Polarisation Effects and Measurements in Optical Fibre Systems

Polarisation
Effects and Measurements

in Optical Fibre Systems
David Ives
Measurement Good Practice Guide No. 67
Polarisation
Effects and Measurements
in Optical Fibre Systems
David Ives
Photonics
National Physical Laboratory
Abstract: This document has been written to give guidance and understanding to the array of
polarisation properties and effects found within optical fibres, optical fibre components and
optical fibre systems. The various polarisation properties and effects are described along with
measurement and calibration techniques.
© Crown Copyright 2004
Reproduced by permission of the controller HMSO
ISSN 1368-6550
April 2004
National Physical Laboratory

Teddington, Middlesex, United Kingdon, TW11 0LW
Website: www.npl.co.uk
This document has been produced for the Department of Trade and Industrys National
Measurement Systems Directorate under contract GBBK/C/011/00005. Extracts from this
good practice guide may be reproduced provided the source is acknowledged and the extract
is not taken out of context.
Approved on behalf of Managing Director, NPL

by S Pollitt, Director of Enabling Metrology Division
Polarisation Effects and Measurements in Optical Fibre Systems
Contents
1. Polarised Light .................................................................................................................1

1.1 Polarised Light ...................................................................................................1
1.2 Polarisers ............................................................................................................3
1.3 Birefringence in Crystaline Materials ................................................................5
1.4 Waveplates .........................................................................................................6
1.5 Optical Activity and Circular Birefringence ......................................................8
1.6 Electro-optic and Magneto-optic effects ............................................................8
1.7 Effect on Reflection and Transmission ..............................................................9
2. Mathematical Descriptions of Polarised Light............................................................12

2.1 Jones Vector and Jones Matrixes .....................................................................12
2.2 Stokes Vector and Mueller Matrixes................................................................13
2.3 Poincaré Sphere................................................................................................14
2.4 Useful Matrices ................................................................................................16
3. Polarisation Effects in Optical Fibres ..........................................................................18

3.1 Birefringence in Optical Fibres ........................................................................18
3.2 Causes of Linear Birefringence........................................................................18
3.3 Causes of Circular Birefringence .....................................................................20
3.4 Local Effects of Fibre Birefringence................................................................20
3.5 Global Effects in Fibre, Polarisation Mode Dispersion ...................................21
3.6 Polarisation Maintaining Fibre.........................................................................28
3.7 Polarisation Dependent Loss............................................................................29
3.8 Polarisation Dependent Response ....................................................................30
3.9 PMD and PDL Combined ................................................................................30
4. The Effects of Polarisation on Data Transmission......................................................31

4.1 Impact of first order PMD................................................................................31
4.2 Impact of Second Order PMD..........................................................................32
4.3 PMD of Installed Links ....................................................................................33
4.4 Optical PMD Compensation ............................................................................33
4.5 Compensation Advantage ................................................................................34
5. Polarisation Measurement ............................................................................................36

5.1 Polarisation State..............................................................................................36
5.2 Birefringence and Beat Length ........................................................................37
5.3 Differential Group Delay and Polarisation Mode Dispersion..........................40
5.3.1 Polarisation State Analyser based measurement of DGD and PMD ...40
5.3.2 Fixed Analyser Method to measure PMD............................................44
5.3.3 Interferometic Method to measure PMD .............................................48
5.3.4 Phase Shift measurement of DGD and PMD .......................................50
5.3.5 Statistical Uncertainty in measurement of PMD..................................52
5.4 Polarisation Dependent Loss and Polarisation Dependent Response ..............53
5.5 Polarisation Extinction Ratio ...........................................................................55
6. Calibration of Polarisation Measurement Equipment ...............................................56

6.1 Calibration of Polarisation State Analysers .....................................................56
6.2 Calibration of PMD Measurement Equipment.................................................57
6.3 Calibration of PDL Meters...............................................................................59
6.4 Calibration of PER Meters ...............................................................................60
7. Glossary and References ...............................................................................................61

