B5 The Orbit.

qxd 9/10/2002 10:35 AM Page 69

69

Chapter 5

The Orbit

I     ,   how rotor dynamic motion

along the measurement axis of a single transducer can be displayed as a time
varying waveform. While the timebase plot can provide important and useful
information, it is inherently limited to one dimension of rotor motion. Since, in
any lateral plane along the rotor, the rotor moves in a two-dimensional path, or
orbit, this one-dimensional picture provided by a single transducer is not ade-
quate.
To measure this motion, a second transducer must be installed perpendicu-
lar to, and coplanar with, the first transducer. Only then will there be enough
information to observe the complete motion of the rotor in that plane. This
information could be presented on two, one-dimensional timebase plots, but it
would be even better if we could display the two-dimensional dynamic motion
of the rotor. That is the purpose of the orbit plot.
The orbit represents the path of the shaft centerline relative to a pair of
orthogonal eddy current transducers. These transducers are usually mounted
rigidly on the machine casing near a bearing; thus, the orbit typically represents
the path of the shaft centerline relative to the bearing clearance of the machine.
(Orbits can be constructed from casing vibration data, but this has limited
application. This chapter will present orbits from shaft relative (displacement)
probes.) Because of its ease of interpretation and extensive information content,
the orbit, with Keyphasor mark, is probably the most powerful single plot format
available to the machinery diagnostician.
In our discussion of the orbit, we will start with a description of how two
timebase waveforms are used to create the orbit. We will discuss the informa-
tion obtained when the Keyphasor mark is included on an orbit, and slow roll
B5 The Orbit.qxd 9/10/2002 10:35 AM Page 70

70 Data Plots

and Not-1X compensation of orbits. Finally, we will discuss the various kinds of
information that can be obtained from an orbit.

The Construction of the Orbit

The orbit combines the timebase waveform data from two, perpendicular,
coplanar transducers to create a single plot showing the two-dimensional
dynamic motion of the shaft centerline (Figure 5-1). The orbit in the figure is
unfiltered, but is predominantly 1X. The data comes from XY transducers, which
observe the motion of the rotor. These transducers are mounted at 0° and 90° R
(relative to the reference direction, “Up,” also known as the orientation angle ref-
erence), but any transducer orientation is possible as long as the transducers are
perpendicular to each other. The signals from the transducers can be displayed
as two, independent timebase plots or can be combined to produce the orbit.
The Orbit plot format is a square, with identical horizontal and vertical
scales and scale factors. A point on the orbit is defined by a pair of X and Y val-
ues, which are obtained from the timebase waveform data. A set of values from
the sampled waveforms can create anything from a portion of the orbit to sev-
eral orbits. The center of the orbit plot is defined by the average values of the X
and Y timebase waveforms. A Keyphasor signal acts like a strobe: the dot shows
the location of the shaft centerline when the Keyphasor event occurs. To com-
plete the plot, a reference direction (for example, “Up” or “West”), the probe
locations, and the direction of rotation are included on the plot.
Note that the direction of rotation cannot be determined from the orbit
without additional information. The best way to determine the direction of rota-
tion is to examine the machine. The direction of rotation may be visible to the
eye, or it may be marked on the machine with an arrow. Another way is to use
slow roll orbits (with caution!), which are almost always forward; thus, knowl-
edge that the machine is at slow roll allows us to determine the direction of rota-
tion by observing the direction of precession. The direction of precession can be
determined by the blank/dot sequence on the plot. Be aware that crossed probe
wiring or an active rotor malfunction that causes reverse precession at slow roll
can render this method invalid.
Positive voltage or position changes on the timebase plot correspond to a
motion component toward the transducer. On the orbit, this positive change
always corresponds to motion toward a transducer, along its measurement axis,
regardless of the transducer orientation. Since the orbit is much smaller than
the transducer, the measurement is always made parallel to the transducer
measurement axis.
In Figure 5-1, points 1 through 5 show the progression of the shaft center-
line around its orbit. Point 1 shows the location of the shaft centerline
B5 The Orbit.qxd 9/10/2002 10:35 AM Page 71

Chapter 5 The Orbit 71

(crosshairs) when the Keyphasor event occurs; that is, when the leading edge of
the Keyphasor notch passes next to the Keyphasor probe (shown below the
rotor).
Points 2 and 4 mark the farthest and closest approach to the X probe (the
minimum and maximum peaks on the X timebase plot). Similarly, points 3 and
5 mark the farthest and closest approach to the Y probe (the minimum and max-
imum peaks on the Y timebase plot).
Often, several cycles of vibration are plotted on the orbit. In the figure, two
cycles are shown in the timebase plots, which means that the orbit also has two
cycles and, in this case, overwrites itself during the second cycle.
The positive peak of the timebase plot always represents the closest
approach of the shaft to the associated transducer. For 1X vibration, the point
on the shaft which is on the outside of the deflected shaft is called the high spot.
The positive peak in a 1X-filtered displacement signal represents the passage of

Up
5
Rotation
Up 1 direction Probe
2 Up
locations
4 Y
Y
5
Rotation 3
direction Down 1

Right
2
5 X
Left

1 X
2 4
4

3
Right
4 3

KØ 3 Down
5
1 50 µm/div
2

Left

Figure 5-1. Construction of an orbit. XY transducers observe the vibration of a rotor shaft. A notch
in the shaft (at a different axial location) is detected by a Keyphasor transducer. The vibration
transducer signals produce two timebase plots (middle) which combine into an orbit plot (right).
At position 1, when the Keyphasor probe detects the notch, the shaft centerline is located at the
position of the Keyphasor dot. Positions 2 through 5 show the correlation of the orbit position
and the timebase plot values. Two revolutions of data are shown in the timebase plots, but only
one can be seen on the orbit because, in this case, the shaft follows the same path.
B5 The Orbit.qxd 9/10/2002 10:35 AM Page 72

