57
\$\begingroup\$

This question covered it for enclosures. However, from the point of view of the fan attached to a heatsink does it matter whether air is blown through the fins, or sucked through the fins. In other words, is the pattern of airflow different enough to matter?

\$\endgroup\$
  • 4
    \$\begingroup\$ Do you want debris blown onto the expensive chip? \$\endgroup\$ – Andy aka May 16 '17 at 8:50
  • 12
    \$\begingroup\$ @Andyaka usually the heatsink covers the chip enough for it not to matter. Kinda hard to get enough cooling otherwise. \$\endgroup\$ – ratchet freak May 16 '17 at 9:09
  • 3
    \$\begingroup\$ Experiment. Try it both ways and see if one works better than the other. Generally, either way, you have an inlet and outlet. One is very likely going to work better than the other. \$\endgroup\$ – Misunderstood May 16 '17 at 9:53
  • 3
    \$\begingroup\$ Age old question. \$\endgroup\$ – Rob May 16 '17 at 10:33
  • 6
    \$\begingroup\$ Why choose? High-end heatsink designs use both. In this design, the two fans are mounted coaxially on both sides of the fins. \$\endgroup\$ – MSalters May 16 '17 at 10:42
107
\$\begingroup\$

This is such a wide subject it really isn't one you can answer with a simple one is better than the other answer.

enter image description here

Standing alone, the blow side of a fan does produce a more concentrated, faster moving, and more turbulent "river" of air compared to the intake side where air is drawn almost equally from all directions. You can test this easily enough with pretty much any fan. Put you hand in front of the blow side and you will feel the airflow and cooling effect. Put your hand behind and the effect is much harder to detect.

The turbulence also greatly improves the efficiency of the heat transfer. Turbulence is in fact your friend.

So from those points of view alone, the blow side does appear the better cooling side.

However, it is not just about the fan.

The geometry of the heat-sink chosen also greatly affects the performance of the fan. A rotary fan slapped on top of your typical linear finned heat-sink will actually be quite inefficient. In fact the region directly under the centre of the fan will get virtually no air movement at all. This of course is unfortunate, since that is normally where the thing you are trying to cool is located.

Further, unless the fins are quite deep the airflow is badly distributed in general. Too shallow, and the resultant back-pressure can actually "stall" the fan. In those circumstances, installing the fan in the "suck" direction can actually improve the situation since the air will enter the sides of the heat-sink more linearly to fill the void in air pressure created by the fan.

enter image description here

Arguably, the heat-sink shown above might be more efficient with longer fins and the fan mounted at one end.

Better designs use radial heat-sinks like the one below. As you can see, the style here is radially symmetric to the airflow on the entire circumference of the fan and consequently delivers a more even heat transfer around the central core.

However, even with this style, the core itself is still badly ventilated. As such it is usually manufactured as a solid high thermal conductance core which acts as a heat-pipe. Even then, looking at the image below, the area around the core in the square section that touches the chip actually is an air void that is quite inefficient. A better design would have that area filled with metal in a rounded conical structure. However, that would of course be impossible to extrude.

enter image description here enter image description here

If fact materials and surface preparations also make a huge difference in heat-sink design. Highly thermally conductive materials are obviously best, but the surface should also be smooth enough not to allow pockets of air to form or to grab at dust particles, but also not so smooth that air passes too easily over it.

One could of course spend years getting that little formula perfect, but in general you don't want a high polish chrome heat-sink. Sandblasted aluminum, or gold coated sandblasted copper, if you can afford it, would work a lot better.

Another serious issue is contamination.

Dust and dirt is going to get into your fan and your heat-sink. Over time this builds up and severely degrades the performance of the unit. It is therefore prudent to design your fan and heat-sink arrangement to be as self flushing as you can.

This is where a blower fan usually wins out. With controlled airflow and if the air coming in can be kept clean, it tends to blow dust out of the heat-sink. Which brings me to the next point.

enter image description here

Air Sourcing and Removal

You can spend thousands of dollars developing the perfect arrangement of fan and heat sink and it will all be for naught if you do not deal with the rest of the air around your cooling system, especially in a tight enclosure.

