Anyone Thought Of A Liquid Cooled Led?

Junk · Jul 19, 2015

Ive built several liquid cooled computers in the past.

I was also thinking about building a light.

Then it occurred to me, if I'm gonna be attaching them to heat sinks anyway, why not liquid cool those heat sinks?

You could drive the 3590 harder, also cut down on heat. I want to drive the 3590 harder than I see plans for, I don't care if I shorten the life cycle. By the time they are dead something new and better will be out. But liquid cooling would allow you to drive it harder with no significant temp increase (I think?)

Any of the more advanced lighting guys see an inherent problem with this before I continue thinking about it? It would be very easy to do. A simple radiator, res, connections, and transfer plates.

Toaster79 · Jul 19, 2015

Totally feasible. Too bad those new Crees are on ceramic substrate. CXAs were on copper so you could reflow them and drive them much harder this way. But yes, liquid cooling is easily achievable. Are you thinking of your own water blocks or something readily available? How about liquid nitrogen? How crazy can you go?

G gnome · Jul 19, 2015

Toaster79 said:
Totally feasible. Too bad those new Crees are on ceramic substrate. CXAs were on copper so you could reflow them and drive them much harder this way. But yes, liquid cooling is easily achievable. Are you thinking of your own water blocks or something readily available? How about liquid nitrogen? How crazy can you go?

Are u some sort of evil genius?

Toaster79 · Jul 19, 2015

G gnome said:
Are u some sort of evil genius?

Sort of :borg:

coloradoBTC · Jul 19, 2015

the whole point of leds is efficiency, doesn't driving them harder decrease efficiency?

Toaster79 · Jul 19, 2015

coloradoBTC said:
the whole point of leds is efficiency, doesn't driving them harder decrease efficiency?

Not if Tj stays in range. Keep the junction temperature low enough and you can drive them as hard as you wish. Just like CPU. Fuck it you can use readymade water cooling sistem for computers.

This one from Zalman looks sexy

coloradoBTC · Jul 19, 2015

Toaster79 said:
Not if Tj stays in range. Keep the junction temperature low enough and you can drive them as hard as you wish. Just like CPU. Fuck it you can use readymade water cooling sistem for computers.

This one from Zalman looks sexy

View attachment 523768

View attachment 523769

how much power does that use compared to a hyper 212 though? the LED itself is running at peak efficiency, but shouldn't the power used to obtain that also count in the efficiency equation?

str8smokn · Jul 19, 2015

Interesting!! I have not heard of water cooling for electronics. This is why I love this place..
STR8

Junk · Jul 19, 2015

The way they are grading the optimal efficiency vs afaik, is light produced vs. heat output. If you control the heat, you can drive the light harder but it's not producing more heat. We are exerting an external cooling force on the led, so in theory, if optimal temp is x, and we can get it to run at temp x, but increase the light output.

I know it can physically be done Toaster (I was hoping you would chime in on this one) like you said it's no different than a CPU, which is what I have experience with.

But would the benefits be there? I guess meaning; it works in theory...anyone with experience think it will work in real life as described?

Sounds like the 3590 won't work with this method. Have another suggestion that would be as good?

I can go pretty crazy, my initial thought would be to use a car radiator. But I would rather not go the liquid nitrogen route. I was thinking of doing a bar setup almost the same as Seaslug's. Same parts, just more LEDs.

Toaster79 · Jul 19, 2015

G gnome said:
Are u some sort of evil genius?

Even the Chinese know it :D :D :D :D :D

Toaster79 · Jul 19, 2015

Junk said:
The way they are grading the optimal efficiency vs afaik, is light produced vs. heat output. If you control the heat, you can drive the light harder but it's not producing more heat. We are exerting an external cooling force on the led, so in theory, if optimal temp is x, and we can get it to run at temp x, but increase the light output.

I know it can physically be done Toaster (I was hoping you would chime in on this one) like you said it's no different than a CPU, which is what I have experience with.

But would the benefits be there? I guess meaning; it works in theory...anyone with experience think it will work in real life as described?

Sounds like the 3590 won't work with this method. Have another suggestion that would be as good?

I can go pretty crazy, my initial thought would be to use a car radiator. But I would rather not go the liquid nitrogen route. I was thinking of doing a bar setup almost the same as Seaslug's. Same parts, just more LEDs.

I'm comming myself from a computer world so that's the knowledge I'm using when were talking LED.

The fact is as long as the junction temperature stays let's say at 85°C (Cree bins their emmiters at this temperature) the efficiency will not suffer. Keeping it lower means higher efficiency (look at Crees datasheets, there are graphs for different efficiency at different temperatures). The biggest problem is the bond between emmiter and cooling surface. Using thermal compounds or even thermal glues presents the bottleneck in thermal management. You can have a super duper copper core heatsink but when you use crappy thermal past or glue, or apply a thick layer your killing thermal mannagement. Thats why I was mentioning reflow technique with LEDs as it is often used in world of flashlights. For example XP-G2 is rated for 1.5A current. Cam you imagine driving it at 5A? Well, it's been done before. If you're willing to waste your time on learning about different crazy approches at thermal mannagement I suggest wisiting Candlepower forums. I spent a few years there learning about all sorts of stuff and building flashlights. I even built a flashlight with a Cree MC-E running at 3.5A.

