Very informative thread. Thank you very much!
Do you plan on adding more LED types to the tests? XR-E R2 (WG) perhaps, and SST-90? 
thanks very much for running these tests. This is incredibly useful for making informed decisions.
I have a question though, that I don't 'think' has come up here - if the emitter(s) were actively, and well cooled, would we see a higher peak output? Ok, it would seem obvious it would increase at least some, but I wonder how much.
Really though, what would be useful, is if it extended the nonlinear relationship of input/output. Instead of plateauing around 2.4-2.8...to say 3.5 
Fair enough question....but, for the tests conducted thus far active cooling via fan and/or water would make no appreciable difference. For the given runtime on the above tests, the thermal capacity of my heatsink setup can be considered infinite (as confirmed via thermocouple/fluke monitoring). The bottleneck is the pcb.
Now having said that, I'm staring at a box containing just over half a dozen peltier coolers leftover from a different project. Using one, or better yet, stacking them can yield near cryogenic temps on the top plate where the star could be mounted. Considering each peltier requires 150w, the power requirements (and cost) launches the idea out of the realm of flashlight practicality. Although the results would be interesting for curiosity's sake, I don't see myself getting that bored yet to try.
Thanks very much for the response. I've had an internal debate for weeks on buying copper PCB xmls (instead of aluminum) for a project. I think it will be worth the extra money to buy copper.
I think you're right about the impracticality of a peltier. But let's hope for some new advancements in battery technology to change that :)
Even if they do, it will probably be up to us to build such things - at least on MY budget...
Folks,
Updated with XP-G data in the first post. I also cleaned it up a bit, like fixing that horrible non-linear graph for the xm-l.
-Match
Great. Thanks Match.
This post has really become my reference post when deciding about current levels.
Thanks a lot M8 :-)
That's crazy I have been pushing XPGs to 2.2 amps (8 chip Nanjg driver with 2 removed) because that's were they looked the brightest by eye and it was right on the money.
Many thanks for this addition, Match!
(Standing by for the XR-E ... 
)
I would really love to se an XP-E or XR-E q5 led in the mix. :) I still like them.
Thanks Match! Been waiting for this info to build my XP-G project. Question now is, is it worth the sacrifice of runtime to get a little more light. I'm looking at 600mA to gain about 50 lumens. I don't think it's worth it. Still seams like 1.4A is the sweet spot for runtime vs. output.
-Garry
2.2A xpg is exactly what the N-light B2s stainless steel lights are running on the 14500 ..
always got a ..."WOW that's bright".... from the crowd ...
ooooohh....aaaahh.
And what would be a similarily "bold" current for an XR-E?
oldbobk,
So are you saying the affect on runtime is negligible between running 1.4A vs. running 2.1A? That doesn't make any sense to me.
I also don't know if I'm up to the challenge of removing two of those chips! I'd probably look into the newer KD driver (and deal with a lousy low, or go 2 mode).
-Garry
The difference in run time with a 2400mah battery would be about 100 minutes for the 1.4 and 70 minutes for the 2.1 amp high, but I never use high when I am concerned about runtime.
Send me a PM when you are ready to remove the chips I can walk you through it or I can send you one premade if you want.
Thanks for the offer Erik! I am definitely keeping you in mind for a custom P60 drop-in or even just a custom programmed driver (think I'd like Medium a bit higher than the usual 30%, or even 4 brightness modes) for another mod.
100 minutes vs. 70 minutes sounds considerable (for my application), so therefore I don't see the gain of 50 lumens worth the 30 minute trade-off.
-Garry
Don't want to be a spoilsport but please remember that Match's curves are only valid for a heavily cooled emitter. The top points for the curves in real life lies at a much lower amperage because of the generated heat .
[EDIT]: Calculations using Cree's derating curves give me for a "normally" cooled flashligt a top on the curve at an LED output of:
370-400 lumen @ 1.8A for XP-G R5 and
730-780 lumen @ 3.5A for XM-L T6
but the top 30% of the current only add about 12-15% of light output.
OTL (out of the lens) output normally are about 80-85% of the LED output.
Yet another reason for me to stay back to 1.4A / 1.5A!
-Garry
Possibly the most informative post for us modders on this forum. I've actually photographed the charts to my phone for easy reference.
Thanks Match!