Peltier cooled high power LED

I have recently aquired some small peltier junction chips, 1-5v, 4-6w. One of these thermoelectric devices would easily fit onto the metal PCB of a luxeon or similar LED, possibly two of them.

Problem:
As you run a 1 watt LED at higher and higher wattage, more and more of the electrical input would be turned into heat rather than light tending to damage the LED.

Calculation:
(Note, this calc is not based on real numbers, just a hypothetical situation.) Say your 1 watt LED is being driven at 5 watts, and has the low efficiency of 40%. This means you are getting 2 watts of light and 3 watts of heat (negating electrical resistance of the connections etc).

Solutions:
1. Use a higher wattage LED. However, this isn't always possible.
2. Get a bigger heatsink. Also, not always possible.
3. Use Peltier cooling. This requires a substantial increase in electrical energy, a bigger heatsink and potentially a lot more circuitry. However, for "wow" lights or other unusual applications I can think of, this is potentially a good idea.

Comments:

A patent was recently awarded for a peltier cooled lighting system, complete with temperature sensors etc (The PDF has diagrams). However, this is for stage lighting, not flashlights, which makes sense.

I will play around with this idea today, but would love to hear feedback on this idea.

I bought a couple higher power peltier's.. (12W).. for doing exactly what you describe in a flashlight.. i want to have a 35-50W flashlight and though i know i'll need a fan to cool it.. i've considered using a combination of fan and peltier.. the model i have is only like 2" square and can freeze water on one side and boil on the other at 12W.. i would not need to run it that hard to be useful.

since LED are brighter when cooler.. i can handle if the hot side of the peltier is over 200F if it means my heatsink can be closer to 100F or less..

The trick is how to dump the extra heat! in my case i have it solved because i have a big internal airspace that is vented to the outside.. in most flashlights this would be a problematic thing.

-awr

I have run the math on this several times, because it seems like a good idea. In the end, it usually does not work out for battery applications. Uou gain more light from running either more LEDs or more power into the LED, then you gain from running the LED and power into the cooler. There Peltiers use a substantial amount of power to move the heat.

It might make sense for a demonstration project where the ultimate goal was to get as much light as possible from a single LED vs max light per watt.

Peltier modules are very inefficient. To remove 1 watt probably takes 2 watts of power to the Peltier module. So now your run time is 1/3 of what it originally was. And you need to cool the hot side of the Peltier or it will not be able to keep the cold side cold. So your heatsinking requirements just went up.

There is a company (coolchips.com) that is working on an improved cooler. Quote "Cool Chips use thermotunnel technology to deliver up to a projected 55% of the maximum (Carnot) theoretical efficiency for heat pumps. Conventional refrigerators operate at up to 45% efficiency and current thermoelectric systems (Peltier Effect) operate at 5-8% efficiency".

If they ever get it commercialized then cooling LEDs might be worthwhile.

Greg

This thread has some comments re: use of Peltier devices in lights.

If you have everything you need, I think it is worth persuing, if only for the experience, but unless the application is pretty extreme (such as AWs 35-50W goal) I can't see them having a useful application in modest lights. I mean, if you want 1W output with more effeciency and less heat, it would be easier to just underdrive a 3W LED rather than to overdrive a 1W and have a Peltier.

And the 5W LEDs don't seem to have a stunning record of reliability when driven at 5W.

This article http://www.overclockers.com/tips45/ has some good stuff if you want to calculate the temperature difference and Peltier power.

Greg

Thanks for all the great responses so far!

One very well done walkthrough of a peltier-cooled beer cooler/tea warmer is here. This guy goes to extraordinary length building and documenting his creation. He also supplies some basic calculations, although he assumes the peltier is 100% efficient, which is obviously not the case, as greg_in_canada pointed out.


My purposes are not to produce a better flashlight, just to experiment with a novel method for fun.

More updates to come.

I came to similar conclusions but noted that with my megasonic i am able to get heat sink temps into the 140F range w/o much difficulty, and have a lot of battery overhead (running 12W light from 60WH battery).. it could be worth the effort.. a fan is far more efficient, but in the case of jamming 35W of emitters in the light.. a fan might not get enough heat out of the sink.. and i don't mind if the fan is blowing 200F air out if it means i can keep the heat sink 20-30F cooler.

