(Updated July 16) Review/Test: Illumination Supply Direct-Bonded 16mm Copper Star with Cree XM-L2 U2 1C

Cool thread. Thanks for the testing, Keep up the good work. You are saving many of us money with this data.

Where do you see the 0.7V?

I would like to see a comparison before and after dedoming

watching closely.

The 16mm boards have the lowest resistances because, well the traces are shorter.

Insanely good data, good job relic!

That makes sense.

Relic … could you test the resistance on a 16mm Noctigon?

Funny you should say that. I have been asked by Hank from Intl-Outdoor to do just that. Since I do not have any 16mm XM Noctigons, he is sending me samples.

Hikelite, you are correct, 16mm boards do have shorter traces. However, good trace design can greatly reduce the resistance. Take the two 16mm samples I provided above. The Sinkpad has a thinner trace than IS-DCB and it shows in the test results. The same can be said about the 20mm boards.
As I noted with an edit in the original comment, it is not entirely fair to compare a 20mm board to a 16mm board.

I have completed the testing, and added several graphs and comparison information. I am awaiting feedback from Calvin on the potential reflow issue (i.e. could it be a single unit or a larger issue). I will update the OP when we know more.

WOW insane information, thanks for the effort relic!

Thanks again relic38. I doubt any off us are doing any better on the reflows so consider your data ok. It would be interesting to see these tests conducted on a flashlight with no external cooling and comparing the different stars on different pills in the same light. We need a mule light. It appears that when you get to around 2.8 amps plus that the copper star would be worthwhile. You could possibly get away running an led on a copper star at 2.5 amps with the same output as an aluminium star and led at 3 amps.

Thanks MRsDNF, I think you are right about the aluminium star when driving the emitter to spec. The gains are significant, but not huge. When you want to drive well over spec, the direct bonded stars do their job.

One thing to note is how lapping does nothing, or does it worse. Theoretically it should do nothing since those layers are so thing that for one's mind is hard to even comprehend such thinness, like 0.002mm and such.


1A -Stock-440lm
2A -Stock-796lm
3A -Stock-1080lm
4A -Stock-1304lm

1A -Lapped-431lm
2A -Lapped-781lm
3A -Lapped-1062lm
4A -lapped-1287lm

I would not say lapping makes it worse. I estimate my measurement tolerance to be at least 5% from test to test, maybe more. These numbers are well within that tolerance.
In theory, lapping away a gap of 0.1mm down to 0.02mm does have an effect, even if it is too small to measure.
I recommend lapping to ensure the surface of the pill and star are flat. Any unseen bumps or machining ridges could hold things apart more than necessary.

Tolerance may be there, the results do show lower numbers, testing such effect several times would give an better idea. Now I am not saying all lapping is bad and one should use surfaces with off road terrain on them. Sinkpad has a big area that does not make full contact and still shows how it handles things very well.

I really like you putting a lot of effort into these tests, but you need some 3.5digit multimeters at least (4000-5000 count at least). I for one will not go trusting too much for that last digit on X.XX only readings.

I do not have time for running multiple tests on each configuration. I barely have time for what I’m doing.

Re: 3.5 digit meters, you are referring to the Vf readings? I will not be going out and buying new meters to add a millivolt digit. I use the same meters for the same test each time, so things should at least compare to each other.
The tests I (and probably most of us) do are not controlled enough to be scientifically compared in an absolute fashion. I just try to control what I can (set things up as close as possible to the same each time) and go from there. The variation in each setup combined with equipment tolerances is where I estimate that 5% (or more) from.

For the 3.5 digit meters, not to add a milivolt to the data results, to make that second milivolt digit more reliable a result. As you know there is not only an accuracy tolerance but also a last digit tolerance on multimeters. Just wanted to clear that up.

great job on the testing Relic, that must have been a lot of work. Looks like the DCB stars are more or less equal (at sane currents at least :laughing:, which is great for the modding community. Interesting results with the lapped star - could be that the massive actively cooled heatsink provides such a constant temperature difference that small differences in thermal resistance between star and heatsink aren’t that important. I wonder what effect it would have under more thermally challenging conditions?

Thanks for your perseverence, going to the bottom with your tests and find out what really is happening, that is a lot of work!

Thanks everyone. I still have two more threads to start for the other stars (I think it’s fair that each product gets it’s own thread with a comparison to each). I’ll be doing an update or two here as well when I have more answers to some of the questions and concerns.

btw I just took the U2 1C that I got from IS and moved the emitter from the 16mm board to a Noctigon 20mm, and found the same issue with the solder paste not covering the pad at all.