Crash-testing a XM-L2 and a XP-G2 on copper Sinkpads

Iā€™m going to use my go to driver for the first attempt. The Manafont ā€œ3T6ā€ driver. People have been resister modā€™ing those for multi-emitter lights with good success.

Hmm, 7 amps is very close to the current that blowed the led. Mind that in the test the time at which the led was at that high current was perhaps 2 minutes, no one knows if the led can do the trick continuously without burning through. The temperature in a actual flashlight will be higher too. And I guess there is some variation between individual leds too.

All in all, well, perhaps just give it a go and tell the world

djozz, I managed to repair a single broken bond wire on an XM-L with Liquid Solder from Radio Shack.

Perhaps a very narrow strip of wax paper could be laid on top of the gap between the plates, the liquid solder placed to contact the remaining tabs, then the wax paper removed after it cures? Iā€™ve used aluminum for separation when soldering and it works just fine as the solder doesnā€™t stick to it.

Donā€™t recall it being said but the bond wires are brass.

Nice work with the charts, appreciate it! :wink:

Just a thoughtā€¦

Wow, that sounds like a delicate job well done. I am still curious if the xm-l2 can be driven a bit harder than 8/9 amps with extra thick bond wires, this may be a way to do this :-) .

I am told that the bond wires were gold, but that may be a myth among flashoholics...

But I would just about swear I read in Cree documentation that the bond wires are brass.

I used a fine seamstressā€™s pin (German made, thinner stronger steel) to lay the smallest possible bead of the black liquid solder in place. When I took a macro image of the work it looked like something a giant splashed on it by stomping in a mud puddle. lol


In this instance, 2 of 3 bond wires were pulled when de-doming. The wires were still there, I just had to move them back into position and bond them to their counterparts. Pretty easy, considering and compared to what you were facing with the wires blown off completely. But it might be quite possible to use a thin wire as a replacement, if placed into the liquid solder such that a small drop is on each end, perhaps the wire could be placed in position and the liquid solder would bond it there effectively re-wiring the damage. ??

Perhaps the bond wires are gold plated brass? They seem stiff and brittle, much like brass would, but look like gold. Seems funny, looking at their tiny size, that everyone is stuffing larger wires in lights to conduct the power they crave when it all has to come through these teeny tiny wires in the end! :slight_smile:

How many Amps does it take to vaporize Phosphor? :stuck_out_tongue:

My bet is, thereā€™s some folks here that aim to find out! lol

Could a thin piece of brass sheet, or copper sheet, be curved into a ā€œhalf pipeā€ then bonded in place over the connection points on either side of the gap using the liquid solder? Hmmmmā€¦ā€¦.

Ok, this really got me brainstormingā€¦.why couldnā€™t a thin strip of the very dome thatā€™s been removed be sliced with an XActo knife to fit over the gap and cover the emitter side exposed plate so the liquid solder wonā€™t short the 2 sides, then use the liquid solder in a solid sheet right over the top of the silicone strip joining the connecting plate to the die? The liquid has a tendency to pool, a needle point can be used to drag ā€œtracesā€ where they need to go. The material dries to a hard but not unbreakable solid, they call it a glue but it doesnā€™t have any real elasticity to it and will flake/break if stressed or pried on. I think itā€™s infused with graphite. It does stick fairly well, but not like an epoxy.

This could be used to re-enforce existing wires so they wouldnā€™t blow under higher amperage. Or so it seems to me.

Edit: For the record, the gap doesnā€™t need bridging as is evidenced on this XM-L, but the layout is different on the XM-L2 and I assume the XP-G2 as well. If the 2 sides cannot be connected, then this bridging method may work to provide an answer.

You know what? I was just thinking of the ' half pipe' and then I saw you writing it down :-) . It will be difficult still...

I guess there is a maximum current /die surface area, since the xm-l2 has a surface area that is about twice that of the xp-g2, I expect a maximum output at about 12 amps when the bond wires are reinforced so that they do not limit the current.

