205:

Theory of Vibration Measurement

Peter H. Sydenham
GSEC Pty Ltd, Adelaide, South Australia, Australia

2 PHYSICAL FEATURES OF VIBRATION

1 Definition of Vibration
2 Physical Features of Vibration
3 Mathematical Description
4 Amplitude and Frequency of Vibration
5 Importance of Damping
6 Application to Vibration Measurement
7 Data Processing
8 The Literature on Vibration
Related Articles
References

1391
1391
1394
1394
1395
1396
1396
1397
1397
1397

1 DEFINITION OF VIBRATION
Vibration here is the oscillatory mechanical motion of an
object. It is a dynamic state. Several other measurable
parameters fit under the term vibration.

Relative position
Velocity
Acceleration
Jerk (the derivative of acceleration)
Dynamic force.

For each parameter it may be the instantaneous value,

average value, or some other descriptor that is needed.
Vibration, in the general sense, occurs as a periodic
oscillation, a random movement, or as a transient motion,
the latter more normally being referred to as shock when
the transient is large in amplitude and brief in duration.

Vibration can occur in linear or rotational forms of motion,

the two being termed translational or torsional vibrations
respectively. In many ways, the basic understanding of
each is similar because a rotational system also concerns
displacements in space. Translational forms are outlined in
the following description. An equivalent rotational system
usually exists for all arrangements described.
In vibration measurement, it is important to decide
whether a mechanical sensor can be physically attached
to the test object called a contacting or intrusive method.
If mass cannot be added, then a noncontacting technique
is necessary.
Detailed measurement of vibration can be a most complex measuring task. The basic requirement is to determine
features of the motion of a point, or more practically an
extended object in a three-dimensional space relative to a
reference framework; see Figure 1.
A point in space has three degrees of freedom. It can
translate in one or more of the three directions when
referred to the x, y, z Cartesian coordinate system shown.
Rotation has no meaning for a purely point measurement. To monitor free motion of a theoretical point object
three measurement-sensing channels are required to capture all three motions so that the vector of direction can be
reconstructed.
If the object of interest has significant physical size, it
must be treated as an extended object in which the rotations
about each of the three axes described above provide a
further three rotational degrees of freedom. Thus, to monitor
the free motion of a realistic object we may need up to six
sensor channels, one for each degree of freedom. These
may be built into a single sensor housing and processed to
give absolute values.

Handbook of Measuring System Design, edited by Peter H. Sydenham and Richard Thorn.
2005 John Wiley & Sons, Ltd. ISBN: 0-470-02143-8.

1392

Common Measurands
+z
+x

a
y
f
0
q
x

Object at
center of
Cartesian
coordinate
framework

+y

x
y
z
a
f
q

Translation
components
Rotational
components

Figure 1. Possible motions of an extended object in space relative to a Cartesian framework.

In practice, some degrees of freedom can be nominally

constrained (but are they really?) possibly eliminating the

need for some of the six sensor channels. Practical installation should always contain a test that evaluates the degree
of actual constraint possible because vibration sensors will
usually produce some level of output for the directions of
vibration they are not primarily seeking to measure. This
is called their cross-axis coupling, transverse response, or
some such terminology.
In many installations, the resultant direction of the
motion vector might lie in a constant fixed direction with
time. In such cases, in principle, only one sensor will be
required provided it can be mounted to sense in exactly that
direction. If not, as is often the case, more than one unit
will be required, the collective signals then being combined
to produce the single resultant.
The potentially available frequency spectrum of vibration parameters extends, as shown in Figure 2, from very
slow motions through frequencies experienced in machine
tools and similar mechanical structures to the supersonic
megahertz frequencies of ultrasound.
1

0m

0
10

0m

0m

10

0m

1.

1m

100 000

1000 000 0.

1m

10 000

100

1000

Acceleration 'g '

10

100

1.0

10

0.1

1.0
0.1

0.001

0.01

0.0001

0.001

1.

10

0.01

0.

m
1

100

1000

10 000

0.0001

0.00001
100 000

1.0

m
1m

0.

0.0

mm

m
0

1.

m
1

0.

m
1

0.0

Frequency Hz

0.

01
0.

0.

m
m

1.

m
1

0.
0

10

m
00
1
m
0.
00
01
0.
m
00
00
1
0.
m
00
00
01
0.
00
m
00
Di 001
sp
m
lac
em
en
t

1.0

0.

01

lo
Ve

0.000001
0.1

cit

0.00001

mm

Acceleration ms2

1000

1.
0

100 000 0.0

10 000

0.
1

Figure 2. Frequency spectrum and magnitude of vibration parameters. (Reproduced by permission of Bruel & Kjaer.)

Theory of Vibration Measurement 1393

It is not possible to cover this range with a single
general-purpose sensor. Each application will need careful
consideration of its many parameters to decide which kind
and model of sensor should be applied in order to make the
needed measurement.

The complicating factor in vibration measurement is the

distributed nature of mechanical systems that enables the
energy causing a vibration to flow from a point. This
leads to complex patterns of vibration, requiring care in
the positioning of sensors.

Head (axial model)

(ca. 25 Hz)

Eyeball, intraocular
structures (30 80 Hz)

Shoulder
girdle
(4 5 Hz)

Lung
volume
Chest wall
(ca. 60 Hz)
Lower arm
(16 30 Hz)

Handarm

Spinal
column
(axial
model)
(10 12 Hz)

Abdominal
mass (48 Hz)

Seated person
Hand grip
(50 200Hz)
Legs
(variable from
ca. 2 Hz with
knees flexing
to over 20 Hz
with rigid posture)

Standing person

Figure 3. Mechanical systems can be modeled in terms of springs, masses, and dampers; model of the human body. (Reproduced by
permission of Bruel & Kjaer.)

1394

Common Measurands

3 MATHEMATICAL DESCRIPTION

Fixed framework

Mechanical systems, including the human body given as an

example in Figure 3, comprise mass, spring compliance (or
stiffness), and damping components.
In the simplest case, where only one degree of freedom
exists, linear behavior of this combination can be well
described using linear mathematical theory to model the
time behavior as the result of continuous force excitation, or
some initial position displacement see Article 60, Zeroorder System Dynamics, Volume 1; Article 61, Firstorder System Dynamics, Volume 1; Article 62, Secondorder System Dynamics, Volume 1 for a general analysis
of linear motion.
Vibration can be measured by direct comparison of
instantaneous dimensional parameters relative to some adequately fixed datum point in space. The fixed point can
be on an independent measurement framework (fixed reference method) or can be a place that remains sufficiently stationary because of its high mechanical inertia (seismic system) that holds it relatively constant in
position.
The accelerometer is the main vibration sensor see
Article 207, Acceleration Measurement, Volume 3. It is
formed with a mass, and a spring having some damping.
If operating in its linear range, it will exhibit a secondorder linear system output response, qo , that is related
to an input function, qi , by the generalized differential
equation
a dq
a2 d2 qo
+ 1 o + ao qo = qi
2
dt
dt

(1)

For the specific mechanical vibratory system given in

Figure 4, this becomes
2

cdx0
md x0
+
ks x0 = qi
dt 2
dt

(2)

where m is the effective mass (which may need to include

part of the mass of the spring element or be composed
entirely of it), c the viscous damping factor, and ks the
spring compliance (expressed here as length change per
unit of force applied).
In practice, these three components may be distributed
to a varying extent across an object. Often, they can be
adequately considered using the lumped model shown.
Where the degree of distribution is significant, then analysis
becomes much more complex and is studied using such
computer-based methods as the Finite Element Method
(FEM).

Spring
compliance
ks

Damping
device
c

Motion of mass
of distance x o
Effective mass m

Input
function
qi

Force exerted F

Figure 4. One degree of freedom, spring-mass-damper system

model.

4 AMPLITUDE AND FREQUENCY OF

VIBRATION
Where the damping effect is negligible, the system will
have a frequency at which it will naturally vibrate if excited
by a pulse input. This natural frequency n is given by


n =

ks
m

(3)

Presence of damping will alter this vibration frequency

value, but as the damping rises the system is less able to
provide continuous oscillation.
The static sensitivity is given by the spring constant,
either as ks , the spring compliance, or as its reciprocal term
stiffness that is expressed as force per unit extension.
The influence of damping is readily described by a
dimensionless number, called the damping ratio, given by
c
= 
2 ks m

(4)

It is usually quoted in a ratio form that relates its magnitude

with respect to that at = 1.
These three important parameters are features of a spring
system as a whole formed from the object under test, plus
any contributions from an added sensor, any protective
cover, and its wiring. The parameters are independent of
the input driving function.
Such systems have been extensively analyzed when
excited by the commonly seen input forcing functions, the
step, impulse, ramp, and sinusoid. A general theory for handling any input function other than these is also available.
In practice the step, impulse, and continuous sinusoidal
responses are used in analyses, as they are reasonably easy
to apply in theory and in practical use.

