Dr. Mei Wu is the founder of Mei Wu Acoustics (www.mei-wu.com). She has been working in the field of acoustics and noise control for over 20 years. She has experience in both consulting and research. As a consultant she provides services in architectural acoustics, mechanical system noise and vibration control, and environmental noise control. As an R&D engineer she has worked on active noise cancellation, specialty silencer development, and finite element and boundary element modeling. She has been in charge of hundreds of projects, including many projects for the microelectronics industry. Dr. Wu has published over 30 technical papers in professional journals and at conferences.
It often happens that the noise level in cleanrooms served by Fan Filter Units (FFU’s) significantly exceeds the level expected by the design team. Sometimes it even exceeds the level measured in a mock-up cleanroom. This paper examines the reasons for the discrepancy. It shows by theoretical analyses and laboratory test results how some factors affect the noise generated by FFU’s and the noise in cleanrooms served by FFU’s. These factors include the flow rate and total static pressure of FFU fans, and the amount of sound absorption and number of FFU’s in a cleanroom. It also explains the common mistakes with mock-up tests. In the end, the paper recommends several ways to specify FFU noise and provides methods to predict cleanroom noise levels.
Fan filter unit, fan powered filter, fan-powered HEPA, noise, vibration, recirculation air system, cleanroom.
Fan Filter Units (FFU's) are used increasingly by the microelectronics industry to provide recirculation air in cleanrooms. Such units usually consist of one or more direct-drive fans discharging into a small plenum then at a HEPA or ULPA filter. They are used in mini-environments and general cleanrooms, especially in retrofit cleanrooms with limited height. The units are located on a ceiling suspension system. There is no space for external noise mitigation. The critical step to ensure that a cleanroom will meet the chosen noise criteria is to select units with acceptable noise characteristics. However, the popular noise specifications provided by manufacturers provide littler information in determining the resultant noise level in a cleanroom served by the units. This paper analyzes the shortcomings of the popular noise specifications, recommends more precise specifications, and provides methods to predict cleanroom noise levels based on the recommended specifications.
The most popular way used by FFU manufacturers to specify FFU noise is to claim a dB(A) (A-weighted sound pressure level) or NC (Noise Criterion) rating. Cleanroom design teams often take these ratings as the resultant noise level in a cleanroom served by the FFU’s. They get surprised when the noise level in a finished cleanroom to be around 70 dB(A) and NC-65 [Reference 1], while the manufacturers’ ratings are about 55 dB(A) and NC-45.
Such large discrepancy between the manufacturer data and actual cleanroom noise levels are caused by several reasons. One of the reasons is that the manufacturers’ data are usually obtained from tests in an acoustical environment, which yields lower sound pressure readings than in a cleanroom, when the noise generated by each FFU is the same. Other reasons may include that the tests are conducted so that the FFU fan operates at a lower total static pressure than in a cleanroom, which causes the fan to generate lower noise. The following sections show the effects of acoustical environment and fan operating point.
Many manufacturers test their FFU’s in a large room with carpet and acoustical ceiling, or even in an anechoic chamber. Usually there is one unit operating during the test. The distance between the microphone and the FFU discharge is usually about 3 feet. This environment is very different from a cleanroom environment. In a cleanroom there is barely any sound absorptive material and there are many FFU’s operating, sometimes several hundreds covering the whole ceiling. Take for example a 100x80 square feet cleanroom with 95% of its ceiling covered by FFU’s. According to our calculations, the noise level inside this cleanroom is 14 dB higher than the noise level measured in a test condition described above, if the noise generated by each FFU is the same. This may change the dB(A) rating from 55 dB(A) to 69 dB(A) and change NC rating from NC-45 to NC-65 (if the critical frequency band is the 125 Hz band).
It has been established in Reference 2 that the noise generated by an FFU depends on the flow rate of the fan, or the face velocity at the filter discharge. Noise level increases with face velocity. The amount of noise increase depends on the characters of the fan and other factors.
Figure 1 shows the sound power levels measured from a 2x4 square foot FFU with an ebm fan. It was tested at 480, 600, 720, and 880 cfm (at face velocity 60, 75, 90, and 110 fpm). The unit was operating in the center of a large room with no additional pressure load applied. The flow rate was controlled by adjusting the rotating speed of the fan. It can be seen that the noise generated by the unit increased by about 10 dB in the high frequency bands when the face velocity increased from 60 fpm to 110 fpm. This explains why it is seen in cleanrooms served by FFU’s with velocity control that the rooms are noisier when the FFU’s are operating at higher speed.