7.1 Glossary of Terms ............................................................................................61
7.2 NPL Publications..............................................................................................65
7.3 References ........................................................................................................66
1. Polarised Light
1.1 Polarised Light
The solution of the electro-magnetic wave equation shows light to be an oscillating electric
and magnetic field. For a propagating wave where the electric field vector has a defined
orientation the light is said to be polarised[1][2]. The simplest case is where the electric field
vector is confined to a plane containing the direction of propagation. This is known as linear
polarised light, see Figure 1.
z
x
Ey
Ex
E
Figure 1 Electric Field Vector evolution for linear horizontal polarised light.
In contrast, natural light has no preferred orientation of the electric field and is said to be
unpolarised. An example is light emitted from a hot thermal source where each photon has its
electric field orientated differently. The degree of polarisation is defined as the polarised
intensity divided by the total intensity and is 1 for polarised light and 0 for natural
unpolarised light.
For the solution of the electro-magnetic wave equation in x, y and z space, where the wave is
propagating in the z direction, we can consider the components of the electric field aligned
with either the x axis or y axis. This describes two orthogonal states of polarisation, the linear
horizontal and linear vertical states of polarisation. By combining two orthogonal polarised
waves the resultant wave can form any state of polarisation depending on the exact amplitude
and phase between the two waves. Thus by combing linear horizontal and linear vertical
polarised light with the same amplitude and phase a linear polarised wave is produced with
the electric field orientated at 45º, see Figure 2, and by combining a linear horizontal and
linear vertical polarised light with the same amplitude and π/2 phase difference a circular
polarised wave results, see Figure 3. In a circular polarised wave the electric field vector
rotates about the z-axis over a wave period with constant amplitude where as in a linear
polarised wave the electric field remains in the plane containing the propagation direction and
1
the amplitude varies sinusoidally during a wave period. For other combinations an elliptical
polarised wave results where the amplitude and angle of the electric field vector evolve
during the wave period.
y
z
x
E Ey
Ex
Figure 2 Electric field components Ex and Ey combine in phase to give a linear polarisation
state at 45º.
y
z
x
Ey
Ex
E
Figure 3 Electric field components Ex and Ey combine in quadrature to give a left hand
circular polarisation state.
2
1.2 Polarisers
Unpolarised or natural light can be polarised by passing it through a polariser. This allows the
transmission of one state of polarisation while absorbing or reflecting the other. For an
unpolarised light wave this results in only half the power being transmitted. For the case of a
linear polariser the resulting output wave will be linearly polarised. If this wave is incident on
a second linear polariser the amount of transmitted light depends on the relative orientation of
the two polarisation axes. The second polariser can be thought to pass the electric field
component orientated parallel to its polarisation axis.
I(θ) = [E 0 cos(θ)]2 = I(0) cos 2 (θ) (1)
where I(0) is the intensity of the input wave and θ the angle between the axis of the two linear
polarisers, this assumes ideal lossless linear polarisers. A second polariser used to determine
the component of a beam aligned with a state of polarisation is often referred to as a
polarisation analyser.
y
x
z
Axis(1)
θ Natural Light
Axis(2) Linear Polariser(1)

E
E cos(θ)
Linear Polariser(2) / Analyser
Figure 4 Effect of two linear polarisers crossed at an angle θ.
Polarisers can be made from dichroic materials that absorb one polarisation state in
preference to its orthogonal state. This is true of the wire grid type polariser where a grid of
fine wires allows the polarisation state with its electric field perpendicular to the wires to be
transmitted while absorbing the polarisation state where the electric field is parallel to the
wires. Wire grid polarisers can be made using a grid of fine wires for long wavelengths but
for visible and near infrared optical radiation a similar effect exists in long conducting
crystals and in Polaroid sheet.
Polarised light can also be formed from reflections from dielectric surfaces at Brewster’s
angle. At Brewster’s angle the light polarised with its electric field vector in the plane of
incidence (the plane that contains the direction of propagation and the normal to the surface)
is not reflected. This leads the reflected beam to be fully polarised with the electric field
vector perpendicular to the plane of incidence. In non-magnetic materials it can be considered
that light polarised with its electric field in the plane of incidence is not reflected since the
electric field of the transmitted beam would be in the direction of propagation, where no
3
electromagnetic propagation is supported. At boundaries where the magnetic permeability is

