Heatsink Design

[ABroodryk]

I've gotten hold of a couple of 100W COB leds and I've scoured the web trying to find a heatsink design advice with nothing concrete that I can follow. I've been playing with some ideas. Passive cooling, Active cooling, Water cooling or a combination of the above.

I'm working on a worst case scenario of 100 watts of dissipation needed. For a standard heatsink selection it will be straight forward but i would like to design a waterproof housing with the heatsink as part of the body.

So the stats and requirements:
1. 100W disapation requirement
2. Tamb = 50'C max
3. Trise above ambient = 10'C
4. Natural convection only (still air but not in a box)
5. Body Material = Aluminium or Aluminium alloy
6. Robust as it could get covered in stuff eg mud
7. Heatsink temperature feedback to lower the current in case of overheating.

What I can gather so far from my research the orientation of the fins makes a difference to the heatsink performance, Surface area, Amount of fins, Heatsink Material, Airflow etc.

The questions:
1. How do I calculate the performance of a custom heatsink EG say I have a solid block of copper 150mm x 100mm x 50mm what will the thermal resistance be in still air for a given heat in?
2. Suggestions where I can find the formulae for the calculations
3. Free or opensource heatsink simulator
4. Calculations or information on calculating water block performance
5. Other suggestions.

Thank you in advance

[T3sl4co1l]

1.25 mW/(cm^2 K) I believe is the typical figure given for convection in air.  So a big block won't dissipate much (maybe 5 or 10W in that size?), while a lot of fins will.

Radiation isn't usually important to add in, around modest temperatures, so don't worry too much about coatings or colors, either.  If you're going to go with anodize or paint (which sounds like a good idea!), matte black will give you ever so slightly better performance.  Although it will be hotter in the direct sunlight, too, if this thing ever gets mounted outdoors in the desert, say...

Tim

[IanB]

The questions:
1. How do I calculate the performance of a custom heatsink EG say I have a solid block of copper 150mm x 100mm x 50mm what will the thermal resistance be in still air for a given heat in?
2. Suggestions where I can find the formulae for the calculations
3. Free or opensource heatsink simulator
4. Calculations or information on calculating water block performance
5. Other suggestions.

The short answer is, "it's complicated".

Here's an article that gives an introduction to the subject:

http://www.electronics-cooling.com/2001/08/simplified-formula-for-estimating-natural-convection-heat-transfer-coefficient-on-a-flat-plate/

It might give you a starting point for further research.

[dom0]

10 K rise above ambient is a very low value which means you'll need a very large heat sink with very low thermal resistance.

I usually do it the other way around when choosing heat sinks, so my design procedure looks like this:

1. Pmax of device?
2. Heat resistance junction -> case + case -> heat sink?
3. Tambj = 50 °C
4. Internal heat rise for the device will be (Rthj-case + Rthcase-sink) * Pmax = Tinternal
5. Tjunction = 115 °C
6. Now we can calculate the permissible temperature rise of the heat sink: Tjunction - Tinternal - Tambj = Ths
7. And thus we can calculate the maximum permissible thermal resistance of the heat sink: Ths / Pmax

Example:
1. 40 W
2. 1 K/W
4. Tinternal = 1 K/W * 40 W = 40 K
6. 115 °C - 40 K - 50 °C = 25 K
7. 25 K / 40 W = 0.62 K/W

Quick check? 50 °C + (0.62 K/W + 1 K/W) * 40 W = 115 °C junction temperature.

Calculations for multiple devices on one heat sink are similar, you just need to consider that the heat sink must sink the heat from all devices, the rest (Tinternal etc) stay the same.

General suggestion: It might be a quicker route to success to use a COTS heat sink mounted to your case in a water proof way (e.g. rubber/silicone/PTFE gasket)

[rob77]

my 2 cents..

i usually prefer the more practical way - "eyeball" the needed heat sink size, then measure the temperature above ambient during max. heat dissipation and then either choose a heat sink with bigger surface or just keep the one i have if the temp. raise is acceptable.

[JohnnyBerg]

@dom0: how do you take a fan into account?

I divide the found value of the heatsink by 3 ~4 , but that is more a rule of thumb and some practical experience. Any thoughts?

[T3sl4co1l]

Depends, how big/fast of a fan do you have, and how good is the thermal transfer?

Heatsinks made for very efficient forced-air (or forced-fluid) operation generally have a great many fins, spaced closely; they would be worthless at convection (too much drag), but heat the air column uniformly, therefore maximizing dissipation for a given flow rate.

