Proper group for heatsink question?

Hi. This is an appropriate group.

A good newbie starter is GC Electronics P/N 10-8109 Zinc Oxide and Silicone-based Heat Sink Compound. It's got a silicone and zinc oxide base, it's a highly visible white, and it's relatively inexpensive. This will allow you to do a visual inspection as you develop skill in applying the right amount of compound between the IC and the heat sink.

Make sure both the surface of the IC and the heat sink are absolutely clean, and that the heat sink is free of burrs, warping, or other visible imperfections. I like to do an alcohol wipe with a lint-free cloth before application of heat sink compound to remove any dust or grit.

Apply a very small dab of the white zinc oxide/silicone heat sink compound to your very clean finger (or ideally, with a fresh finger cot over your index finger) . Rub the compound over the mating surface of the IC, such that you can barely see through the heat sink compound to the surface. Then evenly place the heat sink on the IC, and evenly press down with the clip or fastener.

Now that you've done the work, undo it. Carefully remove the clip or fastener so that the surface of the heat sink doesn't slide around on the IC. Then pull it straight off. Look at your work. You should see a kind of fine fish-scaling of the heat sink compound across the entire surface of the heat sink like lacework, with a tiny amount of the compound extruded out past the edges of the IC. If you don't see the fish-scaling, you (probably) put too much compound on the IC. It's always way too thick the first time. Start over and do it again, first wiping the IC and the heat sink clean with alcohol and the lint-free cloth (it's sometimes better to just scoop off most of it with a paper towel first, then use the alcohol and the lint-free cloth).

Once you do this successfully a couple of times, you'll be able to do it well consistently in the future. It's like riding a bike -- you don't forget. Practice with each type of IC and transistor package. If you use a TO-3 or other heat sink that requires multiple screws, you should tighten evenly a little at a time on each side, ideally using a torque indicator to get the screws tightened down evenly. Just remember that way too much heat sink compound is actually worse than none. And by the way, make sure to clean off the heat sink compund carefully -- it's not good for you.

It also might be a good idea, once you're in practice, to use another type of heat sink compound if your application is high reliability or you're dealing with a lot of heat. Plain zinc oxide-based heat sink compounds tend to dry out and lose their effectiveness over time (years). Others tend to extrude out over time with thousands of thermal expansion-contraction cycles. Once you're in practice, and if you need the reliability, there are a number of translucent silicone-based compounds that have good thermal conductivity if applied properly, will not dry out over time, and are viscous enough to stay put. You might want to look at

for a variety of good products, and some good technical papers and app notes.

And of course, there's always the elusive search for the perfect heat sink compound. The last time I was walking along that path, I started seeing Monster Cables and other audiophile excesses lurking in the bushes, so I turned back to avoid brain cramp. But I've heard there are a number of very expensive silver-loaded heat sink compounds that have somewhat better thermal conductivity than silicone-based, and measurably better thermal conductivity than plain-Jane zinc oxide-silicone. For applications right on the bleeding edge (where the extra 0.2 degree C/W makes all the difference), you just might need something this good. But usually, you just put in a bigger fan or go to water cooling instead.

You might get a more specific answer if you describe what you're doing, and some more about your project requirements.

Chris

the integrated circuit. I'm working on a contract/extern basis for a well-known company that makes oscilloscopes and data probes. They are planning for the future with this R&D project- currently, their DAQ devices need to be able to dissipate about a Watt of heat. My assignment is to design a thermal management system that can dissipate 5W of heat in a quite small package (no larger than current data probes). I have a solution and design that will work, I need now to concentrate on test and build and retest; thermal compound is part of this thermal management equation and it's probably the part that I know least about, material-wise. I've done lots of reading...not only does this have to work but it has to be (reasonably) economical to manufacture. I've so far optimized the size and type of heatsink, figured out the rest of the management package, and need now to find a really good thermal compound that's not horribly expensive. regards, karinne

what else would you like to know? I can't discuss the exact solution we've come up with,(NDA stuff) but I can probably specify other things. For instance, this package is .013*.013*.114 m in size. junction temp no higher than 100C, case temp no higher than 60C. Need to dissipate 5W.

