. Regarding the graphs, would it be possible to enlarge or bolden the numbers ever so slightly?
- LEDs & Other Stuff - - UI CheatSheets - - CRX Flashlight Builds - - CRX Stuff for Sale -
Hi TA, about your set-up,
1) you may want to check for integration quality of the device.
If that is not good, your calibration will only work for emitters with comparable emission profile. For example, a flashlight that uses optics to make a radical different beam from a bare led, will read different from a bare led. Every type of emission pattern will require its own calibration.
If integration is good however, it does not matter what type of light source you measure, a fixed calibration is inherent to the design.
The easiest way to check integration is using a small zoomie in spot modus on low setting (actually the goal is a constant output, and flashlights are more constant on low settings, best is a constant current setting, like the lower settings on the BLF-A6 driver that uses an 7135, same for an old-school AK-47 or similar). Shine it into your device at various angles and record the reading. My sphere gives a maximal variation of 3% for any angle that the zoomie is pointed in. See here.
2) you may want to check for the influence that the to-be-measured lightsource in the entrance hole has on the calibration.
Unfortunately, with better integration, this influence increases. Your lightsource is part of the inner surface of the device where photons bounce around. In a well integrating device, photons will bounce around multiple times before being absorbed somewhere, so the photons have multiple chances to ‘meet’ the entrance hole. The entrance hole is a photon ‘leak’ which is is no problem at all because it is accounted for in your multiplier. However, if this leak is not constant (i.e. a shiny bezel shines light back into the sphere, while a black bezel absorbs the light) your multiplier will be influenced by that.
There is two ways to deal with that. Either you can accept this variation in which case you want to have an idea which order of magnitude this error is, or you are going to measure what the error is for each light source and correct the multiplier for that. you then need a build-in constant light source in your device for measuring the correction.
In my big 46cm sphere with 30mm entrance hole, that I use for emitter measurements, I find that entrance hole effects are minimal, I have a fixed multiplier for all emitter tests. For bigger flashlights I open up the hole to 80mm, in which case I do have to correct for the lightsource, if I do not do that, the error can be up to a few percent.
In my small 16cm sphere I need to compensate for the different reflectivities of what is in the 40mm hole every time, the error would be too large for my standard if I do not do that.
link to djozz tests
Xp-l2, maybe you are not for thrower, but you are a little monster
Forgot my pen
I really like the addition of wattage column! Not every LED has same Vf to begin with. Wattage is a better parameter to define an LED performance. Great job Tex.
- Clemence
www.virence.com
Physical and thermal properties of solder alloy - Shooting Beamshots With HDR - Nichia E21A Tint Shots
Its Okay To Be Smart - SmarterEveryDay - Deep Look - Veritasium - Tech Ingredients
True, but I always like using overkill cooling solutions so it’s not like my flashlights will perform like regular flashlights
Liquid cooling would still be able to keep temps lower than 85C when the LED is producing nearly 100W of heat or more.
You can get a lot lower with phase change though, sub-0 cooling for an LED
The OPTOFIRE - 4.63Mcd aspheric LED flashlight The SYNIOSBEAM - 10Mcd recoil LED flashlight List of the farthest throwing flashlights
Nice tests!
) topic on that tube!
Yeah please make a (pic heavy
It is done when it is done

How my BLF Specials generally work, please read before asking Qs
The Q8, Exciting a groupbuy for a Premium BLF special high lumens soda can light!
The GT, A BLF special GIGA thrower
The FW3A, a TLF BLF special small, elegant, powerful triple
Lithium Ion safety 101 important read
Sorry grammarlovers, I am a real King Typo.
What he said!
KB1428 “Live Life WOT”
Thanks, I will see about making a thread for the sphere in the coming weeks.
Yeah, I liked this setup, it was how I had it setup in Excel and figured a screen shot of the whole thing could work well.
I am thinking about adding some more numbers that will extrapolate the readings out for higher/lower bins based on the data sheet specs of 7% per bin. Anyone seen an example where this doesn’t hold true in the real world?
Look at the bright side, we now have a nice test to show what they can do!
Thanks, I am very impressed with the output, hope to build a light with the other one to see how the beam looks later.
Yeah, if people keep asking I will just have to get off my butt and do it. I didn’t think there would be a lot of interest with so many threads already floating around on how to make a sphere.
Hmm, have you clicked the impages to view them full size? You might have to click them again after they come up full screen to zoom in all the way depending on your screen size. On my screen anyways the numbers are readable if zoomed into full size.
I can try to make them a bit larger if needed though.
Indeed they are! Now to find some 90+ CRI versions or even some higher bin versions, ~2500 lumens should be possible with a W2 bin.
Thanks, I figured it would not hurt to toss those in since they are just a matter of simple math.
Liquid cooling in the classic computer sense is mostly about space savings then actual cooling ability. I for example have always used air cooled setups in my overclocked computers and while it can be very hard to fit the heat sinks the final temps are only slightly higher then those using water cooling with the same setups.
Same with a turbo cars intercooler. after the system is heat soaks all the matters is how fast it can transfer heat to the ambient air. This is determined but surface area and ambient temperature more then anything else. In most cases the air cooler actually has the advantage in surface area (unless using a massive waterblock) and this makes up for some of being inside the computer where the ambient temps are higher.
lol, well if people really want to see it then I guess the people must be appeased. I will see if I can find the time in the next few weeks.
