You'll burn your hands before you melt the solder. Dont ask how I know. :P
So, you configure this board for parallel and just use a straight Nanjg/Qlite stacked for more output? Interesting... I got no experience with these small triples and the lens to jury rig into a host. I'm thinking a custom pill or custom mod to the pill and/or host would be needed to properly fit and align things with the lens?
The plastic will melt way before the solder. Here's a reference chart.
http://www.plasticoptics.com/optical-plastic-materials.html
So let's pick a number. 100°C. Relic38 didn't measure half that in his 16mm copper star test.
The Carlclo TIR has three support feet that fit in the three holes in the board to exactly align the optics (that is: if the reflow is done well ;-) )
Since we had a lot of question regarding coming with LEDs or not I will also post a answer here.
They will come reflown, we know it is not that is not easy to reflow 3 LEDs at the same time on a 20mm diameter board.
The boards are quite thick and you do not have to overdrive the LEDs, packing 3 LEDs at 1A-1.5A each in such a small space is enough to have the Vf drop because of the high heat, again the space is very small.
Thanks!
Hank
Ah, thanks for the chart, it is something I was curious about for a while :-)
Relic38's tests did take place with bare boards and extreme heatsinking, the temperature in a flashlight will become much higher than the 42degC that he measured. How high I don't know but I guess there's no harm in heatsinking very well.
So the LED wires go thru the center hole - I assume you would have to drill a hole through a stock pill top for that. Do you have to address the overall height and width issues of the lens? Do you get lucky and find a perfect host, or have to build up the pill topor something else? I've often seen these triple mod threads but I never could find the magic solution of how to get everything fitted. Am I just missing something here?
I was hoping he had measured the temperature in his torture tests too, but I didn't see that in the MT-G2 test. I'd expect a flashlight to get hotter, but not be such a poor conductor that the light won't get unbearably hot to hold long before 100°C.
this is very cool. What a smart way to get the 2 configurations out of the same star, I really like that.
Hank, will you be offering these bare? I understand that reflowing triples isn’t trivial, but it would be neat to have the option.
Tom - best bet would be a host with a 20mm lens and a screw in pill (Convoy S2, that C20 host I used?). Then you just have to take out the lens, drill the pill for the wires coming out the center, assemble pill and optic then screw it in. You might need to raise the springs at either/ both ends or use a spacer to take up the slack in the battery compartment, but that’s about the only hitch I can think of.
Yes, you need a hole in the middle of the pill. As for the host, I guess you need some creativity.
The only triple I have build is a 20mm triple Nichia219 with these (narrow) Carclo optics (parallel wired with 3.04A qlite driver). I used this Convoy S5 host from Fasttech, it has the right diameter to fit these boards/optics, although a small slice of light (~10%) of each TIR is blocked by the front o-ring. Because a reflector is much deeper than the TIR I have put a cilindrical block of aluminium between the led-board and the pill as a spacer (can store some heat as well). I could make a picture of the light tonight..
The optics are 20mm, lens in something common like the little Convoy lights is exactly 20mm. So the diameter isn't an issue. Cut out the entire center of the stock pill, press in a piece of 5/8" bar stock, trim to length, then stick the MCPCB on top with epoxy if you cheaped out and used aluminum bar or solder if you got all fancy and used copper bar.
It's probably an easy fit in the short reflector lights (Convoy S3), a little more work to make a taller extended pill for the deep reflector Convoys, and it'd be tough and require a lot of luck to do it in a light without a removable pill (Roche F12).
Ahhh - thanx matt and djozz! Awesome setup - and I see, takes some tinkering for sure. This Noctigon is what I believe brings these little triples to a new level. Even if you have 3 XP-G2's driven at 1A-1.5A each, you gotta be getting benefits from the heat relief the copper direct path board brings.
Hank - so with 3 XP-G2's you don't feel the high Vf will be much of a problem on a single cell with a linear driver such as a Qlite?
In theory, these triples are an ultimate pocket flooder because for a single, we all know uping amps from 1A to 3A does not result in a triple increase in lumens, or 1A to 2A in 100% increase, while running triples at 1A each, you should be get closer to achieving triple lumens, specially on direct path copper.
