50%? I'll have to dig up some references, but I think that's pretty far off. Closer to zero.
Hmm.. Indeed seems I'm behind on where things are at.
https://en.wikipedia.org/wiki/Luminosity_function
The theoretical limit for white light seems to be about 250 lumens per watt, and 683 even for green light. So 50% might not be so far off! (you can pretty well fit a triangle in that spectrum and probably have a pretty high cri bulb, so half of 683 looks like a reasonable estimate of the white light limit to me, so around 350 at highest). Although you can actually waste even more power than that as light without even counting it as lumens in an un-optimal spectrum, and the brightest leds have a bunch of blue.
It still comes down to lumens per watt though mostly. Seems to mean there isn't that much room for improvement in LED tech, other than reducing heat. I mean once the LED is 90% efficient you won't get much more lumens per watt ever, but at that point you can still make it 99% and get ten times less heat.
Coming back to that driver efficiency, for linear it's just Vout/Vin regardless of how non ideal the circuit is.
I’m not really sure what you are getting at in relation to my question.
I couldn’t see where the OP stated which driver is currently in use. There are a lot of numbers being thrown around but if it hasn’t been stated how does anyone in this thread know what the OP considers to be too hot? There is a fairly large difference between a 3x and 8x equipped S2+ in both heat and light output.
Yep. That’d be roughly 3W into an XP-L to get 400lm… ish.
Concentrated in a small volume of metal, it’ll feel quite warm. Spread out to a bigger chunk of metal and lots more surface area, it’ll be spread out and be “watered down” across a much bigger area.
And beware if a light doesn’t get uncomfortably warm when pushing a few W through an LED. That usually means a lousy thermal path and keeps the heat bottled up in and around the LED itself without wicking it away to the case. Great for your hand, downright lousy for the LED.
Nah, as soon as that happens, some boffin will go pushing twice the power through the poor LED to make up for it.
probably
400 lumens is an arbitrary number anyways 
An X6 has more mass and surface area to absorb and dissipate heat so an equal wattage output will not feel as hot. A triple emits more lumens than a single at the same current so I’d reccomend a triple X6 on a Noctogon 32 with cute- 3 optics and XM L2 emitters T6 or better with 1.05A shared between them. In theory 3X T6 should give 420 lumens, U2 450 lumens, and U3 480 lumens thought losses in the light will lower OTF numbers for all of these to below 400 actual. The BEST driver is a Ledforpower LD-02 with the sense resistor properly chosen for your maximum current, no pwm means better efficiency than any 7135 driver and your not limited to multiples of 350/380mA but only by the resistor values obtainable so you could opt for more values between 1.4A and 1.75A.
XM-L2 U2 Bin: 2.7W = 428 Lumens
XP-G3 S5 Bin: 2.6W = 435 Lumens
4 x XP-G3 S5 Bin: 2.2W = 458 Lumens
The quad would be the most efficient, saving you about .5W, not a huge difference but you’d also pick up 20-30 Lumens over the singles, and 458 would put you over 400 OTF. Not sure if the cost difference would be worth the reward.
Edit: With the quad wired in parallel you’d need around 800ma of current. 2 of the 380ma 7135s would put you close at 760ma. Or use one of the 1000ma drivers and change the sense resistor.
ok, well he said 400 lumens. 3x vs 8x you mean leds? 7135's? I didn't know you can squeeze more than 3 leds in a s2. Yes that improves efficiency but has little to do with the driver I think. As for 7135's... again, he said 400 lumens. That takes whatever curent it takes for a certain led setup. My point is if delivering that current resistively, no driver is better than any other. The other option is PWM but that has similar losses to resistive anyway. I might be missing the point of the question still, probably am.
Has to be 3x 7135 vs 8x 7135 and it’s very true of both efficiency and total heat.
Back around 2007-2008 that was what I was taught as well, 250 lm/W, and there was a thread discussing it over at CPF about the same time, but Cree’s passed 300 lm/W now.
I found this:
https://www.dial.de/en/blog/article/efficiency-of-ledsthe-highest-luminous-efficacy-of-a-white-led/
It works out the integral which defines lumens as a function of actual watts of light, for various typically spectra. It looks like a typical cree spectrum is a bit over 300 lumens per light-watt, not so far from my by-eye triangle integral answer. So at around 320 lm/W, 100% of the energy is light.
So Walkintothelight has a great point there, still not an area we can really change.
Yeah I read that a couple of places, then I found that actual definition, and like I said, looked like about 350, and then the integrals above are for actual cree spectra.
