PMD Issues in Advanced,

Very High-Speed Networks

This pocket guide provides a comprehensive review of polarization mode dispersion (PMD).
PMD has been causing headaches for network operators for more than a decade now, but why is that?
What is PMD and will it continue to plague the networks of tomorrow?
This guide explains the fundamentals of PMD as well as all related issues, like polarization, birefringence
and DGD. It covers the tolerance specifications of various systems and transmission speeds, the
different international standards that must be adhered to and the most common testing techniques used
on the market.
The purpose of this guide is to review the PMD phenomenon in detail, its characteristics, its
consequences, the nature and effects of its interaction with other critical parameters, its mitigation, along
with examples of test results, generalities and limitations.
PMD Issues in Advanced, Very High-Speed Networks EXFO 1
Table of Contents
1. Introduction ...................................................................... 2
1.1 The Phenomenon ...................................................................................4
1.2 Pulse Broadening, ISI and BER .............................................................6
1.3 Higher Bit Rates .....................................................................................6
1.3.1 Longer Single-Mode Fibers .......................................................7
2. Polarization ....................................................................... 8
2.1 States of Polarization ............................................................................8
2.1.1 Linear Polarization .....................................................................8
2.1.2 Circular Polarization ................................................................13
2.1.3 Elliptical Polarization ...............................................................15
2.2 Unpolarized Light ................................................................................16
3. Birefringence, Causes and Effects ..............................17
3.1 Intrinsic Stresses ................................................................................ 17
3.2 Extrinsic Stresses ...............................................................................19
3.3 Birefringence Planes and Axes ..........................................................20
4. Principal States of Polarization ....................................21
4.1 Definition .............................................................................................. 21
4.2 DGD, PMD and Pulse Broadening ....................................................... 21
4.2.1 No Pulse Broadening.............................................................. 22
4.2.2 Fixed DGD, PMD and Pulse Broadening ................................23
4.2.3 Random DGD, PMD and Growing Pulse Broadening ............24
4.2.4 Random DGD, PMD and Growing Pulse
Broadening as a Function of Fiber Length ..........................25
4.2.5 Pulse Broadening, ISI and BER ..............................................24
5. PMD Specifications and International Standardization .. 26
5.1 PMD Cable and Link Design Specifications ......................................26
5.2 System PMD Specification ..................................................................28
5.3 DGD
max
Specifications for Various
Applications and Modulation Formats ...............................................29
5.4 Proposals for DGD
max
Values for Various
Applications and Modulation Formats ..............................................31
5.5 DGD
max
Specifications for Ethernet Applications ..............................34
6. PMD-Induced Pulse Broadening and Penalty ............36
7. PMD Tests and Measurements ....................................37
7.1 Description of the Available Test Methods ....................................... 38
7.1.1 Fixed Analyzer Fourier Transform ..................................... 38
7.1.2 Interferometric Method (TINTY / GINTY) ................................41
7.1.2.1 Traditional Analysis ..................................................... 42
7.1.2.2 Generalized Analysis ....................................................43
7.1.3 Scrambled State-of-Polarization Analysis ............................ 44
7.1.3.1 General Theory ............................................................. 44
7.1.3.2 Experimental Implementations ..................................45
7.1.3.3 Measurement of PMD as a Function of Distance .... 46
7.2 Uncertainty in PMD Tests and Measurements .................................47
7.3 Calibration of PMD Test and Measurement Instruments ................47
7.3.1 Applicability Matrix for the Available PMD Test Methods .... 48
2 EXFO PMD Issues in Advanced, Very High-Speed Networks
1. Introduction
The continuous demand for more capacity in telecommunication networks has led to the use of fiber optics. However,
this increase in fiber-optic network capacity has led to new limitations, and the principles and understanding of their
effects are of interest to the scientific community.
Problems such as attenuation and dispersion are now well-understood. However, at very high bit rates (VHBR),
parameters related to polarization, such as polarization mode dispersion (PMD), polarization dependent loss (PDL)
and polarization dependent gain (PDG), interact with each other, along with chromatic dispersion (CD) and nonlinear
effects (NLE). This interaction brings additional constraints and further considerations. Table 1 summarizes the various
issues related to VHBR transmissions.
PMD Issues in Advanced, Very High-Speed Networks EXFO 3
Table 1 Issues Related to VHBR Transmissions
Parameter Issues
Attenuation
Dirty connector
Excessive bending
Dispersion
CD
Stochastic phenomenon when interacting with PMD
Residual CD becomes critical
PMD
Additional constraints when interacting with PDL,
PDG, CD and index-related NLEs
Second-order becomes a major issue
NLE
Self-phase modulation (SPM) and
cross-phase modulation (XPM)
Detrimental at high power when interacting
with CD and PMD
Raman optical amplification
Stimulated Raman scattering (SRS) generates
double-Rayleigh backscattering (DRBS)
and multiple-path interferences (MPI)
Four-wave mixing (4WM) Always present in high-power WDM transmissions
Given the shift towards 40 and 100 Gbit/s data rates using new advanced modulation formats, a lot of work has been
done on PMD and differential group delay (DGD).
4 EXFO PMD Issues in Advanced, Very High-Speed Networks
1.1 The Phenomenon
It is now accepted that PMD causes an optical pulse to statistically spread and possibly get distorted in the time
domain. When the broadening becomes too wide, the stream pulses start to overlap and may produce inter-symbol
interference (ISI); the eye pattern starts to close and the bit error rate (BER) increases significantly, which is an
indication of serious signal degradation. After a threshold has been reached, the system initiates a communication
failure and outage.
The biggest issue with PMD is the fact that it is a stochastic phenomenon (i.e., statistical in nature) and can only be
quantified using sampling, distribution and averaging. Similar to any statistical polling, an infinite number of samples
would be required to obtain an absolutely accurate result. Since this is impossible, a manageable number of samples
must be considered. However, the average value calculated from these samples comes with some level of uncertainty.
Since the phenomenon is also subject to time variations, this makes things all the more difficult to understand.
PMD needs a frequency or wavelength to be characterized, like all dispersion. Similar to CD, the basic characterizing
parameter is the index of refraction (IOR), which is the index of propagation of the medium in which the signal is
transported, such as in an optical fiber. The subsequent variation of this index as a function of frequency or wavelength
leads to the group delay as a function of frequency or wavelength.
However, there are two fundamental differences:
Polarization of the propagating signal
Stochastic behavior, not deterministic
PMD Issues in Advanced, Very High-Speed Networks EXFO 5
Since the PMD phenomenon is related to polarization, it is therefore related to the propagation axes (i.e., two axes
with different indexes of propagation), which leads to birefringence (i.e., difference in the index of refraction). Unlike for
CD, there is not just one group delay, but two. A difference in group delays or DGD varies statistically as a function
of frequency or wavelength. The DGD variation may follow a regular, smooth function or be totally random.
Whatever the characteristics of the variation, there will be a maximum and a minimum value as well as an average value
over the widest possible frequency or wavelength range.
Of course, the phenomenon may change depending on whether it is applied to:
a long or short single-mode fiber (SMF)
a simple or complex active or passive optical component in a sub-system or in a low- or high-speed network
a combination of the above in length, size and quantity
At VHBR, the phenomenon is the same except that the transmitted pulses are closer to each other in the time domain.
The effect is that it has faster and more dramatic consequences on statistical broadening. This is why PMD at VHBR
is one of the most critical and crucial phenomenon to take into account.
6 EXFO PMD Issues in Advanced, Very High-Speed Networks
1.2 Pulse Broadening, ISI and BER
The impact of pulse broadening due to PMD on network
operation is similar to the effect of CD. If the pulse
broadens too much, consecutive pulses can overlap in
the bit stream and inter-symbol interference (ISI) may
occur, increasing BER to the point of causing a service
outage. The phenomenon becomes even more detrimental
at high bit rates, long SMF lengths and/or with stronger
birefringence, especially in the case of legacy SMFs and
extrinsic random stresses.
1.3 Higher Bit Rates
At lower bit rates, even if the stress is strong and the DGD
is large, there may be no PMD effect, such as in Figure 1.
However, when the bit rate (br) increases, the bit period
(bp) decreases and the pulses in the bit stream get closer
together until they overlap. This leads to ISI and increasing
BER until a traffic outage occurs, as shown in Figure 1.
Figure 1 Effects of Bit Rate Increase
(bit period decrease) on ISI
PMD Issues in Advanced, Very High-Speed Networks EXFO 7
1.3.1 Longer Single-Mode Fibers
When the SMF length increases, the probability of cumulative stress and pulse overlap increases, as well as ISI and
BER, as shown in Figure 2.
Figure 2 Effects of SMF Increase (constant bit period) on ISI
8 EXFO PMD Issues in Advanced, Very High-Speed Networks
2. Polarization
The concept of polarization in optics is analogous to the one used in sociology. For instance, in a random group of
people arriving at a public meeting, no precise, defined or characteristic behavior is perceived; only background noise
from individual conversations. The population is said to be non-polarized or unpolarized; this is the definition of noise.
Each individual in the population has his/her own opinion and all the opinions are simultaneously present. However,
from an outsiders perspective, the population seems chaotic with no defined opinions or no opinions at all. If a strong,
biased speaker gives a presentation, the audience will start showing interest, i.e., the population becomes polarized.
This population can be said to be slightly or strongly polarized depending on the percentage of people showing
interest. The same is true in optics.
Light, as a transverse electromagnetic wave, is comprised of orthogonal magnetic and electric fields travelling in
the same direction called the axis of propagation. Since common electronic detectors/receivers (Rx) respond to
the electric field (E-field) effects of electrons in materials, and not the magnetic field effects, only the E-field and its
propagation in a glass medium, like an optical fiber, will be considered herein.
2.1 States of Polarization
Polarization is a property of light. In fact, the lightwave is said to be polarized when its E-field vector is at a specific
angle to the propagating z, t axis. The state of polarization (SOP) is determined by a transmitter (Tx), but defined from
the Rx standpoint or from an observer looking at the transmitter or the source of light.
2.1.1 Linear Polarization
The E-field vector may propagate in the x - z, t plane only and the lightwave is then said to be linearly horizontally (LH)
polarized. This is because when looking along the z axis from the Rx to the Tx, the in-coming E-field vector is moving
back and forth in the horizontal plane on a straight line. This case is illustrated in Figure 3.
PMD Issues in Advanced, Very High-Speed Networks EXFO 9

