how to find out what size heatsink you need???

[aljsk8]

is there any rough way of finding how much heatsink you need

as in some kind of mass per watt or surface area per watt

basicly i want to run 4 mce's its not a flashligh so im not limited by size
but i do want it to be big enough to keep the leds confortable

the leds and heatsink will be in open air and mainly with ambient temperatures below 10 degrees C

they will be run at spec (700ma per die)

any ideas?

 

[gadget_lover]

I am not a physicist, so do NOT take this as gsopel.

The heatsink needs to move heat from the source to someplace where it can be removed. If the heatsink is thermally isolated them it just soaks up heat, slowing down the time that it takes for the LED temperture to rise.

Since the slug on the LED is only a few MM across, that interface will limit the transfer of heat. I read somewhere that adding mass to a heat sink will rapidly reach the point of diminishing returns, since the forther from the heat source, the greater the accumulated reseistance to heat conduction.

In essence, I recall that somone who did know the science said the heat sink thickness should only need to be about as thick as the contact area is wide. Note that the use of a "star" does not change the fact that the LED's slug is pretty small.

Remember, I have a poor memory, so that may be wrong.

Daniel

 

[wquiles]

Interesting topic as this is one of my pet-peeves ... the size/material of the heatsink in a handheld flashlight is NOT the real problem. The problem is how to remove heat from the body of the flashlight, once the heatsink has done its job (regardless of how efficient/fast) and moved the heat away from the LED. I have done a LOT of overclocking on PC's, and that is the best possible way to learn about heatsinking, and how to remove heat, or more appropriately to exchange heat/energy.

The problem with any handheld flashlight is that you don't have a forced way or "active" way to remove heat. On a PC you have heatsinks with a fan, the fan moves the heat which increases the temp of the air inside the PC, but then the PC also has fans that circulate air in/out of the PC, and since the PC is most always the hottest thing in the room, you can transfer a LOT of heat from the PC into the room, plus the room is tipycally air conditioned, etc. - you get my point. There is a forced way to remove heat from the device (the microprocessor in this case), AND we are using a lot of energy (to drive all of those fans) to move heat away from the device.

On a flashlight, there is no "forced" method, like a fan, since that would take energy away from the batteries and give us a shorter runtime. So basically we have only convection, through air, or through our blood:
1) The air surrounding the light. Some minor improvements can come from fins, to increase the efficiency of heat transfer to the surrounding air (again a lesson learned from CPU heatsinks), but of the two methods this is still the least efficient way since we don't have a fan blowing through those fins.
2) Convection to the hand holding the light - the blood in your body acts as a cooling system removing heat from the flashlight. This is the most efficient way, for a handheld light that is not under water (like a diving light).

Both methods will remove heat up to a point, and then the temperature will keep increasing since there is only so much the air/blood can remove. Anyone who has used a powerful LED/incandescent flashlight has experienced the flashlight getting warm and eventually hot, to the point that it is no longer comfortable to hold.

Therefore, note that the problem with heat is not necessarily the heatsink used for the LED. Once you move heat from the LED to the body of the flashlight, you are done with all that you can do - the heatsink did its job. The larger the heatsink and the larger the body, the larger the body/heatsink can absorb heat (up to a point), and heatsink material size will have an effect on the speed at which heat is transfered (again lessons learned from CPU heatsinks that get MUCH hotter since they have to dissipate 75-100 watts!), but inevitably you go back to the premise of my discussion - once the heat has been transfered to the body of the light, how do you remove it from the body of the light? Only through air and blood. Of course, a diving light surrounded by water will be of course "water cooled", but those would be the exceptions as most lights here in the forum are hand-held.

So what does all of this manbo-jambo means:
- A single MC-E or P7 driven at spec in a handheld Mag body will eventually get too hot to handle, regardless of the heatsink used. It is simply physics: you have approx 3.5Volts at 2.8 Amps for approx. 10 watts of power that have to be dissipated.

- ALL LED's are rated for Lumens at a relatively very low temp, usually lab-controlled, at around 25C/77F (per the many data sheets that I have seen). That is basically room temperature. As soon as you turn ON your LED light, the output will start dropping since the temp on the heatsink will start going above room temperature.