constant Brewster’s angle, θP, is given by
n
tan(θP ) = t (2)
ni
where ni is the refractive index of the incident medium and nt the refractive index of the
transmitting medium. This effect is usefully used in Polaroid sunglasses to block the
sparkling reflection from water.
θp
Air ni=1.000
Silica nt=1.444
Electric Field
Polarisation Direction
Electric Field
Polarised Into Page
Figure 5 Light reflected at Brewster’s angle from a surface is fully polarised with the electric
field polarised perpendicular to the plane of incidence.
x
z
Figure 6 Scattering of polarised light by a small dipole.
4
Scattered light can also be strongly polarised. Scattering from small particles and molecules
forms a dipole radiation pattern with zero scattering in the direction parallel to the incident
electric field. Thus if we look at light scattering at right angles to the incident beam there will
be partial polarisation with the preferred electric field perpendicular to the incident and
scattered directions.
The highest quality linear polarisers are produced from birefringent crystals arranged to
spatially separate the two polarisation states. This can be as a result of different refracting
angles or by selectively reflecting one polarisation near Brewster’s angle, see Figure 7.
Optical Axis
Optical Axis
Figure 7 Wollaston prism acting as a polarising beam splitter.
1.3 Birefringence in Crystaline Materials

Many crystalline materials are not optically isotropic; this is due to an anisotropy in their
crystal structure. The anisotropy in the crystal structure leads to an anisotropy in the dielectric
polarisability that in turn leads to an anisotropy in the refractive index. A material that
displays two different indices of refraction is said to be birefringent. The simplest form is a
uniaxial crystal, for example Calcite and Quartz, where the crystalline structure has a single
axis of symmetry. The electric polarisability of the crystal is different for electric fields
parallel and electric fields perpendicular to this axis of symmetry, known as the optical axis.
As a result the refractive index for electromagnetic waves with their electric field parallel and
electromagnetic waves with their electric field perpendicular to this optical axis are different.
In the simplest case where the optical axis is perpendicular to the propagating beam, the input
optical beam is split into two, that with its electric field perpendicular to the optical axis and
that with its electric field parallel to the optical axis, these two beams, known as the ordinary
and extraordinary rays propagate at different velocities. This leads to a change in polarisation
state of the beam by changing the relative optical phase between the two beam components.
If the optical axis is not perpendicular to the direction of propagation then the electric field
component that is parallel to the optical axis will experience a different polarisability to the
electric field component perpendicular to the optical axis. This will cause the beam
propagation direction to be deviated. The effect is observed in calcite as a double image
where the ordinary and extraordinary images become separated, see Figure 8.
5
e-ray
o-ray
Optical axis
Figure 8 Double images in Calcite
In isotropic media, isotropic crystals and glasses there is no inherent birefringence as there is
no asymmetry. External influences such as stress can induce a birefringence through the
stress optic effect. This effect is important to polarisation effects in optical fibres.
1.4 Waveplates
Birefringent crystals cut such that their optical axis lies in the plane of the surface are used to
form waveplates. An electromagnetic wave propagating perpendicular to the surface will be
split into two components, one with its electric field parallel to the optical axis and one with
the electric field perpendicular to the optical axis. These two components will propagate at
different velocities giving an optical phase shift between them. This leads to a change in the
state of polarisation of the emerging wave.
y
Quarter Waveplate
z x
Linear Polarised
Circular Polarised Optical Axis
Input
Output
Figure 9 A quarter waveplate converts linear polarised light to circular polarised light when
the input polarisation is at 45 degrees to the optical axis.
6
For example if a linear polarised wave is incident on a quarter wave retarder (that is a crystal
plate where the phase retardance between the fast and slow axis is one quarter wave) at 45º to
the birefringent axis then the emerging wave will have a circular state of polarisation, see
Figure 9. For other retardances and angles of the retardance axis, the incident wave is
decomposed into two orthogonal component waves that are aligned with the fast and slow
axes of the crystal. The emerging waves are recombined taking into account their amplitudes
and new phases to obtain the output state of polarisation. In general this will be an elliptical
state of polarisation.
A quarter waveplate made from birefringent material is tuned to the operating wavelength.
The thickness of the plate is adjusted so that the difference in phase velocity causes a phase
shift of π/2 radians. In quartz the thickness required is extremely thin and so many quarter
waveplates are either multi order that is the phase shift is equal to (2n+½)π radians or are
composed of two plates arranged with there fast and slow axis perpendicular such that the
combined phase shift is only π/2 radians. Multi order waveplates show a stronger wavelength
dependence. Some waveplates are produced from multiple layers of plastic film such that the
various dispersions of each layer cancel giving a π/2 radians phase shift over a wide
wavelength range.
Quarter Wave Rhomb
Half Wave Rhomb
Figure 10 Fresnel Rhombs as achromatic retarders
It is also possible to create achromatic quarter waveplates based on the phase change on
reflection in a Fresnel Rhomb. The angles are chosen such that each reflection gives a π/4
radian phase change between the two polarisation components. Such Rhombs are achromatic
except for material dispersion. For a glass air interface the glass must have a refractive index
greater than 1.497 to obtain this phase difference. This is possible for BK7 for an angle of
incidence of 53.4º at 1550 nm.
7
1.5 Optical Activity and Circular Birefringence