NOTE: The eye represents the direction in which an observer or receiver is looking NOTE: The z axis is pointing off the page
(a) SOP in x - z, t plane (b) Propagation in (a) as seen by the observer or receiver
Figure 3 E-Field Vector Propagation of a Linear Horizontal SOP
y
x z
E
x
E
x
E
x
x
y
z, t
10 EXFO PMD Issues in Advanced, Very High-Speed Networks
The lightwave may also propagate vertically in the y - z, t plane and is then said to be linearly vertically (LV) polarized.
This case is illustrated in Figure 4.

NOTE: The eye represents the direction in which an observer or receiver is looking NOTE: The z axis is pointing off the page
(a) SOP in y - z, t plane (b) Propagation in (a) as seen by the observer or receiver
Figure 4 E-Field Vector Propagation of a Linear Vertical SOP
y
x z
E
y
x
y
z, t
E
y
E
y
PMD Issues in Advanced, Very High-Speed Networks EXFO 11
The wave can also be defined as the combination of an x - z, t plane wave and a y - z, t plane wave, such as the ones
shown in Figure 5. The resulting wave propagates at a certain angle such as +/4 (+45
o
) or /4 (45
o
) or any other
angle as shown in Figure 5.
NOTE: The eye represents the direction in which an observer or receiver is looking NOTE: The z axis is pointing off the page
(a) /4 (45
o
) SOP from both linear orthogonal SOPs in the y - z, t and x - z, t planes (b) Propagation in (a) as seen by the observer or receiver
Figure 5 Angular E-Field Vector Propagation of a Linear +45
o
SOP
y
z
E
x
x
E
y
x
y
z, t
E
x
E
E
E
y
12 EXFO PMD Issues in Advanced, Very High-Speed Networks
Depending on the relative amplitude of both E-field vectors, the resulting E-field vector will have a proportional
polarization angle as shown in Figure 6.
y
z
E
x
x
E
y
y
z
E
x
x
E
y
y
z
E
x
x
E
y
y
z
E
x
x
E
y
y
z
E
x
x
E
y
NOTE: The z axis is pointing off the page
Figure 6 Linear Angular E-Field Vector Propagation with Different Amplitudes and Angles
PMD Issues in Advanced, Very High-Speed Networks EXFO 13
2.1.2 Circular Polarization
When one orthogonal wave propagates out of phase by +/2 (+90
o
) from the other, the resulting wave is said to be
circularly polarized. The direction of rotation of this circular polarization, either clockwise or counterclockwise, depends
on the sign of the relative phase shift between the two waves (x - z, t wave and y - z, t wave). The resulting wave is
clockwise circularly polarized when the relative phase shift is +/2 (+90
o
). In this case, the x - z plane wave lags behind
the y - z plane wave by +/2 (+90
o
) as shown in Figure 7. The polarization is also said to be circular left-hand (Clh).

