In the 2014 brochure, Osram specifies following numbers for cellphone camera flashes OSLUX – LUW FQ6N.

If	350mA	500mA	700mA	1000mA	1.5A	2.0A
Φv (typ.)	98lm	130lm	169lm	218lm	300lm	350lm
Ev avg. at 1m	86lx	114lx	149lx	192lx	264lx	308lx
Uf (typ.)	3.25V	3.3V	3.45V	3.55V	4.0V	4.2V
Max. Pulse duration [Ta=25°C, D=5%)	> 10s	8s	2,5s	600ms	200ms	20ms

Well, 350 lm isn't that bad for a "selfie" in the dark place but still far from being usable as a regular camera flash. Luxeon Flash 7 goes up to 500 lm.

I googled a bit and a typical DSLR built-in flash should give off something about 30,000 lm for about 1/1000 s. (Wikipedia is distinguishing more times according the flash intensity which has a ramp up.)
From the table above it looks that even those special LEDs have their limits and their efficiency drops from 86 lm/W at the lowest intensity to 41.6 lm/W at maximum intensity.
In the plastic cellphone case, there's almost no heat dissipation. A DSLR body would probably have better surface for cooling but even with that, it looks pretty unrealistic to create 30,000 lm for 1/1000 s with LEDs provided the flash should remain about the xenon flash size. Or, am I wrong?
     
     
     
     




            
               
               

I'm unsure of how iPhone LEDs are powered, as far as current goes, but they seem pretty bright. My iPhone has been in use since 2012 and I don't notice any major color discretion of the flash. In my experience of overdriving individual led diodes themselves, at around 5v .5a, they'll be bright, then slowly dim down to a deep blue color. The leds I speak of are just plain Chinese 6500k diodes.

There are two things about a camera flash:
First what matters isn't that much the maximum output in lumens, but the energy of the light pulse. The discharges had very high flux output, but very short time (it can not do otherwise), the LED's are not capable of such high intensity, so they are operated longer in time. But with their limits, you have quite some freedom to vary the time/intensity ratio for still the same exposure energy.

And second thing is the heat: The power appears to be large for a design with no heatsink, so one (with experience with LED lighting) would say it should overheat. Well, the point is, the LED is ON for just a fraction of a second and during the time just the mass of the device is sufficient to swallow all the heat. Of course, when you would leave the power ON for longer, the device will virtually explode from overheating, but that is, why the datasheet specifies the pulse time limit for given current (power) level. You may well compare this with the Xenon discharge flash tubes: The peak power there is in kW range for a 15mm long, 2mm  diameter tube. With steady state load the tube would evaporate within fraction of a second, yet it operates at that power levels for the flash pulses...

It is a question, what is better, it probably depend on the scene you want to snap. Definitely tthe led's offer more than 10x lower battery power need. Part of it is they are a bit more efficient than the tube (about 20lm/W for the Xenon, vs 40lm/W for the LED), but mainly the LED auxiliary is way more efficient than the discharge auxiliary (about 50..80% for the LED ballast, vs no more than 10% for the Xenon discharge; the losses for the Xenon system are mainly in the capacitor ESR)

First what matters isn't that much the maximum output in lumens, but the energy of the light pulse. The discharges had very high flux output, but very short time (it can not do otherwise), the LED's are not capable of such high intensity, so they are operated longer in time. But with their limits, you have quite some freedom to vary the time/intensity ratio for still the same exposure energy.
ischarge; the losses for the Xenon system are mainly in the capacitor ESR)

Definitely. 30,000 * 0.001 = 30 and 350 * 0.02 = 7 (or 10 in case of those Luxeons). Not the same but still comparable.
Moreover camera sensors are continually improving in terms of sensitivity and SNR so higher ISO values are not so noisy. (And there's the promise of those organic sensors in the future that should make camera flashes redundant /as some say/.)
Also optical image stabilization makes those 20 ms (1/50 s) acceptable for many cases, except of sport, animals etc. as you probably suggested.

So something like a very strong 1/1000 s pulse is not feasible with LEDs. Even if they survived because of the very short time, they wouldn't give off much more light as their efficiency radically drops with the high current.


Solanaceae: It's probably the short periods of your iPhone flash operation what prevents LEDs from damaging.

Thanks, Medved.
Although, I use the flashlight feature to seek things in the dark and go into the basement to grab stuff (this takes no longer than 5 mins), the LED heats up, but not enough to worry about.

The "Flashlight" feature powers the LED on way lower current than in the "flash" mode, so the LED can stand the dissipated power. Typical are currents in the order of 50..100mA. With that you get about 15..30 lm of light, pretty sufficient for many uses (and indeed quite bright when you look at the LED)...

in a phone there is not much to heatsink the led.also looks like the thermal resistance from the die to the trace it is soldered to is high.so the limit is set by the mass of the die and the thermal resistance.the game changes when you have a good path to dump the heat.like me driving a cree xm-l at 6a with the die pad soldered directly to the copper heatsink.or the lumileds altilon at 2a on a copper cpu cooler.double the absolute maximum of 1a.this is impractical in a car headlight so they are derated.we can have a huge copper heatsink on a steampunk themed lamp though so we get away with it.the guy i built it for never turns it off and its over 2 years now.

In my phone - Samsung U900 there is literally a heat SINK - just a metal mass soldered to the same connection intended to sink the heat of the LED in the short "photo flash" mode by its sheer thermal capacity, then slowly dissipate it