Heat Sink Compound - Got to love this stuff

[K9BF]

Hello
Been doing some electronics repair. I bought a big tube of heat sink compound and just finished it off. Whenever I replace a component that had heat sink compound between the component and the heat sink, I really pile in on. I love to see all the extra squish out along the sides. I am thinking about building a wall around some components so I can completely bury them in heat sink component. Its got to be better than just leaving all that air around them, right?
Ben

[The Doktor]

No.

[rdl]

The general idea is that the compound fills in the tiny surface irregularities between the device and the heat sink, which would otherwise just be full of air. Using as little compound as possible while still accomplishing that is what is wanted.

[orbiter]

In many cases too much compound can be just as bad than not enough. Totally covering components in heatsink compound will likely result in those components
overheating and dying prematurely.

[uncle_bob]

Hi

Resist the urge to ice cupcakes with the stuff. Yes it's hard, but do try to resist. The stuff is not at all healthy to eat (as in wash up after using it).

Yes people have made sculptures out of heatsink compound. The next amazing discovery is that the stuff separates eventually. The pretty sculpture takes on an odd look as part A moves away from part B. This is also a good reason to stir the stuff if you buy it in gallon pails. It is also another reason to use it in small quantities / thin layers.

As wonderful as the guys who sell the stuff make it sound, it's not the solution to all of life's problems. From the advertising, one would think that water cooling blocks would do better with heatsink compound flowing in them. While this *sounds* silly, yes I have seen a setup like that. It was decided that water was an "unsafe compound" to use in the system. Yes those guys *really* bought the advertising. No the system did not work well at all. In fact there are neat tables of various physical properties of "stuff". Mr Google is good at finding them. When you look at the numbers ...... errrr .... the advertising is at least as accurate as the guy on TV who's going to stop selling furniture later today.

Yes, it's good stuff. It's vital that it be used. It makes an absolutely horrible mess. Use it with caution.

Bob

[v81]

Hello
Been doing some electronics repair. I bought a big tube of heat sink compound and just finished it off. Whenever I replace a component that had heat sink compound between the component and the heat sink, I really pile in on. I love to see all the extra squish out along the sides. I am thinking about building a wall around some components so I can completely bury them in heat sink component. Its got to be better than just leaving all that air around them, right?
Ben

I don't know if i was wrong to do so, but i took this as a troll / joke.

Everyone knows that thermal paste is used only to fill voids between surfaces right?
Thinnest layer possible between surfaces without introducing air gaps is the go.

[CatalinaWOW]

You are trying to get the heat out of the chip and ultimately into the air.  Any excess thermal compound is just adding an insulator into the path.  Sure it is a bad insulator, but it is an insulator.  The only reason to use it at all is that that bad insulator replaces the better insulator (air) in the voids under the chip.  That air can't go anywhere to remove heat, but the air around the edges can, so you don't want to replace it with paste.

[Srbel]

Time Lord is right.

[Brumby]

Bottom line is this - Heat sink compounds are not terribly good conductors of heat - but they are immensely better than air!

There are a number of variables in sorting out thermal solutions - but to give you a (really rough) idea, a good thermal compound might be around 25 times worse than aluminium, but it's 300 times better than air!

Moral of the story - you want to make sure you have as little thermal compound as possible, but ensuring there are no trapped air pockets and have it spread over the maximum area available.


This will take practice and experience.  If you ever want to pull the heatsink off and check, you will have to completely re-do the whole thing.  Putting a heatsink back without doing so will guarantee you will include air pockets.

Monitoring temperatures is the way to go.

[timb]

Personally, if I *have* to use a thermal compound, I use Arctic Silver. Have for, damn, at least 10 years now? Works great as it has silver and ceramic particles in it, so it has much better conduction than your standard white goop. As a bonus, a little goes a very long way. (My original tube last several years and about a year ago when Radio Shack was closing, I got 5 packages for $1 a pop, so I'm set for life basically!)

