Vibration School

VIBRATION SCHOOL
What Is Vibration ?
Vibration is a "back and forth" movement of a structure. It can also be

referred to as a "cyclical" movement
What Is Vibration Caused By ?

Imperfections in the Machine:
Design Assembly
Manufacture Operation
Installation Maintenance
What Are Some Common Machine Problems
That Generate Mechanical Vibration:
Misalignment Unbalance
Worn belts & pulleys Bearing Defects
Hydraulic Forces Aerodynamic Forces
Reaction Forces Reciprocating Forces
Bent Shafts Rubbing
Gear Problems Housing Distortion
Certain Electrical Problems Frictional Forces
What Are Some Common Machine Problems
That Amplify Mechanical Vibration (But Don't Cause It):
Resonance Looseness
Why Do We Measure Vibration ?

1. Assess the condition (primarily the bearings) of a machine. By
performing this task effectively, we can eliminate catastrophic failures
due to component degradation.
2. Diagnose the root cause(s) of any excessive (destructive) vibration.
By performing this task effectively, we can extend the life of bearings
and other components that are absorbing the stresses and fatiguing
forces that are causing the symptom of excessive vibration.
It should be understood that short-term event-based failures (i.e. loss
of lubrication, sudden fracture of a component, etc.) are not protected
against via any program that only collects data periodically. The time
between onset and failure in these cases - which are rare - may take
only minutes (in extreme cases), hours, days or weeks. For example,
many programs are based on monthly data collection. Any event
occurring during that month interval may cause failure prior to the
next data collection. This is not a failure of the program or the
technology any more than driving a fork truck into a machine and
destroying it is. The good news is that the vast majority of potential
and actual failures will NOT fall into this category and DO lend
themselves to being detected, monitored and corrected through a well-
run vibration program.
What Does The Transducer Actually Detect ?
Actual Bearing A Transducer Mounted A Transducer Mounted

Movement: Vertically "Sees" Only Horizontally "Sees" Only
Elliptical Vertical Movement Horizontal Movement
What Vibration "Characteristics" Do We

Measure ?
Amplitude Tells Us:
How Much Movement Occurs

Measure ?
Frequency Tells Us:
How Often The Movement Occurs

(How many "cycles" in a period of time: a second or a minute)

Measure ?
Phase Tells Us:
In What Direction Is The

Movement
(Relative To Other Locations On The Machine At A Given
Moment In Time)
What Are "Conventions" ?

Conventions are standards that you set or adopt that apply to
every machine and application in the program. These
conventions simplify training of new personnel and make sure
everyone involved in the program is on the same page. There
are at least three basic conventions that should be set up.
They are:
1. Bearing Numbering ("Positions")
2. Position Naming
3. Direction Definitions
Common Bearing Numbering Conventions

 Most programs use numbers to identify bearings. Some
use letters instead. By FAR the most common convention
is to begin the numbering (or lettering) on the non-drive
end of the "driver" component (motor, turbine, etc.).
That bearing will be denoted as Position "1", or Position
"A".
 Following the drive train in a logical manner, the next
bearing reached will be Position "2" or "B". That bearing
will be found at the drive end of the motor or turbine (the
driver component).
 Continuing to logically follow the drive train, the next
bearing reached will be Position "3" or "C". This bearing
will almost certainly be on the "driven" component.
 With a typical 4-bearing machine, the last bearing
reached would be Position "4" or "D".
The image below shows a typical 4-bearing, belt driven fan

with the bearings numbered.
Common Bearing Numbering Conventions
What about components that have more than 2 bearings

such as gearboxes. The image at right (which does NOT
show the motor containing bearings 1 & 2) shows a typical
numbering convention for components with more than 2
bearings. Again, it is based on following the drive train in a
logical manner. The important thing is to set up a simple,
straight forward convention and adhere to it consistently.
You can use other people's ideas and conventions or develop
one yourself that makes sense to your people for your
machines.
The gearbox shown here is a "double reduction" gearbox (it
has two separate gearmeshes). This gearbox has three (3)
shaft speeds (the input shaft speed, the speed of the shaft
supported by bearings 5&6 and the output shaft speed) and
two (2) gear mesh frequencies.
Common Position Naming Conventions

Although bearing numbering is (and should be) the primary
identifier of the position of the reading, bearing names are
often used as well. Since there are a variety of common names
used for naming the different bearing positions - several for
each position, in fact - a list of the common ones is put forth
here. There are no right or wrong ones - the only important
aspect is complete consistency in your usage.
 Common Position 1 Names
o Outboard motor bearing
o Opposite drive end motor bearing
o Opposite shaft end motor bearing
o Inboard motor bearing
o Drive end motor bearing
o Shaft end motor bearing
o Inboard ?? bearing (the "??" will be fan, pump, etc.
depending on what the driven component is)
o Drive end ?? bearing
o Shaft end ?? bearing
o Outboard ?? bearing
o Opposite drive end ?? bearing
o Opposite shaft end ?? bearing
You may have applications that do not fall neatly into the 4-
bearing machine category. Long drive lines with dozens of
bearings, gearboxes like the one shown on the previous page,
multi-stage machines, etc. may each require their own unique
solution for bearing naming. In the case of a long drive line,
the bearing may be named to coincide with the piece of
equipment along that drive line that bearing is closest to.
Conversely, you may decide to rely strictly on position
numbers in that case and not use position names at all. Terms
such as "Intermediate Shaft" may be used. There is no single,
universal naming convention that will apply to all machine
configurations. Remember the objective:
Consistency Is The Key
Common Direction Naming Conventions

Directional conventions are also of the utmost importance to
set up and use. A simple, common sense convention insures
that whomever is collecting the data is aware of the correct
transducer location and direction. It also means the analyst, if
different than the data collection technician, can analyze the
data with confidence. This convention goes to the heart of one
of the most important aspects of a vibration program - the
repeatability of readings from one data collection to the next
(what good is a trend without good repeatability ?). Its
importance goes hand-in-hand with the importance of making
sure the exact transducer locations are clearly marked. The
convention begins with ONE hard rule:
Axial Direction is always, Always, ALWAYS parallel to shaft

axis
Let's start with horizontal, direct drive machines. These

machines are the most simple to define.
1. Axial Direction - Runs along the axis of the machine's shaft
(parallel to the shaft & ground).
2. Vertical Direction - The shortest line possible connecting the
machine shaft and the machine base.
3. Horizontal Direction - A line drawn out from the machine
shaft that runs exactly parallel to the ground.
Belt drives can be oriented in any direction and thereby require

a directional convention. The convention shown here has been
adopted for some very simple reasons which help illustrate not
only its usefulness but the importance of conventions in
general:
 Reason #1: The belt (reaction) forces are usually directed
in-line (parallel) with the belts.
 Reason #2:In order to capture these belt related
vibrations and separate closely matched frequencies,
higher resolution readings are set up in the database
parallel to the belts. Details of this subject - spectrum
resolution - are covered in the "Spectrum" section.
 Reason #3: To make it easier to both set up the database
and to analyze collected data, these higher resolution
readings are always taken horizontally.
For these reasons, a convention is used that defines "parallel

to the belts" as horizontal is used. By default, that leaves the
measurement taken perpendicular to the belts defined as
Vertical. By adopting this convention, both collection and
analysis are simplified - a stated objective of using
conventions.
Vertical machines present another opportunity to assign a

directional convention since parallel to the shaft (axial) is now
straight up out of the ground. Since we must adhere to our
one hard and fast rule for directional conventions, the axial
direction remains parallel to the shaft (perpendicular to the
ground; what would be defined as vertical on a typical,
horizontal direct-drive machines).
That leaves vertical and horizontal to be defined. For reasons
similar to those discussed previously for the belt drive
convention, it is recommended that horizontal be defined as
parallel to the discharge of the machine. That would leave
vertical as being defined as perpendicular to the discharge
(and parallel to the ground).
Vibration Characteristics:
Amplitude
What Does Amplitude Tell Us ?
Before we answer that question, let's keep in mind what

exactly we are measuring. Everyone taking this class has
touched a part of an operating machine (even if it is your car
or even your lawn mower) and felt the back and forth
movement. Because that movement is back and forth, it is
defined as "cyclical", or "sinusoidal", and we call it "vibration".
Obviously, we would want to quantify the amount of
movement. That measurement is known as "amplitude".
However, there are several different ways of quantifying the
amount and that is what we will discuss in the following
pages.
What Does Amplitude Tell Us ?
 The 'amplitude' is a measure of the amount of

movement.
 The amount of movement is related to the severity of the
vibration.
Simply put, it measures:
How Much ?
What Are Amplitude "Units" ?

There are several different ways to measure "how much".
These different ways are known as amplitude "units". The
transducer is the mechanism we use to measure vibration
and, in the case of rolling element bearings (i.e. ball bearings,
etc.), we can assume the transducer, being affixed to the
bearing housing, is going to move very close to the same
amount as the shaft itself since a rolling element bearing has
very small internal clearances. The first amplitude unit is the
simplest to visualize and understand:
 Displacement - measures the total distance the
transducer (bearing) travels back and forth during one
'cycle' of movement (a 'cycle' is the process of moving
from one extreme to the other and back again to the
starting point).
To understand the second amplitude unit, you must first

understand that if a bearing is going to move back and forth a
certain distance in a certain amount of time (the 'period'), it is
going to achieve a certain maximum, or 'peak', speed
(velocity) during that cycle. That speed is constantly changing
as it goes from 0 (when the bearing is displaced a maximum
amount in one direction and has stopped momentarily to
reverse direction) to some maximum value it achieves as it
passes the center point in the movement. Once it passes that
center point, the bearing starts slowing down until the speed
again reaches 0 as it reaches the maximum displacement in
the opposite direction.
 Velocity - measures the maximum speed the transducer
achieves during a cycle.
To understand the third amplitude unit, you must understand

that to change velocity, a body must be accelerated or
decelerated. To speed a car up, you press the "accelrator"
pedal. To slow it down, you remove your foot from the
accelerator and let frictional forces (wind resistance, road
surface, brakes) take over. On machines, this "acceleration
force" can be visualized as the reaction of the bearing housing
and surrounding structure to being pushed (displaced) in one
direction - the housing pushes back because it wants to
assume a neutral, or "at-rest", position.
 Acceleration - measures the force(s) that are causing the
back and forth movement.
Now let's look at each of these units more in-depth and see
how they are inter-related.
Measuring "How Much" in Displacement Units
Displacement measures the length of the "trip" back and forth
from (in this case) +X to -X (2X would be the total distance
travelled - the "peak-to-peak displacement"). One of these
"trips" is known as a "cycle" of vibration. The sequence of
images about to be presented show the bearing at various
important points during a single cycle with the transducer
oriented vertically (remember, the transducer only "sees"
movement in the direction of its orientation, or axis).
Since this movement must occur over time, when we measure
it we plot the amount (amplitude) on the y-axis and the time
taken (period) on the x-axis. The resulting shape, in its
simplest form, is called a "periodic signal", a "sinusoid" or a
"sine wave". That is the S-shape you see below the word
"cycle" in the image at right. Mechanical vibration generates a
wave shape that is rarely as simple as what we see here but
the intricacies of processing more complex data will be
covered in later sections.
 The "At-Rest" position ("0") is the position the bearing
would assume if the machine was not running.
 During a single (1) cycle, the bearing passes this position
twice - once travelling towards +X and once travelling
towards -X.
 We'll begin our sequence of images with the bearing just
passing the '0' point moving towards '+X'.
 The red ball (seen at the very ends of the sine wave)
indicates the amplitude level during the cycle.
 The bearing has just reached the '+X' position and has
stopped momentarily to reverse direction.
 The bearing is 1/4 of the way through the cycle (1/4 of a

shaft revolution).
 The bearing is now passing the '0' position again moving

towards '-X'.
 The red ball has completed 1/2 cycle (1/2 shaft

revolution).
 The bearing has now reached the '-X' position - its maximum
displacement in the '-' direction.
 The bearing has again stopped to reverse direction.
The red ball has completed 3/4 of a cycle (3/4 of a shaft

revolution).

 The bearing is now back where it started, having

completed one "cycle" of movement.
 We want to know the total length of the "trip" being
made by the bearing.
 Travelling from the "+X" position to the "-X" means the
total distance travelled = 2X). That is known as Peak-to-
Peak signal detection and that is how displacement
amplitudes should be measured.
 Since displacement measures the amount a component is
being bent back and forth, it is a measure of the STRESS
that the bearing structure is being subjected to. It is, in
other words, an amplitude unit that is particularly
sensitive to the likelihood of a stress failure occurring.
 Stress failures are most likely to occur on slow speed

equipment and are not a very common failure mode
(hence we do not normally use displacement amplitudes
as a primary monitoring tool).
The "Displacement Sine Wave"
English or 'Imperial' Units: Mils (1 mil = 0.001")

Metric Units::Microns (1 um = 0.001 mm)
Conversion::1 Mil = 25.4 um
Measuring "How Much" In Velocity Units

 What is a "velocity" amplitude ? Technically, velocity
measures how much the displacement is changing over a
period of time.
 The units of measure are inches per second or millimeters
per second.
 With sinusoidal motion the velocity constantly changes as
the displacement changes.
 Since the velocity is constantly changing over time,

measuring velocity amplitude over time generates a sine
wave just as measuring displacement amplitude does.

The "Velocity Sinusoid" vs. The "Displacement Sinusoid"
As we began discussing the relationship between velocity and
displacement on previous pages, let's now look at a direct
comparison between the two and see how they relate to one
another at the most significant points in the cycle:
 The bearing is shown here at its maximum displacement
of '-X'.
 Velocity must be zero at this moment since the bearing
must stop momentarily to reverse direction.
 From this moment until the bearing passes the "at-rest"

(0) position, the bearing will be speeding up (the velocity
amplitude will be increasing).
The bearing is now passing the "at-rest" position as it moves

towards the "+X" position:
 The displacement amplitude is '0'.
 Velocity is at a '+' peak since the bearing has reached its
maximum speed (remember, it has been accelerating
since leaving the "-X" position.
 From this moment until the bearing reaches the "+X"

position, the bearing will be slowing down (velocity
amplitude will be decreasing).

The bearing has reached its maximum displacement in the '+'
direction:
 The displacement is "+X".
 The speed (velocity) is 0 since it has (again) momentarily
stopped to reverse direction.
 From this moment until the bearing passes the 'at-rest'

position, the bearing will be speeding up (accelerating -
the velocity amplitude will be increasing).

The bearing is passing the "at-rest" position again moving

moving in the "-" direction:
 The displacement amplitude is "0".
 Velocity is at a "-" peak since the bearing has reached its
maximum speed (the speed is the same as previously
reached when travelling in the "+" direction - only the
direction has changed).
 From this moment until the "-X" position is reached, the bearing will be slowing
down ("decelerating" - the velocity amplitude w

 With displacement, we were concerned with the total

distance travelled (stress-related failures).
 With velocity, we are only concerned with the
maximum speed (velocity amplitude) reached during
that trip. It matters not which direction the bearing is
moving in.
 Since we are concerned only with the maximum
velocity reached, we use Peak signal detection (not
Peak-to-Peak).
 Velocity measures the how often (frequency) the stress
(displacement) is being applied.
 Velocity is measure of the likelihood of FATIGUE being
the mode of failure.
 Fatigue failures are by far and away the most common
cause of general machinery failures.
 Velocity is the best monitoring tool for general
machines.
 Speciality machines, components or specific problems
may not be best monitored by velocity.
 Examples of machines, components and problems not
suited to velocity are gears and certain electrical
symtoms (i.e. very high frequency vibration: >120,000
cycles per minute) and very slow-speed equipment (<
100 rpm).
 It should be noted that even though velocity is suited

to monitor problems in the 60kcpm - 150kcpm range, it
may be advantageous to use our third amplitude unit -
acceleration - in those cases.
 ill be decreasing).
The "Velocity Sine Wave"
English or "Imperial": Inches per Second (ips -or- in/sec)

Metric: Millimeters per Second (mm/sec)
Conversion::1 ips = 25.4 mm/sec
Measuring "How Much" In Acceleration Units
What exactly is "acceleration" ? Acceleration measures the rate
of change of velocity (how quickly the velocity is changing).
There are two methods of applying an acceleration force
(deceleration is simply a negative acceleration):
1. A Pushing Action similar to you compressing a spring
between your hands. The more you compress a spring, the
more force it pushes back with. If you were to push a pillow
block bearing away from its "at-rest" position, it would push
back. If you went farther, it would push back more. That is a
simple way to visualize the acceleration force we are
measuring. The amount of movement at a particular frequency
is a combination of the force being generated by the rotation
of the rotor (unbalance, for intsance, is simply a centrifugal
force due to a heavy spot on a rotor), the stiffness and masses
of the materials and structure involved and other variables that
are structure related.
2. A Striking Action similar to you hitting a nail with a hammer.
This action can be extremely destructive since it can cause
structural flaws (cracks, for instance) to develop. Consider a
jack hammer. It is the striking action that breaks up the
cement. One example In the case of a rolling element bearing,
a rolling element may pass a defect on one of the races and an
impact results similar to hitting a pothole with your car.
What Is The Difference Between

Pushing Forces And Striking Forces ?
1. Both are destructive but one is far more destructive - the
striking action.
2. The Pushing Action causes sinusoidal motion. Since the
velocity changes steadily (creates a sine wave), the
acceleration also changes steadily and plotting it generates a
sine wave just as displacement and velocity do. This is the
type of acceleration we will be discussing in the following
pages.
3. The Striking Action causes instantaneous, transient motion.
Consider striking something with a hammer. The velocity
undergoes a nearly instantaneous increase when the hammer
strikes. Any movement then dampens out until the next
impact. This type of acceleration must be measured differently
and will be covered in a later section on "enveloping signals".

The "Acceleration Sinusoid" vs. The "Displacement Sinusoid"
Let's see how displacement and acceleration relate to one

another at the significant points in the cycle:
 The bearing is shown here at its maximum displacement
of "-X". If you were pushing the bearing housing down to
this position, in which direction would the bearing
housing be pushing back ?
 The housing structure in this position will be pushing back
in the "+" direction since it is trying to return the bearing
to the "at-rest" position.
 As mentioned before, the "amplitude" we measure is a
combination of many variables but what do we actually
measure ?
 The bearing is achieving a certain peak velocity once per
cycle. That velocity is a combination of the amount of
movement (displacement) and the time it takes for one
complete cycle (from which we calculate frequency). The
less time a cycle takes, the higher the frequency of the
vibration and the more force it requires to generate a
particular peak velocity. In other words, going from 0
velocity to 1 in/sec (25 mm/sec) 1000 times a minute
requires a lot less force than doing it 100,000 times per
minute. The forces being applied to make that happen
may destroy a component before metal fatigue (what
velocity is sensitive to) even becomes a factor.
 That makes acceleration an amplitude unit that is
particularly sensitive to the likelihood of a component
failing due to the forces being applied to it due to the
machine's operation.
 When either displacement peak is reached, an
acceleration peak is reached in the opposite direction.
 From the moment shown until the '0' position is reached,
the acceleration amplitude decreasing.

The bearing is passing the "at-rest" position moving in the

"+" direction.
 At the instant the bearing passes "0", the acceleration
force (amplitude) is 0 since the bearing is located in its
at-rest position.
 From this moment until the "+X" position is reached

the bearing acceleration amplitude is increasing to a
peak value in the "-" direction (remember, as it is
being displaced increasingly in the "+" direction, it is
increasingly pushing back towards the at-rest position).
The bearing has reached the "+X" position (the "+"

displacement peak):
 The acceleration force (amplitude) is at a maximum
(peak) pushing DOWN towards the "0" position (it has
reached its maximum value in the "-" direction).
 From this moment until the "0" position is reached the

bearing acceleration amplitude is decreasing as the
bearing approaches its at-rest position.
Measuring "How Much" In
Acceleration Units
The bearing is passing the at-rest position moving in the

"-" direction.
 At the moment the bearing passes the at-rest
position, the acceleration force (amplitude) is 0.
 From this moment until the "-X" position is reached

the bearing acceleration amplitude is increasing to a
peak value in the "+" direction (remember, as it is
being displaced increasingly in the "-" direction, it is
increasingly pushing back towards the at-rest
position).
Measuring "How Much" In
Acceleration Units
 As with velocity, we were concerned only with the

maximum value reached - not the direction.
 As with velocity, we use Peak signal detection.
 Acceleration measures how rapidly the velocity is
changing.
 Acceleration is measure of the likelihood of APPLIED
FORCE being the mode of failure.
 Applied force failures occur at higher frequencies -
almost invariably 60,000 cycles per minute and higher.
There are a limited number of high frequency
generating machinery problems.
 Those problems include primarily rolling element

bearing defects and gears.
The "Acceleration Sine

Wave"
English or Metric - G's (1 g = force of gravity)
Frequency
What Does Frequency Tell Us ?

 By taking the reciprocal of how many seconds a single
cycle takes (reciprocal of "seconds per cycle" is "cycles
per second"), the number of cycles occurring in a given
period of time such as a second or minute can be
calculated.
 For example, if a cycle takes 1/50th of a second, the
frequency is 50 cycles per second (50 "Hertz"), or 3000
cycles per minute (3000 cpm or 3kcpm).
 That number - how many cycles occur in a given time

period - is known as the vibration "frequency".
Simply put, it measures:
How Often ?
What Does Frequency Tell Us ?

 Machines will generate mechanical vibration at multiples
(harmonics) of their running speeds. This type of
vibration is called "synchronous" vibration.
 For example, unbalance causes a force that moves the
bearing (causes vibration) in any direction (plane) at a
rate of once per revolution (1x RPM). That movement
occurs at exactly once per rev - not 1.1x per rev, not
0.9x per rev - ONCE per revolution.
 A pump with 5 vanes on the impeller can generate
hydraulic pulses (which can be measured as mechanical
vibration) at 5 times per rev (5x rpm) - not 4.9x per rev,
not 5.1x per rev - exactly 5 times per revolution.
 Different mechanical problems (unbalance,
misalignment, etc.) tend to generate their own
characteristic vibration 'patterns'. Because the effect
each problem has on the vibration signal we measure,
they each tend to generate vibration at specific (rpm
related) frequencies that the analyst learns to recognize
and look for.
 Other vibration generators may not be tied specifically to
the machine's rotational speed.
 Bearing problems and electrical problems, for example,
tend to generate vibrations at specific frequencies other
than exact multiples (harmonics) of running speed. This
type of vibration is referred to as "non-synchronous" or
"sub-synchronous" (below 1x rpm) vibration.
 By correctly linking the frequency to the various possible

sources, the source of the problem can be identified.
Frequency Identifies The Vibration Source
How is Frequency Calculated ?
Measure the time it takes for 1 cycle:
(Seconds / Cycle)
Take the reciprocal of that to get the frequency:
Cycles / Second (Hertz)
Hertz x 60 = Cycles / Minute (CPM)
Since cpm is easier to relate to RPM, it is more commonly used
and recommended for ease of use.
How Does Frequency Relate to Amplitude ?

The frequency of any periodic signal is mathematically related
to each of the amplitude units: displacement, velocity and
acceleration - if you know any two of these variables, you can
mathematically calculate the other two. For instance, if you
know:
1. How far a bearing is moving back and forth (the
displacement amplitude), and
2. How much time it takes to complete the trip (the 'cycle',
from which the frequency is derived)
Then armed with the proper mathematical formula, you can
calculate the peak velocity reached during that trip. For
instance:
 A bearing vibrates 10 mils (254 microns) in 0.1 seconds.
We know two of the variables:
o The pk-pk displacement is 10 mils.
o The period is 0.1 seconds (time required for 1

cycle).
Frequency is represented as the number of cycles during a
certain period - a second or a minute. The bearing will make
10 of these trips in one second. Since the total distance
travelled in one second is 100 mils, your average speed is 100
mils/sec (0.1 inches per second or 2.5 mm/sec). Of course,
that is the average speed. Since you are constantly speeding
up or slowing down, your peak speed would be about twice
that average speed (0.2 in/sec or 5 mm/sec).
Let's look more closely at the relationship between these 4
variables.
How Do Displacement, Velocity & Frequency
Relate ?
As stated, there are some simple mathematical formulas that
relate the frequency of the vibration, the amount of movement
(displacement), the speed of movement (velocity) and the
force generated during the movement (acceleration). The
mathematics involved is always handled by the software and
hardware but it is illustrative to understand the simple
principles involved. Let's first look at the relationship between
frequency, displacement amplitude and velocity amplitude.
Consider:
 A bearing moves back and forth 10 mils (pk-pk
displacement).
 The bearing moves at 10 cycles per minute (the
'frequency).
 By setting those two variables, you establish a third -

namely the speed at which the bearing must travel to
satisfy those two conditions.
Consider:
 Another bearing is moving at 10 mils pk-pk.
 This bearing moves at 70 cycles per minute.
What is the speed of this bearing vs. the first bearing to

satisfy those conditions ?
 This bearing must have a peak speed of 7x the first
bearing. The relationship between the 3 is linear (e.g. if
the frequency increases 7x and the displacement remains
same, the velocity must increase 7x).
So increasing the frequency 7x while leaving the displacement

the same results in increasing the velocity at which the
bearing must move by 7x.
The point here is simple. There is a direct relationship between
the frequency, the displacement and the velocity. Knowing two
- any two - allows the third to be mathematically calculated
(along with a constant value). Without getting into further
examples, the same direct relationship exists between
frequency, velocity and acceleration. Therefore, all four of
these variables are related - knowing any two allows the other
two to be calculated.
Let's look at the practical aspect of the relationship in a more
graphic way.
Practical Effect On A Bearing Of Velocity Vs.