72 Data Plots

the high spot next to the displacement transducer, and this concept extends to
the orbit. For a 1X-filtered, circular orbit, the orbit represents the path of the
rotor high spot, as well as that of the shaft centerline.
Transducer mounting orientations are usually measured relative to the ref-
erence direction for the machine. For a horizontal machine, the reference direc-
tion is usually “Up.” For vertical machines, the reference direction can be any
convenient reference, for example, “North.” Bently Nevada orbits are always
plotted with the reference direction at the top of the plot. The actual transduc-
er locations are indicated on the edge of the plot. This provides a uniform visu-
al reference along the machine train, regardless of transducer mounting orien-
tation. The orbit on the plot is oriented as an observer would see it when posi-
tioned with his head in the reference direction, looking along the axis of the
machine in the viewpoint direction. (By Bently Nevada convention, this view-
point is usually from the driver toward the driven machine.)
Figure 5-2 shows two examples of an orbit display with different transducer
orientations. In both cases, the rotor orbit is the same, only the transducer
mounting orientations are different. The orbit plots show the same orbit orien-
tation relative to the “Up” reference. Note that the probe labels on the orbit plots
show the actual probe mounting orientations.
At the bottom of the figure, equivalent oscilloscope orbits are shown.
Because the XY axes of the oscilloscope on the right do not match the actual
transducer locations, the oscilloscope must be physically rotated 45° CCW to
display the orbit with the correct orientation. In this orientation, the horizontal
and vertical oscilloscope axes are aligned with the actual transducer orienta-
tions. It is important to remember that, when viewing orbits on an oscilloscope,
the XY axes of the oscilloscope must match the actual transducer mounting ori-
entations, or the displayed orbit will be rotated and will not appear as it does on
the machine.
Filtered orbit plots, like filtered timebase plots, are not constructed directly
from waveform data. A filtered timebase waveform plot is constructed (synthe-
sized) from a filtered vibration vector; a filtered orbit is constructed from the
synthesized waveforms from a pair of vibration vectors. See Appendix 2 for
details of this process.

The Keyphasor Mark

The blank/dot sequence on the orbit is called a Keyphasor mark. The mark
represents a timing event, the Keyphasor event, that occurs once per shaft revo-
lution. The timing signal comes from a separate (Keyphasor) transducer that is
mounted at a different axial location. The timing signal is combined with the
orbit so that the timing of the Keyphasor event can be seen clearly. The
B5 The Orbit.qxd 9/10/2002 10:35 AM Page 73

Chapter 5 The Orbit 73

Up Up

Y Y X

Up Up
Y Y X

Orbit plot Orbit plot

Os
cil
lo
Up

sc
e

op
Y op

e
X
sc

Ri
lo

gh
cil

t
Os

Oscilloscope display Oscilloscope display

Figure 5-2. Probe orientation and the orbit plot. On the left side, the Y and
X probes are mounted at 0° and 90° R, respectively. The orbit plot and oscil-
loscope display show the same view. On the right, the probes are mounted
at 45° L and 45° R. The orbit plot is automatically rotated so that it is prop-
erly aligned with the “Up” reference direction. The oscilloscope, however,
must be physically rotated 45° CCW to display the correct orbit orientation.
Note that the Keyphasor marks on the orbit plots show the plot conven-
tion, where the dot marks the event; the bottom displays show the event
at the beginning of the blank, one possible oscilloscope display.
B5 The Orbit.qxd 9/10/2002 10:35 AM Page 74

74 Data Plots

Keyphasor mark on the orbit shows the location of the shaft centerline at the
instant when the once-per-turn mark on the shaft passes next to the Keyphasor
probe. On circular, 1X-filtered orbits, the Keyphasor dot marks the location of
the rotor high spot at the instant of the Keyphasor event. The blank/dot
sequence shows the direction of increasing time.
Figure 5-3 shows an orbiting rotor (the size of the orbit is greatly exaggerat-
ed) that is viewed by two orthogonal transducers. As the rotor rotates, the shaft
centerline also moves (precesses) along a path which defines the orbit. A
Keyphasor probe is installed to detect a once-per-turn mark on the shaft, in this
case a notch. When the leading edge of the notch passes next to the Keyphasor
probe (position 3 in the figure), the shaft centerline is located at the Keyphasor
dot on the orbit. The Keyphasor signal is like a strobe that briefly illuminates the
shaft as it travels in its orbit. Even though, in the figure, the rotor is on the oppo-
site side of the orbit from the Keyphasor probe, the notch on the shaft is in a
position to be sensed by the Keyphasor probe.
On all Bently Nevada plots, the Keyphasor event is shown as a blank/dot
sequence (as shown in Figure 5-2 in the orbit plots), and the dot represents the
instant that the Keyphasor event occurs. The Keyphasor mark on an oscillo-
scope, however, may be a blank/bright or bright/blank sequence depending on
the type of shaft event and the type of oscilloscope used. Figure 5-2 (bottom)
shows a blank/bright sequence; the beginning of the blank marks the event.
In timebase plots, the time between two Keyphasor marks represents one
revolution of the shaft. In orbit plots, the rotor moves along the path between
two Keyphasor marks during one revolution of the shaft. This path may be quite
complicated. A Keyphasor mark will be plotted every time the rotor completes
one revolution. If several revolutions of data are plotted on an orbit (Figure 5-4),
several Keyphasor marks should be visible. However, in nX-filtered orbits, where
n is an integer, successive Keyphasor marks will plot on top of each other. In the
figure, the orbit on the right is an unfiltered orbit, but consists primarily of 1X
vibration. The Keyphasor dots plot almost on top of each other.
The Keyphasor mark on an orbit plot adds important information. It can be
used to determine the instantaneous direction of motion of the rotor and to
estimate the absolute phase, the vibration frequency in orders of rotative speed,
and, with multiple orbit plots, the mode shape of the rotor.