The heat not only has to be removed from your device to air, but that hot air then needs to be removed from the vicinity. Failing do to so will just recirculate the hot air and thermal failure will still occur on the device you are trying to protect.

As such your cabinet needs to be vented and you should also include cabinet fans to draw in cool air from outside the enclosure. These fans should always include removable mesh and or foam filters to control the amount of ambient dust sucked into the unit. Open grill type exhaust panels are acceptable, however, for best operation a positive pressure should be maintained within the cabinet so airflow is maintained in the out direction to again limit contamination entry.

Special Cases

Wherever the unit is to be installed in an extreme environment special measures need to be taken. High dust environments like floor mills etc., or high ambient temperature environments will require either ducted air direct to the chassis, or a sealed unit and a two stage, possibly liquid, cooling system.

Critical Cases

If your system is controlling something critical then it is prudent to include thermal sensing and possibly active fan control as part of your heat-sink system. Such systems should include the feature of going into a safe state and warning the user to clean the filters or otherwise reduce the ambient heat around the system when necessary to prevent critical failures.

One More Point

You can spend a half years development money getting the best heat-sink design in the world with expensive fans and a perfect air distribution system all locked down then burn out devices for the lack of 2 cents worth of thermal compound.

Getting the heat from the device you are trying to protect into the heat-sink can often be the weakest point in the system. Components not properly mounted to the heat-sink with an appropriate thermal bonding material kills more units than the rest of the issues combined.

Your manufacturing process and procedures should be developed to give those aspects first priority.

For example, if say you are using three or four TO220 style transistors mounted to a single heat-sink, it is prudent to mechanically mount them to that heat-sink, and if appropriate, the heat-sink to the board, BEFORE going through the soldering process. This ensures the thermal connection takes priority.

Either thermally conductive pastes, creams, gels and or electrically isolated thermal pads should always be included between device and heat-sink to fill any air gaps caused by non-flatness, or bumps on either the device or the heat-sink surface.

And keep it clean. A contaminate the size or a grain of salt, or even a stray hair, can cause thermal failure.

\$\endgroup\$
  • 2
    \$\begingroup\$ Also, you forgot to mention that different metals conduct the heat in different ways. And you didn't mention heatpipes. But, you gave a solid answer. Regarding contamination on heatsinks, they are very good at collecting dust, dirt, animal furr and human hair. \$\endgroup\$ – Ismael Miguel May 16 '17 at 13:47
  • 3
    \$\begingroup\$ @DmitryGrigoryev.... what good would a plastic heat-sink be.. It's about transferring heat from the source through the heatsink to the air. The heat-sink needs to be highly thermally conductive. \$\endgroup\$ – Trevor_G May 16 '17 at 14:01
  • 2
    \$\begingroup\$ @Trevor Either I fail to read the question properly, or it indeed is just about the fan. \$\endgroup\$ – Dmitry Grigoryev May 16 '17 at 14:21
  • 3
    \$\begingroup\$ The dust reminds me of a customer who smoked indoors by their PC. The dust acted as a filter for their cigarette smoke over several years. I am scarred for life... Should have denied helping them, but I needed the cash at the time. \$\endgroup\$ – Chris Schneider May 16 '17 at 16:01
  • 3
    \$\begingroup\$ I have almost no interest in CPU airflow at all, but I'm upvoting this with a vengeance because it's probably the best answer I've ever seen :) \$\endgroup\$ – Whelkaholism May 17 '17 at 10:31
5
\$\begingroup\$

The pressure pattern will differ.

When blowing into, the pressure on heatsink surface (parallel to blades) will be higher, which means higher thermal conductivity on the surface.

When sucked through fins, the pressure on the fins' surface orthogonal to the air flux will be higher.

So I think the correct airflow direction depends on the heatsink's dimension ratios & weighting them with thermal spreading pattern. Empirically one could say that when its ampleness is much larger then its depth, definitely blow through is better.


Addition after andresgongora's comment...

