Yeah, I haven’t found any constant current power supplies that can do 50-60A, but if you read the speds it does say on the meanwell PSUs that they limit current at their rated maximum (so 50A or 60A) so maybe it would work even though you can only manually reduce the voltage to 4.5v.
The first two I linked would be great for a stationary searchlight though.
I think one solution to get ~40A and an easy driver solution is to run 4s2p NiMH in 4/3AF (~18670) size. They are about 4500mAh and can do 4C generally pretty easy. Panasonic hhr-450a (now bk450a) and others are quality cells that can be had around $8 a piece from mouser/digikey. Powerizer has a 3.8Ah version that goes for half that. Should fit a GT without modification. Would result in 4.8V nominal. A large diameter FET driver should handle this no problem.
Those tests are done at full load with a heavily overclocked CPU and still around 50°C core temperature with best air coolers that have 140mm fans
this is with a whisper fan at 1500RPM
2) If it was an i5, then no wonder the temps are so low.
3) that is a 7 year old CPU, go look at some recent tests of actually “high end” CPUs made in the past few years.
4) the D15 is literally the best air CPU cooler on the market. And it still can’t reach 45 degrees.
5) the CPU is not overclocked either, it is at stock
6) Even though the silent wings 2 is one of the best quiet fans, it is not “whisper quiet” at 1500rpm, I would say maybe at 400-600rpm.
Enderman, I think you are overestimating the practical effects of your special heatsinking methods. The gain you managed to get with the Osram Black Flat for example was rather small and not really worth the added cost from a luminance standpoint ( I’m talking about the liquid metal paste).
sma’s cooling is not perfect, but it’s not bad either. His actively cooled CPU heatsink is way better than every standard LED light ever made (just like your light canon).
Nobody has tested this LED except him. I don’t understand why everybody is complaining.
60A might be possible with extreme, unrealistic cooling, but it wont be worth it. There will be no perceiveable brightness increase. Only wasted energy and thus more heat.
I didn’t apply the liquid metal paste to the Optofire yet…
It hadn’t arrived when I did the build, I still have to switch the regular paste for it.
When I do switch it I will redo my measurements, although it may be lower now since the LED has undergone many hours of use.
My point is that no a CPU heatsink is not some amazing piece of magic that will bring your 200W LED temps down to 45C. It can’t even get 96W CPUs down to 45C.
As an enthusiast, a slightly higher number does matter to me, because small things add up.
If you think that 4.5Mcd is not any visually different from 4.0Mcd, and you would go with 4Mcd, then you can also say that 3.5 is no different from 4, or 3.0 is no different from 3.5.
In the end you will end up with a mediocre flashlight and nothing special.
You don’t innovate by “compromising” and not pushing further because “oh there is no perceivable difference”
Yes his heatsink is much better than a normal flashlight, but the person I was talking to is one of the creators of STO, and neither he nor I build “normal flashlights”
We are discussing the absolute limits of what this LED could do with “proper” cooling, which I was estimating to be 50-60A.
“Liquid Metal” Hmm… Anything that is liquid evaporates sooner or later :question:
Please then also measure the light in it’s current state. You don’t need to even do a real throw measurement in this case. Just a controlled test of relative output.
How much do you think you will gain? I don’t think it will be much because you are already using a copper pcb which distributes the small amount of heat (24W) onto a much larger surface with a very low thermal resistance. The thermal pad of the LED has an area of 2.7mm^2 and the XP32 pcb has an area of slightly less than 804mm^2. So you already reduced the power density from 8.9W/mm^2 to 0.03W/mm^2 by a factor of almost 300! (lets just ignore the holes of the pcb here).
I think there are more gains to be had by optimizing the solder procedure (type of solder used, amount, pressure, temperature profile) and binning multiple LEDs under indentical, high-current conditions.