Compensation of Orbits
As we discussed in the chapter on timebase plots, the primary objective of
compensation is to remove unwanted signal content (noise) that is unrelated to
the machine behavior that we want to observe. Like timebase plots, both filtered
and unfiltered orbits can be compensated.
B5 The Orbit.qxd 9/10/2002 10:35 AM Page 75

Chapter 5 The Orbit 75

Y X Y X Y X Y X
1 2 3 4

Keyphasor
KØ KØ event KØ KØ

Y X
3

Figure 5-3.The motion of the shaft around a 1X, circular orbit. The Keyphasor event occurs
at position 3, when the Keyphasor probe detects the notch. At that instant, the shaft center-
line is located at the position of the Keyphasor dot in the orbit plot.

Up Up
Y Y

X X

0.2 mil/div 14.3 krpm 0.2 mil/div 14.3 krpm

Figure 5-4. Two unfiltered orbits from a 14 ,300 rpm compressor. Each orbit plot shows
eight revolutions of data. The Y probe is mounted at 22° L, and the X probe is mounted at
68° R. These directions are indicated by the small, black labels on the plot. The left plot
shows multiple Keyphasor dots because of a subsynchronous frequency component due
to a fluid-induced instability. The right orbit shows predominantly 1X behavior at a time
when the instability is absent or very small. Note the clustering of Keyphasor dots into one
group, indicating dominant 1X behavior.
B5 The Orbit.qxd 9/10/2002 10:35 AM Page 76

76 Data Plots

Filtered orbit plots can be slow roll compensated using 1X, 2X, or nX slow
roll vectors. While compensation of a timebase plot requires one slow roll vec-
tor, compensation of an orbit plot requires a pair of slow roll vectors. Each trans-
ducer has a slow roll vector that is subtracted from its original vibration vector;
the resulting pair of compensated vibration vectors is used to synthesize the fil-
tered orbit. The end result is a filtered orbit plot that is slow roll compensated.
Figure 5-5 shows plots of an uncompensated (left) and compensated (right)
1X-filtered orbit. The slow roll vectors are relatively large and in phase with the
vibration, so that, after compensation, the orbit is significantly smaller. Also, the
change in position of the Keyphasor dot shows that the phase of the orbit has
significantly changed, moving the observed high spot location about 45° coun-
terclockwise. This is a significant difference that is important for efficient and
accurate balancing.
Another type of compensation, waveform compensation, can be applied to
the unfiltered orbit. Each unfiltered timebase waveform used to construct an

1X uncompensated 1X compensated

X Vector: 1.87 mil pp ∠320˚ X Vector: 0.54 mil pp ∠277˚

Y Vector: 1.95 mil pp ∠240˚ Y Vector: 0.55 mil pp ∠189˚
Up Up
Y Y

X X

0.2 mil/div 1200 rpm 0.2 mil/div 1200 rpm

Figure 5-5. Slow roll vector compensation of a 1X-filtered orbit. The

orbits are constructed from the pairs of vibration vectors shown above
the plots. Slow roll vectors of X = 1.2 mil pp ∠324° and Y = 1.4 mil pp
∠231° are subtracted from the original vectors to produce the vectors
used in the compensated plot. Note the significantly smaller size of the
orbit and the change in position of the Keyphasor dot by more than 45°
counterclockwise from the uncompensated position. This correction pro-
duces an orbit that displays only the dynamic response of the system
and is very important for efficient and accurate balancing.
B5 The Orbit.qxd 9/10/2002 10:35 AM Page 77

Chapter 5 The Orbit 77

orbit consists of a sequence of digitally sampled values. For each transducer, a

suitable slow roll waveform is selected from the slow roll speed range. Each of
the slow roll sample values can be subtracted from its corresponding value in
the original waveforms (the Keyphasor event is used as a waveform timing ref-
erence). This method has the advantage of removing most, if not all, of the slow
roll component of the signal.
Waveform compensation will remove all components with frequencies up to
the Nyquist sampling frequency limit (½ the sampling rate). Thus, 1X, 2X, ..., nX
(n an integer), and all subsynchronous and supersynchronous frequencies (to
the Nyquist limit), will be removed from the orbit. This includes most of the sig-
nal artifacts due to shaft surface defects.
Figure 5-6 shows how waveform compensation works. At the left, two
uncompensated timebase plots from a steam turbine are combined to produce

Uncompensated Slow roll waveforms Compensated

Y Y Y

3600 rpm 138 rpm 3600 rpm

X X X

3600 rpm 138 rpm 3600 rpm

X X

Y Y

3600 rpm 3600 rpm

Figure 5-6. Slow roll waveform compensation of an orbit from a steam turbine. At left, two
uncompensated waveforms are used to construct the uncompensated orbit. Two slow roll wave-
forms are subtracted from the original waveforms to produce the compensated waveforms and
orbit at right, which are much clearer.
B5 The Orbit.qxd 9/10/2002 10:35 AM Page 78

78 Data Plots

an uncompensated, and rather noisy, orbit. To compensate the orbit, slow roll
waveforms are subtracted point-by-point from the original uncompensated
waveforms. The result is a pair of compensated waveforms at the right, which
produce the compensated orbit at bottom right. The shape of the orbit is much
smoother, less noisy, and more indicative of the dynamic response of the rotor.
Unfiltered orbits can also be notch filtered by compensating both of the
original waveforms with a synthesized, filtered waveform. Each compensation
waveform is reconstructed from a nX-filtered vibration vector that is sampled at
the same time as the original waveforms were sampled. The reconstructed, nX
waveforms are then subtracted from the original vibration waveforms, and the
two resulting waveforms are used to construct the Not-nX orbit.