Think the air pressure as voltage and air velocity as current, the obstacles orthogonal to flow as resistance, the resulting convection of heat as power. Or think the pressure the mass interacts with heat per unit time, which is being refreshed by airflow rate.

So the pressure pattern won't give an exact picture of what is happening there, full convection pattern will be complicated but it gives a good idea about the better direction of airflow.

enter image description here

\$\endgroup\$
  • \$\begingroup\$ If it is wider than deeper, shouldn't it be sucking? And when deeper than wider then blowing? I think I don't get it, could you please elaborate? After all, it's not pressure we are after, but airflow. \$\endgroup\$ – andresgongora May 17 '17 at 9:03
  • \$\begingroup\$ @andresgongora see the addition. \$\endgroup\$ – Ayhan May 17 '17 at 14:40
4
\$\begingroup\$

Heat is transferred by conduction, radiation and convection. To cool an IC, all three modes are used - conduction from the die to the heatsink, radiation from the heatsink to the surrounding environment, convection by moving the air. The Boyle and Charles laws give us \$PV = kT\$, where \$P\$ = pressure, \$V\$ = volume, \$k\$ is a constant and \$T\$ is the absolute temperature. Now if we want to track temperature change over time, we can differentiate this equation. This gives:

$$V \frac{dP}{dt} + P \frac{dV}{dt} = k \frac{dT}{dt} + T \frac{dk}{dt}$$

If you want to move air through a fixed volume, \$V\$, eg, a computer case or its power supply, then \$\frac{dV}{dt} = 0\$; and of course \$\frac{dk}{dt} = 0\$. So, the equation simplifies to:

$$V \frac{dP}{dt} = k \frac{dT}{dt}$$

In other words, if you increase the pressure over time the temperature will rise, and vice versa. To help you understand this principle, consider these two examples:

  1. when you pump up the tyres on your push-bike using a hand pump, the end of the pump nearest the outlet gets quite warm. This heating effect is altered by the P.dV/dt term, which is not zero.

  2. if you have a cubical room in your house with windows and doors on all four vertical walls, and you have a hot wind coming from the North, you can cool the room by opening the window / door on the North wall by say 50 to 100 mm, and opening the windows / doors on the other walls by say 200 to 500 mm. This will lower the pressure inside the room and lower the temperature.

Now to the matter of turbulence.

The highest amount of heat transfer from the heatsink (or other hot components) occurs under laminar fluid flow. When the air flow increases, you may eventually reach a point where the airflow becomes turbulent. The effects of turbulence are:

  • the effective area of the fan decreases - ask any prop-plane pilot about the effect on propulsion when increasing prop speed beyond the RPM red line
  • noise increases = energy lost
  • vortices form, depositing air-borne debris in the lower velocity areas
  • the efficiency of the fan falls and its temperature may increase
  • cavitation occurs giving rise to areas of zero airflow, and hence
    rampant temperature rise.

So, turbulence is definitely NOT your friend.

You may try to reduce the fan speed to reduce the turbulence; if the fan was well designed, the angles of the fan blades will be continuous curves to take account of the increase in air velocity as the air passes over the blades. So, slowing the fan means that the curvature of the blades is no longer correct for laminar flow. This effect is overcome on aircraft and large ship propellors by varying the 'pitch' of the blades, including reversing pitch. This is usually not possible with the size of cooling fans used in electrical equipment.

Shrouding of fans

If there is an unimpeded, continuous air path from the down-side (high-pressure, or outlet) to the up-side (low pressure, or intake), the higher pressure air merely travels over the shortest path back to the inlet and the downstream flow is reduced. You see this all the time - aircraft propellors, marine propellors (see the latest propulsion design of the Spanish warships supplied to Australia), cheap home cooling fans. To overcome this loss and thereby increase the effectiveness of the fan, the better designs have close-fitting shrouds around the tips of the fan blades. Frank Whittle's PhD included using shrouded fans in his jet engine - very much more efficient than open propellors and good for rapid temperature rise to increase the exhaust gas velocity.

