This section provides generalized information on chassis cooling, noise generation and fan failure issues.

Note
Military Embedded Systems published a paper regarding this material in their
May, 2008 issue - Thinking Inside the Box: Boosting The Effectiveness of Air Cooling.

Table Of Contents

• Introduction
• Fan Flow Rate
• Chassis Temperature Rise From Installed Cards
• Chassis Air Flow Impedance Curve
• Fans versus Blowers
• Example Data for Model 417-MB Chassis
• Fan Operation at High Altitudes
• Fan Manufacturer Resources
• Chassis Noise
• Fan Failure Detection
• Cooling Device Manufacture

If high ambient temperatures or noise are not an issue, you can skip this section.  Otherwise, the following information will demonstrate the maximum ambient temperature any industrial chassis can safely operate in as well as help with chassis noise reduction.  You will see most manufacturers' operating temperature and power supply specifications are unrealistic and not supportable.

Additional cooling allows for higher power boards to be installed plus extends component life by limiting the maximum temperatures the components are exposed to.  In general, a 10 deg C temperature reduction will provide a 2:1 increase in MTBF.

There are two methods for selecting a fan.  The first is to simply install as many of the biggest fans available as will fit.  This works well if cost and noise are not factors.  In general, a single fan, regardless of size, will cost about $20 with installation, grills, etc.  A 4-½"  fan will cost about $5 more than a 3-½" fan.  This is not significant compared to the total system cost.  However, four fans will cost about $80 or 25% of the chassis cost but a small percentage of the integrated system cost.  System cooling performance should be determined to assure cards are not operated above their published temperature limits. However, noise is often a limiting factor for chassis selection with conflicting requirements for operating temperature versus low noise generation. This is becoming more pronounced with the high speed processors such as dual Xeon which require 200Watts just for the motherboard.

We are seeing more specifications that include a noise limit for normal operation.  A single chassis doesn't seem very loud.  But if you package six chassis in a rack and put several racks in a room, the combined noise can be deafening.  The problem with bigger fans is they generate significantly more noise than smaller fans.  "Bigger" used in this context refers to the speed and free air flow rating of the fan. Often, the bigger fan is simply stirring the air and providing no real additional cooling benefit.  Because the chassis pressure impedance curve (pressure required to move air) is nonlinear, doubling fan capacity does not double the air delivery. Fans that are physically larger typically are more quiet because you can get the same volume of air flow with a lower fan RPM. Its the RPM that generates the noise.

Please keep in mind the following discussion is directed at applications where noise is of some concern.  If noise is not a limiting factor, then installing fans with the most flow will assure the lowest internal temperatures and longest component life.  However, every chassis has a maximum allowed power loading that can easily be calculated.

For the most part, all industrial computer chassis are constructed the same way with very similar layouts.  This discussion will apply to any of these chassis.  You should ask your potential supplier for a copy of their chassis impedance plot to determine actual chassis cooling performance in the intended environment.

Fan Flow Rate

Cooling fan selection for an industrial chassis is very straightforward.   Installed boards generate heat that must be removed from the chassis.  The maximum allowable internal chassis temperature is determined by the component with the lowest operating temperature specification.  The temperature rise inside a chassis will be a constant factor depending only on the installed boards.  This means the installed boards generate a temperature rise that is simply added to the inlet temperature to determine the temperature inside the chassis.  Thus, if you measure the temperature inside a chassis at room temperature, the internal temperature will rise by the same amount as an external temperature rise.  If your internal temperature is 100 deg F at an outside temperature of 70 deg F (30 deg internal rise), the internal temperature will be 150 deg F when the ambient temperature is 120 deg F.  150 deg F is generally well above published operating limits for installed drives and boards.  This means you can easily measure the temperatures inside the chassis in your office at room temperature and then calculate what the internal temperature will be when the chassis is installed in the field.

The one caveat to this is fans are less efficient at higher temperatures because the air density is lower at higher temperatures.  Thus a given fan will move less air at an elevated temperature.  If your calculations show a marginal condition at the expected operating temperature, then you should measure the actual internal temperatures with the chassis operating at an elevated external temperature.

Power is power.  There is no difference between digital heat and resistive heat.  A system with a 300 watt power supply at full capacity is generating 300 watts of heat.  In addition, the power supply is generating additional heat because of it's inherent inefficiency.  Power supplies are usually about 70-75% efficient, so a 300 watt supply at full load generates 416 watts of heat.  As a comparison, think of your 1000 watt hair dryer.  For UL certification, a chassis will have resistive load resistors installed on each of the power supply outputs with the load equal to the rated output.  Thus there will be 300 watts of heat generated inside the chassis.  An important consideration in this discussion is UL and other safety agency ratings.  UL will require a system be tested at the full power supply rating and at the rated maximum operating temperature.  You can see from the following discussion that the internal chassis temperature will exceed the rating for the power supply and the chassis will fail the UL testing.  The only alternative is to specify a limited high ambient temperature or derate the power supply.