NOTE: The eye represents the direction in which an observer or receiver is looking NOTE: The z axis is pointing off the page
(a) -SOP in both y - z, t and x - z, t planes (b) Clockwise rotation of the wave vector from
(a) as seen by the observer or receiver
Figure 7 Clockwise Rotation of the Wave Vector
y
x z
E
y
E
x
1
3 4
2

E
x
x
y
z, t
E
y

x
14 EXFO PMD Issues in Advanced, Very High-Speed Networks
The resulting wave may also propagate counterclockwise when the relative phase shift is /2 (90
o
); x - z plane
wave lagging behind the y - z plane wave by +/2 or +90
o
. In this case, the resulting polarization is said to be circular
right-hand (Crh) as shown in Figure 8.

NOTE: The eye represents the direction in which an observer or receiver is looking NOTE: The z axis is pointing off the page
(a) SOP of circular right-handed polarized wave (b) Counterclockwise rotation of the wave vector from
(a) as seen by the observer or receiver
Figure 8 E-Field Vector Propagation of Circular Right-Hand SOP
y
x z
E
y
E
x
1
3
4
2
E
x
x
y
z, t
E
y

y
PMD Issues in Advanced, Very High-Speed Networks EXFO 15
2.1.3 Elliptical Polarization
In the more general case of an arbitrary, non-zero relative phase shift between the x - z plane wave and the y - z plane
wave, the resulting wave will be elliptically polarized. A general representation in the x - y plane of a polarized wave as
seen by the observer or the receiver is shown in Figure 9.
NOTE: The eye represents the direction in which an observer or receiver is looking NOTE: The z axis is pointing off the page
(a) SOP of an elliptically polarized wave (b) Elliptical rotation of the wave vector from
(a) as seen by the observer or receiver
Figure 9 E-Field Vector Propagation of an Elliptical SOP
Of course, depending on the amplitude relationship of the E-field vectors in the x - z, t plane and the y - z, t plane and
their respective phase relationship, an infinite amount of elliptical polarizations can be observed.
y
z
a

