Minimalistic multichip DIY LED build

[DrawnToBeaches]

Searching ebay I found a gem I haven't seen used, and many like it. Tempting considering how well the limias work. http://vi.raptor.ebaydesc.com/ws/eB...8729&category=66954&pm=1&ds=0&t=1447385016694

Too bad powering something.like that is quite annoying

 

[DrawnToBeaches]

Only fair, my friend, since most of my knowledge was gleaned from this thread! I'll post pics of a few sps that I just introduced this morning, shortly.

How are the lumia 5.1s working out for you? How is the look of it? Is anything washed out? I'm looking for a full sunlight look with a hint of violet and royal blue. Lumias are temping but ebay has some nice Cree multichips with a nice mix of colors.
This one just for example, little too big for a single led for me.
http://vi.raptor.ebaydesc.com/ws/eB...8729&category=66954&pm=1&ds=0&t=1447385016694

 

[JACOXVIII]

Ok this is a old thread but a good one. So I jumped in on this before the Dream chips cam out. looking to convert my old set up to a 5 Ch chip. So if I'm reading this right my MeanWell HLG-185H-36B is 18-36v & 5.2 A around 186 W
All I would need is the LDD in 700ma to 1000ma
do I need to worrier about CH 4 27-29V

and I want to run something like this
Channel 1: EPISTAR ES-CADBV45P 45*45mil 4500K x 20 chips (32-34V @ 700mA-1000mA)

Channel 2 : EPISTAR ES-CADBV45P 45*45mil 455nm Royal Blue x 20 chips (32-34V @ 700mA-1000mA)

Channel 3 : EPILEDS EP-U4545K-A3 45*45mil 430nm Violet(UV) x 20 chips (32-34V @ 700mA-1000mA)

Channel 4 : EPILEDS BH-R4242D-A1 42*42mil 660nm x 10 chips + Epistar ES--CADBV45P 490nm x 10 chips (27-29V @ 700mA-1000mA)

Channel 5 : EPILEDS EP-U4545K-A3 45*45mil 410nm X 20 chips (32-34V @ 700mA-1000mA)

 

[herring_fish]

I was looking for something else on youtube and ran across this video on mineral oil cooling for a 100 watt LED with no external fan so it would be quiet.

<iframe width="600" height="400" src="https://www.youtube.com/embed/oNIu8qwhE7I?rel=0" frameborder="0" allowfullscreen></iframe>

I apologize for not keeping up with this thread for a long time so this post may be out of line. I did a quick search and didn't find much other than an entire computer being cools this way and it was rather involved/complicated. http://www.youtube.com/watch?v=XGcW7Ii_BKo It did use a small simple fluid pump to run the oil to an external heat exchanger that could be useful.

This is just a simple demo for cooling an LED. Has anyone discussed this and found its flaws? Rather than wait for the down sides of this from more experienced people on the thread, I will just lay out how this could be used. There must be a good reason that this has never been done before.

It seems that you could have localized heat sink cooling or have a local internal fan for each light but daisy chain more than one LED together and pump the oil to a bigger remote heat sink and fan. It sounds like it would be very quiet. Having oil in front of the phosphors would mess with refraction a little but I want to do something different with the disbursement anyway. In the "Realistic" Multi-Chip LED Build thread post #26 there was a lens adapter with side slots for air or oil to circulate through.

 

[herring_fish]

Talk about quiet.

<iframe width="800" height="600" src="https://www.youtube.com/embed/rk2Pfy45WmY?rel=0" frameborder="0" allowfullscreen></iframe>

 

[karimwassef]

Interesting but why not just do conduction cooled liquid (water) in aluminum tubing?

The oil just creates a mass to absorb the heat. The liquid cooled can expel it too.

 

[herring_fish]

I really don't know much about this stuff. That is why I am asking about this and whether this is an anomaly that would not work for a reef tank.

What I like about it is that oil does not appear to conduct electricity, you can allow the entire chip to be exposed to fluid on both sides and fluid carries off heat far better than air.

To me, this means that while you might want a small fan to move the fluid around, there would be no need for and external fan so it would be cool and quiet.

 

[karimwassef]

Heat needs to flow away to reduce the temperature. The oil acts like a giant capacitor for heat by day and then cools down overnight. This works with certain photoperiods, room temps, power levels, etc...