3590 or other CXB series COBs are on ceramic substrate, but that doesn't mean they can't be hard overdriven because of that. We just need to look outside of the box. Read this and youll see where I'm going with this:

http://www.s-bond.com/blog/2011/04/04/ceramic-metal-bonding-part-one/

So I'd say everything is possible with some research and will.

I just built a light system with 4 3070 running at 75W each. The heatsink barely gets warm to touch so I'd say I did a good job. Its a bit different approach than seaslug and SupraSPL are using but works great.

Will have to look deeper into ceramic to copper bonding and build a unit with a heavily overdrive COB just for testin purposes. Mine 3070 are rated for 1900mA but are running at 2100mA ( 2500mA max?) And those heatsinks are rated for 90W. Like I said, ju can hold your finger on that copper core and it wont burn so I'd say everything is serving its purpose.

G gnome · Jul 19, 2015

Toaster79 said:
I'm comming myself from a computer world so that's the knowledge I'm using when were talking LED.

The fact is as long as the junction temperature stays let's say at 85°C (Cree bins their emmiters at this temperature) the efficiency will not suffer. Keeping it lower means higher efficiency (look at Crees datasheets, there are graphs for different efficiency at different temperatures). The biggest problem is the bond between emmiter and cooling surface. Using thermal compounds or even thermal glues presents the bottleneck in thermal management. You can have a super duper copper core heatsink but when you use crappy thermal past or glue, or apply a thick layer your killing thermal mannagement. Thats why I was mentioning reflow technique with LEDs as it is often used in world of flashlights. For example XP-G2 is rated for 1.5A current. Cam you imagine driving it at 5A? Well, it's been done before. If you're willing to waste your time on learning about different crazy approches at thermal mannagement I suggest wisiting Candlepower forums. I spent a few years there learning about all sorts of stuff and building flashlights. I even built a flashlight with a Cree MC-E running at 3.5A.

3590 or other CXB series COBs are on ceramic substrate, but that doesn't mean they can't be hard overdriven because of that. We just need to look outside of the box. Read this and youll see where I'm going with this:

http://www.s-bond.com/blog/2011/04/04/ceramic-metal-bonding-part-one/

So I'd say everything is possible with some research and will.

I just built a light system with 4 3070 running at 75W each. The heatsink barely gets warm to touch so I'd say I did a good job. Its a bit different approach than seaslug and SupraSPL are using but works great.

View attachment 523808

View attachment 523810

View attachment 523811

View attachment 523813

Will have to look deeper into ceramic to copper bonding and build a unit with a heavily overdrive COB just for testin purposes. Mine 3070 are rated for 1900mA but are running at 2100mA ( 2500mA max?) And those heatsinks are rated for 90W. Like I said, ju can hold your finger on that copper core and it wont burn so I'd say everything is serving its purpose.

Awesome work pal

Junk · Jul 19, 2015

Toaster79 said:
I'm comming myself from a computer world so that's the knowledge I'm using when were talking LED.

The fact is as long as the junction temperature stays let's say at 85°C (Cree bins their emmiters at this temperature) the efficiency will not suffer. Keeping it lower means higher efficiency (look at Crees datasheets, there are graphs for different efficiency at different temperatures). The biggest problem is the bond between emmiter and cooling surface. Using thermal compounds or even thermal glues presents the bottleneck in thermal management. You can have a super duper copper core heatsink but when you use crappy thermal past or glue, or apply a thick layer your killing thermal mannagement. Thats why I was mentioning reflow technique with LEDs as it is often used in world of flashlights. For example XP-G2 is rated for 1.5A current. Cam you imagine driving it at 5A? Well, it's been done before. If you're willing to waste your time on learning about different crazy approches at thermal mannagement I suggest wisiting Candlepower forums. I spent a few years there learning about all sorts of stuff and building flashlights. I even built a flashlight with a Cree MC-E running at 3.5A.

3590 or other CXB series COBs are on ceramic substrate, but that doesn't mean they can't be hard overdriven because of that. We just need to look outside of the box. Read this and youll see where I'm going with this:

http://www.s-bond.com/blog/2011/04/04/ceramic-metal-bonding-part-one/

So I'd say everything is possible with some research and will.

I just built a light system with 4 3070 running at 75W each. The heatsink barely gets warm to touch so I'd say I did a good job. Its a bit different approach than seaslug and SupraSPL are using but works great.

View attachment 523808

View attachment 523810

View attachment 523811

View attachment 523813

Will have to look deeper into ceramic to copper bonding and build a unit with a heavily overdrive COB just for testin purposes. Mine 3070 are rated for 1900mA but are running at 2100mA ( 2500mA max?) And those heatsinks are rated for 90W. Like I said, ju can hold your finger on that copper core and it wont burn so I'd say everything is serving its purpose.

