Thermal conductivity calculation ?? for comparing thermal interface materials

"John"

** I think you are being tricked by figures that are non comparables.

The K10 material is soft and conforms to the shape of the device and heatsink surface leaving no air gaps - so the thermal conductivity figure is directly usable.

An Alumina insulator is hard needing thermal grease on both sides to fill voids and work efficiently.

IME with high powered TO3 devices - thin mica with a smear of thermal grease both sides is still the best.

........ Phil

Right.

Yup.

Just be aware that the thermal conductivity of AlO is about as stated, and that the Bergquist numbers are usually wildly optimistic, even if you apply insane mounting pressures. Another factor of 2 might be prudent.

The best thing, as a single insulator, would be aluminum nitride with grease on both sides. It looks pretty much like AlO, isn't toxic, but conducts heat about six times better. It's available in suitable shapes.

For reasonable voltages, we like to have aluminum heatsinks hard anodized 1 mil thick, and then just use silicone grease. That will be tens of times better than the Silpad or the thick AlO.

Best yet is to bolt the transistor to a copper heat spreader and insulate *that* from the main heatsink.

John

First off you should get Motorola's AN1040

Graham

I've heard this mentioened before.

I've had good results with Warth's synthetic 'rubber' insulators.

You don't have to worry about the grease 'drying out' either.

Graham

John,

The "m" in the thermal conductivity constant is really the ratio of the area to the thickness. The units come out as meters, but it's really square meters per meter.

Thermal resistance is in the form of degrees C (or Kelvin) per watt.

So, to get from thermal conductivity (let's call it therm_cond) to thermal resistance (therm_resist):

therm_resist = thickness of the material in question [in meters] / ( contact area of the material [in meters] * therm_cond )

For example, if your materials therm_cond is 1.3W/m-K, and the contact area is 9mm * 8mm, and the thickness of the material is 2mm, then:

therm_resist = 2mm / (9mm * 8mm * 1.3W/m-K) = 0.02 degrees C / watt

If your thermal conductivity is 15.06W/m-K and the material thickness and contact area is the same as above:

therm_resist = 2mm / (9mm * 8mm * 15.06W/m-K) = 0.0018 degrees C / watt

The temperature delta, for 15W transferred through the material, will be :

delta T = watts * therm_resist = 15W * 0.02C/W = 0.3 degrees C (for the 1.3W/m-K stuff)

= 0.03 degrees C (for the 15.06W/m-K stuff)

If you're putting the Bergquist and Aavid stuff in series, then the temperature delta, for the combination, will be 0.33 degrees C.

In order to really find out what the silicon temperature is going to be (this is the only thing that really matters), you need to know "theta jc" for the IC or transistor package, you need to know the "theta sa" for the heatsink (at the given amount of airflow), and you'll need to know the ambient temperature of the airflow.

The Bergquist website shows you how to do these calculations (iirc).

Bob

Thanks guys!

Phil, I forgot to mention that we would use our favorite phase-change compound (Aavid's Ultra-Stick) between any hard insulator and the device/heat sink.

John, I looked high and low for aluminum nitride insulators but could only find recommendations to use them. Do you know a source for TO-247AC insulators?

I really like the hard anodizing idea. Top priority is low thermal resistance, not cost. The Cool Innovations pin heat sinks we're using in the prototypes are plain (no plating or anodizing) and aren't offered with anodizing in the quantities we're looking for (100pc. lots). Do you have a recommendation for company that does hard anodizing?

We had considered using a copper heat spreader between the four TO-247AC devices and the 2.5" square heat sink (4 heat sinks per prototype) but they were already pretty well spread out on each heat sink. And we'd still need an insulator for the spreader. :-)

Graham, That On Semi app note is a nice one. Better than the ones from IR and TI that we have. Thanks!

I'll check the specs on Warth's insulators too.

-- remove SPAMMENOT for e-mail responses --

Ahhhhh....OK, now it's starting to make some sense. :-) We have the theta-jc for the devices, 0.32 degrees-C/W (IRFP2907), We have the theta-sa for the sinks, 0.22 degrees-C/W (Cool Innovations

3-252517R) for our selected fan.