Displacement
These animations graphically illustrate the previous example.

The displacement in each is the same but the bearing on the
right is completing 7 cycles for each cycle completed by the
bearing on the left.
But what we are really interested in, of course, is the effect of
the vibration on the bearing's life and the machine's health.
Again, knowing nothing about vibration analysis and using
only your common sense and knowledge of machines, which
bearing will fail in a shorter period of time ? It doesn't take a
vibration 'expert' to recognize that it will be the bearing on the
right. But since the displacement (a measure of stress) is
constant, the determining factor must be something else.
It is actually two failure modes that are increasing in likelihood
with the frequency:
 The fatiguing effect on the bearing components.
 The forces being applied to make the bearing move.

But since the units are all related, why don't we just use a
single amplitude unit and simplify things ?
How Amplitude Units Relate to Different Failure

Modes
The reason has to do with each amplitude unit's sensitivity to
different modes of machinery failures. In other words, each
unit has a specific usefulness in monitoring machinery health.
There are three types of failure causing effects that we are
monitoring with vibration:
 Stress (bending a component excessively causes it to
fail)
 Fatigue (something simply wears out over time)
 Force (the 'pushing' and/or 'striking' actions being
applied to cause the movement)
The graph below shows the sensitivity of each amplitude unit

to the likelihood of a failure over a wide range of frequencies.
 Notice that at low

frequencies
(primarily below 300
cpm or 5 Hz),
displacement is the
most sensitive unit to
the likelihood of a
failure. That is due to
the fact that a stress
failure (something
being bent back and
forth until it breaks)
is the most likely
failure mode at those
low frequencies - the
fatigue and applied
forces become, as
frequency
approaches 0, simply
too low to cause a
failure.
 Once you get above

300 cpm, the most
likely failure mode
increasingly becomes
the 'fatigue' mode
(to which velocity is
the most sensitive
unit).
Fatigue failures basically occur when a component simply
wears out - it tires of the repeated back and forth movement
(even a relatively small total distance) over an extended
period of time and many, many cycles. Between about 300
cpm and 120,000 cpm (5 - 2000 Hz), fatigue is the most likely
failure mode and therefore velocity is the most effective and
reliable amplitude unit to monitor with.
 Once you reach 120,000 cpm (2000 Hz), the most likely
failure mode is 'force'-related. What is a force-related
failure ? When you reach these very, very high
frequencies (remember, you are dealing with moving an
entire rotor structure back and forth 2000 times per
second or more), you are dealing with massive amounts
of force to move that structure back and forth even a
tiny distance at such a tremendously high frequency.
Therefore, it is that tremendous pushing or striking
action that causes the failure.
It must be understood that there are areas of the chart where
the units overlap and two conditions (stress and fatigue
effects, for instance) exist.
Stress Failures vs. Fatigue Failures

The chart at right shows the
number of failures vs. the
number of running hours.
Notice that a relatively high
number of failures occur
during the first hours of
runtime. These failures are
known as 'infant mortality'
because they occur shortly
after start-up. In other
words, a machine that is
new or rebuilt is started up
and has severe problems.
Within a few hours, days or
possibly weeks, a
catastrophic failure occurs.
If the failure is mechanical
in nature (it could also be
electrical or lubrication
related), stress will often be
a primary cause of failure -
components being bent
back and forth so much that
something simply breaks.
However, once a machine runs for a certain number of hours
(rotations), it becomes stress relieved and the likelihood of
failure changes to fatigue - a component wearing out. Of
course, if the movement (vibration) is high but not quite high
enough to cause an 'infant mortality' stress failure, the fatigue
failure will still occur in a relatively short period of time (which
is one reason why the number of failures on the curve doesn't
ever quite get to zero).
So if displacement is sensitive to stress, and velocity is
sensitive to fatigue, where do the acceleration amplitude units
fit in ?
What Failure Mode Are Acceleration Units

Sensitive To ?
Acceleration amplitude is the trickiest to understand. To begin
with, you must understand that due to the nature of sinusoidal
motion (the back and forth action), the velocity is constantly
changing. It goes from zero to a peak back to 0 back to the
peak and so on. To change the velocity of something,
acceleration must be applied. To speed your car up, you apply
the accelerator. To slow your car down, you apply the brake.
 Acceleration measures the rate of change of velocity.
 Velocity is changed when a PUSHING or STRIKING action
is applied.
 Pushing or striking something is applying a 'force' and
acceleration is, of course, force.
So why is acceleration used in the high frequency range ?

 The rate of change in velocity (acceleration) is more
affected by frequency - how often something is changing
direction - than displacement - how far it is moving.
 Components moving at high frequencies will never fail
due to stress (displacement) because the displacement
amplitude is very small.
 Although there are frequencies where velocity and
acceleration overlap in their sensitivity to failures, the
higher the frequency involved (especially above about
120,000 cpm), the less likely a fatigue failure is and the
more likely it is that the forces being applied that are
causing the movement will be responsible for any failure
that might occur.
 Acceleration is sensitive to the likelihood of a FORCE

related failure. In other words, a failure due to the
pushing and/or striking action the component is being
subjected to.
Let's look at one simple example to illustrate where
acceleration is of use.
Where Are Acceleration Amplitudes Useful ?

Consider a high speed centrifugal compressor. This machine,
through its normal operation, generates some incredibly high
vibration frequencies - well over 1,000,000 cpm (16.67kHz) in
certain cases depending on the specifics of the machine
components. These vibrations are generated by the gear teeth
meshing together and referred to as gear mesh frequency.
Let's briefly look at the following example:
 Consider a machine that generates a gear mesh
frequency of 1,080,000 cpm (18kHz).
 Let's assume that there is some vibration (movement)
occurring at the gear meshing frequency due to the
interaction of the gear teeth.
 Let's further assume that the amount of physical
movement - the distance back and forth, so to speak - is
3 millionths of an inch (0.003 mils, or 0.076 um).
 Everyone can visualize the tiny, tiny amount of
movement generated.
 However, to cause the structure (rotor) to move back
and forth even such an incredibly tiny amount 18,000
times per second requires a force equal to 50x the force
of gravity. That's right, 50 g's.
 The failure will occur due to that force being applied

repeatedly.
Let's look at some examples relating the different amplitude
units to the likely failure mode.
Recommended Frequency Ranges for Different

Amplitude Units
Displacement Units: < 600 cpm (< 10 Hz)
There are instances where the use of displacement amplitudes is essential
even at very high frequencies.
Velocity Units: 300 - 120,000 cpm (5 - 2,000 Hz)

Acceleration Units: > 60,000 cpm (> 1,000 Hz)
When To Use Which Amplitude Unit(s)
Every machine has certain operational characteristics which must be
considered when creating the database. Similar machines have similar
characteristics and similar (many times identical) database point set-
ups. The critical question that must be asked for every machine for
which you are creating a database is simply what problems may develop
on this machine and what vibration frequencies will be generated by
each of these problems.
You may need multiple measurements on a

particular location to get the level of
protection you would like.
In other words, you must create each database point with a specific
purpose in mind.
Consider a motor driving some component connected with a coupling.
What problems may occur on the motor and what vibration frequencies
will each generate ?
 Mechanical influences (unbalance, misalignment, etc.) at 1x, 2x,
and 3x rpm (also be referred to as orders).
 Pumps can generate hydraulically-related vibration at the number
of vanes x rpm - vane pass frequency.
 Compressors do likewise at lobe pass and vane pass frequencies
(to name only 2 types).
 Fans can generate at blade pass frequency.
 With rolling element bearings, vibration at 30kcpm - 50 x rpm (up
to 150kcpm) is typically generated during stages leading up to
failure.
When To Use Which Amplitude Unit(s) ?

Compressor Example
Consider the following example for which we will discuss the
frequencies encountered: a direct driven screw compressor with an
input speed of 3580 rpm. The motor directly drives a bull gear with 48
teeth which drives a pinion gear with 36 teeth. The rotor being driven
by the pinion gear has 4 lobes while the driven rotor has 6 lobes. To
determine what frequencies the potential problems may create, we
need to specifically lay out the frequencies that will be generated on
this machine and consider what problems can develop from the
machine components. The machine schematic is shown here:
Let’s calculate exactly what
frequencies need to be monitored
on the compressor end only:
Frequency
Calculation Frequency
Name
4-Lobe 4773 rpm
3580 x
Rotor
48T/36T
Speed
4-Lobe 4773 x 4
19092
Pass
cpm
Frequency
6-Lobe 3182 rpm
4773 x
Rotor
4T/6T
Speed
6-Lobe 3182 x 6
19092
Pass
cpm
Frequency Compressor Schematic
Bull Gear Motor Speed = 3580 rpm
3580 rpm 171,840
Mesh
x 48T cpm
Freq:
Pinion 171840
4773 x 36T
GMF cpm
343,680
2x GMF 2 x GMF
cpm
515,520
3x GMF
cpm
So we need to monitor the compressor bearings over a range of
frequencies spanning 3182 cpm (1x 6-lobe rotor) to 515,520 cpm (3x
gear mesh frequency). Although it can technically be done with a
single reading, using only one amplitude unit would be a problem
since velocity is no good at 515,520 cpm and acceleration is no good
at 3182 cpm.
But why would using only one amplitude unit be a problem ? The
answer goes back to the relationship between the units and their
sensitivity to different failure modes.
Let's look at some examples.
Formulas Relating Amplitude & Frequency
As stated earlier, there are formulas that relating each of the
amplitude units to one another through the vibration
frequency. The following just lists a few of the possible
variations. You should note that the software carries out the
formulas - the following pages attempt to illustrate the
concept only.
Imperial Units: Metric Units:
Displacement = mils Displacement = um
Velocity = in/sec Velocity = mm/sec
Acceleration = g's Acceleration = g's
Frequency = cycles/min Frequency = cycles/min
Displacement = (19,231 x Displacement = (19,231 x V) /
V) / F F
Velocity = 0.000052 x D x F Velocity = 0.000052 x D x F
Acceleration = 0.00027 x V x Acceleration = 0.0000107 x V
F xF
When To Use Which Amplitude Unit(s)

Every machine has certain operational characteristics which
must be considered when creating the database. Similar
machines have similar characteristics and similar (many times
identical) database point set-ups. The critical question that
must be asked for every machine for which you are creating a
database is simply what problems may develop on this
machine and what vibration frequencies will be generated by
each of these problems.
You may need multiple measurements on a

particular location to get the level of
protection you would like.
In other words, you must create each database point with a

specific purpose in mind.
Consider a motor driving some component connected with a
coupling. What problems may occur on the motor and what
vibration frequencies will each generate ?
 Mechanical influences (unbalance, misalignment, etc.) at
1x, 2x, and 3x rpm (also be referred to as orders).
 Pumps can generate hydraulically-related vibration at the
number of vanes x rpm - vane pass frequency.
 Compressors do likewise at lobe pass and vane pass
frequencies (to name only 2 types).
 Fans can generate at blade pass frequency.
 With rolling element bearings, vibration at 30kcpm - 50 x

rpm (up to 150kcpm) is typically generated during stages
leading up to failure.
Formulas Relating Amplitude & Frequency

As stated earlier, there are formulas that relating each of the
amplitude units to one another through the vibration
frequency. The following just lists a few of the possible
variations. You should note that the software carries out the
formulas - the following pages attempt to illustrate the
concept only.
Imperial Units: Metric Units:
Displacement = mils Displacement = um
Velocity = in/sec Velocity = mm/sec
Acceleration = g's Acceleration = g's
Frequency = cycles/min Frequency = cycles/min
Displacement = (19,231 x V) / Displacement = (19,231 x V) /
F F
F xF
Example #1:
A Bearing Vibrates 100 Mils Pk-
Pk @ 30 cpm
Displacement @ 1x Displacement @ 1x
rpm = 100 mils rpm = 2540 um
English Units: Metric Units:
Velocity = Velocity = 0.000052
0.000052 x D x F xDxF
V = 0.000052 x 100 V = 0.000052 x
mils x 30 cpm 2540 um x 30 cpm
V = 0.16 ips V = 4 mm/sec
Acceleration = Acceleration =
0.00027 x V x F 0.0000107 x V x F
A = 0.00027 x 0.16 A = 0.0000107 x 4
x 30 x 30
A = 0.0013 g's A = 0.0013 g's
Failure Likelihood:
Stress = HIGH
Fatigue = Low
Force = None
What unit is best to monitor with at this
frequency ?
Displacement
Example #2:
A Bearing Vibrates 10 Mils Pk-
Pk At 1000 cpm
Displacement @ 1x rpm = 10 Displacement @ 1x rpm = 250
mils um
V = 0.000052 x 10 mils x V = 0.000052 x 250 um x
1000 cpm 1000 cpm
V = 0.52 ips V = 13 mm/sec
F xF
A = 0.00027 x 0.52 x 1000 A = 0.0000107 x 13 x 1000
A = 0.14 g's A = 0.14 g's
Failure Likelihood:
Stress = MODERATE
Fatigue = MODERATE
Force = None
What unit is best to monitor with at this frequency ?
Velocity
Why not displacement ?
Because it can also be used to monitor higher frequencies on
this machine.
Example #3:
A Bearing Vibrates 3 Mils Pk-Pk At 9,000 cpm
Displacement @ 9,000 cpm = Displacement @ 9,000 cpm =
3 mils 75 um
V = 0.000052 x 3 mils x 9,000 V = 0.000052 x 75 um x 9,000
cpm cpm
V = 1.404 ips V = 35.1 mm/sec
F xF
A = 0.00027 x 1.404 x 9,000 A = 0.0000107 x 35.1 x 9,000
A = 3.41 g's A = 3.41 g's
Failure Likelihood:
Stress = None
Fatigue = HIGH
Force =Low
Velocity
Example #4:
A Bearing Vibrates 0.2 Mils Pk-Pk At 60,000
cpm
Displacement @ 60,000 cpm Displacement @ 60,000 cpm
=0.2 mils = 5 um
V = 0.000052 x 0.2 mils x V = 0.000052 x 5 um x
60,000 cpm 60,000 cpm
V = 0.62 ips V = 15.6 mm/sec
Acceleration = 0.0000107 x V
Acceleration = 0.00027 x V x F xF
A = 0.00027 x 0.62 x 60,000 A = 0.0000107 x 15.6 x
60,000
A = 10 g's A = 10 g's
Failure Likelihood:
Stress =None
Fatigue =Moderate
Force = Moderate
Velocity or Acceleration
Choice will be determined by what other problems
(frequencies)
need to be monitored for developing problems.
Example #5:
A High Speed Compressor Rotor Shaft
Vibrates 0.003 Mils Pk-Pk At 1,080,000 cpm
Displacement @ 1,080,000
Displacement @ 1,080,000
cpm = 0.077 um
cpm = 0.003 mils
(7.7 millionths of a
(3 millionths of an inch)
centimeter)
V = 0.000052 x 0.003 mils x V = 0.000052 x 0.077 um x
1,080,000 cpm 1,080,000 cpm
V = 0.17 ips V = 4.32 mm/sec
Acceleration = 0.00027 x V x F Acceleration = 0.0000107 x V
xF
A = 0.00027 x 0.17 x A = 0.0000107 x 4.33 x
1,080,000 1,080,000
A = 50 g's A = 50 g's
Failure Likelihood:
Stress = None
Fatigue = Low
Force = REALLY HIGH
Acceleration
General Equipment Amplitude Guidelines:
Before discussing amplitude guidelines it should be clearly
understood that these are only guidelines - a starting point to
begin from. The best vibration analysts get to know the normal
vibration characteristics of their machines and look for
deviation from that norm. The values that follow here are
generally regarded in the categories shown but they are and
should be considered guidelines.
Displacement Units:
All Must know frequency to assess severity. At very
Frequencies: low frequencies, even displacement amplitudes
of 40 or 50 mils pk-pk or even more can be
only moderately harmful to the bearing and the
structural components affected.
Note: The values listed here as guidelines for velocity and
acceleration are 'peak' amplitudes. Equivalent RMS values are
30% lower.
Velocity Units:
Within Frequency Range: 300 - 120,000 cpm
Hard Supports Condition Soft Supports
English Metric Very Good English Metric
< 0.10 < 2.5 < 0.15 <4
in/sec mm/sec in/sec mm/sec
0.10 - 2.5 - 7.5 0.15 - 0.45 4 - 12

Satisfactory
0.30 ips mm/s ips mm/s
0.30 - 7.5 - 11.5 0.45 - 12 - 17
Fair
0.45 ips mm/s 0.67 ips mm/s
0.45 - 11.5 - 16 0.67 - 17 - 23

Rough
0.60 ips mm/s 0.90 ips mm/s
> 0.90 > 23
> 0.60 ips > 16 mm/s Destructive
ips mm/sec
Acceleration Units:
Frequencies: <Below 30kcpm, the problem will be better
30,000 kcpm:monitored with velocity as the amplitude unit
of choice.
Frequencies:Between 30kcpm and 120kcpm, you must
30k-120kcpm:know frequency to fully judge the severity.
Velocity can confidently be used as a partner
unit in this range.
Frequencies: > < 2.00 g's Good
120kcpm:
2.0 - 5.0 g's Fair
5.0 - 10.0 g's Rough
> 10.0 g's Very Rough
Amplitude Guidelines For "Other" Equipment
Types:
Slow Speed: Typically generates lower amplitudes. For shafts
< 300 rpm, Time Domain plots should be used.
Machine Tool or "Precision" equipment: Typically tolerates
much lower amplitude levels. Guidelines (vibration alarms) for
each machine must be established. Since this equipment
usually involves keeping the finish quality within certain
tolerances or specifications, establishing a vibration level just
below which those machines go "out-of-tolerance" can be a
very effective method. Bearings should be monitored
regardless of the overall machine condition.
Complex Vibration Generators: Typically generates higher
amplitude levels: Refers to machines that have large forces
normally or a lot of vibration sources. High pressure, lobe-type
blowers (Roots, for instance) often involve motor frequencies,
belt frequencies, lobe pass frequencies, 2-rotor speeds, gear
frequencies and aerodynamic forces as well as loaded and
unloaded conditions. The resulting vibration patterns can be
high relative to the General Machine amplitude references and
yet normal for your machine. Be careful in over-reacting. At
least one manufacturer's vibration guidelines are as follows:
English Units: in/sec Metric Units: mm/sec Classification:
< 0.45 < 11.4 Excellent
0.45 - 0.90 w / no 11.4 - 22.8 w / no
Good
peak > 0.45 peak > 11.4
0.45 - 0.90 w / 1+ 11.4 - 22.8 w / 1+
Fair
peaks > 0.45 peaks > 11.4
> 0.90 w / no peak > > 22.8 w / no peak >
Rough
0.90 22.8
Peak(s) > 0.90 Peak(s) > 22.8 Very Rough
Program Needs: Programs must be set up based on the needs

of the individual pieces of equipment. Unless your program has
loads of similar or identical pieces of equipment, a broad brush
cannot necessarily be used. The actual frequencies being
generated on the machines must be determined or at least
estimated reasonably well. That information should then be
used to specify the plots and data collected based on that. The
successful analyst will also get to "know" the machines and
their typical vibration patterns. That knowledge is possibly the
analyst's strongest line of defense against unexpected failures.
Phase
What Does "Phase" Tell Us ?
 Phase enables us to compare the relative direction of

movement of various locations on a machine.
Simply put, phase tells us:
What Direction ?
Still not clear ?
What Is Meant By What Direction ?

 Imagine a snapshot of a machine operating. In the
snapshot, imagine being able to see arrows drawn at
each bearing indicating what direction that bearing was
moving in at that moment in time.
 From the snapshot, you can thereby determine if the

bearings are moving in the same direction at the same
time (in unison or 'in phase') or not.
Phase helps determine how different locations on the machine
(different bearings, usually) are moving relative to one
another
Rules For Measuring Phase
Rule #1: Phase is a relative measurement.
 Readings at different locations on a bearing are analyzed
relative to one another.
 Readings at the same angle (direction) on adjacent
bearings are analyzed relative to one another.
 Readings at different times at the same exact location
are analyzed relative to one another.
 These comparisons give us information on how the
components behavior - namely how they are moving
relative to one another.
 There is only one use for a single phase reading which
we will discuss shortly.
Rule #2: Phase is collected at ONE frequency at a time.

 The analyzer must be told what frequency. This is done
by tuning the analyzer with either a strobe light or a tach
pulse from the shaft.
Rule #3: The analyzer must be able to detect a vibration

signal at that frequency.
 Since the angle is calculated based on the arrival of a
sinusoidal peak from the vibrating component, there
must be an amplitude peak to get a phase angle.
Simple Phase Analysis
Data Conventions
 In order to discuss phase at the most basic level, we
need a convention to use. Phase is represented by the
360° of a circle. Since a high degree of accuracy is not
required in a simple, general phase analysis (what we
will be discussing here), phase is most easily referred to
in terms of clock face numbers: 1 o'clock through 12
o'clock. The phase "angle" is based on where the mark
appears (the red key in the example shown here) on
the clock face. It is shown here at 12 o'clock but could
appear at any angular location.
 Applications requiring more precise detail (i.e. balancing
or phase monitoring on a turbine) will require the use of
specific angular references (0 - 359°).
Clock Face Numbers Use For Simple

Phase Analysis
What Methods Are Used for Measuring Phase ?

 A 'Phase-Triggering' Strobe Light. This type of strobe
light differs in two ways from a conventional, hand held
strobe light:
o It either feeds frequency information (its flash rate)
into the analyzer or has its flash rate set by the
analyzer.
o It can be triggered by a vibration signal from the
analyzer.
 A Photoelectric Eye (or related mechanism that provides

a 1x rpm pulse from the shaft).
How Is Phase Measured With A Strobe Light ?

