Indeed a better result, so much of an improvement that I'm inclined to think we are not apples vs apples in the example.
3Cdegrees would result from better than 0.1C/W
11Cdegrees would result from ~0.34C/W
However I have seen figures of ~0.2C/W for direct mounting of To247 to the heatsink. Is it possible that introducing a conformal pad between the metal to metal interface can reduce the thermal resistance by ~50%?
I like Granny Smiths.
3Cdegrees would result from better than 0.1C/W
11Cdegrees would result from ~0.34C/W
However I have seen figures of ~0.2C/W for direct mounting of To247 to the heatsink. Is it possible that introducing a conformal pad between the metal to metal interface can reduce the thermal resistance by ~50%?
I like Granny Smiths.
But, that 0.2C/W is not for an air filled interface.
0.2C/W applies to a well assembled direct mechanical fixing using thermal compound to ensure no air in the interface. The average thermal compound thickness of the >95% of the interface is likely to be around 0.0001inch (1/10 of a thou).
From my experience:
ARCTIC MX-2 (aluminum oxide based) thermal compound is 5 times more thermal conductive than mica. There's new one MX-3 which almost two times more thermal conductive than MX-2. Both compounds are not electrically conductive/capacitive.
Looks like Kerafol is the best choice but where can I buy it in North America?
In my F5 I'm using two 50x60mm intermediate aluminum heat transfer pads between FETs and heat sink.
FETs are mounted to pads directly w/o any electrical insulation and pads are mounted to heat sink via 0.25mm MX-2 compound layer.
It gives a very good thermal transfer from FETs to heat sink I measured only 1C difference between heat sink and FET cases.
ARCTIC MX-2 (aluminum oxide based) thermal compound is 5 times more thermal conductive than mica. There's new one MX-3 which almost two times more thermal conductive than MX-2. Both compounds are not electrically conductive/capacitive.
Looks like Kerafol is the best choice but where can I buy it in North America?
In my F5 I'm using two 50x60mm intermediate aluminum heat transfer pads between FETs and heat sink.
FETs are mounted to pads directly w/o any electrical insulation and pads are mounted to heat sink via 0.25mm MX-2 compound layer.
It gives a very good thermal transfer from FETs to heat sink I measured only 1C difference between heat sink and FET cases.
Last edited:
Maybe we can do a small group buy to get some TO-247 Kerafol pads from Zhoufang. They're $1 each and shipping is free with a $50 order.
http://www.diyaudio.com/forums/vend...therm-red-86-83-isolators-247-1-ea-stock.html
http://www.diyaudio.com/forums/vend...therm-red-86-83-isolators-247-1-ea-stock.html
No, The pads are 8mm thick pure aluminum. They are electrically isolated from main heat sink with 0.25mm aluminum oxide thermal compound MX-2. FETs are mounted on pads directly w/o electrical insulation, only with MX-2 as well. Pads have 0.25mm thick frame along the perimeter to get a gap between them and heat sink. I'm using only one 6-32 bolt to keep FET, pad and heat sink all together.
Heat transfer via such pad 10 times better than heat transfer using to-247 size 0.05mm mica.
Last edited:
> The average thermal compound thickness of the >95% of the interface is likely to be around 0.0001inch (1/10 of a thou).
The flatness of a fly-cut surface (heatsink) over the small area is probably in the order of 5~10µm, with a roughness of about 5µm pk-pk. The flatness of a piece of punched sheet metal (transistor) is more like 15~20µm, perhaps with a slightly better roughness. Adding on top the filler particle size of the thermal compound (say 5µm), there is a gap which is quite a bit larger than 1/10 of a thou, or 2.54µm.
On top of that, the 2SK1530 has a metal backing which is quite a bit larger than a TO247. So the factor of two is not difficult to explain.
A thermal "grease" probably has a filler content of less than 40%, or the grease would become too viscous. Extruded ceramic elastomer, on the other hand, can reach up to 65% filler content, especially when using a bi-modal or tri-modal mix. The high extrusion pressure allows a level of compaction that is not easily reached by other means.
Don't take my word. It only costs US$1 to try it out.
Patrick
The flatness of a fly-cut surface (heatsink) over the small area is probably in the order of 5~10µm, with a roughness of about 5µm pk-pk. The flatness of a piece of punched sheet metal (transistor) is more like 15~20µm, perhaps with a slightly better roughness. Adding on top the filler particle size of the thermal compound (say 5µm), there is a gap which is quite a bit larger than 1/10 of a thou, or 2.54µm.