The amount of required air is dependent on the heat generated by the installed cards, the temperature of the inlet cooling air, and the maximum internal temperature allowed.  Any particular fan will only move so much air through a chassis limited by the chassis flow resistance.

One item that has a tremendous effect on the system resistance is the air filter. If your application is in a relatively clean environment, you probably don't need an air filter. On the other hand, if there is conductive dust floating about, you want to keep that out of the chassis. You should choose an air filter based on what you are trying to filter, the available maintenance, and allowable resistance. See Universal Air Filter for information on various filter media.

One last item to keep in mind is the location in the chassis of the heat generating components. Assuming relatively smooth flow through a chassis, the air temperature will rise as it passes hot components. Thus, if the processors are located at the front of the chassis, any component down stream will be immersed in hot air. Conversely, if the processors are at the rear of the chassis, most of their heat is exhausted directly out of the chassis. The actual temperature rise of the air as calculated below is the exit air temperature. Thus, sensitive components, such as drives, should be placed upstream of the heat generating components. On the other hand, the new high speed drives can be some of the worse offenders in generating heat.

A good alternative to heat control is exhibited by our XPT Single Board Computer with dual 3.06MHz Xeon processors. That board includes an air tunnel so the heat from these processors is exhausted directly out the rear of the chassis. See the XPT Cooling White Paper for an explanation of this technology.

Chassis Temperature Rise From Installed Cards

A simple conservative rule of thumb for cooling flow requirements, discounting such effects as heat loss through the chassis walls and laminar versus turbulent flow, is:

• CFM = 3.16 x Watts / allowed temp rise deg F

This is the air flow required, not the fan rating.  It is a conservative number because there is some radiated cooling and turbulent flow is more effective at removing heat from components. 

For example, a typical chassis with 100 watts of load, 130 deg F max internal temperature in a 100 deg F environment (a 30 deg temperature rise):

• CFM = 3.16 x 100 watts / (130 - 100) = 10 CFM

10 CFM of flow through a chassis will limit the internal temperature rise to 30 deg F. 

For an extreme case, calculate for 300 watts, same 130 deg F max internal limit (typical for commercial components) operating in a 120 deg environment (typical chassis specification):

• CFM = 3.16 x 300 watts / (130 - 120) = 94 CFM

94 CFM must be moved through the chassis to limit the internal temperature to 130 deg F.

Or you can reverse the equation given a flow rate:

• Watts = 42 CFM x (130  120) / 3.16 = 136 watts

The difficult part of selecting a fan is determining chassis flow impedance or how much air a particular fan will move through the chassis.   This impedance is the friction the air sees passing through the front panel slots, air filter, internal structures, and out the rear slots and power supply.  There is no method to empirically deduce this number by looking at a chassis design.  System impedance can only be determined by measuring actual flow through a chassis against the pressure required to generate that flow.  Click the photo at right for a chassis under test.  Once this relationship is determined, a plot can be made of the chassis impedance with flow on one axis and pressure on the other.  Various fan curves can be laid on top of this curve to determine actual airflow in CFM through a chassis.

The chassis impedance plot is used by finding the intersection of the fan curves with the chassis impedance curve.  The fan curves are given in the fan manufacturers' literature for each fan.  In the above plot for the Chassis Plans Model 417-MB, a single 45 CFM fan will actually move 27 CFM of air.  Two 45 CFM fans move 42 CFM.

Note that adding an additional fan did not double the flow through the chassis because of the rising tail on the impedance curve.  As the flow increases, the pressure in the chassis increases.  That limits the output of each fan a little more than if only one fan was operating.

Note that once the system power load and maximum internal temperature are determined, the axis labeled CFM can be relabeled as Maximum Ambient Temperature as shown in the calculations above.  The intersection of the various fan curves with the chassis impedance curve will give the highest allowable ambient temperature.

Remember one point of this discussion is noise control.  After determining the chassis factors, fan curves can be used to find a fan with the lowest airflow that meets the cooling requirements.  This fan (or other low volume fans) will presumably be the quietest fan also.

Published fan specifications are rated at zero differential pressure across the fan with a curve showing actual flow versus differential pressure.  The actual flow through the fan is then dependent on the pressure drop generated by mechanical openings (front and rear panel openings) and the air filter (area, thickness, efficiency, and dirt loading).  Each chassis will have a System Resistance Curve of air flow versus delta-P where delta-P is the measured pressure drop across the fan.  Differential pressure is dependent on the velocity or flow rate (CFM) squared so that if the flow rate is doubled, the differential pressure goes up by a factor of four.  Thus, installing a fan with a free airflow twice that of the original fan will not increase system cooling by a factor of 2, but only by 1.4.