b
E
x
E
x
E
y

E
y
x
y
z, t
E
y

x
16 EXFO PMD Issues in Advanced, Very High-Speed Networks
2.2 Unpolarized Light
When the E-field vector is propagating in any random orientation around the z, t axis, at any point along the z axis and/
or at any point in time, this wave is said to be unpolarized; much like the previous audience analogy (refer to page
8 Polarization). This is shown in Figure 10.
y
x z
NOTE: The z axis is pointing off the page
Figure 10 Propagation of the E-Field Vector of an Unpolarized Wave
PMD Issues in Advanced, Very High-Speed Networks EXFO 17
3. Birefringence, Causes and Effects
Birefringence means two (bi) refraction indexes (refringence). This is caused in optical fibers by imperfections and
perturbations in the fiber core, thus creating a polarization dependence in the fibers index of refraction (IOR). These
imperfections or perturbations may be random or imposed intrinsic stresses or random extrinsic stresses. Intrinsic and
extrinsic sources of stress produce birefringence in the fiber core.
3.1 Intrinsic Stresses
Intrinsic stresses are produced during the design and manufacture of preform, fiber and cabled fiber. Examples of
these sources are shown in Figure 11. Random intrinsic stresses create a fiber-baseline birefringence, which is always
present, relatively weak and to some extent, controllable (manageable). Imposed intrinsic stresses create a relatively
strong birefringence. Examples are shown in Figure 12.
(a) Cluster of
dopants
(b) Core non-
circularity
(c) Core-cladding
non-concentricity
(d) Radial variation of core axis (microbending)
Figure 11 Examples of Random Intrinsic Stresses on the Fiber Core
18 EXFO PMD Issues in Advanced, Very High-Speed Networks
(a) Polarization-maintaining fiber (b) High-birefringence
(Hi-Bi) fiber
Figure 12 Examples of Imposed Systematic Intrinsic Stress Causing Linear Fiber Birefringence
Twisting the fiber, as shown in Figure 13, is not going to produce any axial stress and consequently, no PMD.
Figure 13 Examples of Extrinsic Stress (twisting-torsion) Causing Circular Fiber Birefringence
PMD Issues in Advanced, Very High-Speed Networks EXFO 19
3.2 Extrinsic Stresses
Extrinsic stresses are produced during cable installation and by the environment during network operation. Since they
are random in nature and difficult to mitigate, these extrinsic stresses are the worst contributors to PMD in installed,
cabled fibers. The magnitude of the birefringence will depend on the nature and conditions of these stresses.
It is important to note that PMD in the field is directly proportional to the extrinsic stress and to the length of the
installed cabled-fiber sections, spans and links. The terms are defined as follows:
An installed cabled-fiber section is the distance between two splices
A span is the distance between two optical inline amplifiers
A link is the distance between Tx and Rx
A local, stressful event produced over a short distance (meters), such as the pinching, bending and twisting of a fiber
in a manhole or removing the fiber from a trench, is unlikely to contribute significantly to the randomly accumulated
stresses over a long distance (kilometers). This will barely or not at all increase the overall PMD.
20 EXFO PMD Issues in Advanced, Very High-Speed Networks
3.3 Birefringence Planes and Axes
As seen on the previous page, birefringent materials will exhibit two different indexes in two different planes that are
perpendicular to the axis of propagation (z plane). These planes (of polarization) will not have equal IORs. The larger
IOR will create a slower phase velocity along that plane. This axis does not have to be the x or y axis as shown in
Figure 14; it depends on the direction of perturbation (stress).
Figure 14 Difference Between x, y Axes and Fast, Slow Axes
The wave propagating in the plane of perturbation with the lowest phase velocity is said to be retarded with respect
to the other wave and its polarization axis is called the slow axis; the fast axis corresponds to a smaller IOR and a
faster phase velocity. This retardance is due to birefringence.
PMD Issues in Advanced, Very High-Speed Networks EXFO 21
4. Principal States of Polarization
This chapter will examine one of the most critical parameters responsible for PMD: the principal states of polarization.
4.1 Definition
In a birefringent medium, such as an optical fiber, there
are two states of polarization (SOP) called the principal
SOPs (PSP).
One PSP is called the slow PSP. It is aligned with the
slow axis (higher IOR or propagation index) and yields
the slowest group velocity and consequently, the longest
propagation delay. The other one is called the fast PSP.
It gives the fastest group velocity and consequently, the
shortest propagation delay.
These two PSPs are typically orthogonal.
4.2 DGD, PMD and Pulse Broadening
Pulse broadening is related to the PSP split and how the pulse SOP is launched with respect to the PSP axes. The
PSP split in the time domain is related to the PSP difference in group velocity and difference in their group delay
or DGD.
(a) Input SOP aligned with the fast-axis PSP
(b) Input SOP aligned with the slow-axis PSP
Figure 15 Unchanged Output SOP for Input SOP Aligned with PSP
22 EXFO PMD Issues in Advanced, Very High-Speed Networks
4.2.1 No Pulse Broadening
Figure 16 shows that if a pulses SOP is aligned with a PSP axis, the pulse travels from the input to the output
undisturbed. For example, a racecar that is travelling alone on a perfect racetrack will travel without interference.
Figure 16 Pulse Whose SOP is Aligned with a PSP Experiences No DGD and No Pulse Broadening
PMD Issues in Advanced, Very High-Speed Networks EXFO 23
4.2.2 Fixed DGD, PMD and Pulse Broadening
Figure 17 shows an input pulse SOP launched
with its energy shared half-half by two PSPs. If this
pulse is launched in a Polarization Maintaining Fiber
(PMF), the pulse travels from the input to the output
disturbed by the fixed stress imposed by the PMF
design. For example, a racecar is driving on one
lane of the racetrack with rough pavement along its
edge, while another car is racing on another lane
with no disturbances. There is therefore a large
fixed stress and one car lags behind the other by
a fixed, constant delay, until the end of the race.
This example illustrates the following conditions:
PMF case
Systematic constant stress applied to one PSP
axis/plane
Launch SOP aligned equally between both
PSPs (equal PSP energy in the pulse)
PMD value (mean or RMS DGD) depending
on the length.
Figure 17 Pulse Whose SOP is Aligned with Both PSPs
Experiences Fixed DGD and Pulse Broadening
In the above example, the stress (birefringence)
is constant over the full length of the SMF.
24 EXFO PMD Issues in Advanced, Very High-Speed Networks
4.2.3 Random DGD, PMD and Growing
Pulse Broadening
In Figure 18, the stress (birefringence) varies
randomly in magnitude and length over the full
distance of the fiber. The input pulse SOP is
launched with a certain amount of energy shared
over both PSPs. As a consequence, the pulse
experiences a continuous broadening caused
by that random stress from input to output.
For example, two racecars are travelling along
a racetrack and both are experiencing bumps
varying in size and length over the full distance
of the racetrack. At constant velocity, each car
suffers a delay that is comparable to the other car
and the statistical accumulation of these delays
will determine the overall delay at the finish line.
This example illustrates the following conditions:
Conventional telecom fiber case
Random, variable stress applied to both PSP axes/planes
Launched SOP shared between both PSPs
PMD value (mean or RMS DGD) depending on the SMF
length, magnitude of individual stress and degree of
randomness (ideal or semi-random)
Figure 18 SMF Case with Random Coupling, DGD
and Growing Pulse Broadening
PMD Issues in Advanced, Very High-Speed Networks EXFO 25
4.2.4 Random DGD, PMD and Growing Pulse
Broadening as a Function of Fiber Length
In Figure 18, the pulse broadens along the
entire SMF length. Figure 19 illustrates this
property for various SMF lengths.
According to Figure 19, a short-length PMD
case is established on the basis of a cable
section ( 6 km) especially considering the
constant manufacturing improvement in
lowering PMD in SMFs. The case does not
apply to patchcords, jumpers or any short
cable assemblies.
Figure 19 Growing Pulse Broadening with SMF Length
26 EXFO PMD Issues in Advanced, Very High-Speed Networks
5. PMD Specifications and International Standardization
With the advent of 40 and 100 Gbit/s using new advanced modulation formats, a number of publications on PMD
and DGD specifications have been written by IEEE 802.3, the ITU-T and its Study Group 15, as well as the IEC
Technical Committee TC 86.
5.1 PMD Cable and Link Design Specifications
A PMD link design value, PMD
Q
, is used as a PMD coefficient (PMD per unit of distance) for cables/links. The
PMD
Q
(coefficient) is used as an upper limit for the PMD coefficient of a long optical cabled SMF within a defined
concatenated link of M cable sections. This limit is defined in terms of a probability level, Q, which is the probability
that the PMD coefficient value of that long cabled SMF exceeds the PMD
Q
(coefficient). For the values of M and Q
given in Table 2 (see page 27), the corresponding values of PMD
Q
(coefficient) are not to be exceeded.
PMD Issues in Advanced, Very High-Speed Networks EXFO 27
Table 2 Recommended (standardized) Values of the Maximum PMD Coefficient
Number
of cable
sections M
Probability
level Q
SMF
PMD
Q
(coefficient)
[ps/km]
ITU-T IEC
Type Category 60793-2-50
20 1 x 10
4

