Friday, May 22, 2009

Air management techniques can reduce costs and optimize fab performance

Fan type and configuration are important considerations in air recirculation

Advances in fan technology make it possible for fab owners to realize significant savings in operational costs compared to approaches employed in the past. Numerous air-management concepts have been devised over the years using different types of fans to recirculate air in semiconductor fabrication facility cleanrooms. In these facilities, three types of air-moving apparatus predominate: (1) fan-filter units, (2) distributed recirculating air-handling units (RAHs) and (3) fan tower units.

We have attempted to establish a uniform, equitable basis by which to compare these options. To do so, we defined a standard cleanroom module to which each option could be applied. Within the standard module we further defined two cases of filter coverage: 100 percent filter coverage, such as one might encounter in a photolithography area; and 25 percent filter coverage as is typical for a fab equipped with minienvironments.
The standard cleanroom module is 20 feet wide by 80 feet long. This module is intended to represent one segment of a larger ballroom space consisting of similar modules set side by side and mirrored about a central aisle. All filters are mounted in the ceiling and individually supply air at 70 feet per minute (0.35 m/s), regardless of the filter coverage. Air flows vertically in the cleanroom and passes downward through a perforated raised access floor.
In the 100 percent filter-coverage case, the air continues downward through the structural support floor into a clean-classified subfab space. In the 25 percent filter-coverage case, the access floor is made high enough to permit all of the air to return beneath the access floor, allowing the subfab to be non-classified. For both cases, return air then rises vertically at the outer perimeter of the cleanroom. The air-moving apparatus is installed in the outer vertical air passage, or above the cleanroom ceiling. All air-moving systems are acoustically treated as necessary to achieve Noise Criterion 55, which equates to 62.5 dBA. The end result of this uniformity is that from inside the cleanroom there is practically no way to discern the type of air-moving system that is installed.

We compare the three air-management concepts using two criteria: energy effectiveness and capital cost. To evaluate the energy effectiveness, we sought a measurement standard that would apply equally and fairly to a chosen air-management concept and that would be applicable to all locations at any time. The measurement we chose is expressed in terms of power per unit airflow, in this case watts per cfm. The power required to operate a given fan, measured in watts, is constant regardless of geographic location or of the passing years, even though the cost of that power will vary. The airflow rate, measured in cfm, is independent of the fan system in that it is dictated by filter coverage. The fewer watts that are required to move the required air volume the better the energy effectiveness.
We have attempted to include all possible cost impacts of each air-management concept, including mechanical, electrical, instrumentation and control, and building and structural costs. Capital costs are expressed in relative terms, with the expectation that the cost rankings will hold true in other geographical locations and over time.
Conventional approachesFirst let us review what, until recently, has been the state of the art for air-moving apparatus used in the three air-management concepts of this study.
Fan-filter units. Fan-filter units typically consist of a centrifugal plug fan driven by a fractional horsepower motor, mounted in a sheet metal box in whose lower surface is mounted a HEPA or ULPA filter. Sound attenuating material is generally installed inside the box, as the fans tend to be noisy unless acoustic treatment is applied. The motor often has a multi-step or continuously variable speed controller.
Fan-filter units do not develop much static pressure. After air passes through the internal acoustic passages and the final filter, there is comparatively little motive force left to offset external static pressure losses. Care has to be taken to properly design each component in the external airstream to minimize these losses, which are caused by air friction. Often this means enlarging the air passages and over-sizing the sensible cooling coils to keep the air velocity, and thus the resultant air friction, to an acceptable level.