Itā€™s a very thin liquid that sets up pretty fast. Easy to get it where it doesnā€™t belong, Iā€™d guess. I have an XM-L2 I was thinking of de-doming, will see how it would work to re-enforce the bond wires on this bare sample.

Iā€™m thinking that the dome top can be sliced off, then a secondary layer sliced to approximate the thickness of the die board and this layer used to cut the strip for the bridge. Then a light scoring of the remaining layer will facilitate the gasoline removal method for a clean de-dome in as short a time as possible.

Vinh Nguyens bare emitter heating then slicing method might be preferable here to soften the dome a bit for an easier cut.

Hey Ferb, I know what weā€™re going to do todayā€¦

(yes, I have a child that watches too much television)

Edit: Wonder if the thin copper sheet could be curve matched to the bond wires, then slid under them so as to allow a visual on the wires laying on top of their re-enforcement, then the liquid solder could fairly easily be placed atop the thin sheet of metal bonding the wires to it. Minimally invasive and least possible shadow thrown when the light is on.

I wonder if beefing up the bond wires would demonstrably lower Vfā€¦something for Tom E to try for his monster C8 builds :wink:

very nice thread, ty djozz

Thank you very much djozz for posting all these. XM-L2s and XP-G2s are the LEDs that Iā€™m currently very interested in and the information in this thread is just awesome. :slight_smile:

I just put an XP-G2 R5 2B on a 16mm Noctigon, with a huge aluminum heat sink and wired up to a Q-Lite 3.04A driver. With a single AW IMR18350 it will pull 2.99A. But stacking chips on the Q-Lite give it nada. No gains. Why is that? With 4 chips stacked and 4V reading on the AW , I got a Vf of 3.43V on the emitter. So why are the stacked chips doing nothing?

Maybe try a different battery. Voltage drop may be too much

I tried an 18650 Efest IMR, also an 18650 Samsung 20R. The chips didnā€™t allow extra amperage over the standard 3.04 board.

Therefore we canā€™t stack additional chips to the 3.04 Amp drivers? Iā€™d better stay away from them then and use the old reliable 2.8 Amp Nanjgs.

Regular solder paste won't jump the gap between the anode and cathode planes on the substrate, even if applied across the gap it will separate when it melts. There's nothing in the gap it can stick to. The connection onto the die is the tricky part to deal with.

I did something similar lately: stacked 4 extra 7135's (380mA) onto a qlite (a very neat solder job I must say, the chips well heatsinked against the pill) and put it in a flashlight with a (dedomed) xp-g2, with a fresh cgr18650ch IMR battery. I expected about 4.5A, but I got only 3.6A tailcap reading (so without the switch resistance). So a bit more than the stock driver but not the full potential of the 12 chips. I did do all the possible resistance reductions, braided springs, thick wires etc. (I got 170klux throw out of the light, so I was happy anyway :-) )

Scroll up to post#55. You'll need a buck driver and 2 or more cells in series to go past what it will do in direct drive.

Ok, my driver wants to deliver 4.6A, to feed that to a XP-G2 it needs to deliver that at 3.8V (my own measurements). The driver itself eats 0.1V (or a little bit more?), so it needs at least 3.9V input. The cgr18650ch at 5A delivers that voltage only in the first minute (HKJ's measurements), so it does not stay in regulation very long . So yes, the desired 4.5A is a bit wishful thinking, but I could see that the first minute unless the switch eats away voltage as well (or in case of tailcap reading the DMM+leads). Pity, you are right comfy, I did not count on the higher voltage of the XP-G2, a XM-L2 needs 0.2V less for 4.5A current, so that should work a bit better.

I'm getting just over 4 amps on a Nanjg with 4 extra 350 7135's with a de-dedomed XP-G2 on copper. Also I've added 380 7135's stacked on QLite drivers - no problem. The Pana CH cell is good, but not as good as Samsung INR 20Q or 20R or SONY 30A rated cells. 4.5A may not be achievable though. Cells must be freshly charged - it ain't gonna last too long... I always pre-test my stacked drivers on a XML T6 mounted on a big heat sink to verify the amps it puts out -- much easier with XML's than XM-L2's because of the Vf issue.