Arctic Alumina also works well and is a bit cheaper. They also have thermal epoxy versions, which I actually use way more often than non-adhesive compounds. (It's great for attaching those small square copper heat sinks to things like the SoC on a Rapsberry Pi or U800 on a Tek 2465B.)

Now, why do I rarely use thermal compound? Because it's messy and an enormous pain in the ass. I restore a lot of vintage test gear and, as a matter of course one of the first things I do is remove any transistors attached to the case or a heat sink, throw away the mica insulator, wash both surfaces with a degreasing dish liquid and dry with 91% Iso to remove the thermal grease.

Then I reattach the transistor to the case/heat sink, only instead of using a new mica insulator and thermal grease, I use these amazing little things called Sil Pads! They're these grey, flexible, rubbery little pads made out of silicone rubber with a fiberglass weave to provide reinforcement. When the device is tightened against the heat sink, the pad is crushed between the two and because it's silicone rubber, ends up filling any small scratches, voids and unevenness in the metal (the role thermal grease would normally provide). The silicone also provides heat transfer (initially 1.1c-in^2/W, improving with age) and electrical isolation.

It's the ultimate solution, really. You can get TO-220 versions new on DigiKey or eBay, and I've found that, since they're a bit oversized they'll fit similar package sizes (you can trim them with scissors if needed). TO-3 versions can be found as New Old Stock on eBay. You can also buy raw 10"x10" pads there as well. I keep these on hand for odd packages, like those smaller versions of the TO-3 package; I recovered a mica insulator from one and use it as a stencil to cut them when needed. (One of these days I'll scan that insulator and draw it in some vector software, then use my digital cutting machine to make a stack of them out of the Sil Sheets.)

You can also find ceramic insulators for TO-220 packages on eBay. Not sure how well they work as I've never used them. They're quite thick, too.

As an aside, I recently parted out an old Fluke Voltage Calibrator from the late 60's (this one had early 80's date codes on the parts). Anyway, the voltage reference card had an oven on it, made from 1/4" metal plates. They stacked about 6 of these plates; the first 4 had the centers cutout and the top 2 were solid. Two 25W 500Ohm metal Dale resistors screwed to the top and acted as the heater.

Now, the whole thing was bolted to the PCB, with the components to be ovenized in the middle of it. However, it looks like Fluke had absolutely *filled* the center cavity of the oven with thermal compound, so that it completely covered all the components. When I opened it, it was like pulling apart a 40 year old Twinkie. Most of it had long since dried out, so it was either in hard clumps or dust form. The only part still in the original state was what had oozed between the plates before the assembly was tightened (I guess this kept it out of the air and fresh).

In the long run, I'm not entirely sure how effective filling the cavity with thermal grease was, compared to just letting natural convection work its magic.

[mikerj]

Personally, if I *have* to use a thermal compound! I use Arctic Silver. Have for, damn, at least 10 years now? Works great as it has silver and ceramic particles in it, so it has much better conduction than your standard white goop. As a bonus, a little goes a very long way. (My original tube last several years and about a year ago when Radio Shack was closing, I got 5 packages for $1 a pop, so I'm set for life basically!)

The expensive silver compounds have better thermal conductivity than the white stuff, but the point that many people seem to miss is that the actual active layer of thermal compound should be extremely thin if the heatsink and device being mounted are flat.  This means even with the standard white grease the temperature drop will be very small, and the silver will give little improvement.

If the mating surfaces aren't flat then the silver stuff will definitely do better, but it would be better still with flat surfaces.

[Jeroen3]

There are multiple classes of thermal compound:
- Metal-Oxide material paste, cheap white stuff, electrically non-conductive.
- Metal material paste, expensive grey stuff, electrically conductive.
- Stickers, basically buttery type stuff that will melt and settle when heated the first time. Often pre-applied on pc heatsinks. Sometimes it also glues a bit.