- The temperature will keep increasing until some equilibrium point, depending on the air temperature, or how much pain you can stand while holding the hot light, etc.. - note that in the case of LED's this equilibrium point will almost always be higher than the rated temp at which the LED was giving its "rated" output. Those 700-900 lumens are not possible long term on any handheld light that is convection cooled (air/blood), since the temp of the LED will quickly go over the 25C.

- A light with any multiples of MC-E or P7, HAS TO GET EVEN HOTTER, and will get hotter much quicker than the light that has only one MC-E or P7. You are basically adding roughly 10 watts per each of these high-power LED's. The degrading on LED output happens even quicker when using multiple MC-E/P7 since the "shared" heatsink gets multiple times the heat, so the output gets lower and lower with time.

- The larger the heatsink size/size of the body, the more heat can be absorbed by the body, up to a point. This is why you can hold much longer in your hand a Mag-size P7 light than a single CR123-size light driving the same P7. If the host/heatsink size is the same, you will reach this "darn! - this thing is too hot!" point sooner with more LED's.

So, you can spend all the time you want figuring out the largest, meanest, best material for a heatsink (I hear Unobtanium is the best heatsink material!), but in the end you still have the problem of removing the head away from the body of the flashlight, and it only gets worst the more high power LED's you use, regardless of the heatsink used. You can't escape physics.

If you want to remember only one thing from all of this: the job of the heatsink is simply to transfer heat from the LED to the body of the light. That is all. Unfortunately, once the heat reaches the body of the light, the real problem begins, and this happens regardless of the heatsink, if you are using convection (air/blood) to cool the light.

OK, end of my rant.

Will

 

[Linger]

Q)How to find out?
A) experiment

run 4 mce's its not a flashligh

Gotcha.
Go with Will's comments re: heat sink large enough that it gives you a method of clearing heat from the heatsink itself.
Are you gonna consider active cooling? (little pc cpu fans opperate from 3v - 12 v, are very small, and do wonders for cooling your heatsink -- I've gotten the free when I ask computer stores for any scrap heat-sinks)

Best answer for you - experiment. Set up a test rig and see how it does. Use your final design idea, or something comperable if you don't have the real thing available, and try it out. track / post the results.

For example, I'd say a cpu heatsink, 3" x 3" x 3", solid core and finned sides, is sufficient for a single mc-e, and maybe two, but if it'll do three you have to tell me.

Best,
Linger

 

[Jonster]

Just by way of an example I've got a heat-sink (see pic below) that I want to turn into a bike light.
The surface area on this is incredible and even without the forward motion of a bike I think this would dissipate heat admirably.
I've not seen it mentioned so far, so remember that hot air rises and in that sense vertical fins be marginally better than horizontal fins.
Obviously on a bike I will have some good forward motion so that will play a large part in dissipating heat.

I'm not sure where this heat sink is from - it was lying around at work with no takers!!:thumbsup:
My best guess is that it may be from a capacitor in a big machine tool. If anyone has any ideas then maybe we could get hold of some for people to play with.

Note the ID is around 26mm.

 

[Jonster]

Will - great summary on heatsinking.

I hear Unobtanium is the best heatsink material!

I've heard of another great material.
Just ask in your local store for some Dictonite!!

Ahhem - sorry to lower the tone!

 

[wquiles]

Just by way of an example I've got a heat-sink (see pic below) that I want to turn into a bike light.
The surface area on this is incredible and even without the forward motion of a bike I think this would dissipate heat admirably.
I've not seen it mentioned so far, so remember that hot air rises and in that sense vertical fins be marginally better than horizontal fins.
Obviously on a bike I will have some good forward motion so that will play a large part in dissipating heat.

I'm not sure where this heat sink is from - it was lying around at work with no takers!!:thumbsup:
My best guess is that it may be from a capacitor in a big machine tool. If anyone has any ideas then maybe we could get hold of some for people to play with.

(snip pic)

Note the ID is around 26mm.

Man, that is an awesome looking heatsink. Lots of surface area, and as you said, on a bike it would be a great combo. Any ideas as to how you would use it? LED Driver? What LED (or LED's) you want to use?