Another way to look at this square effect is doubling the chassis power usage (100 watts increasing to 200 watts) requires a fan (or fans) with four times the free air flow rating to keep the temperature the same.

Fans Versus Blowers

A fan, commonly a "tubaxial" fan, used in the typical chassis uses aerodynamic lift from the blades to move the air through the fan. The term "tubaxial" comes from the tube shape and the air is moving along the axial direction or along the axle of the fan motor.

A relatively new addition, at least in the industrial computer market, is the blower. Blowers use a combination of aerodynamic lift and centrifugal force to move the air. Air flows into the middle where the blades grab it and throw it out radially. The case contains the air and there is an exit where it all flows.

Fans that interest us are available ranging in size of 40x10mm (40mm x 40mm x 10mm thick) blowing about 5CFM with a max pressure of 0.11 inches of water to 120x38mm blowing upwards of 200CFM and 0.6 inches of water. Some of the 40mm fans, running at 13,000 rpm can blow up to 24CFM at an inch of water but they're very noisy.

Blowers are available from little 30x10mm 1.66CFM .370 inch water to 159x40mm 61.8CFM 1.4 inch water behemoths.

To do an apples to apples comparison, let's look at a 120x38mm fan compared to a 120x32mm blower.

 

	CFM	Pressure	Noise (dB)	RPM	Power (watts)
Fan	190	0.700	59	4000	24
Blower	39.55	1.299	56.5	3100	15

Ignoring the form factor issue, you can see that a fan will blow a lot more air, 4 times as much, compared to a blower. On the other hand, a blower will generate much higher static pressure.

This would lead to the appropriate application for each device. In an application where there is plenty of vertical room, such as a 4U chassis, AND, the chassis impedance is low, then a fan (or multiple fans) will move much more air. On the other hand, for a chassis that is height limited, such as a 1U, OR, the chassis impedance is high, then a blower is the more obvious choice.

As you will recall, chassis impedance, or air flow resistance, is made higher by air filters and number of inlet/outlet openings and their size and shape.

A natural application for a blower is in a 1U chassis with a noise limitation where there is not much room for air vents and only 40mmfans will fit vertically. In the typical 1U, there is room for 4 40mm fans for a total flow of 40CFM but a max pressure of only 0.25 inches. With the chassis impedance, that flow would drop to 15-20CFM through the chassis. A 97x33mm blower would blow 25CFM at 1.00 inch so the actual airflow would drop to about 20CFM or so. The airflow would be fairly equivelent, but the blower would be less expensive and quieter. Some of the 13,000 rpm 40mm fans would provide good pressure in this application but would be extremely noisy..

A potential downside to using a blower in this application is the fans only require an inch of chassis depth while a blower would require almost 4 inches. These are the tradeoffs that engineers get paid so much to balance.

Example Data for Model 417-MB Chassis

Follows is actual data for the Model 417-MB 4U chassis.  Various configurations were tested to determine where the pressure drop was occurring.

It can be seen that replacing the 30ppi filter with a 10ppi material realizes a 15% improvement in flow.  Opening the door in front of the filter only improved the flow by about 15%.  This shows that re-engineering the slots in the door would not greatly improve the flow.  Removing the plug-in cards only accounted for about ½ CFM increase.

Fan Operation at High Altitudes

Compensating the calculations for operation in airplanes at extreme altitudes (greater than 10,000 feet) is straightforward.  Heat removal is a function of the mass of the cooling air, not the volume.  However, fans are constant volume, not constant mass devices.  Because the density of the air is lower at higher altitudes, more volume must be moved to provide the same mass to achieve the same cooling effect.  It is the mass of the air that provides cooling.  If more volume is not available, the reduction in cooling efficiency can be calculated by simply multiplying by the ratio of the reduced pressure over sea level pressure. 

For example, if the calculations showed 200 watts could be cooled at sea level, then only 100 watts could be accommodated at 18,000 feet where the pressure (or mass flow) is ½ that at sea level.  The lower ambient temperature at that altitude should be of some benefit.

Discounting air conditioners, what else can be done to improve system cooling performance?  Additional small spot fans can be provided inside the chassis to provide additional cooling for high power devices such as the CPU.  The only other alternative to installing bigger fans for more flow is reducing system flow impedance.  Anything done to make it easier to move air through the chassis will help.  This can be done in steps.  The table above shows the results of various system changes.  What is the requirement for filtering the air?  Can the filter be removed or replaced with a material with less restriction?  Can the front panel slots be increased in number or size?  Can additional exit slots be provided?  Can slots be added to the lid or chassis?  These are all part of the design requirements determined by the intended environment, FCC and UL requirements, and product appearance. 