Theory of Vibration Measurement 1395

5 IMPORTANCE OF DAMPING

Note the resonance buildup point at n , and that it

is limited by the existing degree of damping. Thus,
the damping of the system to be measured by addition of the sensor itself, if it is of the seismic kind,
can be of importance as a modifier of likely system
responses.
A well-known measurement error situation is where an
oscillatory response is being plotted with a mechanical chart
recorder. If the response of the chart drive mechanism
(itself a second-order system) is not adequate, the record
is attenuated as the oscillation frequency of the measured
input rises. The same effect can arise with poorly set up
vibration-measuring systems.

From the plots of different damping ratios for a step input

given in Figure 5, it can be seen that as the damping factor
increases the response to a transient step force input
(applied to the mass) can vary from a sinusoidal oscillation
at one extreme (called underdamped ) to a very sluggish
climb to the final value (overdamped ). The nominal value
of = 0.7 is called the critical damping state as this is the
fastest rise to its final value.
In the case of a continuous sinusoidal force being applied
to the input, the system frequency response varies as shown
in Figure 6.

2.0

Damping
ratio

Increasing
damping

1.8
1.6
0.25

1.4
1.2
Final
Output
1.0
value
displacement
x0
0.8

0.5
0.75
1.0
2.0

1.5

0.6

2.5

0.4

5.0

0.2
Initial value
0

Normalized time wn t

Figure 5. Displacement responses of a second-order system to an input force step.

Theoretical peak
at
2.0

Increasing
damping
0.5

0.3
0.4

Instantaneous
amplitude x 0

1.0
0.7
1
0.5

0.2

2
D
amp
ing

5
ra
t

io 10

0.1

0.1
w
wn

0.2

0.5

1.0

Natural
frequency
wn

Figure 6. Displacement responses of a second-order system to a continuous sinusoidal force input.

10

1396

Common Measurands

As damping increases, the system response takes on the

form of the lower first-order exponential response system,
and it cannot oscillate.
The above discussion, with respect to vibration of the
measurand is also the basis of understanding the operation
of seismic vibration sensors see Article 208, Amplitude
and Velocity Measurement, Volume 3.
Second-order measuring systems, therefore, will have a
natural frequency of vibration. This is the frequency at
which they vibrate when given impulse energy that is not
overridden by continuous forced vibrations. Thus, a sensing
system that is second order and not damped will (because
of noise energy inputs) produce outputs at its natural frequency that are not correlated with frequencies occurring
in the system of interest. Use of vibration sensors must,
therefore, recognize these limitations by selecting designs
that position their resonant frequency correctly, and that
possess the right degree of damping. Such issues are covered in Article 206, Practice of Vibration Measurement,
Volume 3.

6 APPLICATION TO VIBRATION
MEASUREMENT
In practice, it is also often more convenient to sense vibration by an indirect means and obtain the desired unit by
mathematical processing. For example, accelerometers are
conveniently used to obtain motional forces (from force =
mass acceleration) and hence stresses and strains. Acceleration signals can be twice integrated with respect to time
to yield displacement. Sensors that operate as velocity transducers can yield displacement by single integration.
Integration of data is generally preferred to differentiation
as the former averages random noise present to a smaller
value compared to the signal, whereas the latter, in reverse,
will deteriorate the signal-to-noise ratio.

7 DATA PROCESSING
Signals obtained in vibration measurements will take many
forms depending on the vibration characteristics of the
object under test. These will include

repeating pulses,
sine waves,
complex waveforms,
noise.
Their nature can be

continuous, having the same signal waveform,

varying waveforms with time,

modulated amplitudes,
modulated frequencies,
short bursts,
random.

The kinds of information to be extracted from the raw

signal can
be frequency and magnitude of an oscillation;
be identification of the source of pulses and unwanted
noises;
impact behavior as objects collide or are deliberately
impacted;
be the frequency spectrum of a signal that establishes
relative amplitudes and energy distributions; in vibration, spectral analysis is also taken on a spectrum;
be a prediction of the future condition of equipment.
These signals are used to establish such things as

wear in bearings,
noise inside a vehicle,
dangerous vibrations that can cause mechanical failure,
response of loud speakers,
where to attach balancing weights,
machine health monitoring,
earthquake research,
building vibration in wind,
where to apply noise dampening materials, and so on.

Thus, the signal processing of vibration signals often has

to use the full power of signal processing in both the analog
and digital domain.
The methods used for general analog and digital signal
processing, in general, are applicable see Article 142,
Z-transforms, Volume 3; Article 143, DFT and FFTs,
Volume 3; Article 144, DSP Chip Sets, Volume 3; and
Article 145, DSP Tools, Volume 3.
Texts devoted to vibration measurement cover the special methods used, Harris and Piersol (2001), McConnell
(1995), Thomsen (2003), and Wowk (1991).
Today, as much as possible this signal processing is carried out in the Digital Signal Processing, DSP domain.
These methods are very powerful and make use of dedicated DSP chips and processors. Tool kits are commonly
available and have become so widespread that many of the
analysis tools are now routinely built into equipment, test
sets, and into the vibration sensor system itself.
The sensor, however, is usually analog by nature so
the extraction of signals in high noise environments is
often essential see Article 176, Signals and Signal-tonoise Ratio, Volume 3; Article 178, Noise Matching and
Preamplifier Selection, Volume 3; Article 180, Bandwidth Reduction of Baseband DC Signals, Volume 3;

Theory of Vibration Measurement 1397

Article 181, Amplitude Modulated Signals: The Lockin Amplifier, Volume 3; Article 182, Boxcar and Signal
Averagers, Volume 3; Article 183, Correlators in Signal
Extraction, Volume 3; and Article 184, Photon Counting, Volume 3.

RELATED ARTICLES
Article 180, Bandwidth Reduction of Baseband DC Signals, Volume 3; Article 206, Practice of Vibration Measurement, Volume 3; Article 207, Acceleration Measurement, Volume 3; Article 208, Amplitude and Velocity
Measurement, Volume 3.

8 THE LITERATURE ON VIBRATION

There exist many general books on the kinds of transducers that are used to measure the various vibration variables. Some of the titles published have become standard
sources carrying forward considerable unchanging knowledge, Bruel and Kjaer (1982), Trampe-Broch (1980), and
Harris and Piersol (2001).
Whereas the data processing has changed greatly over
the past decades, the sensing principles and nature of
vibration remain the same, making older works valid for
many aspects of vibration.
Few general instrument reference texts, such as Webster
(1999), Dyer (2001), address the subject of vibration as a
distinct topic. Relevant material will be found there under
such headings as velocity and acceleration measurement,
accelerometers, position sensing, and piezoelectric systems.
There are, as would be expected, works entirely devoted
to vibration and related measurands. The following will
be of value to those who require more than the limited
introduction that an article such as this can provide,
Crandall (1959), Harris and Piersol (2001), Ling and
Shabana (1996), McConnell (1995), Mobley (1999), Smith
(1989), Steinberg (2000), Thomsen (2003), Timar (1989),
and Trampe-Broch (1980).
The various trade houses that manufacture vibrationmeasuring and testing equipment often provide extensive
literature and other forms of training aids to assist the uncertain user. Their application engineers also offer assistance
in vibration system design and set up.

REFERENCES
Bruel & Kjaer. (1982) Measuring Vibration-Elementary Introduction, Bruel & Kjaer, Naerum.
Crandall, S.H. (1959) Random Vibration, Wiley, New York.
Dyer, S.A. (2001) Survey of Instrumentation and Measurement,
Wiley, New York.
Harris, C.M. and Piersol, A.G. (2001) Harris Shock and Vibration Handbook, McGraw-Hill, New York.
Ling, F.F. and Shabana, A.A. (1996) Theory of Vibration: An
Introduction, Springer Verlag, Berlin.
McConnell, K.G. (1995) Vibration Testing: Theory and Practice,
Wiley Interscience, New York.
Mobley, R.K. (1999) Vibration Fundamentals, ButterworthHeinemann, Oxford.
Smith, J.D. (1989) Vibration Measurement and Analysis, Butterworth-Heinemann, Oxford.
Steinberg, D.S. (2000) Vibration Analysis for Electronic Equipment, Wiley Interscience, New York.
Thomsen, J.J. (2003) Vibrations and Stability: Advanced Theory,
Analysis, and Tools, Springer Verlag, Berlin.
Timar, P.L. (1989) Noise and Vibration of Electrical Machines,
North-Holland, Amsterdam.
Trampe-Broch, J. (1980) Mechanical Vibration and Shock Measurements, Bruel & Kjaer, Naerum.
Webster, J.G. (1999) The Measurement, Instrumentation, and
Sensors Handbook, CRC Press, Boca Raton, FL.
Wowk, V. (1991) Machinery Vibration, McGraw-Hill, New York.

