Formulating scientifically based specifications to control floor
and tool vibration in IC fabs
Vibration. The word is almost a mantra that sometimes is used
to scare facility owners into justifying additional funding to "fix" or
"prevent" vibration-related issues in any project, be it a tool installation,
a fab retrofit, or the construction of a new facility. Few people in the
construction field are actually trained in the science of the structural
response to vibration, and fewer still have the necessary experience to
effectively deal with the effects and mitigation of vibration-related
The traditional approach to vibration-related problems is to assume
that architects, facility engineers, or tool suppliers will deal with
them. Many simply defer to specifications proposed by such organizations
as IEST or SEMI in the hope that simply incorporating such specifications
into their designs will solve the problem. This approach has not proven
to be effective and can be costly if it is implemented during a production
ramp-up. Many specifications and methods of measuring, responding to,
and correcting for vibration do not necessarily reflect the most optimal
practices or follow well-established scientific principles. Nor do they
often solve the problem.
This article provides theoretically and experimentally verified recommendations
for dealing with vibration issues in the semiconductor fab. These recommendations
are cost-effective and increase the performance of a fab's dynamic system,
including its structure, flooring, walls, infrastructure, and tools. Based
on established scientific practices and experience derived from nonsemiconductor
industries, the article challenges many current methods for dealing with
vibration.Bruce Huling, Motorola
Although modern concepts of vibration and structural dynamics for addressing
lithographic tool installations have been known for several years, the
semiconductor industry has been willing to accept demonstrably incorrect,
poorly defined theories and criteria for fab floor designs.1
As noted in the literature, the industry's failure to apply correct vibration
theories to fab installations has resulted in ill-advised structures,
such as minuscule 10 x 10-ft structural bays, and led to astronomical
construction costs.1,2 The IC industry has begun to change,
but it may still take years for it to fully appreciate a modern approach
to vibration control.
Equipment manufacturers have also been slow to address their floor vibration
needs in easily understandable, scientifically grounded terms. Moreover,
excitation sources are almost never defined and certainly not well understood.
Contrary to widespread belief, properly constructed, reasonably balanced
rotating machinery installations are rarely the major source of excitation.
For example, after 29 large fan towers were direct-mounted to the supporting
floor immediately adjacent to a large fab, their excitations were found
to be almost undetectable on the fab floor. Regardless, the view prevails
that such fan towers must be supported on ineffective "isolation" systems,
although that is not the case for residential fans. Low-stiffness isolation
systems often tend to cause balancing problems and the premature demise
of the fans.
In fact, primary fab floor excitations are typically associated with
cart and personnel movements (footfall), tool and associated equipment
operation, and the motion of suspended ducting and pipes. The validity
of this view has been confirmed by an unpublished study by a major computer
chip manufacturer. Since some fabs' vibration specifications make no mention
of tool operation or cart and personnel movements, such specifications
pertain only to an empty building and are therefore clearly of no value.
Such specifications have apparently been generated to minimize responsibility
and liability rather than to address floor system vibratory performance.
Equally unfortunate is the little-known fact that most manufacturer-provided
floor-vibration criteria are actually not based on scientific studies.
Typical examples are given in Figure 1, which shows stepper floor-vibration
specifications. Although it is reasonable to assume that the steppers
in the illustration would have similar floor-vibration specifications,
they do not. Indeed, some of the specified motions are unrealistic on
nonempty, operational fab floors, where there are people and tools. For
example, the maximum displacement specification for stepper 1(a) of ~5
µin. at a frequency of 40 Hz is simply unachievable.
|Figure 1: Floor-vibration specifications, peak-to-peak displacement
components, for four different steppers.
It should be recognized, however, that it is difficult for toolmakers
to provide valid floor-vibration specifications for structural systems
they have not designed, and virtually impossible using a frequency componenttype
specification for which the excitation is not even defined.
In contrast, a "total vibratory system" is comprehensive because it
includes information on the supporting soil, columns, floor system, tool
pedestal, and tool components.1 Based on the premise that the
source of the vibration can be isolated, this comprehensive view has often
been ignored. Consequently, tool manufacturers, in some cases, have simply
incorporated erroneous fab floor-vibration criteria (which have been generated
by others) into their specification information.
Vibration consultants often formulate incorrect vibration-control specifications
because they have not received adequate scientific training. The field
of structural dynamicsthe study of the dynamics of structures, such
as fab floor and tool-framing systemsis typically a graduate engineering
curriculum requiring an undergraduate background in structures, structural
design, and vibration. However, undergraduate courses in vibration, acoustics,
and electrical or mechanical engineering do not provide any background
in structural engineering. At the same time, a structural engineering
curriculum almost never includes courses in elementary vibration or even
mechanical engineering, much less structural dynamics. Consequently, the
most accurate recommendations for fab and tool-pedestal structural systems
are made by those who have not merely received on-the-job training in
the field of vibration control but who have an academic background in
the science of structural dynamics.