 

[Jonster]

Man, that is an awesome looking heatsink. Lots of surface area, and as you said, on a bike it would be a great combo. Any ideas as to how you would use it? LED Driver? What LED (or LED's) you want to use?

Yeah I'm really chuffed with this little find.
I'm planning on lopping it in half and putting a Cree MC-E (4P) in each with the two lights wired in series.
Then I'll have a little battery pack with 6-qty 18650 batteries (wired 3S2P if that makes sense) and a buck driver. I've not chosen which driver yet. My preference is a hipFlex or hipCC but cost-wise I may just go for a cheapie from KD.
I'll be putting pic's up as and when I get it running - but don't hold your breath!!!

 

[wquiles]

Yeah I'm really chuffed with this little find.
I'm planning on lopping it in half and putting a Cree MC-E (4P) in each with the two lights wired in series.
Then I'll have a little battery pack with 6-qty 18650 batteries (wired 3S2P if that makes sense) and a buck driver. I've not chosen which driver yet. My preference is a hipFlex or hipCC but cost-wise I may just go for a cheapie from KD.
I'll be putting pic's up as and when I get it running - but don't hold your breath!!!

Two things to keep in mind:
- Any time you have LiIon cells in series you might/will get them unbalanced as no two cells are identical, so please make sure you have a way to recharge them individually. Also, it is much safer if those cells have built-in protection (also called "protected" cells).

- Driver-wise, having a cheap driver from KD or other "sources" usually means that you end up with lower efficiency and a higher chance of something going wrong, as the driver itself will get very warm/hot since it has to dissipate quite a bit of energy (this also robs you from runtime as you are wasting energy as heat in the driver). I have personally been using a hipCC for a while now, and I can attest to being a great driver for this specific application. In fact, in my current project (see link in my signature below) I am yet again using a hipCC to drive a warm tint MC-E from a supply composed of two 12 volt SLA batteries in parallel - I was measuring over 80% efficiency in the quick test I did last night.

I look forward to pictures from your project

Will

 

[aljsk8]

cheers for the response guys

good info wquiles athough drifting a bit from what i asked

you said...

interesting topic as this is one of my pet-peeves ... the size/material of the heatsink in a handheld flashlight is NOT the real problem.

but i never said it was a problem and i also said it wasnt a handheld flashlight.

i just want it to be sufficient, after all just like you have mensioned its got to be big enough to do the job and then beyond a certain size there is no benifit but that is still "a size" even if its not getting it as cool as we would like

im not trying to argue just get my thoughts across so sorry if ive crossed the line :grouphug:

so i guess what im asking is what is that sweet spot that max size that will give the best results within the anoying constraints we all have to live with

i was after a generic ballpark type rule some none scientific area calculation but if there isnt one then thats ok

im just askin in my aplication which is not a flashlight is a heatsink and leds in open winter (-5 to 10 degrees C) air with possable wind of 1 - 5mph, with 4 x mce at 700ma per die - then what size will give me the lowest temp achiavable using only fins and convection

oh and i cant experiment as i just want to order all the parts then build

so i really want to get it right 1st time

this is my old light and im basicly doing the same thing but upgrading it - so that will help you understand the aplication

http://www.candlepowerforums.com/vb/showthread.php?t=179661

hope all that makes sence

oh and one more thing - mce / p7 are total 10w (i think) so not all that needs to be dissapated as some of that 10w is going to make the light and not just heat

 

[bretti_kivi]

if we're talking british winters, outside, then that changes the requirements quite a lot.

Mine would change it even more - down to -20 - but yours are damp and cool. Which makes any heatsink even more effective.

You have a choice of materials: alu, copper; vertical fins will work better because of convection - heat rises. Adding horizontal channels help again for moving air around.

You don't actually *need* to limit to fins and convection, but are you intending to make it water resistant?

Most commercial heatsinks will have a C/W rating; you're dissipating 10W per MC-E into the sink, so you can work from that - 40W into a 0.25C/W sink implies it will go up in temperature by 10 degrees. You've probably got 20-30 to play with, but I'd suggest running on the sensible side.

You're probably pushing around 800mA to the crees at the moment, so you won't be increasing the power very much at all. Use a bigger baseplate (preferably copper) and add another CPU sink on the top, should be enough.
How many hours you get before it's dead? If 7A --> 3 hours should see the battery close to dead.