For added electrical isolation you can use:
- Mica, horrible thermal conductivity, but it's cheap.
- Sil-pads, reasonable thermal conductivity, not very good though. Also cheap.
- Aluminum oxide plates, cold white material. Very good thermal conductivity (that is why is feels cold). You can combine these with oxide based paste. They are a bit thick, because the oxide is fragile. You can get them sub-mm, but they crack very easy. Get them at >1.5 mm. A 5mm one still performs better then a sil-pad. But they are expensive to buy and mount.

Note: there also is a maximum torque at which you can fasten an to-220. Over-tighten it and it will gap, the compound cannot fix this. See appnote: https://www.fairchildsemi.com/application-notes/AN/AN-4166.pdf.

[Kleinstein]

... a good thermal compound might be around 25 times worse than aluminium, but it's 300 times better than air!

The cheap thermal compound might be around 300 times worse than aluminium, but this is still better than air.

[AlxDroidDev]

One thing I always do with my computers is to lap the heatsink with fine sandpaper (starting at 800 grit, finishing at 2000 grit), until I have a mirror finish look, way better than it was before. I have done this to a few LM317, LM2941CT and a few other voltage regulators. I'll do this to the MOSFETs I'll be using in my electronic load and to the coolers.

Neverthless, even on lapped surfaces, I still use HS compound, and I am also a fan o Arctic Silver 5, and I also have a couple tubes of IC Diamond. A small tube of either one goes a long way, since all I use is a drop the size of a rice grain at the middle of the CPU/GPU. The pressure applied by the cooler/heatsink is enough to evenly spread it over the surfaces.

In my current desktop, an i7-920 overclocked from 2.66 to 3.65GHz, I am using Indigo Xtreme, with the best HS I could afford (can't remember the model, but was one of the top 5 according to Tom's Hardware at the time I built the machine). It's even better than AS5, but a lot more expensive and a PITA to apply. It comes with a cool set of lab-grade gloves, 'though!

[timb]

Personally, if I *have* to use a thermal compound! I use Arctic Silver. Have for, damn, at least 10 years now? Works great as it has silver and ceramic particles in it, so it has much better conduction than your standard white goop. As a bonus, a little goes a very long way. (My original tube last several years and about a year ago when Radio Shack was closing, I got 5 packages for $1 a pop, so I'm set for life basically!)

The expensive silver compounds have better thermal conductivity than the white stuff, but the point that many people seem to miss is that the actual active layer of thermal compound should be extremely thin if the heatsink and device being mounted are flat.  This means even with the standard white grease the temperature drop will be very small, and the silver will give little improvement.

If the mating surfaces aren't flat then the silver stuff will definitely do better, but it would be better still with flat surfaces.

Personally, if I *have* to use a thermal compound! I use Arctic Silver. Have for, damn, at least 10 years now? Works great as it has silver and ceramic particles in it, so it has much better conduction than your standard white goop. As a bonus, a little goes a very long way. (My original tube last several years and about a year ago when Radio Shack was closing, I got 5 packages for $1 a pop, so I'm set for life basically!)

The expensive silver compounds have better thermal conductivity than the white stuff, but the point that many people seem to miss is that the actual active layer of thermal compound should be extremely thin if the heatsink and device being mounted are flat.  This means even with the standard white grease the temperature drop will be very small, and the silver will give little improvement.

If the mating surfaces aren't flat then the silver stuff will definitely do better, but it would be better still with flat surfaces.

Exactly. My rule of thumb is, if the surface is flat and you can see it squish out around the edges, you've applied too much.

You'd be surprised that even the pros get it wrong! I recently took apart a 1980s vintage HP power supply and it seems like they applied the thermal grease almost randomly. One TO-3 transistor had what appeared to be a quarter sized squirt applied to both sides of the mica pad and just stuck on; it was everywhere around the transistor. The next TO-3 transistor had a tiny dab covering maybe 1/4 of the surface area. It wasn't an aftermarket repair, either.