It is unlikely that an elementary vibration consultant without a background
in structural design and engineering can team up with a structural engineer
without a background in vibration to achieve efficient, effective solutions.
Because it is likely that neither expert will really understand what the
other is talking about, except in the most general terms, the requisite
checks and balances will be lacking.
Another significant source of inaccurate specifications is the vibration
measurement taker, that person or company that takes measurements with
a vibration instrument but which has little or no knowledge, beyond the
basics, of how to run the instrument as explained in the operating manual.
Inaccurate specifications may be supported by the warranty statement
accompanying a tool purchase: "The customer must supply verification of
compliance with the vibration specification upon request. Failure to do
so will void the warranty." But what if the specification is invalid?
How is a fab supposed to comply with it? In fact, toolmakers' floor-vibration
specifications are rarely satisfied, leading most toolmakers to accept
precise floor-vibration specifications arrived at on the basis of scientific
Generating Faulty Specifications Based on Frequency-Related Criteria
If vibration specifications for sensitive tools have no scientific basis,
they should not be included in tool information packages. Intensely promoted
but demonstrably incorrect, frequency-related component motion concepts
are a major part of the problem. These concepts result in such inappropriate
specification terms as one-third-octave acoustic band, fast Fourier transform
(FFT), power spectral density (PSD), velocity or acceleration, and frequency-related
components. This practice can be partially attributed to the output provided
by some, but not all, commercially available measurement instruments.
Such representations are often completely confusing to those who must
use them, such as equipment engineers and even tool manufacturers' own
These specifications ignore the following considerations:
- Frequency-related component motion criteria have not been accepted
by qualified structural dynamicists.
- Although toolmakers claim that such criteria are applicable to floor
structural systems, no structural or structural dynamics textbook mentions
them. While Fourier analysis is briefly addressed in elementary textbooks
on vibration, one-third-octave acoustic bands are not. Even if a floor-system
vibratory situation were truly random and broadband, which it is not,
frequency-related motion components must still be appropriately evaluated
for their dynamic responses, although not by means of the methods that
have been presented to the semiconductor industry.3,4
- System frequencies are often not of major importance. Dynamic-response
evaluations are frequently made without even computing frequencies,
even though such frequencies can be easily determined if necessary.1
- Equipment and structural engineers preparing static designs for advanced-technology
facilities typically have no understanding of such concepts as one-third-octave
acoustic band, FFT, or power spectral density and do not know how to
use them for preliminary designs. Such representations are thus of little
or no practical value. Moreover, structural dynamicists do not generally
use these methods. All engineers, however, can understand and measure
total floor displacement.
- The sole use of components-of-motion criteria for evaluating vibratory
response situations was proven to be inadequate many years ago in several
structural dynamics studies at major universities. It was shown that
summations of mathematically valid response spectra components, much
less FFT representations, do not provide precise solutions.4
- Typical operational-fab excitations caused by such factors
as moving carts and personnel cannot achieve the minuscule, ~1 µin.,
displacements associated with component criteria based on 100-µin./sec
velocity. This is easily proven by considering a typical fab floor having
a fundamental frequency of 30 Hz. The displacement for a 100-µin./sec
velocity is 100/(2 x x 30) =
0.53 µin., or approximately 1.1 µin. peak to peak. This value
cannot be achieved even by a well-constructed slab-on-grade floor system.5
Excessively large structural systems are the end result of attempting
to achieve the unachievablea suspended fab floor structural system
with the stiffness of granite.1
Achieving Floor Stability and Vibration Resistance
The stability and vibration resistance of the supporting floor system
must be addressed in some fashion by toolmakers as well as structural
and equipment engineers. There are certainly limber floors and large bay
sizes that are not satisfactory for supporting vibration-sensitive tools.
The criteria for these floor systems, however, must be easily understood
and usable by all parties. Unfortunately, very few understand the frequency-related
motion component criteria that have often been advanced to deal with this
Using floor-vibration criteria based on a broad range of motion-component
frequencies, such as 0100 Hz, completely ignores the fundamentals
of structural dynamics response and frequency pertaining to floor systems
as well as the fundamentals of higher mathematics.1,3,4 For
example, such component measurements do not accurately reflect floor vibrations
caused by footfall or cart excitations.
Most frequencies within the 0100-Hz range are irrelevant to fab
floor-system vibrations. Structural dynamicists know that floor systems
act as "response filters"that is, floors only respond significantly
to excitations that correspond to their fundamental frequencies, filtering
out most of the excitations that might exist in the 0100-Hz range.3
Figure 2 illustrates this fact in its depiction of the actual vertical-response
amplitudes of a fab floor subject to harmonic unbalance excitations from
0100 Hz. As is almost always the case, the maximum response is concentrated
at the floor fundamental vertical frequency of ~31 Hz, which means that
a broadband, frequency-related component criterion has little or no value.
|Figure 2: Floor-vibration response "filtering": peak-to-peak
displacement for various harmonic excitation frequencies. The maximum
magnification factor is only 4, even at the floor fundamental vertical
frequency of 31 Hz.