Texas Avenger Driver Series
My LED Test series - XP-L2 V5 - Nichia 219C 90+ CRI - Latticebright "XM-L" - XHP35 & PWM efficiency - XHP50 - XP-L V5 - XM-L2 U2 - XP-G3 S5 - XP-L HI V2 - Oslon Square & direct comparison to Djozz tests - Nichia 319A - Nichia 219B 9080 CRI - Nichia 219C D320 - Nichia 229AT - XHP70.2 P2 - XHP50.2 J4 - Samsung LH351D
Easy comparison tool for all my LED tests
All good points.
1: Yes, this was my biggest issue with the “normal” PVC setup, where you put the light would change the reading drastically and that was unacceptable for me. Thats why I set out to improve it and it is now truly integrating as good as the josh sphere anyways. It doesn’t matter where I point the light inside the sphere the readings only change by ~3-4% tops.
Although I have centering rings 3D printed that ensure the light is always in the same place and this means that the readings are even more stable in real world use, the depth that the light is inserted into the sphere plays a FAR larger role then where it is pointed. That can effect the readings by 10% or more easily.
2: This is an almost non-issue for me as I have 3D printed up a wide selection of “centering rings” based on the design that josh uses. These do several things at once. First they are perfectly sized to the head of the flashlight I am testing, allowing no light to escape. Second since they are covered in reflective foil that all act like a large reflector so the reflector in the light itself has minimal effects.
The light also bounces around a lot before reaching the meter so it helps balance out small variations. Now lights that have a thick black bezel do effect things a bit but there is nothing that can be done there.
Texas Avenger Driver Series
My LED Test series - XP-L2 V5 - Nichia 219C 90+ CRI - Latticebright "XM-L" - XHP35 & PWM efficiency - XHP50 - XP-L V5 - XM-L2 U2 - XP-G3 S5 - XP-L HI V2 - Oslon Square & direct comparison to Djozz tests - Nichia 319A - Nichia 219B 9080 CRI - Nichia 219C D320 - Nichia 229AT - XHP70.2 P2 - XHP50.2 J4 - Samsung LH351D
Easy comparison tool for all my LED tests
smart to cover the inserts with reflective foil, it was one of my objections to Josh’s sphere that he used inserts that alter the reflectivity of the sphere, in your case I agree that it will create a fairly constant large reflector for any (edit: reflector- )flashlight.
Looking forward to your thread!
link to djozz tests
A 280mm radiator AIO performs better than any cpu heatsink that exists though…
The point of the water is to move the heat away faster than copper and heatpipes do, not to soak up heat.
The coolant reaches a constant temp within minutes, there is very little and it does not absorb much energy, usually stays below 40C.
The OPTOFIRE - 4.63Mcd aspheric LED flashlight The SYNIOSBEAM - 10Mcd recoil LED flashlight List of the farthest throwing flashlights
I think josh changed his design to include the reflective tape based on you if I had to guess. My sphere showed up with the reflective tape when I got it and it is where I got the idea to use it on the PVC version.
The change combined with the rings made a massive difference in consistency. The other big change was the internal baffle. Although that is a bit hard to explain here although it is quite simple really and works amazing.
Texas Avenger Driver Series
My LED Test series - XP-L2 V5 - Nichia 219C 90+ CRI - Latticebright "XM-L" - XHP35 & PWM efficiency - XHP50 - XP-L V5 - XM-L2 U2 - XP-G3 S5 - XP-L HI V2 - Oslon Square & direct comparison to Djozz tests - Nichia 319A - Nichia 219B 9080 CRI - Nichia 219C D320 - Nichia 229AT - XHP70.2 P2 - XHP50.2 J4 - Samsung LH351D
Easy comparison tool for all my LED tests
Thanks for the tests! Looks very promising…
About the improving the chart, I do have a few small suggestions:
-Place lumen on the primary axis and Vf on the secondary for easier reading
-Reduce Vf axis range to 0-5V, and remove the unused decimals for current and vf
-Bigger font, adding axis titles optional
After this you’ll have a very nice chart that you can use for future tests, and adding more lines just requires a few clicks.
USB power meter/tester thread
1: I swapped the axis’s because most people are more interested in the high end range of the lumen curve vs the low end. This makes the high end easier to read I figured. The right side axis is what all the guidelines are based off of. If people would prefer the lumen axis on the left side that is easy to swap.
2: I increased the Vf range so as to move the VF line away from the lumen line. It was confusing when they were overlayed in the same range although that was before I added the colors. I could knock the VF down some to put it more in the center of the chart if people think that would be easier to read.
3: Which font are you referring to? I should add axis titles, I forgot to do that.
Texas Avenger Driver Series
My LED Test series - XP-L2 V5 - Nichia 219C 90+ CRI - Latticebright "XM-L" - XHP35 & PWM efficiency - XHP50 - XP-L V5 - XM-L2 U2 - XP-G3 S5 - XP-L HI V2 - Oslon Square & direct comparison to Djozz tests - Nichia 319A - Nichia 219B 9080 CRI - Nichia 219C D320 - Nichia 229AT - XHP70.2 P2 - XHP50.2 J4 - Samsung LH351D
Easy comparison tool for all my LED tests
For that reason I let the Vf axis not start at zero but a bit under the minimum measured voltage.
link to djozz tests
Pages