Fan Manufacturer Resources

• Comair Rotron 619-661-6688
• Delta Electronics
• Nidec 860-482-4422
• Globe 513-228-3171
• Control Resources 508-772-4043

Chassis Noise

So why not just put the biggest fan available, or several of them, into a chassis?   The more air the better, right?

Any system specification will provide operating temperature limits and many specifications provide noise limits for installed equipment.  These two requirements are diametrically opposed.  For a chassis to operate at a high ambient temperature, large volumes of air must be moved through the chassis.  This requires large fans operating at high speeds that generate noise.  The quiet fans won't move enough air to adequately cool the chassis at high temperatures.

A realistic specification for operating temperature must be determined.  Various fan curves can be analyzed to find the fan that will move the right amount of air to keep the chassis within the allowable temperature limits.  The fan spec can be checked for noise to find the fan with the lowest noise for a given air flow.  This technique will produce a chassis with the lowest generated noise.  It is entirely unrealistic to expect an industrial chassis to operate at 120 deg F but only generate 40dB of noise at that temperature.  Realistically, the noise limit is for a shirt sleeves environment with the higher operating temperature specified for the infrequent case where the room air conditioning fails and a noisy chassis is not a concern.

Noise measurement is based on a log scale such that a change of 10dB is a doubling of the perceived noise. One problem is that published noise specifications for various manufacturer's fans can be misleading, or, at worse, outright marketing garbage. There are standards for measuring noise such as ISO07779 but not all manufacturers follow the standard.

Chassis noise is generated by the fan blades plus turbulence in the air flow into the fan.  Higher speed fans will generate considerably more noise than lower speed fans because of the higher blade speed plus the higher flow rate.  A fan with twice the flow will generate much more than twice the perceived noise.

There is a formula for getting a percieved noise comparison between two devices.

x = bel rating of fan #1 or other component (louder device)
y= bel rating of fan #2

The perceived relative noise level is then defined as:

(3^x)/(3^y) = ratio

For example, if Fan(x) is 40dB (4.0bel) and Fan(y) is 25dB (2.5bel)

(3^4.0)/(3^2.5) = 5.2 or Fan(x) would seem to be 5 times as loud as Fan(y).

Note that a bel equals 10 decibels.

If the temperature specification does require operating at high temperatures, then temperature controlled fans can be used.  These fans have the same size and mounting as normal fans, but use a temperature sensing device to reduce the fan speed at low temperatures and increase it at higher temperatures.  Thus, at a shirt sleeve temperature, the fan will operate at a low speed, low noise RPM.  At high temperatures, the fan speeds up, providing additional cooling, but with increased noise.  Some fans have the sensor built into the fan so that as external temperatures rise, fan speed increases.  Other fans provide a remotely mounted sensor that can be mounted either at the chassis air outlets or at the hottest location in the chassis.  Thus, instead of inferring increased internal temperature, the actual component temperatures can be monitored.  A pair of temperature compensated fans provide redundancy with the surviving fan increasing output to make up for the lost cooling and higher temperatures caused by a failed fan.

The fan manufacturers listed in Chassis Plans Manufacturing Documentation Packages offer drop in replacement temperature controlled fans and can help you with a substitution.

Fan Failure Detection

Another possible feature in an industrial chassis is fan failure detection.  Two companies (Control Resources - 978-486-4160 and PC Power and Cooling - 800-722-6555) are marketing inexpensive fan failure detection modules that may be installed in a chassis to provide this feature.  A 555 integrated circuit can be configured in a missing pulse mode to provide a very inexpensive failure circuit.  A single printed circuit can provide chassis temperature alarm, fan failure detection and speaker amplification.

However, fan failure is a rare occurrence when quality ball bearing fans are used.  The low end of fan MTTF (Mean Time To Failure) is 5 years of operation.  Failure of the fan failure sensing circuit or susceptibility to electrical noise are much more likely. The additional cost and complexity of fan failure circuits is generally not warranted by the marginal added benefit.  The chassis also provides redundancy with multiple fans, the power supply fan plus good radiative cooling through the metal chassis construction.  For true redundancy, two fans should be installed.  Preventative maintenance may be used to replace the fans every couple of years to preclude fan failure.  Instead of fan failure detection, the real concern is excessive chassis temperature.  Install a temperature alarm in place of the fan failure circuitry.  In any case, the air filters need routine servicing.  At that time, the fans can be inspected for operation.  The odds of simultaneous fan failure are small enough to be discounted.

Cooling Device Manufacturers

There are an untold number of fan manufacturers. The two listed below provide a good cross section of available devices. Chassis Plans primarily uses Mechatronics.

Updated 11/3/04