Therefore, any extra cooling we can squeeze out of any part of this is essential. Your response is very helpful, and if you don't mind, I will print it to share with my other ME team member who has less hands on experience than I do with any electronic stuff, and I don't have a ton- limited to home PC builds and hobby RC and robotics projects.

can you indicate why "too much is worse than too little"? We need to minimize contact resistance as much as possible; we have access to a totally kick-ass machine shop for the prototype of the heatsink. Water cooling is not an option, unfortunately, nor is thermoelectric cooling, but we've come up with something that should work; we now need to prototype and test.

I've done this with home PC's, but nothing more demanding than that; I need something that performs really well for reasonable cost, that I can access an amount of to use in testing.

and...aavid is a great site...as is another one I found,

oops- I thought that said "Whats *AN* IC?" We don't know. We are designing a general thermal solution for future need. We can assume size of about 1 cm^2 to 1.2 cm^2 for the chip itself. regards, karinne

this might help:

What's the IC ?

Graham

I worked with an Infiniband switch from Agilent that had an absolute max junction temp of 110C (but it quit operating properly at 100C anyway). That was a fair sized device (about 28mm x 28mm square) flipchip, so junction to case/heatsink thermal resistance was almost 0. Had a big heatsink on it as it dissipated about 18W, 12 of them inside a 1U enclsure, so managing even the ambient temp was painful. Obviously, that system had fans.

I agree that without knowing the device for junction to case thermal resistance and ambient, the OP will have an impossible time designing heat management for it.

Cheers

PeteS

Oh. Homework. Should've said so. Try sci.electronics.design.

Good luck Chris

You can ? How so ? What package would that be ?

It sounds like you're seriously in the dark about thermal management to me.

What is it you're actually trying to acheive ?

'General solutions' makes it sound like someone hasn't a clue what they're doing.

You might care to check out a heatsink manufacturer's range. See Aavid for example

Graham

The OP hasn't got an IC that you can screw / bolt to, so the above is irrelevant.

Most likely they need some thermal epoxy.

Since the OP doesn't even know the package dissipation there's very little help that can be offered.

Either the Op gives some relevant info or shuts up.

Graham

What package is that ?

Unusual spec. Normally it's 125C or 150C. It would help if you told us what the IC actually is.

Irrelevant. The difference between case temp and junction temp is set by the junction-to-case thermal resistance. The junction temp is the one to target ( case temp is not important ).

You'll need to specify the max operating ambient temp too.

Graham

From that page:

"Today, there are far more advanced thermal pads available, made by companies such as Power Devices, Bergquist or Chomerics, to name only a few. For links to the web sites of these companies, check out the links page. The performance of these pads can be roughly equal to standard thermal compound."

That is, in nearly all situations, grossly untrue. The thermal pads, or the phase-change stuff, wind up being 2-10 mills thick, even if you manage to apply insane contact pressures. Regular thermal silicone grease or thermal epoxy will squish down below 100 microinches with mild pressure, so will have an interface thermal resistance of 1/20 or more likely 1/100 of the pads. At, maybe, 2% of the cost.

And all the thermal pad vendors lie their asses off on thermal resistance specs. If you want to blow up power fets, use Bergquist pads.

John

They've given a fairly large grant for this to be considered 'homework'. The fact that I work at a university rather than in industry does not automatically imply homework. I am simply the ME on the team. I don't know as much about some of this as the 3 EE's on the team, regarding the electronics. My work is in thermal-fluid sciences, so that's what they've got me for.

I know more specifics on this device, but am constrained from discussing some of them. Therefore, any vagueness or confusion due to such is completely my fault.

Thought I could get some general suggestions here for a thermal compound that needs to function in an enclosed space of prev. posted size at a T_amb, enclosure of 60C, T_amb, external of 20 C, T_j of 100C, etc. - that's all I was looking for. I did get a few good suggestions, but on the whole, it appears that this was not the right group to query. Will take your (generic) advice, shut up, and go elsewhere.

regards

kw

How can you do that without even knowing the ambient temperature ? Have you made one and measured its thermal resistance ?

What's this 'management package' ?

