Pyrolytic Graphite Sheet, high performance thermal interface material

[BravoV]

When browsing Digikey for thermal pads, found this so called pyrolytic graphite sheet (PGS) which has an impressive thermal conductivity property at 700 to 1750 W/(m-K), for comparison, pure copper is below 400 W/(m-K). 

Reference -> Panasonic PGS web site



Example list of Panasonic PGS products that are available at Digikey -> Panasonic PGS , its damn expensive thermal pad there. 

To anyone that had used this PGS product, curious what kind of applications need this kind of ultra high performance thermal interface material ? Cooling laser ? Please share your experience if you have any.

[kizzap]

Screw the PGS stuff, I want pure diamond. In fact, I demand it!

-kizzap

[amyk]

That's a rather large area... you only need 1/100th of that for something like a MOSFET, and at that area it's <$0.50.

Some extreme overclockers might be interested...

[BravoV]

That's a rather large area... you only need 1/100th of that for something like a MOSFET, and at that area it's <$0.50.

Some extreme overclockers might be interested...

You're right, that price is nothing for those crazy extreme overclockers that used to spend a fortune just for cooling gadgets.

But as you said, even the smallest piece available can be cut into many-many smaller pieces that will suitable for high powered semiconductor.

Actually I'm currently thinking of using it at those poor power mosfets that dissipate lots of heat in linear region like in the dummy load circuit + a high performance overclockers grade cpu heat sink, this is quite tempting ...

[mamalala]

Slight problem for the PGS only variant: it is also electrically conductive. There are variants available that are non-conductive, but those use a insulating layer of a different material, which will probably bring down these numbers. Plus, those withstand "only" 100°C max, as per datasheet.

Greetings,

Chris

[BravoV]

Of course it conducts electricity, basically its a Graphene or graphite like the pencil's lead as we all know for quite sometimes now.

What interest me (personally I just knew it recently) that its now available for sale as retail product, and this opens opportunities to exploit it's special property for avg people like myself.

The Digikey link I posted is those bare PGS (S type) that without those extra bonding material that can withstand the temperature up to 400 C, there are other types too like you mentioned at DK, but I'm not very impressed with those low temp limitation.

[mamalala]

Of course it conducts electricity, basically its a Graphene or graphite like the pencil's lead as we all know for quite sometimes now.

Yes, i was just curious if they may had a special mixture or something that added some kind of insulation. After all they talk about polymers, so there could have been a chance.

The Digikey link I posted is those bare PGS (S type) that without those extra bonding material that can withstand the temperature up to 400 C, there are other types too like you mentioned at DK, but I'm not very impressed with those low temp limitation.

Yea, the 100°C of the other types are really low.

Greetings,

Chris

[mrflibble]

Holy crap! I didn't realize that this stuff was commercially viable already. And even at doable prices.

What would be interesting to know is how it performs over time, and over many thermal cycles. If it retains it's properties over lets say 3 years time of regular PC usage then it looks like a viable solution even for the moderate non-hardcore overclockers amongst us. Or even without overclocking. Just a reliable low maintenance thermal interface that's high performance to boot is interesting. A 25 x 25 mm square isn't even all that expensive at single units. Let alone the high volume if cpu manufacturers adopt this.

[M. András]

yeah but with the poor machine work on the heatsinks on the pc heatsinks you have to polish the thing to get this to work unless somehow the thicker sheet fills up the gaps in the copper baseplate, as for semiconductors there are some package which is isolated ixys makes some but the price is premium level. i found these when looking for materials to put between a sot227b isotop package and a machined polished copper waterheatsink to make a few kw load, which i didnt manage to get made yet copper and those ixys linear fets are damn pricy so as the pumps and radiators

[amyk]

Holy crap! I didn't realize that this stuff was commercially viable already. And even at doable prices.

What would be interesting to know is how it performs over time, and over many thermal cycles. If it retains it's properties over lets say 3 years time of regular PC usage then it looks like a viable solution even for the moderate non-hardcore overclockers amongst us. Or even without overclocking. Just a reliable low maintenance thermal interface that's high performance to boot is interesting. A 25 x 25 mm square isn't even all that expensive at single units. Let alone the high volume if cpu manufacturers adopt this.

The thermal pad Intel has been using on its stock heatsinks is graphite-based and it works very well - if assembled correctly.

[mrflibble]

I doubt we're talking about the same stuff here. Either that, or intel recently started doing this. Last stock cooler from Intel I got the thermal pad definitely was thicker than 100 um. It was the usual moderate performance grey stuff.

[M. András]

Holy crap! I didn't realize that this stuff was commercially viable already. And even at doable prices.

What would be interesting to know is how it performs over time, and over many thermal cycles. If it retains it's properties over lets say 3 years time of regular PC usage then it looks like a viable solution even for the moderate non-hardcore overclockers amongst us. Or even without overclocking. Just a reliable low maintenance thermal interface that's high performance to boot is interesting. A 25 x 25 mm square isn't even all that expensive at single units. Let alone the high volume if cpu manufacturers adopt this.