We were pretty comfortable with the equations for figuring out the junction temperature for a given power level (just needed the theta-cs), but we were getting dizzy trying to convert the thermal conductivity numbers we had to thermal resistances (for the insulators). Thanks for taking the time to explain that.

Time to do a bunch of calculations!

-- remove SPAMMENOT for e-mail responses --

These people seemed willing to do AlN insulators...

I'll check when I get to work. If it's a cast pin-fin sink, you should probably machine it flat first, too, as they tend to be wavy.

Most anodized heatsinks are cosmetic soft anodize, not a reliable insulator. Make sure it's "hard" anodized, 0.5 to 1 mil thick, and burr-free.

2.5" square is pretty small. Do all four transistors always heat equally? If not, there's a big advantage to going to one bigger heatsink, rather than four small ones.

You do need an insulator for the spreader, but the footprint is potentially a lot bigger so theta is correspondingly less.

John

This equation isn't quite right - though it used more or less correctly below.

therm_resist = thickness of the material in question [in meters] / ( contact area of the material [in SQUARE meters] * therm_cond )

Pay attention to the units - either

therm-resist = 0.002m /(0.009m * 0.008 m * 1..3W/m.K) = 21.37 degrees Kelvin per Watt

or therm_resist = 2mm / (9mm * 8mm * 0.0013W/mm-K) = 21.37 degrees Kelvin per Watt

In fact the Berquist material is only 0.156 mm thick (0.006 inches), so the correct answer would be

= 1.67 degrees Kelvin per Watt

if Berquist's figures were to be trusted

By the same argument, actually 1.8 degrees C/watt

Using the corrected numbers

= 15W * 1.67C/W = 25.05 degrees C (for the 1.3W/m-K stuff)

= 27 degrees C (for the 15.06W/m-K stuff)

If you're putting the Bergquist and Aavid stuff in series, then the temperature delta, for the combination, would be 52 degrees C.

It's pretty easy to make conceptual slips in working out these sorts of equations - a reality check is often a very good idea.

--
Bill Sloman, Nijmegen

Damn! Looks like I picked the wrong week to stop beating my dog.

What's a factor of a thousand between friends, anyway?

Thanks for the correction.

Bob

I was running the equations last night and was wondering why the numbers were looking so incredibly great until I noticed that the K10 thermal resistance was almost exactly 1000 times smaller than all the other silicone pad numbers I had gotten. I added some zeros after that. :-)

Here are the numbers for my original two top choices for a TO-247AC device with a thermal pad area of 206 sq. mm:

Bergquist K10 silicone pad = 0.57 degrees-C/W Aavid 4180 Aluminum Oxide pad = 0.66 degrees-C/W

That Bergquist pad is pretty impressive.

Warth (now Laird Technologies) has an even more impressive pad, the KTP series (particularly the KTP127, 0.127mm thick), but it's been tough trying to find someone here in the States to get a quote from. Another day or two should do it.

The numbers for Aluminum Nitride insulators are incredible, but you gots to pay for it. About $7 a device...ouch. Pretty sure the performance won't be worth it, especially if we can hard anodize the heat sinks.

-- remove SPAMMENOT for e-mail responses --

I didn't think of looking into buying a sheet of AlN (perhaps the size of the heat sink). I just concentrated on finding TO-247 shaped pads. Thanks for the wake-up call.

I checked into those three sites.... Valley Design had some prices, it would cost about $6-$10 per device. Wow. The thermal resistance of AlN is incredibly low but I have to look hard at the numbers to see if it's worth the price. Good to know about though.

They're forged, according to Cool Innovations. Amazing that they can do that with the pins so close together. The sinks are sent already machined on the base and the tops of the pins.

I did a LOT of checking around and found dozens and dozens of places that do "Type III hardcoat anodizing" (I think that's what I'm looking for) with standard thicknesses of 1 mil surface coating, 1 mil penetration. Or thicker if I wanted. Lots of good info on the American Anodizers Council web site

But, a recommendation for an anodizer would be great, thanks. I'm not looking forward to cold calling a bunch of places looking for someone who understands what's needed.

