Everything You Know Is Wrong January 2002

Answers to those Doggone Thermal Design Questions

By Tony Kordyban

Copyright by Tony Kordyban 2002

 

Dear Tony,

Which is better, mounting a fan at the inlet of my chassis so it pushes the air in, or at the outlet, so it draws the air out? 

Hobson from Quandary, Minnesota

 

Dear Hobson,

Two weeks of holiday partying are over, I have to go back to work, and I didn’t get that “Dexter’s Laboratory” video game that I specifically asked Santa for, so I’m a little grumpy. So let’s just settle this “fans-on-the-inlet vs fans-on-the-outlet” question once and for all.  I have been asked this question about once a month for the last 12 years, and I am tired of giving the wishy-washy answer of “it depends on the details of your product design.”  Here is the clear, simple, unequivocal answer you are looking for:

Always mount your cooling fans on the inlet side of your electronics chassis.  Always.  ALWAYS!*  It allows you to direct the highest velocity air to the areas of greatest need, the cool inlet air keeps the fan motors from overheating, and the higher air pressure inside your chassis forces dust out through any accidental leaks, rather than sucking dust in.  Those reasons should be good enough for anybody!  And  stop ending your sentences with prepositions!

*Always, unless, of course, you have a good reason to do it the other way.  Some possibilities for “good reasons” are:  you prefer the more uniform, laminar flow found on the inlet side of a fan; you don’t want the heat dissipation from the fan motors to pre-heat the air flowing over your delicate electronics; there is no room for a fan on your front panel, and you don’t wish to exhaust air into the face of the customer.  The overall air flow rate is pretty much the same either way.

—————————————————————————————————————

Dear Mr. Hot Air,

The heat sinks on display at the trade shows are always red or blue or gold or black.  The cute guys in the booths tell me that this improves the radiation.  What’s the best coating?  Should I use black anodizing, or should I go with black or white paint?   The expected life for my product is 30 years, so I am concerned about corrosion protection, too.

Cindy from Cincinnati

 

Dear Cindy,

Let’s settle this heat sink coating thing once and for all (sorry, my post-holiday headache has not gone away yet.)  My mom was frugal.  She would wash the plastic forks and spoons from family picnics to re-use the next next year.  Of course we always forgot to bring them and had to buy a new set on the way to the picnic site, but the point is that I was raised to be cheap.  My favorite heat sink coating is NONE.  That’s right, just bare aluminum.  (I’m assuming aluminum here, since you didn’t mention your material.  Copper or other materials are another story.)

It’s true that a clean, fresh, shiny aluminum surface is terrible at radiating heat.  It’s emissivity is something like 0.01 or lower, depending on how well-polished it is.  But I don’t care.  Why?  Here are a couple of reasons a cheapskate would love:

1.  After some amount of time (from a few days to a few months) the unprotected surface of aluminum corrodes to form a very thin layer of oxide (some chemist will probably correct me on that.)  But the corrosion layer is self-protecting.  Once the layer forms, corrosion stops.  You have probably seen it.  The aluminum looks dull, as if it has been rubbed with soap.  This layer has an emissivity of about 0.5 or 0.6.  That’s not too bad, considering it’s free.  Paint might get the emissivity up into the range of 0.7 to 0.9 (1.0 being a perfect emitter of thermal radiation.)

2.  I don’t care about radiation from heat sinks.  I design mine to get rid of heat by convection to the air.  If there is any additional heat loss by radiation, fine, that gives me a bit more margin.  Why ignore radiation, though?  Isn’t that “free” heat transfer that could make my heat sink a little smaller?  Most of my circuit boards these days are cooled by forced convection, which means that radiation accounts for something less than 10% of the total heat transfer.  The faster the air flow, the less important radiation is.  Then there is the problem that your heat sink has to have something to radiate to, which is cooler than the heat sink.  If your heat sink is exposed to the outside world, or at least to the relatively cool cover of the chassis, then maybe there will be a big enough temperature difference for radiation to have some effect.  My heat sink seems to always be facing another hot circuit board.  If that other board is hotter than my heat sink, then radiation may actually make my heat sink hotter, not cooler.  The last problem is that a heat sink designed to work well in forced convection (lots of fins close together) is not optimized for radiation, which wants all of the surfaces facing outward (not towards each other).

So when you ask me which coating is the best for radiation, I shrug my shoulders.  One coating probably has a better emissivity than the others, but I think the overall effect it will have on component temperature will be insignificant (unless you are cooling in a vacuum.)  I will also guess that after 10 or 20 years of service, the coating of dirt on your black anodized heat sink will give it the same emissivity as my dirty, corroded, bare aluminum one.

By the way, does anybody want 37 jars full of used (but clean) plastic knives, forks and spoons?

—————————————————————————————————————

Dear Mr. Kordyban,

You are my hero.  I know you’ll settle this office dispute, which has led to a black eye, a split lip, and one severe Silent Treatment.  Can you tell us what is the difference between Typ Power and Max Power for a component?  Which one are we supposed to use in CFD thermal simulations, or even estimating junction temperature by hand using the package thermal resistance?  Sometimes it makes the difference between a feasible design and an infeasible one. 

Theodore Cleaver from Beaverton, Oregon

 

Dear Beav,

Hero?  Breaking up fights?  What am I, your father?

Don’t worry, I am not mad at you.  I agree with you.  This business of Typical and Maximum Power from component data sheets is very frustrating for us thermal engineers.  It is frustrating, because there is no simple answer I can give you that would be right.

But since I am being Mr. Cranky today, I will give you a simple answer.  Use Typical Power and most of the time you will be closer to the truth than if you use the Maximum Power value.  If this answer satisfies you, then you deserve the trouble you’ll get by following this rule of thumb blindly.  If it makes you suspicious, please read on.

Typ and Max Power are listed in data sheets for the electrical designer to size the power supply, fuses and traces in the board.  They are not there to help the thermal engineer.  The component manufacturer feels it has done its duty by giving, perhaps, a gross overestimate of the power consumption.  It’s safe to draw too little power, in their view, but a bad problem if the part draws more than the data sheet claims.

That is why you often see in data sheets a value for Max Power, but under Typ, just a blank.  Typical varies all over the place for lots of reasons, so it’s hard to give a single number.  But they can always find a power value so high that 99.999% of all the parts will never exceed it.  Even if 99.99% never get close to it in actual use.

Why does power vary?  Understanding that will help you decide which brand of power rating to use for CFD calculations.  But it depends on the type of component.  Using the simplest component as an example, power dissipated by a resistor depends on two things:  how it is used in the circuit, and how much its resistance varies.  Nominally, its power only depends on the voltage applied across it, which may not be constant with time.  In addition, its resistance can vary due to manufacturing tolerances, its operating temperature, age, and so on.  The same can be said of any component.  There can be variation due to how the part is applied in a circuit.  Then there is variation among components with identical part numbers.  Under the exact same conditions, their power may not be the same.

This gives me headaches bigger than the one I have right now.  I once had to size a heat sink for a custom part made using Gallium Arsenide (GaAs).  This chip technology had a particularly wide process variation.  The chip designers predicted it would have a Typical Power of 1.5 watts (W), but that to get acceptable yield, the worst case power would be allowed to vary all the way up to 2.8W.  I did my best to design a heat sink for 2.8W.  But how could I test it under worst case conditions to see if it worked, when all the chips they gave me to test drew only  0.8W?

Luckily for me, this question of Typ or Max Power is only important for a handful of components on any particular circuit board.  Run-of-the-mill digital logic parts are so low in power that I don’t worry about getting their temperature very accurately.  I do have to pay more attention to relatively high-power components, or those that are very sensitive to temperature.  For them I try to beat up the electrical designer and the component vendor to get what I call “realistic power”, which might be abbreviated Real. Pow.  By realistic power I mean the power that will be dissipated by a component when the product is in use by the customer in a realistic, worst case thermal situation.

But what is realistic?  Let’s take a memory device as an example.  An SRAM might draw 100 mW when it is sitting idle, but when accessed by the processor might take 3W, at least for a few microseconds.  The conservative designer would tell me the SRAM draws 3W continuously, because he knows it can never draw more than that.  But in reality, the processor does not access every memory chip on every clock cycle.  It has a duty cycle less than 100%, that depends on the system architecture, and how the software works.  My experience is that memory is accessed less than 10% of the time, so that the “realistic” power is closer to 300 mW than to 3W.  Don’t quote me on that.  Maybe my electrical guys just design in way too much memory.  But for memory I never use the Max Pow from the data sheet.

How about a fiber optic laser transmitter with a built-in Peltier cooler?  The Typ Pow might be listed as 500 mW (at 25°C ambient), with a Max Pow given as 5W (at 70°C ambient).  Is this the same situation as the SRAM?  Not at all.  The Max Pow for this device has nothing to do with duty cycle and everything to do with the Peltier cooler.  It is there to keep the laser diode at a fixed, low temperature.  At 25°C ambient, the diode doesn’t need much cooling, and the power drawn is that needed to run the laser alone — the Peltier cooler is not active.  But at higher ambient, the Peltier cooler switches on and starts drawing more power.  The increase in power from Typ to Max is all used by the Peltier cooler.  For this laser transmitter I would definitely use the Max Pow value, because it would “really” be dissipated at the maximum ambient.

These are two examples of components that I understand.  I wish I could give you a complete list of component types with recommendations for estimating Real. Pow.  Unfortunately, I don’t know how most components work, how their power varies with application, so I am in the same boat as you.  I am at the mercy of the electrical designers and component vendors.  I think the best you can do is to explain to them that if they want a “realistic” temperature prediction, they have got to provide you a “realistic” power value, not the Typ or Max.

That should settle the Typ vs. Max question once and for all!  Or at least I hope this explanation works better for you than this aspirin is working for me right now.
—————————————————————————————————————

Isn’t Everything He Knows Wrong, Too?

The straight dope on Tony Kordyban

Tony Kordyban has been an engineer in the field of electronics cooling for different telecom and power supply companies (who can keep track when they change names so frequently?) for the last twenty years.  Maybe that doesn’t make him an expert in heat transfer theory, but it has certainly gained him a lot of experience in the ways NOT to cool electronics.  He does have some book-learnin’, with a BS in Mechanical Engineering from the University of Detroit (motto:Detroit— no place for wimps) and a Masters in Mechanical Engineering from Stanford (motto: shouldn’t Nobels count more than Rose Bowls?)

Tony Kordyban head shotIn those twenty years Tony has come to the conclusion that a lot of the common practices of electronics cooling are full of baloney.  He has run into so much nonsense in the field that he has found it easier to just assume “everything you know is wrong” (from the comedy album by Firesign Theatre), and to question everything against the basic principles of heat transfer theory.

Tony has been collecting case studies of the wrong way to cool electronics, using them to educate the cooling masses, applying humor as the sugar to help the medicine go down.  These have been published recently by the ASME Press in a book called, “Hot Air Rises and Heat Sinks:  Everything You Know About Cooling Electronics Is Wrong.”  It is available direct from ASME Press at 1-800-843-2763 or at their web site at http://www.asme.org/pubs/asmepress,  Order Number 800741.