Heat sink area for 10 watt LED?

pretty large. a TO220 package has about 70K/W and are about 200mm^2 and you want 2K/W so atleast 9000mm^2 - about the area of both sides of a credit-card but only if you use really thick copper. I'd start with a piece of

Massive.

I've bought quite a few of those LEDs on eBay and they get HOT. So hot in fact that I usually run them at half the stated power without active cooling.

(Output is a curve so they put out about 70% of full light at half power).

Something like this heatsink is the minimum you'll need for that LED:

Even so you'll probably need a fan to keep it below 60c at that30c ambient (assuming you run it at 10W).

Copper PC board? Nah.

I'm guessing that an infinite sheet of 1 oz, single-sided copperclad FR4 wouldn't be enough. The lateral heat conductivity of 1 oz copperclad is about 70 K/w per square.

Wow! That bad? Good number to know, though I now have a peck basket full of heatsinks from the JVC projection set tear-down...

Cooling in air requires a large area, fins, etc., because of the low heat transfer coefficient of air, ~ 10 watts/m^2-C.

For some stationary applications instead of going for a huge surface area it may be useful to use gas - particle bed fluidization with shorter fins.

This would get the heat transfer coefficient up to 200 w/m^2-C.

The fan or blower might need to output a somewhat higher pressure but the column shouldn't be very high in most electronics applications so a few kPa would do.

The optimum particle dia. for maximum heat transfer is about 10 microns -- dust -- but larger sizes will also work.

Fluidized particle bed cloumns are easy home brew.

On the down side

1. this requires gravity so the orientation must be pretty much upright.

1. it would need to be visible or monitored somehow to make sure it was working properly.

Bret Cahill

e.

You're right. I found a large aluminum sink that keeps the sink side of the LED barely warm to the touch at half power. But the lighted face of the LED gets very hot. I can only touch it for a couple seconds.

Any idea how efficient these LEDs are in terms of light output verses heat?

-Bill

I accidentally dropped a piece of paper on one once and it burned right through it in seconds.

You don't want to touch the front with your finger...

It's a curve that drops off as you increase the current going through them. That's why I don't try to run them at full power - those last few hundred milliamps mostly get converted to heat, not light.

That number makes no sense.

A small fan directed properly at a small heatsink gets down to 3 K/w easily, fine to sink a 10 watt LED.

Crazy impractical.

Great! Let's fill our electronic racks with dust. We could reverse our dust filters and use them to keep the dust *in*

The heat transfer coefficient is more or less proportional to the conductivity of the air / thickness of laminar boundary layer.

You are comfortable [sustainable] sleeping at 22 degrees C with no blanket. Your 2m^2 skin area has an average temp of 30 degrees C.

160 watts is about the heat a human needs to sink sleeping.

With a fan the boundary layer gets thinner and heat transfer coefficient increases but then the delta T would decrease and you could save some money on the AC.

The heat sink in the photo looked like it was about 17 cm in diameter.

The "distributor" basically an air hockey type plenum, would need more of a pressure drop than an axial cooling fan but a centrifugal blower should be enough.

I made a column using glass micro spheres in a 7 cm dia. glass tube, about 45 cm high. The air was from a shop compressor regulated way down. Glass bubbles don't transfer much heat but the packed bed increased in height by 5X and behaves almost like a white liquid, i.e., ripples and waves.

There's no reason it couldn't scale down to 1 cm diameter items.

Three phase gas/solid/liquid fluidization has the all time highest heat transfer coefficient for cooling, millions of watts/m^2-C -- comparable to the magnitude of the coefficients in steam burn condensation.

There will be some limits to the scalability because optimum diameter of the solid particles for maximum cooling is about 1 cm.

Are there any base ball sized electronic devices that would become possible with megawatts of cooling? This might be what is necessary to get that circuit to last more than a few milliseconds.

Bret Cahill

Heat pipes have been used successfully for many years without chunks in them, they just use a liquid-gas phase change fluid (eg: freon, ether or propane) the fluid boils at the hot and the vapour rises into the radiator where it condenses,

Water cooling is common with very high power densities, too.

or

A place in Tampa made water cooled transformers for Disney World light displays. Instead of wire coils they coiled up 1" - 2" copper tubing and pumped coolant inside of the [electrical] conductor.

Ice + air + water would definitely make those puppies cooler although I haven't done or heard anything about 3 phase _internal_ flow.

Bret Cahill

or

What's the maximum heat transfer coefficient?

Bret Cahill

Big deal. Water cooled tubing is a very old technology. Take a look at the 'cooling pond' outside trhe old 500 KW WLW transmitter site, then look at the date it was built.

Mainframes were water cooled in the '70s - '90s. The Hudson River was IBM P'ok's cooling pond. CMOS made that unnecessary around the mid '90s, though it's probably still used to cool the air in the buildings.

I saw a new Unisys system being scrapped that was water cooled. It was a prototype for a navigation system for a nuclear sub. It was in a round rack that was made to fit through the hatch and had a large eyebolt on the top of the aluminum rack. I tried to buy it for the scrap price, to put it in my shop but they insisted on destroying it.

or

The heat transfer coefficient of evaporation or boiling water is going to be less than half of that of condensation of steam -- about the

-- and it drops further with organic fluids.

Heat transfer is supposed to be the happen' field in mechanical engineering yet you can search the ME stacks at a good university library all day long and maybe find one footnote on fluidized particle bed heat transfer. Multi-phase is exclusive to process engineering where an inelegant empiricism rules.

To get an idea of just how much we still live in the dark ages consider a book published about 10 years ago, _Rheology of Foams_ Heller, et. al., said that the petroleum industry tried to get viscosities of a lot of foams [the fluffy gas-liquid mixtures] but almost all the numbers were bad if not useless.

Part of the hold up is the cloistered nature of chemical engineering where the oil bidness is about as open and transparent as the mafia. Part of the problem might be a legacy of the cold war where heat transfer was mistakenly singled out as a strategic field when it isn't any different than anything else in engineering as far as national security is concerned.

In any event, it's still the wild west in multiphase. If anyone ever wants to do a paper requiring lots of lab work, just email.

Bret Cahill

Just had a look around ebay and I see this 100 watt RGB LED selling for around $80. Probably needs a fan to keep it cool?

-Bill

Nice!

At that power level you'll need something like a mid-range PC CPU cooler (with heatpipes and stuff) to keep it cool.