You simply want thermal transfer 'grease' ?

Graham

No good deed goes unpunished -- didn't really say that at all. First, I gave you quite a bit of _very_ specific advice for a newbie or hobbyist, which is the assumed audience for s.e.b., and could have been assumed given your lack of specificity in your original post. Something vague like your original post is charitably assumed to be the result of lack of knowledge rather than lack of effort. If you'd spent more than two minutes on your original post (from further posts, there obviously was quite a lot more you _could_ have said without bumping up against non-disclosure), or read the first couple of paragraphs of my first response, you'd see that and not get offended.

Second, your problem extends a little beyond just heat sink compound (5 watts dissipated by an 0.5" X 0.5" X 4.5" heat sink in free still air keeping t(j) below 100C and t(sink) below 60C). This seems to be a very ambitious engineering puzzle, for which there is no easy answer. Several engineers at s.e.d. have described things they've done with heat sinks that approach witchcraft, and getting their advice might be a good idea.

Also, there happened to be some good advice on the heat sink compound to use in my first post. If you'll look at the website I suggested, you'd see the Ultrastick phase change product, which has thermal conductivity ratings better than the others, and looks like it might be a good starting point (and will probably be cost competitive with the higher-priced silver-loaded stuff). With something like this, it might be best to start with the best, then see what kind of margin you have to play with (unfortunately, I have the suspicion that your margin is negative already, and getting a really good heat sink compound will just eliminate that as a source of dither). The manufacturer also will provide samples, especially if they're talking to someone who is a potentially good customer.

Nobody told you to shut up, although I still feel you might get better advice elsewhere. Sorry you feel that way.

(By the way, look closely at your mathematical modelling, check your assumptions, and make sure you can actually accomplish what you want first. I'm not too sure even a perfect heat sink of the size you specify could do the job you need in non-computer-modelled free still air.)

Good luck Chris

That was the *space* and it's enclosed ?

You don't have a hope in hell of dissipating 5W with a delta T of a few 10s of degrees in that space.

You still haven't even mentioned if the 'thermal compound' has to provide mechanical adhesion ( as in a glued-on heatsink ) or is simply required as an interface filler where the mechanical aspect is provided by nuts and bolts or clips..

If you simply *won't* provide any meaningful information *no-one* can help you.

It's clear to anyone who knows about these things that the minimal info required is in no way going to create trouble over NDAs or whatever.

You're being deliberately abstruse.

Please don't repeat the same lame question in s.e.d. You'll get a similar respsonse.

Graham

Hey, John......

I realise I came over as being frustrated at the OP's lack of detail.

I have my reasons though.

The initial post suggested that it was just an interest in a thermal transfer compound but I dunno, I smelt something.... In any event it wasn't clear if it was a grease or an epoxy that was required ( hinted at ) and I don't like to give bad advice.

Read the follow-up posts carefully and the horrible truth finally comes out. The transfer compound is the last of their problems and likely only was though about on account of anecdotal comment ( e.g CPU coolers ) . The following data I kind of extracted as best I could from the follow-up posts.

The OP's data aquisition module is 13m x 13mm x 114? mm

It contains an IC that's ~ 1cm^2 dissipating *5* Watts !

The IC is inside the module which is unventilated.

( There is simply no hope ever of cooling this IC as required )

In fact the determining characteristic thermally is the enclosure dimension since it's actually *the enclosure* that will need to dissipate the heat.

This explains the posted 'package' temp of 60C. The OP didn't understand that I was originally referring to the unspecified *IC package* and gave the data probe package dimensions instead it seems...

The reason for 60C is that IEC regs only allow a delta T of 40C for accessible parts and if the ambient temp is 20C then the temp of the probe package must not exceed 60C.

In short it was like trying to get blood out of a stone to ge this far.

The OP's DAQ seems destined for the big trash can in the sky since I know of no way ( I'm sure there is no way ) to get 5W out of such a small pacakge with a delta T of 40C.

Typical idiot academics. The first thing they should have considered was the thermal constraint.

I'll bet that IC has been designed around some power-hungry PLD or similar. They'll need to consider a new technology. In short, instead of ending, their project has merely begun.

Graham