Crystals that have a twist or spiral structure exhibit optical activity. The plane of polarisation
of a linearly polarised optical input rotates as it passes through the material. Crystalline
quartz is optically active for waves propagating along the optical axis. Also solutions of many
organic substances show optical activity for example a solution of sugar. In such systems an
input linearly polarised beam is split into its left and right hand circular polarised components
and it is these states that propagate at differing velocities. When the left and right hand
circular polarised components are recombined they form a linear polarised output and the
phase shift between them determines azimuth of the linear polarisation. Since the difference
in propagation velocities is between left and right handed circular states the phenomena is
sometimes known as circular birefringence.
1.6 Electro-optic and Magneto-optic effects

In a number of crystalline materials an external static electric or magnetic field can modify
the refractive index properties. There are a number of different effects.
The Kerr effect is observed in an isotropic medium where the electric field causes the
medium to become birefringent with the optical axis aligned with the applied electric field.
The birefringence δβl in rads.m-1 is given by
δβl = 2π BK E 2 (3)
where BK is the Kerr constant and E the electric field strength. It should be noted that in the
Kerr effect the birefringence is proportional to the square of the electric field.
The Pockels effect occurs in crystals that do not have a centre of symmetry. The electric field
modifies the refractive index through the electro-optic tensor. With crystals such as KDP the
effect is through a cross term in the tensor such that for the applied electric field along the
crystals z-axis, the birefringence occurs along an axis at right angles in the x-y plane. This
allows the applied electric field to be in the direction of propagation of the optical beam or
perpendicular to the optical beam. However with crystals such as LiNbO3 an applied electric
field along the crystal z-axis directly affects the refractive index for optical beam with their
electric field polarised parallel to the z-axis. The birefringence is changed with the optical
axis parallel to the z-axis.
In the case of KDP and similar 42m crystal structures the change in birefringence is given by
2π 3
δβl = n 0 r63 E (4)
λ
where r63 is the element from the electro-optic tensor and E is the electric field applied along
the crystals z-axis. The birefringence change has an optical axis in the x-y plane
In the case of LiNbO3 and similar 3m crystal structures the change in birefringence is given
by
δβl =
2π 1
λ 2
[ ]
E r33n 3e − r13n 30 (5)
where rij are the elements from the electro-optic tensor and E is the electric field applied
along the crystals z-axis. The birefringence change has an optical axis along the crystal z –
axis.
8
It should be noted that with these effects the change in birefringence is proportional to the
electric field.
The Faraday effect occurs in materials when a strong magnetic field is applied in the
direction of propagation of the optical beam. The magnetic field causes a circular
birefringence, a velocity change between left and right circular polarised light. This causes
the plane of polarisation of a linear polarised beam to rotate as it propagates. The direction of
rotation of the plane of polarisation depends on whether the optical beam is propagating in
the direction of the magnetic field or against the direction of the magnetic field. A circular
birefringence δβC rads.m-1 is induced, given by
δβC = 2 VV H (6)
where VV is the Verdet constant.
The electro-optic effect is used to produce optical modulators and polarisation controllers.
The Faraday effect is used to great effect in optical isolators that only allow light to propagate
in the forward direction.
1.7 Effect on Reflection and Transmission