The thermal pad Intel has been using on its stock heatsinks is graphite-based and it works very well - if assembled correctly.

if they would use this stuff. peeps wouldnt need to decap their 300bucks worth cpus to change the cheap chinese shit paste they was using ffs 20C difference when replaced with an arctic cooling mx2? that stuffs is around 5W/(mK)

[mrflibble]

Heh, yeah that was just silly of Intel. Luckily on i5-2500k the core is still soldered to the IHS.

As for lapping, you don't need to go totally overboard with that. A bit of work on the cpu and a bit more work on the comparatively crappy surface of aftermarket coolers, decent paste and you get nice price/performance.

Right now I've got no real incentive to muck about with it (if it works, don't touch it ), but with these prices I'll definitely keep PGS in mind for a future PC build. Mosfets however, mmmhh...

[peter.mitchell]

Holy crap! I didn't realize that this stuff was commercially viable already. And even at doable prices.

What would be interesting to know is how it performs over time, and over many thermal cycles. If it retains it's properties over lets say 3 years time of regular PC usage then it looks like a viable solution even for the moderate non-hardcore overclockers amongst us. Or even without overclocking. Just a reliable low maintenance thermal interface that's high performance to boot is interesting. A 25 x 25 mm square isn't even all that expensive at single units. Let alone the high volume if cpu manufacturers adopt this.

The thermal pad Intel has been using on its stock heatsinks is graphite-based and it works very well - if assembled correctly.

if they would use this stuff. peeps wouldnt need to decap their 300bucks worth cpus to change the cheap chinese shit paste they was using ffs 20C difference when replaced with an arctic cooling mx2? that stuffs is around 5W/(mK)

Yeah, why don't they do what that did on the higher end C2D line and solder the die to the IHS, GREAT thermal conductivity!

[BravoV]

... as for semiconductors there are some package which is isolated ixys makes some but the price is premium level

Which Ixys part number is that ? Really interested.

[M. András]

... as for semiconductors there are some package which is isolated ixys makes some but the price is premium level

Which Ixys part number is that ? Really interested.

isoplus247/220, and the isotop aka sot227 packages
http://www.digikey.com/product-search/en/discrete-semiconductor-products/fets-single/1376381?k=isoplus
you can get the appnote and the technology white paper from their website 2.5kv isolation without compromising thermal conductivity

[BravoV]

... as for semiconductors there are some package which is isolated ixys makes some but the price is premium level

Which Ixys part number is that ? Really interested.

isoplus247/220, and the isotop aka sot227 packages
http://www.digikey.com/product-search/en/discrete-semiconductor-products/fets-single/1376381?k=isoplus
you can get the appnote and the technology white paper from their website 2.5kv isolation without compromising thermal conductivity

But none of that selection are IXYS Linear L2 series mosfets ?

I'm actually looking at the linear fet at these selections -> http://www.digikey.com/product-search/en?FV=fff40015,fff804c8,ffec545c , at the price of $37.4 a pop. 

Does this type miniBLOC/SOT-227 like below has the isolation at it's bottom huge & thick copper tab ?

[Conrad Hoffman]

I'm confused. How is an electrically conductive pad of any thickness better than polished surfaces and a drop of silicone oil? Get the interface to near zero thickness and the thermal conductivity of what's in there doesn't matter much.

For many things I've switched over to the Aavid "phase change" compound, which I think is just lightly loaded wax. The first time you heat it up, it goes liquid and the joint goes to near zero, assuming the parts mate up well.

[Rufus]

I'm confused. How is an electrically conductive pad of any thickness better than polished surfaces and a drop of silicone oil? Get the interface to near zero thickness and the thermal conductivity of what's in there doesn't matter much.

When the material is more thermally conductive than the materials it is joining it doesn't matter (within reason) how thick it is. The real question is how well does this material conform to the the typical surfaces to be 'joined', how well does it accommodate gaps and surface textures at what pressures? I could not find any hardness information on the Panasonic web site - I gave up looking after 5 minutes.

Why is probably the second most important quality of this material not mentioned? because it sucks?

[mrflibble]

Your confusion confuses me (I am easily confused). Everything depends on the characteristics of the pad, and of the interface between the pad and materials A and B.

When the thermal conductivity of the pad/paste/whatever is orders of magnitude lower than that of materials A and B, your confusion is not so confusing. Because then indeed the first order of business is to get a reaaaally thin film in there. Because in this case the goop is the biggest contribution to thermal resistance.

However when the thermal conductivity of the pad is en par with or even high than that of materials A and B, this rule of thumb is not all that applicable. Then it is not so much the thickness of the stuff, but purely the quality of the interfaces.

Reading back, not sure if I explained it all that well, but we can only hope... 

Edit: Heh, I see Rufus beat me to it.

And yes, I tried to find that in the datasheet as well. GRRRR. With this stuff your main concern becomes the surface interfaces, aka how well it conforms to the tiny bumps in your materials A and B. And also, how stable this interface is over time.

Regardless, I am suitably impressed by the price level of this stuff. Now if only we could get some customer option dopants in it for a small price increase I'd be pleased as punch. 