Using a hand to detect cooling

The cooling you feel when downstream of a fan is mostly the effect of vaporisation of the fluid water resident on your skin - the loss of 540 cal / gram through vaporisation will certainly 'feel' cool. But the effect on electronic / electrical components that do not have water on their skins is zilch. So, using your hand to detect temperature fall is the wrong model.

IN SUMMARY:

Sucking is better than blowing for lowering temperature. Laminar flow is the most efficient means of convecting and conducting heat away. Shrouding of fan blades increases effectiveness and efficiency of the fan.

\$\endgroup\$
  • 2
    \$\begingroup\$ I thought that the problem with laminar flow over heat sinks was that close to the heat sink, you have a boundary layer of still air which acts as an insulator. High performance single fan CPU heat sinks almost always push air in rather than pull air out. I thought that the reason they did this was to break up that boundary layer air and get the airflow closer to the heat sink itself. It's the same argument with smoothness. You want your heat sink fins to be neither too smooth nor too rough. \$\endgroup\$ – Mark Booth May 18 '17 at 8:23
  • \$\begingroup\$ @MarkBooth A lot of CPU coolers I've seen recently arelined up with a vent in the case so they draw outside air into the case and directly over the heatsink. This is simpler (cheaper) to arrange with the fan between the vent and the heatsink. \$\endgroup\$ – Chris H May 19 '17 at 9:41
  • \$\begingroup\$ I presume that you're talking about integrated systems @ChrisH, I haven't seen that with after market coolers, where you don't have any guarantee about where your CPU is in relation to your case vents. \$\endgroup\$ – Mark Booth May 19 '17 at 17:11
  • \$\begingroup\$ @MarkBooth mainly systems sold pre built but I've seen it on cases. And custom built systems. The CPU position is quite standard these days - just pick your standard. \$\endgroup\$ – Chris H May 19 '17 at 18:37
  • 1
    \$\begingroup\$ I have to agree with @AliChen As I recall turbulence increases forced convection heat transfer. This is why a rough surface is better than a smooth. \$\endgroup\$ – Misunderstood Jun 6 '18 at 16:18
3
\$\begingroup\$

I think it is design dependent. The main factors are:

  • to source colder air and to outflow hotter air into intended direction. If you suck out of heat sink, heat sink's air inflow may be located near other heating elements, and thus air sourced may not be of needed low temperature, or, inflow temperature may change with operation, negatively changing efficiency of the cooling system;
  • dust getting into small holes of the heat sink. If you blow in, as many commentators said, you have single point of air entrance and it may be covered with filter, or air simply may be sourced from cleaner location by design. If you suck out, air source is most probably located very close to PCB surface and other components, sucking accumulated dist from them.
  • There's another way to design cooling system. If you open contemporary notebook or high end PC, you may find it has water or other liquid cooling, and fan may not need to be located near the chip; it can be put in any location designer thinks conveniently serviceable and most clean.

Thus I vote for inflow, but again, all depends on the device design.

\$\endgroup\$
  • \$\begingroup\$ "If you open contemporary notebook", it uses heat pipes. Water is a great danger in portable systems since they can be damaged. \$\endgroup\$ – ivan_pozdeev May 17 '17 at 19:29
  • \$\begingroup\$ @ivan_pozdeev heat pipe is having some liquid inside anyway en.wikipedia.org/wiki/Heat_pipe. Maybe you are right, I never disassembled heat pipe to see what exactly is inside. \$\endgroup\$ – Anonymous May 17 '17 at 19:48
  • 2
    \$\begingroup\$ Heat pipes are not liquid cooling, the principle is different (the only thing in common is the use of a cooling agent). And they use a substance with a lower boil temperature rather than water. \$\endgroup\$ – ivan_pozdeev May 17 '17 at 20:16
  • \$\begingroup\$ And the fill in a heat pipe tends not to be a good conductor or a good solvent (unlike water). \$\endgroup\$ – Chris H May 19 '17 at 9:39
3
\$\begingroup\$

I work for an Optical Networking (Telecom) technology company, and always dealing with cooling and EMC. Excellent comments for a basic design decision for card/shelf based equipment question - to put the fans at the air filter intake or exhaust side.