I'm inclined to try 3.5A initially and if you can get extra heat sinking in there in a host like the S3 or S5, maybe you can get a couple of mins of runtime before you get that burning sensation in your hand 
.
Adding multiple LEDs in parallel will pull more out of the same cell as compared to a single LED. It doesn't scale 1:1 but it's pretty dramatic. It's been a while since I measured one but I think it was something like direct drive to a single XML was around 4A, same cell direct drive to two XMLs in parallel was 5.8A or so. The Vf vs. cell voltage is less of an issue. I imagine it would be fairly easy to get 6A total from a decent cell into 3x in parallel.
some quick calculations on a direct drive single 18650 (IMR pana cgr18650ch) flashlight build with triple xp-g2 parallel : taken into account 0.1V voltage loss in the system you can get 3x 2.5+ amps (2100 OTF lumens) for at least 5 minutes :evil:
Nice board, lot’s of possibilities!
Does anyone/everyone know that mattaus makes silver plated copper pills for the triple in a P-60 drop in? Center hole already in place by design.
The EDC+ Triple XPG-2 uses a similar silver plated copper pill (they worked with Matt on this) and they drive it conservatively for a measured out the front 720 lumens. It’s only pulling like 2A.
I just put an XP-G2 R5 2B in my HD2010 on a Noctigon, with a half inch thick 3/4” wide copper heat-sink under it and it hit’s 610 yds fairly easily pulling 4.25A all by it’s little self. 
That’s a Q-Lite with 6 chips.
I knew I had some pictures of the pill assembly (TIR is not in picture) already somewhere on my phone, just uploaded them :-) :
These boards sure get crowded when you parallel wire them (these wires were tiny already!) 
I have one of the EDC+ Nichia triples and it is a very nice dropin indeed but under-driven as you say. I sometimes think the 50% medium might be better at 30% but its not a deal-breaker.
Q-lite has 8 chips going by the photo on intl-outdoor. Dumb question but how did you get 4.25A out of it? Did you mean 6 extra chips perhaps?
you’re welcome 
Thinking about it a bit more, this would take a bit more tinkering than I originally thought, as the lights I suggested take 16mm stars. I guess you could deal with that by cutting the lip off. The TIR I have in my XM-L2 equipped C20 (?) fits the front O-ring perfectly though, so that part at least is simple.
I think a triple XP-G2 NW would make an awesome EDC light - pretty floody but with enough through to be useful at distance too. A friend of mine likes building 20mm triple bike lights and he uses XP-G2 triples for the bars (flood and through) and XP-E2 triples for the helmet (throw). I reflowed a 2x XP-G2 1x XP-E2 (or possibly the other way round) triple for him that he really likes too.
With Salsa please! My memory is horrible. I’m pretty sure I put 6 chips on there, but the light has been done for quite some time, I just replaced the emitter/star with the XP-G2/Noctigon.
Yes, Q-lite has 8 chips at the 380mA binning for 3.04A right out of the gate. I modified 4 of those for a friend day before yesterday by stacking 4 chips on each board, and tested them on a mounted XM-L U2 for 4.14A.
It might be interesting to note, I also stacked chips on a QLite for an XM-L2 on Noctigon in my new M8, and different batteries yield drastically different results! With the Samsung 20R’s it pulls 5.02A, Efest V1 pull 4.83A, V2 pulls 4.65A, Panasonic PD’s pull 4.61A and the AW 2600mAh cell pulls 4.11A. So yeah, the cell used makes a lot of difference!
It might also be of interest to note that 3 or 4 months ago I couldn’t solder for crap! Now I’m stacking chips and adding teeny tiny 200 ohm resistors that look like a flake of pepper from my pepper mill grinder!
Hakko 888 soldering station and .031 Kester solder has made worlds of difference.
This is awesome stuff!! Looking forward to building a light or two around this baby. Thanks a ton, Hank!!
Any chance of an XM-L version anytime soon?