DIY Heatsink design (with copper cap and copper tubes for 3W leds)

richon

Newly Enlightened
Joined
Oct 11, 2011
Messages
30
Location
Viña del Mar - Chile
Hi, I want a DIY Heatsink design (with copper cap and copper tubes for 3W leds)

so far with a 3/4" copper cap and a 3/4" copper tube (with 4 pices 3/4" long) I get a 11,77 square inches of disipation area.

I know I would need at least 36 sq in for a 3W led.

would this be a good idea as I can add later more pieces of cooper tube for increasing the disipation area?

I plan to use this as a steady house light


heatsinkdemo2.png
 

xul

Enlightened
Joined
Aug 28, 2011
Messages
336
Location
MD
If the LED datasheet gives {forward voltage @ a specified I} vs. {junction temperature} you use the LED itself as a thermistor to determine the performance of your heatsink.
 

MikeAusC

Enlightened
Joined
Jul 8, 2010
Messages
995
Location
Sydney, Australia
Arctic Alumina has a thermal conductivity about ONE-TWENTIETH of copper. If you want the joints between the tubes to be effective, you need to bend the ends so that the gap is a faction of a millimetre and is tightly clamped together.
 

deadrx7conv

Enlightened
Joined
May 5, 2010
Messages
621
Location
USA
How is the light going to be mounted?

If you're doing a ceiling or wall mount, you can run a straight pipe, within the wall vertically or horizontally in the ceiling, whatever length is needed to be to cool the LED. 3w isn't too hard to cool. You can also cap both ends of the pipe, and fill it halfway with coolant. A couple of t-fittings, 90 or 45 degree angle fittings, and you can tuck a lengthy 'pipe heatsink' into almost any wall or ceiling. Sweat-solder the fittings and enjoy.

And, for you chopped up tubes, if you go that route, I'd prefer soldering/welding over any 'epoxy' to join the stack of tubes. I would definitely end cap the tube that the LED is on, after filling it 1/2-way with coolant.
 

Kinnza

Enlightened
Joined
Nov 15, 2005
Messages
223
Location
Spain
Have you thought on using a larger diameter cap piece and mount inside it some smaller pieces? Nicer and better for thermal dissipation. Some small holes in the front cap to allow air to pass would get better passive convection rates.

On the other hand, there is too aluminium tubes for cheaper, which won't get as corroded as copper as time passes. Anodized aluminium has enhanced surface area (due microscopic roughtness), making it better for dissipation.
 

Th232

Flashlight Enthusiast
Joined
Dec 25, 2008
Messages
1,064
Location
Sydney, Australia
I'd vote for using solder instead of the AA as well, as Mike has pointed out, compared to metals AA has an absolutely lousy thermal conductivity.
 

richon

Newly Enlightened
Joined
Oct 11, 2011
Messages
30
Location
Viña del Mar - Chile
well you got me all confused:

first I wanted a simple way to build a effective heatsink... but the soldering part is way better than AA (although hard to solder al the part without loosening the others at the same time, right?

the tube with coolant to the wall y a super effective design.. but my walls aren't bricks wall (these are "not heavy" light house building)

I found that using a Turbine Kind of heatsing has even better Area over the "flower kind" of heatsing

deffinitly I would have to think how to fasten the parts to the copper cap.

any ideas are welcome.


PS: is really 12 sq inch a must per Watt disipation in LEDs?
 

xul

Enlightened
Joined
Aug 28, 2011
Messages
336
Location
MD
the Led has a Vgs of 3,5V and a 21°C/watt (it has 3W and runs at 800mA)

how would I determine the performance?
Post your datasheet. With a given current, the voltage across a silicon junction changes -2.2 mV per degree C and maybe LEDs are close to this.
 

blasterman

Flashlight Enthusiast
Joined
Jul 17, 2008
Messages
1,802
Tried copper pipe and end caps for heat sinking and while it works, it's very limited in efficiency.

Biggest problem is the cap to LED star interface. Copper end caps tend to be rough and somewhat rounded so they don't make very good contact with the LED star. This needs to be smoothed / lap'd, which you seem to be aware of.

Next problem is the cap to pipe interface is a problem and you need to have thermal compound here as well.

Last, the inside of the pipe isn't nearly as efficient as removing heat unless it's totally open and in a vertical position. I actually found that glueing a star to the side of a 4" long piece of 1" diameter copper pipe worked better because it's one solid chunk.

Clustering pipes as in your drawing will also have little thermal advantage because of the mimimal contact area. Only the center cap the LED is attached to is getting the actual heat load.

Long story short, I'd rather pay $10-15 for a dedicated radial heat sink like a Wakefield because they are much more efficient per size.

12^ inches per watt is a bit overkill for Cree, but it assumes all the surface area is getting absolute convection.
 

richon

Newly Enlightened
Joined
Oct 11, 2011
Messages
30
Location
Viña del Mar - Chile
this is the datasheet link of the LED: http://www.dled.cl/download/?file=L...tZW50cy8yMDEwMDYwMjE2NDY1MV9ocDN3IHd3LnBkZg==

heatsinkdemo3.png


This will give me:
1" diameter tube = 3" and a little of perimeter
with 4" long tube = 12,5" sq Inch of disipation area.

Was this your example? was it with 1W led or 3W led?

how hot was it?

How about using the 4" long tube, flatened some in the middle to glue the Star/led and cuttind the edges and oppening like a "flower" ??? would this be better than just a plain tube?
 

xul

Enlightened
Joined
Aug 28, 2011
Messages
336
Location
MD
It's not on the datasheet. Try this
http://www.google.com/search?client...perature+mv+coefficient+LED&ie=UTF-8&oe=UTF-8

It seems to be more than -2.2 mV.

The procedure would be put some current through an LED and immediately measure the Vf. Let's say at 1.000A it's 3.000V.

Wait 10 minutes for the temp to stabilize and remeasure V and I.

If the I is still 1.000A and the coefficient is -4 mV per degree C (or per degree K) and the new V is 2.800V then the new temp is 50C more than the old temp.

Then your junction to ambient thermal resistance is 50/(~3) = ~17 C/W.
 
Last edited:

deadrx7conv

Enlightened
Joined
May 5, 2010
Messages
621
Location
USA
I want to assume the user is trying to build with available materials. Recycled computer heatsinks are worth searching for.

I have plenty of 3w LEDs, many in flashlights. They will run non-stop until the batteries die, several hours. The flashlight dimensions aren't much more than what is shown in the diagrams.... aluminum tube 3/4"-1" diameter and 3-6" long.

If you want to use copper pipes, just make sure you file the copper caps flat for the star to sit on. And, use a cap larger than the star. Don't want the star to overhang the cap curve. You might need to find a 30mm or 1.25-1.50" cap to work with. Since the cap might not send heat to your copper pipe, USE A LIQUID! How many air cooled cars are being sold today? Liquid will require 'thought' when mounting your LED to the wall/ceiling.

With that type of LED that you're using, you can also consider mounted the LED withOUT the star. This way, you have direct contact from the LED's thermal pathway bottom to the copper pipe or copper cap.
 

xul

Enlightened
Joined
Aug 28, 2011
Messages
336
Location
MD
On second thought, to avoid frying the LED, run the current at 0.1 A at first.
 
Top