The fan approach is like this:

LED -> heatsink / fan-> forced air

So the heat goes from the LED to the exhaust forced air. It is limited by the ability of air to take on heat and the thermal transfer efficiency of the heatsink/fan to the air. Air is a poor thermal conductor. The downside is the fan sound and the relatively weak thermal performance (higher temp, lower PAR).

Here's how liquid cooling works:

LED -> water in Al pipe -> chiller/heat exchanger/fan -> circulating pump

This allows the chiller to be physically distant from the LEDs. So, like an air conditioner, the heat exchange element can be place outside and can use much more powerful fans/larger exchangers. Water is a substantially better conductor of heat (24x) and the cooling solution can be much larger & more efficient. This reduces noise, size, and improved efficiency and therefore PAR. This is what I do.

The oil looks like this:

LED -> mineral oil stores heat (slows temp rise) ..... At night, the oil slowly cools down.

This would not work for high photoperiod operation. It needs a balance of off-time and cooler air temp to relieve the heat captured. Also, the temp is not uniform. It's coolest when first on, and heats up until off. PAR would drift as temperature rises. Also - taken from overclockers forums - Mineral oil has a specific heat of 1.67 kj/kg°k and a thermal conductivity of .133 w/m°k. That's better than air, but it can't compete with water at 4.19 kj/kg°k and .58 w/m°k. http://www.overclockers.com/forums/showthread.php/520492-Mineral-oil-instead-of-water

So... There's many ways to skin a cat but they're not all equal.

One note - water cooling is best with DI water but that eventually would corrode the aluminum frame - just a maintenance item that needs to be addressed. If you're willing to accept slightly poorer performance, you can add corrosion resistance compounds.

 

[karimwassef]

In general, if you really want to learn how to really cool chips (LED is just a chip), check out the DIY overclocker forums. These guys spend a ton of time testing and sharing how to best cool high power output chips.

 

[herring_fish]

Thank you very much for adding your informed feedback.

In post #4408
Now I done want to put words in your mouth but it sounds like we need to know what the temperature at the end of the day (photo period) is as well as at the beginning to see if oil can absorb enough heat to last throughout the day and if enough heat dissipates (through the glass?) during the night.

In post #4407
Now does this video look like heat is properly transferred to the heat sink via mobilized oil?
OR am just I barking up the wrong tree in your opinion.

 

[karimwassef]

Both videos are informative. The first one doesn't show the steady state operation reaching thermal equilibrium. While the oil is good at pulling heat from the LED, the oil's ability to transfer heat through the acrylic box to the metal heatsink fins is quite poor.

Basically, he's trying to emulate the conduction cooled example but without a circulating loop because oil is hard to pump, so he puts it in container with a small fan to get some motion. If you wanted a large sink for the heat, you could bolt the LED (with thermal grease) to a massive block of copper. The larger the copper, the more heat it will absorb. The larger it's surface area, the more it will release into the air. The more air you can force against that surface area, the more heat it releases and the cooler it gets.

Basically, metals are better at moving heat than water, water is better than oil, and oil is better than water.

The end point of cooling is "rejection" into the outside air. The problem is that LEDs are very small and hot point sources of heat, but the best heat exchangers are large with big surface areas and strong fans.... Everything in between is a way to bridge from the point source of the heat to the efficient rejection (or expulsion) of heat to the air far away from the source.

If you just move the heat to the air around the LED, you still need to cool that air with an air conditioner, etc... If you can move the heat far away, preferably to the outside, then you get the most efficiency out of your cooling system.

 

[herring_fish]

Thank you for indulging my here.

In post #4407

You may be correct in that he is trying to transfer heat through the acrylic but I originally assumed that he made a hole in the acrylic wall and covered it with the metal heat sink so that oil would get in direct contact with the metal. If not, that is a bad thing for sure. He did talk about the chip coming to equilibrium which I also assumed was alluding to a steady state condition.

Let us set aside what he really did and design our own theoretical box that does have oil directly being exposed to the entire flat side of the heat sink and the fins exposed to the air. It seems that oil being exposed to the chips on both sides and then being blown to the heat sink with a fan, would be better than having only one side being exposed to the metal through the thermal grease and the other side of the chip being only exposed to the air"¦and in many cases this air cannot circulate with the ambient air that surrounds the enclosure.

Wild being more complicated, would it cool better?

 

[karimwassef]

IMO No. The thermal conductivity of metal and thermal grease is substantially higher than mineral oil.

