Heatsinking a 3W Luxeon

How is your math?

OK, lets start: Your 3w rise the temp of the led, without forced cooling, to a level that it fails? So your heat sinking arrangement has to remove heat to surrounding air faster than it is generated to keep the temperature in working range. "L", if you leave it in your car in "Ozzie, Ozzie, oy, oy, oy!" summer day the air can be 100 deg C. On the other hand if you use it in air conditioned comfy room then the air is 23-27 deg C which makes a lot of difference. The thickness of the plate is more important on its inertia to sudden changes in temp and the cooling effect is determined by the area cooled by the moving air and quantity of air moving.

Have fun

Stanislaw Slack user from Ulladulla.

You didn't say the more important information: will be the flashing led working continuous (say 24h/day) or not ? Continuous 3W can't be dissipated from a closed plastic enclosure, even you have an aluminium heatsink inside, only if you let the hot air flowing somehow from the inside to the outside, or using a metal enclosure.

greet> I want to use a single 3W Luxeon Star LED. but this is primarily a

Hi, Jason -

If I was doing this, I would look up the max junction temperature (or case temperature) and subtract that from the max expected ambient temperature. Then divide by the 3 Watts. This gives you the maximum thermal resistance you can tolerate from junction (case) to ambient. If you're working with junction temp, subtract the junction-to-case thermal resistance, then subtract the case-to-sink thermal resistance. You now have the required thermal resistance of the heatsink.

For example, suppose you can tolerate a junction temperature of 100C. Further, suppose the total thermal resistance from junction to sink is 5C/W and your max ambient temperature is 40C. Therefore, you need about ((100-40)/3)-5 or about 15C/W.

I am probably wrong about the thermal resistance of a flat aluminum plate, but I will use the value of 100C/W for one square inch of exposed plate (one side). The value depends on the plate's orientation anyway. So, you might need about 100/15 or about 7 square inches of exposed sink.

So, your lid is just about what you need. Or it may be considerably less than what you need. Put your LED (or a transistor or some other device) on the lid and measure the temperature rise. Be experimental.

For some (better, perhaps) information on this subject, see:

Not your case exactly, but it might be educational. Also, Lumileds may have application notes on heatsinking of their devices.

This is a group composed of professional engineers. If I have misled you, then we (you and I) will both be gratified by my being corrected. Perhaps your question would have been more appropriate on sci.electronics.basics.

Good luck.

John

That's a little thin for good heat spreading; use something thicker if you can. Or put a heat spreader (a poker-chip-sized at least piece of copper or aluminum) directly under the led, clamp-epoxied to the thinner metal.

Sounds good. If you screw the c-folds to the sides of the plastic box, heat transfer will be a little better... squash out the air gaps.

Better yet, put the c-folds on the *outside* of the box. The better you heatsink an led, the brighter it will be and the longer it will last.

But 3 watts really isn't a lot of heat.

I wouldn't drill holes in the aluminum. It would just impede heat spreading and reduce the effective air-transfer surface area. Holes in the plastic box would help a tiny bit. Disappearing the plastic box would help a lot.

That helps.

Yup. Try it, and back off the current if it seems too hot, but it will probably be OK.

John

It will be steady for a few seconds, then flashing for a few seconds (cycling). Will be running not more than say 5 hours at a time.

The aluminium heatsink is outside the enclosure, not inside. The plastic enclosure is ventilated as well.

Jason.

I won't be using it in the car (lol) so don't worry about that! It will not be used in temperatures beyond, say 25 degrees C. Most likely in cooler temps (22 and below). So what you're saying is the thicker the better? That shouldn't be too much of a problem. You see, I got nowhere to put a fan, that's the reason why I'm posting this!

Thanks Jason.

God forbid, never mentioned anything like this, we are on a group that has _design_ in its name, so I treat it as such. For every solution there is some calculated result and you can use this as a minimum or add as much safety factor as you wish, but know what you are doing.

Example: look at any manufactured heat sink, be it on your CPU or the one surrounding the piston on your "Harley Davidson". You will notice that a lot of material is removed to create larger surface for more efficient transfer of heat to air.

But the air has to flow freely, no magic about it.

No need of forced air movement, enough surface of the heat sink in air and the warmer air will move up by itself to be replaced by the cooler air from the surroundings.

Have fun

Stanislaw Slack user from Ulladulla.

OK. I had a deep thought about it all, and came up with a major design change that *can* include a fan... I previously thought it was impossible (too hard to explain). Anyway I got this old "486" CPU fan and heatsink combo (about 400 x 400mm) that I could attach directly onto the luxeon. Remember my casing is well ventilated to allow single direction air movement. Would this do it?? I'm thinking that if this heatsink and fan actually cooled down an old 486 "CPU", why couldn't it cool down a 3W luxeon? Even better, could I use one of those heatsinks designed specifically for a LUXEON (stick-on pin grid) and use my CPU fan on that?

Jason.

