I don’t like that they seem to have the two-pad footprint so the usual DTP-board mounting does not work.

Hey guys. I’m not as knowledgeable as a lot of the others on this forum, but AFAIK, if you make an array of these in an odd-numbered configuration, only the one emitter in the center will contribute to throw. In other words, whether you have 1, 9, 25, 81, etc. of these, you will have the same amount of throw. As far as the LED is concerned, throw only depends on the brightness at the center of the focal point. There’s only one focal point, so adding more LED’s won’t change the amount of throw.

lol, yeah I totally messed that up. I was thinking 9 total dies when doing the math and it must have slipped into the post.

If they are rated for 600 lumens then that is another story. The only thing I saw was something like 200 lumens each die.

Like djozz said, without the thermal pad I am worried that they will not see the overdriven benefits we see in other LED’s. Much like the E21A or any LED that is not used with a DTP MCPCB.

I would love to be proven wrong though, a 3x3 with 6000 luemns is nice but would be basically the same as the XHP70.2 in a performance aspect. The key difference would be the ability to run it at 3V, which would be nice.

Although with boost driver tech improving this is less important then it used to be.

Overall I would say I am firmly neutral on these at this point. Simply not enough hard data to make any conclusions. It could go either way equally IMHO.

This type of problem really drives me crazy. There must be a simple solution to this.

I’m thinking for a 3V/parallel application there’s a fairly simple fix to get DTP performance. Series array applications would get complex/practically impossible. But if we used Aluminum “DTP” boards where one of the pads was “DTP” and then the base was anodized we may see decent results. You could then use a typical copper overpour as a heat spreader for the other pad.

:FACEPALM: wait… this makes me wonder. How do they make the Al to solder junction on regular Al DTP boards??

You could make a DTP pcb for one of the pads on these LED’s with some driver tweaks to apply either the ground or V+ through the flashlight body. They would need custom drivers though and as we learned with the E21A, only making one of the pads DTP doesn’t help all that much.

Even with both pads DTP with the E21A clemence found very little to be gained by over driving them. I see no reason to expect much difference with the cree version.

Was this data/experimentation posted somewhere? I didn’t see this. That’s a bummer then. I agree that the two platforms are likely going to share the same issues. Here’s for hoping they go up to at least an XB footprint then. XD25 anyone? O:)

The other thing that has me wondering is they claim to have improved the alignment properties while soldering so could they have done something wild with the solder pad geometry to cause this?

I found the thread you are referring to! It’s funny to see that clemence had the same thought - 4 months earlier haha :wink: I’ll shut up now.

Also, I see now that Nichia used an interesting solder pad geometry themselves…

Found this data today… By the looks of it we should see tint shift be nearly eliminated and surface brightness that rivals or beats all previous CREE products. That last part is highly dependent on how underdriven the specs are relative to previous generations of LEDs. The trend seems to be that they are getting more aggressive with max ratings, leaving less room for improvement when BLF-ing them.

It also says: neutral thermal pad: no.
So I assume a two-pad footprint.
Makes BLFing much less succesful too.

Yes. This is true. But even running them within the factory specs makes for interesting possibilities. I’d be shocked if you couldn’t get another 33-50% out of them too. Historically we can get 66-100% additional output from overdriving Cree products. I’m not saying they’re going to blow everything out of the water, but where I think they will make the greatest impact is with system integration. Optical efficiency and flexibility here is what excites me the most. Not record breaking lumens or throw, though they may get close. I’m keeping an open mind here. I think about a 5x5 array that’s capable of a *rated*16000 lumens in a footprint similar to an MT-G2 and surface brightness on the heals of and XP-G2 at any voltage configuration you want and utilizing a single optic (aka 100% aperture utilization vs a triple or quad which run ~65%) without tint major tint shift… Sounds pretty nice. Even running within specs. I’ll take the efficiency and add LEDs and get a broader hotspot (albeit with some heavy MT-G2 style artifacts)

Just because the individual dies have a higher surface brightness the total package has a higher surface brightness compared to the xhp50.2.

BTW should make an interesting mcpcb design to put the LEDs so close to each other.

I think these would be better in multi-emitter optics than as arrays. Masters of neither throw nor output, but a good combination of both.
I’m thinking about something like that:

These would hopefully enable better throw, much better cd/W than XP-G2 while having only slightly lower output. And hopefully slightly better lm/W too, though probably not.
It would throw slightly less than a quad array with a single optic of the same diameter, but be smaller and (am I right here?) put more lm in the hotspot.

I wouldn’t recommend a 2x2 array. Only odd numbers so that you can have one die in the center perfectly focused. A 3x3 would still be smaller than the XM/50.2 footprint and be factory rated at 6000 lumens 4860 lumens.

EDIT 11/28/17: semi-official max lumen specs found. Previously stated numbers where a calculated guess from other claims made previously.

Also, it should be noted that the XHP50.2 maxes out at about the same lumens when maximally driven.

Each die is rated at 650 lumens? Wow, that is impressive. Although they always seem to put space between the dies, would it still be smaller then an xhp50 die?

If so then these new dies are a lot more interesting then I first thought. a 4x4 then should be about the same as an XHP70 except with the ability to make 10k lumens with a 3V LED? Now that is an interesting prospect.

Although I worry that they will suffer the same issues as the E21 from nichia. Where they just can’t make much output in the real world due to the lack of thermal pad.

“The first product available in the new family of XD LEDs is the XLamp® XD16 LED that delivers a breakthrough lumen density of up to 264 lm per square-millimeter”

Quote from their press release months back.

1.6*1.6*264 = 675

I don’t know if that’s going to hold true or not, but that’s what they claimed to the world. That leaked data sheet that I linked though shows max power at 4W. Which would put max output closer to 500lm based on typical efficacy at rated max current for XPG3

How about immersion cooling then? :slight_smile:

A thought:
3x3 array, but wired to 2 separately driven groups.
Group 1 is only the central emitter.
Group 2 is the rest.

You can drive only group 1 to have a thrower or drive both for a flooder.
Would it be possible to make a good PCB for such light? If no, would cutting corners and making 4+1 enable such thing?

I think there’s a slim chance to get a white CSP LED without phosphor blanketing the sides of the die. Without the phosphor the blue light would leak without first converted to white light. I’m curious about this NX footprint.
Cree has different approach here. The NX phosphors cover the entire sides, while E21A blocking and
reflecting the side spill using white reflective material. This perhaps explain why NX produce more output. E21A has the narrowest beam among current CSPs.

- Clemence

No new info, but apparently Arrow is getting ready to sell; the XD16’s.

The price is attractive. XD16? Does that mean 16 dies?

I think it was 1.6x1.6mm die/led size