-awr

20% seems to be the maximum efficency of such a device and apparently there is a "sweet spot" temperature. Anything above or below this gives lower performance.

whoa on the 20% figure that does present a big problem for making this a useful venture.. that's not a bit deal if you are talking about a one-time event like cooling a cup of water... but if you want to draw 20-30W out of a heatsink means 100-150W of energy to do so.. and well.. i think the peltier i have is like 12W meaning i can expect 2-3W of heat transfer.. which well that will get quite warm / cold only if it doesn't have much heat transfer in/out of it.

regardless, i will definitely test out the modules i have.. if for nothing else i can use them on the bench to keep my test LEDs cold, even if the other side is flaming hot.

-awr

What if you were to take a high performing luxeon, say a Ubin luxIII or W bin LuxV, and have a VERY large reflector, and mount it to a peltier. Crank up the power as much as you can without melting the bond wires (is it 2amps, or 2.5?). You could have a "lantern" design with a handle and free air to the peltier or with small fan too. Yes, runtime would suffer a lot, but you'd have perhaps the worlds farthest throwing LED setup.

Actually, maybe a tighter spot could be achieved via NewBie's asphyrical lens setup. It would be all custom so why not?

It may not be practical to have peltier LED flashlight, but a Peltier LED spotlight? Hmmmm!! Throw in a SLA with a buck circuit for the LED and you're set!!

that's brilliant actually. I think it is primarily heat that causes problems with too much current.. with 12w devoted to cooling i could probably keep the junction temp close to room temperature even at 1.5 to 2A on the emitter.. and the host i'm thinking i'd put it in is a 28V 3AH battery solution.. thanks for that idea very cool.. well maybe if i can get another of my megasonic lanterns that has a much bigger reflector so that might be a better host.

Peltier devices are horribly inefficient energy transporters. In addition, you have grossly overestimated the efficiency of the LED. For purposes of estimating heat produced by Luxeon, start by assuming the efficiency is ZERO. 1 watt is 500+ lumens (varies according to wavelenght), so even at 50 lumens per watt, a Luxeon would be 10% efficient.

How bad are Peltier devices? Most mechanical refrigerators have a coefficient of performance in the range of 2-3. a COP implies that the use of 1 joule will transport 3 joules of heat.

An Air condition with an SEER of 10 has a COP of 3. My experience with Peltier devices is that by the time you provide forced air cooling for the hot end, you are lucky if you can get a COP of .3 . In other words it will take 3 joules to move 1 joule of heat with a Peltier device.

So if you drive a Luxeon(s) with 5 watts, it will for all practical purposes you will dissipate 5 watts.
A Peltier device then consume another 15 watts getting rid of the 5 watts, so you need 20 watts to drive the Luxeon at 5 watts.

The only real electronic requirement on a Peltier device is protection. If the cooling on the hot end fails, and you don't cut off the power, most will fry in a lot less than 1 minute.

Peltier devices are attractive if you are severely constrained on space, but have lots of power available. Otherwise, you are much better off to consider using a heat pipe to move the heat. Even a small heat pipe (6mm/.25 inch) can move 100 watts. a Heat pipe is basically a tube filled capillaries that contain a working fluid such as Ammonia, some form of Freon, or even water.
At one end, as it heats, the liquid absorbs the latent heat, and vaporizes. The Vapor then diffuses to the cold end, where it gives up the latent heat, and liquifies. The Liquid then returns to the hot end either via gravity or Capillary action.

It works because latent heat is very high.
For example 100 calories will convert 1 gram of water a 32F to 1 gram of water at 212F. But it takes about 270 calories to convert 1 gram of water at 212F to 1 gram of steam at 212F!

A heat pipe allows you to efficiently transmit heat from a place that you don't want it to a place where you can get rid of it.

As an example I have two portable electric refrigerators. One is an Igloo Peltier device. I have measure it with about 55 watts in, pumping about 50btu/hour. It doesn't have a thermostat because it is simply incapable of freezing anything.

It is just about impossible to get colder than about 35F below ambient.

It cost about $50, and weighs about 8 pounds. It works far better as heater than cooler. As a heater, for every 3 joules (watt seconds) that you drive it with, you get 4 joules of heat.

