Cree XD16 measurements

Could you explain what you believe is the mechanism causing this phenomenon then?

You’ll find most of the information here:

- Clemence

Interesting results but sadly about what I expected, in fact a bit better I would say. I only expected around 800 lumens max based on the info I had.

Although you also have a DTP mcpcb, which most could not do or use with flashlights. So expect performance to be worse in actual use with a normal mcpcb.

Sadly like many have said, not very interesting for flashlight use. Although I am surprised that it still suffers the tint shift issues, wasn’t that one of the big marketing points that it fixed the tint shift issues?

A single XD16 is not interesting, but a 2x2 or 3x3 array can have respectable performance (good output and no donut hole) with the possibility of being 3V. If my speculation is correct about the output going up by 15% upon making into an array then it can compete and even beat the XHP70.2 or XHP50.2. Of course the lack of proper MCPCBs would still limit the performance.

Are you going to run your set up in series or parallel?

If in series this would be similiar to the XHP35.

I’m planning a 3x3 array all in parallel. Should have a 4.8mmx4.8mm light emitting surface.

A progress shot of my 3x3 XD16 parallel MCPCB.

Interesting looking, and not simple lol.

I would say “simple, but not easy” :wink: :innocent:

Interested in seeing the results! Thank you for your contributions!!

Yes, simple but time consuming. :sunglasses:

I just placed the dies on to see how it will look. The copper is about 20mm in diameter. When I light it up with my DMM the contact is intermittent and only some dies light up because they are just resting there, but I can see there is good “light sharing” going on. The dies surrounding a lit die light up evenly just from the light that was escaping out the side when it was alone. So I’m hopeful about a nice boost in efficiency/output when these are in array form.

When Clemence tested this with the E21 they burned the edges of the LED’s, so at higher currents it could work in reverse.

Although apparently if you put some clear silicone over the entire surface and make it thin enough to go between the dies, it will fix the problem? At least I think that was the idea.

Interested to see how this turns out.

Whatever happens, I salute you for the effort! :beer:

Cool! I know Clemence’s prediction but nothing beats an actual experiment! :slight_smile:

Good job with the file! I know it’s very time consuming but let’s do it.
If you can bond those sides using any kind of optically clear silicone then the photon cross talks should disappear.
EDIT: with optical bridge most of the photons will escape to the adjacent LED without being bounced back to heat the die it exited before. I had the same thought but not many people would want to carefully glue each module by themselves. I decided not to continue with the gapless design for some reasons:

- Gapless design as is (with no post processing) limits the max. output (limited max current)

- Cleaning the flux residue requires special method which would not be easy task for most people, let alone mass production.

- Designed for Nichia E21A. Gapless vs gapped design beam profile don’t differ much. Donuts with either SMO/plain TIR optics.

  • Boost drivers are getting widely available

With XD16 higher output capability, you might be able to overcome what E17A/E21A can’t do previously.

- Clemence

Thanks guys.

I’m not too worried about excessive heating of the phosphor due to “photon crosstalk”. The E21A has exceptionally bad cooling of the phosphor, especially at the package edges. The edges of the E21A burned first even when it was operated alone, so it’s not surprising that a neighboring die would exacerbate the burning. I wouldn’t necessarily expect the same results with the XD16.

The “light sharing” that I observed with the XD16 in array form is a different phenomenon than the E21A “photon crosstalk”. The XD16 flipchip design has the die exposed at the edges so the unconverted blue light escapes from the sides of the clear sapphire/SiC. When in array form I suppose this light enters the neighboring dies through their respective exposed sapphire/SiC edges. This results in the entire neighboring die being illuminated practically evenly, not just the phosphor near the edge. The resulting light collection/conversion efficiency looks to be much improved compared to the lone XD16 I tested. I assume this mechanism is part of cree’s design.

I am, however, concerned with flux residue seeping up the cracks and causing burns on the surface. Any special tips for cleaning the flux residue?

Interesting that there are such high deviations in light flux compared to stated bin. Thanks for the test. Maybe I can get my hands in the next time on some samples, too.
With the XP-L2 and XHP70.2 (links to my tests in sig) I experienced the same phenomena in lower flux than expected, and also with the Lumileds Luxeon V (from two different suppliers, but not so extreme than of the Cree emitter).

In summary, the smaller the LED the worse the heat dissipation is. The trend to design smaller and smaller emitter is a big problem for high-power junkies like us…

Any update on the array testing?

The board is done, and hopefully I will reflow the LEDs and do some preliminary testing tonight.

Well, I ran into an unexpected snag. I had used arctic silver thermal adhesive to fill in the gaps around the fingers. I used JB weld to adhere the two pieces together, but I wanted to maximize the heat spreading around the fingers so I used a thermal adhesive. The arctic silver was not electrically conductive after it cured, but after I heated up the board for a few minutes to tin the fingers (I don’t use solder paste) I discovered there was a short between the two sides. I checked for any stray solder or wire connecting the sides and couldn’t find any, so I concluded it must have been silver particles in the arctic silver that aligned just right when it was heated. The resistance between the two sides was around 10 Ohms. I painstakingly scraped the adhesive from between the fingers and the short disappeared. So I will have to clean it up and get some ceramic-based thermal adhesive before I reflow the LEDs.

Here is how the board looked before all this.

That is come impressive work.