Distributed return air handlers (RAHs). Distributed air handling systems have typically been built around packaged air handling units consisting basically of a sensible cooling coil and a fan. Plug-type centrifugal fans have been employed more often than not due to their relative compactness compared to scroll fans. Sound attenuation is almost always required at the fan discharge. Centrifugal fans can develop as much static pressure as is needed to move the air through the various components of the recirculation loop. However, as the total airflow increases, so does the fan size, the fan cost, and the amount of noise that the fan generates. Therefore, multiple, smaller air-handling units, installed in parallel, typically are used. Figure 1 contains plan and section views of the standard cleanroom module with a conventional distributed RAH air-management system.
Fan Towers. Fan towers are nearly always built with vaneaxial fans. Vaneaxial fans offer the advantage that they can efficiently move large volumes of air against comparatively low static pressure. This is typically the set of conditions under which any cleanroom recirculating fan must operate. Vaneaxial fans also are relatively compact and efficient and, hence, are fairly inexpensive.
A disadvantage of these fans is that a large part of the total pressure generated is velocity pressure. This becomes a significant factor in the selection of the fan. Vaneaxial fans are typically the most noisy of the three fan types in this study. Extensive sound attenuation measures must be taken, which add cost and additional static pressure loss. Figure 2 again depicts the standard cleanroom module, but this time with a conventional vaneaxial fan tower system.
Historically, cleanroom air-moving systems have accounted for a large percentage of the energy budget. Referring to Figure 3, note that the energy effectiveness of traditional fan systems is approximately 0.4 watts per cfm. For a typical large fab with a recirculation rate of 2,000,000 cfm, operating 24 hours per day, and an electricity cost of $0.08 per kWh, this amounts to an annual cost of over $500,000.
Recent advancesIn the last few years, fan systems have been introduced that make much more effective use of fan energy and thereby significantly reduce the annual operating cost.
Fan-filter units are now available with the fan, the motor and the motor controller engineered together as a package for optimum efficiency. One such package features a brushless, electronically commutated dc motor with an external rotor. The fan impeller is fitted directly onto the rotor. This mounting configuration allows heat transfer from the motor to be optimized. A Hall-effect sensor is used to detect the position of the rotor magnet each time it rotates. Control circuitry then precisely adjusts the motor voltage to match the torque requirement of the fan, thereby minimizing inefficiencies due to slip. Overall, the resultant motor efficiency is 75 percent to 80 percent, compared to less than 40 percent for phased split capacitor or shaded pole motor designs. With this improved efficiency comes the byproduct of quieter operation.

Because the fan uses a dc motor, its speed is infinitely variable. The controller can be set up so the rotational speed of the fan is remotely monitored and controlled. The on-off status of each fan-filter unit also can be remotely controlled and monitored.
Figure 4 depicts plan and section views of a fan-filter unit installation for the 100 percent filter coverage case. Note that the air passageways are quite large, which has a corresponding impact on the overall building size.
The 25 percent filter coverage case is depicted in Figure 5. The lower airflow allows the size of the air passageways to be reduced, so the effect on the building geometry is also greatly reduced.
Vaneaxial fan packages are now available that combine advanced fan engineering with aerodynamically and acoustically engineered sound attenuators. This type of package may be applied in either a fan tower or a distributed RAH to yield a quiet, efficient system.

To achieve the desired level of performance, the vaneaxial fan must be manufactured to close tolerances; the fan barrel must be nearly perfectly cylindrical so that the fan-blade clearance is minimized. Unlike centrifugal fans, two fan-selection variables can be manipulated to achieve optimum performance: rotational speed and blade-pitch angle. Finally, the inlet and outlet sound attenuators are custom engineered to maximize sound-attenuation performance without imposing the high static pressure penalty of off-the-shelf attenuators.

Click here to enlarge image

In addition to these performance enhancements, the vaneaxial fan tower package can also be arranged so as to simplify maintenance. Each fan is spring-mounted on a movable dolly that is normally locked into place and attached to the inlet and outlet sound attenuators with a simple drawband. Ordinarily the fans are maintenance-free—there is only one moving part. However, when it is necessary to maintain or replace the fan, it is disconnected from the sound attenuators and wheeled as an assembly out of the air passageway and into an adjacent maintenance bay.

Figure 7 is a plan and section view of a distributed RAH installation for the 100 percent filter coverage case.
If needed, vaneaxial fans can be selected with additional static pressure capacity to accommodate such items as prefilters or chemical filters. Fan-filter units generally do not have this capability. For uniformity, these items are not considered for any of the fan options analyzed in this paper.

Figure 6 depicts the physical configuration of a distributed RAH using a vaneaxial fan. Air enters the outer annulus, passes through a cylindrical cooling coil and enters the fan. The air is then discharged into the supply plenum via a sound attenuator. Access to the fan is gained through a hinged door or removable panel on the outer case of the RAH.
The 25 percent filter coverage case is shown in Figure 8. Distributed RAHs require the installation of a structural fan deck to support the equipment. The interstitial space is part of the return air stream and must therefore be designed and operated to clean protocols.

Figure 9 contains plan and section views of a fan tower installation for the 100 percent filter coverage case. A tall, wide vertical chase is needed to accommodate the equipment. Air is drawn into the apparatus from the clean subfab and is discharged directly into the supply plenum. The fan itself resides in non-classified space.
The 25 percent filter coverage case is shown in Figure 10. Here the fan is mounted horizontally on a structural fan deck in the non-classified interstitial space. An air duct directs the supply air to the pressurized plenum directly below the deck.
Energy effectiveness—Figure 11 is a graph of the relative energy effectiveness for the 100 percent filter coverage case, applying the newer fan technologies to the three air-management concepts.
For each air-management option the static pressure loss, measured in inches water column, was calculated across the recirculation loop. The static pressure losses vary depending on the nature of the air passageways, and each option has unique characteristics. In all options the static pressure was intentionally kept to a practical minimum. Fans were then selected which would meet the airflow and static pressure requirements with the lowest power input. The static pressure requirement for each option is plotted against the corresponding energy effectiveness expressed in watts per cfm. The operating point for each option is indicated. Vaneaxial fans are in reality selected based on total pressure, which is the sum of static pressure and velocity pressure; however only the static pressure value is indicated to serve as a standard basis for comparison with the fan-filter unit option (fan-filter units use centrifugal fans which are selected based on static pressure only). The watts per cfm reported for the vaneaxial fan options includes the power necessary for static and velocity pressure.