or 0.01%
G.652 A and C
1
B1.1 0.5
B and D
1
B1.3 0.20
G.653 A B2 0.5
B 0.20 (Larger values can be agreed upon by manufacturers and users)
G.654 A B1.2 0.5
B and C 0.20
G.655 A and B B4 0.5
C, D and E 0.20
G.656 B5 0.20
G.657 A B6 0.20
B Not essential as the SMF supports the access network
installation with very small bending radii
1
G.652.C and G.652.D SMFs are also called low water-peak SMFs
It is important to remember that the PMD
Q
specification can only be used for cabled SMFs in production, and installed
links, spans and cable sections, with careful consideration for PMD measurement uncertainties, as discussed below.
28 EXFO PMD Issues in Advanced, Very High-Speed Networks
5.2 System PMD Specification
The maximum DGD (DGD
max
) is used as a PMD
specification in transmission systems. DGD
max
is defined
as a DGD value corresponding to the probability that the
transmission system will experience a DGD value larger
than DGD
mean
over a duration specified in Table 3.
Due to the statistical nature of PMD, a relationship
between DGD
max
and DGD
mean
can be established and
defined probabilistically using a ratio S of DGD
max
to
DGD
mean
, as shown in Table 3.
International standards organizations provide documents
on DGD
max
specifications for various applications and
bit rates (br). The following sections provide a summary
of DGD
max
specifications with a 1-dB penalty, except
otherwise indicated.
While most test equipment measure DGD
mean
(or DGD
rms
),
systems use DGD
max
. Table 3 helps translate a system
requirement into a testing requirement, based on an
acceptable BER.
Table 3 Ratio of Maximum to Mean DGD and Corresponding
Probability
DGD
max
to
DGD
mean
Ratio
Probability of
DGD
mean
being
over DGD
max
Time per year
of DGD
mean

