Conformity Magazine

As electronic equipment grows more complex, it runs hotter. Cooling—often by forced air—is necessary for today’s products to operate reliably, and sometimes, even to operate at all. In this article, we survey some of the issues associated with forced air cooling and, in addition, the way cooling issues interact with the NEBS (Network Equipment-Building System) telecommunications equipment environmental requirements of GR-63-CORE.

Looking back, who would have thought that miniaturization would make products dissipate more power rather than less? Heat has always been a problem with electronic equipment since the antediluvian days of vacuum tubes, and it isn’t getting any better. As semiconductor size and chip complexity scale exponentially according to Moore’s Law (roughly, that the number of transistors per unit area doubles every 18 months), circuits run faster… and hotter… and there are more of them, due to increased product functionality.

Put another way, shrinking an individual transistor lowers the power it requires to run, all other things being constant, but those other things are hardly ever kept constant. A smaller transistor can run at the same speed as a larger one while operated at lower voltage and current. But, it is also a faster component, and can be operated at a higher speed, which brings the current drawn up from the new, lower value, giving back much of the power savings. Miniaturization makes both increased product speed and increased functionality possible through higher levels of integration, and both occur. Thus, the combination of faster operation and higher component density overcomes the potential heat savings of miniaturization. The net trend is toward greater, rather than lower, power use per unit of equipment volume.

Heat is a perennial problem, not only at the equipment level, where most designers (and this article) are focused, but also at the installation level, where it must be removed. In telco Central Offices, there are guidelines for heat per unit area of the installed equipment footprint.

Noise, too, is a concern. There are also limitations on the noise that equipment can generate to satisfy both NEBS and OSHA (Occupational Safety and Health Administration) requirements, as well as European Union requirements such as those of the Machinery Directive, which are designed to keep the probability of noise induced hearing loss in workers at an acceptable level. Since higher air velocity is generally associated with both increased cooling and increased noise, the requirements for heat removal and noise control are opposed to each other.

Thermal issues also play an important factor in product reliability, and reliability is important in the Central Office, where “five nines” (99.999%) operational up-time is expected. For semiconductors, junction temperature is the single most important factor in determining the failure rate. Higher junction temperatures yield higher failure rates and shorter MTBFs (mean time between failures). Heat is detrimental to other electrical and electronic components as well. Hotter operation is inversely related to the MBTF lifetime of capacitors, resistors, inductors, and transformers by a variety of aging and failure mechanisms.

Air Flow and Cooling – Issues and Rules of Thumb
You’ve got a hot product, and you need to remove heat from it. What to do? Your tools are the three basic mechanisms of heat removal: conduction, radiation, and convection/advection.

Conduction occurs when heat flows through media across a temperature gradient. It is due to molecules jostling each other at an accelerated rate as they heat up. Conduction is the mechanism that gets heat from the silicon to its enclosing package and any attached heat sink.

Radiational cooling occurs from any heated surface as it radiates infrared electromagnetic energy into the environment. Radiation is the mechanism that conveys energy from the sun to the earth, or from a fireplace to someone sitting across the room. It can be increased by increasing surface area, or by improving the efficiency of surface emission (“paint it black”).

The last mechanism, convection or advection, is the transport of heat energy through a moving fluid (gas or liquid) medium. There is overlap in the usage of the two terms. Advection is a general term for transport of heat by a moving fluid; convection is more often used to imply that the fluid movement is driven by heat-induced differences in density.

Without forced transport of cooling air, a product will eventually reach some thermal equilibrium (assuming it can operate at the elevated equilibrium temperature) through a combination of these mechanisms. With forced cooling, a much greater amount of heat can be removed, resulting in lower temperatures inside the equipment and at its surface.

A detailed thermal analysis model can be quite involved. It must take into account a number of factors beyond the scope of this article including:

Detailed location of all heat sources
Modeling of heat source coupling to cooling medium (e.g., package and heat sink thermal conductivity, orientation)
Geometry of the enclosure, fans, and significant internal components, which can both channel and block cooling air
Fluid dynamics – laminar versus turbulent flow, velocity distribution throughout product and near heat generating components, back pressure, variations in atmospheric heat capacity with altitude and humidity, etc.