So you're are just using box CPU coolers...liquid cooled would be an instant adaptation.

I should start a kickstart account

How are people currently using heat sink setups bonding their lights?

What I would like to do is make is a genuine 1k led. Not "equivalent" (I guess in performance) but a 1k draw (roughly) Buckets of intense light to cover roughly a 5x5

Agreed on the thermal paste, I always just bought the best in bulk so I kind of take it for granted. I still have a few of the cheaper ones in the garage.

Toaster79 · Jul 19, 2015

I'm using Arctic Silver products

I would really have a deep look in some nubers comparing copper core heatsink with copper heatpipes filled with gas and water block with coolant running through silicon piping and have the heat removed by forced air just like the box CPU heatsink.

Look at the construction of this heatsink:

You have a copper block with heatpipes soldered to it and futher to aluminum part of the heatsink. I'd say this type of construction outperforms liquid cooling. Solder your COBs directly to the block and you have the best possible heat path and fastest heat removal.

Junk · Jul 19, 2015

I don't remember the name, they all have arctic or silver in the name lol.

So your saying those heat sinks are enOugh to do the jobs themselves. Do I even need a heat sink if I'm doing 4" wide heat sink bar as the structure?

Toaster79 · Jul 19, 2015

You can also go pasive way with that 4" bar, but make sure it's rated for high enough power dissipation. You also wont be overdriving your COBs as the heat removal from the core wont be fast enough. Aluminium has much higher thermal resistance than copper. Also keep in mind that with pasive heatsinking you need at least 4sqr"/W of surface. Also heatsinks for passive or active cooling have different fin structure. Active ones have higher count of thinner fins as the heat is being removed by air forced through them where passive heatsinks rely solely on natural convection so the fins are further apart so the heat can escape vertically and doesn't get trapped between.

Junk · Jul 20, 2015

The whole point of the heatsink is to transfer heat no? The principle being that you are giving the heat more metal to offset it, and more surface area to radiate and cool itself.

I was going to use the .485 ones from heat sink USA.

Junk · Jul 21, 2015

Also, those prepackaged cooling systems are ok for someone who doesn't want to mess with ordering different parts...meaning they want a kit. But it's not going to be as good as one you could put together yourself.

Toast,

I was reading an article once about how to apply the paste. I had always done it like you posted above. Apparently, it leaves small air pockets. The most heat transfer they achieved was by putting a thick strip down the middle and then squishing it down so that it comes out the sides. I think the next best pattern was an x connecting diagonal to diagonal. A circular blob in the middle was also good.

Toaster79 · Jul 21, 2015

Junk said:
The whole point of the heatsink is to transfer heat no? The principle being that you are giving the heat more metal to offset it, and more surface area to radiate and cool itself.

I was going to use the .485 ones from heat sink USA.

Basically yes, but what I'm trying to say here is the speed at which the heat is being removed from the core is crucial. Copper has higher conductivity than aluminum, silver even higher, meaning the heat travels faster through the material. Once that is accomplished, radiative surface comes in play. We need the heat to be removed as fast as possible from the source. More metal doesn't necessarily mean more heat will be removed. If the mass is too large and surface insufficient the heat will actually be accumulated inside the metal with nowhere to escape, raising the core temperature until material gets saturated. That's why I don't like large heatsinks since they actually need to heat up to certain point (get saturated) before they start radiating excess heat. Another thing to keep in mind is black heatsinks radiate heat better than silver ones and god forbid polished ones (excep contact surfaces). That's another reason I like to stick to just box heatsinks.

Applying thermal compound:

I like to apply a blob, spread it over the surface than scrape off the excess so the layer is as thin as possible. The idea of using thermal compounds is to fill the gaps between two materials where surfaces aren't perfectly flat so the air (really good insulator) isn't being trapped between. Then I put my COB on that layer, pres firmly and slide around so the excess is beeing pushed out on the sides. That photo i posted is a photo of thermal adhesive which is two part. So I apply small blob of each part, mix it on the surface and spread it then do the press, slide, rotate thing with the COB. The problem is that shit gets thick fast and you only have about two to three minutes until it sticks.

seaslug · Jul 22, 2015

I finally recovered the name of this early water cooled light--Theoreme Innovation SmartBar

https://www.thcfarmer.com/community/threads/led-growing-with-the-ti-smart.1839/

A slow cure thermal epoxy:

Anyone Thought Of A Liquid Cooled Led?

Anyone Thought Of A Liquid Cooled Led?

Junk

Toaster79

G gnome

Toaster79

coloradoBTC

Toaster79

coloradoBTC

str8smokn

Junk

Toaster79

Toaster79

G gnome

Junk

Toaster79

Junk

Toaster79

Junk

Junk

Toaster79

seaslug

Privacy & Transparency

Privacy & Transparency