This involves several steps - each one designed to satisfy one
of the rules for measuring phase. We must start with a couple
of assumptions.
 We have a machine with a 'problem' - higher than
desired amplitude at a particular frequency.
 The frequency we are concerned with in this case (for

simplicity sake) is 1x rpm.
Step #1 - Mount Transducer
 The first step in collecting a phase
reading is to place the transducer
on the bearing in the direction
desired.
 The transducer provides the
analyzer with the vibration signal
necessary (as per Rule #3).
Step #2 - Locate a Reference Mark

 Commonly a key or keyway, this
mark is what we will be using as
our 'clock face' reference when it
shows up under the strobe flash
(as per Rule #4).
Step #3 - Tune the strobe to 1x rpm
(the pertinent frequency)
 Tuning the strobe light properly
will freeze the shaft with 1 mark
as shown here. Be careful you
haven't tuned it to 1/2 rpm since
that will also show 1 mark.
 By tuning the strobe light, you
have told the analyzer at which
frequency we will be measuring
phase (as per Rule #2).
 With a strobe light (as opposed to
a photoeye). This allows the
analyzer to focus on a specific
vibration signal - a relatively
clean sinusoid (since all other
frequencies are being filtered out)
that is occurring at or very close
to (+/- 0.75%) the tuned
frequency.
 You can compare this to what a

radio does - it tunes to one
frequency and filters all others
out.
What Is Phase Telling Us ?
So far, nothing. The strobe light is being triggered by an
internal trigger on either the analyzer or the light itself - it is
acting exactly like a typical, hand-held strobe light. There is
one final step which must be performed:
 A switch (or keystroke) on either the analyzer or the
strobe can be activated which changes how the strobe is
being triggered.
Once this switch activated, the strobe stops to using its

internal trigger to flash and starts to use the vibration sinusoid
being detected. Therefore:
 Every time a peak signal is detected by the analyzer, it
sends a signal to the strobe light to flash.
 The strobe light therefore flashes at exactly the
frequency of the vibration being detected.
 If the vibration is being mechanically generated by the
shaft (rotor), the strobe will flash at an exact harmonic of
running speed (1x rpm in our example but it could be 2x,
3x or any higher harmonic) and the shaft (mark) will
freeze under the strobe flash.
 If the vibration is being generated by some other source
- any other source (bearings, electrical, other machines
nearby, etc.), it will not flash at an exact harmonic of
running speed and the shaft will (mark) will not freeze
under the strobe flash.
Let's look at a graphic example of how this process works and

why it is important.
Radial Phase Analysis Around A Single Bearing

Phase Angle #1
What is happening in the animation here ?
 The transducer is mounted vertically so the movement
(vibration) being measured is vertical only.
 The red dot represents the "at-rest" position of the shaft.
 The black dot represents the center point of the shaft.
 There is a "heavy" spot on the rotor that causes a
centrifugal force to be generated that results is causing
the shaft center point (the black dot) to rotate around the
'at-rest' center point (the red dot).
 The "+"peak' occurs when the bearing is displaced the
maximum amount towards the transducer. In other
words, as the black dot passes the 12:00 position
(passes the transducer).
 The location of the heavy spot can be determined at any
moment in the shaft's rotation by imagining a line drawn
from the red dot directly through the black dot. This is, of
course, not able to be seen in real life on a real machine.
 What we can see is some "mark" (usually a key) that
becomes visible under the flash of the strobe light once
per revolution (the yellow dot). This is our reference
mark. The mark is 45° behind the heavy spot.
 In real life, we would not know where the mark is located

relative to the heavy spot. However, we do know that
unless we do something to change the location or size of
the heavy spot, those two positions will not change
relative to one another as we move the transducer. They
will remain 45° apart.
Radial Phase Analysis Around A Single
Bearing
Phase Angle #1
 With the transducer mounted vertically, the peak signal
arrives when the black dot is at the 12:00 position (0°).
 The strobe light is set up to flash when an amplitude
peak arrives so it will flash at that moment.
 When the strobe flashes, the key shows up at 10:30
(315°).
We now have 1 piece of phase data on this bearing at a

frequency of 1x rpm:
 With the transducer mounted vertically, our phase angle
is 10:30 (315°).
Radial Phase Analysis Around A Single
Bearing
Phase Angle #2
What has been changed in this animation ?
 The transducer is mounted horizontally so the movement
being measured is horizontal only. Another way to say
this would be that the transducer has been moved 90°.
So now what is happening ?

 The peak signal will still be detected when the bearing is
displaced the maximum amount towards the transducer -
as the black dot passes the 3:00 position (90°).
 The strobe light is set up to flash when an amplitude
peak arrives so it will flash at that moment.
 The mark remains 45° behind the heavy spot so when
the strobe flashes, the mark will show up at about 1:30.
We now have 2 pieces of phase data on this bearing at a

frequency of 1x rpm:
 With the transducer vertical, our phase angle is 10:30
 With the transducer horizontal, our phase angle is 1:30
What Do These Phase Angles Mean ?

We have checked the radial vibration at 2 angular locations
around a single bearing. This has provided us with two
valuable pieces of information. The first is:
 When triggering the strobe light from the vibration signal,
the shaft appeared frozen. In other words, the mark was
not rotating - it remained stationary (even if it wobbled
back and forth a bit).
What does this mean ?

 It means the vibration IS coming from this shaft. This is
now established as FACT - something only a phase
reading can do (as we will see later in the 'Plots' section).
What is the second piece of information ?

 When moving the transducer to a new angular location
(moving from horizontal to vertical is 90°), the location of
the mark (the phase angle) shifted the same amount
(90°).
What does this mean ?

 The shaft is vibrating in a more circular orbit that
typically indicates unbalance.
Let's examine each of these facts more closely and see how
they were arrived at from the phase measurements we
recorded.
Facts Established by Phase:

Frequency Confirmation
Fact #1: The vibration IS (or IS NOT) coming from the shaft
 First, you must remember that the strobe light is actually
being triggered by the vibration signal. Every time the
analyzer detects a peak signal from the transducer, it
instructs the strobe light to flash. The strobe flash will,
therefore, flash at exactly the same rate as the vibration
is occurring.
 Second, mechanical vibration - i.e. vibration being
generated by the rotation of the rotor - occurs only at
exact multiples (harmonics) of the rotational speed
(rpm). These vibrations are known as synchronous
vibrations.
 Third, sources of vibration other than the rotor - belts,
bearings, electrical vibrations and other non-synchronous
vibration sources will not generate vibration at exact
multiples of the rotational speed. This is true regardless
of how close they are to being synchronous. Even if the
adjacent machine is running at 0.1 rpm different speed,
it is still a different speed - it is still non-synchronous.
Since the strobe is flashing at exactly the vibration frequency

being generated, whether or not the mark (shaft) appears
frozen under the strobe light reveals whether the vibration is
synchronous or non-synchronous. This test will be referred to
as a:
Frequency Confirmation
Frequency confirmation simply means you are confirming the
source of the vibration is the rotor that appears frozen under
the strobe flash and it is the ONE use for a single phase
reading.
Frequency confirmation is a simple test that requires only a
few seconds to perform but can be crucially important to the
successful diagnosis of a machine's problem.

Rough Orbit Shape
Fact #2: Comparing the phase angles at 2 separate
transducer locations (vertical and horizontal in our example)
allows you to make the rough judgement of the shape of the
orbit - one that is more circular or one that is flatter in shape.
Before explaining this, you must
understand what is meant by the
'shaft orbit'. Consider the following:
 A shaft is generating vibration
at only 1x rpm.
 The amplitude horizontally is 'X'
ips or mm/sec.
 The amplitude vertically is
exactly the same - 'X' ips or
mm/sec. Figure 1
 In fact, no matter what
direction the transducer is
oriented (pointed), the
amplitude at 1x rpm is 'X' ips or
mm/sec.
 By plotting the recorded
amplitude values in the
appropriate direction (Figure
1), you can plot the 'orbit', or
shape, of the shaft center
point.
 In this particular example, the

orbit is a circle. In the 'real'
world, this would be virtually
impossible.
 Another shaft is generating
vibration at only 1x rpm.
ips or mm/sec.
 The amplitude vertically is 80%
of the horizontal amplitude -
0.8(X) ips or mm/sec.
appropriate direction (Figure Figure 2
2), you can again plot the
'orbit', or shape, of the shaft
center point.
 In this example, an elliptical

orbit is plotted. These relative
amplitudes would not be
considered unusual.
 A third shaft is generating
vibration at only 1x rpm.
ips or mm/sec.
 The amplitude vertically is 50%
of the horizontal amplitude -
0.5(X) ips or mm/sec.
appropriate direction (Figure Figure 3
3), you can again plot the
'orbit', or shape, of the shaft
center point.
 In this example, an elliptical
orbit is plotted. These relative
amplitudes would not be
considered unusual.
When the amplitudes are relatively equal, you can see how the
'orbit' approaches a circle. But let's look at the other extreme.
What if the horizontal amplitude were 'X' and the vertical
amplitude were very, very low - even 0.00 ips or mm/sec.
That 'orbit' would be a straight line - linear movement
(vibration).

Rough Orbit Shape
transducer locations (vertical and horizontal in our
example) allows you to make the rough judgement of the
shape of the orbit - one that is more circular or one that is
flatter in shape.
So we have examined the extremes that are possible:
 Circular orbits (where the amplitudes are exactly the

same regardless of transducer direction)
 Linear orbits (where the amplitude is zero in one

direction).
These extremes are equally unlikely. Everything in between is
an ellipse. We can therefore realistically use the rule of thumb
that all orbits are elliptical.
However, it can be helpful in
determing the nature of the
problem to know whether the
orbit is approaching circular or
linear (which, it should be noted,
can be better "quantified" by
plotting amplitude values).
 When the transducer was in
the vertical position, the
peak signal arrived at a
specific moment and caused
the strobe to flash at that Vertical Transducer Location
moment. Under this strobe

flash, the key showed up ay
the 10:30 position (315°).
 When shifted to the

horizontal position - 90°
away - the moment the
peak signal arrived changed
by 1/4 rotation - the same
90°. Under this strobe flash,
the key shows up at 1:30
(45°) - a shift of 90° from
the first phase reading.
Because the movement shown
here is very circular in nature, the
phase angle will change based on
the angular location of the
transducer. Regardless of the
angular location you place the
tranducer at, the peak signal will
arrive as the black dot passes the
transducer location.
This type of orbit - an ellipse that
approaches circular - is
characteristic of one of the most
common vibration problems:
unbalance. When excessive
vibration is detected at 1x rpm
(the only frequency unbalance
generates vibration at),
Horizontal Transducer Location
unbalance is one of several
possible sources. This type of
phase test can help differentiate
between unbalance and the other
potential sources of the high
vibration levels.
Rough Orbit Shape
transducer locations (vertical and horizontal in our
example) allows you to make the rough judgement of the
shape of the orbit - one that is more circular or one that is
flatter in shape.
 We have discussed what it means when the phase shifts
an angular amount very close to the angular amount you
move the transducer - a rounder elliptical orbit.
Now let's examine what a 'flatter' (more
linear) orbit might reveal during a phase
analysis:
 The animation at right is similar to
the previous without the 'strobe
flashes'. The red dot still represents
the 'at rest' shaft center point and
the black dot is still the actual shaft
center point during operation.
 Note the black dot passes just above Linear Vibration

and below the red dot. The amplitude
difference is on the scale of 10:1 with
horizontal being higher.
 The red arrow actually represents the phase angle. It
follows the direction of displacement and is always
pointing from the red dot towards the black dot. In the
previous animation (where we had a circular orbit), we
didn't show an arrow from the red dot to the black dot -
we showed a once per revolution 'flash' that occurred
based on where the transducer was but the flash
occurred only when the red dot, black dot and transducer
were in that order in a straight line.
 You can see that the arrow remains relatively steady for
relatively long periods of time (as it passes 3:00 and
9:00) during the revolution.
 Under this condition, you could place a transducer
anywhere between about 1:00 and 5:00 and observe
basically the same phase angle - the arrow moves very
little as the black dot passes that area.
 Likewise, you could place a transducer anywhere
between about 7:00 and 11:00 and observe basically the
same phase angle but it would be 180° different than if
the transducer were between 1:00 and 5:00.
 If the transducer were located between 5:00 and 7:00

(150°-210°) or between 11:00 and 1:00 (330°-30°), you
would get an unsteady phase reading although it would
NOT rotate - only wobble back and forth.
What Is The Significance Of The Orbit
Shape ?
A more circular ellipse (i.e. when you move the transducer a
certain angular amount, the mark shifts an equal amount)
usually indicates:
 More of an 'unbalance' type vibration force (centrifugal
force that is fairly equal all the way around)
A flatter ellipse (i.e. when you move the transducer a certain

angular amount, the mark either does not shift or shifts 180°)
usually indicates:
 Flatter, more 'linear' motion indicates problems that
affect a specific direction such as looseness, resonance,
and (on belt or chain drives) bent shafts or eccentricity.
These are Rules of Thumb and a number of variables such as

structural strength (which can influence amplitude values in
one direction versus another) much be considered. Phase
analysis, however, can provide some important information
related to how the components are moving.
Radial Phase Analysis Across Two Adjacent

Bearings
Up to now, we have only analyzed
radial phase readings taken on a
single bearing at different angular
locations. What about comparing
adjacent bearings ? Phase is still
used as a "relative" reading.
 When comparing phase
angles between bearings,
you compare phase angles
with the transducer mounted
in the same angular location
(vertical to vertical,
horizontal to horizontal,
etc.).
Analysis will reveal how the

bearings are moving relative to
one another at the frequency
being measured. Bearings "In-Phase"
 In the animation here, the yellow balls represent heavy

spots on the rotor. Since they are located at similar
angular locations, the forces generated will cause the
bearings to move together, or "in-phase".
 If a transducer were mounted vertically on either of the
two bearings, the strobe would flash as the heavy spot
passes the 12:00 position since the peak signal arrives at
that moment.
 Our mark (the red key on shaft) is in the same position
(i.e. same phase angle) regardless of which bearing the
transducer is mounted on.
 The bearings are in-phase.
Radial Phase Analysis Across Two Adjacent

Bearings
 In the animation shown
here, the heavy spots are
opposite one another.
 The heavy spot on the left
hub affects the left bearing
more than the right.
 The heavy spot on the right
hub affects the right bearing
more than the left.
 If the transducer were Bearings "Out-of-Phase"
mounted vertically on the
right bearing, the strobe
would flash as the heavy
spot on the right passed the
12:00 position. That occurs
when the key is just past
12:00.
 If the transducer were
mounted vertically on the
left bearing, the strobe
would flash as the heavy
spot on the left passed the
12:00 position. That occurs
when the key is just past
6:00.
 The bearings, therefore, are

180° 'out-of'phase'.
What is the significance of this ?
 The significance is in balancing the rotor. Unlike the
previous example, a 2-plane balance such as this
requires 2 separate corrections (solutions) - one for the
left side heavy spot and one for the right side heavy spot
- while the 1-plane (the previous example) requires only
1 correction.
Axial Phase Analysis Around A Single Bearing

Phase analysis can also reveal some
important information when performed in the
axial direction. Let's examine what is
happening in the animation here:
 There are 6 transducers mounted
axially - the movement being measured
is axial only.
"Planar" Motion
 We are measuring phase in the same
manner as for radial phase - namely
triggering the strobe light at a peak
amplitude and freezing some reference
mark (a key ?).
 With the transducers mounted in this manner, the peak
signal will arrive when the bearing is displaced its
maximum to the right.
 Each of the 6 transducers are displaced the maximum
amount to the right at the same moment.
 Therefore, regardless of which transducer you are
analyzing from, the phase angle (location of the mark)
will be the same.
Just as our radial phase analysis provides important

information on the 'shape' of the movement radially, an axial
phase analysis provides important information on the 'shape'
of the movement axially. In the above animation, we have
found there is no phase shift as we move the transducer
around the face of the bearing. But how else could it be
moving ?
The Bearing Is Moving Axially In A "Planar" Motion (Not

Twisting On The Shaft)
What is the significance of this ?
The Source Of The High Axial Vibration Is Not Caused By

Incorrect Installation Of This Bearing
Axial Phase Analysis Around A Single Bearing

The bearing could be moving as shown here
- a 'twisting' action on the shaft or in the
housing. Let's examine this animation:
 The transducers are mounted in
identical positions to the last animation.
 The peak signal will arrive at each
individual transducer location when the
"Twisting" Motion
bearing is displaced its maximum to the
right at that location.
 That peak arrives at a different moment
for each transducer.
 Since the transducers are 60° apart,
the phase angle measured at each will
be 60° different than the adjacent
transducers.
How can this information help the analysis ?

The most likely source of 'planar' axial motion (as in the
previous animation) is:
 Misalignment
The most likely sources of 'twisting' axial motion (as shown

above) are:
 A bearing cocked on the shaft
 A bent shaft through the bearing
 Housing distortion (i.e. soft foot) causing twisting motion
The phase analysis thereby helps in differentiating between

machine problems that cause similar vibration symptoms (aids
in diagnosis of root cause).
Axial Phase Analysis Across Adjacent Bearings

An 'axial phase analysis' is a procedure that is conducted one
bearing at a time. On a smooth running machine, all axial
phase readings (on any bearing at any angular location) will be
similar. On a machine with high axial vibration, the following
procedure should be used if possible:
 Each bearing should be checked for planar vs. twisting
motion. Any twisting motions detected should be
corrected.
 The bearings should be compared to one another. Any
significant (> 60°) phase shift should raise a flag that
something is not right.
 Most commonly, a phase shift >60° will be detected
when comparing the bearings closest to the coupling and
will typically indicate misalignment.
 When a phase shift is detected between bearings on the
same component (i.e. motor), housing distortion such as
soft foot should be checked.
However, transducer orientation must be accounted for (see

the next page).
Understanding Transducer Orientation

The "orientation", or direction, of the transducer is extremely
important. The reason for this is simple.
 The '+' and '-' directions are defined by the transducer.
Either '+' is towards the transducer or away from it.
 The strobe will be triggered by a '+' signal.
 If you change the orientation, you change the definition
of '+'.
The animation here shows

both transducers oriented
(pointed) in the same
direction:
 The '+' direction is
defined as 'towards'
the transducer.
Are These Bearings "In-Phase" ?
 The strobe will trigger
on a '+' peak so each
transducer will cause
the strobe to flash
when they reach their
maximum
displacement to the
left (towards the
transducer).
 The strobe flashes

when the reference
mark is just past 6:00.
 The transducers have the same orientation and generate
the same phase angles so they are in-phase.

It is common to collect phase readings across a coupling. In
that case, you will almost always have the transducer
orientation shown below - the transducers are oriented in
opposite directions.
 The transducer
mounted on the left
bearing will cause the
strobe to trigger when
the bearing is at
maximum
displacement to the
right (towards that
transducer). At that
moment, the reference
mark is approaching
the 12:00 position.
 The transducer
mounted on the right
bearing will cause the
strobe to trigger when
the bearing is at
maximum
displacement to the
left (towards that
transducer). At that
moment, the reference
mark is approaching
the 6:00 position (just
out of sight).
 Since they have phase angles 180° different, the
bearings may appear out of phase - but are they?
 The bearings are moving in the identical fashion as the
previous page and we established they are 'in-phase'.
 The difference is the transducer orientation. It must be
accounted for in the final readings.
 Since the orientation of the two transducer is 180°

different, a 180° adjustment must be made to one of the
measured phase angles. Once that adjustment is made,
the phase angles become equal - the bearings are
moving "in-phase" with one another.

 In the animation at
left, the peak signal is
arriving at each
transducer
simultaneously.
 Regardless of which
transducer orientation
is being used, the
strobe is being
triggered with the
mark approaching
12:00. But are they
"in-phase" ?
 Of course not - not if
you make the required
180° adjustment to
one of the angles to
account for transducer
orientation.
 The bearings are 180°

'out-of-phase'.
The Most Common Use
of Phase:
Diagnosing
Misalignment
 Phase is a common and reliable way to diagnose
misalignment.
 The procedure involves conducting an axial phase
analysis around bearing faces as well as from bearing to
bearing.
 The point at which you find a phase shift of > 30° is
typically the source of a problem (possibly not the only
problem).
 Note in high axial vibration in the animation shown here.
These components are 'angularly misaligned'.
 Are the faces of the machine "in-phase" or "out-of-

phase" with one another ?
How would the transducers be oriented in this case ? Almost
certainly 180° opposite (one pointing left and one pointing
right). You will have to make an adjustment for transducer
orientation.
Are the machine faces "in-phase" or

"out-of-phase" with one another?
How Amplitude Units Affect Phase Angles

Once you start a phase analysis, you should never change
the amplitude units you are using. Although we have been
creating examples using displacement units, velocity units
are the most versatile and commonly used. Let's review how
phase angles are determined:
 The strobe flash is triggered by the arrival of the peak
amplitude signal from the transducer.
 The location of the reference mark is determined by the
moment of the strobe flash.
 The timing of the arrival of the peak amplitude signal
determines the location of the mark.
The key to why amplitude units affect phase angles is that:

 The moment the peak signal arrives is determined by
which amplitude unit is being used.
Using displacement units, the peak will arrive when the
transducer is displaced the maximum amount in the '+'
direction. The mark is at 10:30 (315°).

 Finally, let's look at acceleration units. When will the peak
arrive in that case ?
 Using acceleration amplitudes, the peak will arrive when
the bearing housing/structure is pushing (applying force)
the greatest amount in the "+" direction.At that moment,
the mark is at 4:30 (135°). That is 90° different from
what we measured with velocity units.
 By changing from velocity to acceleration, we have

induced another 90° phase shift (180° from the phase
angle using displacement units).
Displacement Sinusoid Velocity Sinusoid Acceleration Sinusoid
Comparing these images (from the 'Amplitude' Section) will

help you understand why the peaks arrive at different
moments for different amplitude units.
 Velocity is 1/4 cycle (90°) ahead of displacement and 1/4
cycle behind acceleration.
 Acceleration is 1/2 cycle (180°) away from displacement.
When is this information important ?

 When doing any work with phase (general analysis,
balancing, etc.), don't change the amplitude unit with
which you are working.
How Is Phase Measured With A Photoeye ?
The procedure for collecting phase with
a photoeye is somewhat different than
with a strobe light. Let's measure phase
at 1x rpm as shown here.
Step #1 - Mount Photoeye & Trigger
 The first step in measuring phase

is to properly set up the photoeye.
It must be mounted rigidly next to
the shaft so it can detect a trigger
mark rotating on the shaft. The
mark is often a piece of reflective
tape. With some modern detectors
(like 'Lasertachs'), pattern
recognition is used and reflective
tape is often not needed. The
trigger gives the analyzer a 1x
rpm pulse (gives it the frequency).
Step #2 - Mount Transducer

 Mount the transducer at the
location and direction desired.
Step #3 - Instruct Analyzer to collect a

phase measurement.
 A keystroke will tell the box to
collect a phase reading.
How does the photoeye calculate the phase angle ?
 The 1x rpm trigger provides the box with the period of
the frequency being measured - the period of 1 cycle at
1x rpm.
 The box takes that period (measured in seconds) and
divides it by 360.
 Once all the math is completed, the box is ready. When
instructed to collect a phase angle, it waits for the trigger
mark to pass so it can begin a count up to 360.
 After beginning it's count, it awaits the arrival of a peak
signal from the transducer. The moment it has that, it
stops the count. That is the phase angle.
 For the sake of accuracy, several angles are checked and

an average is what is eventually displayed as the phase
angle.
What Is "Time Synchronous Averaging" ?

It is a procedure that differentiates between synchronous and
non-synchronous frequencies. When applied properly, it is a
powerful tool.
 The analyzer and photoeye are set up as outlined on the
previous page.
 The analyzer is instructed to collect a spectrum. The
specifics on how to perform this are different for each box
and exact instruction will be left to the various
manufacturers.
 The analyzer will collect a spectrum that contains ONLY
synchronous vibration. That is, only vibration frequencies
that are exact harmonics of the trigger rate (usually 1x
rpm).
 Any non-synchronous frequencies - no matter how close
they are to being synchronous - are filtered out and
disappear from the spectrum. This allows for comparison
with a normally collected spectrum and subsequent
analysis of whether a particular frequency is coming from
the rotor or some other (non-synchronous) source.
For more information on this powerful troubleshooting tool,

see the 'Field Tests' manual.
End Of Phase Section:

End Of Vibration Characteristics
You have now completed the "Vibration Characteristics"
section of our training. If you feel satisfied with your
understanding of the subjects, you should now move on to:
 How vibration is measured and how a database should be
set up, or
 Plots: The trend plot, the spectrum plot, the "envelope"

plot and time domain plot.
Acquiring and Displaying Data

How Data is Displayed:
The Trend Plot
Y-Axis Data:
Amplitude
X-Axis Data:
Time (typically days or months)
 A 'Trend' plot is simply a number of amplitude values -

snapshots of the total vibration (vibration at all
frequencies) - over a period of time.
 The interval between readings will be the time elapsed
between those readings. That time interval could be
anything from months to milliseconds depending on the
specifics of the vibration program and system(s)
involved.
 A trend plot offers limited analysis tools (there is no

identification of specific frequencies, for instance) but can
be an important indicator of developing problems.

Time Domain Plots
Y-Axis
Units:
Amplitude
X-Axis Units:
Time (seconds or milliseconds)
The first process the data collected is put through is to convert

what is an analog signal (the transducer moving with the
bearing) to a digital signal - seen here. This is a "time domain"
plot. Typically, the length of a time domain plot will be very
short - commonly in milliseconds. It is common to want to
capture 5-7 revolutions of a shaft. To capture 5 revolutions of
a shaft running (for instance) 3000 rpm (50Hz, or revolutions
per second), you would need 5/50 = 0.1 seconds = 100msecs.
On that plot, you should be able to see 5 sine waves for 5
revolutions of the shaft plus any wave shape distortion (is it a
perfect sine wave or an unusual shape) plus any higher
frequencies that might be occuring (electrical, bearings,
gears).
 A 'Time Domain' plot displays amplitude vs. time.
 However, unlike a trend plot, the amplitude is a
continuous representation of the amplitude value.
 For instance, if the amplitude unit for the above plot were
displacement, the line would represent the actual bearing
location as it moves back and forth.
 Also unlike a trend plot, the values can be negative or
positive since, for instance, the displacement can be on
either side of a neutral, or 'at-rest' position, and velocity
or acceleration amplitudes can be in one direction or the
other (defined as the '+' and '-' directions depending on
the direction the transducer is pointing).
The time domain is more difficult to analyze than the next plot
we will discuss - the "Spectrum" - but under certain conditions
it can provide insights and information not available on the
spectrum plot.

