Why a good thermal path really matters

[bshanahan14rulz]

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

As long as people use the stars in their stock form without cutting or sanding or grinding, they are completely safe. I myself have a pressurized gas chamber made mostly of BeO in crucial areas, and I am not worried about Berylliosis. Unless I were to drop it and break it, causing BeO dust that may become airborn. But yeah, it's mostly safe.

On a separate idea, what if you cut out the insulating layer below the slug and used, say, indium solder or indium metal to bond it to aluminum. Anybody know if indium will wet aluminum or slightly oxidized aluminum?

 

[MikeAusC]

. . . . . Even the unit in the link below has a gap between the emitter and copper chunk that would need to be filled with solder. . . . .

You must have something between the LED base and the star - and solder seems too be the best thermal option that's well ahead of any alternative.

 

[AaronM]

I understand that.
I just object to the rather large gap that needs filling.
The least thickness of something, the better.

 

[Kinnza]

The key to lowering thermal resistance is to avoid the thermal interface for isolating purposes. The beryllia star accomplish it directly, as it is not electrical conductor, so LED can be soldered directly to the cooper printed on top. You cant do that with a copper star, as you need to isolate the electrical contact from the copper.

Still if you need a thermal interface between beryllia and heatsink, then the full star surface acts transferring all the heat, without the bottleneck that usually happen at the dielectric layer. To get best performance with a copper star, you need one having the area for the thermal slug raised to the level of the circuit on the dielectric covering the star but not the thermal slug (but only valid for LEDs with electrically isolated thermal slug). Once you have the heat spread along the star without any thermal barrier (just the solder interface), thermal impedance is very reduced because you have a surface area many times larger than the initial heat generating source and it is way easier to transfer it away. The key for having the lowest thermal resistance is to be able to spread the heat as most as possible before it must go through a layer with lower thermal conductivity.

 

[saabluster]

I really don't think this is the best stuff to be using for stars but it would be a fantastic material for LED package construction.

 

[MikeAusC]

If you're mounting an XM-L or similar LED that has 3 pads on the underside, and the "star" can be connected to LED-ve, you will get maximum thermal coupling if you can solder the large central pad and the -ve pad to the "star".

The +ve pad can then be thermally coupled via an electrical insulator that's as thin as possible.

 

[Quest4fire]

Can't resist bumping this thought provoking thread. Anything new to report saabluster?

 

[D0do]

@Saabluster
How did you remove the residue of the mold release agent?
With a Q-tip and some water? Or should one just ignore it?

 

[saabluster]

Can't resist bumping this thought provoking thread. Anything new to report saabluster?

Not really. Have not had any spare time to play recently. I will say that I noticed Cutter seems to be trying to use better boards as my last order of XP-E had a very nice ceramic isolator instead of the fiberglass.

@Saabluster
How did you remove the residue of the mold release agent?
With a Q-tip and some water? Or should one just ignore it?

Isopropyl alcohol. Don't ignore it though. It could contaminate the dome and turn black over time.

 

[Shooter21]

i wonder how well a diamond heat sink would work instead of using solid copper?

 

[AnAppleSnail]

i wonder how well a diamond heat sink would work instead of using solid copper?

In theory, industrial diamond would work rather well (22 W/cm*K vs 4 W/cm*K). I wonder about the differential expansion, though, being 1/17th that of copper. I guess the low temperature range of flashlights (relatively speaking) keeps this from being too troublesome. However, artificial diamond is no fun to machine, I bet... but if you could, the heatsink and optic could be one piece!

 

[hellokitty[hk]]

Well the thing is diamond is not usually a conductor so you could do some very interesting things, possibly mounting the LED die directly on diamond.