Software tools that can generate a detailed thermal model are available, but they are rather expensive both in terms of product cost, training, and modeling time. Fortunately, there are some crude rules of thumb that can get the designer in the ballpark. In fact, some designers of rack-based equipment deliberately over-design their cooling systems to allow for lack of modeling detail, provide reserve cooling in the event of partial system failure (high ambient, broken fan), and particularly to allow for the additional heat removal capacity that may be required by future product upgrades and expansion.

The amount of heat that can be removed by forced air cooling bears an obvious relation to the volume of air that flows. At any point, the volume per unit area of air transport is given by the air velocity. One crude rule of thumb models the heat problem by presuming even heat distribution and focusing simply on an estimate of the overall volume of air that needs to be moved to remove a given quantity of heat for a given rise above ambient temperature.

According to this rule of thumb, a first cut at fan sizing is given by:

Required Volume (CFM)

This may need to be adjusted upwards to account for fan inefficiencies as pressure increases, and variations in atmospheric pressure.

Noise and OSHA Workplace Requirements
To minimize hearing damage, OSHA has workplace noise requirements. Workplace noise has to be below 85 dBA to avoid the requirement that workers wear protective hearing gear. The units “dBA” merit a brief explanation. They are the ratio, in decibels, of a sound level to the normal threshold of hearing, measured with an “A-weighted” filter.

The A-weighting curve used during the measurement roughly follows the sensitivity of the human auditory system. From approximately 800 to 8000 Hz, the “A” weighting is nearly flat and minimally attenuative. This is the range where human hearing is most sensitive. Outside this range, the A-weighting attenuates. The A-weighting curve is down nearly 20 dB at 100 Hz, and approximately 6 dB at 20 kHz. Table 1 shows the dBA levels of some familiar sounds.

Sound Source	Level in dBA
Gunshot, threshold of pain	140 and higher
Jet take off, heavily amplified music	135
Chain saw, jack hammer, snowmobile	120
Tractor, farm equipment, power saw	100
OSHA occupational exposure limit	90
Average listening radio level, vacuum cleaner	75
Normal conversation	60
Rustling leaves, soft music	45
Whisper	30
Weakest audible sound, acute threshold of hearing	0

Table 1: Intensity Levels of some common sounds
Note: The OSHA occupational exposure limit of 90 dBA is based on the level that, with 8-hour daily lifetime exposure, yields a 25% estimated probability of noise induced hearing loss.

How is Air Flow Measured?
The key measures of air cooling capacity at any location inside a product are the local air velocity and the temperature difference between the cooling air and the heated components. While there are a number of computer programs that can aid in calculating a volume profile of the air velocity throughout a product, sometimes there’s nothing like direct measurement for confirmation. Two common methods of air velocity measurement are the “hot wire” anemometer and more traditional mechanical anemometers, such as a turbine.

The hot wire probe consists of a thin (5 to 10 microns) heated wire, commonly tungsten or platinum. The wire is heated, and the amount of heat conducted away from it is a measure of the velocity of the cooling medium for a fluid of given characteristics (water, air of a given density, etc.).

The heat that is transferred to the flowing medium can be calculated in one of two ways, both of which measure the resistance of the wire, which varies with temperature because of the material’s thermal coefficient of resistance. One technique is to keep the current through the wire constant, and measure the change in resistance, and hence temperature, that occurs. An alternative is to use a feedback loop to keep the wire’s resistance constant, which also keeps the wire at a constant temperature. The current necessary to do this is monitored, which allows calculation of the heat passing into the cooling medium (air). This in turn allows calculation of the cooling medium’s velocity.

Because of the small dimensions of the wire, hot wire probes can be made very compact, which means that they can yield high spatial resolution of measurement—which allows precise mapping of localized air velocity. Also, because of their small mass, they have low thermal inertia (heat capacity) and can be designed to be capable of surprisingly quick (sub millisecond) response times.

The hot wire anemometer is a solid-state instrument. A mechanical anemometer can also be employed. We’re all familiar with the cup anemometers used in weather stations. Another mechanical anemometer type is the turbine—basically a fan operated in reverse, a windmill passively spun by moving air. Regardless of the mechanical details, a mechanical anemometer transforms linear airflow to an axial mechanical motion that is then measured, typically with an optical or magnetic transducer.