206:

Practice of Vibration Measurement

Peter H. Sydenham
GSEC Pty Ltd, Adelaide, South Australia, Australia

1 Mass-spring Sensors
2 Areas of Application
3 Cross Coupling, Cabling, and Amplifiers
4 Influence Effects on the Sensor
5 Loading by the Sensor
6 Duration of use of Sensor in Tests
7 Amplitude Calibration
8 Accelerometer Calibration
9 Shock Calibration
10 Force Calibration
Related Articles
References

1398
1402
1402
1404
1404
1404
1405
1406
1406
1406
1406
1406

1 MASS-SPRING SENSORS
Many vibration measurements make use of the massspring, seismic, sensor system. The seismic sensor is mostly
described in the literature as the accelerometer, but this use
is not the general case.
Other methods of vibration measurement used include
inductive, capacitive, and optical methods that are primarily
noncontact displacement measuring methods having a fast
response and small range capability see Article 191, Displacement and Angle Sensors Performance and Selection, Volume 3. The strain gauge, a contact method of
displacement measurement, is also able to deliver vibration data; this is discussed in Article 192, Strain Sensors,
Volume 3.
Given the appropriate design of spring-mass-damping
combination a seismic sensor attached to a vibrating surface can yield displacement, velocity, or acceleration data.

The principles and theory of this kind of sensor are covered

in Article 205, Theory of Vibration Measurement, Volume 3. Unfortunately, the conflicting response needs of the
three do not enable a single design to be used for all three
cases. It is, however, often possible to derive one variable
from another by mathematical integration or differentiation
of the data.
The design of vibration sensors can be most sophisticated. Fortunately, the user of vibration sensors rarely
becomes involved in its design but must appreciate the
design features in order to avoid incorrect use. Minor
digressions from best practice can lead to metrological error
that is not easy to detect.
Two forms of seismic sensor exist. The first, called
the open-loop sensor makes use of the motion of a mass
moving relative to the support case to operate either a displacement or a velocity sensing transducer. These possess
some undesirable properties that often rule them out of practical use.
The second form closes the loop (and is, therefore,
referred to as a closed-loop or servo seismic sensor) using
the output signal to produce an internal force that retains
the mass in the same relative position with respect to the
case, the magnitude of the force needed to effect balance
being a measure of the vibration parameter. As will be
explained, this method while being more sophisticated is
much more effective.

1.1 Open-loop sensors

The fundamental schematic arrangement of the open-loop
seismic sensor form is as given in Figure 1. There the mass,
spring, damper, and sensing element are clearly identified.

Handbook of Measuring System Design, edited by Peter H. Sydenham and Richard Thorn.
2005 John Wiley & Sons, Ltd. ISBN: 0-470-02143-8.

Practice of Vibration Measurement 1399

Mass moves relative to case

Damping
Sensor
case

Mass

Displacement or
velocity sensor
Electrical
signal output

Spring
Input
movement
Plate of interest
that is vibrating

constancy of spring rate with time

temperature
influence effects
suitability to be manufactured and packaged to produce
a suitable sensor unit.

Mounting arrangement
(one kind)

Figure 1. Schematic layout of open-loop, seismic form of vibration sensor.

In a real design, the case and the mass are also part of
the systems stiffness and they have to be made adequately
stiff. Actual construction can vary widely depending upon
how the spring force and damping are provided and upon
the form of the sensor used.
The spring element is sometimes produced as a distinct mechanical assembly. Figure 2 is an example made
using a flexure-strip spring suspension. Alternatively, perforated membranes, helical coils, torsion strips, disks, helical
springs, solid compliable pieces, and the like can be used,
see Article 85, Elastic Regime of Design Design Principles, Volume 2.
Often the compliance of the mass itself may be the spring
element, for example, in the piezoelectric crystal, which
also acts as the sensing element and contributes some of
the damping needed.
Important design parameters of a spring element used for
measurement are its
compliance
amplitude range
fatigue life

All masses have compliance to some extent. This has

to be allowed for in the support packaging design or else
the case can modify the measurement amplitudes obtained.
Except for sensors with the higher natural frequencies, the
masses used to form the support case can be regarded
as completely rigid compared to the compliance of the
spring element.
Design of these sensors uses the best of fine mechanical element design techniques as are covered in
Article 80, Principles of Fine Mechanics Kinematic
and Elastic Designs, Volume 2; Article 81, Principles of Fine Mechanics Systems Considerations,
Volume 2; Article 82, Kinematical Regime Members
and Linkages, Volume 2; Article 83, Kinematical
Regime Fasteners, Bearings, Volume 2; Article 84,
Kinematical Regime Rotary Motion, Volume 2; Article 85, Elastic Regime of Design Design Principles,
Volume 2; Article 86, Elastic Regime of Design Spring
Systems, Volume 2; Article 87, Elastic Regime of
Design Plates and Bimorphs, Volume 2; and Article 88, Error Sources in Fine Mechanics, Volume 2.
Rotary equivalent forms of the linear arrangement shown
in Figure 1 are also available.
Sensing methods that have been used in vibration sensors
include the electrical-resistance sliding potentiometer, variable inductance (see, for instance, Figure 2), variable reluctance, variable capacitance, metallic strain gauges (metallic
bonded and also unbonded use as shown in Figure 3),
semiconductor strain gauges, piezoelectric crystal and magnetostrictive elements, position-sensitive optical detectors,
Linear variable differential
transformer LVDT senses
mass motion

Sensitive
direction
of
accelero
meter

Accelero meter
housing
attached to
subject of
interest

Cantilever
spring

Cantilever
spring
Seismic mass
(core)
Sensitive axis
of sensor

Figure 2. Spring vibration sensor using a parallel flexure-strip spring suspension and inductive sensor of mass displacement.

1400

Common Measurands

sing

Sen

PSD, and electromagnetic principles that can provide direct

velocity sensing.
Sensors of the mass-spring form are usually encapsulated
and set up to be used as an accelerometer. The encapsulation takes many forms, ranging from miniature units of total
weight around just 1 g through to 0.5 kg units where the
sensing mass must be physically large to get the response
down to low frequencies. Simultaneous measurements in
two, or three, orthogonal directions are often required. To
meet this need, seismic sensors are made that consist of
two, or three, units that are mounted in different directions,
as shown in Figure 4.
Temperature invariably affects the metrological performance of sensors and here also it changes, to a varying degree, the characteristics of the displacement sensing method, the spring rate, the damping coefficient, and

axis

Seismic mass

Figure 3. Displacement sensing of mass motion using unbonded

strain gauges.

M
M

P
B

S
C

Center mounted
compression
(CM)

Inverted center
mounted compression
(ICM)

P
M
M
P

R
C

F
Annular shear
(AS)

F
Delta shear
(DS) (3 components)

S = Spring M = Mass B = Bass C = Cable

P = Piezoelectric element R = Clamping ring
F = Fastening surface

Figure 4. Examples of small size single- and three-axis accelerometers using piezoelectric displacement sensing. (Reproduced by
permission of Bruel & Kjaer.)

Practice of Vibration Measurement 1401

the case dimensions. Effective design see Article 85,
Elastic Regime of Design Design Principles, Volume 2
and Article 88, Error Sources in Fine Mechanics, Volume 2 can reduce these effects considerably, but to get
the best from a design of high performance vibration sensors can require compensation for temperature change.
This is either performed in the electronic circuitry
as a correction to the data using a temperature sensor
placed accordingly or by incorporating some form of
thermomechanical device into the spring-layout to reduce
the error at source.

1.2 Servo vibration sensors

The performance of open-loop seismic sensors can be
improved with respect to their sensitivity, accuracy, bandwidth, and output signal amplitude by forming the design
into a closed-loop operation.
Figure 5 gives the schematic diagram of a simple form
of closed-loop system, which is based upon a moving-coil
actuator and a capacitance position sensor. This particular
kind is described here, as the elements of the closed-loop
system are clearly evident.

The inertial mass upon which the acceleration is to be

exerted is able to rotate on the end of a freely supported
arm. This is attached to the electrical coil placed in the
permanent magnetic field supplied by the magnet assembly.
Acceleration applied to the mass attempts to rotate the arm
thereby causing displacement of the arm. This unbalances
the capacitive displacement sensor monitoring the relative
position of the pendulous mass. The displacement signal
produces an input to the difference-sensing amplifier. The
amplifier drives a corresponding electric current into the
coil, causing the arm to rotate back to the null displacement
position. Provided the loop response is rapid enough, the
mass will be retained in a nearly constant place relative
to the displacement sensor. Acceleration variations are,
thereby, converted to variations in coil current.
In this way, the displacement sensor is used in the
preferred null-balance mode see Article 126, Electrical
Bridge Circuits Basic Information, Volume 2 and
Article 127, Unbalanced DC Bridges, Volume 2
wherein error of linearity and temperature shift are much
reduced. In this design, only the more easily achieved
proportionality between coil current and force is important.
Servo instruments are further described in Article 113,

Acceleration
input
Position

Inertial
mass

Position
sensor

DC amp

Demodulator

EO R
L

Restoring force

(a)

Torque
generator

Electrical
output

Feedback current

Damping
network

Pivot axis
Pivot

Permanent
magnet
of torque generator

Supporting
arm

Input
current to
moving coil

Position
sensor
Pendulous
inertial
mass
(b)

Jewel
bearing

Acceleration
input

Figure 5. Component layout of a simple form of closed-loop accelerometer.

