Everything You Know Is Wrong July 2004

Answers to those Doggone Thermal Design Questions

By Tony Kordyban

Copyright by Tony Kordyban 2004

Greetings, O Cooler of Circuits,

We make a home entertainment product that normally sits on top of the customer’s TV set.  The AdUtainment System offers satellite TV service for a very low price.  The way we can do this is that our box records all your favorite programs overnight, speeds them up by about 10%, and inserts more commercials into the extra time slots.  Plus, the remote has no fast forward button.

Clever, no?  Now my thermal question.

When the electronics are working their hardest, during the night when the processor is time-compressing the video, and if the room temperature is at its allowed maximum of 35°C, then aggressive fan cooling is required.

When the room temperature is more comfortable, say, 20°C, and the customer is just playing the video back, fan cooling is not so critical.  I’d like to use a thermostatically controlled fan, so that it runs at full speed only when it is needed, and more slowly at other times.  That would reduce fan noise for the customer, and make the fan live longer.

Where is the best place to put the temperature sensor?  Inlet air?  Exhaust air?  And is there a formula that tells me how many degrees I get for every RPM of fan speed?  I could use your help.  There is a huge backlog of unwatched commercials waiting for me to finish this project.

B. L. Zebub from Madison Avenue

*** March 2014 Update ***

Does this one look a lot like June 2004? It is, but this time I actually answer the question about how to slow down a fan to make less noise, and how it affects temperature.

*** March 2014 Update ***

Dear B. L.,

Where do I sign up for this service?  And are you planning an upgrade in which there is no programming at all with commercials 100% of the time?  How much would that one cost me?

That would be my solution.  Run commercials ALL the time.  Because their audio volume is louder than ordinary programs, they will cover up the fan noise and you won’t need to slow it down.

But I suppose the sponsors would like their ads to be heard in all their subtlety, so I should answer your question as you have asked it.

First question:  Where to measure temperature for the purpose of controlling fan speed?  You suggest a choice between inlet and exhaust air temperature.  Although both of these have been done many times in the past, I don’t recommend them.

You can’t use inlet temperature, because, as you’ve already told me, the power dissipation changes between day and night.  A sensor at the inlet wouldn’t detect any increase in power dissipation or internal temperature, so the fan would not speed up even when the components got hotter.  I could only respond to changes in ambient.

An exhaust temperature sensor might work better in that regard.  But I find that exhaust air temperature is a weak predictor of trouble inside the box, too.  A component temperature could shoot up 30°C, and the exhaust air might go up only 2 or 3 degrees.  Your sensor might see that as just “noise.”  You are trying to detect the patient’s fever by asking him to blow on the thermometer.

What you really want to control is component temperature, so I would chose a “representative” component and attach the sensor to its case.  What does representative mean?  It depends on what components are in your box.

Is there one component that is the “bad boy?”  Maybe you have a microprocessor that uses 80% of the power of the whole box.  It gets much hotter than all the other components, and is much closer to its operating limit than anything else.  It will always be the first component to exceed its limit if the ambient gets too high.  That is an obvious choice for the “representative” component.

What if you have two microprocessors, and they take turns getting too hot, depending on whether you are compressing or watching video?  Maybe you need two sensors.  Or three or four.  You need enough sensors, and enough understanding of what is happening thermally inside your box, to make sure that when you slow down the fan, none of the components will overheat.  I don’t think you can design that kind of sensor system with simple rules of thumb.  It takes detailed testing.

Second question:  How much does temperature increase when you decrease fan RPM?

You don’t sound like a thermal ingenue, but for the sake of the First Graders at the Kordyban Kooling Skool, I’ll start out with this statement:

The slower the fan turns, the hotter the components will get.

Beyond that simple statement, the relationship between component temperature and fan speed starts to get complicated.  Let’s assume you know that we need to talk about component temperature rise, and not its temperature.  If you cut the fan speed in half, the component temperature doesn’t double.  The temperature rise doubles.  That is, the difference between component temperature and room air temperature doubles.  Maybe.  But not exactly

In forced convection, component temperature rise is a function of the air velocity.  The convective heat transfer coefficient (h) increases with air velocity near the component surface.  But cutting the velocity in half does not necessarily cut h in half.

If your component is anything like a flat plate, h changes with velocity (V) something like this:

h2 = h1 (V2 / V1) n

n is 0.8 for turbulent flow, and 0.5 for laminar flow, and probably some other number for the weird flow pattern inside your box.

Is the flow inside your box turbulent or laminar?  Maybe both — the flow is not uniform everywhere.  Turbulence develops with longer flow lengths and higher velocities, both of which vary all over the place in a typical electronics assembly.  So you probably have regions of turbulent flow, such as right near the fan, and other regions of laminar flow, where the air velocity is slow, or even dead.

Maybe at full fan speed, the flow is turbulent, but when you cut it in half, the flow becomes laminar.  If you change from turbulent to laminar in the middle of your speed change, how can you predict how h changes?  I don’t know.

I don’t even want to get into the subject of flow separations and re-attachments, and how they have a strong effect on local temperatures, and how those effects move around when you change the fan speed.

So you can’t say that component temperature rise doubles when you cut the velocity in half.  It goes up by roughly that order of magnitude, but you won’t know exactly how much unless you measure it.

There is another complicating factor.  Fan speed and air speed are not locked together one-to-one either.  So far all I told you was how temperature rise changes with air velocity (and I didn’t even tell you that!), not fan speed.  Now I need to explain how air velocity changes with fan speed.

Flow rate through your box depends on two things:  the shape of the fan curve and the shape of the box flow resistance curve (the system resistance curve.)  I have explained this a million times before, so I won’t show you that picture again.  It is on the cover of my book, if you need to see it.  Where the fan performance curve intersects the system resistance curve is the operating point, which tells you the average air flow through the box.

When you reduce the fan speed (I sometimes call it fan RPM, or Revolutions Per Minute), the fan curve slides down and to the left, according to the Fan Laws:

G2 = G1 (RPM2 / RPM1
P2 = P1 (RPM2 / RPM1)2

G is the flow rate and P is the pressure.

Here is an example:

turbulentRPMI am always shocked by how much the pressure values of the fan curve go down when you cut the fan RPM.  You need pressure to move air through the resistance of the box.

I have plotted a couple of typical resistance curves on here.  The laminar flow curve is a straight line, and the turbulent flow curve goes up with the square of the flow rate.


In this example, if the flow is turbulent, when you cut the fan RPM in half, the flow rate goes down from about 18 to 9 liter/sec.  The flow rate is cut in half, which seems intuitive.

But if the flow is laminar, as in the second graph, and you cut the fan speed in half, the flow rate goes down from about 23 to about 7 liter/sec.  It goes down about 67%!

I am not saying that will always happen with laminar flow.  What I want you to realize is that you won’t know what happens until you measure it.  Put your thermal sensors on your representative components, cut the fan RPM, and be ready to be surprised at what you get.

Let’s sum up this complicated explanation.

When you cut fan speed, you cut air velocity, but you can’t be sure exactly how much.

When you cut air velocity, you increase component temperature rise, but you can’t be sure how much.  Different components could go up different amounts

Neither inlet nor exhaust air temperature is a good place to put your fan speed control sensor.

To do this right will take a lot of testing.  If you have a CFD model that you trust, it may help to reduce the number of tests.

I don’t think you’ll have any trouble with these complex relationships.  You probably have experience with the graphs that tell you how much your customer base goes down every time you increase the number of commercials per hour, plotted against how much your revenue increases when you stuff in the extra commercials.  It is a juggling act like that.


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.