The four transistors on each heat sink are spaced as best as we can and all run to within 1% of each other (power dissipation).

I should probably get out the calculator again. The spreader would get us cloer to the "ideal" heat source (spread evenly across the heat sink) but there would be another thermal interface between the spreader and sink and I'd want to make sure the cost and performance would be worth it.

Hmm...a large 1/4" thick copper spreader, the size of the 4 heat sinks, might be a good idea for mounting though. It would be easier to bolt the 4 sinks to the spreader and just have one solid assembly to attach the transistors to. And I'd get that lower theta too.

Maybe even thermal epoxy to attach the sinks to the spreader? Have to check those numbers too. Weren't computers supposed to make my life easier by now? All I'm doing every day is banging away at my HP calculator, working thermal equations!! :-)

Time to hit the McMaster-Carr web site for Alloy 110 prices!

Thanks again John

-- remove SPAMMENOT for e-mail responses --

Be careful about thermal epoxy. Most of the stuff I looked at was loaded with little glass balls to guarantee enough epoxy filled gap to provide a - fairly high - level of insulation.

I've used an elastomeric thermal adhesive which didn't insulate particularly effectively, but was thin enough to offer a low thermal resistance.

--
Bill Sloman, Nijmegen

We use Santa Clara Plating, in Santa Clara California; they do consistantly good work.

Fab drawings should say "Finish: deburr, heavy etch, 1 mil hard black anodize."

We pay something like a $75 lot fee plus a few dollars per part for small stuff.

Hey, these are slick:

*532maxclip07*&terms=532-maxclip07&Ntt=*532maxclip07*&Dk=1&Ns=SField&N=0&crc=true

John

Thanks for the reference John!

You read our mind with those clips. The heat sinks we have in mind don't support the standard clips (in a slot) but we were thinking of using the MAX07 (screw mount) clips. With 18lbs of force, they look good.

But, how necessary are clips for mounting a TO-247AC case? With the screw going thru the body of the TO-247, it's not going to pivot up like using the tab on a TO-220, is it?

Oh, wait, with the clip I can use the Super-TO247 (or whatever it's called) case that doesn't have the screw hole and gain another 10% decrease (or so IIRC) decrease in thermal resistance due to a bigger thermal pad.

Soooooo much to consider when you're trying to pack as mamny watts as you can into a given space. :-)

-- remove SPAMMENOT for e-mail responses --

We use both the regular and super 247's with clips. I figure not having a hole in the heatsink reduces theta even more! Hmmm, have to do the math on that claim some day.

The clips don't press nearly as hard as a screw, so aren't appropriate if you use a sil-pad. These clips *will* squash a big blob of regular Dow Corning-type thermal grease down below 100 microinches thickness, if the surfaces are that flat.

Oh, the solid phase-change things, the ones that look like sil-pads, suck.

Hard anodize does sort of mess up tapped holes. I think we tap them a tad oversize to prevent the screws from galling. You can chase the holes with a tap afterwards, but it wrecks taps.

This thermal stuff is a lot more complex than it first appears.

What's the gadget you're building?

John

That's somethng important for us to consider. If we didn't hard anodize, we'd go with the Bergquist K10 pads or AlN (expensive!) sheets. The K10 pads would then need to be screw mounted.

I wasn't really impressed with the specs myself, especially considering how great the Aavid UltraStick phase-change compound is.

What were your experiences with the phase-change pads?

I was wondering about that yesterday. We buy taps by the box full (and can afford to go through a few of them) but it would save a lot of time if we didn't have to.

An active load used for testing power sources. I have two of them going at 350W each, but I really need to get up to the 1KW+ range.

350W in a 5" cube wasn't too hard. Tripling that power level gets interesting! It might not be possible, without liquid cooling, etc., but a goal nonetheless. :-)

John

-- remove SPAMMENOT for e-mail responses --

They don't ever flow out from below the device, so they stay several mils thick. Theta galore.

John

< snip >

The hole isn't central. The header can indeed 'lift' slightly.

I like to use a *large* washer ( commonly generically called a bicycle washer ) when mounting TO-247 to distribute the pressure.

Graham