Uncompensated 1X-filtered Not-1X

Y Y Y

4510 rpm 4510 rpm 4510 rpm

X X X

4510 rpm 4510 rpm 4510 rpm

Y Y

X X

4510 rpm 4510 rpm

Figure 5-7. Not-1X compensation of an orbit. At left, two uncompensated waveforms are used to
construct the uncompensated orbit. 1X waveforms, synthesized from vibration vectors at running
speed, are subtracted from the originals to produce the Not-1X waveforms and orbit at right.
These waveforms and orbit contain all frequencies except 1X. The remaining vibration is primarily
1/2X from a rub.
B5 The Orbit.qxd 9/10/2002 10:35 AM Page 79

Chapter 5 The Orbit 79

For example, using this technique, you can examine an orbit with all of the
1X vibration removed (Figure 5-7). A Not-1X orbit is created by subtracting two,
synthesized 1X waveforms from the original unfiltered waveforms. The result-
ant orbit reveals any frequency information that may have been obscured by the
1X response. This can be helpful for identifying vibration characteristics associ-
ated with a variety of malfunctions. In the figure, note that the ½X vibration,
which is the dominant remaining component, is clearly visible.
The warning about compensation mentioned in the last chapter is worth
repeating here: compensation is an art as well as a science. There are many vari-
ables that can change the compensation vectors or waveforms. It is possible, by
using incorrect compensation, to produce orbits that convey the wrong impres-
sion of machine behavior. Initially, it is always best to view data without any
compensation. Then, when it is used, compensation should always be done with
caution.

Information Contained in the Orbit

An orbit plot, especially with Keyphasor marks, is a powerful diagnostic
tool. It can be used to measure the amplitude of filtered or unfiltered vibration
in any radial direction; the relative frequency of filtered vibration versus rotor
speed; the relative frequency of X versus Y unfiltered vibration; and the direction
of precession. Using the orbit, the absolute phase of filtered vibration can be
estimated. Additionally, the shape of an unfiltered or filtered orbit can provide
important clues to the behavior of machinery, highlight significant changes in
response that one-dimensional timebase plots cannot, and help identify where
a problem may be occurring in relationship to the components of the machine.
Recall that timebase signals are plotted about a mean value and contain no
dc information. These plots contain only dynamic (ac) information. Because the
orbit is constructed from timebase plots, it also has no dc content. The orbit
only displays the shaft motion relative to the average position; there is no infor-
mation in the orbit about the average position of the rotor. To obtain the aver-
age shaft position, use the average shaft centerline plot (Chapter 6).
Multiple orbit plots can be created from the same location at different rotor
speeds to show evolution of rotor vibration over speed; or, they can be created
from different axial locations at the same speed to show the mode shape of the
rotor.
Like the timebase plot, the orbit has many of the characteristics of a basic
oscilloscope display. Both oscilloscope orbit displays and computer-based orbit
plots can be used to make a number of measurements. The following discussion
applies primarily to orbit plots, but it can be extended to oscilloscope orbit dis-
plays.
B5 The Orbit.qxd 9/10/2002 10:36 AM Page 80

80 Data Plots

While many software packages have the ability to process and display this
machinery vibration information, a machinery diagnostician should be able to
analyze the information without using a computer, for two reasons. First, an
oscilloscope, with an orbit display, may be the only instrument available.
Second, it is always good to verify the numerical results that come out of a com-
puter. A quick glance at the orbit plot can provide a useful sanity check on any
other numerical results.

One of the most basic measurements that can be made on an orbit plot is
the peak-to-peak amplitude of vibration. This measurement can be done on
either filtered or unfiltered plots. To measure the peak-to-peak amplitude
(Figure 5-8),

1. Select a transducer. In general, the amplitude will be different

for each transducer. In the figure, the X transducer is chosen.

2. Draw a line (the measurement axis) from the transducer loca-

tion (the X mark on the perimeter of the plot) through the cen-
ter of the plot. The line must extend well beyond the limit of the
orbit itself.

3. Construct two lines (3 and 4) that are perpendicular to the

measurement axis and tangent to the orbit at the maximum
and minimum peaks of vibration with respect to the transducer
(the red circles).

4. Measure the distance between the two tangent lines in a direc-

tion parallel to the measurement axis (a plot scale has been
added for convenience). Calculate the peak-to-peak amplitude
using Equation 5-1:

 units 
pp amplitude =(number of div pp)
 div  (5-1)

The peak amplitude is one-half of the peak-to-peak amplitude.

In the figure, there are a just under 12 divisions between the two measure-
ment lines. Applying Equation 5-1, the peak-to-peak amplitude of the orbit as
viewed by the X probe is
B5 The Orbit.qxd 9/10/2002 10:36 AM Page 81

Chapter 5 The Orbit 81

 5 µm 
pp amplitude =(12 div pp)  = 60 µm pp (2.4 mil pp)
 div 

There are two key points to remember when using this technique. First, the
peak-to-peak measurement must be made parallel to the measurement axis of
the probe. Measuring simply vertically or horizontally, in this case, would pro-
duce a different and incorrect result.
Second, the peak-to-peak measurement is made between the tangents to the
orbit that are also perpendicular to the measurement axis of the probe.
Remember that the orbit is small compared to the transducer size, so that the
measurement axis is always in the same direction. It will help to remember that,
on the scale of the plot, the face of an eddy current transducer would be about
1 m (3 ft) in diameter!

5
Y X

1
Figure 5-8. Measurement of peak-to-peak ampli-
tude on an orbit. Here, the X transducer measure- 2
ment axis is drawn together with perpendicular
lines that are tangent to the maximum and mini-
mum points on the orbit (with respect to the X
transducer). A plot scale is drawn parallel to the 3
measurement axis for convenience.

5 µm/div 4
B5 The Orbit.qxd 9/10/2002 10:36 AM Page 82

82 Data Plots

An orbit can be used to determine the direction of precession of the rotor.

The Keyphasor mark on the orbit (the blank/dot sequence on Bently Nevada
orbit plots), shows the direction of increasing time, which is the direction that
the shaft is moving (the direction of precession). On oscilloscopes, however, the
sequence can be a blank/bright or bright/blank, depending on the type of oscil-
loscope and whether a notch or projection is used as the Keyphasor mark on the
shaft. When using an oscilloscope, it is always a good idea to verify the configu-
ration of the event that is being used in your equipment by examining a time-
base display.
Once the Keyphasor mark sequence is determined, it will show the direction
of increasing time. This is the direction that the rotor moves in the orbit, regard-
less of the direction of rotation. The direction of precession can be compared to
the direction of rotation to determine whether the precession is forward (pre-
cession direction same as rotation direction) or reverse (precession direction
opposite rotation direction).
On complex orbits, the rotor may undergo forward precession over part of
the orbit and reverse precession over another part of the same orbit. In the ½X
orbits in Figure 5-9, note how inside loops maintain forward precession (left),
while outside loops show reverse precession (right).