Tek is the only company I've see do a proper job applying thermal paste.

There are multiple classes of thermal compound:
- Metal-Oxide material paste, cheap white stuff, electrically non-conductive.
- Metal material paste, expensive grey stuff, electrically conductive.
- Stickers, basically buttery type stuff that will melt and settle when heated the first time. Often pre-applied on pc heatsinks. Sometimes it also glues a bit.

For added electrical isolation you can use:
- Mica, horrible thermal conductivity, but it's cheap.
- Sil-pads, reasonable thermal conductivity, not very good though. Also cheap.
- Aluminum oxide plates, cold white material. Very good thermal conductivity (that is why is feels cold). You can combine these with oxide based paste. They are a bit thick, because the oxide is fragile. You can get them sub-mm, but they crack very easy. Get them at >1.5 mm. A 5mm one still performs better then a sil-pad. But they are expensive to buy and mount.

Note: there also is a maximum torque at which you can fasten an to-220. Over-tighten it and it will gap, the compound cannot fix this. See appnote: https://www.fairchildsemi.com/application-notes/AN/AN-4166.pdf.

Some metal pastes aren't electrically conductive. Silver based pastes (Arctic Silver, Silveria, etc.) are but alumina-oxide pastes
 (Arctic Alumina, MX-4, etc.) aren't. There are also silicone based pastes that are non-conductive and perform well.

Sil Pads may not perform as well as straight paste, but they generally perform better than mica insulators + paste. They also have a burn-in period (up to 100 hours, much lower at higher temps) wherein they form to the surface, providing better contact and filling gaps. (This is one reason it's recommend not to reuse them.)

I've never used the ceramic (alumina) plates, but everything I've read certainly says they're great. The only issue I see is having to still use paste (to fill tiny scratches and voids in the metal) and the thickness. The latter is why I've never used them; most of the stuff I work on just doesn't have 5mm between the component and heat sink/case to work with.

Though, I am working on the design for an electric load and have been thinking of using them for mounting the FETs to the waterblock. Combined with an alumina based paste it seems to provide maximum thermal transfer while still providing isolation. (I may end up mounting the FETs directly to a small metal block and then clamping that block to the waterblock with a custom cut alumina sheet between the two,  not sure yet.)

[Halvmand]

Timb: I think you got it upside down. Mica insulators is roughly half as thick. I seem to remember thermal resistance is less to for a given thickness.
My personal experience tells me they're better to. I've changed the sil pads for mica in my tti power supply because the two 2n3055 reached - 110 deg c at worst case. Now they reach to about ~ 100 deg c with the mica.

[timb]

Timb: I think you got it upside down. Mica insulators is roughly half as thick. I seem to remember thermal resistance is less to for a given thickness.
My personal experience tells me they're better to. I've changed the sil pads for mica in my tti power supply because the two 2n3055 reached - 110 deg c at worst case. Now they reach to about ~ 100 deg c with the mica.

Nope, I meant it the way I said it. In your case, it may have simply been down to the type of thermal paste you used and how it was applied that net you the decrease in temperature. Or the Sil Pads may have been improperly installed.

I just measured a TO-220 Sil Pad and it's only about 30% thicker than a mica washer. Keep in mind too that the Sil Pad gets thinner when compressed.

I used to measure temperature as well when I first started using them and in every case it was always better than the mica + paste it replaced. That said, the paste was a lot of times poorly applied, so I suspect that if I would have used a new mica insulator along with a thin layer of Silver or Alumina thermal compound, I could do better than the Sil Pad.

But honestly, that's a lot of time and effort for a few extra degrees of cooling that's not really needed. Most TO-3 and quite a few TO-220 transistors will happily take well over 100c. I've seen power supply transistors run 25 years at 130c no problem.