being over DGD
max
2.5 1.5 10
3
13.1 h
3.0 4.2 10
5
22 min
3.1 2.0 10
5
10.5 min
3.2 9.2 10
6
5 min
3.25 6.19 10
6
3.2 min
3.3 4.1 10
6
2.15 min
3.4 1.8 10
6
56.6 s
3.5 7.7 10
7
24 s
3.6 3.2 10
7
10.1 s
3.7 1.3 10
7
4.1 s
3.75 8.21 10
8
2.6 s
3.775 6.5 10
8
2.0 s
3.8 5.1 10
8
1.6 s
3.9 2.0 10
8
0.63 s
4.0 7.4 10
9
0.23 s
4.1 2.7 10
9
0.09 s
4.2 9.6 10
10
0.03 s
4.3 3.3 10
10
0.01 s
4.4 1.1 10
10
0.0035 s
4.5 3.7 10
11
0.0012 s
4.6 1.2 10
11
0.00038 s
PMD Issues in Advanced, Very High-Speed Networks EXFO 29
5.3 DGD
max
Specifications for Various Applications and Modulation Formats
DGD
max
specifications are listed in Tables 4.a and b for synchronous digital hierarchy (SDH)/synchronous optical
network (SONET) non-return to zero (NRZ) and optical transport network (OTN) applications. It is assumed that at bit
rates lower than those in the table, DGD
max
becomes too large to have a significant effect on power penalty due to PMD.
Table 4.a DGD
max
Specifications for SDH/SONET NRZ Applications
NRZ applications Bit rate [Gbit/s] DGD
max
[ps]
STM-x OC-x Exact Nominal
4 12 0.622 0.622 480
8 24 1.244 1.25 240
16 48 2.488 2.5 120
64 192 9.953 10 30
256 768 39.813 40 7.5
(some SMF categories
have a PMD coefficient
too high to guarantee
this DGD)
Table 4.b DGD
max
Specifications for OTN Applications
OTN applications Bit rate [Gbit/s] DGDmax [ps]
NRZ OTU1 + FEC 2.666 120
NRZ OTU2 + FEC 10.709 30
NRZ OTU3 + FEC 43.018 7.5
(some SMF categories
have a PMD coefficient
too high to guarantee
this DGD)
NOTE: OTU: optical transport unit; FEC: forward error correction
30 EXFO PMD Issues in Advanced, Very High-Speed Networks
Table 4.c DGD
max
Specifications for NRZ 25G (OTN NRZ OTL4.4) Application
Parameters Units OTN NRZ OTL4.4
1
+ FEC
Nominal bit rate
Gbit/s
25
OTN bit rate 4 x 27.953 (111.810)
Wavelength window nm 1310
Frequency range THz 229.0 + (0.8m), m = 0 to 3
Source type SLM
Channel spacing GHz 800
Number of channels 4
SMF type ITU-T Rec. G.652 [77]
Maximum BER 1x10
-12
Maximum path penalty dB 1.5 2.5
Maximum attenuation dB 6.3 18
Reach km 10 40
DGD
max
ps 8 10.3
1
Optical channel transport lane (OTL) 4.4 (OTU4 signal running on 4 channels also called lanes) =
255/227 x 24.883200 Gbit/s = 27.952493 Gbit/s per lane or 111.810 Gbit/s total
PMD Issues in Advanced, Very High-Speed Networks EXFO 31
5.4 Proposals for DGD
max
Values for Various Applications and Modulation Formats
Tables 5 and 6 list the various proposals for system PMD specifications at 40 and 100 Gbit/s respectively. The
DGD
max
values should not be interpreted or used as system PMD specifications. The information is given only to
demonstrate the effort of the international standardization community to understand the effect of PMD at very high
bit rates and in various transport mechanisms. This continuous and evolving effort may eventually lead to system
PMD specifications.
Table 5.a Proposed DGD
max
Values for 40-Gbit/s OTN Applications Using Various Modulation Formats
OTN application 40G OTU3 + FEC
Parameters Units ODB/PSBT NRZ-DPSK
NRZ-p-DPSK
66 GHz FSR
P-DPSK
DP-QPSK
(coherent)
Bit rate Gbit/s 43.018 4 x 10.7545
Wavelength range nm 1530 - 1565 (C-band)
SMF type reference ITU-T Rec. G.652 [77] and G.655 [80]
DGD
max
(1-dB OSNR penalty) ps 5.5/7 8 7 6 75
ODB: optical duo binary; PSBT: phase-shaped binary transmission; DPSK: differential phase-shift keying; FSR: free spectral range; DP-QPSK: dual
polarization-quadrature phase-shift keying
32 EXFO PMD Issues in Advanced, Very High-Speed Networks
Table 5.b Proposed DGD
max
Values for 40-Gbit/s OTN Applications Using Various RZ-Based Modulation Formats (return to zero)
OTN application 40G OTU3+ FEC
Parameters Units RZ-QPSK
RZ-DQPSK
(coherent)
OPFDM-RZ-DQPSK RZ-AMI
Bit rate Gbit/s 43.018 2 x 21.509 2 x 21.509 43.018
Wavelength range nm 1530 - 1565
SMF type reference ITU-T Rec. G.652 [77] and G.655 [80]
DGD
max
(1-dB OSNR penalty) ps 9 18/20 20 9.5
DQPSK: Differential QPSK; OPFDM: orthogonal polarization frequency-domain multiplexing; AMI: alternate mark inversion
Table 6.a Proposed DGD
max
Values for 100-Gbit/s OTN Applications Using Various Modulation Formats
OTN application 100G OTU4
Parameters Units NRZ ODB/PSBT
Bit rate Gbit/s 4 x 27.953 (111.810) 3 x 43.018 (130) 43.018 4 x 27.953 (111.810) 3 x 43.018 (130)
Wavelength range nm 1530-1565 (not mentioned but presumed)
SMF type reference ITU-T Rec. G.652 [77] and G.655 [80]
DGD
max
(1-dB OSNR penalty,
BER = 1 x 10
4
)
ps 2.9 2.5 7 2.7 2.3
PMD Issues in Advanced, Very High-Speed Networks EXFO 33
Table 6.b Proposed DGD
max
Values for 100-Gbit/s OTN Applications Using Various Modulation Formats
OTN application 100G OTU4
Parameters Units RZ-DQPSK DPSK
Bit rate Gbit/s 43.018 4 x 27.953 (111.810) 3 x 43.018 (130) 4 x 27.953 (111.810) 3 x 43.018 (130)
Wavelength range nm 1530-1565 (not mentioned but presumed)
SMF type reference ITU-T Rec. G.652 [77] and G.655 [80]
ITU-T Rec. G.652 [77] + DCF
(80 km + 12.8 km)
DGD
max
(1-dB OSNR penalty,
BER = 1 x 10
4
)
ps 19 7.3 6.3
9 7.7
Ratio DGD to symbol duration = 10%
DCF: dispersion-compensating fiber
Table 6.c Proposed DGD
max
Values for 100-Gbit/s OTN Applications Using the Most Advanced Modulation Formats
OTN application 100G OTU4
Parameters Units DP-QPSK DQPSK DP-DQPSK
Bit rate Gbit/s
4 x 27.953
(111.810)
3 x 43.018
(130)
4 x 27.953
(111.810)
3 x 43.018
(130)
4 x 27.953
(111.810)
3 x 43.018
(130)
Wavelength range nm 1530-1565 (not mentioned but presumed)
SMF type reference ITU-T Rec. G.652 [77] + DCF (80 km + 12.8 km)
DGD
max
(1-dB OSNR penalty,
BER = 1 x 10
4
)
ps 27 23
18 15.4 36 30.8
Ratio DGD to symbol duration = 10%
34 EXFO PMD Issues in Advanced, Very High-Speed Networks
5.5 DGD
max
Specifications for Ethernet Applications
The tables below show the Ethernet system PMD specifications for various bit rates.
Table 7.a DGD
max
Specifications for 40-Gbit/s Ethernet Serial Application [95]
OTN application
40GBASE-FR
Parameters Units
Signaling rate Gbit/s 41.25
Channel spacing nm
Wavelength range
Tx nm 1530 to 1565
Rx 1290 to 1330 1530 to 1565
SMF type IEC 60793-2-50 type B1.1, B1.3 , B6 [96]
Reach km 0.002 to 2
DGD
max
(2-dB penalty, BER = 1 x 10
-12
) ps 0.5
PMD Issues in Advanced, Very High-Speed Networks EXFO 35
Table 7.b DGD
max
Specifications for 40-Gbit/s and 100-Gbit/s Ethernet Applications [97]
OTN application
Parameters Units
Signaling rate Gbit/s
4 lanes x 10.3125 GBd
(41.25 Gbit/s)
4 lanes x 25.78125 GBd
(103.125 Gbit/s)
Channel spacing
nm 20 (CWDM)
GHz 800 (DWDM)
Center wavelength
(wavelength range)
nm
1271 nm (1264.5 to 1277.5)
1291 nm (1284.5 to 1297.5)
1310 nm (1304.5 to 1317.5)
1331 nm (1324.5 to 1337.5)
Center frequency
(wavelength range)
THz
231.4 THz (1294.53 to 1296.59)
230.6 THz (1299.02 to 1301.09)
229.8 THz (1303.54 to 1305.63)
229.0 THz (1308.09 to 1310.19)
SMF type IEC 60793-2-50 type B1.1, B1.3 , B6
Reach km 0.002 to 10 0.002 to 10
0.002 to 30
0.002 to 401
DGD
max
(2.6-dB link penalty,
BER = 1 x 10
12
ps 10 8 10.3
1
Links > 30 km with the same power budget are considered engineered links. Attenuation for such links needs to be less than the worst case specified
for B1.1, B1.3 or B6A SMF
36 EXFO PMD Issues in Advanced, Very High-Speed Networks
6. PMD-Induced Pulse Broadening and Penalty
DGD
max
is set to allow no more than a specified power penalty. The worst case power penalty is also affected by the
transmission format: NRZ or RZ.
For 40-Gbit/s NRZ applications, a 1-dB penalty allowance corresponds to a DGD
max
of 7.5 ps, which is the limit on
the DGD at the receiver. If half the penalty is allowed, then the DGD
max
decreases, whereas if twice as much penalty
is allowed, then the DGD
max
increases, giving the system more PMD tolerance.
Figure 20 shows the mean PMD-induced power penalty as a function of PMD.
Figure 20 PMD-Induced Mean Power Penalty as a Function of PMD and Bit Rate
16
14
12
10
8
6
4
2
0
0
2.5 Gbit/s
5 Gbit/s
10 Gbit/s
40 Gbit/s
20 40 60
M
e
a
n