The filtered orbit can be used to estimate the absolute phase of the two
component signals. Phase estimation will be most accurate for circular orbits,
and less accurate for elliptical orbits (Figure 5-10). This is because, in a circular
orbit, the shaft moves along the orbit with constant angular velocity (equal time
intervals and equal angles between dots). In elliptical orbits, the orbital angular
velocity is not constant (equal time intervals and unequal angles between dots).
Because phase is a timing measurement, this angular velocity variation causes
inaccuracies when trying to estimate the phase with respect to each transducer.
On 1X orbits, it also causes the high spot on the shaft to oscillate as the rotor
traverses an orbit. This effect, when combined with bending, actually produces
2X stress cycling in a shaft with a 1X, elliptical response.
This phase estimation technique is best used to quickly confirm the validi-
ty of phase data from some other source. Usually, a visual inspection of the orbit
with a Keyphasor mark will provide a good cross-check of the data. For accurate
measurement of both absolute and relative phase, the timebase plot is a better
choice.
B5 The Orbit.qxd 9/10/2002 10:36 AM Page 83

Chapter 5 The Orbit 83

Y Y
Forward
Forward

X X

Reverse

Figure 5-9. Direction of precession in orbits. In the orbit plot, the shaft moves from the
blank toward the dot. The orbit plot on the left illustrates that an inside loop is always
forward precession. At the top of the right orbit, the shaft moves in a counterclockwise
(X to Y) direction, which is the same as the direction of rotation. Thus, the precession is
forward. At the bottom of the right orbit, the path forms an outside loop, and the shaft
moves for a short time in reverse precession.

Y Y
18˚ 5˚

62˚
X X

Figure 5-10. Shaft orbital angular velocity in circular and elliptical orbits. The orbit dots
show the changes in position of the shaft centerline in equal time intervals. During
each time interval, the shaft rotates 18°. In a circular orbit the shaft centerline has con-
stant orbital angular velocity and moves through equal 18° increments in equal time
intervals. In elliptical orbits, the shaft centerline has different orbital angular velocities in
different parts of the orbit. In this example, the shaft orbital angular velocity varies from
5° to 62° per interval. This changing angular velocity makes it difficult to estimate phase
from highly elliptical orbits.
B5 The Orbit.qxd 9/10/2002 10:36 AM Page 84

84 Data Plots

To estimate the absolute phase of a signal (Figure 5-11),

1. Be sure that the orbit is filtered to a harmonic of running speed

(1X, 2X, 3X, etc.). An unfiltered signal can be used to estimate
the phase if the orbit is dominated by a single frequency. Select
the desired transducer. In the figure, the Y transducer is select-
ed.

2. Locate the Keyphasor mark on the orbit.

3. Determine the direction of precession. This will be the direction

that the rotor is moving in the orbit and is indicated by the
blank/dot sequence (red arrow).

4. Absolute phase is the fraction of the vibration cycle, in degrees,

from the Keyphasor event to the first positive peak of the signal
with respect to the selected transducer. On the orbit, this will be
from the Keyphasor dot, in the direction of precession, to the
point on the orbit that is closest to the transducer (the red cir-
cle in the figure).

5. Draw lines from the Keyphasor dot to the center of the orbit and
from the closest approach point to the center of the orbit. The
angle between these lines is the estimate of the absolute phase.

The relative phase of the two signals can also be estimated. The relative
phase is the fraction of the vibration cycle between the point of closest approach
to one probe and the closest approach to the other probe. For a circular, 1X orbit,
this will be 90°. For an elliptical orbit, this number can range from nearly 0° to
nearly 180°, depending on the orbit ellipticity and orientation (Figure 5-12).
While the orbit can provide an estimate of the relative phase, the timebase plot
is a better choice for accurate measurement.
B5 The Orbit.qxd 9/10/2002 10:36 AM Page 85

Chapter 5 The Orbit 85

1
Y X
Figure 5-11. Estimating absolute phase with respect to
the Y transducer on a 1X-filtered orbit. The angle
between the Keyphasor dot and the closest approach
2
to the probe (red circle) is an estimate for the absolute
4 3
phase of the Y transducer signal. Here, the measured
angle is about 65°. Because the orbit is elliptical, the 5
rotor shaft will be moving along this part of the path
relatively quickly (see Figure 5-10) and will cover the
angle in somewhat less time. Thus, the absolute phase
will be less than 65°; it is actually about 60°.

Y Y

X X

Figure 5-12. Estimating relative phase in orbits. Two 1X-filtered, highly

elliptical orbits are shown. The positive peak of the X signal is shown
with a green, vertical line, and the positive peak of the Y signal is shown
with a blue, horizontal line. The relative phase, in this case, is the portion
of the vibration cycle between maxima of the X and Y signals, measured
in the direction of precession (blank to dot, increasing time). In the left
plot, the relative phase is about 160°. In the right plot, the relative phase
is about 20°. Because of timing issues with highly elliptical orbits, time-
base plots are a better tool for determination of relative phase.
B5 The Orbit.qxd 9/10/2002 10:36 AM Page 86

86 Data Plots

The unfiltered orbit can be used to determine the relative frequency of vibra-
tion versus running speed. This technique can be used for both supersynchro-
nous or subsynchronous vibration and uses the Keyphasor marks that are gen-
erated once per turn of the rotor.
To determine the relative frequency versus running speed:

1. Plot an unfiltered orbit, and select a reference transducer.

2. Start at a Keyphasor dot on the orbit. Trace a path around the

orbit, counting the number of complete cycles of vibration rela-
tive to the reference transducer, until you return to the original
Keyphasor dot. This may require several revolutions worth of
data.

3. Count the number of shaft revolutions that occur during the

cycles of vibration.

