He's having a contest. You don't win contests by taking people to dinner.
I feel you missed the point, that it doesn't have to be a short-lived, overheating result. That's why specs exist, so we can go from there building not just some contest winner but something with lasting value.
My contrast was one of winning some contest by building a flashlight that isn't viable for (literally) 20 years or more, versus the effect of wowing someone for a moment. If the wow factor was all that was important, the $100 dinner wins. If long lasting wow were more important, build a flashlight to be superior for all the virtues of it being LED-based since brightness is always the compromise but offset by other virtues. Trying to pump as much current as possible through an LED regardless of efficiency or lifespan is clearly a misuse of LEDs, is like negating everything we hold useful about them.
I'm claiming that the XR-E driven at 2A will last longer than he wants to use it, since it does not fail catastrophically even when run for 30 minutes. The onus is on you to disprove my claim.
Again, we have no evidence that it:
1) Runs at over 2A for 30 minutes. It will require a constant current supply or constant measurement to determine this, certainly not 4 X AA. If 4 X AA can power it for 2A at the end of 30 minutes, they will be exceeding 2.5A at the beginning. Science is really important here, a glance at a graph is not the same thing as considering all variables.
2) We have no evidence that when running at excessive current, the LED will not be dropping not only in lumen output but permanently aging, prematurely. Let's suppose when the LED is brand new, you can momentarily get 350 lumens out of it. Is that really winning a contest if you need 1650/350 = ~ 5 LEDs to do it, but then if you kept trying to use that flashlight then it drops lower and lower in output when the halogen version still continues outputting full light? That's a bit like cheating, nobody builds a flashlight that can't meet real-world uses.
No, if it runs long enough to be cost-effective against a halogen, remains reasonable efficient, remains reasonable usable. Otherwise it's just an arbitrary concept, not worthwhile as a light. It's a shame you can't trust the very makers of an LED, their own specs in a highly competitive market. They might be a bit conservative, but not by 200%.
There's a little tool I like to call the inequality. According to this Energizer datasheet, at a discharge rate of 1C (2.5 A), a typical Energizer 2500 mAH NiMH rechargable can sustain at least 1.2V for 30 minutes...
At 100% full charge, when brand new, at optimal temperature, etc.
Now let's get back to reality. The fact is, when draining a NiMH AA at this rate, one of two things has to be true. Either it started at significantly higher than 2.5A, or it ended up significantly lower than 2.5A after 30 minutes.
Look again at the datasheet and see that for the first (1/3rd or so) of that 30 minutes, voltage ramps down significantly. One of two things is true, either the LED was ran at far above 2.5A (which would fry it), or it was not far above that figure and the rest of the time it was at lower current.
That's why a constant current source has to be used to make a claim about current:time.
multiply that by 4 batteries gives a total 4.8V... referring back to evan's post, we see that Vf at 2A is only around 4V. 4 AA batteries are definitely capable of direct driving an LED for 30 minutes at over 2.5A, it's probably closer to 3.
http://data.energizer.com/PDFs/nh15-2500.pdf
While this seems simple enough, it directly contradicts the claim of 2.5A. IF
1.2V/cell had been maintained, Evan's data shows the current would be higher than 2.5A.
I would like to pause for a moment and remaind you that anyone can build a circuit that fries itself because it wasn't properly designed. There is no trick to that. Recognizing what works for a flashlight is a different matter, and if spending up to $100 as the original post suggested is a budget, trying to squeeze every last lumen out of fewer LEDs just doesn't make sense.
I have already disproven the latter part of this paragraph, as for the first half. This is a contest, again, it only has to last for the duration of the contest, which is measured in minutes of run times. If it lasts 50,000 hours, then great, we have a bonus, but if it doesn't, it doesn't subtract from it's original purpose as a tool to outthrow a halogen spotlight. We do not care about how fast about the LED dies because OP is in a throw contest with a halogen spotlight, not a throw contest over 50,000 hour with a halogen spotlight (not to mention halogen spotlights don't last 50,000 hours). The OP did not specify that he wanted in the feature set that the flashlight has to last 50,000 hours.
I don't feel a term of the contest was that it's ok if it's not usable later, unless there was a monetary bet. WHY would you insist it has to be a wasteful destructive design? It's not as though there isn't another way, this is not a hard thing to do. It is just a bad design to push parts to the point of frying for no reason.
Again, the market demand for XR-Es driven at 2A is not very high.
Again, if all they had to do was state "use a larger heatsink", then they would have done so, as they had already stipulated the importance of heatsinking at lower current.
The reduced lifetime would be prohibitively expensive. Multi-die emitters are more cost effective. Also, if Cree included graphs of their XR-Es at 2A, it would overlap the multi-die emitters and be competing with itself.
They don't provide higher current graphs because it's not designed to withstand that. Take any other semiconductor as an example, it is clear the manufacturer rates for max current and that in that scenario suitable heatsinking is required. Datasheets exist for a reason, and that reason is to counter people who have not done the extensive testing the manufacturer does. THEY know their product and the whole point of the datasheet was to provide this info. If everyone were as cavalier as you seem, we'd have the same situation with LED datasheets as we have with generic power supplies or computer speakers rated for 400W PMPO but cost only $8.
An LED is not an end onto itself. If you can't appreciate their virtues and take advantage of them without losing efficiency and lifespan, then there was no point. A one-time 30 minute run or 5 minutes at a time is not proof of a good result, anyone who bought a flashlight that could only do that much would consider it defective.
The main point is that there is no good reason to sacrifice life or efficiency just to use fewer LEDs, as it is certainly impossible that the contest be won with only one or two, or actually as the math suggests, even 4 cannot win such a contest. Therefore, the idea of using only 2 or 3 LEDs is gone, it has to be a custom flashlight and in that context, no reasonable implementation that tries to push each in output when it can be smaller, lighter, and more efficient merely using more LEDs.
I think we need to get more info from the thread originator about whether he'd rather throw away a $100 flashlight just to use fewer LEDs, or if it'd be nice if it was usable for many years instead.