The "FFT" Spectrum Plot
Y-Axis
Units:
Amplitude
X-Axis Units:
Frequency (cpm or Hz)
A "Spectrum" is plot of amplitude vs. frequency. The above

plot is a spectrum that was created from a time domain plot
using a mathematical principle called the "Fast Fourier
Transform", or "FFT". This plot is often simply referred to as an
"FFT".
This principle states that any periodic signal (what we
measure with vibration) can be broken down into a series of
simple sinusoids that, when combined, will generate the
periodic signal we have just analyzed. In practical terms that
means this process can generate the spectrum we see here
from a time domain signal it has analyzed. By plotting
amplitude versus frequency (instead of time), it becomes far
easier to analyze. By relying on complex mathematical
processes, however, it also becomes susceptible to generating
what can be misleading information. The plot displays a certain
number of amplitude values (400, 800, 1600, etc.) over a
range of frequencies. The plot seen here tells the analyst that
there is:
 'A' amplitude at a frequency of approximately 3534 cpm
(58.9 Hz)
 'B' amplitude at approximately 7084 cpm (118.07 Hz)
 'C' amplitude at approximately 10,633 cpm (177.22 Hz)

and so on.
This plot is the most commonly used analysis tool since, by
enabling frequency identification, it allows for preliminary
identification of the source of the vibration.

The "Envelope" Spectrum Plot
Y-Axis
Units:
Amplitude
X-Axis Units:
 An 'Enveloping Spectrum' plot is identical to a
conventional spectrum in the way it displays data:
amplitude vs. frequency.
 However, there is a significant difference in the way the
raw data is processed.
 Whereas conventional FFT processing extracts simple
sinusoids from complex motion, enveloping spectrum
signal processing looks for transient (instantaneous)
impacts - the "striking" action discussed earlier versus
the "pushing" action the FFT works well with.
 These impacts do not usually survive the conventional
FFT process and therefore are kept hidden from the
analyst.
 However, knowledge of the frequency of these impacts
provides extremely important information towards proper
diagnosis of machinery problems at an earlier stage than
would be possible with conventional FFT analysis.
 This additional analysis tool is particularly useful in

diagnosing bearing faults.
This plot, which is a spectrum in its appearance, is extremely
useful in the identification of rolling element bearing defects
and a number of other impact generating vibration problems.
How Are 'Trend' Amplitudes Determined ?
Practical Definition:
 The 'Overall' Amplitude is the sum of all of the vibration
energy occurring between 0 cpm and the data collector's
maximum frequency (its "Fmax" - well over 1,000kcpm).
In other words, if you were to collect a spectrum over a
frequency range of 0 - collector's Fmax and added all of
the amplitude peaks together, that would be your
'overall' amplitude. Although not technically correct, that
is one way to look at it.
Technical Definition:
 Consider again collecting a spectrum from 0 - the
collector's Fmax. Take every amplitude value, square it,
add them together and take the square root of that sum.
Although not exactly perfectly accurate, that is closer to
what actually goes into calculating an 'overall' amplitude.
That is known as a "Root Mean Square", or RMS, value.
 All data is collected as an "RMS" value. That is an ISO
convention and applies no matter what collector you may
have.
 The RMS value was settled on in order to guard against a

transient spike in the signal distorting the true value
significantly.
But there are several different 'signal detection' units available
- RMS, peak and peak to peak. How exactly are they related
and calculated ?
The Difference Between RMS, Peak and Peak

to Peak Amplitudes
Again, all data is collected as an "RMS" value ("Root Mean Square").
That is an ISO convention and applies no matter what
manufacturer's collector you may have.
The "RMS" value is calculated by simply multiplying the peak
amplitude (shown in the graphic below) by 0.707:
RMS = Peak x 0.707
But is that technically

correct ?
o Technically,
the RMS value of
a pure sinusoid is
equal to the area
under the half-
wave.
o That
corresponds to
(peak x 0.707)
on a pure
sinusoid.
o Pure
sinusoids are
rarely
encountered in
the world of
mechanical
vibration.
o Due to that,
the trend
amplitude value
is calculated by
simply
multiplying the
peak amplitude
by 0.707.
But some analysts prefer to display the amplitudes as a 'peak'
amplitude (usually because the vibration severity charts they are
using are in peak amplitude values). So how is that value
calculated ?
 To calculate the 'Peak' amplitude, we divide the RMS value by
0.707.
 For those of you paying close attention, you may have noticed
that, from start to finish, the way a peak amplitude is
calculated is:
o Measure the peak amplitude.
o Multiply by 0.707 to obtain the RMS value.
o Divide by 0.707 to obtain the Peak value.
In some cases - specifically the use of displacement amplitudes

where the analyst wants the total amount of bearing movement back
and forth - the 'Peak to Peak' value is desired by the analyst. In that
case, the peak value is simply multiplied by 2. Velocity and
acceleration units are not typically displayed as peak to peak since
the important piece of information is what the maximum (peak)
value is.
So let's summarize:
 The highest (peak) value of collected data is multiplied by
0.707 to obtain the ISO standard of RMS amplitude. The
majority of programs world-wide use this value.
 For those programs that wish to display the data as a 'peak'
value, the RMS value is then divided by 0.707 to obtain the
peak amplitude.
 For those situations where peak to peak amplitudes are desired,

the peak amplitude is simply multiplied by 2.
Summary: It doesn't matter which value you use - RMS or Peak - so
long as you are CONSISTENT (another very important convention) !!
Amplitude is, after all, simply a number from which we make certain
generalizations about the machine condition. There is no exact,
precise number above which disaster awaits and below which you are
safe.
Be aware of the conversion if comparing values between programs
that use RMS and Peak values. This applies to spectrum plots as
well.
Now let's look at the vibration sensors - transducers.
Transducers
A vibration transducer is the instrument that measures or
senses the vibration and is commonly referred to as a pickup
or sensor. The basic understanding and proper selection of an
appropriate transducer is important. This text will not deal in
extensive detail with all the different types and technical
characteristics of transducers but will attempt only to provide
a basic understanding of the important aspects. There are
basically three types of transducers commonly used. They
are:
 Accelerometer
 Velocity Transducer
 Displacement Transducer
Transducers:
Accelerometers
Accelerometers are by far the most common and versatile
types of transducers in use. The seismic, or piezoelectric,
accelerometer produces a output charge when held against a
vibrating surface due to the characteristics of the piezoelectric
disks that are inside the transducer. These transducers
contain no moving parts and as such are quite rugged. The
charge that is generated through the compression and
expansion of the piezoelectric disks is proportional to the
amount of vibration acceleration (force).
 Generated voltage is very small -
most models have built in hi-gain
electronic amplifiers.
 Output charge is proportional
acceleration force accelerometer is
exposed to.
 Accelerometers are the only
transducer capable of reliably
measuring high frequency vibration
that is related to problems such as
bearing defects and gear problems.
 Accelerometers have fair to poor
accuracy at low frequencies where
acceleration amplitudes can become
One Design of a
quite small even in the presence of Piezoelectric
high displacement amplitudes. Accelerometer
 Some are somewhat susceptible to
radio frequency interference
 Some can have problems with
thermal growth (putting the
transducer on a very hot bearing)
which can cause false amplitude
values at low frequencies unless
they are given time to thermally
stabilize.
 Transducer should have sensitivity

of 100 mv/G or higher. The more
mv/G the accel generates, the more
accurate it will be in the low
frequency (< 120 cpm) roll-off
ranges.
Recent years have seen the development of affordable
accelerometers that are reasonably accurate to frequencies
even well below 120 cpm. Low frequency will remain,
however, a weakness of accelerometers due to the nature of
acceleration amplitudes at low frequencies. Accelerometers
are also useful because they are much more compact and
lightweight than velocity transducers and therefore can be
used in more environments and applications. They are also
not as susceptible to magnetic fields generated by electric
motors or other magnetic sources as velocity transducers.
A final advantage of using an accelerometer is the ability to
integrate the results in order to obtain velocity readings and
double integrate the results in order to obtain displacement
readings. This ability means that a single transducer can be
used to register and quantify all three of the primary
amplitude units of acceleration, velocity and displacement
over a wide range of common frequencies.
Transducers:
Velocity
There are two primary types of velocity transducers, the
moving coil type and the piezoelectric type.
Moving Coil Type -
 Internally is made up
of a coil supported by
springs and a
permanently fixed
magnet.
 Coil movement
relative to the
magnet generates a
voltage proportional
to the velocity of that
movement. Moving Coil Velocity Transducer
 Generally require no
output signal
amplifiers.
 They are larger than other types of transducers.
 They are not reliable for low frequency measurements
without the use of a correction factor because the coil
tends to follow the magnet at low frequencies.
 Balancing at lower frequencies ok since the relative
amplitudes recorded will be consistent.
 Moving coil type velocity transducers are also affected by

magnetic fields that may be present around electric
motors or other magnetic sources and may require
magnetic shielding in order to be used in such an
environment.
Piezoelectric Velocity Transducer -
 Similar to the piezoelectric accelerometer in design with
the obvious difference of the output charge being
proportional to the velocity of the transducer.
 The charge is created by the piezoelectric disks
responding to the stress of whatever applied forces there
may be.
 Generally require a built-in amplifier due to the extremely
small signal generated.
 Piezoelectric velocity transducers are generally not
affected by magnetic fields and can also measure
accurately down to 60 cpm or less.
 There are thermal sensitivities and ambient temperature

limitations involved.
Transducers:
Displacement
Generally known as
proximity probes, these
transducers are limited in
application but where
applicable they are
extremely useful and
important. They are
generally used where a Proximity Probe
smooth running rotor is
very critical. For instance,
turbines in power plants
will have permanently
mounted proximity probes
monitoring turbine
movement and possibly
phase. High speed
compressor rotors are
another example of a
machine that will almost
always be monitored with
proximity probes.
 Does not actually touch the shaft or bearing - it is
mounted an exact distance from a clean, smooth area of
the shaft and fixed in place.
 Externally applied high-frequency carrier signal emits a
magnetic field from a coil in the tip of the probe towards
the shaft.
 The shaft absorbs some of the magnetic energy present.
 The closer the shaft to the magnetic field source, the
more it absorbs. The more magnetic energy it absorbs,
the more it uses up the carrier signal and reduces the
carrier signal strength.
 The further away the shaft is from the probe, the less it
absorbs. The less magnetic energy it absorbs, the less it
uses up the carrier signal and the higher the carrier
signal strength remains.
 The probe produces 2 signals:
o AC signal proportional to the vibration (shaft
movement).
o DC signal proportional to the size of the gap.
Generally, the applications are:

 Relatively lightweight rotors in massive housings and
rigid (generally “sleeve” or “babbit”) bearings.
 High speed rotors on machines such as turbines,
centrifugal compressors and pumps (where shaft speeds
can exceed 50,000 rpm).
 Very slow speed machines on which it is difficult to obtain
reliable data with velocity transducers or accelerometers.
Machines such as these can fail catastrophically in an

extremely short period of time (hours, minutes or even
seconds) from the initial onset of the problem and on-line
monitoring with permanently mounted proximity probes is
essential to both machine and human safety.
So you've got a collector, transducers and you understand the
basics. How do you set up a database ?
Database Setup
Database Setup
Introduction
A crucial step for any monitoring program is, unfortunately,
right at the beginning - database setup. This is unfortunate
because it is a crucial task that is performed when the people
doing it are most likely inexperienced and of limited education
in the field. Once created, it is very difficult to develop new
ideas or drastically change things without creating a colossal
mess.
The parameters selected and programmed into the PC-based
database will be fed into the data collector and will determine
the characteristics and usefulness of the data you collect. Most
data collectors also have the capability of allowing the user to
program in many of the most important parameters on the
spot while performing an in-depth analysis. That, however, is a
separate function from what has been loaded in from the
database and can be extremely cumbersome and time
consuming compared to having the information pre-
programmed into the data collector and merely having to walk
the route stopping at collection points and pressing store on
the collector.
A monitoring program tracks the condition of your equipment
through the measurement of vibration amplitudes on a regular
interval. Deviation from the norm then triggers further
investigation to determine the source and correct the actual
cause of the excessive vibration.
Database Setup:
Purpose of a Monitoring Program
What is the purpose of a vibration monitoring program ? Well,
the operative word is “monitoring”. The purpose is to monitor
the vibration levels. This means collecting data that will alert
you to any developing problems while not over-burdening you
with collecting overly detailed or excessive amounts of data. A
monitoring program is not intended nor is able to do is allow
for specific diagnose of every problem that arises while sitting
behind your computer.
To be sure, a well designed vibration monitoring program will
nearly always give you a specific direction in which to search
for a problem and it will help you prioritize work by gauging
the severity and source of most of the problems you will
encounter. However, further analysis, or at the very least
confirmation checks, are always a wise course of action to take
before proceeding with any corrective actions.
Database Setup:
Selecting Point Parameters
As an example let us look at a simple AC motor. The details
surrounding each of the frequencies we have to monitor are
found in the "Troubleshooting Charts" (accessed from the main
menu). The question we will address is - what are the different
potential problems we need to be looking for:
Frequency(s) to
Explanation
monitor
1x, 2x, 3x rpm Running speed harmonics
2x AC Line Frequency Torque Pulse Frequency (variation in motor
air gap)
1x, 2x, 3x Rotor Bar Pass Electrically related frequencies equal to the
Frequency (RBPF) number of rotor bars or winding
or Winding Slot Pass slots x rpm. These can reveal certain
Frequency (WSPF) electrical problems and/or conditions.
30kcpm - 120kcpm Frequency range in which symptoms of
rolling element bearing defects usually show
up initially.
So the questions are - what are the Fmax's, how many lines of
resolution should be used, what amplitude units should we
use, what is the best way to analyze the data and basically,
how can we make our database work for us as efficiently as
possible ? How can the need for occasional high resolution
readings be squared with the need to monitor relatively high
frequencies ? Accuracy vs. speed of data collection.
Database Setup:
Selecting Point Parameters
The solution, of course, is that several different measurements
must be taken at each position with each one addressing one
or a few specific problem areas (and hence a specific
frequency range).
Mechanically Caused Vibrations
 A complete protection set-up would collect a reading
horizontally and verticallyon each bearing plus one axial
reading on position 2 with an Fmax of 120000 cpm in
order to monitor potential bearing defects developing and
the common problems occurring at 1x, 2x and 3x rpm.
That's five readings.
Electrically Caused Vibrations

 In addition to those two readings, you may wish to take a
high resolution reading on one motor bearing (typically a
12000 cpm Fmax with up to 1600 lines of resolution).
This reading would separate the electrical and mechanical
vibrations at the low frequency end.
 Finally, you may want to collect a reading with an Fmax
of 720000 cpm to look for the high frequency electrical
problems (loose windings, loose or cracked rotor bars).
The reason the word may is used is because electrical
problems are relatively uncommon and extra electrical
readings can be collected separately if a problem is
suspected. A full set of baseline readings should be
collected in any event to initially check for any such
problems. The only one of these readings that should not
be collected with velocity amplitude units is the 720000
cpm Fmax reading which should utilize acceleration units.
 That's two more readings.
Bearing Protection:
 An enveloping spectrum should also be taken on each
bearing to check for any frequencies at which impacts are
occurring. That's two more readings. What you are left
with is this:
However, let's get back to the word monitoring - as in a
monitoring program. You could, depending on the proactivity
of the program personnel, eliminate a few of the readings and
rely on the others to tip you off that something is wrong. You
could then go into full-blown anaysis mode and collect lots of
data. Using that logic, the readings could be divided into
"necessary" and "optional" readings as follows:
Necessary
Readings Max Freq # of Lines Units Direction Position

1 120kcpm 400 Velocity horiz Brg 1
2 60kcpm 400 Enveloping horiz Brg 1
4 60kcpm 400 Enveloping horiz Brg 2
5 60kcpm 400 Velocity axial Brg 2
The necessary readings shown here monitor for early stage
bearing defects as well as providing general information on
machine health. In-depth analysis may be difficult, depending
on the specific problem, and a problem developing may
require more and better data be collected.
Optional
Readings Max Freq # of Lines Units Direction Position
6 720kcpm 1600 Acceleration horiz Brg 1
7 120kcpm 400 Velocity vertical Brg 1
9 120kcpm 400 Velocity vertical Brg 2
The optional readings shown here monitor in additional planes
(vertical) and for more specific problems with high frequency
(720kcpm, 1600 lines) and high resolution (12kcpm, 1600
lines) spectra.
Database Setup:
Further Example of Selecting Point Parameters
Let's look at a component other than the motor - a screw
compressor (note that we have already discussed what is
needed to adequately monitor the motor). You would need:
 To collect a reading
with an Fmax of
120000 cpm on each
of the four bearings in
order to monitor
potential bearing
defects developing and
the common problems
occurring at 1x, 2x and
3x rpm and
aerodynamic (# lobes
x rpm) pressure-pulse
related frequencies.
 An enveloping
spectrum to help in
analyzing any impact
generating
frequencies. Screw Compressor:
48 x RPM = Gear Mesh Frequency
48/36 Gear Ratio = 1.33 Speed Increase
 By calculating the gear
mesh frequency at 48 4 x 1.33 = Lobe Pass Freq.
teeth x rpm, you (Aerodynamic)
would want to collect a
reading with an Fmax
of 3x gear mesh
frequency
(3x48=144), or
approximately 150x
rpm to look for the
high frequency, gear-
related problems.
Necessary
Readings Max Freq #Lines Units Direction Position

All
1 120kcpm 800 Velocity Horizontal compressor
bearings
All
2 60kcpm 800 Enveloping Horizontal compressor
bearings
All
150 x
3 1600 Acceleration Horizontal compressor
RPM
bearings
Note that there are no low Fmax readings being taken (lowest
Fmax is 60kcpm). The simple reason for this is that none are
necessary. There are no frequencies so close as to be
inseparable on a normal spectrum. The expected dominant
frequencies on the spectrum will be related to the aerodynamic
pulse frequency and the motor rpm. The use of these three
readings on each bearing (12 readings on the compressor side
alone) will protect against occurrence of a unforeseen problem
-assuming, of course, proper interpretation of the data as well
as proper operation and general maintenance of the machine.
Note that implementing a global default change from 400 to
800 lines is a good technical change with practical advantages
and, with the speed of processors and amount of memory
available, there are minimal downsides.
Database Setup:
How Much Data is Enough ?
So you see that a simple AC motor may require as many as
nine separate spectra in order to catch all the potential
problems that may develop on it. The driven component as
many or more. How do you collect all of this data ? More
importantly, how do you analyze all the data you've collected
in an efficient yet productive way ? That is the main problem.
Of course, some problems are much more likely to develop
than others and a reduction in the number of readings may be
unavoidable in order to decrease data collection and analysis
time.
As with many other maintenance or management decisions,
trade-offs between time and accuracy of information must be
made.
 Good news: The more readings you have, the more
information you will have, the more protection you will
have, etc.
 Bad news: The more readings you take, the longer your
data collections will take, the longer your analyses will
take and the more paperwork you will have.
Once again, let’s return to the basic rules governing the

purpose of a monitoring program. Although it is possible, very
few types of problems will show up in only the vertical
direction without also affecting the horizontal readings. So we
could eliminate some of the vertical readings with a slight
decrease in the protection level. Note that only one reading is
taken axially. Typically, bearing 2 is where the axial reading is
taken due to easier access but that is not always the case.
So how much data should your program collect ?
The answer is actually fairly simple:
 It depends on how proactive your maintenance
department is relative to a PdM program.
 In other words, the faster and more efficiently you
investigate and solve problems, the less data you can get
by collecting.
I've personally seen tremendously successful programs that

collect a bare minimum of data periodically - trend (overall)
values and spectrums on overall alarm only. The secret to
their success is simple - the moment something goes into a
(well set up) statistical alarm, the collector is triggered to
collect a spectrum. When the data is uploaded, they analyze
the spectrum (no spectrums are collected that are not in
alarm) and immediately send someone out to investigate and
correct as necessary. Average plant-wide vibration around
0.16 in/sec (3.9 mm/s) - not bad for a plant that runs 25+
routes a week and has two full-time analysts.
Database Setup:
Trading Time Versus Accuracy
There is one good arguument against bothering to collect the
two special electrical readings:
 Electrical problems are often apparent through means
other than a velocity spectrum. Many problems, including
high 2x line frequency and broken/cracked rotor bar
problems, will almost always cause an audible, pulsing
hum in the motor that anyone can notice without
vibration equipment.
Can we therefore collect these two readings only when a

problem is suspected and not every data collection ?
 For a purely monitoring purpose, we would then be down
to only five readings, reducing data collection time by
more than 50% since we would be eliminating the most
time consuming of these readings.
 To take this reduction to an even further extreme, we
could collect only overall values for each of these points
and trend amplitude changes (increases) only, collecting
spectra only when the trends indicate it is necessary.
What data should be collected on this motor ?

