Results: Direct bonded XM-L emitter to Copper


Here are my results from testing a Cree XM-L T6 that I directly soldered to a pure copper slug. Other than eliminating the star, this test was performed much like my other emitter tests.

And without further ado, here are the results:

I have super-imposed the new data on top of my previous xm-l star mounted data for an easy comparison.

To say the least, I found the data shocking! I knew these pcb's were holding emitters back, but never thought it would be to this extent. Also, I only expected to see a noticable divergence above @3 amps, but the data shows there is a difference even as low as 1.2 amps.

Not only that, but the emitter keeps chugging along all the way up until it maxed out my test bench power supply.

All datapoints are lumen snapshots 30sec after power on. Once the test was complete, out of curiosity I set the power supply to 5 amps and let it run. After 5 minutes, lumens dropped from 1367 to 1357 for a 10 lumen drop... in a word - Wow!

Now, how pratical is this for flashlight use? Having used my modded 1C mag with a very similar set-up as this I can say it definately has its merits. True, the body and head of a typical maglite will get heatsoaked well before the 25lb block of aluminum used in this test, but still the difference is dramatic. One could now have a true turbo mode to be used for brief periods and actually get the increased output from the power spent.


Hopefully the info below will help answer some of the questions I've seen pop up in the comments.

* The procedure I used to create the copper/emitter bond is exactly as shown here.

* The dimensions of the above copper slug are 21.5mm diameter and 5mm thick. This in turn was lapped to the aluminum test block and secured.

* At 6.4 amps, initial turn-on was a bit higher than the 30sec snapshot, and continued to drop slowly after 30sec. With the current test set-up, I'm not sure how much more power could be supplied without running into issues such as the bond wires melting. (Yes, active cooling such as dry ice or TEC modules would help :) ).

* Both the emitter and copper heatsink were pre-tinned then sanded before reflow to ensure a very thin solder layer.

* Despite some places now offering copper pcb's, I'm not sure that they would mimic my results. The reason being that even though the pcb is copper, there is still a very thin layer of high dielectric bonding ply under the emitter that limits optimum heat transfer.


Nice work Match. Really pretty remarkable results.

Do you have a chromaticity meter or anything? I would have to assume that at those currents you are introducing some pretty large color/tint shifts into the LED. Also, you are definitely shortening the lifespan of the LED.

With that said, those are some damn impressive results.

WOW WOW WOW.... 1500lumens... never ever expected this! And that hasn't even reached the negative return point!

Thanks Match. Wonderful effort. Now modders has their basis to squeeze more lumens out of XM-L.

Definitely impressive results.

Taken from CREE's XM-L datasheet:

"The maximum forward current is determined by the thermal resistance between the LED junction and ambient. It is
crucial for the end product to be designed in a manner that minimizes the thermal resistance from the solder point to
ambient in order to optimize lamp life and optical characteristics."

According to CREE it seems that cool whites are those that are going to shift more with more current.Understandable if you look a the low CRI specifications of cool white emitters.

This data from CREE seems mostly ignored. I realize that most do not look for Energy Star, but still.

"Cree currently recommends a maximum drive current of 2000 mA for XLamp XM-L white in designs seeking the ENERGY
STAR* 35,000-hour lifetime rating (≥ 94.1% luminous flux @ 6000 hours) or 25,000-hour lifetime rating (≥ 91.8%
luminous flux @ 6000 hours)."

Great work Match. I am shocked by the result

Awesome. If you mind me asking... how did you go about bonding the LED to the copper? and is it a viable alternative to people with only "normal" equipment?

And that was a T6 bin.... I imagine you would see up to another 5-10% increase with a true U2 binned LED?

Now if only someone would start mass producing these with U2's at $20 a poop. 8)

Re: The above comments concerning lifespan, is lifespan dependent on the temperature the LED reaches or the current that flows though it? Surely if the LED is maintaining the same temp. at a higher current the lifespan will be very close to its original rated value?

Copper is not cheap.