Common Measurands

Force-feedback Sensors, Volume 2, in Jones (1982), and

sections in Webster (1999).
Microminiaturization has provided considerable scope to
form mass-spring systems in microelectro mechanical systems (MEMS) devices see Article 85, Elastic Regime
of Design Design Principles, Volume 2; Article 163,
Uses and Benefits of MEMS, Volume 3; and Article 164,
Principles of MEMS Actuators, Volume 3. The complete seismic spring system can be formed at micrometer dimensions, meaning that the mass is very much less
than constructions formed of mixed materials as shown in
Figure 4.
The spring element is made in many forms cantilever,
double-ended supported beam, vibrating strings, and disks.
With such small masses, these are able to operate at
much higher vibration frequencies and shock rates than
traditional forms.
They are used in many applications such as airbag
sensors and also in accelerometer sensors for generalpurpose use. Solid-state manufacture, however, is not able
to fully compete with the performance of the best traditional
forms of sensor because of production variations and
parameter drift over time.

Vibration level

1402

Figure 6. Vibration levels observed are recorded for comparison

with a normal frequency spectrum signature.

beginning to fail and when it will reach that state. It can

then be conveniently replaced before a major untimely
break down occurs. Figure 7 shows the set of equipment
commonly needed to set up one of these systems.
Further information on machine health condition monitoring can be found in Reeves (1999), Rao (2000), and
Mobley (2002).

2.2 Energy exploration

2 AREAS OF APPLICATION
When searching for information about a measurement
technique, it is often helpful to have an appreciation of
the allied fields that use the same equipment. Vibration
will be of interest in many different applications, but some
can be singled out as the main areas to which commercial
marketing forces have been directed to enhance certain
sensor characteristics.

When a very robust sensor is needed, the energy exploration

application field is a good place to start looking. The phrase
Measurement While Drilling (MWD) has been coined for
their in-process measurements.
For this market, ruggedized vibration sensors are needed
to withstand the working conditions existing deep down
drilling boreholes, wherein shock, high temperatures, and
pressure demand much more protection than does a laboratory unit. There temperatures up to 185 C and shock levels
of 1000 g can arise in practice.

2.1 Machine health monitoring

A significant field of interest is that of machine health
monitoring, or condition monitoring. Failures in machines
can often be avoided by listening to the sounds and
vibrations made by the system and making changes early
enough. An example is shown in Figure 6, where the
position of vibration sensors and their typical frequency
spectra are given. Vibration and other forms of sensor are applied to the operating system, first while running in its early life to provide a base calibration for
ongoing comparison, and then at periodic intervals during life.
If the frequency/amplitude data (the so-called signature)
has changed from the baseline shape, then this can provide
diagnostic information, suggesting which component is

2.3 Space
In space navigation and control and inertial grade navigation, some key requirements are small in size and weight,
high in reliability, and low in noise levels. As an example,
one unit offered for this market has an RMS noise level up
to 10 Hz of 0.6 g.

3 CROSS COUPLING, CABLING, AND

AMPLIFIERS
With vibration measurement, it is all too easy to produce
incorrect data. This and the following section headings

Practice of Vibration Measurement 1403

Vibration transmitter

Transmitter/monitors

Human-machine interface and A to D

converters

Low-cost accelerometers

BNC junction boxes

and switch boxes

4 Channel vibration monitoring

system

Figure 7. Range of equipment needed to instrument a machine health monitoring system. (Courtesy STI, Sales Technology Inc.)

address several important installation conditions that should

be carefully studied for each new application.

3.1 Cross-coupling and cross-axis sensitivity

A vibration sensor is usually designed with one principal measuring axis that is perpendicular to its mounting surface.
Transducers will exhibit cross-axis coupling to some
extent. This arises because the sensing system will also
be partially sensitive to components of transverse motion.
This means maximum sensitivity is not exactly aligned
with the axis of symmetry of the mounting case but
will be the resultant vector of the cross-axis sensitivity for two orthogonal planes through the center of the
sensor.
Best practice measurements include a test that vibrates
the sensor in a direction perpendicular to the direction of
normal use.
Once the cross-axis sensitivity is established, a value
judgment is needed to decide if this error exceeds the
allowable limit for the application to hand; special care
is needed where complex vibration is involved.
Rotational (torsional) sensitivity may also be important.
These tests can be avoided each time they are used if
the sensors are precalibrated for this source of error and, of
course, are still within calibration. An upper limit of 3 to

5% is offered by suppliers as the upper bound; most sensors

will calibrate under this magnitude.
Sensors that have no cross-axis sensitivity quoted should
not be used until it is established that it does not impact
on the measurement needed. For example, conducting
a comparison test of two objects placed into the same
vibratory conditions will not be influenced much by any
cross-axis sensitivity, unless its value is way above the 5%
level, whereupon the signal could become saturated and
thus limited.
In the case of a triaxial accelerometer (three units
mounted together on each of the three-dimensional axes),
any high magnitude cross-axis output sensitivity can corrupt
data if the principal motion is at right angles to the
base.
A quality sensor will be calibrated and usually marked
on the sensor with a red spot placed at the point of
maximum transverse sensitivity. A polar plot of the output
might also be supplied that is tied into the position of the
spot.

3.2 Coupling compliance

The compliance of the bond made between the sensor and
the surface it is mounted on must be adequately stiff. If not,
the surface and the sensor form a system that can vibrate in
unpredictable ways. As an example, an insufficiently stiff

1404

Common Measurands

mounting can give results that produce much lower frequency components than truly exist. In extreme cases, the
sensor can be shaken free as it builds up the unexpectedly
low-resonance frequency of the joint to dangerous amplitude levels. As a guide, the joint should be at least ten
times stiffer than the sensor mounting so that the resonant
frequency of the joint is well above that of the sensor.

Vibrations apparently occurring at electric mains frequency

(5060 Hz) and harmonics, thereof, are most suspect.
Measurement of mechanical vibration at these frequencies
is particularly difficult because of the need to separate true
signal from influence error noise signals.

3.3 Cables and preamplifiers

Vibration sensors contain mass. As this mass is made

smaller, the sensitivity of the sensor usually will be reduced,
so units with larger mass are more desirable. However,
addition of mass to a vibrating system can load the mass
of that system, causing shifts in frequency and amplitudes
from the normal state of operation.
For this reason, manufacturers offer a wide range of
attached type sensors. Provided the mass added is, say, 5%
or less of the mass of interest, then the results will usually
be acceptable.
Cables can reduce the overall mechanical compliance
by forming a mechanical bridge from its holding down
point to the sensor case. This can corrupt the true signal
magnitude. They can also increase the damping, thereby
shifting frequencies.
Where a system is particularly sensitive to external
loading, the use of noncontact, fixed-reference methods may
be the only way to make a satisfactory measurement.

Certain types of sensors, notably the piezoelectric (PZT)

kind, are sensitive to spurious variations in capacitance and
charge. Sources of such charges include the tribo-electric
effect occurring in the bending motion of the vibrating connecting cables; mechanical stress in the material produces
electrical signals as the cable bends. Special kinds of connecting cable are used that allow for movement of the cable.
Another source of potential error arises from any varying
relative humidity, as this can alter electric field leakage; this
becomes important when designing long-term installations.
The preamplifier stage is also important to get right for
the PZT sensing element has very high output impedance.
Such issues are largely overcome by integrating the preamp
with the sensor unit.

4 INFLUENCE EFFECTS ON THE

SENSOR
Ideally, the sensor should operate in a perfect environment wherein sources of external error, called influence
parameters, are insignificant see Article 16, Reduction
of Influence Factors, Volume 1.
In vibration sensing, possible influence effect error
sources include temperature variation of the sensor, possible
magnetic field fluctuations (especially at radio frequency),
and existing background acoustic noise vibrations. Each of
these might induce erroneous signals.
A good test for influence parameters is to fully connect
the sensor system, observing the output when the measurand of interest is known to be at zero level, such as by
clamping it in a very stiff manner. The output will then
comprise internal noise of the sensing channel and internal
defects, such as system mains supply hum, plus contributions of any impacting influence effect. If the sensing noise
is not observable, then more gain is needed to ensure the
test is valid.
Where practical, the important error inputs can be
systematically varied to see the sensor response. Many
a vibration measurement has finally been seen to be
worthless because some form of influence error turned
out to be larger than the true signal from the measurand.