This is actually based on the definition of lumens though. It's not like this is theory related to how led's work. This isn't actually theory at all, it's just that 1 watt of green light is DEFINED as 683 lumens. There were probably some rough estimates, maybe they asumed nobody would ever get all the light into the perfect spectra shape, ok, but 350 or so(might be a little more) is a hard limit if you want a high cri light. 1 watt of electricity cannot make more than 1 watt of light, well, not without cooling something (there is an article.... never mind).
Yes, and we should see that sooner than later since the 300lm/W barrier has been breached. On thing to note however, is that the figures Cree and other manufacturers use are typically well below the current at which most everyone drives the emitters, especially here.
True, it’s much farther off for high-CRI emitters, if ever, but the emitter Cree used in the 303 lm/W record was what most here would classify as Neutral White, around 5K IIRC.
Edit: It was about 5K, 5150K to be exact, but the surprising part is the drive current, 350ma, that’s awesome. We may actually see production emitters get there in a year or two.
http://www.cree.com/News-and-Events/Cree-News/Press-Releases/2014/March/300LPW-LED-barrier
Yes, but again, 400 lumens is 400 lumens. For a given led that's going to require let's say about 4 watts, and at 3.7V that's going to require about 1.1 amps of current, which is 3 7135s. It's not 1 and it's not 8 [1]. And you could do it with a resistor or an op amp, or with narrow traces or bad springs instead and the efficiency will still be 3.7V/4.2V or whatever exact voltage that current draw works out to.
[1] ok, like I said you can do it with PWM, using 8 instead of 3. Yeah that's a bit worse, but my bets are not so different at that power level. Yeah, maybe enough to matter. Indeed, best to use 3, and I guess that's the point both of you are trying to make. I'd agree with that.
You could get 480 lumens using only 2.2W with a quad XP-G3 setup. That’d put you somewhere around 420-440 lumens OTF. A CC 1000ma driver resistor modded down to 800ma would probably net a very steady output.
I think we are on different pages. The OP stated that they wanted something "with less heat and less brightness" and that they wanted something with 400+ lumens, if the OP tells us that the current driver is 8x7135 it would be reasonable to suggest that the OP tries a 3x7135 or a 4x7135 S2+, the 3x isn't that far off of 400lm and only gets what I consider to be slightly warm. OTOH if the OP tells us the current light is 4x7135 it would be reasonable to suggest the OP gets a light with more mass.
Not covered well is “How hot is hot?” In that lies the real issue here. I run a 10440 in a BLF348 and to me it’s very,very warm but not quite hot, yet many others call the same combo “too hot”. Of several lights I have that will do 400L, none of them ever feel hot to me at that level; not even close but the OP obviously perceives that same heat differently than I do.
DavidEF is on track with the “wattage = heat” message. The coolest light will be the one which uses the least current while delivering those 400 lumens 
How quickly it begins to feel warm will be relative to the mass of the light, the efficiency of the thermal path from the elements producing heat, and the efficiency of the host at dissipating heat. If left on until the heat ’normalizes’ completely, all hosts using the same components will be at the same temperature - that’s simple physics and the only differences between hosts will be how long it takes for the normalization point to be reached.
So my question to the OP would be regarding runtime, given that eventually all will heat equally eventually. If runtimes will be short anything will do but more mass will let the light feel cooler. If runtimes are long, the ability of the host to dissipate it’s heat (finning etal) will matter more. For average use the highest mass + the best heat dissipation + the lowest amperage to give 400L + the poorest acceptable thermal path for the components inside = equal the coolest feeling light.
Phil
Yes, and a larger surface area will also dissipate heat more quickly. Win/win.
Eventually that heat will find its way out of the internals and into the body of the light, so the body will eventually get hot. It’s just by that time, you may have cooked your LED and driver.
Overall, though, I’ve found more of a problem with lights heating up battery compartments, and damaging the batteries. (Though I don’t have any small lithium-ion 1000 lumen pocket-rockets.) One of my lights, a 4sevens Quark 2xAA XPG2 design, heats up a lot for a modest 250 lumens on max. (It’s an inefficient driver, not the LED.) It dumps the heat to the body just fine, but that really starts to heat up the battery compartment.
After just a few dozen cycles, the Eneloops I use in that light have lost about 10% of their capacity and the wrappers show signs of peeling. I can actually smell the batteries cooking when I leave it tailstanding, on max, for 30+ minutes.
Holding the light, and having that heat dumped into my hand, prevents the issue. But tailstanding, it dumps heat too slowly to keep up with the internal heating. It gets quite hot to the touch. All from only about 6 watts being powered from the batteries. But, probably at least 5 watts of that is waste heat. A small 2xAA light form just isn’t big enough to cope with that, unless there’s a breeze on it, or it’s cool outside, or you’re holding it.