[cdev]

Make sure its squeaky clean first. Spreading it thin in one application is difficult to get right without practice. I actually rub a little bit in first hard using a coffee filter, then scrape most of that off, and then put a little dab on and spread that. You should then test it with a stress test program, currently I use a program called "cpuburn" ( https://patrickmn.com/projects/cpuburn/ ) on Linux.

I have a tube of this that has lasted me for a while http://www.arcticsilver.com/ceramique.htm

[Gyro]

Timb: I think you got it upside down. Mica insulators is roughly half as thick. I seem to remember thermal resistance is less to for a given thickness.
My personal experience tells me they're better to. I've changed the sil pads for mica in my tti power supply because the two 2n3055 reached - 110 deg c at worst case. Now they reach to about ~ 100 deg c with the mica.

Mica washers, as supplied, are normally far too thick and also tend to be different thicknesses between suppliers. It's very easy with a razor blade to cleave them into several thinner ones (obviously taking into account the required isolation voltage, Mica has a dielectric strength of 118kV/mm approx). This drastically improves their thermal resistance to way better than Sil Pads or Kapton. 

[uncle_bob]

Hi

The main reason all this stuff is "to thick" is that there is no absolute guarantee of perfect flatness (no burrs) on the parts you are mounting. The thickness is there to accommodate the "worst case" bit of crap that you might run into mounting the part.

Bob

[Deus]

Personally, if I *have* to use a thermal compound, I use Arctic Silver. Have for, damn, at least 10 years now? Works great as it has silver and ceramic particles in it, so it has much better conduction than your standard white goop. As a bonus, a little goes a very long way. (My original tube last several years and about a year ago when Radio Shack was closing, I got 5 packages for $1 a pop, so I'm set for life basically!)

What arctic silver are you talking about?
As Arctic Silver 5 is pretty expensive but surely isn't the(ir) best.
Their Ceramique2 is much cheaper, performs almost the same (+/- 1 a 2c difference only)

Arctic MX-2 and MX4 beat Arctic Silver 5 in both price and performance.
AS 5 also has to replaced every few (2-3?) years as it drie out.
Others last much longer.

Not saying AS5 (or AS 2 or 3) is bad, just that there are better and cheaper now.

There are better ones, but more expensive and only very small difference.
Not worth the price difference.

[Gyro]

Hi

The main reason all this stuff is "to thick" is that there is no absolute guarantee of perfect flatness (no burrs) on the parts you are mounting. The thickness is there to accommodate the "worst case" bit of crap that you might run into mounting the part.

Bob

Yes, I agree, surface preparation of both the heatsink and the device are important. This is the difference between one-off or critical applications looking for lowest possible thermal resistance and less critical volume applications where sticking in a Sil-pad and a clip will do. Even Sil-pads can be punched through by swarf or PCB burrs, in fact that's one issue I've seen with them in screw clamped applications with TO-220 packages (well not quite puncturing) - because all the clamping pressure is at the tab, the body tends to lift a bit with the compliance and the top corners of the tab tend to dig-in over time. Compliant insulators, in fact all insulators, work better with more central pressure from a sprung clip or clamping bar.

[station240]

I hate heatsink compound, so hard to clean it off things like heatsinks, ICs, my fingers when I accidentally touch it. 10x worse when it's dried out.

I usually use paper to wipe it up when it's still wet, takes a few passes but it also polishes the metal heatsink as a bonus.
Not figured out a good way to remove the dried stuff, that doesn't involve solvents that make me ill.

[timb]

Personally, if I *have* to use a thermal compound, I use Arctic Silver. Have for, damn, at least 10 years now? Works great as it has silver and ceramic particles in it, so it has much better conduction than your standard white goop. As a bonus, a little goes a very long way. (My original tube last several years and about a year ago when Radio Shack was closing, I got 5 packages for $1 a pop, so I'm set for life basically!)