5 LOADING BY THE SENSOR

6 DURATION OF USE OF SENSOR

IN TESTS
When damping of a structure is very small, as in many
mechanical structures, the time taken for a resonance to
build up to its peak value is long. When using forced
vibration frequency scanning to seek resonances, it is
important not to sweep the excitation input frequency
too rapidly.
As the Q Factor rises, the time to reach full amplitude
of the resonance can take many minutes in large-size
mechanical systems.
One often-used test is that for obtaining the vibration
frequency spectrum of a machine tool as its tool post
is excited, as it would be when machining metal. The
time taken to sweep through the complete spectrum can
take many tens of hours. The issue then is to be certain
that the scanning and measuring equipment maintains its
declared specifications over such long continuous periods
of operation.
Some tests are best done in the time domain using an
impulse or a step increase, the time domain data resulting
being transformed into the frequency domain. Impulse tests
need special care to avoid occurrence of zero shift and

Practice of Vibration Measurement 1405

ringing errors, caused by nonlinearities, in electrical aspects
of amplifiers and the sensor for which residual mechanical
strain can also arise if the spring element is overloaded
during such shocks.

7 AMPLITUDE CALIBRATION
Here, attention is drawn to the issue of sensor calibration.
Whereas the fixed-reference methods do have some relevance in the practical measurement of vibration, the need
for a convenient fixed datum is very often not met.

Static amplitude (displacement) is easily calibrated using

a standardized micrometer, displacement sensor, or optical
interferometer. The motional element of the sensor is
deflected a known amount and the output measured.
Note that some types of accelerometer, such as the
PZT type, are not DC-connected and thus have zero static
output at any position; there has to be motion present, at a
relatively rapid rate, to generate an output.
Dynamic calibrations may be made either by comparison, using a technique of known accuracy and frequency
response, or by using a calibrated vibration generator.
Dynamic amplitude of the non-DC-connected types requires use of a noncontacting method of measuring the

Accelerometer
to be
calibrated
Actuate

Vibration control
generator

Drive
coil
Velocity
coil
Vibration exciter

Sense

Comparator

(a)
Measured
signal

Accelerometer
to be
calibrated

Reference
signal

Comparator

Reference
accelerometer
Vibration
Ag sin wt
(b)
Accelerometer
to be
calibrated
Vibration
Ag sin wt

Mirror

Beam
splitter

Helium neon laser

l = 632.8 nm

Photo
detector

True RMS voltmeter

10 000

Frequency counter

(c)

Figure 8. Calibrating accelerometers. (a) Calibrated vibration exciter shaking accelerometer at calibrated levels; reciprocity method.
(b) Back-to-back calibration of a calibrated accelerometer against one to be calibrated; comparison method. (c) Absolute measurement
using optical interferometry.

1406

Common Measurands

displacement taking place. This is viewed on a display

screen along with the sensor output. Viewing the signal is
recommended as this allows the wave shape to be checked.
Once the nature of the signal has been validated, the
data can be processed into more directly useful charts or
numerical values.

8 ACCELEROMETER CALIBRATION
Figure 8 shows outlines of three methods of calibrating
accelerometers and other vibration-measuring sensors. Calibration is normally performed at 500 rad s1 . Figure 8(a)
is the reciprocity method. A calibrated vibration exciter
shakes the accelerometer at calibrated levels. Figure 8(b)
shows the comparison method. This uses back-to-back calibration of a calibrated accelerometer against the one to
be calibrated. Figure 8(c) is a noncontact method using
optical interferometry; this gives an absolute not relative
determination.
Other methods that can be used are to subject the
accelerometer to accelerations produced by the Earths
force. Simple pseudostatic rotation of an accelerometer in
the vertical plane will produce accelerations in the 0 to
1 g range (g is used here for the Earths acceleration
value). Larger values can be obtained by whirling the
accelerometer on the extremity of a rotating arm of a
calibrating centrifuge. Alternatively, it can be mounted on
the end of a hanging pendulum.

9 SHOCK CALIBRATION
Short-duration acceleration, as produced by impact, requires
different approaches to calibration. Accelerations in such
applications as explosive devices can exceed 10 000 g and
last for only a few milliseconds.
A commonly used method is to produce a calibrated
shock by allowing a steel ball to free-fall on to an anvil
on which is mounted the sensor. This method provides
an absolute calibration but, as with all of the methods
described, has uncertainties associated with the practice of
the method. In this case, one source of error is caused by
the difficulty of releasing a ball to begin its downward path
without imparting a small negative or positive velocity at
time zero.
Another method uses a long pendulum carrying a large
mass. This is allowed to swing down to impact on the
object carrying the sensor under test. Physical consideration
allows the rate of energy release to be calculated and thus
the impact forces, but it is not that accurate. It is useful
for relative calibrations where the sensor under test and a
standardized unit are mounted in the same impacted object.

10 FORCE CALIBRATION
Static forces can be calibrated by applying dead weights to
the force sensor, the weights being calibrated masses.
Dynamic forces arising in vibration can more easily be determined using the relationship force = mass
acceleration. A shaking table is used to produce known
accelerations on a known mass. In this way, the forces
exerted on the accelerometer can be determined along with
the corresponding output voltage or current needed to produce transducer sensitivity constants.
At the completion of any calibration, the determination
and support data about the method of calibration must be
recorded according to the organizations policy. This ensures
traceable results in the event of dispute arising later see
Article 45, Calibration Process, Volume 1. The period
between calibrations is covered in Article 46, Calibration
Interval, Volume 1.
Space does not permit greater explanation; detailed
accounts of vibration sensors are available in the literature,
Trampe-Broch (1980), Bruel and Kjaer (1982), Smith
(1989), Timar (1989), Wowk (1991), McConnell (1995),
Ling and Shabana (1996), Mobley (1999), Steinberg (2000),
Dyer (2001), Harris and Piersol (2001).

RELATED ARTICLES
Article 57, Static Considerations of General Instrumentation, Volume 1; Article 59, Introduction to the
Dynamic Regime of Measurement Systems, Volume 1;
Article 205, Theory of Vibration Measurement, Volume 3; Article 207, Acceleration Measurement, Volume 3; Article 208, Amplitude and Velocity Measurement, Volume 3.

REFERENCES
Bruel & Kjaer (1982) Measuring Vibration-Elementary Introduction, Bruel & Kjaer, Naerum.
Dyer, S.A. (2001) Survey of Instrumentation and Measurement,
Wiley, New York.
Harris, C.M. and Piersol, A.G. (2001) Harris Shock and Vibration Handbook, McGraw-Hill, New York.
Jones, B.E. (1982) Feedback in Instruments and its Applications,
in Instrument Science and Technology, (ed. B.E. Jones), Adam
Hilger, Bristol.
Ling, F.F. and Shabana, A.A. (1996) Theory of Vibration: An
Introduction, Springer Verlag, Berlin.
McConnell, K.G. (1995) Vibration Testing: Theory and Practice,
Wiley Interscience, New York.

Practice of Vibration Measurement 1407

Mobley, R.K. (1999) Vibration Fundamentals, ButterworthHeinemann, Oxford.

Steinberg, D.S. (2000) Vibration Analysis for Electronic Equipment, Wiley Interscience, New York.

Mobley, R.K. (2002) An Introduction to Predictive Maintenance,

Butterworth-Heinemann.

Timar, P.L. (1989) Noise and Vibration of Electrical Machines,

North-Holland, Amsterdam.

Reeves, C.W. (1999) The Vibration Monitoring Handbook, (Coxmoors Machine and Systems Condition Monitoring Series),
Coxmoor Publishing Co., Oxford.

Trampe-Broch, J. (1980) Mechanical Vibration and Shock Measurements, Bruel & Kjaer, Naerum.

Rao, J.S. (2000) Vibratory Condition Monitoring of Machines,

CRC Press.
Smith, J.D. (1989) Vibration Measurement and Analysis, Butterworth-Heinemann, Oxford.

Webster, J.G. (1999) The Measurement, Instrumentation, and

Sensors Handbook, CRC Press, Boca Raton, FL.
Wowk, V. (1991) Machinery Vibration, McGraw-Hill, New
York.

207:

Acceleration Measurement

Peter H. Sydenham
GSEC Pty Ltd, Adelaide, South Australia, Australia

1 Use of Seismic Sensors

2 Optimal use of Seismic Sensor for
Acceleration Measurement
3 Typical Accelerometers
4 Response to Complex Waveforms
5 The Piezoelectric Sensor
6 Amplifiers for Piezoelectric Sensors
7 Microaccelerometers
8 Signal Conditioning in Microaccelerometers
9 Measurement of Shock
Related Articles
References

1408
1408
1409
1409
1410
1411
1411
1411
1412
1413
1413

1 USE OF SEISMIC SENSORS

Of the many measurands that appear under the heading of
vibration see Article 205, Theory of Vibration Measurement, Volume 3 the most measured is acceleration.
The fixed-reference method of measuring acceleration is
rarely used, determinations being made more easily with
the mass-spring, seismic form of sensor. This method uses
the inertial mass as a short-term fixed reference: the test
object moves whilst the mass stays sufficiently stationary
in space for short periods of time.
For the seismic sensor system, the mass and the spring
compliance are both fixed entities. Consideration of the
F = m a (force = mass acceleration) law with spring
compliance being linear shows that displacement of the
mass relative to the sensor case is proportional to the
acceleration of the case.