Can any heatsink protrude from the top of your box? i.e can you mount an additional heatsink on top of the aluminium lid? If so then that should work just fine. Otherwise put a larger heatsink inside your box mounted on the underside of the metal lid. Or simply get the biggest heatsink you can to fit in the box and leave the lid off. Any limit on how much your box can weigh? ;-)

Dave :)

You just have to watch the maximum ratings.If the rating was for a 25 deg. C sink, then you can never achieve this without peltier added.

greg

Hi Greg, So you couldn't tell me, off hand, if my old CPU heatsink and fan might do the trick? This is where I'm struggling... I got no idea what my old 50x50mm 486 CPU heatsink's thermal resistance is :-( It's just a spare part I had lying around with no info, with a small fan attached.

Peltier? If it comes to that, I'll ditch the luxeon idea altogether... too much trouble; and power.

Jason.

Its a good heat sink, but I would lower the drive current fo maximum longevity. These maximums are at a junction temp of 25 deg. C. These lights can be in really harsh environments like busses. With sun and heat, the junction temps could be really high. Got to derate. Figure if you could keep the star heat sink at 25 deg. C, there is another metal layer before the junction. You have two insulated layers btween the stars main sink, and the actual junstion. Placing the star on top of another heatsink just adds more differentials.

greg

On my 5 watt Luxeons, the little plate which the Led resides is only 3 mm in diameter. The metal conducting the heat is only 3 mm dia at the very beginning and 3 mm long. It ajoins a base plate of about 6 mm in diameter.

greg

OK. Lets take a stroll in "very poor neighborhood", find someplace a broken TV equipment and take a look at "non forced cooling" of power components, active and passive, You will see that your 3W power "waste" is trivial and can be done in many ways. (Scratching rusty memory) Don't remember off hand what was the power consumption of the 486 CPU but for sure it was over three watts. Last 486 that could survive Australian summer without forced cooling or heat sinking was 486DX2/80 with 40MHz bus.

Have fun

Stanislaw Slack user from Uladulla.

I can assure you my light won't be operating in ambient temperatures beyond

25 deg C ... and definately not in a hot car or bus! lol.

Sorry, I'm full of questions as usual! haha ........

Because I am still in the 'design' stage, I am thinking of slightly lowering the drive current anyway, down to 800/900mA? I suppose this would relieve a bit of pressure already right? I purchased a '1000mA BuckPuck' that has an inbuilt pot to adjust the current to a lower setting if required... still awaiting delivery of it. About the metal layers, I don't know what you're talking about because I don't have one to look at (yet), but I will question your comment once I get my hands on one. Would you recommend joining the star's heatsink to my larger 'foreign' heatsink using '25% silver thermal compound'? I read that its heat transfer is better than the normal stuff.

Jason.

Ohhh yes.... those were the days! And look what sort of cooling is needed nowadays!!! It's amazing, but it's only going to get worse right? So are you trying to say that my CPU heatsink and FAN doesn't even come close to keeping the 3W LED cool enough? Even if I lower the current drive a bit? Even if the LED will alternate every few seconds between steady on and strobing? Sorry, I might be reading you wrong... I'm just trying to grasp what you are trying to tell me =)

Jason.

If it goes like it did, it will get "worse". Since the time of the 486 the number of transistors in CPU's multiplied (in millions) and the frequency of transitions (the heating state) went from 40Mhz to 3+GHz. Then it was 150W power supply and single fan in it, today 400-500W supply with fans on CPU, wideo card, heat sinks on memory, fans on hard drives, it is amazing that the box is not flying.

For your LED a heat sink the size of CPU heat sink with forced by fan air flow will create freezing conditions, just make sure that there is good heat transfer between the components.

Have fun

Stanislaw Slack user from Ulladulla.

Thanks for your help. I will make sure there is great heat transfer between components, don't you worry =)

Regards Jason.

I have played with Luxeon stars enough, and a usual CPU heatsink should be adequate for a Luxeon III.

According to the DS46 datasheet, the thermal resistance from junction to mounting flange surface is 17 degrees C per watt for blue, green and white versions. Any old heatsink compound should add no more than one or two degrees C per watt.

At 700 mA, the maximum forwrd voltage is 4.47 volts, meaning a power dissipation of 3.13 watts. Multiply that by 19 degrees C per watt, and the junction could foreseeably get 60 degrees C hotter than the most-nearby surface of the heatsink.

The maximum junction temperature is 135 degrees C. That means you need the heatsink to get no hotter than 75 degrees C - which I find easy for a usual CPU heatsink to achieve with 3.13 watts (or more likely a little less power) being dissipated into it.

Keep in mind that white LEDs last longer at lower junction temperatures. The phosphor degrades, and does so faster with more light and with higher temperature. If you need the LED to last tens of thousands of hours, you probably want to keep the junction temperature down to around 100 degrees C or less.

One more thing - if you measure the temperature of a bare metal object with an infrared thermometer, put a piece of tape over it - bare metal has low emissivity in the thermal IR range and the temperature reading will be somewhere between that of the metal and that of surfaces whose reflections can be "seen" (even if blurred) in the metal surface. But masking tape, paint and a decent black anodizing layer will give good temperature readings.

- Don Klipstein ( snipped-for-privacy@misty.com)