4. Express the result in the form nX, where n is a fraction and is

defined by

number of cycles of vibration

n= (5-2)
number of shaft revolutions

5. Reduce the order of the fraction to the lowest common denom-

inator. For example, ²⁄₄X can be reduced to ½X.

Figure 5-13 shows several examples of orbits with different frequency ratios.
There are several items of note:

1. For integer ratios, the number of Keyphasor dots in the orbit is

always equal to the denominator of the frequency ratio. Both
the ¹⁄₃X and ²⁄₃X orbits have three Keyphasor dots. Similarly, ¼X
and ¾X would have four.

2. Frequency ratios with one in the numerator (¹⁄₃X, ½X, 1X) have
orbits with clearly defined blank regions before the Keyphasor
dots.
B5 The Orbit.qxd 9/10/2002 10:36 AM Page 87

Chapter 5 The Orbit 87

3. Frequency ratios with an integer other than one in the numera-

tor can have poorly defined or obscured blank regions. For
example, in the 2X orbit, there are two complete cycles of vibra-
tion for each revolution of the shaft. Thus, there will be one
Keyphasor mark for every two cycles of vibration. If the orbit has
no other frequency components, it will cover up the blank of the
Keyphasor mark during the second cycle of vibration. Thus, at
first glance, a 2X or 3X orbit will look very similar to a 1X orbit;
however, the higher order orbits will paint over the blanks of the
Keyphasor marks on an oscilloscope. Some software plotting
packages have a special provision to avoid this.

Using these rules, it is possible, with a little practice, to determine simple

frequency ratios of vibration with a quick observation of an orbit.

Y X Y X Y X Y X Y X

1/3X 1/2X 2/3X 1X 2X

Figure 5-13. Vibration frequency relative to running speed. Several simulated, unfiltered orbits are
shown with different frequency ratios relative to running speed. The paths of the 2/3X and 2X
orbits are shown slightly offset to emphasize the path and Keyphasor dot behavior. The number
of vibration cycles yields the numerator of the frequency ratio, the number of Keyphasor dots
determines the denominator.
B5 The Orbit.qxd 9/10/2002 10:36 AM Page 88

88 Data Plots

We have discussed this behavior for vibration at simple integer ratios. What
if the vibration frequency is slightly less or more than an integer ratio? Each
orbit in Figure 5-14 shows several revolutions of data. The middle orbit shows
vibration that is exactly ½X, the orbit on the left shows vibration that is slightly
below ½X, and the orbit on the right slightly above ½X.
When the frequency is an exact integer ratio, each Keyphasor dot plots in
the same place every vibration cycle: the Keyphasor dots are said to be “locked.”
When the frequency is slightly above or below a simple integer ratio, the
Keyphasor dots will move slowly around the orbit. The direction that they move
will depend on whether the vibration frequency is above or below the fractional
ratio. How fast they move depends on how far the frequency is from the integer
ratio.
When the frequency is slightly below a simple integer ratio (left orbit) the
orbital motion of the rotor is slower, so the rotor completes less of the vibration
cycle per revolution. The Keyphasor dot occurs a little earlier in the cycle and
appears to move slowly around the orbit in a direction opposite to the direction
of rotor precession.
When the frequency is slightly above a simple integer ratio (right orbit), the
orbital motion of the rotor is slightly faster, so the rotor completes more of a
cycle of vibration per revolution. Thus, the Keyphasor dot occurs a little later in
the cycle and appears to move slowly around the orbit in the same direction as
the direction of precession.

Y X Y X Y X

2 2

1
1

<1/2X 1/2X >1/2X

Figure 5-14. Keyphasor dot behavior in unfiltered orbits. The direction of

rotation in all three orbits is Y to X (clockwise), and the numbers show
successive Keyphasor dots. In the left orbit, the dominant frequency of
vibration is slightly less than 1/2X, and each successive Keyphasor dot
plots in a new position located slightly against the direction of rotation.
The middle orbit shows exactly 1/2X vibration; the Keyphasor dots are
locked. The right orbit shows vibration that is slightly greater than 1/2X,
and successive Keyphasor dots move in the same direction as the direc-
tion of rotation. The Keyphasor dot is acting like a strobe.
B5 The Orbit.qxd 9/10/2002 10:36 AM Page 89

Chapter 5 The Orbit 89

Because of the sensitivity of this visual effect, the orbit can be superior to the
spectrum plot for determining whether vibration is an exact integer ratio of run-
ning speed. Spectrum plots have an inherent limit in resolution which is deter-
mined by the span and number of spectral lines. For example, a 400-line spec-
trum with a span of 500 Hz will have a resolution, or frequency bin width, of
500/400 = 1.25 Hz = 75 cpm. If a machine is turning at 3600 rpm, ½X vibration
has a frequency of 1800 cpm, and 0.49X vibration has a frequency of 1764 cpm.
These frequencies differ by only 36 cpm, less than the spectral resolution. Thus,
at this resolution, both frequencies would plot in the same spectral line, and it
would require a much higher resolution spectrum to discriminate between these
two frequencies.
Yet this frequency difference can be very important. Under the right condi-
tions, rub can produce exactly ½X vibration. On the other hand, fluid-induced
instability typically occurs at a subsynchronous frequency a little below ½X and
is unlikely to occur at exactly ½X. Determining this seemingly small difference
in frequency can be very important to properly diagnosing a machine malfunc-
tion. The orbit would clearly show locked Keyphasor dots for pure ½X vibration
and Keyphasor dots moving against the direction of precession for vibration at
slightly less than ½X.

The unfiltered orbit can be used to determine the relative frequency of vibra-
tion in the X direction compared to the frequency of vibration in the Y direction
(Figure 5-15):

1. Plot an unfiltered orbit.

2. Start at a Keyphasor mark and move around the orbit. Note how
many positive (or negative) peaks are encountered with respect
to one of the transducers.

Y
1
Figure 5-15. Frequency ratios from orbits. Start at the
Keyphasor dot and move around the orbit. Count the
number of positive peaks that occur in the X direction
and in the Y direction. In this example, the X transducer X
sees two positive peaks (one is smaller than the other), 1 2
and the Y transducer sees only one positive peak. The
X:Y frequency ratio is 2:1.
B5 The Orbit.qxd 9/10/2002 10:36 AM Page 90

90 Data Plots

3. Repeat this procedure with respect to the other transducer. The

relative frequency can be stated as

f X : fY (5-3)

or the reverse.