As might be expected, the vaneaxial options tend to out-perform the fan-filter unit option because the vaneaxial fans and motors are inherently more efficient. Figure 12 is a graph of the energy effectiveness for the 25 percent filter coverage case. Again, the performance of the vaneaxial fan options is superior to that of the fan-filter units. In both the 100 percent case and the 25 percent case, the energy effectiveness is about double that of the older technologies, which means that the annual operating cost is halved.
Capital costs—Turning to capital costs, we must look not only at the cost of the air-moving equipment, but at the total cost that each air-management concept imposes on the facility in which it is installed. Tables 1 and 2 summarize these facility impacts.
Comments—The equipment cost of quality fan-filter units is about half-again higher than either distributed RAHs or fan towers. The basic reason is quantities. Even though the unit cost of a fan-filter unit is low compared to the other air-moving options, in our cleanroom module there are 200 such units and the total cost quickly exceeds that of the others.

The fan-filter units require larger air passageways in order to reduce static pressure losses. This increases both the overall building width and height. Distributed RAHs require a greater building height, similar to what is required for fan-filter units, but less building width. Fan tower units require a greater building width, but do not add to the height. A structural fan deck is required to support the distributed RAHs, but is not needed for the other two options.
Greater cooling coil surface is needed for the fan-filter units than for the other two options in order to keep the static pressure loss down. Each distributed RAH is supplied with a sensible cooling coil, so cooling water must be piped to each unit.

Fan-filter units can be set in a gasketed ceiling. We have assumed pressurized supply plenums for the other two options, which require a gel-track ceiling.
The electrical distribution cost is highest for fan-filter units. Even though the motors are much smaller, there are many more of them.

Finally, automation costs are higher for fan-filter units, again due to the large quantity of fans. Fan-filter automation cost can be and often are lowered by eliminating the monitoring functions, but this is done to the disadvantage of those who need to maintain the system. Even so, the comparative cost ranking of the three options remains intact.
Comments—The cost spread for the three fan options is all but eliminated in the 25 percent filter coverage case.

The lower total airflow allows for air passageways whose dimensions are similar for each equipment option. However, since the subfab is no longer used as an air passageway, the fan tower unit is now too long to mount vertically. Instead, it is placed horizontally overhead on a structural fan deck. This eliminates the need for a wide vertical return air passageway. The building interstitial height in the case of fan-filter units is driven only by the need for the plenum to be accessible for maintenance purposes. The other two options require greater interstitial heights to house equipment and for the pressurized supply plenums. A structural fan deck is required to support the distributed RAHs and the horizontally mounted fan tower.
Greater cooling coil surface is still needed for the fan-filter units than for the other two options. Cooling water must still be piped to each distributed RAH.
The electrical distribution cost is now highest for distributed RAHs due to the reduced quantity of individual fan-filter units that need power wiring. The same relationship also now applies with respect to automation costs.
Figure 13 is a graph of the total relative costs of each option. For the 100 percent filter coverage case there is a clear distinction between the three air-management options. Facility costs are highest for the fan-filter unit option and lowest for the fan tower option. By reducing the amount of automation, the cost of fan-filter unit option draws closer to that of distributed RAHs.
For the 25 percent filter coverage case, the capital costs are approximately the same, with or without reduced fan-filter unit automation, although distributed RAHs are slightly higher.
ConclusionFan-filter units with electronically commutated dc motors operate much more efficiently than do models with conventional motors. Vaneaxial fans can now be obtained that are both efficient and quiet. Properly engineered vaneaxial fan systems also offer overall facility capital cost savings compared to other systems. Vaneaxial fan systems are simple in concept and are relatively easy to maintain. Vaneaxial fan systems also can be selected to meet additional static pressure requirements imposed by such items as prefilters or chemical filters.
AcknowledgmentsThanks to Michael O'Halloran for coming up with the basic idea for this paper, and to Rod McLeod and Willy Kohne for reviewing the draft manuscript. Thanks also to Steve Dikeman of AcoustiFLO Ltd and Zareer Cursetjee of Cleanpak International for help with the vaneaxial fan selections.
Editor's note: Variations of this article were presented at SEMICON Singapore 1999 and at CleanRooms Asia 2000.

No comments: