Everything You Know Is Wrong December 2002

Answers to those Doggone Thermal Design Questions

By Tony Kordyban

Copyright by Tony Kordyban 2002

Dear Doggone Thermal Answer Guy,

Enough with the suspense already!  Last month you told us that Theta j-c was no good for figuring out the junction temperature of a component.  You really blew it out of the water, or should I say, the stirred, constant-temperature bath.  Then you get to the punch line, “So how SHOULD we figure out the junction temperature?” 

But you weaseled out and didn’t give an answer.

So now I’m asking, how do you figure out the junction temperature of a component?

Chip from Silicon Valley



Dear Chip,

Here are the reasons why I didn’t give the answer before:

  1. I don’t want you to do something just because I do it that way.  Maybe I seem like a credible authority in the field of Electronics Cooling.  So what?  Technology isn’t like writing a term paper in English lit.  Engineers don’t cite recognized authorities to prove something — we go into the lab and test it to make sure it works.  Just because Tony figures out Tj (junction temperature) a certain way doesn’t make it right.  I am trying to get you to understand the physics of heat transfer in electronics applications, not to memorize some official procedure.  Take the hints I have given you in my resistor diagrams and spreadsheet and figure it out for yourself.  That will be much more valuable that me just telling you how I do it.
  2. My method of determining Tj is so novel and valuable, that my employer considers it a trade secret, and I am not allowed to publish it.
  3. The above two reasons are just just smokescreens to cover up the fact that I don’t have a good, simple way to find Tj accurately (although the first reason is still true!)  It’s pretty easy to demolish something that is flawed, but it is hard to build something better to replace it.

So that’s the real reason I didn’t give an alternative to using Theta j-c for figuring out Tj.  There isn’t one that is as simple as multiplying the total power by Theta j-c and adding it to the case temperature (Tc).  What are the candidates for an alternative method?

Computer Modeling.  Computational Fluid Dynamics (CFD) or Finite Element Modeling (FEM) software tools can be used to build highly detailed models of component packages.  You can include as much internal detail as you like (assuming you can find out the internal construction of the packages from the component vendors), and then you can plop them down onto a detailed model of your own circuit board in you own chassis.  If you get the geometry, material properties and power dissipation right, these computer tools can calculate junction temperature for you directly.  This method, I think, has the promise of giving us the temperature information we need, and conceivably could be extremely accurate.  The main drawbacks are:

  1. building all the detailed models for hundreds or thousands of components, and then producing the solutions, is a heck of a lot of work
  2. component vendors are leery of revealing the internal details of their packages, perhaps for good business reasons.  They don’t mind sharing a few generic numbers, like Theta j-c, but they don’t want to give away their manufacturing secrets
  3. there are still doubts about the accuracy of CFD and FEM in practice, for a variety of reasons, ranging from the skill level of the tool user, to large uncertainty in the input data, to the choice of turbulence model.

Summary:  Computer modeling might give you accurate Tj, but it is a lot of hard work and requires a lot of information you probably don’t know.

Measure Tj directly.  This obviously solves the problem of calculating anything.  But for the typical component package, this is still difficult and expensive to do.  You might be able to drill a hole into a component package without wrecking it, and attach a tiny temperature probe directly to the die.  If you have a lab capable of doing that kind of microsurgery on working components, then I should be writing to you for advice on Theta j-c, and not the other way around..

I have argued in the past that component vendors should design-in temperature sensors on all their new chips, and bring out the leads so we can read junction temperature directly while the chip is working in our applications (see my article “Thermal I/O” in Electronics Cooling magazine.)  The idea has yet to catch on, since it reduces the work for only a few thermal engineers, and nobody else.

Some very large, high power components do have thermal sensors on board.  I am thinking of Intel’s Pentiums, which report their own junction temperature to the computer operating system.  This allows the computer to shut itself down in the case of a cooling system fault.

Summary:  direct measurement would work and can be very accurate, but it is tricky, expensive, and probably not practical for most thermal engineers.

Keep using Theta j-c, but be aware of its limitations.  Let’s recall what the original question was from last month.  Is Theta j-c “conservative”?  When you use to Theta j-c to estimate Tj, can it ever be too low?