In the figure, two positive peaks are encountered in the X direction, and one
positive peak is encountered in the Y direction. Thus the X:Y frequency ratio is
2:1, and the components are 2X and 1X. (There is also a 1X component in the X
direction, as revealed by the different horizontal positions of the two peaks.) As
seen in this example, the shape of an unfiltered orbit can provide important
information about the frequency content of machinery vibration.

Shape can also provide clues to the presence of excessive radial loads that
may exist. High radial loads tend to push the rotor to high eccentricity ratios in
a fluid-film hydrodynamic bearing. The oil film that supports the rotor produces
very high spring stiffness at high eccentricity ratios. This stiffness is highest in
the radial direction and lowest in the tangential direction. Figure 5-16 shows two
unfiltered orbits from two different steam turbines with high radial loads. Note
that the two machines turn in opposite directions (black arrows). Both orbits
display remarkably similar shapes; in fact they are nearly identical when cor-
rected for rotation direction. The arcs indicate the location of the bearing wall,
and the red arrows indicate the likely direction of the applied radial load.
So far we have discussed measurements that can be performed on single

Figure 5-16. Effect of radial

load on orbit shape. Orbits are X Y X
shown from two different
steam turbines with opposite Y

rotation directions (black

arrows). Both machines are
experiencing high radial loads.
The red arrows indicate the
approximate direction of the
applied radial load. The arcs
represent the probable orien-
tation of the bearing wall.
B5 The Orbit.qxd 9/10/2002 10:36 AM Page 91

Chapter 5 The Orbit 91

orbits. However, multiple orbits can greatly increase the power of the orbit plot.
Multiple orbits can be created over speed, over position, over changing load, or
over a changing machine process parameter. The changes that appear in multi-
ple orbits help illuminate various aspects of machine behavior.
Multiple orbits over speed are created using data from a single measurement
plane in the machine. During a startup or shutdown, at appropriate speed or
time intervals, a data sample is taken that is used to create an orbit. Several such
plots can be combined to show how rotor behavior evolves over speed.
Multiple orbits of this kind can be used to identify a balance resonance.
Figure 5-17 shows a series of 1X-filtered, compensated orbits taken during the
shutdown of a 75 MW steam turbine. Remember that the Keyphasor dot shows
the position of the rotor high spot when the Keyphasor event occurs. As the
machine passes through the resonance, the relationship between the unbalance
location (the heavy spot) and the rotor response (the high spot) changes. The
heavy spot location does not change (it is fixed in the shaft relative to the
Keyphasor notch), but, during a shutdown, the high spot lags the heavy spot less
and less as the machine passes through the resonance. The orbit shows these
changes. First, as speed decreases, the high spot location (which is shown by the
Keyphasor dot) typically moves around the orbit in the same direction as rota-
tion. At the same time, the orbit size increases, reaches a maximum as the
machine passes through the resonance, and decreases again below the reso-
nance. Note also that the orientation of the orbit ellipse changes as the shaft
goes through the resonance, due to the anisotropic stiffness (the stiffness is not
the same in all radial directions, see Chapter 13) of the system. These effects are
a result of the rotor system dynamic response to unbalance (see Chapter 11).

Shutdown
Y X Y X Y X Y X Y X

1800 rpm 2020 rpm 2220 rpm 2400 rpm 2600 rpm

Figure 5-17. Multiple orbits over speed from a 75MW steam turbine generator. All orbits are plot-
ted at the same scale from 1X-filtered shutdown data from the HP turbine bearing. The orbits
show the machine passing through a balance resonance at around 2220 rpm. With decreasing
speed, the Keyphasor dot on the orbit moves in the direction of decreasing phase lag (the direc-
tion of rotation).
B5 The Orbit.qxd 9/10/2002 10:36 AM Page 92

92 Data Plots

Multiple orbits over position are created when the data is taken from several
measurement planes at the same time. Because the unfiltered orbit shows the
path of the shaft centerline, multiple orbits of this kind give us a three-dimen-
sional picture of the motion of the rotor along the length of the machine train.
This technique works best when machines are rigidly coupled; flexible couplings
reduce the influence of machines on each other.
Figure 5-18 shows a series of 1X-filtered orbits from a 125 MW steam turbine
generator set. The Keyphasor dots mark the location of the shaft in each plane
at the instant the Keyphasor event occurs. These dots can be linked to obtain an
estimate of the three-dimensional mode, or deflection, shape along the rotor.
Note that, on a simple orbit plot, we cannot do this for the remainder of the
orbit. Shaft motion along the orbit occurs at different rates in different parts of
the orbit. Without additional timing marks, we cannot say for certain where the
shaft is at any particular time. The Keyphasor mark gives us the timing infor-
mation for one particular point on each orbit. If we can access the individual
digital sample points on the orbit, then we can connect simultaneous points on
different orbits and obtain mode shape shape information for those points.
Note also that the deflection shape of the shaft is an estimate. Other, more
complicated shapes are possible, and it would require more planes of measure-
ment to be sure of the deflection shape at this speed. This is specifically the

HP/IP Turbine LP Turbine Generator

1 2 3 4 5 6

1 2 3 4 5 6
Y Y Y Y Y Y
X X X X X X

Figure 5-18. Multiple orbits over position from a 125 MW steam turbine generator set. A series of
1X-filtered orbits at the same scale were captured at the same moment in time. For each orbit,
the Keyphasor dots show the location of the shaft when the Keyphasor event occurs. The
Keyphasor dots can be linked (red curve) to obtain an estimate of the deflection shape of the
rotor shaft.
B5 The Orbit.qxd 9/10/2002 10:36 AM Page 93