p
o
w
e
r

p
e
n
a
l
t
y

[
d
B
]
PMD [ps]
80 100
PMD Issues in Advanced, Very High-Speed Networks EXFO 37
7. PMD Tests and Measurements
PMD measurement has been the subject of studies and publications including lengthy discussions in international forums
since the early 90s, when operators began to observe the impact of PMD at 10 Gbit/s as random network outages.
Since then, quite a few PMD measurement and test methods have been proposed. The following is a list containing
the main ones, in alphabetical order:
Fixed analyzer
- Extrema counting (FA-EC) (standardized; N/A)
- Fourier transform (FA-FT) (standardized;
commercially available)
Interferometric method:
- Generalized interferometry (GINTY) (standardized;
commercially available)
- Traditional interferometry (TINTY) (standardized;
commercially available)
Modulation phase shift (MPS) (standardized; N/A)
Poincar sphere arc method (PS) also called SOP
method (standardized; N/A)
Polarization phase shift (PPS) (standardized; N/A)
Scrambling SOP analysis (SSA) (standardized;
commercially available)
Stokes parameter evaluation:
- Jones matrix eigenanalysis (JME) (standardized;
commercially available)
- Poincar sphere analysis (PSA) (standardized; N/A)
NOTE: N/A means that the method is either only published or not commercially available.
Only the methods that are available and usable in field instruments for installed-link PMD testing are discussed hereafter.
38 EXFO PMD Issues in Advanced, Very High-Speed Networks
7.1 Description of the Available Test Methods
7.1.1 Fixed Analyzer Fourier Transform
FA-FT uses one SOP from a BroadBand Source (BBS) or a Tunable Laser Source (TLS) and one SOP from a
corresponding optical spectrum analyzer or power meter depending on the setup, as shown in Figure 21.
Figure 21 Schematic of FA-FT Test Method Setup
The method measures the statistical power variation of the SOP going across and changed by a DUT during a
wavelength scan (this is where the term wavelength scanning originated).
polarizer
polarizer
(0
o
or 90
o
)
DUT
A
n
a
l
y
z
e
r

S
o
u
r
c
e
B
B
S
T
L
S
S
O
P
l
S
O
P
o
O
S
A
P
M
PMD Issues in Advanced, Very High-Speed Networks EXFO 39
Similar to TINTY, which will be discussed below, FA-FT must respect stringent requirements (these stringent
requirements will not be repeated in the TINTY section to lighten the text):
If a BBS is used, its spectrum must be Gaussian with no spectral power ripples
The mode coupling must be random (no mixed coupling measurements are allowed); Figure 22 illustrates Random
Mode Coupling (RMC)
Figure 22 Schematic of FA-FT Test Method Setup
0
Gaussian fit
5 10 15
Delay [ps]
20 25 30 35 40 45
40 EXFO PMD Issues in Advanced, Very High-Speed Networks
The SMF must be very long (kms)
The resulting interferogram (half) must be ideally Gaussian (half) with a large number of fringes (100s to 1000s)
The resulting interferogram (half) must extend to zero by at least three times its RMS half-width
The PMD must be large (no fraction-of-a-ps measurement is allowed); no cable section measurements are allowed
No measurements of optical amplifiers or narrowband components or links containing them are allowed with the BBS
The TLS Degree of Polarization (DOP) must remain high during the measurement period
The frequency increment/spacing must be constant
The source spectrum (BBS) or range (TLS) is limited to a predefined fixed window, so the statistical averaging is
limited to a finite amount of wavelengths (or window) and consequently, the mean or RMS value comes with an
uncertainty that is directly proportional to the spectral windowing.
In the case of random mode coupling (RMC), like any other PMD test method, the light source window must be
as wide as possible (e.g., typically 200 nm, theoretically to infinity) in order to get the largest possible amount of
statistical sampling. This allows the statistical average to be determined with a minimum amount of uncertainty. This
is why it is difficult to obtain an accurate PMD value with short SMF and low PMD: the uncertainty simply becomes
unacceptably high.
PMD Issues in Advanced, Very High-Speed Networks EXFO 41
7.1.2 Interferometric Methods (TINTY/GINTY)
The interferometric PMD test method is divided into two different analyses:
A traditional analysis restricted to a number of stringent conditions for getting DGD
RMS
in RMC regime (large PMD
value, long SMF) as defined in the previous section
An unrestricted, generalized analysis for getting DGD
RMS
in any mode coupling regime (any SMF type and length
and any PMD value).
42 EXFO PMD Issues in Advanced, Very High-Speed Networks
7.1.2.1 Traditional Analysis
The traditional interferometric PMD test method, TINTY, is based on a linearly polarized BBS and an interferometer
(in which orthogonal SOPs interfere) as well as a polarizing analyzer that is used at the input, as shown in Figure 23.
Figure 23 TINTY Schematic and Typical Test Result Interferogram
Source DUT Analyzer
SOP
l
polarizer
SOP
O
polarizer
(analyzer)
TINTY interferogram
Delay [ps]
BBS Interferometer
PMD Issues in Advanced, Very High-Speed Networks EXFO 43
7.1.2.2 Generalized Analysis
In case of GINTY, there are no restrictive conditions and the method applies to any situation starting at the lowest
PMD (zero) of a very complex DUT or link with mixed mode coupling or to any light source shape and spectrum.
Figure 24 GINTY Schematic and Typical Cross-Correlation Interferogram Test Result
Source
DUT
Analyzer
SOP
l
scrambling
SOP
O
scrambling
Polarization
diversity
detector
BBS
Delay [ps]
GINTY interferogram
Interferometer
P() P||()
Data
44 EXFO PMD Issues in Advanced, Very High-Speed Networks
7.1.3 Scrambled State-of-Polarization Analysis (SSA)
7.1.3.1 General Theory
SSA measures the power from the DUT at two closely spaced frequencies (i.e., a frequency pair) with k = 1 to N,
where N is the total number of pairs across a selected frequency range. Every pair is associated to a set of randomly
and uniformly scrambled in/out-SOPs in order to obtain the DGD or one randomly scrambled in/out-SOP to obtain
the PMD.
Figure 25 illustrates the concept of frequency pairs, the amount of center frequencies, frequency spacing and
frequency ranges in an SSA implementation.
NOTE: I/O-SOPs are randomly and uniformly scrambled at every pair
Figure 25 SSA Concepts of Frequency Pair, Spacing and Range and I/O-SOP Scrambling
PMD Issues in Advanced, Very High-Speed Networks EXFO 45
From a measurement standpoint, a large number of I/O-SOPs (in the thousands) improves the uncertainty, but it
requires a long measurement and averaging time. On the other hand, a small number of I/O-SOPs (in the tens)
requires a shorter measurement and averaging time, but has increased uncertainty.
7.1.3.2 Experimental Implementations
With the above approach, SSA can be used in an end-to-end forward implementation or in a single-end roundtrip
implementation. Each implementation has its own experimental configuration and application conditions. Table 8
describes the various SSA implementations, conditions and results.
Table 8 SSA Experimental Configuration Matrix Using TLS
Parameters
SSA implementation
End-to-end forward Single-end roundtrip
SOP Independent randomly and uniformly scrambled I-SOP and O-SOP Combined randomly and uniformly scrambled I/O-SOP
Light source CW TLS Pulsed TLS
Detection Polarization diversity detector
46 EXFO PMD Issues in Advanced, Very High-Speed Networks
7.1.3.3 Measurement of PMD as a Function of Distance
Another SSA implementation is the quantitative
measurement of PMD (DGD
RMS
) as a function of
distance along an installed, cabled SMF using random
scrambling polarization optical time domain reflectometry
(RS-POTDR).
Since the implementation is based on OTDR, the
single-end roundtrip configuration is used, as shown in
Figure 26.
The same SSA theory applies here, except that in this
case, the average is calculated as a function of distance
using a certain distance interval from the installed SMF
related to the selected OTDR pulse width.
The PMD(z) value is calculated from the value of
local differences between pairs of OTDR traces,
which correspond to random pairs of closely-spaced
frequencies/wavelengths.
Figure 27 provides an example of a bidirectional test
result showing the cumulative PMD of an 18.9-km SMF
with splices and PMD emulators. Connectors are also
used at both ends.

