Thermal compounds

[BIGWOOD]

I'm currently using Fujik Silicone Thermal Glue but I heard that there are better stuff out on the market for transferring heat from the LED to the heatsink like the Arctic Silver products. Is this true and someone tested? And what about potting the driver? What would you use is this case? I took a look at their website and there are quite a few different products. Now I'm confused as to which one would be the best for different applications.

1. Arctic Silver 5
2. Céramique 2
3. Arctic Alumina

and then there are the adhesives & epoxies...:thinking: Any suggestions would be much appreciated.

 

[ma_sha1]

Arctic Silver 5 is not a glue, I use it for inserting heat sink to flashlight body, so it's removable.

The two parts Arctic Silver is excellent thermo glue, better than Fujik or Arctic Alumina, I use it for glue down led star

Fujik is non conductive, I use it for potting drivers.

I don't know Céramique 2, never used it.

 

[BIGWOOD]

ma_sha1,

Thanks for the clear explanation.

 

[The_Driver]

DIEMAT DM6030Hk is much better than anything listed above (see the table of page 2 of the datasheet I linked too). It even noticeably better than soldering an emitter directly to a copper heatsink which is generally regarded as the best thing one can do to improve heat transfer.
Saabluster uses it for most of his lights (look at this post of his and also this one).
Unfortunately the minimum order amount of 50g costs $350... :( (you also need an oven)

 

[BIGWOOD]

WOW - That's quite an eye opener in terms of how far some of the members go to improve heat transfer. Thanks for sharing.

For my purposes it's a little too hardcore for me and not to mention out of my price range. I rather swap out the $5 LED for a new one once it burns out. :o

 

[CTS]

DIEMAT DM6030Hk is much better than anything listed above (see the table of page 2 of the datasheet I linked too). It even noticeably better than soldering an emitter directly to a copper heatsink which is generally regarded as the best thing one can do to improve heat transfer.
Saabluster uses it for most of his lights (look at this post of his and also this one).
Unfortunately the minimum order amount of 50g costs $350... :( (you also need an oven)

I suppose it's somewhat ironic that while I was writing a post about thermal issues this post appeared.

You must be very cautious about spec sheets. These look an awful lot like scientific documents but don't think for a moment these get out the door without the marketing department getting their fingerprints on them.

I read a study a while back done by a university, sponsored by ASME, that addressed thermal adhesive conductivity. In that test the high silver filled adhesives outperformed other materials by a margin of 10 to 1. That was at an application thickness of approx 1/2 millimeter. At the absolute minimum thickness they could bond the surfaces the least expensive, most low-tech material was a couple percentage points better than the hyper-expensive high silver content epoxy. I believe that material was a filled RTV silicone. I will try to look for the document.

Their rating of 60 w/m K is right around the transfer ability of silver solder. Tin/silver solder is about 20 points higher, just a tick shy of indium.

In this context, keep in mind that solder isn't affected by the bond thickness. It conducts the same amount of heat per given area regardless of how thick it is (within reason- at some point you have the resistance issue). If one were to use a compound such as Diemat, they would be faced with the challenge of positioning the diode, then pressing it to achieve an adhesive thickness of 0.005"

 

[The_Driver]

I suppose it's somewhat ironic that while I was writing a post about thermal issues this post appeared.

You must be very cautious about spec sheets. These look an awful lot like scientific documents but don't think for a moment these get out the door without the marketing department getting their fingerprints on them.

I read a study a while back done by a university, sponsored by ASME, that addressed thermal adhesive conductivity. In that test the high silver filled adhesives outperformed other materials by a margin of 10 to 1. That was at an application thickness of approx 1/2 millimeter. At the absolute minimum thickness they could bond the surfaces the least expensive, most low-tech material was a couple percentage points better than the hyper-expensive high silver content epoxy. I believe that material was a filled RTV silicone. I will try to look for the document.

Their rating of 60 w/m K is right around the transfer ability of silver solder. Tin/silver solder is about 20 points higher, just a tick shy of indium.

In this context, keep in mind that solder isn't affected by the bond thickness. It conducts the same amount of heat per given area regardless of how thick it is (within reason- at some point you have the resistance issue). If one were to use a compound such as Diemat, they would be faced with the challenge of positioning the diode, then pressing it to achieve an adhesive thickness of 0.005"

Thanks for the insight, I never thought about it that way. I remember saabluster mentionining somewhere that he also uses it because it is easier to epoxy emitters with this stuff than directly soldering onto the copper heatsinks. I'm guessing it's easier because the emitter doesn't float around (requiring re-centering it) like it does when being soldered...

It would be nice if he would give some insight on this topic. He must have very good reasons for using this super expensive stuff...