German modder Vinz made himself some synthetic diamond based thermal glue a few years ago (high diamond content, very expensive) and he did achieve some benefits in combination with some other measures (pcb and heatsink flattend and perfectly polished, LED soldered under pressure using very expensive solder with very high silver content, LEDs with especially low Vfs etc.), but in the end the differences were limited. With all his tricks he managed to get an XM-L2 to run at 7.5A.
Granted, back then all the LEDs (XM-L2, XP-G2) died at some point because the bond wires melted at very high currents (he even strengethend the bond wires of an XM-L2 in one light). I guess his measures would have a greater effect on todays LEDs.
He used all his tricks to make the Mjölnir (Infos here, here here).
Well yes, obviously there are physical limits to what a heatsink will do. The thermal resistance of the LED alone will prevent 45°C at 40A. And, yes, I now who you are talking to.
It doesn’t matter what I say is visually different. I don’t decide how the human eye works. 
If you want a small visual difference you’ll need about 20% more cd.
In my eyes going from 4Mcd to 4.5Mcd is not worth it if this makes the light unstable, dangerous, unreliable, overheat too early or much more expensive.
Going from 3.5Mcd to 4.5 would be worth it to me if it’s possible in a resonable way.
But, if I were already at 4Mcd and now need to pay hundreds of dollars to get a little bit more even though the LED has no chance of becoming much brighter, then I would stop at some point and focus my efforts on other aspects of the light (reliability, size, runtime etc.).
But I get your discussion. It’s always good for someone to push the limits.
Vinh put this LED in TN42 with direct drive. 3800 lm and 900 kcd sounds OK, doesn’t it? He says heat is not a problem.
I guess it runs at 20A or slightly more.
http://www.candlepowerforums.com/vb/showthread.php?442055-TN42vn90
Nice result. Guess that will be a good seller, with the baddest-ass LED around.
Holly cra ck! This would work exceptionally well in aspherics so Mr. Enderman it is no big deal to invest 100$ to improve light zooka?
Hm, I will think about it.
It would make the beam a lot more visible with all those lumens, which I like, so maybe if I can get it to close to black-flat intensity it will be worth it.
Not sure how effective my cooling system will work at 200W though, it may require a redesign.
Oh don’t worry about cooling… You see that Vinh is not worried so why you should be ? Run it for shorter times.
Just do it… I really saw from your posts that money is not issue for hobby you like… It will worth every penny even if it will fall somehow in candela performance but I guess It will not drop on performance and wavien collar will double intensity…
Edit; Just imagine the beam… Big fat juicy beam… Zombie, and vampire killer zooka 
Not to mention how our Jaws will drop when we see 2-3mcd light with massive hotspot
Do that and you’ll be in jail soon (you will really have to find more suitable place for testing after such modd )
Yeap., i am one of the people who buys 25k or 40k lumen light… but i can also run 40k lumen at 10 % mode which is 4000 lumen for the long time. Heat is not an issue.
talking about CPU cooling, it depends on very many things
for example: you live in tropical country, where Avg temp is >30*c, bring the OCed cpu down to 45*C is almost impossible, but if you live in like Greenland, or North Pole(:P), i think it’s not the problem
the other thing is cpu bin(like led bin, but you can’t choose cpu bin), sometime if you are lucky you can get a cool CPU that can go 5ghz with decent cooling, and if you are unlucky, you may even cant go higher 4.8ghz or else you get BSOD
I don’t like lights that drop intensity over time.
If I build it, then it should run at 100% continuously
Can’t really use a wavien collar with the mirror design I’m using which I recently posted about.
Honestly it’s a trade off, either 2x the lumens, or 2x the intensity.
I’ll definitely consider this LED for future projects though.
Enough for cooling without a fan ?
This driver with a 6S Lippo is planned:
Regards Xandre
No, probably not. It is a chimney design, turn it horizontally and performance will suffer greatly. Though it has high thermal mass, so one may simply use small enough batteries and it will sink the heat.