 10 spectra and overalls, or . . .
 5 overalls ONLY ?
These are decisions that must be made by the people in

charge of executing the program. They should be based on a
number of criteria including machine criticality, manpower
availability, instrumentation used (speed and capability) and
available software. Each reduction in the amount of readings
taken increases the potential for an unforeseen failure. Any
use of vibration analysis, however, is an improvement over
reactive maintenance, which merely reacts to breakdowns or
imminent problems.
How does Vibe-Assist address this dilemma ?
Database Setup:
An Example Of An Effective Database Setup
The exact nature of your database setup and the specifics
must be addressed according to the vendor you purchased
your software from. Most programs fail because usable,
worthwhile data can not be extracted from the database. Why
does this happen ? The software usually has certain useful
features such as reporting capabilities that can be accessed - if
the database is created in such a way as to not only take
advantage but to make the best use of those features.
By way of a single example - and their are numerous ways to
set up a database - the service company Vibe-Assist looked at
two ingredients that went into the collection and analysis of
data - what information they wanted vs. what the reporting
capability of the software was (Entek's Odyssey™) - and came
up with database setup templates that are used for creating an
effective database structure that uses software features of
Odyssey to provide information we want. By generating a
database structure that takes maximum advantage of a
powerful software reporting feature, Vibe-Assist reduced their
analysis time by an incredible 80% or so. This new database
structure did not generate this huge improvement by reducing
machinery protection or analysis accuracy. On the contrary,
the new setup improved the reliability of the analysis and
improved the level of protection possessed by the machines
they monitor. In order to illustrate the value of an effective
database setup, Vibe-Assist has agreed to share a few of their
component setups along with an explanation of the logic
behind the database structure. Shown below are examples of a
direct drive, 4-bearing machine and a belt-drive, 4-bearing
machine. The links below under the "Type" of reading will take
you to explanation pages for each of the readings.
 Direct Drives
 Belt Drives
 Explanation of Readings
 Sample Report
Direct
Drive
Bearing 1
# of
Reading Type Max Freq Units Direction
Lines
Trend &
1 30kcpm 800 velocity horiz
Spectrum
120kcpm
Trend &
2 - or - 800 acceleration horiz
Spectrum
60kcpm*
Spectrum 20x rpm

3 800 envelope horiz
Only up to 60kcpm
5-9 revolutions acceleration
2048
4 Time Domain or horiz
bytes
relevant period or velocity
Trend &
5 240x rpm 800 acceleration horiz
Spectrum
* - For the acceleration spectrum, use 120kcpm for machine speeds over
1200 rpm and 60kcpm below that.
Direct
Drive
Bearing 2
# of
Lines
Trend &
Spectrum
120kcpm
Trend &
Spectrum
60kcpm*
Spectrum 20x rpm

Only up to 60kcpm

2048
bytes
Direct
Drive
Bearings 3
&4
# of
Lines
Trend &
Spectrum
120kcpm
Trend &
Spectrum
60kcpm*
Spectrum 20x rpm

Only up to 60kcpm
5-9
revolutions or acceleration
2048
4 Time Domain horiz
bytes
relevant or velocity
period
Top
Belt Drives
Explanation of Readings
Sample Report
Belt Drive
Bearing
#1
# of
Lines
Trend &
Spectrum
120kcpm
Trend &
Spectrum
60kcpm*
Spectrum 20x rpm

Only up to 60kcpm

2048
bytes
Trend &
5 240x rpm 800 acceleration horiz
Spectrum
Belt Drive
Bearing
#2
# of
Lines
Trend &
Spectrum
120kcpm
Trend &
Spectrum
60kcpm*
Spectrum 20x rpm

Only up to 60kcpm

2048
bytes
Belt Drive
Bearings 3
&4
# of
Lines
Trend &
Spectrum
120kcpm
Trend &
Spectrum
60kcpm*
Spectrum 20x rpm

Only up to 60kcpm
5-9
revolutions or acceleration
2048
4 Time Domain horiz
bytes
relevant or velocity
period
Top
Direct Drives
Belt Drives
Sample Report
There are numerous other component setups in the templates -
these are for generic, rolling element bearing machines running at
normal speed (1000 - 3600 rpm). But this seems like an extreme
amount of data - how does this help with analysis ? Well, first, each
reading has a specific job to do:
 Reading 1: Velocity Trend - Tool that is sensitive to fluctuation
(increase) in low - mid frequency ranges where mechanical
problems develop (1x - 5x rpm) and, to a lesser degree,
higher frequency problems such as rolling element bearings.
 Reading 1: Velocity Spectrum - Analysis tool for low - mid
frequency problems. Higher resolution readings collected on
belt drives and inboard motor bearings.
o Note: By properly setting some simple statistical alarms
on the trend plot, this spectrum may be turned off
because the analyst will be alerted by the overall alarm
to any situation where a spectrum is required and
eliminate many of these most time consuming readings.
 Reading 2: Acceleration Trend - Tool that is not affected by
influences at 1x - 5x rpm but is increasingly sensitive to
problems developing above 30kcpm such as gears, bearings
and high frequency electrical. Another good candidate for
statistical alarms.
 Reading 2: Acceleration Spectrum - Analysis tool for high
frequency problem detection and analysis. Used in conjunction
with the next reading . . .
 Reading 3: Envelope Spectrum - e.g. gSE, ESP, Peakvue, HFB
and more. This plot gives information on transient impacts
occurring that may be related to a bearing defect or several
other sources.
 Reading 4: Time Domain - Can be de-activated if the analyst
prefers but is an important tool to use at times. Can be set to
whatever time sample the analyst chooses.
 Reading 5: Acceleration Spectrum - On motors only, looks for
2x and 3x rotor bar pass frequency and winding slot pass
frequency - very high frequencies.
So what are we left with in the best of environments ? By turning

off the time consuming velocity spectrum and rely on trends only to
monitor the low-frequency end of the spectrum, we are left with:
 Velocity overall
 Acceleration overall
 Acceleration spectrum
 An enveloping spectrum
That is a total of about 5 seconds of data collection. If the velocity

trends into alarm, we turn on the velocity spectrum for analysis. If
an acceleration overall trends into alarm, we check the enveloping
spectrum for impact frequencies and try to match up harmonics on
the acceleration spectrum. This database structure protects against
all possible trendable problems (as opposed to "event-based"
problems that can lead to short-duration failures).
Top
Direct Drives
Belt Drives
Sample Report
The key, however, is in the simple reports that can be run. An
amplitude threshold is set for the report (perhaps 0.2 in/sec or 5
mm/sec). A line is triggered for the report for every single peak
that exceeds the pre-set amplitude threshold. The line on the
report, shown below, includes all the location information, the
Fmax, peak amplitude and frequency of the peak as well as the rpm
and date. This makes initial review of the data simple and brief.
This report addresses the low-mid frequency ranges that require
velocity units for effective monitoring (> 30kcpm). This is the
frequency range in which mechanical problems will show up: 1x -
10x rpm or so. This report does not address bearings, gears,
certain electrical vibrations or any other high frequency vibration
sources.
To address the high frequencies, an identical report is created to

handle the acceleration spectral data. The amplitude trigger for
acceleration units will typically be about 1G - a very safe, fairly low
amplitude threshold. Any line on the acceleration report is
investigated by first looking for any impact frequencies on the
associated envelope spectrum and then by making an assessment
of the vibration source. Further investigation may be necessary but
the report(s) give easily accessed, easily analyzed information that
prompts immediate investigation in the problem areas. Much of the
data does not get looked at but that is simply because the vibration
levels too low to be concerned with.
Database Setup:
Summary
The creation of a database, with all the associated decisions
that must be made, is a job that requires a very good technical
understanding of vibration analysis, machinery operational
characteristics and good doses of common sense. Always
question what you are doing, especially whether or not the
desired goals are being achieved and, if not, what other
methods might be used to achieve those goals.
The database setup shown on the previous page that Vibe-
Assist is successful with was not presented as an example of
the only way a database can be setup. It was included only as
an example of a setup that is known to provide high levels of
protection while being simple to manage and easy to analyze -
three very important aspects. There are certainly other
database structures that are successful. No matter the
database setup, the individuals involved are of crucial
importance; the machines, the maintenance environment and
numerous other factors also affect the success. However, the
importance of the task of setting up a database cannot be
overstated. Although doing it properly does not guarantee
success, doing it improperly guarantees failure. There are
many variables to consider and options to choose from.
The particular database setup we've just gone through utilizes
a strength of the Odyssey™ software - the ability to generate
a report with the specific information we wanted. Your
software may or may not have that ability. You may need or
want to focus on some other strength of your software (since
it is the key determining factor in what database setup will
work for you). The database is comprehensive yet manageable
and scaleable to the user's needs. If you have Odyssey™
software and are struggling with your database setup, feel free
to copy the database setup laid out above and customize the
report as you saw on the previous page. You can also, for a
small fee, obtain the templates (including the different
components, a written manual the reports, etc.) directly from
Vibe-Assist.
Plots:
Trends
What Is A "Trend" Plot ?

Y-Axis
Units:
Amplitude
X-Axis Units:
Time (typically days or months)
 A 'Trend' plot is simply a number of amplitude values -

snapshots of the total vibration (vibration at all
frequencies) - over a period of time.
 The interval between readings will be the time elapsed
between those readings. That time interval could be
anything from months to milliseconds depending on the
specifics of the vibration program and system(s)
involved.
 A trend plot offers limited analysis tools (there is no

identification of specific frequencies, for instance) but can
be an important indicator of developing problems.
How Are 'Trend' Amplitudes Determined ?
Practical Definition:
 The 'Overall' Amplitude is the sum of all of the vibration
energy occurring between 0 cpm and the data collector's
maximum frequency (its "Fmax" - well over 1,000kcpm).
In other words, if you were to collect a spectrum over a
frequency range of 0 - collector's Fmax and added all of
the amplitude peaks together, that would be your
'overall' amplitude. Although not technically correct, that
is one way to look at it.
Technical Definition:
 Consider again collecting a spectrum from 0 - the
collector's Fmax. Take every amplitude value, square it,
add them together and take the square root of that sum.
Although not exactly perfectly accurate, that is closer to
what actually goes into calculating an 'overall' amplitude.
That is known as a "Root Mean Square", or RMS, value.
 All data is collected as an "RMS" value. That is an ISO
convention and applies no matter what collector you may
have.
 The RMS value was settled on in order to guard against a

transient spike in the signal distorting the true value
significantly.
But there are several different 'signal detection' units available -
RMS, peak and peak to peak. How exactly are they related and
calculated ?
The Difference Between RMS, Peak and Peak to
Peak Amplitudes
Again, all data is collected as an "RMS" value ("Root Mean Square").
That is an ISO convention and applies no matter what
manufacturer's collector you may have.
The "RMS" value is calculated by simply multiplying the peak
amplitude (shown in the graphic below) by 0.707:
RMS = Peak x 0.707
But is that technically

correct ?
o Technically,
the RMS value of
a pure sinusoid is
equal to the area
under the half-
wave.
o That
corresponds to
(peak x 0.707)
on a pure
sinusoid.
o Pure
sinusoids are
rarely
encountered in
the world of
mechanical
vibration.
o Due to that,
the trend
amplitude value
is calculated by
simply
multiplying the
peak amplitude
by 0.707.
But some analysts prefer to display the amplitudes as a 'peak'
amplitude (usually because the vibration severity charts they are
using are in peak amplitude values). So how is that value
calculated ?
 To calculate the 'Peak' amplitude, we divide the RMS value by
0.707.
 For those of you paying close attention, you may have noticed
that, from start to finish, the way a peak amplitude is
calculated is:
o Measure the peak amplitude.
o Multiply by 0.707 to obtain the RMS value.
o Divide by 0.707 to obtain the Peak value.
In some cases - specifically the use of displacement amplitudes

where the analyst wants the total amount of bearing movement back
and forth - the 'Peak to Peak' value is desired by the analyst. In that
case, the peak value is simply multiplied by 2. Velocity and
acceleration units are not typically displayed as peak to peak since
the important piece of information is what the maximum (peak)
value is.
So let's summarize:
 The highest (peak) value of collected data is multiplied by
0.707 to obtain the ISO standard of RMS amplitude. The
majority of programs world-wide use this value.
 For those programs that wish to display the data as a 'peak'
value, the RMS value is then divided by 0.707 to obtain the
peak amplitude.
 For those situations where peak to peak amplitudes are desired,

the peak amplitude is simply multiplied by 2.
Summary: It doesn't matter which value you use - RMS or Peak - so
long as you are CONSISTENT (another very important convention) !!
Amplitude is, after all, simply a number from which we make certain
generalizations about the machine condition. There is no exact,
precise number above which disaster awaits and below which you are
safe.
Be aware of the conversion if comparing values between programs
that use RMS and Peak values. This applies to spectrum plots as
well.
Now let's look at the vibration sensors - transducers.
Analyzing A Trend
Analyzing a trend is not typically a difficult thing to do . . .
providing you know what it is you are trending.
 Are you trending high frequency vibration ?
 Not if you only use displacement.
 Are you trending the typical, general equipment speed
ranges of 1000 - 3600 rpm and associated mechanical
problems.
 Not with displacement or acceleration - only with
velocity.
 Are you trending very low frequency vibration - below
100 rpm, for instance ?
 Only with displacement and, if the frequency is low

enough, only by changing type of transducer.
With vibration, a trend headed in the wrong direction almost
certainly means it is increasing. With a properly set up
program, properly collected data and proactive personnel, a
program can be successful with collecting little more than
trend values. For whatever reason, however, that is rarely the
case.
In any event, trending can be useful but it is not usually
enough to make specific judgements. A determination of the
exact frequencies involved is usually required and should be
attempted in any event to avoid a possibly embarassing and
expensive mistake. A determination of frequencies can be
made two separate ways - viewing the data as a raw time
domain plot or putting it through an FFT process to generate a
spectrum. An overview of the process of analyzing a spectrum
will now be presented followed by an in-depth look at how, in
a practtical sense, the FFT process works. Understanding that
will help you view the spectrum with an appropriate amount of
respect and skepticism.
Plots:
The FFT (The Spectrum)
What Is An FFT (Spectrum) Plot
Y-Axis
Units:
Amplitude
X-Axis Units:
Frequency (number of cycles per minute or per second)
 Whereas a Trend is amplitude values versus time, a
"Spectrum" Plot is amplitude versus frequency.
 A spectrum, a.k.a. an "FFT", allows you to assess
severity (with the amplitude) and helps identify the
source (with the frequency).
 This is the most commonly used analysis tool and is

usually sufficient protection for general speed machinery.
FFT Tools
Vibration spectra provides important clues to machine
problems. There are tools provided in all software packages
that help with this analysis. The most important are:
 Moveable Cursor - A "base" cursor that can be moved
to any frequency and identifies the amplitude at that
frequency.
 Harmonic Cursors - Activating this tool creates
additional cursors (as many as are required) that
appear at integer multiples of the base cursor. If the
base cursor is located at 1x rpm, harmonics will appear
at 2x, 3x, 4x, etc. This is the most important analysis
tool available. All harmonic cursors are at higher
frequencies than the base cursor.
 Sideband Cursors - Activating this tool creates
additional cursors at frequencies to either side of the
base cursor. If the 1st (closest) sideband cursors are
located 50 cpm to either side of the base cursor,
additional sideband cursors (as many as required) will
each be located an additional 50 cpm away. For
instance, the 2nd sideband cursors will be 100 cpm
away from the base cursor, the 3rd will be 150 cpm
away, etc.
These three tools are sufficient for the beginner and for 99%
of most analyses. There are, of course, other useful tools
such as " "labels" and frequency groups but that is more
vendor specific and should be learned in a software class and
through experience.
 Remember, ALL spectral analysis is limited in its use
and accuracy by the spectrum resolution.
How An "FFT" Plot Is Created ?

First, the vibration is "sampled" (collected) over a pre-
determined period of time. The period of time used for the
sample will be based on parameters programmed into either
the database (for interval-based, route data collection) or the
analyzer (for in-depth, or "spot", analysis).
 Although
sometimes a
relatively simple
sine wave, it will
far more often be
a complex signal
with a number of
different
frequency
components.
 The "complex"
signal shown
below (still
simplistic
compared to data
collected on most
real machines) is
made up of a 1x
rpm component
(e.g unbalance)
and a 5x rpm
component (e.g.
number of vanes
on the impeller -
"vane pass"
frequency) being
generated by the
machine.
 There can be (and

usually are) far
more influences -
background
(frictional) noise,
misalignment,
bearing problems,
soft foot,
looseness,
frequency
modulation,
amplitude
modulation, etc.,
etc., etc.
What the transducer actually 'senses' is an analog signal - one
that mirrors the actual movement of the bearing at the
location of the transducer. The signal processing that follows
the analog signal collection consists of a couple of
mathematical processes:
 A/D converter - Converts the analog signal to a digital
one.
 Fourier Transform - This process is based on the principle
that any periodic signal (e.g. vibration) can be broken
down into a series of simple sinewaves that, when
combined, result in the shape of the original signal.
Using the above "complex" signal as an example, in practical

terms the principle means that the FFT process can deduce the
two frequencies (1x and 5x) that were present to create the
signal we see. The signal is fairly simple, though. Despite the
presence of two signals, even we could make that judgement.
That principle, however, can be extended to any periodic
signal. For each signal the FFT analyzes, there is one and only
one mathematical solution to the problem - a specific series of
simple sinewaves of precise amplitude values and phase
relationships (which do NOT show up on the resulting plot but
ARE considered by the FFT as we will see later) that, when
combined, create the precise shape of the signal the FFT is
analyzing.
The FFT process is an extremely complex mathematical
process that is being applied to mechanical vibrations.
Although a fairly reliable and useful tool, it MUST be
understood that a spectrum is always suspect because these
mathematical processes (A/D and FFT) often cause either or
both of the following to happen:
 Vibration peaks get added (like sidebands and
harmonics) that don't actually exist. That is not to say
either are to be ignored - they can still provide valuable
clues to the analyst.
 Occurrences (events) that may be obvious when viewing
the raw time domain signal are completely lost.
It is the signal shape that is being analyzed and deviations

due to any mechanical problem from purely sinusoidal motion
can cause the above phenomenon (harmonics, sidebands) to
occur. For these reasons, it is strongly recommended that at
the very least the time domain be used where it is most
useful and the spectrum is the weakest:
 Slow Speed Equipment
 Gear Applications
 Sleeve (Plain) Bearings
The reasons for this lie in what the FFT process actually does
and what factors influence its output (the spectrum).
What An "FFT" Is Actually Made Up Of ?

One of the "parameters" that must be programmed into the
database or the analyzer is known as the "Number of Lines"
(of resolution). This parameter determines how many
individual amplitude values will make up the final FFT plot.
That is what a spectrum is made up of - a certain number of
amplitude values (e.g. 800) that each measure the vibration
found in a relatively small frequency range. This parameter -
number of lines - works in conjunction with your Maximum
Frequency, or "Fmax", to establish your "Spectrum Resolution"
- a critically important subject. The Fmax divided by the # of
lines equals the spectrum resolution. The units are: "CPM per
Line of Resolution"
Fmax
Spectrum Resolution =
Number
of Lines
Close-up Look At FFT Plot Shows Individual, Frequency-Based
Amplitudes
With Straight Lines Connecting The Dots
The spectrum shown here shows 15 of the amplitude

ranges that make up this spectrum.
 Each frequency range in this spectrum is 30 cpm
wide. Each red circle is labeled as an exact multiple of
30 except the peaks.
 The frequency values shown at the tops of the peaks

are able to be calculated more accurately.
 The y-axis value of each red circle is the amplitude for
that frequency range. For instance, the y-axis value for
the red circle with the number 2130 above it is the
amplitude of vibration detected by the FFT process
between frequencies of 2101 - 2130 cpm.
 The number of lines of resolution can be 100, 200, 800,
1600, 3200 (and others) but typically is 400, 800 or
1600.
 The maximum frequency shown on the plot is called the
Fmax. If you select an Fmax of 60,000 cpm (60kcpm,
1000 Hz, 1kHz - all mean the same thing), your
spectrum will cover a frequency range of 0-60kcpm.
With 400 lines, for instance, each line of resolution will
be 150 cpm wide (60k/400=150).
 The width of each frequency range is called the
Spectrum Resolution.
 It is of CRUCIAL importance to understand spectrum

resolution.
How Is An FFT Analyzed

Analysis is performed by asking one seemingly simple question:
 "Are any of the amplitude values higher than desired ?"
o If the answer is "No", move on. Remember, this

includes all areas of the spectrum from 1x rpm thrugh
harmonics out to high frequency, bearing related
frequencies while keeping in mind the amplitudes in
these ranges should be judged by different standards.
For instance, you should become alarmed at much
lower amplitudes when analyzing bearing-related
vibrations than when analyzing 1x rpm amplitudes.
Sample Spectrum: Q1 - Are Any Amplitude Values Higher Than Desired ?

o If, however, the answer is "Yes", you must move on to
a second question:
 "Where are they coming from ?" More technically, what is
the source of the "excessive" amplitudes that you are
concerned with ?
The source of the vibration is judged by the

o
frequency(s) of the amplitude peak(s) being examined.
Spectrum resolution must be kept in mind during this
process. The better the resolution, the better your
judgment.
Since the frequencies displayed actually cover frequency ranges,
the best judgment of frequency you can ever make from a
spectrum is an estimate.
Spectrum resolution goes to the heart of a very important
question - how accurate is the data (frequencies) you are using
for your analysis ? Without fully understanding the implications
spectrum resolution has for the data you are analyzing, you will
never be a good analyst.
How Is An FFT Analyzed
Sample Spectrum: Q2 - Where Are (Any) High Vibration Amplitudes Coming

From ?
This sample spectrum is taken from a machine as described
above. How do we analyze ?
 There is no indication of bearing problems (in the "Bearing
Defect Frequency Harmonics" area) so we would have no
indication of a problem developing in that range.
 Does the peak at 1800 concern us ?
There are no amplitude values shown here. The answers to the

questions are part of a judgment must be made by the analyst
based on their understanding of general vibration and that
specific machine's normal operating (vibration) characteristics.
So if the answer is YES, further judgment must be made of the
source (unbalance, misalignment, soft foot, etc.) and the problem
must be corrected. Part of the judgement will be an assessment
of the severity of the problem based on the vibration levels seen,
type of problem, criticality of the machine, etc.
Important Facts To Remember When Analyzing
An FFT
1) Frequencies on a spectrum are always estimates. Therefore,
judgements you make are always based on assumptions.
2) Turning an estimate into a fact can only be done with field
testing (phase readings, etc.).
3) A spectrum is not necessarily an accurate reflection of what
the machine is doing - it should always be looked at with a certain
level of skepticism and supplemented with time domain analysis.
Spectrum Resolution
Max Frequency
Spectrum (Fmax) Spectrum Resolution Units:
Resolution = # of Lines of "CPM per Line of Resolution"
Resolution
Common Choices for: Common Choices for:
Fmax: Lines of Resolution:
3,000 90,000 720,000 200
6,000 120,000 900,000 400
12,000 180,000 1,200,000 800
24,000 240,000 1,500,000 1,600
30,000 300,000 1,800,000 3,200
60,000 600,000 2,400,000 6,400
The Importance Of Spectrum Resolution

Question: What Does Spectrum Resolution Actually Do ?
Answer: It determines the accuracy of the frequency data you

are analyzing.
Question: Why is that important ?
Answer: Frequency is how we identify the source of the

vibration.
Lets look at a couple of examples where understanding

spectrum resolution can be the difference between a correct
diagnosis and an incorrect diagnosis.
Understanding Spectrum Resolution:
Direct-Drive / Example 1
The FFT shown here displays data from 0 - 120kcpm. You have
made the judgement that one of the peaks shown in the low-
mid frequency range (1800-7200 cpm in this case) is
excessive. How reliably can you answer the following
questions:
What is the source of the vibration at 1,800 cpm ?
 Obvious choice - 1x rpm: Since it is a direct drive
machine and we will make the assumption that there is
no other vibration feeding into this machine from
another, you have a very good probability of being right
(note that it is not a 100% probability) if you say that
peak is the vibration 1x rpm.

 Obvious choice - 2x rpm: Since the machine is running
around 1800, chances are again fairly good you will be
right if you say this peak is related to 2x rpm.

 Obvious choice - 4x rpm: This could be the 4x running
speed harmonic and many people would think that is the
obvious choice - but that would be very risky because
there is another vibration source very close to this range.
2x AC line frequency in the U.S. is 7200 cpm (6000 cpm
in countries with 50 Hz power supply) and could very well
be any proportion of that amplitude.
 For example, if the machine is running 1780 rpm (1480
rpm in 50Hz countries), 4x that is 7120 cpm (5920 cpm
in 50Hz countries). That is 80 cpm away from 2x line
frequency (7200 cpm or 6000 cpm) - much less than the
300 cpm/line spectrum resolution. That means the
energy from both of the vibrations will be quantified in
the same line of resolution (i.e. combine into one, single
peak).
Are you monitoring bearing defect frequency harmonics

sufficiently ?
 Yes. With a machine running 1800, collecting an FFT to
120kcpm is sufficient to monitor the full range of
frequencies in which bearing-related peaks will appear.
Summary: This is a good general purpose spectrum for fully
monitoring the frequency ranges in which likely problems will
show up but with a spectrum resolution of only 300 cpm/line,
that could be a problem for any detailed analysis. Let's
continue by reducing the Fmax.

questions:
 Obvious choice - Still 1x rpm: Since it is a direct drive
machine and we will make the assumption that there is
no other vibration feeding into this machine from
another, you have a very good probability of being right
(note that it is not a 100% probability) if you say that
peak is the vibration 1x rpm.

 Obvious choice - Still 2x rpm: Since the machine is
running around 1800, chances are again fairly good you
will be right if you say this peak is related to 2x rpm.

 Obvious choice - ?: This could be the 4x running speed
harmonic but could also be 2x AC line frequency and
could also very well be a combination of the two
vibrations since the vibration energy from each falls
within the same line of resolution.

sufficiently ?
 No. With a machine running 1800, collecting an FFT to
60kcpm is probably NOT sufficient to monitor the full
range of frequencies in which bearing-related peaks will
appear.
Summary: Not the best choice for fully monitoring the
frequency ranges in which likely problems will show up since
not only do you have mediocre spectrum resolution (150
cpm/line) but you also are not monitoring the full range of
vibration frequencies in which bearing problems will show up.
Let's again reduce the Fmax.

questions:
 Obvious Choice - Still 1x rpm: Notice that the frequencies
listed have changed. Now that each line of resolution is
only 30 cpm wide (12000/400), the accuracy of the
frequencies displayed has improved as well. As the
resolution improves, the likelihood that the peak is 1x
rpm continues to improve but will never reach a 100%
certainty.

 Obvious choice - Still 2x rpm: See the explanation for 1x
rpm.

 Obvious choice - ?: At this point, you may notice the
harmonics not lining up perfectly with the4x rpm peak.
Note that it is also labelling a bit higher (7190 cpm) than
4x rpm would be. This may well be 2x AC line frequency
and could still be a combination of both vibrations since
the vibration energy from each falls in adjacent lines of
resolution.

sufficiently ?
 No. Not even close. This is a spectrum useful only for
monitoring low-mid frequency sources.
Summary: A reasonable choice for monitoring the lower,
mechanical frequency ranges in which likely problems will
show at running speed harmonics. There is no monitoring of
vibration frequencies in which bearing problems will show up.
Let's now increase the number of lines instead of reducing the
Fmax.
The FFT shown here displays data from 0 - 12kcpm but with
1600 lines of resolution instead of the 400 lines each of the
previous FFTs contained. How reliably can you answer the
following questions:
 Obvious Choice - Still 1x rpm: Notice that the frequencies
listed have changed. Now that each line of resolution is
only 7.5 cpm wide (12000/1600), the accuracy of the
frequencies displayed has improved as well. As the
resolution improves, the likelihood that the peak is 1x
rpm continues to improve but will never reach a 100%
certainty.