On top of that, the 2SK1530 has a metal backing which is quite a bit larger than a TO247. So the factor of two is not difficult to explain.
A thermal "grease" probably has a filler content of less than 40%, or the grease would become too viscous. Extruded ceramic elastomer, on the other hand, can reach up to 65% filler content, especially when using a bi-modal or tri-modal mix. The high extrusion pressure allows a level of compaction that is not easily reached by other means.
Don't take my word. It only costs US$1 to try it out.
Patrick
There are ways to increase filler content and still maintain low viscosity. By formulating to produce non-newtonian flow properties (thixotropic) you can quite easily formulate to over 70% filler content, and still maitain low viscosity (under shear).
I was considering formulating a thermal paste using nano-diamond particles about 8 years ago, but I gave up before even attempting. I made the assumption no one would want to pay that much for thermal compound. I never did the cost anlysis so I don't even know how much it would cost to make.
Anyway, I agree with most things Patrick has said here.
I was considering formulating a thermal paste using nano-diamond particles about 8 years ago, but I gave up before even attempting. I made the assumption no one would want to pay that much for thermal compound. I never did the cost anlysis so I don't even know how much it would cost to make.
Anyway, I agree with most things Patrick has said here.
For a mono mode powder, the theoretical maximum packing density is about 65% by volume (part of my Ph.D. topic). To go over that, you need to use bi-modal or tri-modal (2 or 3 nominal powder size). If you put a thixotropic paste below the transistor, you would need a very high pressure to squeeze the excessive out through the (hopefully very) small remaining gap.
This is where phase change compound like Aavid Ultrastick comes in. Though thixotropic, the wax base melts at about 60°C, allowing the compound to flow. It still takes a lot of pressure to squeeze out the excess.
Patrick
This is where phase change compound like Aavid Ultrastick comes in. Though thixotropic, the wax base melts at about 60°C, allowing the compound to flow. It still takes a lot of pressure to squeeze out the excess.
Patrick
I never said you would use nano-particles to make thixotropic compounds.
However you can combine the two different types of particles in different ratios to get thixotropic flow properties as well as the benefit of high thermally conductive particles such as diamond.
The good thing about the nano-particles is you don't need a lot (about 10% volume content or less if dispersed well) to get particle-to-particle contact. Unlike carbon nano-fibres/tubes, nano-diamond particles disperse quite easily.
You end up with a compound like whipped cream. It normally does not flow or sag but as soon as you apply some force/shear it flows very easily.
I never did any measurements to know how good it was, at that point in time I was only interested in getting the most nano-diamond particles I could get in there and still maintain nice flow properties. So you might end up with 20% diamond and 50% thixotropic particles. Obviously the thixotropic particles/pigments also need to have good thermal properties
Anyway, having said that and without doing any measurements, my gut feeling is that the keratherm pads would be better.
Sorry, if I came across a bit harsh, Patrick. I never intended to offend you. If it makes you feel any better I turned down a PhD scholarship, to take up a job offer.
Edit: Where you see volume content I am referring to pigment to binder ratio. There are also some level of volatile solvents (not included in the volume) which evaporate once applied.
However you can combine the two different types of particles in different ratios to get thixotropic flow properties as well as the benefit of high thermally conductive particles such as diamond.
The good thing about the nano-particles is you don't need a lot (about 10% volume content or less if dispersed well) to get particle-to-particle contact. Unlike carbon nano-fibres/tubes, nano-diamond particles disperse quite easily.
You end up with a compound like whipped cream. It normally does not flow or sag but as soon as you apply some force/shear it flows very easily.
I never did any measurements to know how good it was, at that point in time I was only interested in getting the most nano-diamond particles I could get in there and still maintain nice flow properties. So you might end up with 20% diamond and 50% thixotropic particles. Obviously the thixotropic particles/pigments also need to have good thermal properties
Anyway, having said that and without doing any measurements, my gut feeling is that the keratherm pads would be better.
Sorry, if I came across a bit harsh, Patrick. I never intended to offend you. If it makes you feel any better I turned down a PhD scholarship, to take up a job offer.
Edit: Where you see volume content I am referring to pigment to binder ratio. There are also some level of volatile solvents (not included in the volume) which evaporate once applied.
Last edited:
I have no problem with open technical arguments. You should know.
> Where you see volume content I am referring to pigment to binder ratio. There are also some level of volatile solvents (not included in the volume) which evaporate once applied.
