Well, just wanted to say this comes foisted by the lack of proper boost (and/or boost/buck) drivers; hope we soon have some of them. 
Cheers ^:)
My mod requires access to the positive and negative contacts of each die. With the XHP70 (and 50) the positive bond wires are visible for each die. I’ve looked at pictures of the XHP35 and I don’t understand how it is layed out. There are only 2 bond wires visible, one on one die and one on another…
Theoretically I could do the mod on the XHP50, but I probably won’t. I do plan on making some 2x26350 and dedomed XHP50 flashlights.
Missed this thread as well, great surgery Easyb! And the option for high drain 26650 cells is enough reason to want those dies in parallel.
I’m tempted to copy this mod! 
Thanks. I would enjoy seeing your version of the mod.
Regarding 6V with 2x26350 vs 3V with 1x26650: the available 26350 cells have more than double the IR of the good 26650 cells, but because half the current is needed for the same output, you will get comparable performance because the greater circuit resistance multiplied by the reduced current gets you a similar voltage drop.
Djozz, I don’t know if you saw my thread about calculating current in direct drive setups. Along with your forward voltage measurements one can predict the current flow in these different setups.
For example, with 2x26350 cells with 0.072 ohms IR each: about 6.5A will flow through an XHP50 with fully charged cells.
About 8A should flow through an XHP70 with the 2x26350, so this is actually higher performance than my 3V XHP70, which flows 13.7A.
Another update on this mod: the non-optimal solder joint between the enamel coated wire bridge and the package surface failed. I noticed the light slightly flickering when on 100% and the next highest mode. At first I thought it was a driver issue because it would stop flickering in about 5 seconds and all four dies were lit fine when in moonlight mode. But I happened to look at the LED through solar viewing glasses and saw one of the dies was flickering. I probed the solder joint and it was loose. It might have happened when my cat knocked the light off my desk onto a wood floor. The wire bridge was epoxied down so the joint to the bond wires was still intact; only the joint to the package surface was loose (in the first picture, the joint at the top right).
I resoldered the joint and it is fixed for the time being. So anyone attempting this mod should pay close attention to this joint.
I had always guessed something like this might be possible, but never got around to trying it. I think I’m going to have to now. The Convoy M1 is known to make a decent host for an XHP50, I believe.
Wow… im impressed. Especially at the skill to solder at such a micro scale to achieve that. Nice work! :+1: Now if only Cree would build factory 3V versions of the XHP series…
I agree that would be a cool setup. Give it a try! Definitely get a hands-free magnification setup. I used a 20x loupe held by one of those “third-hand” soldering helpers.
Nice mod. Any updates?
That solder joint I talked about a few posts up failed again, and I decided to abandon the mod. It was just too fragile.
It was a cool project though! :sunglasses:
This is an old thread, and I stumbled upon it when I started Googling for 3V XHP50.2s when I saw that such thing exists in Cree datasheet.
Look at what I found!
Cutter link
A bare emitter costs $7.5 AUD. That’s really tempting but I wonder how much is the shipping.
Yes, it is very cool that it exists now, and it did not go unnoticed : 3 volt XHP50.2
In the meantime also available at Kaidomain and ASflashlights.
Thanks for pointing that out. Now that I know it, I feel like I shouldn’t dig up such an old thread. 
Digging up old threads is ok at BLF, and this is a nice one, pretty cool led surgery in the OP by EasyB
It’s funny you brought this thread back up because I was just doing some exploratory surgery on a XHP50.2 last night. My thought was that, with no bond wires, both pos and neg connections are on the bottom of the die. Then if I could separate the dies from the package I could essentially reflow the dies on a package of my design, making a 3V XHP70.2 or even a 3x3 array of 2mmx2mm dies.
This is indeed possible in principle. I was able to separate the dies pretty easily by wedging a razor blade under them at room temperature. It looks like the dies are attached with some sort of metal bonding, and it doesn’t surprise me that the XHP50.2 had reliability problems with how easy the dies came up. The majority of the underside is the negative contact, and the positive contact is a thin strip on one side (75 microns wide). I tested and this metal surface is solderable.
So if we made a MCPCB we could make any sort of array we wanted, if the reflow process or bonding process was figured out. A 3V XHP70.2 would be cool. A 3x3 or larger array would make for a super powerful single reflector light, but a multi-reflector light could achieve similar output and beam intensity, so the array development is probably not worth the work.
The original motivation for the 3V XHP70 and then for my work with XD16 arrays was to have a very powerful single reflector, single cell (so compact and light weight) light. But now with the 3V XHP50.2, or good 18350 cells and XHP70.2, I can have my light-weight high output thrower. So I will probably not pursue any more array development.
That’s a nice discovery. I never expected a 3V mod is possible for a gen 2 XHP50. I’ve always wanted a 3V quad die setup but my skill level is not enough for that. Before that I was using an improvised setup in a single reflector light…
I got some of those 4x MCPCBs, but I never actually tried one because I think the dies are just too far apart. It sure will be nice to have those 3V XHP50.2s.
I have put one of these quad 3535s into a Nitecore P16. With the medium sized SMO reflector (orginally designed for an XM-L2), there is a quite large donut hole.