The way heavy computing chips are liquid cooled involves connecting the chip (like our LED) to a copper plate that has cold water flowing into it. The warm water flows back out in a loop to a heat exchanger and then a reservoir with a pump to push the circulation.

This is a good primer of AIO liquid cooling (All-In-One)
http://www.anandtech.com/show/7738/closed-loop-aio-liquid-coolers

and shows the basic components of how this is done.

There are overclockers who play with the concept of "immersion" cooling, which is what the idea of mineral oil is really about. This is generally because it is cheaper and some consider it cooler (more hip? not lower temperature).

The origins of immersion cooling date back to old electric ballasts that were fully encapsulated and immersed in thick oil/tar inside of a metal box. The idea was for the tar/oil to absorb the heat away from the hot electrical components and then the metal box can relinquish that heat to the surrounding air, or larger metal frame (if it's bolted to one).

 

[karimwassef]

since youtube seems to be a good communication forum - try these. I don't like linking "marketing" from manufacturers but corsair is pretty established and their videos are pretty educational.

<iframe width="560" height="315" src="https://www.youtube.com/embed/uQ1p9QbZ9tc" frameborder="0" allowfullscreen></iframe>

<iframe width="560" height="315" src="https://www.youtube.com/embed/bFstDuftJow" frameborder="0" allowfullscreen></iframe>

<iframe width="560" height="315" src="https://www.youtube.com/embed/vG3W6yvqTuw" frameborder="0" allowfullscreen></iframe>

 

[karimwassef]

and because every DIY has its "Extremes"....

<iframe width="560" height="315" src="https://www.youtube.com/embed/QmSBaizEqkk" frameborder="0" allowfullscreen></iframe>

and for the other side, here is immersion cooling...

<iframe width="560" height="315" src="https://www.youtube.com/embed/XGcW7Ii_BKo" frameborder="0" allowfullscreen></iframe>

<iframe width="420" height="315" src="https://www.youtube.com/embed/Y1WkdXvtEKY" frameborder="0" allowfullscreen></iframe>

and in case you're wondering about cooling using a fridge... nope...

<iframe width="560" height="315" src="https://www.youtube.com/embed/B8bhGw4vUFE" frameborder="0" allowfullscreen></iframe>

The world is a fun crazy place if you're DIY friendly

 

[karimwassef]

I am very tempted to post my build... but it's just not prime time ready yet... but this thread is making it hard to resist.... so I'm sharing what others have done in the general field.

 

[herring_fish]

Those were very good videos. I had looked around for something like that but didn't really know what to look for. I watched the first one twice. Thank you.

I noticed that while rare, they can leak. One reference was to a water based coolant but that green stuff didn't look like my corals would tolerate it. I will start doing some research on that stuff.

In posting those, are you kind of not liking the two sided oil idea?

 

[herring_fish]

I made my last post out of order. I didn't see your lick.

Yes I had seen those when I was doing my searches, including listing the second one in post #4407. But is the theoretical one a bad idea. The computer over clock cooling requires that the CPU chip be affix to the mother board. The COB chip just has two wires going to it so it can be suspended in mid oil but there may be something that I am overlooking.

PLEASE post some of you build "¦even if it is not ready yet.

 

[karimwassef]

It doesn't have anything to do with the number of connections. It has everything to do with thermal conductivity and impedance.

Copper > Aluminum > Thermal Paste > Water > Oil > Air

So, copper, aluminum and thermal paste are the highest conductivity by orders of magnitude. That's why that is the preferred cooling mechanism for very hot chips.

The second big variable is surface area. The larger the surface area, the more heat can transfer. So, using the base of the LED chip (where all the heat is) and bolting it against the cold pipe is optimal.

The third is distance. The smallest distance that the heat has to travel, the more heat can transfer. This is why the thermal paste between the LED and the cold pipe has to be very very thin.

In contrast, you have a thick layer of poorly conductive oil... even though it covers the whole LED, it's still inferior.

 

[karimwassef]

As far the funky green liquids - it's all DI water with some fluorescent dye so it glows under UV ... the green color is aesthetic only. If there is a leak, the concern would be corrosion (pitting) of the Aluminum and the DI water with Al dripping into the tank. It's not that big of a deal if the % is low enough. That's also why I'm not using Copper, even though that is a better pipe material.

Some of those colored waters have some anti-fouling and anti-corrosion chemicals. That's why I just use pure DI water and that has the absolute best thermal conductivity.