Air Flow and GR-63-CORE Requirements
As noted previously, there are a number of ways in which airflow relates to the NEBS environmental requirements of GR-63-CORE. Obviously, the flow of air through a piece of equipment is related to keeping the product operating reliably, but it also affects and is affected by other factors:

Chassis Surface Temperature
Acoustic Noise
Contaminants (immunity to corrosion)
Effects of Ambient Temperature, Altitude, and Humidity
Relation to Fire Test Survivability
Air flow measurements, board upgrades, and fire test requirements

Surface Temperature: Clearly, there is a relation between the amount of air that flows through the product and the balance between convective and radiational cooling. Increased airflow will increase the amount of heat that is carried away by convection, and thereby lower the surface temperature of the product. The NEBS requirements on surface temperature are designed to protect personnel from contact burns, and limit the surface temperature rise to 12 deg. C above ambient.

Acoustic Noise: Generally speaking, the higher the volume of airflow, the noisier the product will be, both because of increased turbulence and higher van rotation speed. The general requirement is that a product’s noise, under worst case (maximal fan power) conditions be under 60 dBA. Verizon and AT&T have recently decided that they will accept the slightly higher level of 65 dBA. And, there is no noise limitation once the ambient rises to 40 deg. C, which would only happen under a failure of the office cooling, and would require maximum effort to keep down the equipment’s temperature rise above ambient.

The single product noise level allowed by GR-63-CORE—of the order of a normal conversation—might seem to be a rather modest level (see Table 1), but in the central office environment where many pieces of equipment operate simultaneously, the noise adds up. The telephone companies are caught between the ever-increasing levels of heat output per unit volume of equipment that must be removed and the noise associated with doing it.

Contaminants: Contaminants come in the form of chemicals and dust, and can be considered to have a volume concentration. In GR-63-CORE, immunity to these is specified in Section 4.5, and is tested in two ways. Section 5.5.3 tests equipment tolerance to “hygroscopic dust”, which measures the likelihood that, over time, accumulating particulates will combine with moisture to form disrupting conductive paths. (Actually, an equivalent test is performed by using a hygroscopic coating to simulate the cumulative effect of dust by providing a mildly conductive coating on the circuit board under test). The “mixed flow gas” test of section 5.5.2 provides an accelerated lifetime airborne corrosion test.

Increasing airflow increases the amount of airborne particulate and gaseous contaminants that flow through the product. Hence, there is some interaction between long-term contaminant resistance, air filtering requirements, and airflow. With respect to the way the two tests are performed, increased airflow adds to the severity of the mixed flow gas test by continually moving fresh contaminants through the product cabinet. It has little effect on the NEBS hygroscopic test, because that test is performed by treating the boards outside of the cabinet and then reinstalling them.

Altitude and Humidity: Air-cooling works because air carries heat. The specific heat of air varies with its density, humidity, and temperature. The denser and more humid the air is, the higher its heat capacity. For less dense and less humid air, more air flow in terms of volume and flow velocity will be required to maintain a desired equipment operating temperature.

NEBS GR-63-CORE testing incorporates an extensive OPTH (operating pressure temperature humidity) profile. The worst-case situation in the sense of maximum demands on the cooling system air flow occurs at maximum temperature and altitude. A rough estimate is that the cooling efficiency of a given volume of air will drop by 25 to 30% at the highest operating altitude of 13,000 feet.

Relation to Fire Test Survivability: All of the major telephone companies require fire-spread tests as part of their equipment qualification test suite. Although there are some company-specific variations, the test methods are either based on NEBS GR-63-CORE section 5.2 or ANSI T1-319. In these tests, a controlled ignition source—a perforated tube with a known volume-time profile of flaming methane gas—is introduced into the enclosure. The resulting fire, if any, is monitored for duration, intensity (via infra-red sensors), exiting flame, and volume of smoke produced.

Increasing or decreasing airflow can dramatically affect fire-testing results. Depending on the details of the equipment – localized airflow and fuel load relative to the fire injection/start point—very different responses to changes in airflow can be seen.

Some fires are oxygen-starved; increasing airflow accelerates them. For other products, increasing airflow tends to cool the fire down and slow its propagation, or even keep the fire from starting. (For ANSI T1-319 tests, accepted by SBC, the inability to start a fire under specified test conditions constitutes a “pass” of the test).