The other unit is an Engle/Norcold mechanical refrigeraor with a novel linear compressor design. It draws about 45 watts.
It is half the size of the Igloo unit, but weighs almost 30 pounds and cost close to $500. It has a thermostat, because it can easily achieve temperatures of -20C/0F with an ambient temperature of nearly 100F in about 20 minutes, and it can easily make ice. They are common used in remote areas because they are exceptionally tough, and can reliably keep temperatures well below freezing even with ambient temperatures well above 100F.

Peltier cooling for a portable device is frightningly inefficient...

Heat pipes.. very interesting thought.. i remember seeing something not too long ago about using one of those.. that is a faaaar more efficient idea.. i just happen to have..

1) extremely over-abundant power plant
2) host to put it in.
3) small 12W peltier.

It's obviously a experiment of principle not practical.. but with the V.28 battery pack.. it's an 86W pack and if i run an over-driven K2 emitter say at 2A... maybe i can keep it cool enough that the Tj is at only 10-20C over ambient even if it costs me 12+W in cooling, that would be a worthy experiment to me.

The heat pipe idea is extremely interesting though.. i saw a thread not too long ago showing them.. have any information about where some are available to actually purchase that would be small enough to incorproate into a light.. and what requirements are actually involved to keep the circular cooling cycle to go w/o a pump.

-awr

Old Story...

Funded by the Defense Advanced Research Projects Agency, RTI engineers concocted a heat pump using two atomically precise superlattices of bismuth telluride, one layered with antimony telluride, and one with bismuth telluride selenide. The result: a device that conducts electricity and insulates against heat transfer better than anything ever seen. Way better - it's 2.5 times more efficient and 23,000 times faster than the stuff it replaces.

More info on peltier developments:
http://www.nanoforum.org/dateien/temp/ER April 2004 total.pdf?20042005174151

NewBie said:

Old Story...

Funded by the Defense Advanced Research Projects Agency, RTI engineers concocted a heat pump using two atomically precise superlattices of bismuth telluride, one layered with antimony telluride, and one with bismuth telluride selenide. The result: a device that conducts electricity and insulates against heat transfer better than anything ever seen. Way better - it's 2.5 times more efficient and 23,000 times faster than the stuff it replaces.

Click to expand...

They don't say what it replaces but only claim 25% efficency - which is very good for such devices.
Here's a link that is not a pdf.

andrewwynn said:

Heat pipes.. very interesting thought.. i remember seeing something not too long ago about using one of those.. that is a faaaar more efficient idea.. i just happen to have..

Click to expand...

Heat pipes are interesting and fairly old. Most rely on gravity to circulate a liquid that runs in a sealed pipe and might not work well with flashlights. They also add a good bit of mass and volume.

I suspect the immediate future will see better designs that use everything as a heat sink. A good metal reflector would work and designs that dunp more heat into the body so that our hands can offer a better exit for the heat.
The near future can solve the problem with more efficent LED's...

Well, solve it so the general public can have really good lights at reasonable costs.

Unless 100% effeciency is achieved we will always be building things that need to get rid of heat.

Hi Awr,

If you go with the peltier idea, there is something else you should keep in mind. When you turn off the LED, there is still heat which needs to be removed. So you will have to have some way to keep the peltier and fan running for a little while after you turn off the LED.

In the same vein, you should start the cooler before turning on the LED to ensure that it cools the LED in time. Since I'm assuming that the peltier and fan are doing all of the actual cooling, and there will be little to no passive heatsinking, it's a good idea to get the cooling system running first. Otherwise, the LED could be generating significant amounts of heat before the peltier and fan combo can remove it. In startup condition, the amount of lag time is probably on the order of less than a second, but it's better to be safe than sorry.

Look at the CPU cooling world for some tips here. They always fire up their active coolers before the CPU and always have a cool down period after the CPU is switched off.

pb

This thread made me remember this old thread:
http://candlepowerforums.com/vb/showthread.php?t=33596

-Ross

Andrew: glad you like it. Now it's up to you to build the worlds farthest throwing grossly overdriven portable luxeon spotlight!! :nana:

Rossitron: ah, cool you found that thread. I was thinking of it when I saw this thread.