6x 20mm XML-ledboard comparison

Interesting results, thanks for the detailed tests.

I think where this is most valuable is in modifying those oddball size stars that Sinkpad and Noctigon don't have, 25mm in the Aleto 26650, 32mm in (I forget what). Maybe even triplet/quad boards?

Huge difference between direct heat path and one with the dielectric layer though

the basic details are below. a comparative guide between normal dielectric and the DLC

email me if you want details

Excellent djozz. Thank you.

nice graphic

I find these are helpful too

thanks

Looks useful. I hope you'll send samples to one of our members for review. Relic38 has done some nice reviews. Hopefully he'll be up for testing this too.

A couple questions though.

How can I tell from your webpage if a mcpcb is using DLC? Will a triple XM-L board be made?

Looks very promising but, as usual in marketing, the 'old' type board is drawn to perform worse than in reality; the heat when entering the aluminium core of the board will go circular in all directions, not just downward, as the drawing suggests.

In any case, I think the major bottleneck was already overcome by the direct-bond-to-copper-boards, I do not expect a performance improvement as high as when Sinkpad boards entered the market. But the electrical insulation between thermal pad and core is a really useful feature!

Exactly!

All Copper boards can be enhanced to distribute heat better (not really needed), like all AL normal boards have the thermal trace as wide as the boards. For some reason on direct thermal path they decided not to them like that.
I do not see them actually work better than direct thermal pad boards which clearly cannot take advantage of the 390W/m.k of the cooper when you can see Aluminum boards rated at 135W/m.K performing as the 390W/m.K copper boards.
So what I mean that wide spread can be done with normal copper direct thermal path, especially for single LEDs on 20mm.

djozz - Excellent info, great job on this! Lots of time and some cost I'm sure...

Question - didn't see it mentioned, what was the time lapse on your measurements? For example, was it pretty much instant once the amps was set to the point of reading the lux meter? Wondering if under-lying these stats is the dimension of time - what's the instant reading vs. 10/30/60 seconds of runtime. Of course it would take long collecting all the data pts, but would be interesting to know if the copper MCPCB's show any advantage over aluminum.

The reality is copper as a base substrate has better thermal performance than ALU. The issue has always been the dielectrics which are generally high in fibreglass are the issue and direct thermal pads make a big difference. This is more an exercise in replacing the dielectric to make ALU boards a better performer relative to copper as ALU is a lower cost product. I have some initial samples of a multi led XML PCB, happy to donate for test?

The costs were mainly my new power supply. I bought it to do these kind of measurements, and it performs really well. With my old one I could not have done these kind of tests, it does not go over 7 amps, and setting a current is a pain. The new one adjusts really easy and goes up to 20A. I prepared some things beforehand (like initial testing out the power supply and test-leads, and removing the dielectric layer on one of the boards). But the actual tests were done in one session lasting an evening and part of the night.

The time lapse was mentioned, but not in the OP (should have been there). I started each test at 0.5A, going up 0,5A each time and recording voltage and lux, up to 7A. After each rise I waited 10 seconds and then recorded V and lux, I think the led was at each current for about 25 seconds total, but the starting point for each new current of course was only 0.5A less. Indeed if I had waited much longer to let it settle more, the whole test-series would taken more time than I was happy to spend on it (my hobby is squeezed in between work and family, I have to admit both suffer from it, but I just need to have a hobby), but looking at the luxmeter during the tests I never (also not at 7 amps) saw fast declining numbers signifying an unstable siuation, (typical example: when a reading at 10 seconds was 702 lux, at 20 seconds it would be 700, and that did not worry me).

Interesting.... Copper has such a small effect on output and thermal sag over aluminum? Hhmm.... Our custom made copper pills have more volume of metal than the stock pills, so the measured differences may be more related to total mass/volume as opposed to the source material, I suppose.

Then I'm a bit stumped. Where is that thermal performance advantage of copper going to? It's there I suppose, but such a small factor only at higher amps/temps.

That sounds great. Either djozz or relic38 would be great. Djozz didn't mention it in his post at the top of this thread, but he also stress tested a direct bond copper mcpcb for the Cree MT-G2. I think it'd be great if you could send samples to both testers if they're up for it, especially of that multi XML board since that may offer something that no one else does.

The graphic is a bit disingenuous, but it's minor compared to the results of independent testing. If the results are good, I'll be posting the test results in this forum and at least one other, not that misleading graphic.

I would think a thermal imaging of the emitters on the stars, with and without dielectric aluminum vs copper would be a good comparison

Heck I would like to see some of the homebrew budget mods on test as well (the ones where they drill out the aluminum under the star and install a copper slug from a wire)

It's a head scratcher, especially that dielectric that performed surprisingly well. As far as the metal of the pill and flashlight, I think it comes down to emissivity. I bet the best combination is something like this.

Use copper to quickly absorb lots of heat and conduct it to a layer of aluminum with lots of surface area. Maybe the current set up is good enough not to suffer significant sag, but this concept should increase comfort. Where lights typically only get warm around the pill, a longer copper sleeve would spread the heat more evenly throughout the body, which should decrease the heat felt at any particular area since there's more surface area of aluminum to dissipate the heat. Increasing contact area is already something lots of you guys do with P60's...and I'd do it too if I had any of those.

Don’t underestimate the heat SINKING properties of aluminium compared to copper.
You can put much more heat in a given mass of alu than in the same mass of copper, before it actually gets hotter / warmer.

I have tried to explain that one before, but there is the widespread misconception that more mass means more heatsinking. If you compare copper and aluminium by volume (that is what you do in practice when you swap aluminium parts for copper parts) the copper part does win for storing heat, but only by a factor 1.4 compared to aluminium (while it is 3.3 times as heavy).

I didn’t know copper is better per volume as a sink !?
Thanks.

1.4 correlates well enough with this data.

I don’t think we should emphasize mass too much, volume is the factor is most cases, unless you’re flying it to space or something. So copper still have the edge hands down in most applications.