For casual use of heatsinks in a "breezy" environment (a fan circulating air, but not particularly directing it anywhere, or using fancy heatsinks), a factor of 2 to 4 might be typical.

Tim

[dom0]

A factor two (generously spaced fins... maybe 10 mm apart) to a factor of four (fins less than 5 mm apart) is typical in my experience, too.

Heat sinks specifically made for forced-air cooling usually have separate specs for thermal resistance with a specific fan, and/or a diagram showing thermal resistance vs. airflow.

[Circlotron]

@dom0: how do you take a fan into account?

I divide the found value of the heatsink by 3 ~4 , but that is more a rule of thumb and some practical experience. Any thoughts?

Did some tests a while back with 100 watts into a 100mm x 100mm bare aluminium h/s with 4mm thick base and 20mm deep fins, vertical, 1mm thick on 10mm centres. 100mm 12 volt fan right against the fins just wafting along at about 9 volts improved dissipation x4. Fan going way fast at 16 volts, several times previous airflow improved dissipation x6. Very much a law of diminishing returns.

[Psi]

For some comparison, here is my 100W led on a large 300W mosfet audio amp heatshink and large PC fan on top

The heatsink got too hot without the fan, like 70degC. I dont remember what the temp with-fan is but i do remember that adding the fan knocked it back a lot.
Maybe 5-20C above ambient. I can measure if you want.

[JohnnyBerg]

Haha, a lot of confirmation about my rule of thumb 

Anyone a pointer to a real calculation?

If memory serves well, there is only little airflow needed, to get good results.

[bookaboo]

Any of the decent LED floodlights 100w and above use copper heat pipes as part of their heatsinks, the heatsinks themselves are huge aluminium fins 200mm x 200mm.
Cheapo lights just use the mild steel chassis.

[T3sl4co1l]

Static convection is limited by air speed and the diffusion boundary layer.  The boundary layer, in turn, is determined by how fast the air is flowing: more speed means more disruption (whether laminar or turbulent flow), which results in a thinner layer.  Which means the fins can be spaced more closely.

So, it's not really a good comparison to say "X heatsink is Y better under Z airflow", because none of those quantities are designed properly.  If you're stuck with a given (junk box or off-the-shelf) heatsink, it can be handy (or required!) to add a fan, but the ideal case will have all three controlled for best results in a given case.

When all heat is transferred to the air, you get ~1 W/((L/s)*K) heat capacity.  For a modest 20 K temp rise at 100W, you need 5 L/s (300 LPM).  Which seems reasonable.  But, if not all of that flow is fully heated to the given temperature, if some of it is blowing past without being heated (it's outside the diffusion layer -- fins too widely spaced, extrusion not long enough), or just blows right around without going near at all (no ducting, general circulation, etc.), the LPM requirement will be much, much higher.  This figure is only the absolute minimum amount of air required.

Tim

[Circlotron]

For some comparison, here is my 100W led on a large 300W mosfet audio amp heatsink and large PC fan on top

I think an old cpu heatsink and fan assembly would be just the right size, shape and capacity for one of those high power LEDs.

[SeanB]

Just the plain circular Intel heatsink is fine up to 10W uncooled, with it sitting in free air with the find vertical. If you want to go higher then you will need the Intel branded fan, as this is a ball bearing one and which is fine up to 70W if using an old LGA775 part cooler, or if you use one of the thicker versions you can go up to 120W with fan cooling just using the stock unit. If you use an aftermarket one then you can go higher in power.

[lincoln]

Some good info here,
 I also like this writeup: http://ludens.cl/Electron/Thermal.html

For my own projects when I would make heat sinks out of scrap extrusions or other reclaimed material. I will wash oil and dirt off, square off the material in the mill and mount a power resistor(s). The sort with an al hosing and a couple of bolt holes. Some thing that approximates the heat concentration. Some where I picked up a K type thermocouple bench meter that has 8 channels, 4 probes attached. One is a short bit of wire out the back to give an ambient. Next has a washer/ring that I can bolt directly to a to220 or a foot on the resistor. the others are scrounged bits of thermocouple wire that I can wedge in the fins or around.

Driving the resistors with a bench supply, I know how much power i'm putting into the piece. And can calculate about what I can get out of the material. Be sure to turn of the HVAC, or in the case of my lab, LOL!!! what HVAC?

[ABroodryk]

Thank you for all the replies. The links were a good read and i'll try to get my head around it. But as thinks stand i might go to a water-cooled solution... Not ideal but it might have to do