Floor time-history responses (and frequencies) can be accurately computed
using modern finite element theory.1 In contrast to the types
of broadband random-vibration concepts that proponents seek to apply to
the semiconductor industry, random-vibration structural dynamics theories
have been used to analyze multicomponent rocket structural systems.3,4
A case in point is a rocket structural system with a thin-skin panel with
a fundamental frequency of 100 Hz. Pertinent 100-Hz excitations can be
transmitted to the skin via the broadband frequencies associated with
motor detonation. However, although a tool may have a 100-Hz mechanical
component, the supporting fab floor system with a fundamental frequency
of 2040 Hz will filter out any potentially detrimental 100-Hz excitations.
This scenario presupposes a rational floor (not tool) vibration specification.
If the floor does not transmit the excitation to the tool, the tool's
components will not be affected. Tool vibration and structural dynamics
are another matter. These factors must be addressed by multicomponent
studies that are merited, but rarely performed, for very expensive tools.
The means for specifying allowable floor vibrations are summarized in
Figure 3.1 For vibration-sensitive tools, the total vertical
displacement motion at the center of the bay induced by footfall should
not exceed 100 µin. peak to peak. The associated bay center static
stiffness must be at least 1000 kip/in. Structural engineers should be
able to easily calculate the latter for preliminary static designs. For
tools that are ultrasensitive to vibration, the more-stringent total motion
response of 50 µin. peak to peak and a stiffness value of 4000 kip/in.
should be used. These values do not represent structural overkill but
correspond to a well-constructed slab-on-grade floor system.5
Horizontal motions and stiffness rates should be based on the same values.
The horizontal motions of the floor, even with a minimal shear-wall system,
are typically less than vertical motions, but they still merit careful
consideration. Final theoretical response analyses and evaluations should
always be made by a qualified structural dynamicist.
|Figure 3: Graph used for specifying allowable floor vibrations
to determine the recommended fab floor structural system. Response
to foot impulse excitation versus stiffness points are used to generate
the solid line.
The displacement-based criteria provided in this article, unlike frequency-related
criteria, are easy to use. They have a sound theoretical basis, have been
verified by hundreds of on-site measurements, and have been used at several
hundred tool installations. Equipment manufacturers have approved of these
installations. Furthermore, several manufacturers of sophisticated tools
have provided their own floor-system stiffness recommendations. Finally,
no vibration-related difficulties have been reported at any of the tool
installations whose specifications are based on displacement criteria.
The generation of rational, easy-to-use, floor vibration specifications
is long overdue. In some 25 years of dynamic studies for the semiconductor
industry, the author has rarely encountered anyone who knows how to use
any of the vast menagerie of cumbersome floor vibration specifications
provided by tool manufacturers and acoustics consultants. Most of these
specifications do not even define the excitation and thus have little
This article has presented simple, easy-to-use, floor vibration specifications
that are understandable to all engineers and have been successfully implemented
in many tool installations. As an internationally acclaimed engineer once
stated, "If a solution is not both theoretically sound and easy to use,
it is not the best solution."
- K Medearis, "Rational Vibration and Structural Dynamics Evaluations
for Advanced Technology Facilities," Journal of the Institute of
Environmental Sciences 38, no. 5 (1995): 3544.
- PB Ross, "Moore's Second Law," Forbes Magazine (March 1995).
- SH Crandall, Random Vibrations (Cambridge, MA: MIT Press,
- R Clough, "Earthquake Analysis by Response Spectrum Superposition,"
Bulletin of the Seismological Society of America (1962).
- K Medearis, "Fan Foundation SystemsAnalysis and Design Guidelines"
(Palo Alto, CA: The Electric Power Research Institute [EPRI], 1986).
Kenneth Medearis, PhD, is the founder and technical director
of Kenneth Medearis Associates, Fort Collins, CO. Before organizing the
consulting firm in 1969, he served as a professor of engineering and mathematics
at Colorado State University in Fort Collins, where he set up and directed
the computer center. A registered professional engineer in several states
and a member of the International Standards Organization committees on
vibration and dynamics, Medearis has written a number of scientific publications
and a book entitled Numerical-Computer Methods for Engineers and Physical
Scientists. In 1995 he received the Institute of Environmental Sciences's
Maurice Simpson Award for a technical paper on vibration and structural
dynamics evaluations of advanced-technology facilities. Medearis r eceived
a BS in civil engineering and an MS in structural engineering from the
University of Illinois in Champaign-Urbana and a PhD in structural dynamics
from Stanford University in Palo Alto, CA. (Medearis can be reached at
970/484-3553 or firstname.lastname@example.org.)
Search | Current Issue | MicroArchives
Buyers Guide | Media Kit
Questions/comments about MICRO Magazine? E-mail us at email@example.com.
© 2007 Tom Cheyney
All rights reserved.