Bret

 

[wquiles]

cheers for the response guys

good info wquiles athough drifting a bit from what i asked

you said...

interesting topic as this is one of my pet-peeves ... the size/material of the heatsink in a handheld flashlight is NOT the real problem.

but i never said it was a problem and i also said it wasnt a handheld flashlight.

i just want it to be sufficient, after all just like you have mensioned its got to be big enough to do the job and then beyond a certain size there is no benifit but that is still "a size" even if its not getting it as cool as we would like

im not trying to argue just get my thoughts across so sorry if ive crossed the line :grouphug:

so i guess what im asking is what is that sweet spot that max size that will give the best results within the anoying constraints we all have to live with

i was after a generic ballpark type rule some none scientific area calculation but if there isnt one then thats ok

im just askin in my aplication which is not a flashlight is a heatsink and leds in open winter (-5 to 10 degrees C) air with possable wind of 1 - 5mph, with 4 x mce at 700ma per die - then what size will give me the lowest temp achiavable using only fins and convection

oh and i cant experiment as i just want to order all the parts then build

so i really want to get it right 1st time

this is my old light and im basicly doing the same thing but upgrading it - so that will help you understand the aplication

http://www.candlepowerforums.com/vb/showthread.php?t=179661

hope all that makes sence

oh and one more thing - mce / p7 are total 10w (i think) so not all that needs to be dissapated as some of that 10w is going to make the light and not just heat

Actually, in the case of LED's, yes, almost 100% of that energy goes out as heat.

There is no "sweet spot" when refering to heatsink design, because there are too many variables. You basically have to assume some conditions and then decide how efficient a heatsink has to be, or determine if an existing heatsink could do the job at hand. For example, using this terminology:

Tr - Temperature rise = Rth * Ph
Ta - Ambient temperature
Th - Heatsink temperature
Ph - Power applied to heatsink
Rth - Thermal resistance (in °C/W)


lets assume that:

Ta = 10C
Ph = 10 W
Rth = 0.25C/W

Then

Tr = 0.25 C/W * 10 = 2.5C
Th = Ta + Tr = 12.5C

Is this Th too high? You as the designer has to decide that is acceptable - this is why there no such thing as a sweet spot. The way we build flashlights, the Th (temp of the heatsink, and eventually the temp of the host) can get as high as we are willing to hold it. In your application, which is not a handheld flashlight, the circuit/LED can get to higher temperatures, but of course, the higher the temp, the lower the output, so "ideally" you would like to keep the circuit/LED as close to 25C as possible, which unfortunately would require a gigantic heatsink and/or forced cooling.

In reality, you probably want to keep it at something no more than 120F since water temp at 122F for more than one minute will give you a first degree burn:
http://www.armstronginternational.com/files/products/valves/pdfs/ay-699.pdf

So, if we pick a max. temp of 120F (or close to 49C), you have an idea of the max. temp for the heatsink. Working backwards from your ambient (which you know) and the power (which you also know), would give you an idea of the max. Rth (thermal resistance) that you can have in your project.

Since you are making a heatsink from scratch, you really don't know the thermal resistance value (Rth), you have to find out what it is by doing some experimentation:

Tr - Temperature rise
Ta - 10°C
Th - 49°C
Vh - 3.5V
Ih - 2.9A
Ph - Power applied to heatsink
Rth - Thermal resistance (in °C/W)

Tr = Th - Ta = 49 - 10 = 39°C
Ph = Vh * Ih = 3.5 * 2.9 = 10W
Rth = Tr / Ph = 39 / 10 = 3.9°C/W

NOTE: All above assumes fairly still air, no forced cooling. Heatsink efficiency increases with forced air.