 

[CTS]

I just posted some math about the various interfaces in the thermal path in the other thread I started. In looking at what's below the diode, I don't know if doing anything more than using the MCPCB gets you anything. Your off-the-shelf MCPCB is capable of conducting twice the heat generated by the LED. What I mean is twice the heat that the LED is able to thermally conduct through an ideal bond of about 60 w/m K. If you used that 60 bond at 4 sq/mm to bond it to a solid diamond heat sink at 2000W/m K you're still only going to conduct as much heat as the diode joint can carry. To do much better, either the surface area of the diode needs to increase or the bond needs to be improved at the manufacturer level. I have no idea how diodes are made or what the die material is. But if you assumed for a moment that it's a piece of copper, the die could be furnace brazed to a copper board before assembly- or a single piece base.

I'm going to go out on a limb and do some theorizing. If what I'm noticing is true, manufacturers are looking for increases in efficiency and output. If they were able to get vastly increased output by simply making the thermal conductive abilities of the diode more efficient, they would be doing just that. My guess is that the diode itself is the limiting factor- that even if you had 100% thermal conductance from the die, the LED's internal ability to conduct heat to the die is limited.

I would definitely like to learn more.

 

[The_Driver]

I just posted some math about the various interfaces in the thermal path in the other thread I started. In looking at what's below the diode, I don't know if doing anything more than using the MCPCB gets you anything. Your off-the-shelf MCPCB is capable of conducting twice the heat generated by the LED. What I mean is twice the heat that the LED is able to thermally conduct through an ideal bond of about 60 w/m K. If you used that 60 bond at 4 sq/mm to bond it to a solid diamond heat sink at 2000W/m K you're still only going to conduct as much heat as the diode joint can carry. To do much better, either the surface area of the diode needs to increase or the bond needs to be improved at the manufacturer level. I have no idea how diodes are made or what the die material is. But if you assumed for a moment that it's a piece of copper, the die could be furnace brazed to a copper board before assembly- or a single piece base.

I'm going to go out on a limb and do some theorizing. If what I'm noticing is true, manufacturers are looking for increases in efficiency and output. If they were able to get vastly increased output by simply making the thermal conductive abilities of the diode more efficient, they would be doing just that. My guess is that the diode itself is the limiting factor- that even if you had 100% thermal conductance from the die, the LED's internal ability to conduct heat to the die is limited.

I would definitely like to learn more.

Have you seen this thread: Why a good thermal path really matters?

 

[CTS]

Have you seen this thread: Why a good thermal path really matters?

Took a look thru just now. What strikes me is the speed at which these devices are developing. It also serves to illustrate a point that a neophyte like me probably could overlook, yet an experienced builder just assumes- that while most of these products look alike, there are broad variances in quality.

Over the last several years my industry has been permeated with some truly abhorrent Chinese made $h... I have gained an almost reflexive disdain for anything made there, since I have to sometimes live with the results. It makes me wonder just how many new hobbyists are pushed away from this segment of the hobby thinking that it's just too hard when their issues actually stem from lesser quality components. Many are drawn to the appeal of that stunner light that's small yet delivers amazing amounts of light, yet they end up with something that breaks easily and often from something as simple as using a $2 board instead of a $4 piece, not appreciating the differences. This was one of my first thoughts when seeing people reflowing diodes and not discussing the solder formulation. How many people think there are just two kinds of solders- electrical and plumbing?

Thanks for the link- It'll be a good read this evening

 

[The_Driver]

Took a look thru just now. What strikes me is the speed at which these devices are developing. It also serves to illustrate a point that a neophyte like me probably could overlook, yet an experienced builder just assumes- that while most of these products look alike, there are broad variances in quality.

Over the last several years my industry has been permeated with some truly abhorrent Chinese made $h... I have gained an almost reflexive disdain for anything made there, since I have to sometimes live with the results. It makes me wonder just how many new hobbyists are pushed away from this segment of the hobby thinking that it's just too hard when their issues actually stem from lesser quality components. Many are drawn to the appeal of that stunner light that's small yet delivers amazing amounts of light, yet they end up with something that breaks easily and often from something as simple as using a $2 board instead of a $4 piece, not appreciating the differences. This was one of my first thoughts when seeing people reflowing diodes and not discussing the solder formulation. How many people think there are just two kinds of solders- electrical and plumbing?

Thanks for the link- It'll be a good read this evening

Well...
The led to copper bonding/soldering thing is only done by a select few modders here. No flashlight manufacturer does this.
You always have to think about if all these improvements actually make a practical difference. If a normal flashlight is 7% brighter because of very expensive heatsinking techniques than those aren't really worth it.
When leds are driven at moderate current levels they never get thaaat hot in the first place. A PCB thermal-epoxied to a reasonably sized aluminum heatsink will do fine.
A DEFT from saabluster (Cree XR-E overdriven at 1.5A-1.7A) or the Varapower lights are totally different beasts.

 

[CTS]

I very much agree. We're talking about pushing these devices well beyond their design specs. That's the fun part!

 

[Hoop]

Some of the luminus chips such as the CST-90 and CSM-360 etc. are, I believe, directly formed upon their copper PCB's and it greatly helps with heat transfer. Also, the led-tech.de copper PCB's are well regarded as far as XML stars go.