The curves, plotted in Figures 5 and 6 of Article 205,

Theory of Vibration Measurement, Volume 3 for sinusoidal input of force to a second-order system, are applicable for obtaining the output response curves of accelerometers using displacement sensing. In this case, the vertical
axis is interpreted as the relative displacement of the mass
for a given acceleration.

2 OPTIMAL USE OF SEISMIC SENSOR

FOR ACCELERATION MEASUREMENT
The curves, Figure 6 in Article 205, Theory of Vibration
Measurement, Volume 3, show that a seismic sensor will
provide a relatively constant output sensitivity representing
sensor acceleration from very low vibration frequencies
up to around 0.5 of the natural frequency of the springmass arrangement used. Further, the damping ratio can be
optimized at around 0.50.6 and electronic compensation
added (if needed) to raise the upper limit a little further
than the undamped resonance point.
At first sight it might, therefore, appear that a single,
general-purpose design could be made having a very high
resonant frequency. This, however, is not the case for the
deflection of the spring (which is a major factor deciding
the system output sensitivity) is proportional to 1/n2 . In
practice, this means that as the upper useful frequency
limit is extended, the sensor sensitivity falls off rapidly.
Electronic amplification allows low signal output to be used
but with additional cost to the total measuring system. As
the signal level falls, added amplification will also increase
the noise level.
At the very low frequency end of the accelerometer response, the transducers become ineffective as

Handbook of Measuring System Design, edited by Peter H. Sydenham and Richard Thorn.
2005 John Wiley & Sons, Ltd. ISBN: 0-470-02143-8.

Acceleration Measurement 1409

10

103

Acceleration, g

102

10

10

10

104

103

10

10

Velocity amplitude

Operating
range

102

10

101

10

Peak-to-peak
displacement

n
um tio
im ra
ax le
M cce
a

10

n
tio
ra
le
ce
ac

Figure 1. A range of accelerometers is required to cover the full

needs of vibration measurement. (Courtesy, Honeywell Sensing
and Control.)

105

Lower frequency limit

M
di ax
sp im
la um
ce
m
en
t

102

um
im
in
M

As a guide to the range of capability available, one

major manufacturers catalog for traditional accelerometers not made in the more recent microminiature form,
offers accelerometers with sensitivities ranging from a
small 30 V ms2 through to 1 V ms2 with corresponding sensor masses of 3 g and 500 g and useful frequency
ranges of 1000 Hz and 0.2 Hz. As the mass increases,
the frequency range falls or, conversely, to get low frequency response the mass has to increase. Sensors have
been constructed for higher frequencies using the MEMS,
microelectro mechanical system technology as is discussed
in Section 7. A selection of traditional accelerometers is
shown in Figure 1.
The many constraints placed upon the various performance parameters of a particular seismic sensor can be
represented in numerical terms on a single chart, Figure 2,
Harris and Crede (1976).
It is not always appreciated that all materials exhibit
some degree of mechanical compliance and that the modulus of elasticity is a property of the material, not the way
it is used. Elastic structures, such as helical springs, are
used to increase the compliance to suit a particular task
see Article 85, Elastic Regime of Design Design Principles, Volume 2.
To obtain the stiffness needed, accelerometer springs
may occur as stamped rigid plates, as flat cusped spring
washers, or even as a sufficiently compliant clamping bolt.
In applications of piezo-sensitive materials, use is often
made of the compliance of the material itself.

102

10

10

3 TYPICAL ACCELEROMETERS

101

1.0

103

Upper frequency limit

accelerometers because the accelerations produce too small

a displacement to be observed against the background noise
level. By increasing the mass, the units can be made to
respond to lower frequencies see Article 208, Amplitude
and Velocity Measurement, Volume 3.

104

105

Frequency

Figure 2. Useful linear operating range of an individual seismic

vibration sensor can be characterized with this type of chart.

4 RESPONSE TO COMPLEX
WAVEFORMS
The response curves given in Figures 5 and 6 of Article 205, Theory of Vibration Measurement, Volume 3
relate to seismic sensors excited by sinusoidal signals.
To predict the behavior of a certain sensor, such as an
accelerometer, used to measure other kinds of continuous or discrete waveforms, it is first necessary to break
down the waveform into its Fourier components see Article 29, Relationship Between Signals in the Time and
Frequency Domain, Volume 1. The response, in terms of
amplitude and phase, to each of these is then added to form
the resultant response.
It has been stated above that damping can be added to
extend the useful bandwidth of a seismic sensor. However,
where this is done it can increase the phase-shift variation
with frequency; that is, the components are sensed with
different phase shifts to the undamped case and thus the
reconstructed signal is no longer a fully faithful replica of
the original motion.
A signal comprising many frequencies will, therefore,
produce an output that depends largely on the damping
and natural frequency values of the sensor. A number of
responses can be plotted, such as shown in Figure 3 from
Harris and Crede (1976). Generally, the damping value for
best all-round results is that near the critical damping value
of 0.7.

1410

Common Measurands

2.0

Seismometer indicated
acceleration output
Maximum acceleration
of input

Damping
factor
z

0.4
0.7

1.0

Half-sine input
pulse

0.1

0
Seismometer
output responses
for varying degrees
of damping
1.0

1.0t
Time
Natural period of seismometer ~ duration
of half-sine pulse

Figure 3. Example of response, at various damping-factor levels,

of a seismic accelerometer to a complex forcing input-half-sine
wave of similar period to that of the natural resonance period of
the sensor.

5 THE PIEZOELECTRIC SENSOR

Numerous sensing methods have been devised to measure
the motion of the mass in a seismic sensor. Discussed here
is the most commonly used PZT, piezoelectric transducer
method: others are described in Norton (1969), Oliver
(1971), Webster (1999).
Force applied to certain crystalline substances such as
quartz produces, between two surfaces of a suitably shaped
crystal, an electric charge that is proportional to that force.
This charge is contained in the internal electrical capacitance formed by the electrically insulating, high-dielectric
material and its two conducting surfaces deposited on each
face.
The first approximation descriptive mathematical relation
for this effect is
q = a F Ks

(1)

where q is the electrical charge generated by force, F (in

Newtons), applied across the faces of a piezoelectric device
having a mechanical compliance of spring rate Ks (mN1 )
and a material constant a (of dimensions C m1 ).
The constant a, depends on many factors including the
geometry of the crystal, position of electrodes, and material
used. Typical materials now used (natural quartz is less
sensitive and, therefore, less applicable) include barium

titanate with controlled impurities, lead zirconate, lead

niobate; many others are available but their composition
is a trade secret.
The usual PZT sensing material is made from loose powder that, after shaping, is fired at a very high temperature.
While cooling, the blocks are subject to an electric field
that traps a polarization.
Sensitivity of PZT materials is temperature-dependent
on the charge sensitivity and the capacitance value, both
of which alter with temperature. These effects do not
follow simple linear laws. Such materials have a critical
temperature, called the Curie point. They must never be
taken above it for it will then destroy the properties as
calibrated.
The Curie point varies from 120 C for the simpler barium titanate forms ranging up to values close to 600 C. For
the interested reader, detailed explanation can be found in
Article 118, Ultrasonic Sensors, Volume 2, Bruel & Kjaer
(1976), Endevco (1980), Harris and Crede (1976), TrampeBroch (1980), and in the detailed information provided by
the makers of PZT materials in application notes, and many
Internet Web sites.
To read the charge of a PZT sensor as it is vibrated
by the subject under test, an electronic amplifier is used
that converts charge magnitude to a voltage equivalent. It
is the nature of this sensing system that it provides no DC
response; when stationary there will be no output other than
system noise.
In practice, the PZT sensors used to measure acceleration
can be operated down to around 0.1 Hz, dependent on
the amplitude to be measured. With a natural resonant
frequency that can be made relatively very high (up to
100 000 Hz in some designs), PZT sensors provide a useful
frequency range that can cover most vibration needs. The
system response, however, relies not only on the sensor
but also upon the cables and the preamplifier used with
the PZT unit see Article 206, Practice of Vibration
Measurement, Volume 3.
To produce the charge, the PZT material can be used
in its pure compression, shear, or bending modes. This
method of sensor design is amenable to the combination
of three units giving the three translation components of
vibration. Figure 4 in Article 206, Practice of Vibration
Measurement, Volume 3 gives some examples of commercially available PZT accelerometers.
The PZT material itself contributes only of the order of
0.03 of the required critical damping factor. Without additional damping being added, PZT transducers should not be
used too close to their resonant frequency see Figure 6
in Article 205, Theory of Vibration Measurement, Volume 3 or an error will arise.