 Obvious choice - Still 2x rpm: See the explanation for 1x
rpm.

 You can finally differentiate between 4x rpm and 2x line
frequency: 2x line frequency (7200 cpm) is the higher
amplitude vibration. Note that there is also vibration at
4x rpm (7160 cpm) but it is of lower amplitude. The
choice for corrective action if the vibration was 4x rpm
vs. 2x line frequency is completely different. Failure to
recognize the limitations of the previous FFTs and make
this further anaysis could lead to embarassing and
expensive mistakes.

sufficiently ?
 No. Not even close. This is a spectrum useful only for
monitoring low-mid frequency sources.
Summary: A very good choice for monitoring the lower,
mechanical frequency ranges and separating the mechanical
frequencies from the electrical frequencies in that range.There
is, of course, no monitoring of vibration frequencies in which
bearing problems will show up. Let's repeat the process on a
belt-driven piece of equipment.
Note that with a >400 line spectrum, you can "zoom-in" on narrow
frequency ranges and appear to have "normal" spectrum resolution. This
is because the software stores however many amplitude values (800,
1600, 3200, etc.) even though it can only display 400 or so due to
limitations of the CPU monitor.
Zoom-In On Smaller Frequency Range (6k - 9kcpm)

Belt-Drive / Example 1
mid frequency range (< 12kcpm) is excessive. How reliably can
you answer the following questions:
What is the source of any of the vibrations apparent in the
<12kcpm range ?
 Entirely Unclear: As a belt-driven piece of equipment,
there are 3 rotating components (motor, fan, belts) plus
the same 2x line frequency we touched on in the direct-
drive example. If you venture a guess as to the source of
any particular peak, you have about a 1 in 3 or 4 chance of
being correct - not the best of odds.

sufficiently ?
 Yes. With a machine running 1800, collecting an FFT to
120kcpm is sufficient to monitor the full range of
frequencies in which bearing-related peaks will appear.
Summary: A spectrum really suited to bearing monitoring and
low-mid frequency trending only - no specific frequencies are
identifiable in the lower frequency ranges.Let's proceed with the
same sequence of FFTs as shown in the previous example.
mid frequency range (< 12kcpm) is excessive. How reliably
can you answer the following questions:
<12kcpm range ?
 Still Very Unclear: As a belt-driven piece of equipment,
there are 3 rotating components (motor, fan, belts) plus
the same 2x line frequency we touched on in the direct-
drive example. You may have improved your chances of
guessing right but you are still guessing..

sufficiently ?
 Not really. With a machine running 1800, collecting an
FFT to 60kcpm is probably not sufficient to monitor the
full range of frequencies in which bearing-related peaks
will appear.
Summary: A spectrum not really fully suited for either purpose
- bearing monitoring or low-mid frequency trending. Let's drop
the Fmax further still.

made the judgement that one of the peaks specifically in the
low-mid frequency range we are looking at is excessive. How
reliably can you answer the following questions:
<12kcpm range ?
 Much Clearer: Knowing that mechanical vibrations occur
only at exact multiples (harmonics) of running speed
(including 1x), we can assume the following:
o 900 cpm - 1x Fan (high likelihood)
o 1170 cpm - 2x Belts (high likelihood)
o 1770 cpm - 1x Motor (reasonable possibility - could
also be 2x Fan or 3x Belts, couldn't it ?). Most likely
- ???
o 2340 cpm - 4x Belts (high likelihood)
o 3540 cpm - 2x Motor (probable - 4x Fan and 6x
Belts could also be affecting this peak although the
probability is pretty low for either of those). Most
likely - 2x Motor.
o 7200 cpm - 2x Line Frequency (with a resolution of
30 cpm/line and the motor running at 1785, 4x
motor will be 7140 - 2 full lines of resolution away
from 2x line freq.), or 8x fan rpm (8 could be the
number of blades on fan). The most likely is 2x line
frequency but that doesn't make it so.

sufficiently ?
 No. This spectrum does not serve that purpose.
Summary: A fairly good spectrum for resolution at the low-

frequency end - but is it enough ? Let's increase the number of
lines to 1600.

The FFT shown here displays data from 0 - 3.6kcpm. You have
made the judgement that one of the peaks specifically in this
"mechanical" range (dealing with rpm's) we are looking at is
excessive. Because you have sufficient resolution, you can now
incorporate the "process of elimination" into your thought
processes (when a source becomes the only logical possibility).
How reliably can you answer the following questions:
<3.6kcpm range ?
 Much Clearer Than Before: Knowing that mechanical
vibrations occur only at exact multiples (harmonics) of
running speed (including 1x), we now have a much more
reliable assumption than with the e following:
o 1177 cpm - 2x Belts (high likelihood simply via
process of elimination)
o 1770 cpm - Three separate mechanical vibrations
had previously combined to form this peak. Because
of the new spectrum resolution of 7.5 cpm/line,
they are now separated on the FFT once you zoom-
in on that portion of the plot. Now they can be
judged for cause and severity and what might have
mistakenly been called motor unbalance (1x motor)
can correctly be called belt problems (3x belt typical
frequency generated by belt problems - could be
wear, resonance, etc.) along with possibly some
mechanical looseness on the fan (2x fan typical
frequency generated by looseness around the fan
bearings and/or the surrounding structure).The
three are:
 3x Belts is the highest amplitude of the three.
 2x Fan is the second highest amplitude of the
three.
 1x Motor is by far the lowest of the three.
o 2340 cpm - 4x Belts (high likelihood by process of
elimination)
o 2700 cpm - 3x Fan (high likelihood by process of
elimination)
o 3540 cpm - 2x Motor (probable - 4x Fan and 6x
Belts could also be affecting this peak although the
probability is pretty low for either of those). Most
likely - 2x Motor.
What About The 7.2kcpm peak ?

 7200 cpm - Can't see from this spectrum but we can also
zoom-in on that frequency range when done analyzing
this frequency range.

sufficiently ?
 No. This spectrum does not serve that purpose.
Summary: A very good spectrum for resolution at the low-

frequency end - but is it enough ?
But is it possible that the three frequencies in the 1762 - 1807
cpm range could have lined up even more closely ?
Of course. It is not uncommon to find belt-generated
frequencies aligning themselves very closely with driven or
driver component frequencies. Let's discuss a not-to-remote
possibility:
The belts run at 600 rpm and the fan runs at 1192 rpm. 2x
belts - a normal frequency belts will generate when
experiencing wear or other problems - would then be 1200
cpm - a mere 8 cpm apart from one another. Furthermore,
you don't know the belt speed. After all, it is probably more
common to not know the belt speed than it is to know it. How
easy would it be to try to go balance that fan (the most likely
source of 1x rpm vibration; especially on a fan: unbalance).
And how embarassing and expensive could that be for you and
for the credibility of the vibration program.
You should always, Always, ALWAYS confirm your vibration
source before before attempting any significant corrective
actions such as balancing (tightening loose bolts, for instance,
can be done as they are found).
How Do You Know What Spectrum Resolution

Is Required ?
Rule of Thumb:
Spectrum Resolution should be 33% of
difference between the frequencies.
 In previous example, frequencies of interest were:
o 1,762 cpm
o 1,785 cpm
o 1,807 cpm
 Smallest difference is 22 cpm (1,807 - 1,785 = 22).
 Spectrum resolution should be 22/3 = 7.33 cpm/line
Why 1/3 ?
 It means that there will be at least TWO lines of
resolution between the frequencies in question.
 This assures proper separation of the peaks.
Where To Collect "High Resolution" Data

Max Frequency (Fmax)
Spectrum Resolution =
# of Lines of Resolution
Direct Drive [12,000 Fmax / 1,600 Lines]
Machines:  Collected for all >10 HP Motors.
 Will allow separation of mechanical /
electrcial frequencies.
 This is particularly important on 2-pole ac

induction motors since it will be easy to
confuse 2x rpm for 2x line frequency.
Belt Drive [12,000 Fmax / 1,600 Lines]
Machines:  Collected for >10 HP Motors. In addition
to performing the above tasks;
 Will greatly improve chances of being
able to separate motor, driven and belt
frequencies from one another. Note that
it does not say guarantee the separation
of those frequencies.
[6,000 Fmax / 400 Lines]

 Collected on inboard (pulley end) of
driven component bearing for same
(mechanical) reasons as listed above.
 This can be adjusted to meet specific

needs if the actual speeds are
determined.
Beat  Any time there is a cycling or "beat"
Frequencies: frequency is encountered there are two
frequencies very close to one another.
Understanding spectrum resolution will
enable you to calculate what is needed
and use it appropriately.
What Are "Beat"
Frequencies ?
What is meant by the term: "beat frequencies" ? If you work in
a place with lots of machines, you have probably felt or heard
one. Have you ever walked by a machine and felt or heard the
vibration increasing . . . and then going away . . . and then
increasing . . . and then going away . . . etc., etc. That is a
"Beat" vibration.
Beat vibrations are caused by vibrations that have very close
to the same period and similar amplitudes.
 The animation shown here shows two separate sine
waves - the red 1x rpm component of a 2-pole motor
and the line frequency being supplied to that motor.
Either can cause mechanical vibration.
 Note that the amplitudes are the same. The amplitudes
must be at least similar to produce a beat. Otherwise,
the dominant amplitude will be only slightly affected by
the much lower amplitude and the effect will not be felt.
What does this interaction of these signals produce ?

 You can see here that when the peaks coincide (they are
'in-phase'), they add together and create more vibration
than either single signal produces by itself.
 When they oppose one another (they are out-of-phase),
they cancel each other out and the vibration disappears -
for a moment, anyway.
 The "Beat" frequency is simply how often these separate
signals go in-phase with one another. It might be 10
times a minute or 20 times a minute.
 The difference between the two frequencies IS the beat

frequency. By measuring the frequency of the beat (when
you can feel or hear it), you can actually determine what
spectrum resolution you need to separate the 2 peaks on
a spectrum by simply counting the number of beats that
occur in one minute or so (or portion of a minute) and
dividing by 3.
For example, a 2-pole US motor is running at, say, 3580 rpm.
Line frequency is 3600 cpm. The difference is 20 cpm. That
means if the 1x rpm component and the line frequency
component have similar amplitudes, there will be a beat
frequency occurring at 20 cpm. If you have ever walked past a
motor and heard it "humming", it is a beat vibration being
generated by the interaction of mechanical and electrical
frequencies.
Summary
The FFT is without a doubt the most widely used vibration
analysis plot. But it would be a mistake to consider the FFT to
be the infallible or the only analysis tool. Spectrum resolution
is perhaps the single most important plot characteristic to fully
and completely understand. It is also very important to
understand how a spectrum is generated - the concepts
behind the FFT process without getting into the technicalities.
An understanding of the FFT process will further boost your
ability to effectively analyze an FFT.
Plots:
The Enveloping Spectra
What Is An "Enveloping Spectra" Plot ?
Y-Axis
Units:
Amplitude
X-Axis Units:
 The term "enveloping" spectra plot is not always a

technically correct description of the signal processing
involved but will be the term we use for simplicity sake.
 An enveloping spectra is the same in appearance
(amplitude vs. frequency) as a conventional spectrum -
it simply displays different information.
 An enveloping spectra is not sensitive to sinusoidal
motion - unlike the FFT plot that determines what simple
sinusoids combined to generate a complex signal in
displacement, velocity or acclereration units.
 An enveloping spectra is sensitive to impact related
events.
 The ability to quantify both the frequency of impacts and
their intensity is very helpful and important to the
vibration analyst. Although there are machines that
generate impact energy normally (i.e. reciprocating
equipment), most machines don't. Impacts are
destructive forces and normally indicate some type of
problem is developing. Most typically, this plot is used to
detect bearing defects.
How Are Enveloping Spectra Plots Processed ?

What is the enveloping signal and how is it processed ?
 The unit of amplitude measurement is acceleration but
the signal is processed differently than a conventional
acceleration signal is.
 The names for the amplitude unit are manufacturer
specific - they each have their own name and/or
acronym for the unit. A few of the manufacturers are:
o CSI (Emerson) uses Peakvue
o Entek (Rockwell Automation) uses gSE (spike
energy - the original IRD acronym)
o SKF uses HFD (high frequency domain) and ESP
(envelope signal processing - originally a DI unit)
 Filters are used to help process the signal and focus on
any impacts that may be occurring.
 The filters come in two classes:
o Envelope filter - this type of filter sets a frequency
'envelope' that includes a high frequency (Fmax)
and a low frequency (Fmin). Any vibration
occurring outside that range is filtered out.
o Hi-Pass filter - this type of filter eliminates the
Fmax but still sets an Fmin filter below which all
vibration influences are filtered out.
o Each manufacturer sets up its own signal
processing and filters. Therefore, although they
each provide similar information, they are not
directly comparable in the amplitude realm.
 The signal processing focuses on the transient, impact
type events (spikes on the time domain signal) that the
FFT process "misses" (it would be more accurate to say
"makes more difficult to find") due to the way it
processes the time signal.
 If there is a consistent period between impacts (i.e. the
impacts are occurring at a regular interval), that period
will be converted into the desired frequency units (Hz or
cpm).
 The intensity of the impacts will also be assessed. This
is related to the size of the impact spike on the signal
versus any background noise occurring.
 The results are displayed on a spectrum with amplitude

peaks at the frequency(s) they are occurring at.
The enveloping spectrum provides us with valuable
information unavailable on displacement, velocity and
acceleration spectra. It provides another useful weapon for
the analyst.
Enveloping Spectra Plots:

Bearing Defect Multipliers
To understand the envelope plot's importance in diagnosing
bearing defects, you need to understand how bearing defect
frequencies work. To understand bearing frequencies, we
begin with bearing "multipliers". A bearing defect multipliers
is based on the geometry of the bearing. The important
geometric characteristics include the pitch diameter, the
number of rolling elements, the rolling element diameter and,
for ball bearings, the contact angle. There is a multiplier for
each of the four bearing components you see here. The
purpose of each multipler is to tell you how many impacts
(spikes on the time domain plot) will occur for each shaft
rotation for a defect on any of the four different bearing
components. These components are:
 Cage or Train (black)
 Balls or Rollers (dark gray)
 Outer Race (light gray outside)
 Inner Race (light gray inside)

Figure 1 - A Typical Ball Bearing
Figure 2 - Outer Race Defect Being

Impacted As Each Ball Passes
Bearing Defect Multipliers
Since these multipliers are based on the geometry of each
individual bearing, you can obtain specific numbers for any
bearing from numerous sources (vendors, manufacturers,
etc.). To turn a multiplier into a frequency requires applying
the multiplier to the speed of the machine in question. If your
multiplier is 3.05 and the machine runs at 1000 rpm, the
defect frequency is 3050 cpm. That means there are 3050
impacts occurring each minute due to the presence of that
defect.
What are the ranges of these bearing defect multipliers ?
FTF [Fund. Train Freq.] 0.30 - 0.45 x RPM
BSF [Ball Spin Freq.] 1.5 - 4.5 x RPM
2xBSF [2x Ball Spin Freq] 3 - 9 x RPM
BPFO [Ball Pass Frequency Outer] 2.5 - 9 x RPM
BPFI [Ball Pass Frequency Inner] 4 - 13 x RPM
These are typical ranges you will find on common bearings.
Some bearings may have considerably higher defect
frequencies - the determining factor is primarily the number of
rolling elements (which is related to the load rating of the
bearing). The higher the load rating of the bearing, the more
rolling elements there are and the higher these multipliers can
be. The inner race multiplier, for instance, can be well over 20
but that is unusual.
There are some important facts about these defect multipliers
that the analyst should always keep in mind:
 They are based on proper installation (i.e. proper fit) and
proper lubrication. Certain conditions can alter these
multipliers and in some cases actually increase them.
 They can be very close to exact harmonics of running
speed - 3.05 x RPM, for instance. That means if the
machine runs at 1780 rpm, the defect frequency is 5429
while 3x rpm is 5360 cpm - only 69 cpm difference.
These could be easily confused and misdiagnosed.
 It is extremely important to understand that no matter

how close they are to exact running speed harmonics,
bearing defect frequencies CAN NEVER BE exact running
speed harmonics. They are always non-synchronous
vibration sources - a fact vital to their correct diagnosis.
How Does "Impact Energy" Occur ?
Let's examine how impact energy due to a typical bearing
defect occurs:
Figura 1 Figura 2
In Figure 1, as each rolling element passes the defect, an

impact occurs. As we began discussing in the Time Domain
section, if you strike a bell, the bell will vibrate at its natural
frequency. That is true of any structure. The time it vibrates
will be determined by the force of the impact, the mass, the
damping characteristics of the object and other variables. This
is called "free" vibration (as opposed to the "forced" vibration
caused by energizing a machine and keeping it rotating and,
consequently, vibrating). The bearing impact causes the
bearing assembly to "ring" briefly until the free vibration due
to the impact dampens out. There are two frequencies
occurring here that are specifically related to the bearing
defect:
1) The bearing assembly natural, or "resonant", frequency
(based on the period of the bearing assembly resonance).
 Since the impact causes the bearing structure to ring,
there is a sinusoid generated briefly related to the
bearing assembly's resonant frequency.
 Because there is a sinusoid generated, this frequency is
detected by the FFT process and amplitude peaks will be
generated initially on the acceleration spectra (since it is
more sensitive to high frequency vibration) and
eventually the velocity spectra (displacement units are
useless at those frequencies).
 The difficulty lies in the fact that the FFT will have to
mathematically account for the fact that the spike
suddenly appears, briefly rings down and then is gone
until the next impact occurs. It is not a constant
sinusoid, it is transient.
2) The "impact" frequency (based on the period between
impacts).
 The impact frequency itself has no sinusoidal motion
associated with it. In other words, there is no sine wave
that connects the start of one impact to the start of the
next impact - they are individual 'events' that occur.
 These impacts (spikes) are specifically what the
enveloping signal processing looks for and measures.
 It will calculate the intensity of the impact (the size of

the spike) and the frequency (based on the period
between impacts) while filtering out any sinusoidal
motion it finds.

How Does Impact Energy Affect The FFT ?
Let's review how the FFT process works by examining the
following computer generated signal:
Figure 1 - Shows Approximately 9 Shaft Rotations (470 msecs)
The conventional FFT process focuses on sinusoids - namely,
mathematically calculating what series of simple sinusoids
(signals) were combined to generate the signal we see here.
What can we see from the above plot ?
 A low frequency sinusoid that shows about 9 cycles across
Fig. 1. That is the 1x rpm signal.
 Some frequency modulation of that signal (compare the
positive going side of the wave to the negative going side of
the wave).
 A large number of spikes, or impacts, that occur across the
plot and appear to vary somewhat in intensity (the size of
the spike).
Figure 1 is a typical example of a plot that an analyst might

collect - 9 rotations of a shaft. But although the 1x sinusoid is
fairly clear, the impacts are not. Let's zoom in a bit.
Figure 2 - Shows Approximately 2 Shaft Rotations (115 msecs)

Cutting the displayed sample to just over 115 msecs (about 2
shaft rotations), we can now clearly see:
 The frequency modulation of the 1x rpm signal.
 The ringdown frequency of the impacts.
 If we simply count the number of impacts in one cycle (from
30 - 80 msecs, for instance), we would find about 4-5 per
shaft revolution (or "x RPM").
It should be clear to us as analysts that this is an impact

occurring and investigation of the period involved (time between
impacts) should lead us to a diagnosis. But more often than not,
the analyst will not be using the time domain - they will be using
FFT analysis. What does an FFT performed on this signal generate
?
Figure 3 - FFT Generated From Signal In Figure 1

 1x, 2x and 3x rpm peaks. These are probably due to the
frequency modulation present.
 A series of peaks at high frequencies that are spaced about
5400 cpm apart.
 The absence of a peak at or near 5x rpm - the impact
frequency. This is because there is no sinusoidal motion
associated with the frequency of the impacts - only the
ringdown frequency that results from the impacts.
But where do the peaks between 31,000 and 65,000 cpm come
from ? How does the FFT process come to "see" them ?

The answer lies in the mathematics involved in the FFT process.
Let's look again at a time domain plot that is representative of
what a bearing defect will look like:
Figure 1 - A Time Domain Plot With Bearing Related Impacts Occurring

The time domain in Figure 1 is clearly symptomatic of a
bearing defect - impacts at a frequency not related to rpm
(the larger sine wave). The time sample is 333 msecs.
But let's look at the FFT. What does the preceeding signal
generate when subjected to the FFT process ?
Figure 2 - The FFT Resulting From The Signal Shown In Figure 1

The FFT in Figure 2 shows series of peaks out in the 50k - 90k
range. These peaks are the "symptom" of a bearing defect
developing. But why ? Why does the FFT generate vibration at
those frequencies ?
The answer is in the math. There is only one series of simple
sinusoids that would result in the shape of the signal shown
above. Want more proof ?
Figure 3 (below) is a 30 msec slice of Figure 1 - a close-up, so
to speak:
Figure 4 shows the entire series of simple sinusoids that were

programmed in to create the exact signal shape you see in
Figures 1 and 3.
Figure 4
When the signal shown in Figure 1 is put through the FFT
process, that process is asked "what simple sinusoids create
that exact periodic signal".
 The process "sees" the series of sine waves shown in
Figure 4 as the mathematical solution to the question.
 Note the varying amplitude values of the sinusoids in
Figure 4.
 Note the three instants (65-66 msecs, 76-77 msecs and
87-88 msecs) all of the high frequency signals are in
phase (adding together).
 Note the high number of out-of-phase sinusoids at 71

msecs and 82 msecs.
There is only 1 combination of simple sinusoids that will
combine to mathematically create any periodic signal. Alter
the signal in any way and the series of sinewaves creating
that signal will change.
This analysis does not, of course, include the larger '1x rpm'
and '2x rpm' sine waves you can see. Those are 'seen' by the
FFT due to the frequency modulation on the 1x rpm signal (Fig
1).
Since this is such a complex subject, let's go through the
details in a different way.