Now that you have a goal, you can now find an available heatsink that has 3.9 C/W or lower. If your application will allow using of a CPU Cooler power supply, I would try those first since those have very nice/low C/W values, and are relatively cheap. For example:
http://www.dansdata.com/coolercomp_p8.htm

I hope this helps. My apologies if my earlier rant was off-track

Will

 

[aljsk8]

it dosnt need to be waterproof - will only use it when its dry

current setup is actually driving the crees at about 1000ma
it gets to about 120 degrees c if its not in the cold air probably makes about 1800 lumens for over 40w (45l/w)

as ive said your not disipating 10w as some is making the light and the rest heat. if your putting 10w in and making 10w out as heat - where is the light coming from??

appart from the tripod ill be starting from scratch so would preffer 1 heatsink

be using 12v 15a battery this time should see 2500 lumens (60l/w) for over 3 hours could get longer but ruin the battery

edit...

wq it seams you post a reply while i was replying to the other post

that information is super and really helps i suppose by sweet spot i meant the equalibrium in a given situation - i know all situations will give different "sweet spots"

still confused as to the 100% heat thing tho??

no worries about the other post its all good

 

[wquiles]

Yup, LED's have almost no radiated heat like an incandesent bulb does. Think of an LED as a CPU on a computer - you are getting the CPU to do "work" for you (email, programs, etc.), but it all translated to almost pure heat.

As to the max. temp, it would be good to check the max. junction temperature of the LED that you are using. For example, on the CREE XR-E, the max. temp is 150C:
http://www.cree.com/products/pdf/XLamp7090XR-E.pdf

Note also on their data sheet that they have a fairly high/bad thermal resistance (bad compared to a CPU heatsink), at 8C/W - so at max. drive of 1 Amp and a vf of say about 3.6 Volts (about 3.6 watts), the LED will cause the heatsink (or whatever the LED is glue to in the circuit) to go up in temperature almost 30C, and that is above ambient (of course not so bad when your ambient is 10C).

Will

 

[gadget_lover]

Actually, in the case of LED's, yes, almost 100% of that energy goes out as heat.

Are you sure about that? I've not done scientific tests, and all of the web sites I found just now talk of Luminous efficiency, not of power efficiency, BUT....

It looks like white LEDs are up to 22% efficient at converting energy to light. I thought it was even higher, but I can find nothing to support that.

http://en.wikipedia.org/wiki/Luminous_efficacy

That still means that 8 watts out of 10 have to be dissipated somewhere. And that was the point you were making.

CPU's are creating nothing but heat, since they do not (to the best of my knowledge) glow. Wait! On second thought, the definition of a semiconductor includes the requirement that it convert some energy to some light. So I guess the 45watt CPU is probably glowing a little under that ceramic or epoxy or metal cover.


Daniel

 

[wquiles]

That still means that 8 watts out of 10 have to be dissipated somewhere. And that was the point you were making.

Exactly. I was off on the actual percentage, but most of the energy that you apply to an LED is simply "transformed" into heat.

Will

edit: I stand corrected - it is not 100% as heat. I will use 90% as a more reasonable number from now on.

 

[aljsk8]

that sounds more like it - the wiki page sounds about right

so in my case about 7-10% is making light and the rest heat
in that case ill use the 40w anyway as a guide

just want it to be more like the 50 degrees C

insted of the 120 degrees C of my current setup

-------------------------------------------------------

Will - so for all 4 mce's on a heatsink i would need 0.97c/w which by the looks of things
is not that easy without a fan

am i right in thinking size of heatsink is irelevent when looking at the c/w figure? so if a seller quotes a c/w figure then thats all i need to know

problem i see for the cpu type heatsinks is they are designed for a chip in the centre where i want to stick 4 on (not in the centre)

i could use a fan although i wanted to keep things simple but appart from simplicity there is no reason i cant use a fan

how much power does a standard fan use (say a 60 or 70mm fan)

 

[wquiles]

that sounds more like it - the wiki page sounds about right

so in my case about 7-10% is making light and the rest heat
in that case ill use the 40w anyway as a guide

just want it to be more like the 50 degrees C

insted of the 120 degrees C of my current setup

-------------------------------------------------------

Will - so for all 4 mce's on a heatsink i would need 0.97c/w which by the looks of things
is not that easy without a fan

am i right in thinking size of heatsink is irelevent when looking at the c/w figure? so if a seller quotes a c/w figure then thats all i need to know

problem i see for the cpu type heatsinks is they are designed for a chip in the centre where i want to stick 4 on (not in the centre)

i could use a fan although i wanted to keep things simple but appart from simplicity there is no reason i cant use a fan

how much power does a standard fan use (say a 60 or 70mm fan)

Yes, assuming an ambient of 10C, power of 40W, and a max. Th = 50C, your Rth would be about 1 C/W, and that puts you in the territory of fairly large CPU heatsinks. The shape is not uber critical as long as its Rth is about 1 C/W. Of course if you had a fan, you can get away with a smaller heatsink.