I have been told by some electronic module vendors we used, that pulling the air LOSES 10-15% cooling efficiency. Two other observations I have are
1) (big) Fans at the INTAKE unfortunately PREHEAT the air by the friction and fan-motor heat dissipation
2) in trying to add ducting/deflectors in our circuit card to focus airflow FAILS miserably if you are PULLING air thru the PCBA.

It simply blocks the air movement, just like too fine features - air just to goes around the heatsink! I believe the basic difference is PULLING air causes movement by pressure difference only (less turbulence) bit PUSHING air uses active turbulence and pressure difference.

\$\endgroup\$
2
\$\begingroup\$

When the question is shorted down to a [common] heat sink and [common axial blade-type] fan, it deserves a shorter answer. And the answer is, as usual and unfortunately, "it depends".

(1) When a fan is attached on the top of a heat sink in "sucking" direction, the air enters fins (or pins) in laminar matter (at least relative to a larger vortex scales than fin/pin distance). As such, the boundary layer around heat transfer surfaces is thick, and heat transfer is rather poor. More, in one-sided sink construction with a typical fan, there will be a "dead zone" at the center with poor air flow, exactly in the place where the heat is generated beneath the sink.

(2) When a fan blows into the heat sink blades, the output air flow is turbulent, and thermal boundary layer around metal surfaces is thin, so the air flow penetrates deeper into fin structure and close to metal surfaces, providing good heat transfer. And the highest [turbulent] air velocity is around the center of sink, where the thermal "stress" is highest.

So, it looks like the case (2) has clear advantage over case (1). Unfortunately, there is one more factor, which is the fan performance under different surrounding conditions. Unlike blowers who are creating higher pressure relative the ambient space (and used in heat-pipe designs inside laptops), axial fans provide better airflow performance when sucking air out of a tighter space into ambient, so the case (1) has some preference here.

On the other hand, when an axial fan faces high aerodynamic impedance like when blowing in, it can "short-circuit" itself and provide only little to no airflow. So the use of axial fan has some advantage in weak thermal case (1), while performance of the same fan is reduced by running it into a pressurized (but more thermally-efficient) area.

So the case (1) has poor heat transfer but better fan performance, and case (2) has better heat transfer but poorer fan performance. The net result is "it depends", which includes several factors as fin thickness and spacing. And it depends on fan construction. There are three types of axial fans, tube-axial, vane-axial, and propellers, which can have blades optimized for performance in one or other direction. Tube-axial fans also have good pressurizing performance, and are used in blade servers. So the results may vary.

Obviously, the best result can be achieved with a dual-fan design, like this one, where one fan blows in, and another suck the air out.

enter image description here

\$\endgroup\$
1
\$\begingroup\$

If the fan and the heatsink are enclosed inside an air duct, you will get the same airflow on both fan sides, so the heatsink position shouldn't matter much. For a "fan on top of a heatsink" setup, the blow side definitely provides better cooling.

\$\endgroup\$
  • 2
    \$\begingroup\$ Although the amount of flow will be the same, the flow itself will not, with a more laminar flow on the intake side, and a more turbulent flow on the out side. \$\endgroup\$ – Mark Booth May 18 '17 at 8:09
1
\$\begingroup\$

Suck or blow is not the simple answer - it boils down to (no pun intended) the temperature of the air flowing through the heatsink, the speed of flow and the contamination that can build up. So the simple answer is the coolest air, the best airflow and the least contaminate - only really answerable by investigation and experiment.

\$\endgroup\$
1
\$\begingroup\$

In most cases fan in suck mode is much better than blow mode.

If fan is placed in blow mode, wind force would be blocked and spread by heatsink therefore heat would dissipate around heatsink and as a result the same source of airflow would be sucked back in by fan and heat would be recycled.

In suck mode, heat would be blown away in a more concentrated line thus much less heat would be recycled.

An exception would be that the fan is strong enough to blow the heat far enough away from the heatsink so that airflow would not be recycled. Then blow might actually be better cause its more concentrated thus airflows faster (same amount of airflow but faster) and because of that the wind itself would become cooler =)

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.