Lifespan is a bit misinterpreted, actually the LED looses brightness with high currents and high temperature. But even the high currents are depended on, as CREE says: "The maximum forward current is determined by the thermal resistance between the LED junction and ambient"

There are no lifespan ratings at 3A as far as I know.

Sorry to match if I'm jumping in on the question and shouldn't be - but I can answer this. The LED was 'bonded' to the copper with solder. Nothing more exciting than that - He soldered the center heat slug to the copper slug. Not at all difficult, in fact - heat it up and cool it down.

A U2 could be expected to be a few percent more efficient, yes.

The temperature is the primary factor in LED lifespan, but high current levels carry their own degrading factors. I wouldn't expect an LED used like this to die in a day or a week, but I also wouldn't expect it to last 50,000 hours. The high current is going to cause some degradation of the bond wires. How much is anyone's guess, since I don't think any real studies have been done on using an LED in this manor.


Very well done test Match! Results are even better than I expected, thank you for the time and effort you put into this!

Wait....does this mean that...naaaw, can it be??

"You're gonna need a bigger boat."


Great job, I am wondering how it will perform now with a big aspheric lens

Cheers for that..Sorry, I only saw his thread a couple of minutes after posting. I was once again looking in the wrong section.

The awesome thing about this mod is the potential for both extreme brightness and potentially lower draw currents. This mod essentially makes a T6 into a U3 or thereabouts at higher currents.

From a quick calculation the maximum difference in lumen between the T6 and U2 (assuming the high current does not change the margin between the U2 and T6) is about 12.5% (1/8th) (280 vs 320 lumen) with the minimum being ~0% (300 lumen).

As for lifespan... I would be happy with more then 2000 hours at 4A (which is still almost 3 hours a day for two years strait). By the time it burns out you could probably buy something probably very close to (if not over) 200 lumen per watt.

Time to use some dry ice bro....

Really interesting, I'm doing the tests, I'm getting good results with copper, when I have time I will put. With an aspheric be a joy, epro first touches stabilize xm-l in any project, and movement after enrredar with aspherical.

WOW. I'm guessing, by that curve, that 1600, or a bit more, is the peak, but that w/ anything less than your huge heatsink, 6A would be about as far as I'd go for turbo.

This makes me very excited to finish my HF109 w/ aspheric lens

@cheaplite: He drilled a hole through the copper (at least on his Maglite mod) and soldered to the tiny silver squares right beside the dome. Might not be pretty enough for pictures here. :D

Great testing and nice giveaway!

I truly understand how terrible hard it can be to get this to work - at least for me. Having destroyed 1 XML and reluctant to destroy another one since it is the last one I have on this boat I can only bow in respect :-)

May I be a bit cheeky and ask how many tries before you succeded and did you wreck any emitters in the process?

I have another question also.

Lately there are more and more retailers that offer XML bonded directly to copper. and intl-outdoor also added this a few weeks ago. I even noticed there was one on DX at a certain time. Common for all of these is that the "star" is either 14 or 16 mm copper slugs of approx 1 mm thickness.

Now my question: If I were to aquire one of these and wanted to build it onto, IE not reflowing the emitter onto something else like you have done, a bigger slug what would - in your learned opinion - give me the best thermal conductivity to the base of the new slug:

a) Press fit the emitter & star into a copper plate of similar thickness.

b) Press fit the emitter & star into a copper plate of similar thickness & then solder the joint.

c) Press fit the emitter & star into a copper plate of similar thickness and add a bottom disc of copper of same size with thermal compound.

d) Press fit the emitter & star into a copper plate of similar thickness and add a bottom disc of copper of same size with solder.

If solder is to be used would it not be best to use a solder with very high silver content?. Really sorry to ask all these questions and I fully understand if you are hard pressed to answer. But with these remarkable results then... I cant help myself.

BTW. Considering direct mounting a XR-E. Those should be easy with their BIIIG solder pads on top :-) Even I should be able to do it.

So, the bottleneck really is the Star => body, not necessarily the Emitter => Star?