The answer has to do with the transient nature of the impacts
and the principles involved in the FFT process. Looking at the
question from the FFT's perspective, we can re-phrase it as:
 What would cause a sinusoid to appear and then
disappear at regular intervals ?
Since the FFT is based on the principle that any periodic signal
can be broken down into a series of simple sinusoids, there
must be some combination of sinusoids that would produce a
sudden spike followed by a subsequent "ringdown" (high)
frequency followd by nothing until the next spike occurs.
The answer is actually fairly simple. When a series of sinusoids

separated by a common frequency (5400 in this case) are
combined to generate a periodic signal, the signal will appear
as a transient sinusoid (a spike, or impact, followed by a
ringdown followed by nothing until another spike suddenly
appears).
The following list of simple sinusoids were fed into a signal

generating software program. Note that although the
amplitudes are different, the frequencies are all separated by
5,400 cpm. Although there were other variables inputted to
create a more realistic looking signal, this list is, in fact, the
exact series of sinusoids that were combined to create the
transient (impact) sinusoid you saw on the previous page.
 0.05 @ 31,800 cpm

 0.16 @ 37,200 cpm
 0.28 @ 42,600 cpm
 0.30 @ 48,000 cpm
 0.18 @ 53,400 cpm
 0.10 @ 58,800 cpm
 0.06 @ 64,200 cpm
What would the signal look like if we only used the above 7
signals (plus some background noise and amplitude
modulations) ? See for yourself:
Figure 1
 The result is only impacts and background noise.
 What is happening is that this combination of signals will
all come into phase with one another at about the same
time, rings down to the noise level in about 4 msecs as
the signals go out-of-phase with one another and
remains at background noise level for another 6-7 msecs.
 The result is a large, brief amplitude increase (a spike, or
impact) every 11 msecs or so. That equals an impact
frequency of 5,400 cpm (the difference between the
frequencies).
Of course, the FFT does not have the benefit of knowing which
sinusoids went into generating this signal. In fact, that is
exactly it's job - to calculate those simple sinusoids from the
complex signal (including other influences like 1x rpm, other
mechanical vibrations, amp and freq modulation, etc.). So the
process is:
 The above signal is fed into the FFT process. That process
then calculates what simple sinusoids combined to
generate the signal.
 The FFT can deduce that there is a combination of sines
and cosines (signals) that will result in the above
complex signal - the combination listed.
 Adding or removing any signals that are a multiple of
5,400 will alter the appearance by making the impact
either sharper (more signals) or less well defined (less
signals).
 In fact, there is only one solution to each signal - only
one set of simple sinusoids.
 So when the FFT process is presented with the above

signal, what does the spectrum look like
Figure 2 - FFT Generated From Signal Shown In Figure 1
0.05 @ 31,800 cpm

0.16 @ 37,200 cpm
0.28 @ 42,600 cpm
0.30 @ 48,000 cpm
0.18 @ 53,400 cpm
0.10 @ 58,800 cpm
0.06 @ 64,200 cpm
Above: The Signals Used To
Generate The FFT Shown Here
 Note that there is no indication whatsoever of the impact
frequency (about 5400 cpm) on the spectrum.
 Why is there no peak at the impact frequency ? Because
there is no sinusoid associated with it !
There are several problems complicating the analysis of a

velocity or acceleration FFT such as the one shown in Figure
2.
 You must notice the existence of the peaks (separated
by the defect frequency). This may seem silly but keep
in mind that:
o You may be analyzing dozens or hundreds of
machines - thousands of bearings.
o The high frequency peaks you see in Figure 2 will
initially be low amplitude - particularly if you are
using a velocity spectrum (which most people use).
 The analyst relies on these peaks being harmonics of the
bearing defect frequency. You must be able establish
that pattern.
 Before you even try to determine the defect frequency
(which often requires time and effort), you must
somehow notice or sense that there is a harmonic or
sideband pattern that should be investigated.
 Eventually, You must know the defect frequency (the
impact frequency). The peaks you see in the 33k-63k
range are harmonics of the defect frequency (6x - 12x
defect frequency). That is how you will diagnose the
problem - by identifying the source of those peaks
through the use of harmonics. Without knowledge of the
defect frequency, it can be far more difficult.
So now, let's return to the subject at hand. What will the

enveloping spectrum look like and how will it help with the
analysis ?
Enveloping Spectra Plots
What Information Do
They Provide ?
Figure 1 shows an actual enveloping spectrum collected on a bearing
with a defect:
Figure 1 - Enveloping Spectrum
The bearing defect frequency identified is just over 3x RPM. Notice that
there are no significant peaks at 1x, 2x or 3x rpm on Figure 1 (there
were on the velocity spectrum). There are, however, extremely
significant peaks at 1x, 2x and 3x the impact frequency - in this case a
bearing defect frequency (there are other impact sources). The
enveloping signal provides the following:
The impact frequency:
 This piece of information can be used on the velocity or

acceleration spectrum to help determine the condition of the
bearing (how bad is it ?).
 After identifying the defect frequency from Figure 1, inspect your
velocity or acceleration plot and place your cursor on that same
impact frequency and turn on your harmonics.
 If you are able to relate, through the harmonics, the high
frequency peaks to this impact frequency, you have confirmed the
presence of a bearing defect.
 You can then make an assessment of condition based on the
amplitudes present, noise level, etc.
The intensity of the impacts:
 This piece of information
can be used to help
determine how quickly a
bearing can be expected to
deteriorate since the
impacts are so
destructive.
 You can compare this to
hitting a small pothole in
your car or hitting a huge,
sharp edged pothole - the Figure 2 - Enveloping Spectrum From
first is annoying, the Figure 1 on dB Scale
second can destroy your
wheel.
 The assessment can be

made by displaying the
amplitudes on a 'dB' scale
(see Figure 2) and
comparing the peak
amplitude to the
surrounding 'carpet' level
(which is affected by
lubrication and load,
among other things).
 In Fig. 2 (which is the same plot as Figure 1 except the
amplitudes were on a linear scale in Fig. 1 and are on a dB scale
in Fig. 2), the amplitude on the peak is about 125 dB.
 The surrounding carpet level, which is an estimate of the
surrounding amplitudes, is in the 100 - 102 dB range. The
following guidelines can be used:
o Difference of 12-18 dB is a significant level of impacting and

should be watched closely.
o Difference of > 18 dB is a severe level - intense impact
energy, very destructive.

Impact Sources
What are the common impact sources that the enveloping signal and
spectrum are helpful in detecting and what are the frequencies
associated with each. Each of the subjects listed below is discussed
more in-depth in the 'Troubleshooting Charts' section.
 Bearing Defects - Frequency of peaks will be the specific bearing
defect (impact) frequency.
 Looseness - Typically occurring between the shaft and bearing;
the housing and bearing; and/or the internal bearing clearances.
The observed frequency on the enveloping spectrum will be
harmonics of running speed (1x, 2x, 3x, etc. x RPM).
 Electrical Looseness (ac motors) - Looseness in windings, end
turns, loose iron, loose connections, etc. Frequency will be 2x AC
line frequency and harmonics. This also applies to Variable
Frequency Drives (VFDs) but the ac frequency must be
determined for each case.
 Lubrication - Lack of lubrication will drive up metal to metal
contact (high frequency noise). No specific frequencies are
triggered but a general lifting of floor, or 'carpet' level, will occur.
 Reciprocating Equipment - Analyst must determine the specifics
of the machine to determine what frequencies to expect. Running
speed harmonics are common with even numbered harmonics of
even higher amplitude (there are a lot of events at 2x rpm in
typical recips), number of pistons x rpm in some hydraulic
pumps. Impacts are normal in equipment such as this and the
analyst should be looking for change from the norm.
 Gears - Backlash, other impact sources. Frequencies typically will
be related to the number of teeth.
Note that each of these 'problems' generates its own, specific

frequency(s). Each of these subjects is covered more extensively in
the 'Troubleshooting Charts' section but there is one common thread
to using the enveloping spectrum (a word of warning, so to speak):

Words of Warning
 The enveloping spectrum is extremely sensitive. It will pick up
impact energy that is not necessarily a problem or is a very early
stage problem.
 For instance, it can detect bearing defects before they have
migrated to the surface of the bearing. Pulling the bearing at that
point will not reveal a defect and may cost you something more
valuable than money - credibility.
 Enveloping spectra should be used in conjunction with other
analysis tools (velocity and acceleration spectra, thermography,
time domain, your experience, etc.) before performing any
corrective actions. It is a powerful tool but must be used with
care.
 Like other aspects of vibration analysis, experience will help

greatly as it is acquired.
The Time Domain

 What Is A "Time Domain" Plot
 How Do You Analyze A Time Domain Plot
 The Relationship Between The Time Domain and The FFT
o Frequency Modulation (Harmonics)
o Amplitude Modulation (Sidebands)
 Where Time Domain Plots Can Be Invaluable
o Beat frequencies
o Bearing defects
o Gear tooth defects
o Rubs (Truncated Wave Shapes)
 How To Become Comfortable With Time Domain
 Setting Up Time Domain Plot Parameters
 Summary
What Is A "Time Domain" Plot ?
Y-Axis
Units:
Amplitude
X-Axis Units:
The "Time Domain" plot is a powerful tool to use for analysis

since it is as close to the raw (analog) data as you are likely to
get. It also can be quite intimidating to learn to use (with good
reason) and many analysts do not use it at all. Even more
than the FFT, it creates a number of questions for the analyst:
 How is it interpreted ?
 How is one set up ?
 Should I use it everywhere ?
 Why do I need it at all when I have the FFT to use ?
This section of the manual will provide you with information on

how to use and interpret the time domain plot. More
importantly (since it is from this signal that an FFT plot is
produced) this section will attempt to give you a general
understanding of how the FFT process views the signals we
feed into it and how these signals impact what we see when
we look at an FFT.
The time domain, of course, is where the reading begins - an
analog measurement of how the surface is moving. This
analog signal is fed from the transducer to the analyzer where
it is converted to a digital signal - it goes through an A/D
converter. The result of this process can be seen above in the
plot above.
It is important to realize that it is experience (i.e. practice)
that creates a 'comfort' level for the analyst in interpreting the
time domain plot. Experience in setting it up properly and
experience in being able to recognize what you are seeing -
the pattern of what you are seeing. Let's zoom in on this plot.
How Do You Analyze A Time Domain Plot ?

Y-Axis
Units:
Amplitude
X-Axis Units:
The time domain plot shown here is a small portion (time-

wise) of the previous plot - a zoom-in. At first glance, what do
we see ?
 A series of spikes - perhaps 50 or so. Each of these
'spikes' is a cycle of vibration just as the nice smooth
animated sine waves we previously examined were. The
difference, however, is obvious - these are not nice,
smooth sine waves. This is a 'real-world' time domain
plot.
 The amplitude of the 'spikes' varies a great deal.
 There appears, at times, to be changes in frequency of

the spikes (in some areas, there are 2x or 3x as many
spikes than in other areas).

Y-Axis
Units:
Amplitude
X-Axis Units:
But let's look a little more in-depth. What are we actually

seeing here ? It would only be fair to tell you the reading was
collected on a gearbox. Let's break our analysis into the
frequencies and amplitudes.
 Frequency
o The 50 or so cycles occur in a very short period of
time - they are measuring 'high frequency' vibration
(the shorter the period of the cycle, the higher the
frequency). As you probably guessed, it is actually
the gear mesh frequency.
o What is the source of the 'extra' cycles we noticed
in the areas of low amplitude ? They appear to be
occurring about twice per (gearmesh) cycle when
they do occur. One possible explanation would be a
momentary 'chattering' of the teeth (the teeth
bouncing back and forth) due to excessive backlash.
o Also, some of the spikes appear fatter than others.
What this means is that particular cycle (or part of
that cycle) is taking slightly longer (in milli or even
micro seconds) than other cycles. In vibration
terms, this is known as frequency modulation.
When a signal is put through the FFT process,
frequency modulation causes harmonics.We will see
why in a few more pages but the first question
should be - what could cause this ?
o Have you ever tried turning a shaft - even in a
childrens toy - that is easy to turn for part of a
rotation and hard to turn for the other part of the
rotation ? That binding action can occur in industrial
machinery for any number of reasons. But why is it
occurring here ?
So we can find a number of seemingly small details by

analyzing the time scale and how the cycles are occurring.
Let's now examine the amplitude data.
Y-Axis
Units:
Amplitude
X-Axis Units:
What can we see from the amplitudes being displayed ?

 The gear mesh peaks are changing size -
"modulating". That means the amplitudes are not

always the same. But how are they modulating - are
the amplitudes changing a predictable amount within
a predictable period of time ? Or does it seem more
random in nature ?
 If you had the tools necessary to analyze the plot, you
would find the amplitudes get high once per shaft
revolution and low once per shaft revolution. In
vibration terms, the gear mesh amplitudes are
modulating at 1x rpm.
 There is no obvious 1x rpm sinusoid - the dominant
amplitude is at gear mesh frequency.
So putting all of this analysis information together, what
can be seen ?
Y-Axis
Units:
Amplitude
X-Axis Units:
Let's review what we have.

 A dominant gear mesh frequency that is experiencing
both amplitude and frequency modulation.
 'Extra' cycles that appear to occur twice as often as the
dominant gear mesh cycle.
 A low amplitude 1x rpm signal.

Possible Explanation
A possible explanation (only testing and inspection can prove
anything) for these patterns would be improper gear mesh
setting (pitch diameters not in-line at all times). For part of
the gear's rotation, the teeth are too tightly set. That
increases the resistance to rotation and drives up the gear
mesh amplitude. For the balance of the rotation, the teeth
have excessive backlash. That decreases any resistance to
rotation and reduces the gear mesh amplitude but also allows
the teeth to bounce back and forth (chatter).
Action Recommended
Unfortunately, we have only reached the level of an educated
guess at this point. The data points in a direction but there
could be any number of other problems influencing the
patterns we are seeing and analyzing. Our "explanation" is
based on probabilities and, if we have it, previous experience.
We must now go into our LEARN MODE. What is the learn
mode ? It is the mode where we are actively involved in any
investigation and corrective action(s) taken so we can learn
from what we find. If you do not use the LEARN MODE (and
sometimes it requires you making it happen), you will not
learn from your successes or (and more importantly) your
mistakes.
Importance Of Understanding The Relationship

Between
The Time Domain Plot And The FFT
This may all seem a bit overwhelming (ok, it is overwhelming
- at least for beginners) but there is an important point to be
made here and it has nothing to do with analyzing the time
domain plot - it has to do with analyzing the FFT plot.
Even if you don't plan on using time domain plots, it is
important to understand how the FFT is generated from them
and how different wave shapes affect the FFT process and the
spectrum we see.
The Relationship Between The Time Domain

And The FFT
The following pages contain plots that will allow us to look at
the FFT process backwards. These "complex" time domain
signals are the result of programming in a series of amplitudes
and frequencies from an FFT and then seeing what the
resulting time domain looks like. Although mechanical
vibration will always be more complex than what we analyze
here, it is illustrative to see the process working in reverse.
In order to understand the complexities of the FFT process,
let's look at how it works in its most basic form - analyzing a
simple sinusoid. That is what you see in Figure 1. Performing
an FFT on Figure 1 would generate the plot you see in Figure
2 - a single peak at 1x rpm.
Pure Sinusoidal Motion

Result Of Running Fig. 1 Signal Through FFT Process
The FFT process sees a simple sinusoid and calculates the

period (time required) for a single cycle. In Figure 1, the
period is 40 msec (the x-axis units are milliseconds = 0.040
seconds). Then, do the following:
 Period =40 msecs per cycle [0.04 secs / cycle]. We have
seconds per cycle - we want cycles per second since we
are interested in frequency.
 Frequency (Hz) = 1 / 0.040 = 25 cycles per second
[25Hz]. If we want cycles per second:
 Frequency (cpm) = 25 Hz x 60 seconds / minute = 1500
cycles per minute [CPM].
The FFT is created with a peak at 1500 cpm (25Hz). The

amplitude shown will be based on the Window type shown and
whether you have a signal detection of RMS, peak, peak to
peak or true-peak.
Unfortunately, in the mechanical world there are only two
problems that cause such a pure sinusoid to occur (and it will
only be pure if they are the only problems present):
 Unbalance
 Resonance

And The FFT
What about less than pure sinusoidal motion. Any thoughts on
what could possibly cause the signal shown below ? More
importantly, what does the FFT process "see" as the
combination of signals that would create what we see in this
signal ?
Figure 1: A Signal That Is Not Sinusoidal
The FFT looks like thsi when applied to the signal shown
above. Why ?
The FFT Generated By Figure 1
Because the FFT process "sees" the combination of two distinct

signals: one at 1x rpm and one at 2x rpm. The two signals are
in-phase with each other (the "+" peaks line up). When the
"+" peaks line up, the very high peaks occur (0, 40, 80
msecs). When the "-" peak on the 1x wave lines up with the
"+" peak on the 2x wave (20, 60, 100 msecs), you get the
small "bump" at the bottom of the wave.
The Actual Signals Used To Generate Figure 1

And The FFT
Let's look at another signal with a completely different shape.
Can you see in your mind a bearing moving in this manner ?
What types of mechanical problems do you think could cause
the shape of the signal shown here in Figure 1 ? What does
the FFT process see ?
Figure 1 - The Raw Signal
It may surprise you to find that putting this signal through an

FFT will generate an identical spectrum to the one we saw on
the previous page.
Figure 2 - The Resulting FFT
Why ? It is clearly a very different signal. Or is it ? As a matter

of fact, exactly the same combination of signals went into
creating this one with one difference - the 2x component has
shifted 1/4 of a cycle on the time line relative to the 1x
component - the 2x signal is a 90° out-of-hase with the 1x
signal.
Figure 3 - What The FFT "Sees"

And The FFT:
Frequency Modulation
So is it safe to assume that each of the previous signals were
generated by a machine that is generating vibration at 1x rpm
and 2x rpm (i.e. a reciprocating compressor) ? Or could there
be another explanation for the signal shape seen on those
pages (which is really what is being analyzed - the signal
shape) ? Let's return to our discussion of the actual, real-life
vibration signal we looked at a few pages back.
 We discussed how there can be some variation in the
free rotation of the shaft - a momentary "binding" action
that occurs as the shaft rotates through a particular
portion of it's rotation.
 That phenomenon could occur for a number of reasons.
In that situation, we considered the possibility of the
gears being improperly set. That would create more
resistance to rotation when the teeth were bottomed out
than opposite that point. It would momentarily slow
down the rotation.
Let's examine the 'frequency modulated' signal shown here:
Figure 1: A "Frequency Modulated" Signal

Figure 2:
A Bearing Undergoing Frequency
Modulation During Each Rotation
 Let's first examine the positive-going portion of the

signal. The bottom of the cycle (the '-' peak) first occurs
at about 19 msec. The '+' peak occurs at about 32 msec
so it takes a total of about 13 msec to move from the "-"
peak to the "+" peak. The reciprocal of the period will
give us the frequency during that portion of the signal:
o 1/(0.013 x 2) x 60 = 2308 cpm (the 13 msecs is
multiplied by 2 to calculate a full cycle).
 Now let's examine the negative going peak. From the
"+" peak at 32 msec, the signal descends to a "-" peak
by about 53 msec - a total of 21 msec. For that portion
of the signal, the shaft rotates:
o 1/(0.021 x 2) x 60 = 1429 cpm.
 Yet if we simply calculate the total time for one cycle
(peak to peak), we measure from 19 msec to 53 msec -
about 34 msec.
o 1/(0.034) x 60 = 1765 cpm.

This is called frequency modulation. What is happening here
may or may not be evident if we were to analyze the time
domain signal - it will depend on the resolution (yes, time
domain is just as dependent on resolution as the FFT is), the
time sample, number of bytes, etc. But remember, the
question we are discussing here is how will the FFT treat this
phenomenon ?
The FFT only deals in pure sinusoids. So how will it account for
the frequency modulation we see here ? We will
unquestionably have a peak around 1765 cpm but the signal is
not a pure sinusoid - it is distorted by the frequency
modulation we see. How does the FFT mathematically explain
this phenomenon ?
In other words, what combination of simple sinusoids, when
combined, will generate the signal we see above ?

And The FFT:
So the FFT has fed into it the following signal:
Figure 1 - Signal Being Analyzed
So what does the FFT "see" ? This plot was generated from the
signal in Figure 1:
Figure 2 - Spectrum Generated by FFT Process
So we have peaks at 1x, 2x and 3x rpm. But how can the

spectrum we see in Figure 2 be the result of performing an FFT
on the signal we see in Figure 1 ?
If the signal shape in Figure 1 is a result of a "binding" action,
then there are no 2x or 3x components present - only 1x rpm.
Why does the FFT add these peaks ?

And The FFT:
The answer is actually very simple - it is precisely that
combination of simple sinusoids - 1x, 2x and 3x - in exact
amplitude and phase relationships that generates the signal
shape we see. Any change in the amplitude values of any of
these simple sinusoids or their phase relationship to one
another and the resulting signal shape will be altered. Here are
the 3 signals that combined to generate the previous signal:
Figure 1 - The Three Simple Sinusoids
What, exactly, are these signals ?

 1x rpm w/ amplitude of 1.8 (pk-pk) and no phase shift
('+' peak on y-axis)
 2x rpm w/ amplitude of 0.45 (pk-pk) and 90° phase shift
 3x rpm w/ amplitude of 0.05 (pk-pk) and 180° phase
shift ('-' peak on y-axis)
Note that the phase relationships are not displayed on the FFT
- it only displays amplitudes and frequencies although the FFT
process does use phase in its analysis.
So this "distorted" signal shape:
is explained (by the FFT) as the result of

this combination of simple sinusoids:
and the FFT process generates this plot:
So what does all of this mean ? The frequency modulation we

see in the original signal generates the presence of harmonics.
Frequency Modulation creates harmonics.
Are all harmonics the result of frequency modulation ? No.
There are mechanical viration sources that generate vibration
at multiples of running speed (i.e. reciprocating units @ 2x
rpm). But some harmonics - more often than not those
associated with problems such as misalignment - are the result
of frequency modulation. They are mathematically created.

And The FFT
The previous example may or may not have impressed upon
you a rather stunning fact:
 Many (even most, at times) peaks on the spectrum are
not actually being generated by the machine - they are
generated by the FFT process due to the shape of the
signal being processed.
That's a rather depressing statement - many of the peaks are

not being generated by the machine (they don't actually exist).
It shouldn't be a suprise though. This is a process that relies
on a complex mathematical principle to analyze data that
comes from dynamic machines with many, many different
variables being applied to them (do you have unbalance,
misalignment, is the base solid, the pipes lined up, etc., etc.,
etc.).
On the bright side, it also illustrates why it is useless to try to
identify every peak on a spectrum (a trap many analysts,
unfortunately, fall into). Let's look at the situation another
way.
If you knew that when a machine had a misalignment problem
that your computer would blow out a puff of blue smoke - the
more misalignment the more smoke - would you care about
the precise details of why your computer would do that or
would you simply be happy knowing you could count on that
notification ?
The point is that when machines have particular problems -
misalignment, for instance - those machines will vibrate in
certain ways characteristic to the problem they have and those
problems will affect the shape of the signal they generate. This
applies, for instance, to the binding action seen on the
previous page attributed to gears bottoming out. The shape of
the signal being generated due to that particular problem will
be affected in a reasonably consistently way. Under those
conditions, the FFT process will generate a reasonably
consistent result (the plot we see). Subtle changes to the
shape will change the spectrum but on the whole, certain
patterns you learn to recognize on the FFT plot will lead you to
investigate certain problems based on the likelihood of finding
that problem - the more likely (and easier to check) problems
go first and on down the list. Three or four different problems
may each generate similar spectrums - it is up to you to
differentiate between those similar symptoms and solve the
problem.
YOU CAN'T DO THAT SITTING AT A COMPUTER !!
Why is all this important ? Because:
 Understanding the inherent limitations and shortcomings
of an FFT should impress upon you the tremendous
importance of field testing and troubleshooting.
 It must be recognized that the spectrum provides clues
and insights - not facts. It is only one tool you have in
your battle to protect your machines.
 Although the FFT always "works" (we will get a plot),
sometimes we do not understand what it is telling us. To
attempt corrective actions without a thorough
investigation can be . . risky.
 The time domain plot can provide valuable clues and

insights and in some situations will provide information
that is impossible to determine from the FFT.
And The FFT:
Amplitude Modulation
We have discussed frequency modulation and its impact on the
spectrum plot - namely, it creates harmonics. But we have also
touched on amplitude modulation - now let's cover it more in-
depth. Amplitude modulation is a increase and decrease in the
amplitude of a particular frequency at a different frequency. So
for instance your gear mesh amplitude gets high once per
shaft revolution and gets low once per shaft revolution - the
gear mesh amplitude is modulating at 1x rpm. That's simple
enough but what effect does that have on the FFT (i.e. how
doe the FFT explain it) ? Let's look at some examples:
What do you make of the time domain plot shown in Figure 1 ?
Figure 1
You can see a low frequency cycle (occurring 15 times over the
time sample) and a high frequency (occurring many times for
each of the low frequency cycles). One way to describe this is
as "a high frequency riding a low frequency". For analysis, let's
zoom in:
Figure 2 - One Cycle Of Low Frequency
Figure 2 shows only just over one of the low frequency cycles
(one peak to just past the next peak). The high frequency
could be a gear meshing frequency. The low frequency is at 1x
rpm. How many teeth are on the gear ? This is another
advantage of using time domain on gearboxes - you can
actually obtain detailed internal information that you can only
guess at on the spectrum. Count the small peaks from the top
of one low frequency peak to the next. There are 23 teeth.
What does the spectrum look like ?
Figure 3 - The Resulting FFT
Both the amplitudes and the frequencies are constant - there is

no modulation in either. You will only get the peaks that are
actually being generated. Now let's consider some variations
on this "perfect" gearbox.

