Results: Direct bonded XM-L emitter to Copper

Not surprised at all.

Some sellers have been telling us that all the time ...


Haha :bigsmile: I got about half way through the thread and was going to post about the Chinese sellers knowing this all along with their 1600 lumen ratings... but you beat me to it!

But really, how funny is that? The famous "1600 lumen" advertisement is technically spot on. All they need to do is insert an asterisk that explains this can be achieved through simply adding copper

So you're selling these for how much? :)

These numbers are not possible even for a U2 and zero heat resistance.....

Care to explain how you've come to that conclusion? Random thought droppings on a topic like this are just not going to do it...


he drilled 2 holes thru the copper and ran wire thru the holes and soldered the wire to the led?

Maybe he just soldered wires to the two solder pads on top of emitter?

these ones.

Show us your numbers, please. If you've done calculations you should show them, I think we'd appreciate your input. I don't know if it's possible either but until proven otherwise I have to believe these kinds of gains aren't strictly impossible. The removal of electrical contact pads on the underside probably has an effect too (in addition to the absence of an MCPCB).

WOW, awesome results Match, thanks for doing this and presenting the results, great work.

I think I would be a happier man if I had faith instead of doing the mat(c)h .

That's an interesting viewpoint that removing the two electrical metal pads will give better cooling?. I am aware that the MCPCB adds 0.2 degr. C/W to the thermal resistance. That is 2 degr. C for the 10W dissipated at 3 Amp and has only a minor influence on LED temperature.

Well to make a long history short, I'm using a spreadsheet that includes formulas fitted to Cree's data sheet (that is what I have faith in).

Let's use a current of 3 Amp that is inside the specs so all can be verified with the data sheet.

If I solve for thermal resistance from LED chip bottom to ambient to match Match's FIRST measurements on XM-L (giving 881lm @ 3A) I find 2.7 degr. C/W (using 290 lm @ 700mA for T6). That number is very plausible as I found a L2 copy to have 10-15 degr.C/W (if I use this number, the spreadsheet predicts nicely what I can measure from my L2-alikes).

For Match's recent measurement, 1000 lm @ 3 A, I have to put in an un-realisable thermal resistance of minus 6.6 and that is why I am sceptical.

Try it yourself with the spreadsheet. Comments to the spreadsheet are welcome.

It's even easier . . the tech info for my TF light suggests that 1600 only applies to the T6 for a fraction of a second before heat sag on strobe. They aren't claiming it is a sustained rating . . in fact they lower sustained on high to 1000 (if I remember correctly).

Tried the spreadsheet, and I think I still believe match's numbers... If we assume that it's a really 'good' LED that he happend to test (still well within the realm of possibility though).. then I can pretty easily get the spreadsheet to match the test results.

Drive data: case 1 case 2
Led current [mA] 3000 4000
PWM [%] 100.0 100.0
Body data:
heat resistance ['C/W] 0.5 0.5
OTF factor 1 1
LED data:
bin [lm@700mA-25'C] 310 310
Vf-coeff [V/'C] -0.003 -0.003
heat resistance ['C/W] 2.5 2.5
Lum efficiency [%] 30 30
avg Current [mA] 3000 4000
Vf @ 25'C [V] 3.354 3.467
LED effect [W] 9.89 13.56
LED temp ['C] 46 53
heat derate factor 0.962 0.946
Light Output:
LED cold [lm] 1011 1214
LED varm [lm] 973 1149

YES copper MATTERS!!! I have thermal designs at work where the use of copper as a thermal *conductor* SAVED the design.... we could not pass thermal testing without the use of it in place of Aluminum, which is normally our first choice.

Conversely brass, stainless steel and Titanium are the worst thermal conductive metals commonly used in this hobby.

thanks for running this test!!

You're better at this than I am, that's for sure. Every time I get into fooling with bare emitters and reflowing, I kill them 99.9% of the time. Finally got sane (for me) and quit on it.



Put that copper into the fridge for a while. Would be interesting to see how it works ~ -10ºC on copper 8)

Amazing, just amazing.