What arctic silver are you talking about?
As Arctic Silver 5 is pretty expensive but surely isn't the(ir) best.
Their Ceramique2 is much cheaper, performs almost the same (+/- 1 a 2c difference only)

Arctic MX-2 and MX4 beat Arctic Silver 5 in both price and performance.
AS 5 also has to replaced every few (2-3?) years as it drie out.
Others last much longer.

Not saying AS5 (or AS 2 or 3) is bad, just that there are better and cheaper now.

There are better ones, but more expensive and only very small difference.
Not worth the price difference.

Up to my early 20's I was a PC gamer, so naturally built my own gaming rigs. Around the time the P4 came out, AS5 was pretty much the best available if I recall. A single tube last me at least 5 years of doing my own systems, as well as systems for a few friends.

About two years ago, when Radio Shack was closing down most of the stores here in the US, I bought 6 tubes of AS5 for like $1 each, which is the only reason I've got it on hand now.

Though, for restoring vintage test gear, if I can't use Sil Pads, I use Arctic Alumina, because it's non-conductive and has a low capacitance; important properties when you're spreading it around pins like on a TO-2 package.

I use a lot of Arctic Alumina Thermal Epoxy as well for bonding heat sinks to the top of IC packages or the bottoms of PCBs.

But yeah, you're right, there are better formulations. I was just trying to point out there were much better alternatives to standard white goop.

I hate heatsink compound, so hard to clean it off things like heatsinks, ICs, my fingers when I accidentally touch it. 10x worse when it's dried out.

I usually use paper to wipe it up when it's still wet, takes a few passes but it also polishes the metal heatsink as a bonus.
Not figured out a good way to remove the dried stuff, that doesn't involve solvents that make me ill.

Use hot water and a *non-scented* degreasing dish liquid (Dawn here in the US). Fill the sink up with hot water and a tablespoon of the dish detergent and let the heat sink and/or part soak for 10 minutes. Then, scrub with a small nylon soft bristled brush, the dried paste will come right off. Rinse throughly with clean tap water followed by either distilled water or Isopropyl Alcohol (70% will work fine). Drying Times: Distilled Water; let dry overnight, 70% Iso; 3 hours, 90% Iso; 1 Hour.

Now you have parts ready for new thermal compound.

(If you need a source for the brush: In the US, go to the sporting goods section of Walmart, in particular where the shooting supplies are, you'll find a pack of "Gun Cleaning Brushes" that contains three brushes, one with brass bristles, one with harder white plastic bristles and one with softer black nylon bristles, the latter works extremely well for cleaning electronics, as it doesn't generate static so you can use it on the CPU/transistor/IC you've got soaking in the water; the white plastic or even brass brush can be used on the heat sink if the paste is really stubborn, but I've never seen it take more than a few seconds with the black brush).

[Halvmand]

Nope, I meant it the way I said it. In your case, it may have simply been down to the type of thermal paste you used and how it was applied that net you the decrease in temperature. Or the Sil Pads may have been improperly installed.

I just measured a TO-220 Sil Pad and it's only about 30% thicker than a mica washer. Keep in mind too that the Sil Pad gets thinner when compressed.

I used to measure temperature as well when I first started using them and in every case it was always better than the mica + paste it replaced. That said, the paste was a lot of times poorly applied, so I suspect that if I would have used a new mica insulator along with a thin layer of Silver or Alumina thermal compound, I could do better than the Sil Pad.

But honestly, that's a lot of time and effort for a few extra degrees of cooling that's not really needed. Most TO-3 and quite a few TO-220 transistors will happily take well over 100c. I've seen power supply transistors run 25 years at 130c no problem.

I'll agree that most times the mica and paste is not worth the hassle. But those times you're limited in space and heat sink 5 - 10 degC is a lot.

The thinnest I've found on digikey is 0.051 for the mica and 0.080 mm for the sil pap. I don't think they get thinner as you tighten though. If they get thinner, they should squish out the sides of the component, they don't do that.