Figure 3 shows the effects of total static pressure on yet another 2x4 unit with an ebm fan. The unit was tested at three different static pressures while the face velocity was kept roughly unchanged. The sound power levels increased by about 10 dB when the total static pressure increased from 0.45-inch w.g. to 1.2-inch w.g.
Clearly, higher total static pressure causes higher noise, although some fans are more sensitive than others.
From the analyses above we conclude that FFU noise specification should identify the operating point of the fan (flow rate and total static pressure). It should also identify if the noise data are sound power levels or sound pressure levels. If the data are sound power levels, it should identify the size of the unit, because larger units generate more noise. If the data are sound pressure levels, it should identify the size of the unit and the measurement environment.
As a noise control consultant, the author prefers sound power levels over sound pressure levels, because sound power levels describe the acoustical properties of an FFU, and they do not depend on the acoustical environment of the test site. However, sound pressure levels can also be used to specify FFU noise. Sound pressure levels are easier to measure than sound power levels, and they can be converted into sound power levels easily if they are measured in controlled acoustical environments.
A recommended way is to measure sound pressure levels at a few inches from the filter discharge. These sound pressure levels can be converted into sound power levels by adding ten times the 10 based logarithm of the area of the filter in square meters. As long as the ambient noise is significantly lower than FFU noise, the filter is located away from large reflective surfaces, such as walls and floor, and the measurements are conducted within 2 or 3 inches from the filter discharge, the acoustical environment of the test site does not affect the test results much.
The other way is to measure the sound pressure levels at a fixed distance, say 36 inches, from the filter discharge with the filter located in a large room. As long as the room is large enough, the ambient noise is significantly lower than FFU noise, and the filter is located away from large reflective surfaces, such as walls and floor, the test results are not significantly affected by the acoustical environment. These sound pressure levels can be converted into sound power levels, although the procedure is more complicated than that to convert the levels obtained by the first method.
The second method is similar to the popular method used by the manufacturers. The sound pressure levels obtained by the second method are usually lower than the levels obtained by the first method. This may be the reason that manufacturers favor the second method. However, once the data are converted into sound power levels, the levels should be the same, if the conversion is done correctly.
It is impossible for FFU manufacturers to specify a generic noise level for cleanrooms served by their FFU’s, because cleanroom noise levels depend not only on the noise character of FFU’s but also on the acoustical character of the room and number of FFU’s in the room.
It is important that FFU noise specification identifies the FFU fan operating point, such as flow rate (or filter face velocity) and total static pressure. It is also important that the operating point was determined accurately during noise measurements. Our experience is that among rotation speed, flow rate, and total static pressure, the rotating speed of a fan is the easiest to measure, and the static pressure is the most difficult to measure. Because of the compact nature of FFU’s, airflow in the discharge plenum is very turbulent. It is almost impossible to find a location to measure the static pressure in the plenum.
It is also difficult to measure the flow rate of the fan. For the work published in Reference 2, we used a revolving vane anemometer or a Velgrid to measure the flow at the filter discharge. We later found that the readings can vary largely with the ways the readings are taken. For the work of this paper, we used a Pitot tube to measure the flow velocities across a traverse plane in the long straight pipe attached to the FFU inlet. However, we had trouble finding a traverse plane with uniform flow distribution.
Once the sound power levels generated by an FFU have been determined, the sound pressure levels in a cleanroom can be calculated from the following equation:
SPL = PWL + 10 Log ( N ) + 10 Log ( 4/R ).
In the equation, SPL represents the average reverberant field sound pressure level in dB (re 20 µpa); PWL represents the sound power level generated by each FFU in dB (re Watt); N represents the number of FFU’s in a cleanroom; and R represents the total amount of sound absorption in the room in Sabins (square meters). The logarithms are 10 based.
From the equation we can see that sound pressure levels in a cleanroom depend on the sound power levels generated by each FFU, the number of FFU’s in the room, and the amount of sound absorption in the room. As expected, the more noise each unit generates, the more units in a room, and the less sound absorption in the room, the higher noise levels in the room. In the next three sections, we use this equation to demonstrate the effects of number of FFU’s and sound absorption in a room.