Figure 26 Single-End Roundtrip Configurations with Pulsed TLS,
Combined SOP
I
and SOP
O
Scrambling and PDD
Figure 27 Example of SSA Cumulative PMD Test Result
as a Function of Distance
Source
DUT
Polarization
diversity
detector
Circulator
Open connector
or
Optical reflector
Pulsed TLS
P() P||()
Data
Analyzer
SOP
I/O
scrambling
PMD Issues in Advanced, Very High-Speed Networks EXFO 47
7.2 Uncertainty in PMD Tests and Measurements
The uncertainty of PMD test and measurement results
is based on two elements, as shown in Figure 28.
The first element is the measurement uncertainty as
per the selected instrumental implementation and
equipment parameter settings.
The second element is the fundamental uncertainty of
the selected frequency/wavelength range and the PMD
value. This is called Gisins Uncertainty, and this is what,
in many conditions, limits the measurement and/or
accuracy of low PMD values. This uncertainty is fairly
small in absolute terms, but can be significant with low
PMD measurements (relatively speaking).
7.3 Calibration of PMD Test and Measurement Instruments
The calibration of a PMD test instrument, like any other test instrument, must be done using a traceability process and
a set of procedures that involve a number of critical steps.
The first step consists in using standard reference material (SRM), designed by and available from an internationally
recognized independent national metrology laboratory (INML). This SRM is provided with a calibration certificate
stating the guaranteed value of PMD and its uncertainty obtained from DGD measured over a fixed frequency/
wavelength range at precise controlled environmental conditions. To obtain a guaranteed value, the INML designs
and builds its own instrumentation in its own laboratory.
-5
PMD [ps]
10 15 20 25 30 0
Instrumental and roundtrip
uncertainty (5 % to 7 %)
G
i
s
i
n

s

u
n
c
e
r
t
a
i
n
t
y
(
1
0

%

t
o

4
0

%
)
Figure 28 Example of Typical Experimental Uncertainty
48 EXFO PMD Issues in Advanced, Very High-Speed Networks
7.3.1 Applicability Matrix for the Available PMD Test Methods
Table 9 gives an overview of the PMD test methods and their applications.
Table 9 Applicability of PMD Test Methods
Available PMD Test Methods
FA-FT
INTY
SSA
TINTY GINTY
Configuration E2E E2E E2E 1E
Conditions
Applications
RMC RMC Any MC Any MC
Fibers and cables in factory PMD > 1 ps, long fiber X
Passive components in factory
Pumped amplifiers in factory
Aerial links in the field PMD > 1 ps, long fiber X X
Unamplified links in the field PMD > 1 ps, long fiber X X
Amplified links in the field
With TLS/OSA PMD > 1 ps,
long fiber
X
Acknowledgements
This guide would not have been possible without the enthusiasm and teamwork of EXFO staff, particularly the hard
work and technical expertise of the Product Line Management team.
No part of this guide may be reproduced in any form
or by any means without the prior written permission of EXFO.
Printed and bound in Canada
ISBN 978-1-55342-102-3

Legal Deposit-National Library of Canada 2012
Legal Deposit-National Library of Quebec 2012

2
0
1
2

E
X
F
O

i
n
c
.

A
l
l

r
i
g
h
t

r
e
s
e
r
v
e
d
.

P
r
i
n
t
e
d

i
n

C
a
n
a
d
a

1
2
/
1
2

2
0
1
2
0
6
4
5

S
A
P

1
0
6
4
4
8
3
For details on any of our products and
services, or to download technical and
application notes, visit our website at
www.EXFO.com.