Thanks, Match

That's a useful tool-thanks. I assembled a quick one last week (just for know) on Google Docs. It's not as complete as yours but I think it incorporates the requisite data to show how various values for heat transfer can affect relative performance. I based it on Cree's documentation ('Optimizing PCB Thermal Performance'). I went by experimental data found over on CPF to make assumptions (I can't assign best-fit polynomials to data points in Google Docs ). A couple of Saabluster's (link) and Ma_sha1'a threads revolve around similar experimenting. Likewise, Match's results definitively show that a good thermal interface makes a difference. I'd definitely be wary of equating people's raw data with Cree's datasheets because it's difficult/futile to interpret causes for discrepancies like the one you noted. That said, I still wonder what effect the reduced thickness of the thermal/electrical pads has on thermal resistance.

Thanks for the response folks! I've updated the origional post to help clarify a few common questions, but let me address some of the specific ones here:

*Bose301s asked about chromatic color shift and led lifespan - As the emitter was inside the I.S. during the whole test, I wasn't able to observe any color shift directly. Although in use outside the i.s., I haven't noticed any substantial shift. Most color shifts (emitter going blue) that I've noticed stem from high emitter temp, which usually is a result of high current unless heatsinking has been addressed such as it is here. As far as LED lifespan, I can't say how much it would degrade.

*Manual Man asked about the bonding method - There is now a link to the origional post I made with the details on how I did this. PilotPTK also gave a nice summery above.

*Dthrckt asked about peak lumens and turbo mode - I feel that my data above is close to max for passive cooling. With active cooling however, (dry ice, peltier...) I think it would much higher than 1600lm...

*Ledsmoke askes about my reflow success rate and copper pcb's now available - Having done about 6 of these now, I wish I could say it gets easier...but it can still be a pain at times. So far I've only lost 1 emitter during the process. As for the new copper pcb's, without testing I can only speculate. I'm sure they would be better than some of the current ones, but most likely not as good as a solid direct bond (See my thoughts in the update in the first post). As for you question on mounting a star to a copper slug, your best bet would be simply hand lapping the two surfaces along with a good thermal compound and tight pressure holding them together. I wouldn't try to solder the star to the new copper backing.

*Sixty545 expresses his sketicism - As a skeptic myself, I can appreciate your concerns. But a flat out statement of denial can't be taken seriously. Since there is no peer reviewed journal to publish my results in as there exists in the scientific community, I can only publish my data along with the methods from which it was derrived here. I think it would be fantastic for someone to independantly verify or disprove my results. Please keep us informed on your own testing, along with the detailed method and the data collected.

Thanks again all for your comments and interest in this.


I've had great success lapping aluminum pcb to aluminum pill and aluminum pcb to copper pill.

Unfortunately, I lack the equipment to test output. My sole criteria for judging the results is based on soldering leads to the pcb.

w/ unlapped pcb its easy. w/ aluminum lapped to aluminum it is significantly more difficult. with aluminum lapped to a decent size (1" of 3/4" dia.) copper rod, it is VERY difficult, ie impossible w/ my detail iron, and not easy w/ my 240W soldering gun (one must be careful not to 'unflow' emitter).

As far as soldering a copper pcb to copper - I think it would work well, but I'd feel uncertain if I'd used too much solder, or not enough, and if it was even, etc. If you have the copper heatsink, you might as well solder right to that. Grinding the pads off isn't hard - finding the right jig to hold the emitter while soldering leads was the trick for me.

Here's some data on solder vs exposy vs mechanical heatsinking (though obviously not entirely relevant to flashlights...)

@ Match: OK thank you.

@dthrckt: Having already done it i just know it is not something I want to do a whole lot of times. It sucks when soldering to the tiny pads. No matter what. I destroyed on XML and then went on to succeed with a XML.

But it still sucks.

As for judging how much solder: First you asses the tightness of the fit between parts. Then you assess how big an area it is totally. Then you take total area in mm3 * 0,3mm solder if using thin 0,5mm solder. Then you divide by 10. Then you use roughly half of that and then you only used a little more than necessary :-)

Brass is superior in thermal conductivity to Stainless Steel or Titanium.

Brass: 109-125 W/m.K

Stainless Steel: 16 -19 W/m.K

Titanium: 7-21W/m.K