Fires can be divided into those that have adequate oxygen and those that are oxygen-deprived (see Reference 2). Those that are oxygen-deprived sometimes evolve a large amount of flammable, but uncombusted gas that periodically flares up, like the licking tongues of flame in an air-restricted stove or glassed-in fireplace. Variations in airflow can affect this tendency, and also affect how little or how much flame exits the cabinet, which is one of the test criteria.

Another factor in the fire test involves the placement of the fans and their control electronics. Fans are run during the fire-spread test. Fans that are placed at the bottom of the equipment (typically blowing air up through it) are more likely to survive all or most of the fire test intact. Often those that are placed at the top or back of equipment will be damaged during the fire test and stop operating—which decreases the amount of air is fed to the fire. This observation is not meant to imply that a particular fan placement is optimal for fire containment—it merely shows that there are many ways in which the details of air flow delivery can interact with equipment fires.

Maintaining and Augmenting NEBS Certification—Air Flow and Fire Spread
Recent pronouncements from some of the telephone companies have given an increased role to air flow measurements in the extension of certification to modified or add-in boards. At a couple of recent NEBS conferences representatives from Verizon and SBC commented on the desirability of air flow information as part of the documentation required to “leverage” information gained from a prior fire test to cover additional add-in products without the need to repeat a full scale “burn.”

While a full-scale fire test will always provide satisfactory proof of performance for add-on electronics, it is something manufacturers will certainly wish to avoid. The fire-spread test is perhaps the most expensive part of NEBS testing. Because it is a destructive test of valuable equipment, the net test expense to the manufacturer is high relative to most other types of NEBS tests. It is highly desirable that this test not be repeated after the initial product qualification, as a host chassis and a full suite of board assemblies—not just the ones being updated or added to the product—will be required for a full retest.

Fortunately, the effect the new assemblies will have on fire performance may usually be deduced from a simpler set of studies.

In the past, new/updated assemblies were described, for fire-spread purposes, in terms of their performance on a needle flame test and their fuel load, that is, the amount of combustible organic material they contained. This information was then compared to that obtained for assemblies (or PC boards) used in the original product qualification.

Now, the telephone companies are asking for an additional step—airflow analysis. In addition to providing needle flame and fuel load information, manufacturers need to show that airflow measurements taken at various points on proposed new assemblies yield air velocity profiles similar to those obtained with components that were previously tested and then subjected to a full scale fire-spread test.

The additional requirement of airflow measurement adds approximately one day’s additional testing per circuit board, which is a minimal expense when one considers the true costs of a full-scale fire spread test. A rough comparison of the “old add-on”, “new add-on”, and full fire spread test is shown in Figures 1 through 3. The incremental cost of additional airflow tests is negligible in relation to the cost of a full-blown fire-spread test.

Figure 1

Figure 2

Typical H/W Costs	Low	Medium	High
Chassis (Frame, Mother Board, Power Supplies, Filters & Fans)	$1,000.00	$5,000.00	$10,000.00
Number of cards	1	12	18
Card Cost ~$15K (range is $10k to $40K)	$15,000.00	$180,000.00	$270,000.00
Manfacturer’s Equipment Cost	$16,000.00	$185,000.00	$280,000.00
Lab cost for Fire Test	$10,000.00	$10,000.00	$10,000.00
Needle Flame Test (per NEW card)	$1,500.00	$1,500.00	$1,500.00
Air Flow Test (per NEW card)	$1,500.00	$1,500.00	$1,500.00
Old Fire Leverage Cost	$1,500.00	$1,500.00	$1,500.00
New Fire Leverage Cost	$3,000.00	$3,000.00	$3,000.00
Repeated Fire Cost	$26,000.00	$195,000.00	$290,000.00

Figure 3

Concluding Remarks
Cooling considerations are an important part of equipment design, affecting reliability in a number of ways. For telecom equipment, where NEBS compliance is required, forced air-cooling interacts with a number of other required performance parameters. n

About the Author
Tom Naughton heads the Environmental Simulation Laboratory at Curtis-Straus, LLC. He may be reached at tnaughton@curtis-straus.com.

References

GR-63-CORE, “NEBS Requirements: Physical Protection,” Issue 2, Telcordia Technologies
SFPE Handbook of Fire Protection Engineering, 3^rd Ed., 2002, National Fire Protection Association. See Chapter 3-4, “Generation of Heat and Chemical Compounds in Fires,” and Chapter 3-5, “Compartment Fire Modeling”.