Check out the Skinnee Labs thermal compound roundup for a comparison of pc related performance of common thermal pastes.

 

[The_Driver]

Some of the luminus chips such as the CST-90 and CSM-360 etc. are, I believe, directly formed upon their copper PCB's and it greatly helps with heat transfer. Also, the led-tech.de copper PCB's are well regarded as far as XML stars go.


Check out the Skinnee Labs thermal compound roundup for a comparison of pc related performance of common thermal pastes.

The luminus chips with are factory mounted on copper pcbs aren't any better than a loose one soldered to a copper heatsink. They are made for projectors, where they are driven hard, but the manufacturing needs to be efficient and easy. It's easier to screw a package like this (which also has a socket for a special plug on the pcb) into a projector and just connect the plug instead of doing everything by hand or having a reflow soldering machine do it.

The Indigo Extreme which is the best solution in that thermal "paste" comparison is rated at 20W/m K

 

[bshanahan14rulz]

uh-oh.

 

[Hoop]

The luminus chips with are factory mounted on copper pcbs aren't any better than a loose one soldered to a copper heatsink. They are made for projectors, where they are driven hard, but the manufacturing needs to be efficient and easy. It's easier to screw a package like this (which also has a socket for a special plug on the pcb) into a projector and just connect the plug instead of doing everything by hand or having a reflow soldering machine do it.

The copper luminus chips have a thermal resistance of .92*C/W and the loose SST-90 is at .64*C/W. With a thin solder layer I see what you mean; you would beat the copper offering. Compared to a loose XML which has 2.5*C/W, the copper luminus chips look pretty nice though.

The Indigo Extreme which is the best solution in that thermal "paste" comparison is rated at 20W/m K

Indigo Extreme is not a thermal paste, but rather a engineered thermal interface. It is a metal pad which needs to melt and cure. Interesting is that the thermal pastes come close to its performance in those tests when the best thermal paste, Prolimatech PK-1, has a thermal conductivity of 10.2 W/m-℃.

 

[The_Driver]

Indigo Extreme is not a thermal paste, but rather a engineered thermal interface. It is a metal pad which needs to melt and cure. Interesting is that the thermal pastes come close to its performance in those tests when the best thermal paste, Prolimatech PK-1, has a thermal conductivity of 10.2 W/m-℃.

I know ;).
Thats why I put the word paste in apostrophes.

 

[CTS]

I looked at Indigo. I suspected by your reference to it melting that it was a phase-change alloy. It's a great concept but has limitations, one of which pretty much negates its use with LED's. PCA's are very reactive to oxygen. They are only useful in an application where air/water can be sealed out. There's a significant difference in scale- 1600 sq/mm on the CPU vs 7-9 sq/mm with an LED. On a computer CPU, there's plenty of room around the outside for gasketing. If you look at the products offered you'll see that in their packaging. Gallium is another fantastic thermal conductor. It has the advantage of being a liquid at room temperature so it's very easy to use. Unfortunately it gobbles up aluminum like a leaking alkaline battery.

There are some things that are for the most part are carved in stone. Metals conduct heat better than just about anything else. Heat transfers at the atomic level, so metal atoms bonded at the atomic level will conduct heat better than metals touching each other. If you took two copper blocks, one cast and one made of sintered copper particles, the copper block will conduct better. Simple physics. In a real world application, it's not a miracle material that provides the solution. It's going to be an understanding of all the issues and attending to every detail. A copper PCB does nothing for you if it has a lousy contact area with the bottom of the diode. There are no materials that will affix a diode with better thermal conductivity than several readily=available solders. On the flipside, you can cause yourself problems by ignoring that there are dramatic differences in conductivity across the range of solder alloys. Understanding that bond thickness is critically important can have dramatic effects.

On Indigo, before I went to all that trouble of sealing it (and I suspect the pricetag of some gee-whiz consumer product) I'd look at one of the high-conductivity adhesives. Diemat, which I have no experience with, is purported to conduct at 60.

The point I'm trying to make is that to build a high-performance design, you have to have a handle on all the variables and that requires quantifying exactly what conductive values you have at each point.

If your goal is to get triple SST-90's in a stock Mag D head and not glue the boards to the sink, you have design limitations- barriers put into place by the laws of physics. Now if you've got a Peltier cooler and an exhaust fan in the tailcap, you're probably getting closer to something that works.

 

[thijsco19]

I have a small question, does anodisation effect the heat transfer between alluminium?

 

[CTS]

I have a small question, does anodisation effect the heat transfer between alluminium?

Anodizing is an exceptional medium for heat emission. That's where that conductive heat energy is radiated out in the form of magnetic radiation (radiant heat). Most heat sinks are made from aluminum and anodized for this very reason. There are specialty coatings out there used as heat disbursing agents that are based on alumina, which is the ceramic that is formed when the aluminum is anodized.

For conduction, it has, depending on thickness a substantial insulating effect. I've read that it cuts conductivity between ten and thirty times.

The short- great on the outside, bad on the inside.