That is cheating.
> However you can combine the two different types of particles in different ratios to get thixotropic flow properties as well as the benefit of high thermally conductive particles such as diamond.
> So you might end up with 20% diamond and 50% thixotropic particles.
That is exactly the same as what I meant by bi-modal.
If all particles are perfect spheres of the same size, and you can pack them by hand in the tightest possible manner, you can fill up about 65% of the volume. The rest of the space you can now fill up with a particle size (smaller spheres) about a factor of 10 smaller. Again if you have perfect packing, you would fill up say 50% of the remaining 35%, giving you a theoretical fill factor of 80%+.
That is theory. In practice, the particles are not spheres, they are not all the small size, and they do not pack themselves into perfect filling geometry. So a bi-modal mix will get you slightly above 70%. Extrusion is good for packing, compared to the like of isostatic pressing or slip casting, in that you have some lubrication effect of the molten plastic, plus the extremely high extrusion pressure which helps to pack the particles as tight as possible.
The Kerafol 82/86 is not the best they have in sheer thermal performance. I have tried other stuff which are better thermally. But they have such high filler content and hence so fragile that they break into pieces in your finger. So I thought I wasn't going to recommend the public to use any of those.
There are ways to get even better performance (e.g. doubling the pad size by soldering a 3mm copper plate to the TO247 of twice the footprint, and using a larger piece of Kerafol). But you need to know what you are doing. And at a dT of already 3°C (with the 2SK1530, so about 4°C for TO247), the gain is at most 2°C for all the trouble.
Patrick
.
> Where you see volume content I am referring to pigment to binder ratio. There are also some level of volatile solvents (not included in the volume) which evaporate once applied.
That is cheating.
> However you can combine the two different types of particles in different ratios to get thixotropic flow properties as well as the benefit of high thermally conductive particles such as diamond.
> So you might end up with 20% diamond and 50% thixotropic particles.
That is exactly the same as what I meant by bi-modal.
If all particles are perfect spheres of the same size, and you can pack them by hand in the tightest possible manner, you can fill up about 65% of the volume. The rest of the space you can now fill up with a particle size (smaller spheres) about a factor of 10 smaller. Again if you have perfect packing, you would fill up say 50% of the remaining 35%, giving you a theoretical fill factor of 80%+.
That is theory. In practice, the particles are not spheres, they are not all the small size, and they do not pack themselves into perfect filling geometry. So a bi-modal mix will get you slightly above 70%. Extrusion is good for packing, compared to the like of isostatic pressing or slip casting, in that you have some lubrication effect of the molten plastic, plus the extremely high extrusion pressure which helps to pack the particles as tight as possible.
The Kerafol 82/86 is not the best they have in sheer thermal performance. I have tried other stuff which are better thermally. But they have such high filler content and hence so fragile that they break into pieces in your finger. So I thought I wasn't going to recommend the public to use any of those.
There are ways to get even better performance (e.g. doubling the pad size by soldering a 3mm copper plate to the TO247 of twice the footprint, and using a larger piece of Kerafol). But you need to know what you are doing. And at a dT of already 3°C (with the 2SK1530, so about 4°C for TO247), the gain is at most 2°C for all the trouble.
Patrick
.
Last edited:
That is cheating.
.
I know. Ha Ha Ha Ha Ha (evil laugh)
If you can tell me why spherical particles are the absolute worst shape for thermally conductive films/coatings, I might share a few more evil tricks.
If that is the case I reckon I might be able to beat Kerafol at there own game (If you are listening Kerafol make me an offer
).About 8 years ago now I produced films that had somewhere between 80% and 90% (I have to go back and check my notes) pigment to binder content, using specialised elstomeric polymers, that maitained very good cohesive strength and excellent flexibility.
Anyone play with this stuff?
Thermal - Pads, Sheets, Pyrolytic Graphite | Digi-Key
Very expensive, amazingly conductive, possibly not that much more helpful since the package is still a bottleneck... but it could make a heck of a heat spreader! Follow one of the items for a link to check the data sheets.
Hmmm... 20mm x 20mm pads for under $2, if 500 min order is made...
Thermal - Pads, Sheets, Pyrolytic Graphite | Digi-Key
Very expensive, amazingly
conductive, possibly not that much more helpful since the package is still a bottleneck... but it could make a heck of a heat spreader! Follow one of the items for a link to check the data sheets.Hmmm... 20mm x 20mm pads for under $2, if 500 min order is made...
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.