Acceleration Measurement 1411

Input

Charge
amplifier

Accelerometer

Calibration
generator

Integrator
(velocity
&
displacement)

Battery
power
supply

High-pass
filter

Amplifier

External battery charge

Low-pass
filter

Amplifier

Detector
(peakpeak or
RMS)

Lin-log
converter

AC
output

DC lin
output

Meter

DC log
output

External filter

External power supply

Figure 4. Block diagram of vibration measuring system showing functions that may be required. (Reproduced by permission of Bruel
& Kjaer.)

Mounting arrangements within the case also can add

additional damping. Some use an additional spring to precompress the PZT element so that it remains biased in
compression under all working amplitudes: this makes for
more linear operation.
Typical sensor element charge sensitivities range from
0.003 up to 1000 pC m s2 , implying that the following
preamplifier units will also need to vary considerably.

6 AMPLIFIERS FOR PIEZOELECTRIC

SENSORS

7 MICROACCELEROMETERS

An amplifier for reading out the state of the PZT sensor

is one that has very high input impedance, an adequate
frequency response, and low output impedance. Adjustment
of the gain, filtering action and integration time variously
yield velocity and displacement sensing see Article 208,
Amplitude and Velocity Measurement, Volume 3.
Figure 4 is a typical analog conditioning system incorporating most features that might be needed. Some of the
block functions often use digital processing methods, but
the charge amplifier still needs to be analog.
The charge amplifier could be designed to see the sensor
either as a voltage or a charge (current) source. The latter
is preferred for, using modern electronic-feedback operational amplifier techniques, the effect of cable, sensor, and
amplifier capacitances can be made negligible (which in the
voltage-reading method, is not the case). Cable length is,
therefore, of little consequence. This is justified as follows.
Figure 5 is the relevant equivalent circuit for a PZT
accelerometer connected to an operational amplifier (the
preamplifier) via a cable. It includes the dominant capacitances that arise in practice.
It can be shown (see Trampe-Broch (1980), for example)
that the use of feedback in this way with a very high
amplifier gain A gives
e0 =

Sq a
Cf

This shows that the user need only define the sensor charge
sensitivity Sq and the feedback capacitance Cf in order to
be able to relate output voltage from the preamplifier to the
acceleration of the sensor.
If the preamplification shown here is carried out in the
sensor housing, then a special connection cable having
reduced triboelastic effect is not needed from the sensor
to the fixed cable support. Many cases, however, cannot
place the preamplifier in the sensor housing as that adds
mass to the accelerometer, which may not be acceptable.

(2)

Microelectronic forms of the accelerometer made by

the MEMS technology see Article 162, Principles of
MEMS, Volume 3; Article 163, Uses and Benefits of
MEMS, Volume 3; and Article 164, Principles of MEMS
Actuators, Volume 3 can be remarkably small compared
with traditional forms.
For example, one unit reported at the R&D stage by
researchers at Seoul University, Kwang et al. (2003), has
a proof mass of 15 g that is 870 1027 50-m thick.
The accelerometers natural frequency at 14 kHz implies
operation up to 7 kHz or a little more if compensation is
used. The spring elements supporting the lateral capacitive
comb type sensor are 4-m thick and 250-m long. The
sensing gap is 3 m.
Another development unit, Figure 6, shows one of the
many forms of suspension used to support the proof mass,
LLNL (2000).

8 SIGNAL CONDITIONING IN
MICROACCELEROMETERS
The signal processing used in the MEMS forms of accelerometer can be very sophisticated as it allows integration of
the system blocks needed to produce a ready to use robust
output signal. These are

1412

Common Measurands

Feedback
preamplifier

Feedback
capacitance
C1

PZT accelerometer

Input acceleration a
A
Output
Charge q

Ii
Ca

Cc

eo

C1

q = Sq a
Where S q is accelerometer charge sensitivity
a is acceleration
System transfer = e o = S q
a
coefficient
C1

Figure 5. Equivalent circuit for piezoelectric sensor when interrogated, as a charge-generating device, by an operational amplifier
technique.

computer controlled settings,

smart features see Article 160, Smart Sensor System Features, Volume 3.
bias voltages,
wireless link interface.

15 mm

Figure 6. An experimental microaccelerometer sensor.

seismic sensing element,

voltage regulator,
AD converter,
sample and hold,
microprocessor,
switching,
noise reduction,

A good study, readily available to consult on the Internet

is Kulah et al. (2004). Their account shows the intricacies
and level of achievement possible with fully integrated
MEMS units.
Further details of MEM systems are found in
Article 162, Principles of MEMS, Volume 3; Article 163,
Uses and Benefits of MEMS, Volume 3; and Article 164,
Principles of MEMS Actuators, Volume 3.
Signal conditioning is covered in the Section on Data
Acquisition Systems starting at Article 132, Data Acquisition Systems (DAS) in General, Volume 3.

9 MEASUREMENT OF SHOCK
Shock is a sudden short impulse of applied force that can
generate very large acceleration (100 000 g can arise) that is
not recurrent. It can be regarded as a once-only occurrence

Acceleration Measurement 1413

of a vibration waveform, although sometimes it is used to
describe a short burst of oscillation. Examples are objects
subject to explosions and impacts arising because of a
collision.
Understanding the formalized behavior of a given vibration sensor to a shock situation requires Fourier analysis of its response to a truncated wave shape, that is,
a discontinuous wave. The mathematics becomes more
complex.
Theoretical considerations lead to the generalization that
as the waveform becomes more like a single pulse of high
magnitude and very short duration, the frequency band
of the sensor must be widened if the delivered output is
to be a satisfactory replica of the actual input vibration
parameter.
Thus, the accelerometer used to measure shock must
preferably possess sufficient flat bandwidth below its natural frequency point. Special DSP, digital signal processing
methods allow measurement signal extraction beyond this
criterion using advanced algorithms.
Fidelity of output increases as the period of the natural
frequency of the sensor becomes shorter than the pulse
length. An idea of the variation of responses with natural
frequency and damping is available in graphs given in
Harris and Crede (1961). An example can be found in
Figure 3.
Here also the very large force exerted on the transducer
(force = g mass of sensor) requires a total design that
recognizes the need to withstand large transient forces
without altering the mechanical residual strains in the sensor
system during measurement. Mounting circuitry on the
sensor itself adds more room for induced shock error, for
the silicon chips can be shock sensitive.
Well-designed shock sensors can accurately measure
single half-sine-wave pulses as short as 5 s. Some amount
of ringing in the output is usually tolerated in order
to provide measurement of very short duration shocks.
Compensation to increase the bandwidth is also useful in
extending the range of operation.

RELATED ARTICLES
Article 59, Introduction to the Dynamic Regime of Measurement Systems, Volume 1; Article 61, First-order
System Dynamics, Volume 1; Article 62, Second-order
System Dynamics, Volume 1; Article 205, Theory of
Vibration Measurement, Volume 3; Article 206, Practice of Vibration Measurement, Volume 3; Article 208,
Amplitude and Velocity Measurement, Volume 3.

REFERENCES
Bruel & Kjaer. (1976) Piezoelectric Accelerometer and Vibration
Preamplifier Handbook, Bruel & Kjaer, Naerum.
Endevco. (1980) Shock and Vibration Measurement Technology,
Endevco Dynamic Instrument Division, San Juan Capistrano,
CA.
Harris, C.M. and Crede, C.E. (1976) Shock and Vibration Handbook Vol. 1, Basic Theory and Measurements, McGraw-Hill,
New York, (1961, reprinted in 1976).
Kulah, H., Salian, A., Yazdi, N. and Najafi, K. (2004) A 5V
closed-loop second-order sigma-delta, micro-g micro-accelerometer, Center for Wireless Integrated Microsystems, University
of Michigan, http://www.eecs.umich.edu/hkulah/pdfs/HH-02Kulah.pdf.
Kwang, Y., Yeon, J.C., Kim, Y.H., Rhee, S.W. and Oh, S.H.
(2003) High Resolution Silicon Accelerometer Using Eutectic
Bonding, in Nanotech Conference 2003 , abstract http://www.
nsti.org/procs.
LLNL. (2000) MEMS microaccelerometer , Lawrence Livermore
National Laboratory, http://microtechnology.llnl.gov/Devices.
html.
Norton, H.N. (1969) Handbook of Transducers for Electronic
Measuring Systems, Prentice Hall, Englewood Cliffs, NJ.
Oliver, F.J. (1971) Practical Instrumentation Transducers, Pitman, London.
Trampe-Broch, J. (1980) Mechanical Vibration and Shock Measurements, Bruel & Kjaer, Naerum.
Webster, J.G. (1999) The Measurement, Instrumentation and Sensors Handbook, CRC Press, Boca Raton, FL.