[robrenz]

IMO Flatness/conformity of the mating surfaces at operating conditions (very difficult to achieve) is more critical than surface finish If you are using minimal thermal interface materials . If you have perfect flatness/intimate contact then improved surface finish will improve the thermal transfer. That is why thin interface materials that can conform can often produce better heat transfer even though their thermal conductivity is lower.

[M. András]

... as for semiconductors there are some package which is isolated ixys makes some but the price is premium level

Which Ixys part number is that ? Really interested.

isoplus247/220, and the isotop aka sot227 packages
http://www.digikey.com/product-search/en/discrete-semiconductor-products/fets-single/1376381?k=isoplus
you can get the appnote and the technology white paper from their website 2.5kv isolation without compromising thermal conductivity

But none of that selection are IXYS Linear L2 series mosfets ?

I'm actually looking at the linear fet at these selections -> http://www.digikey.com/product-search/en?FV=fff40015,fff804c8,ffec545c , at the price of $37.4 a pop. 

Does this type miniBLOC/SOT-227 like below has the isolation at it's bottom huge & thick copper tab ?

the sot 227 type packages have an  aluminium nitride isolation layer. but none of their dqatasheet says how much voltage can it withstand, their linear fets comes in sot 227 too, as for the other type i remember to see a year ago a linear fet from them in those packages but i cannot find it now

edit: if you have any corporate email. you could ask them about the details on those packages i highly doubt they will answer to my free email address. even to the gmail address
im intrested in the isolation voltage on the sot 227b packages too. plan to use them as i mentioned

[ejeffrey]

It is a bit hard to parse the marketing brochure, especially as the data presented is wrong if taken at face value: the thermal conductivity of copper is 401 W/mK, not the roughly 300 shown in the table, and aluminum is comparably off.  If it were just an error it wouldn't affect the conclusions although it would make you question the accuracy of the rest of the data.  The alternative is that they are measuring something other than the standard thermal conductivity.

Here is what I think is going on.  Graphite is anisotropic: the thermal conductivity is very high in the transverse direction but relatively low in the normal direction.  So the purpose of this is not the same as e.g., a sil-pad or thermal grease which is just filling in the air gap between two surfaces.  Instead this is meant to go between a silicon die and a heat spreader plate.  It is so thin that the normal direction conductivity doesn't matter, but it acts as a local heat spreader, evening out hot spots on the die.  This probably explains the weird thermal conductivity numbers on the plot: they are a surface conductivity rather than a bulk conductivity, but normalized by some speficied thickness to get an 'effective bulk conductivity'.  That is presumably why the conductivity of the PGS varies inversely with thickness.  Almost all the lateral conductivity is done by the first few microns, but as they normalize it by a larger thickness the 'effective bulk conductivity' drops.

[Harvs]

I'm confused. How is an electrically conductive pad of any thickness better than polished surfaces and a drop of silicone oil? Get the interface to near zero thickness and the thermal conductivity of what's in there doesn't matter much.

For many things I've switched over to the Aavid "phase change" compound, which I think is just lightly loaded wax. The first time you heat it up, it goes liquid and the joint goes to near zero, assuming the parts mate up well.

Yeah I was thinking the same when I first saw this thread.  Then I went to the panasonic website and I got a completely different feeling of what this is all about.  I thought the whole point they're marketing towards is as a space constrained heat spreader, not to improve the thermal conductivity of joints.  (E.g. a high performance tiny processor in a phone could get the heat out of the very small packages on the graphene sheet, which would effectively spread the heat across the rest of the case.)

[BravoV]

Sorry if its misleading, it was just a spontaneous reaction in creating this thread when I saw it 1st on sale at DK without further in-depth investigation.

So its more for spreading the heat evenly through that thin sheet (horizontally wise), rather than conducting the heat (vertically) as common thermal pad does ? 

Too bad the minimal shipping cost from DK is $70  , other wise I might order one just for curiosity sake.

Any volunteer ?

[Conrad Hoffman]

OK, now the stuff makes more sense to me. I used to use HPOG (highly oriented pyrolytic graphite) as a target for scanning probe microscopes- you'll see a lot of the atomic scale images use that as a demo, and the stuff consists of loosely attached layers. You actually expose a fresh surface by pulling the top layer with tape. Thus, it should have only so-so thermal conductivity in the normal direction.

I'm still confused, though that's always been the case. It's because you can never get consistent conductivity numbers for all the different pads and such, to make a direct comparison. Even a single manufacturer will use different methods for their different products, just so you can't really tell at a glance what's what. I'd love for some independent person to set up a standard test and publish numbers for all the different pads, mica and different goops, to finally put some sense to it.

FWIW, I use a simple measure- the hotter the heatsink, the better. IOW, if the device is hot and the heatsink isn't, the interface needs work. If the device and heatsink are close to the same temp, life is good. Copper spreader plates are often a wonderful thing, but few people use them other than maybe for CPUs. Being an analog guy, CPUs are of very little interest to me! When you get to the conductivity level of copper, how much could possibly be gained by switching to graphite? IMO, not much.