And The FFT:
Amplitude Modulation (Sidebands)
Remember this plot from the first page in this section ? Well,
Figure 2 shows a computer generated plot that is somewhat
similar:
Figure 1 - Actual Time Domain Signal
Figure 2 - Computer Generated Time Domain Signal
You can see the low frequency (usually 1x rpm) cycle still
occurring but this shape looks somewhat like an Angel Fish.
This shape is typical of an amplitude modulation. Let's zoom in
on the computer generated signal to get a clearer picture of
what is happening.
Figure 3
Figure 3 shows only a bit more than a single one of the low
frequency cycles. It is clear that the high frequency signal (the
gear mesh amplitude) is increasing and decreasing in
amplitude at a rate of once per shaft revolution. Figure 4
shows the two signals involved separated:
Figure 4
This represents what we were discussing before - a modulation

of gear mesh amplitude once per shaft revolution due to a gear
problem. The amplitude varies significantly at a rate of once
per revolution. How does the FFT handle amplitude modulation
? Figure 5 shows you:
Figure 5
Figure 5 shows a peak at 1x rpm and a peak at gear mesh

frequency (GMF) with smaller peaks surrounding it. It you
could measure the frequencies involved, you would see that
the smaller peaks are equally spaced around the large peak -
the separation is equal to the frequency of 1x rpm. There may
be a series of these peaks called sidebands around the gear
mesh frequency. This series of peaks are what would
mathematically cause the amplitude to go up and down
(modulate) as the peaks go in and out of phase with one
another. The difference between sidebands and other noise
around a peak is the equal spacing (1x rpm in this case).
Peaks that are not equally spaced are most likely not
sidebands.
Peaks on the higher frequency side of the large peak will be
located at GMF + 1x rpm, GMF + 2x rpm, GMF + 3x rpm, etc.
On the low frequency side of GMF, the peaks will be located at
GMF - 1x rpm, GMF - 2x rpm, etc. The spacing of the peaks is
the key indicator in where the problem lies. The spacing at 1x
rpm indicates the gear running at that frequency (speed) is the
source of the problem.
Although not nearly as common as harmonics, sidebands are
critically important to learn to recognize for a couple of
reasons:
 Sidebands always indicate a problem (amplitude
modulation is never "normal").
 Sidebands are only generated by three types of
problems:
o Gear-related problems
o Bearing-related problems
o Electrically-related problems
 Sidebands can be significant at very low amplitudes (their
mere presence can be significant).
 Sidebands should be analyzed on a logarithmic scale (as

opposed to a linear amplitude scale) so the low-amplitude
peaks can be detected.
The Addition Of Sidebands Is The FFT's Explanation Of The
Presence Of Amplitude Modulation
The Relation Between The Time Domain And

The FFT:
Frequency And Amplitude Modulation
Now let's have some real fun !! What do you see in the time
domain plot shown in Figure 1 ?
Figure 1 - The "Complex" Signal
The most noticeable characteristic is the 'Angel Fish' pattern

we saw on the previous page. Let's start to zoom in and see
what else we have.
Figure 2 - A Close Up Of The "Complex" Signal (Perhaps 60msecs)
Figure 2 above looks very similar to the signal analyzed on

the previous page except that the gear mesh amplitude
actually drops to about 0 once per shaft revolution - there is
more gear mesh amplitude modulation than we previously
had - so we would expect to see a similar FFT except that
the sidebands would be higher and possibly greater in
number. Let's look at the individual signals that went to
create Figures 1 and 2:
Figure 3 - The "Complex" Signal Components Separated
It is now clear that the 1x rpm signal is nowhere near a pure

(clean) sinusoid ? Note how wide the trough of the wave
shape is compared to the peak. Something is causing the
'binding' action we have discussed (frequency modulation) -
at once per revolution (1x rpm). So we can expect to see 1x,
2x and 3x rpm peaks on the spectrum plus a peak at gear
mesh frequency with sidebands surrounding it at 1x rpm.
That's it, right ? To be on the safe side, let's look even closer:
Figure 4 - A Close Up Of The Gear Mesh Signal (~5 msecs)
We've been hasty again. It is now clear that the gears are
also binding momentarily as they go in and out of mesh -
perhaps they are incorrectly aligned (note that the seemingly
straight green line shown in Figure 4 is actually a small
portion of the 1x rpm signal). The gold signal shows about 3-
1/2 cycles of the gear mesh signal. Clearly, the slope of the
gear mesh signal is steeper in the positive going direction
than in the negative going direction - more frequency
modulation but this frequency modulation is for the gear mesh
signal. Now let's look at the resultant FFT:
Figure 5 - The FFT Generated From Figure 1
We have:
 1x, 2x, 3x rpm peaks
 1x, 2x, 3x gear mesh frequency peaks
 Sidebands around gear mesh frequency harmonics
spaced at 1x rpm
How many mechanical vibration generators are present ? Two:

 1x rpm
 1x gear mesh
Of 12 identifiable peaks on this spectrum, 2 are actual

mechanical vibrations. The FFT has 'created' the other peaks
due to the specific shapes of the signal being processed. The
FFT sees things that we cannot see except in its detailed and
precise mathematical analysis of the data. It also makes
judgements that may be faulty (creating peaks) due to lack of
resolution and data accuracy.
But in spite of that, what kind of action(s) might be performed
if the amplitudes of the peaks shown were considered
unacceptable ?
 Balance (high 1x rpm peak)
 Alignment (harmonics of 1x rpm)
 Mechanical looseness (harmonics of 1x rpm)
 Gear eccentricity and settings (due to sidebands around
gear mesh harmonics)
 Gear wear and alignment (due to gear mesh harmonics)
These are only some of the most likely problems but you will
most probably find the problem during the course of this
investigation. Also, consider that we are assuming with our list
here that it is not just one or two of the amplitudes to be
excessive but all of them. That is quite unlikely.
You are far more likely to be mislead by spectrums if you are
ignorant
of the FFT process than if you have a proper understanding of
it.
Where Time Domain Plots Can Be Invaluable

Obviously, time domain is a powerful tool. But what kinds of
problems and situations are better analyzed with the time
domain plot ?
 Slow Beat  Chipped / Broken Gear

Frequencies Teeth
 Impacts  Amplitude Modulation
 Transients  Frequency Modulation
 Rubs  Slow Speed Bearing Defects

Additionally, you can gather information related to the machine
condition such as:
 Misalignment  Eccentricity
 Looseness  Waveform Shape
 # of Teeth on a Gear  Instabilities

Let's see how the time domain can be of help with some of
these.
Using Time Domain Plots To Find:
Beat Frequencies
A beat frequency occurs when two closely matched frequencies
occur with similar amplitudes. Figure 1 shows an uneventful
spectrum with a resolution of about 25 cpm per line.
Figure 1
But as we discussed in the 'Frequency' section, that spectrum

resolution (25 cpm/line) will be insufficient to separate peaks
less than about 50 - 60 cpm apart - the FFT will combine those
frequencies into a single peak. Figure 2 shows the signal from
which Figure 1 was generated.
Figure 2
We can see from Figure 2 that the beat is occurring over about
4000 msecs (4 seconds) which means it is occurring 15 times
per minute. With a beat frequency such as this, chances are
you would be able to hear and/or feel the beat occurring. By
timing it, you can find the 'beat' rate (15 beats per minute).
From the 'Frequency' section, we know that we need a
spectrum resolution of about 1/3 of that beat - 5 - 6 cpm/line
of resolution - to be able to separate the peaks on the FFT.

Impacts Created by Bearing Defects
An impact can occur due to a number of different problems -
bearing defects being #1 on the list (most common and most
important). But because of exactly how the FFT works, the
impact frequency gets filtered out of the displacement, velocity
or acceleration spectra. Figure 1 shows a outer race defect
occurring on a bearing rotating about 1200 rpm.
Figure 1 - A Signal Showing An Impact Occurring
But why does the impact frequency get filtered out ? It is

simply because there is no sinusoid at the impact frequency
(no sinusoid associated with the rate at which the impacts are
occurring). The only sinusoids present are:
 The 1x rpm sinusoid (we see 3+ cycles of that);
 The sinusoid associated with the ringdown occurring just

after the impact occurs (similar to a bell ringing down
after being struck).
So how does the FFT process the signal shown ? The 1x rpm
sinusoid is no problem but how does it account for the periodic
impacts occurring ?
Impacts Created by Bearing Defects
Figure 1 - A Time Domain Signal Showing Bearing-Related Impacts

Approximately 3 Impacts Occur For Each Shaft Revolution
Even though there is no sinusoid associated with the impact

frequency (about 3 impacts per revolution of the shaft), note
that there is a short-duration sinusoid occurring immediately
following the impact. In fact, it is generated by the impact
itself.
When you strike (impact) a bell, what happens ? It rings
(down) until it dampens out and stops ringing (or until it is
struck again). This is also known as "free" vibration. The same
effect occurs when the bearing component is struck. Notice in
Figure 1 that for each impact, there are several high frequency
(closely spaced) peaks. That is the "ringdown" of the bearing
components and assembly after the impact has occurred. Just
as the bell rings at its resonant (natural) frequency, the
bearing will ring down at its resonant frequency. THAT will
typically be found by the FFT process because it is sinusoidal.
However, that ringdown sinusoid is not a continuously
occurring frequency - it is transient (it comes and goes). How
does the FFT explain the spike suddenly appearing, quickly
ringing down and then disappearing until the next spike
(impact) occurs ?
What will show up on the spectrum is the 1x rpm peak and the
harmonics of the bearing defect frequency that occur closest to
the ringdown frequency of the bearing components. Those
peaks will be at relatively high frequencies since they are
associated with the ringdown frequency of the bearing
components (and that is a high frequency). Figure 2 shows the
FFT generated from the signal shown in Figure 1.
Figure 2 - The Resulting FFT From The Signal In Figure 1
Notice the lack of any peak near the impact (defect) frequency
(about 3.1 x rpm).
The use of time domain for bearing defects is particularly
useful for slow speed equipment. A spectrum will often provide
no warning or very late warning of a defect developing.
Please note that this subject is covered far more in-depth in
the 'Enveloping Spectra Plots' section.

Impacts Created by Gear Teeth
What is the result of a single broken, cracked or chipped gear
tooth ? It is an impact once per shaft revolution (1200 rpm).
What does the FFT look like ? Figure 1 shows the time domain:
Figure 1 - Shows Once Per Revolution Impact
Figure 2 shows the FFT generated from the signal seen in

Figure 1.
Figure 2 - The Resulting FFT Plot
The spectrum shown goes out to about 22,000 and the number
of teeth is 25. The amplitude at 30,000 cpm (GMF - which is
not shown on the plot) is increased on slightly under this
circumstance. So what is the FFT symptom you can expect
with this problem ? The time plot shows a peak amplitude for
for the 1x rpm signal of perhaps 0.8 ips (20 mm/s). The FFT
amplitude is under 0.5 ips (12 mm/s). There may be a slight
increase in GMF.
Just as impacts in a bearing excite the natural frequencies of
the bearing structure, there will almost certainly be some
excitation of the gear train's natural frequency due to the
impacting. That is an important clue and typical of excessive
wear or impacting between gears but unfortunately you won't,
in all likelihood, know what that natural frequency is. Therefore
...
The only effective way to detect this problem from its early
stages is with time domain analysis.

Rubs
What does a rub cause on the spectrum ? Well, it is quite
unpredictable but can, because of the wave shape, cause
extensive harmonics, half-harmonics (0.5, 1.5, 2.5, etc. x
rpm) or even sub-harmonics. But what does the time domain
look like ?
Figure 1
Once the rotor contacts the side (begins the rub), it is

prevented from moving any further in that direction. It will rub
until the forces present pull it away from the contacted surface
in the other direction. This is another example of a problem
where the diagnostic capabilities of the time plot far exceed
those of the spectrum. For instance, the length of the flat
portion (the rub) relative to the length of the entire cycle will
tell you how much of the rotation is rubbing. You should note
that this condition would not be apparent if the was mounted
horizontally because the rub is strictly in the vertical direction.
This is also known as a TRUNCATED wave shape and a rub is
one of the
problems that would generate such a wave shape.

Truncated Wave Shapes
Since we've touched upon the truncated wave shape let's consider
the wave shape shown in Figure 1. What is happening ? A great deal
of movement in the positive going direction (amplitude reaches
about 1.0) and not much in the negative going direction (amplitude
reaches about -0.2). Can you think of any conditions that could lead
to a signal shape like this ?
Figure 1 - A "Truncated" Signal Shape
In case you aren't quite sure yet, let's sum up what we can see from
the wave shape:
 The '+' peak amplitude = 1
 The '-' peak amplitude = - 0.2
 The wave is 'cut off' (truncated) on the bottom.
 There is a little bounce or bump at the bottom.
Let's take a look at the FFT generated by the signal in Fig. 1 in case
you would like that to help with your analysis:
Figure 2 - The FFT Generated From The Signal Shown In Figure 1
The FFT shows a peak at 1x rpm (about 1800 cpm) and 2x rpm
(about 3600 cpm). There is nothing at higher frequencies.
Perhaps the animation on the next page will help.

Truncated Wave Shapes
Hindsight being 20/20, you can see that the time wave shape
was telling you how the bearing is moving. The fact that the
bearing is moving a great deal in one direction and not the
other is a valuable piece of information on how the bearing is
moving if you put it to use in your mind. Figure 1 shows a
bearing moving in this manner.
Figure 1 - One Possible Cause Of The

Previous Page's Signal Shape
Ok, ok, the animation is a bit exaggerated. An unbalance force

(the yellow ball on the rotor you see flashing by) is present. If
properly fastened in place, this (centrifugal) force would be
sufficient to move the bearing housing a certain amount (far
less than what you see). Since, however, the bearing is not
properly fastened in place, the force is able to move the
bearing a great deal further. The impact of the rotor dropping
back onto its base causes a bit of a bounce.
But the wave being shown in Figure 1 is actually very close in
shape to that in Figure 1 on the previous page - the sharper
edges on the previous signal shape are simply due to a loss of
resolution (longer time sample, same number of data bytes).
Figure 2 (below) shows the signal from the previous page in
much greater detail - only a couple of cycles are being visible
(better resolution). The signal shape is virtually identical to the
signal shape in the animation.
Figure 2 - Small Segment Of Signal Seen On Previous Page;
Approx. 70msecs Versus 1200msecs On Previous Page
One of the key methods used in time domain analysis is to use

the wave shape to see (in your mind) how the bearing or
structure is actually moving and using that to think about what
might be happening.
What You Need To Do To Be Able

To Analyze Time Domain Plots
It is extremely important to understand the limitations of the
FFT and the unpredictability of the FFT process when several
problems are present simultaneously. The time domain plot
should be used whenever applicable or in the presence of a
stubborn or unusual problem. But there are three things you
must do to get comfortable with and good at analyzing time
domain plots.
Practice
PRACTICE
PRACTICE
Well, that's not entirely true - you also need to understand
how to set them up. We can't help you with the practice part
but we can help you with the setup.
Setting Up The Parameters For A

Time Domain Plot
To set up a time domain reading can be a bit cumbersome.
This is mainly due to the fact that the setup is often done using
FFT parameters such as Fmax and lines of resolution. This
section will first explain how a time domain is set up and then
provide some easy to use examples. The time domain
equivalent to Fmax and lines of resolution are:
 Fmax = Period (length of time sample being collected)
 Lines of Resolution = Bytes (how many pieces of data are
collected to create the sample)
o 512 Bits is equivalent to 200 lines of resolution
o 1024 bits is equivalent to 400 lines of resolution
As you may already know, a time domain plot is just as

susceptible to resolution limitations as an FFT is. Figures 1, 2
and 3 are each from the same time domain plot - the latter
two are zoomed in on.
FIGURA 3
Figure 1 Figure 2
The plot shown in Figure 1 - an actual, real-life time domain plot
- was collected with 2048 amplitude values (the time domain
equivalent of "lines of resolution"). The length of the time sample
is 0.114 secs.
Figure 2 shows a portion of the same time domain plot with the
sample reduced to 0.04 seconds (by zooming in). This is done in
the same manner as one would zoom in a an FFT. It still looks
pretty good but just as with a spectrum, zooming in has done
nothing to improve the accuracy of the data. Once collected, you
can never improve or in any way change the accuracy of any plot
- time domain or spectrum. The resolution is dictated by the
parameters set up and cannot be altered after the fact.
Figure 3 shows the same plot with the sample reduced to only
0.01 seconds. It is now quite clear that the time domain plot is
generated by compiling a series of amplitude values and
connecting them with lines - the same way an FFT is generated.
This is JUST AS IMPORTANT and must be stressed just as much
as with the spectrum. Zooming in to this level does nothing to
improve the resolution and is about as helpful in viewing the big
picture as looking at a forest with your face 2 inches from a
particular tree would be - in other words, not helpful at all.
So how do we set up a time domain reading using FFT
parameters from this information ?

Time Domain Plot
The first thing we need to do is figure out how long (in
seconds) our time sample needs to be. How do we do that ?
Well, it depends on what we are trying to analyze. Let's take a
machine rotating at 3600 rpm (60 Hz). How many shaft
rotations do you want to see in your time sample ? If you said
5 - 7 for a normal analysis, go to the head of the class. So
we'll shoot for 6 shaft rotations:
 6 rotations / 60 rotations per second = 0.10 seconds
We now have our desired time sample - 0.1 seconds (100

msecs). We also have a formula for future reference (for the
sake of consistency, we'll call the rotation an "event"):
 # of events desired / # of events per second = time
sample desired (seconds)
By using events, we can more easily apply the formula to

anything from gear mesh to bearing defects - not just shaft
revolutions. Shaft revolutions will, however, be the most
common 'event'.
At this point, there are two possibilities which will determine
how you proceed:
 Your analyzer or software requires a time sample length
and the number of data bits desired. If this is your
option, you're about done. Simply choose 0.10 seconds
(or 100 msecs) for the sample length or period and the
corresponding data bits for the number of amplitude
values on the plot you want (512 for 200 lines, 1024 for
400 lines, 2048 for 800 lines, 4096 for 1600 lines). It is
recommended you collect either 2048 or 4096 bits of
data.
 Your analyzer or software forces you to set up the
reading in FFT parameters. In this case, you have a bit
more math to do.
For option 2, choosing the number of lines of resolution we

want is straight forward - simply select the number you want.
It is recommended that you use 800 lines as the minimum and
we will use that in our example. Once you've decided on the
desired resolution and you know the time sample you want,
use the following formula to find the Fmax you must select:
 Fmax = # Lines / Time Sample
 Fmax [Hertz] = 800 / 0.1 seconds = 8000 Hz
 Fmax [CPM] = 8,000 Hz x 60 = 480,000 cpm
If you want 1600 lines with the same length time sample, you
would use:
 Fmax = 1600 / 0.1 seconds = 16,000 Hz x 60 = 960,000
cpm
If you want 400 lines with the same length time sample, you
would use:
 Fmax = 400 / 0.1 = 4,000 Hz x 60 = 240,000 cpm
Note that you can generate the Fmax in cpm directly by using
the # lines x 60 and dividing it by the desired time sample:
 800 lines x 60 / 0.1 seconds = 48,000 / 0.1 = 480,000
cpm
Also note that the shorter the time sample desired or greater
the resolution, the higher the Fmax selected.

Time Domain Plot
Let's run another example where we want to capture 10
bearing defect impacts on a shaft running 1200 rpm. Well, first
let's convert to Hz: 1200 cpm = 20 Hz. Next, we need to know
the defect frequency. For the example, we will use the very
common outer race defect frequency found just over 3x rpm.
That means we will need approximately 3 shaft revolutions to
capture 10 impacts. 20 shaft revolutions per second and we
want 3 - that's 150 msecs (0.15 secs). We'll stick with our 800
lines and go straight to the cpm answer:
 Fmax = 48,000 / 0.15 = 320,000 cpm
How about if we wanted to see 6 revolutions of a shaft turning

at 60 rpm (1 Hz). Well, the fact that you want to see 6
revolutions on a shaft going 60 rpm should tell you that you
want 1/10th of a minute - 6 seconds. Again, we'll stick with
800 lines.
 Fmax = 48,000 / 6 = 8,000 cpm
The table shown below shows some time domain setup

parameters. The table values assume 800 lines of resolution
(2048 data bits).
By using 800 lines and the Fmax shown, you will obtain a time
sample that contains 5 - 10 revolutions of the shaft providing
the machine is in the RPM range shown. In other words, if you
have a machine running 1500 rpm and you want about 7
revolutions of the shaft on your time domain plot, find the
'RPM Range' below that contains 1500 (1314 - 1838) and use
the Fmax shown (180,000 cpm or 3kHz) and 800 lines.
Fmax RPM Range

1,500 1 - 20
3,000 20 - 39
6,000 40 - 79
12,000 80 - 131
18,000 132 - 184
24,000 185 - 236
30,000 237 - 289
36,000 290 - 368
48,000 369 - 473
60,000 474 - 656
90,000 657 - 919
120,000 920 - 1313
180,000 1314 - 1838
240,000 1839 - 2363
300,000 2363 - 2888
360,000 2888 - 3413
420,000 3413 - 3938
480,000 3938 - 4725
600,000 4725 - 5775
720,000 5775 - 7875
1,080,000 5776 - 11288
1,500,000 11288 - 16000
Note that the Fmax's shown in the table above are based on
800 lines.
 If you want to use 1600 lines (4096 bits), double the
Fmaxs used for each speed range.
 If you want to use 400 lines (1024 bits), cut in half the
Fmaxs used for each speed range.
 If you are looking for an event occurring more than once
per revolution (e.g. bearing defects), use the next highest
frequency range listed.
 If you are looking for a "beat" frequency, use the Fmax

listed next to the beat frequency rate (a very low Fmax) -
not the shaft rpm.
Time Domain Plots:

Summary
Time domain analysis is a powerful but intimidating tool.
Hopefully, this section of the manual has helped you
understand some of the secrets of the time domain as well as
some of the secrets of the FFT process. Again, there are
specific areas where we recommend using time domain
analysis without exception:
 Slow Speed Equipment (< 300 rpm)
 Sleeve Bearings (particularly if readings reflect true shaft
movement)
 Gear Applications
However, there are many people who would argue that time
domain is a valuable tool on all applications and we cannot
argue with them - the wave shape can provide information
that you will not get from an FFT. If you are comfortable with it
and have the time to collect it, by all means - it is another way
to look at how your machines are behaving.
We also cannot argue with the people who claim:
 Time Domain is complicated
 Time Domain is difficult to interpret
 Often you can't make head nor tail of a time domain plot
Each of these is also true. So . . . . the sooner you start using

it and getting comfortable with it, the sooner you'll become
proficient at using this powerful tool. Remember the three
things you need:
 PRACTICE
 PRACTICE
 PRACTICE

Vibration School

Uploaded by

Copyright:

Available Formats

Vibration School

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Vibration School

Uploaded by

Copyright:

Available Formats

What are some common causes of mechanical vibration?

What are some common causes of mechanical vibration?

What characteristics of vibration are measured?

What characteristics of vibration are measured?

VIBRATION SCHOOL

Vibration is a "back and forth" movement of a structure. It can also be

What Is Vibration Caused By ?

Why Do We Measure Vibration ?

What Does The Transducer Actually Detect ?

Actual Bearing A Transducer Mounted A Transducer Mounted

What Vibration "Characteristics" Do We

How Much Movement Occurs

What Vibration "Characteristics" Do We

Frequency Tells Us:

How Often The Movement Occurs

What Vibration "Characteristics" Do We

In What Direction Is The

What Are "Conventions" ?

Common Bearing Numbering Conventions

The image below shows a typical 4-bearing, belt driven fan

Common Bearing Numbering Conventions

What about components that have more than 2 bearings

Common Position Naming Conventions

Consistency Is The Key

Common Direction Naming Conventions

Axial Direction is always, Always, ALWAYS parallel to shaft

Let's start with horizontal, direct drive machines. These

Common Direction Naming Conventions

Belt drives can be oriented in any direction and thereby require

For these reasons, a convention is used that defines "parallel

Common Direction Naming Conventions

Vertical machines present another opportunity to assign a

Before we answer that question, let's keep in mind what

What Does Amplitude Tell Us ?

 The 'amplitude' is a measure of the amount of

Simply put, it measures:

What Are Amplitude "Units" ?

To understand the second amplitude unit, you must first

To understand the third amplitude unit, you must understand

 The bearing is 1/4 of the way through the cycle (1/4 of a

Measuring "How Much" in Displacement Units

 The bearing is now passing the '0' position again moving

 The red ball has completed 1/2 cycle (1/2 shaft

The red ball has completed 3/4 of a cycle (3/4 of a shaft

Measuring "How Much" in Displacement Units

 The bearing is now back where it started, having

 Stress failures are most likely to occur on slow speed

The "Displacement Sine Wave"

English or 'Imperial' Units: Mils (1 mil = 0.001")

Measuring "How Much" In Velocity Units

 Since the velocity is constantly changing over time,

Measuring "How Much" In Velocity Units

 From this moment until the bearing passes the "at-rest"

The bearing is now passing the "at-rest" position as it moves

 From this moment until the bearing reaches the "+X"

Measuring "How Much" In Velocity Units

 From this moment until the bearing passes the 'at-rest'

Measuring "How Much" In Velocity Units

The bearing is passing the "at-rest" position again moving

Measuring "How Much" In Velocity Units