208:

Amplitude and Velocity Measurement

Peter H. Sydenham
GSEC Pty Ltd, Adelaide, South Australia, Australia

1 Measurement of Displacement with a Seismic

Form of Sensor
2 Theory of Displacement Measurement Using
a Seismic Sensor
3 Practice of Displacement Measurement Using
a Seismic Sensor
4 Measurement of Velocity
5 Theory of Velocity Measurement
6 Practice of Velocity Measurement
Related Articles
References

1414
1414
1415
1415
1415
1416
1416
1417

1 MEASUREMENT OF DISPLACEMENT
WITH A SEISMIC FORM OF SENSOR
Where a fixed reference method is inconvenient, one of
several forms of seismic sensor system can be employed as
follows.
It is sometimes more convenient to mathematically integrate the signal from a velocity transducer making use of a
fixed reference framework.
Integration is a sound process, for the noise present is
reduced by the integration, averaging process. The integral of an acceleration measurement is velocity and the
second integral is then displacement. Thus, an accelerometer can be used to measure displacement under certain
conditions.

2 THEORY OF DISPLACEMENT
MEASUREMENT USING A SEISMIC
SENSOR
The second-order equations of motion, given in Article 205, Theory of Vibration Measurement, Volume 3,
for a mass that is moving relative to a fixed reference
frame (the mode for studying the movement of vibrating
objects) can be reworked to provide response curves relating the displacement amplitude of the seismic mass to that
of its case.
Figure 1 is the family of response curves showing the
effects of different operating frequency and the degree of
damping. Given that the case is moving in sympathy with
the surface of interest, it can be seen, from the curves,
that for input vibration frequencies that are at least twice
the natural frequency of the seismic sensor, the measured
output displacements will be a true indication (within a
percent or so) of the movements of the surface. This
form of seismic displacement sensor is also often called
a vibrometer.
It is possible to lower the frequency of operation by
using a damping factor with a nominal value of 0.5. This,
however, does introduce more rapidly changing phaseshift error with frequency, which may be important. The
lowest frequency of use above which the response remains
virtually flat is seen to be where the various damping factor
curves approach the horizontal line equal to unity ratio, to
within the allowable signal tolerance.
Given that the chosen damping remains constant and that
a system does follow the second-order response, it is also
possible to provide electronic frequency compensation that
can further lower the useful frequency of operation by a
useful amount.

Handbook of Measuring System Design, edited by Peter H. Sydenham and Richard Thorn.
2005 John Wiley & Sons, Ltd. ISBN: 0-470-02143-8.

Displacement amplitude of seismic mass relative to case

Displacement amplitude of case

Amplitude and Velocity Measurement 1415

to make use of a direct-reading displacement seismic
sensor design.

2.0

0.3
0.4
0.5

1.0

0.7

4 MEASUREMENT OF VELOCITY
1
2

0.5

5
0.2

0.1
0.3

Damping ratio
z

10

0.5

0.81.0
Natural
frequency

10

w
wn

Figure 1. Responses relevant to displacement and velocity sensing with seismic sensors.

Thus, to directly measure displacement amplitudes with

a seismic sensor, it must have a natural frequency set
to be below the lowest frequency of interest in the subjects vibration spectrum. In this mode, the seismic mass
virtually remains stationary in space acting as a fixed
reference point. It is also clear that the seismic sensor cannot measure very low frequencies, for it is not
possible to construct an economical system having a
low enough resonant frequency, as that needs massive
masses and very springy suspensions; these are the features of the large strong motion seismometers that measure ground and building motion and displacement during
earthquakes.

3 PRACTICE OF DISPLACEMENT
MEASUREMENT USING A SEISMIC
SENSOR
The curves are theoretical perfections and would apparently indicate that the seismic sensor, in this case, will
have flat response to infinite frequency. This is not
the case in practice, for, as the frequency of vibration rises, the seismic sensor structure begins to resonate
at other frequencies caused by such mechanisms as the
spring vibrating in modes other than the fundamental of
the system.
Given that the low frequency range of accelerometers
can extend down to less than 1 Hz (see Article 207,
Acceleration Measurement, Volume 3), it may often be
more practical to twice integrate an accelerometer signal,
in order to derive displacement amplitude, rather than

This section covers simple traditional forms of velocity

sensor used to measure the short motions of solid objects
that will possibly also need acceleration determination.
The many other forms for measuring velocity (i.e. flow
speeds) of gases, liquids, solids, and mixed-phase flows are
covered in the following articles:
Laser Doppler velocimeters and other flowmeters
Article 187, Flowmeter Selection and Application,
Volume 3
Angular and displacement devices in general Article 191, Displacement and Angle Sensors Performance and Selection, Volume 3; Article 192, Strain
Sensors, Volume 3; Article 193, Specialty Displacement and Angle Sensors, Volume 3.
Other methods, such as digital encoders, Hall-effect
toothed wheel sensing, tuning forks, magnetohydrodynamics, and optical correlation are variously covered in Webster
(1999), Norton (1969), and Dyer (2001).

5 THEORY OF VELOCITY
MEASUREMENT
The prime method used to generate a direct velocity signal
makes use of the law of electromagnetic induction. This
gives the electrical voltage e generated as N turns of an
electric coil to cut magnetic flux over time t as
e = N

d
dt

(1)

Velocity sensors are thus self-generating; they produce a

voltage output that is proportional to the velocity at which
a set of turns moves through a constant and uniform
magnetic field.
Many forms of this kind of sensor exist. The commonly
used, moving-coil arrangement comprises a cylindrical coil
vibrating inside a magnetic field that is produced by a
permanent electromagnet. A commonly seen arrangement,
Figure 2(a), is that typified by the reversible-role, movingcoil, loudspeaker movement. For this form, the output
voltage Vout is given by
Vout = Blv 109

(2)

1416

Common Measurands

Cylindrical
airgap

Generated
electrical
output

Moving
coil in magnetic
field

Velocity
input
(a)

Fixed coaxial
permanent
magnet
structure

Variable airgap modulates magnetic circuit

S
Ferromagnetic
material

Electrical
output

(b)

Magnetic
circuit

Vibrating input

Electrical
output

Fixed
coil
S

(c)

Magnet moving in coil

in vibration direction

Figure 2. Forms of velocity sensor. (a) Moving coil, (b) variable

reluctance, (c) moving permanent magnet.

where B is the flux density in Tesla, l the effective

length of conductor contributing in total flux-cutting, and
v the instantaneous velocity, expressed in ms1 , of the
coil relative to the magnet. Given that the design ensures
that the magnetic field is uniform in the path of the
fixed conductor length, the sensor can provide very linear
output.

6 PRACTICE OF VELOCITY
MEASUREMENT
These were the sensors adopted in early seismology studies
because of their inherently high output at relatively low
velocities. The coil impedance will generally be low,

enabling signals with good signal-to-noise ratios to be

generated along with reduction of error caused by variations
in lead length and type. They are, however, large with
resultant mass and rigidity problems. Many of the forms
tend to have relatively low resonant frequencies (tens of
hertz), which restricts use to the lower frequencies. It will
be apparent that these sensors cannot produce signals at zero
velocity because no relative movement occurs to generate
flux-cutting.
A second variation of the self-generating velocity sensor,
the variable-reluctance method, uses a series magnetic
circuit containing a permanent magnet to provide permanent
magnetic bias. A part of this circuit, the armature, is made
so that the effective air gap is varied by the motion to be
monitored. Around the magnetic circuit is placed a pickup
coil. As the armature moves, the resulting flux variation
cuts the coil, generating a signal that can be tailored by
appropriate design to be linear with vibration amplitude.
This form of design has the advantage that the armature
can readily be made a part of the structure to be monitored,
as shown in Figure 2(b). This version is not particularly
sensitive, for the air gap must be at least as large as the
vibration amplitude. As an example, a unit of around 12mm diameter, when used with a high magnetic permeability
moving disc set at a 2-mm distance, will produce an output
of around 150 mV ms1 .
A third method uses a permanent magnet as the mass
supported on springs. One example is shown in Figure 2(c).
Vibration causes the magnet to move relatively to the fixed
coil, thereby generating a velocity signal. This form can
produce high outputs, one make having a sensitivity of
around 5 V ms1 .
Where a fixed reference cannot be used, this form of
sensor, instead of a displacement sensor, can be built
into the seismic sensor arrangement. In such cases, the
vibrating seismic sensor will then directly produce velocity
signals. These will follow the general responses given in
Figure 1. From those curves, it can be seen that there is a
reasonably flat response above the natural frequency, which
is inherently quite low.

RELATED ARTICLES
Article 187, Flowmeter Selection and Application, Volume 3; Article 191, Displacement and Angle Sensors
Performance and Selection, Volume 3; Article 192,
Strain Sensors, Volume 3; Article 193, Specialty Displacement and Angle Sensors, Volume 3; Article 205,
Theory of Vibration Measurement, Volume 3; Article 207, Acceleration Measurement, Volume 3.

Amplitude and Velocity Measurement 1417

REFERENCES

Norton, H.N. (1969) Handbook of Transducers for Electronic

Measuring Systems, Prentice Hall, Englewood Cliffs, NJ.

Dyer, S.A. (2001) Survey of Instrumentation and Measurement,

Wiley, New York.

Webster, J.G. (1999) The Measurement, Instrumentation and Sensors Handbook, CRC Press, Boca Raton, FL.