If the 4x MC-E's will be mounted in the same plane, you can use a single larger size heatsink like what is used for power supplies, something like this:

Since you don't want to install 4 MC-E's in the center of the CPU/other heatsink, you could always use 4 smaller heatsinks, one for each MC-E. Do you have a diagram that you can share with us as to how you envision the 4x MC-E's to be mounted?

Will

 

[aljsk8]

well the mce's will be on stars so its not that i dont want to mount them in the centre its just that i cant

for example i have a big heatsink in the shed about 70mm cubed
but it has a little copper bit in the middle about 25mm dia so obviously i cant get 4 stars on there although they would fit on the aluminium part

my old light has the same size heatsink 70mm cubed and with 9 overdriven cree p4s and 3x 3w resistors gets to about 120 degrees (tested using dmm with probe) c in 20 degree ambient (but oddly goes seriously lower in freezing temperatures - not yet tested)

yes they basicly want to be in the same plane and if i can cram all the stars next to each other then thats great but spaced out is ok too just whatever works

and if a fan dosnt use much watts and can be wired easly off the battery then that is an option

ive been looking at some akasa passive cpu heatsinks on ebay
ive emailed akasa for the c/w specs as they are not on the website

how much does a cpu put out in heat? (i know this probably varies a lot) is it less than the 40w im looking at or more - if its more and you can use a passive heatsink on one then thats great - unless they rely heavly on air movement in the case

i could use seperate heatsinks but they would end up linked together as there is only one fixing point on the tripod so heat would pass to the others

all this is giving me a lot to think about

i suppose i have to worry too that although the heatsink might be at 50 degrees C the junction temp will be higher - you have the c/w of the package then the thermal paste

i will probably drill and tap holes for the stars so i can use artic silver insted of artic adhesive

 

[wquiles]

CPU's, specially the more recent ones with multiple cores can put out a LOT more than 50 watts (some in the 140 watt range!):
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3619&p=1

Here are more examples for some (older) CPU's:
* AMD Single Core Athlon 64 Processors dissipate 89 Watts Maximum (All Parts)
* Intel Single Core Pentium 4 Processors dissipate 115 Watts (3.4 GHz, 3.6 GHz and 3.8 GHz Parts)
* Intel Single Core Xeon Processors dissipate 111 to 120 Watts (Not clear on data sheets)

* AMD Dual Core Athlon 64 X2 Processors dissipate 110 Watts Maximum (All Parts)
* AMD Dual Core Opteron Processors dissipate 95 Watts Maximum (Socket 940)
* AMD Dual Core Opteron Processors dissipate 110 Watts Maximum (Socket 939 - New 1XX Series)
* Intel Dual Core Pentium D Processors dissipate 130 Watts (3.0 GHz and 3.2 GHz Parts)
* Intel Dual Core Xeon 2.8 GHz Processors dissipate 150 Watts Maximum

As for passive heatsinks (kind of that you have in mind for your project, it is being done more and more today, specially as folks pursue a very quiet PC (with no/few fans) - these types of heatsinks are VERY efficient, but are also limited in how much heat they can remove, and many times depend on "some" airflow from another fan inside the computer case. In your particular case, since your project will be outdoors, something like this might be very effective for your project:
http://www.frostytech.com/articleview.cfm?articleID=2273


One trick that I have used more than once when I have LED's on stars is to carefully trim the edges in order to get them to sit closer:

Also, using screws and artic silver should give you better heat transfer than the thermal epoxy as you will be able to get a thinner layer, specially is you also lightly sand the back of the stars since they are NOT flat (they are punched out). Here you can see that even commercial heatsinks can benefit from lapping the surface facing the CPU:
http://www.heatsink-guide.com/content.php?content=maxtemp.shtml

Will