If the sound power level generated by each FFU and the amount of sound absorption in a room remain unchanged, every time the number of units doubles, the noise level in the room increases by 3 dB. For a cleanroom with 50% of its ceiling covered by FFU’s, the noise level is 3 dB higher than that in the same room with 25% ceiling coverage. For a cleanroom with 100% ceiling coverage, the noise level is 3 dB higher than that with 50% ceiling coverage, and 6 dB higher than that with 25% ceiling coverage. In actual cleanrooms, since the ceiling area not covered by FFU’s will be covered by sheet metal panels, which provide some sound absorption and reduce noise levels, cleanrooms with high ceiling coverage are a little more noisier than predicted above.
The total amount of sound absorption in a room equals the summation of the sound absorption coefficient of each material multiplies the area of the material in square meters. If the sound power level generated by each FFU and the number of FFU’s in a room remain unchanged, every time the amount of sound absorption doubles, the noise level in a room decreases by 3 dB.
In office spaces, sound absorption is mostly provided by the acoustical ceiling and carpet. The sound absorption provided by gypsum board walls is negligible. In a cleanroom, in the absence of acoustical ceiling and carpet, cleanroom wall panels, which has similar sound absorption coefficients as gypsum board, become important sound absorption providers. HEPA (or ALPA) filters and return air openings are also sound absorption providers. Filters and return air openings have higher sound absorption coefficients than cleanroom wall panels (return air openings have a sound absorption coefficient close to 1.0). But since wall panels cover a very large area and they have a reasonably good absorption coefficient in the 125 Hz band, the panels usually provide an important part of sound absorption in the critical frequency band of 125 Hz band, especially in medium to small cleanrooms.
Since small cleanrooms have more wall area per filter unit than large cleanrooms, they are usually slightly quieter than large rooms. For example, if a large cleanroom (100x200 square feet) and a small cleanroom (20x20 square feet) each has 50% ceiling covered with the same type of filters and return air openings to allow 500 feet per minute airflow, the small room is 3 dB quieter than the large room. For very small rooms, such as mock-up cleanrooms, the difference is even more significant.
For cleanrooms with less than 100% ceiling coverage, the non-filter-covered part of the ceiling can be covered with cleanroom acoustical ceiling panels (panels are available for Cleanness Class 100 to 10,000). This increases the amount of sound absorption in a room and reduces the noise. For example, if 50% of the ceiling of the large cleanroom described above is covered with NRC 0.85 cleanroom acoustical ceiling tiles, the noise level is reduced by 3 dB.
A mock-up cleanroom is usually built to simulate a cleanroom and therefore to predict the noise level in a cleanroom. However, there are cases where the noise level measured in a finished cleanroom turns out to be higher than the level measured in a mock-up cleanroom. A major reason for this discrepancy is that a mock-up cleanroom has more sound absorption than an actual cleanroom.
A mock-up cleanroom is usually a small room, say 4x4 square feet, built with gypsum board, sheet metal, or cleanroom wall panels for walls and FFU’s for ceiling. A gap of several inches is usually left between the bottom of the walls and the floor to allow airflow. The sound absorption in this room is mostly provided by the wall panels and return air opening. Although gypsum board and sheet metal have similar sound absorption coefficients as cleanroom panels, a mock-up room has significantly higher sound absorption than an actual cleanroom, because the area of wall panels per filter unit is appreciably larger than that in an actual cleanroom. And the return air opening area is appreciably larger too. Take for example a 4x4 square foot mock-up room with 100% filter ceiling and an 8-inch high return air opening along all four walls. Compare it with a 100x80 square foot cleanroom with 100% filter ceiling served by the same type of FFU’s. The noise in the mock-up room is 6 dB lower than that in the large cleanroom, because of the extra amount of sound absorption pre filter unit.
Another reason for a low noise reading in a mock-up room may be the difference in FFU fan operating point. If the fans serving the mock-up room operate at a lower total static pressure, or lower flow rate, the noise generated by these fans will be lower than that in an actual cleanroom.
This paper shows through measurement results how fan flow rate and total static pressure affect sound power levels generated by FFU’s. It also shows analytically how acoustical environments affect the noise level in cleanrooms. It suggests ways to specify FFU noise and ways to predict cleanroom noise levels based on the specification.