Chapter 5 The Orbit 93

objective of mode identification probes (see Chapter 12). However, this infor-
mation can be combined with information from a good rotor system model to
confirm the rotor deflection shape.
Knowledge of the rotor deflection shape can be helpful for interpolating the
rotor vibration at midspan points well away from the measurement locations.
This information can help identify possible locations where high vibration of the
rotor might conflict with limited clearances in the machine.
Multiple orbits over changing operating conditions can be created with data
from a single plane (or multiple planes). Changes in load or operating conditions
often produce changes in vibration behavior. The information from multiple
orbits can help clarify the cause of a machine problem.
Figure 5-19 shows a series of orbits that were obtained from a single plane
on a gas pipeline compressor. As suction pressure was varied, vibration behavior
changed dramatically, from normal, mostly 1X behavior to fluid-induced insta-
bility. This information was used to diagnose and develop a remedy for the prob-
lem.
Multiple orbits are important because they open up other dimensions, such
as speed, time, load, or some other changing parameter. The additional infor-
mation can be very helpful for detecting and diagnosing machinery problems.

Suction pressure

Time 12:26 12:27 12:28 12:29

Y Y Y Y

X X X X

14.3 krpm 14.3 krpm 14.3 krpm 14.3 krpm

Figure 5-19. Multiple orbits over changing operating conditions from one bearing of a gas
pipeline compressor. While the machine runs at 14 300 rpm at a constant discharge pressure, the
orbits show that a fluid-induced instability occurred whenever the suction pressure was
increased.
B5 The Orbit.qxd 9/10/2002 10:36 AM Page 94

94 Data Plots

The Orbit/Timebase Plot

The orbit/timebase plot combines the orbit with the two timebase plots that
are used to create it and provides the same choice of display characteristics: fil-
tered, unfiltered, or compensated. Timebase plots are discussed in Chapter 4.
Bently Nevada orbit/timebase plots are created with the orbit on the left
side of the plot. The Y timebase plot is displayed above the X timebase plot to
the right of the orbit (Figure 5-20). The plot contains information on the direc-
tion of rotation, the plot scaling used in the orbit (the same as the vertical scale
in the timebase plots), the speed, and the time scale on the timebase plots.
The figure is an example of how to use an orbit/timebase plot to locate a sur-
face defect on a shaft. The plot shows slow roll data from a boiler feed pump
motor. This orbit has a large amount of glitch, probably due to some shaft sur-
face damage. Surface defects displayed in orbits typically show up as spikes that
point in the general direction of the transducers. The timebase plots help clari-
fy the timing relationships of the spikes and make it possible to determine the
angular location of the surface damage. The spike in the Y plot occurs about
50 ms after the Keyphasor event. The period of one revolution is the time
between successive Keyphasor events and is about 400 ms (actually 392 ms,

X Y

Spike

Y Spike
X

0.2 mil/div 153 rpm 0 200 400 600 800

Time (ms)

Figure 5-20. An orbit/timebase plot from the inboard bearing of a boiler feed pump
motor. This plot shows two revolutions of slow roll data from probes mounted at 90° L
and 0°. The timebase clearly shows the timing of one large positive-going noise spike,
which shows up on the signals from both probes, 90° apart. Note that, on the Y time-
base plot, the spike occurs about 50 ms after the Keyphasor event; thus, the location of
this mark on the shaft can be determined. See the text for details.
B5 The Orbit.qxd 9/10/2002 10:36 AM Page 95

Chapter 5 The Orbit 95

based on the speed of 153 rpm). So, once the Keyphasor event occurs, the shaft
turns approximately

 50 ms 

 400 ms/rev (
360˚/rev )= 45˚

until the defect is under the Y probe. To find the damaged spot, we would align
the Keyphasor notch with the Keyphasor probe, and look 45° against rotation
from the Y probe.
Remember that positive peaks on the waveform represent rotor passage
next to the measuring probe and that the probe mounting locations are shown
on the orbit plot. The Keyphasor marks provide an important guide for estab-
lishing the location on the orbit or timebase plots. The Keyphasor mark repre-
sents the same instant in time on all the plots.
This combination of plots allows us to correlate events on the orbit with
their corresponding events on the timebase plots. The timebase plots act to pro-
vide a time scale that can help establish accurate timing of observed events on
the orbit.

Summary
The orbit shows the path of the shaft centerline. It combines the one-dimen-
sional timebase information from two, orthogonal, coplanar transducers into a
two-dimensional plot of the lateral motion of the rotor shaft. Orbits can be unfil-
tered or filtered.
A Keyphasor mark on an orbit shows the location of the shaft when the
once-per-turn Keyphasor mark on the shaft passes next to the Keyphasor trans-
ducer.
To remove slow roll runout from the orbit, unfiltered orbits can be waveform
compensated, and filtered orbits can be vector compensated. Orbits can also be
effectively notch filtered by compensating the unfiltered orbit with two synthe-
sized, filtered waveforms. The end result is a Not-nX orbit.
Orbits can be used to measure the peak-to-peak amplitude of filtered or
unfiltered vibration and the direction of precession of the rotor. Filtered orbits
can be used to estimate the nX amplitude and phase.
Orbits are useful for various kinds of frequency analysis. The relative fre-
quency content of X and Y signals can be determined. With the Keyphasor
mark, the orbit can be used to determine frequencies of vibration relative to run-
ning speed. The Keyphasor marks are superior to a spectrum plot for establish-
B5 The Orbit.qxd 9/10/2002 10:36 AM Page 96

96 Data Plots

ing whether the frequency of vibration is an exact submultiple of running speed

or not.
Orbit shape can yield information on the direction and relative magnitude
of static radial loads. Also, different frequencies of vibration can produce char-
acteristic and recognizable shapes in orbits.
Multiple orbits can be created over speed, axial position, or change in load
or some other parameter. By adding another dimension, multiple orbits greatly
increase the information available for machinery diagnosis.
Orbit/timebase plots are created from the combination of an orbit with the
two XY timebase plots used to create it. The timebase plots are displayed to the
right of the orbit with the Y plot over the X plot. Orbit/timebase plots can be
used to establish timing relationships for features seen in the orbit.