Buffaloed from Upstate New York was not trying to get an accurate value of Tj.  He was trying to establish an upper bound on the range of its value.  Suppose I have a component with a temperature limit of Tj =100C.  If Theta j-c is “conservative”, and it gives me a worst case estimate of Tj of only 90C, then I know everything is fine.  I don’t really care if the actual  value of Tj is only 75C.  Sure, my estimate of pretty poor, but it is safe.  But it is only safe if I am guaranteed that Theta j-c is “conservative”, that it never predicts too low.

But my article last month proved that Theta j-c is NOT guaranteed to be conservative.  And it was pretty easy for me to find examples of package designs and boundary conditions that would give  predictions of Tj that were too low.

But we can use the kind of analysis I did last month to give us some general guidelines for when Theta j-c is conservative and when it might not be.  When we suspect it isn’t, we can resort one of the other methods to get a better handle on Tj.

Here is the basic idea:  for a typical component package in the JEDEC test that measures Theta j-c, the package is very small compared to the circuit board to which it is soldered.  Just because of that, most of the component heat goes down into the board, and only a small fraction goes out through the top of the package, where case temperature is measured.  The temperature rise from junction to case depends on how much heat goes out through the top, assuming that Tc is measured on the top surface of the package.

In most ordinary applications, most of the heat will still go down into the printed circuit board.  When that is true, Theta j-c will still give a pretty safe estimate of Tj.  It is only when the amount of heat going through the top is a bigger fraction than it was in the JEDEC test that we get into trouble.  The more heat that goes out through the top, the less safe Theta j-c tends to be.  Let’s look at a few examples:

Vacuum (space applications).  When there is no air surrounding the circuit board, no heat can leave the top surface of the component package.  All the heat has to conduct into the board, and from there to the chassis.  With no convection from the top of the package, the case temperature must be equal to Tj.  Theta j-c is very conservative here.  Exception:  This doesn’t apply if radiation from the top of the component is significant.

Natural convection.  When there is no fan cooling, the heat loss through the top is so small, that the case temperature is probably almost the same as Tj.  Theta j-c should be very conservative.

Fan cooling.  When forced air flow across the top of the package is introduced, the heat flow through the top of the package increases.  I think with moderate air velocities (whatever that means) Theta j-c is still conservative.  But as velocity goes up, the probability that Theta j-c is safe goes down.

Heat sink.  The whole point of adding a heat sink to the top of a package is to increase the heat flow from the top of the package.  This increases the heat flowing from the junction to the case, and so it increases the temperature rise from junction to case.  When large heat sinks are combined with forced air flow, it seems very likely that Theta j-c will give a value of Tj that is too low.

Jet impingement, heat pipe, thermoelectric cooler, etc.  Any cooling enhancement that is applied directly to the top of a component, and not applied to the circuit board at the same time, is going to throw off Theta j-c. Don’t rely on Theta j-c to give you a safe estimate of Tj in these situations.

Here is the paradox.  Using Theta j-c is safe only in natural convection or when we don’t use heat sinks or high velocity forced convection.  But our most important problems are high power components, which need fans and heat sinks the most.  So Theta j-c is  useful only those components that we don’t care about already.

If you are planning to clamp a thermosyphon onto that 100 watt processor, I think you’d better find a better way of determining the die temperature than Tc + Q x Theta j-c.

Dear Tony,

I saw a “Heat Sink Pen” in an electronics supply catalog.  What is that for, drawing pictures of heat sinks on schematics?  Or do you autograph your book with it?

Clarence fromBedfordFalls


Dear Clarance,

Your question reminds of of that old children’s book about the kid with a purple crayon.  Whatever he drew on a wall became real and came to life, and caused all kind of problems that could only be solved with ever more clever uses of the purple crayon.  I think pre-school is a bit early to start scaring kids about the Pandora’ Box aspect of new technology.

The Heat Sink Pen is a perfectly good product, with a perfectly stupid name.  It should be called a “thermal grease pen”, because that is what it dispenses when you draw with it.  Thermal grease is an interface material commonly used between heat sinks and components, to improve the contact resistance.  A pen is a convenient way to apply it, if you are doing it in small quantities.

Thanks for mentioning my book.  I usually use a Sharpie razor tip marker for autographs.  It writes well on any surface —  a book, a photograph, or even the face of a most enthusiastic fan.


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?)

In 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.