Light impinging on a dielectric surface is reflected from the surfaces and transmitted into the
bulk. The reflection coefficient is a function of the angle of incidence, polarisation of the
input light and the dielectric medium. For a lossless medium the total power remains
unchanged.
The two polarisation states that are considered are those with the electric field vector in the
plane of incidence, E||, and that with the electric field vector perpendicular to the plane of
incidence, E⊥. The plane of incidence is the plane containing the normal to the surface and
the direction of propagation of the incident beam, see Figure 11.
E|| E⊥
θi θr θi θr
Air ni=1.000
Silica nt=1.444
θt θt
Electric Field
Polarisation Direction
Electric Field
Polarised Into Page
Figure 11 Definition of E|| and E⊥. The plane of incidence is the plane of the paper.
9
For normal incidence on an isotropic medium the reflection coefficient for the two
polarisation states is identical. However there is a difference in the optical phase shift
between the two beams of π radians. This is because the perpendicular polarised beam, E⊥,
will experience an additional phase shift of π radians because the direction of the beam is
reversed thus to the observer a field pointing to the left as looking into the beam will be
pointing to the right after reflection. This is most significant when a circular polarised beam
impinges on the surface as the reflected beam has the opposite handedness.
Solving the boundary conditions at the surface we can obtain the amplitude reflection
coefficient for the electric field parallel and perpendicular to the plane of incidence.
nt n
 Er  µt
cos(θi ) − µ i cos(θ t )
r|| =   = n ni
i
(7)
E
 i || µ cos(θi ) + µ cos(θ t )
t
t i
ni n
E  µi
cos(θi ) − µ t cos(θ t )
r⊥ =  r  = ni nt
t
(8)
 Ei ⊥ µi
cos(θi ) + µt
cos(θ t )
where in most dielectric materials µi = µt = 1.
r||
Amplitude Reflection Coefficient
Amplitude Reflection Coefficient
0 0.8
θp
0.2 0.6
r⊥
0.4 0.4
r⊥
0.6 0.2
0.8 0
r|| θp θc
1 0.2
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90
θi (degrees) θi (degrees)
Figure 12 Reflection coefficient from an air Figure 13 Reflection coefficient from a silica
– silica (n=1.444) surface. (n=1.444) - air surface.
3.14 3.14
E⊥
Relative Optical Phase Shift (rads)
Relative Optical Phase Shift (rads)
E||
2.36 2.36
1.57 1.57 E⊥
0.79 0.79
E||
0 0
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90
θi (degrees) θp θp θc θi (degrees)
Figure 14 Reflection relative optical phase Figure 15 Reflection relative optical phase
shift from an air - silica (n=1.444) surface. shift from a silica (n=1.444) - air surface.
10
The optical phase change on reflection is given by

 δ  n n 2 sin 2 (θi ) − n 2t
tan ||  = i i 2 (9)
2 n t cos(θi )
δ  n i2 sin 2 (θi ) − n 2t
tan ⊥  = (10)
 2  n i cos(θi )
3.14
Optical Phase Difference (rads)
2.36
1.57
BK7
0.79
SiO2
0
0 10 20 30 40 50 60 70 80 90
θi (degrees)
Figure 16 The optical phase difference between || and ⊥ polarised beams on internal
reflection.
Reflections from Metallic surfaces show similar curves to dielectric media. The reflection
coefficient for light polarised perpendicular to the plane of incidence increases to one as the
angle of incidence increases to 90º while the reflection coefficient for light polarised parallel
to the plane of incidence decreases to a minima at the principle angle of incidence (equivalent
to Brewster’s angle) and then increases to one as the angle of incidence increases to 90º.
11
2. Mathematical Descriptions of Polarised Light

2.1 Jones Vector and Jones Matrixes
As stated previously the polarisation state of an optical wave can be described by the
superposition of two orthogonally polarised waves with differing amplitudes and phases.
Jones Vectors[3] provide a two-element vector to describe the amplitude and phase of the
electric field in the two basis polarisation states. The electric field is described by
 a1  iωt
  e = [a1 1 + a 2 2 ] eiωt (11)
 a 2 12
where |1> and |2> are the two basis polarisation states such that <1|1> = <2|2> = 1 and <1|2>
= <2|1> = 0, the orthogonality condition and a and b are complex values giving the amplitude
and phase of the component of the wave in each of the basis states. The vector is normalised
such that a1.a1*+ a2.a2* = 1, where the * denotes the complex conjugate.
It is usual to take the basis states as linear polarised along x and y ie the linear horizontal and
linear vertical polarised states. Thus a linear vertical polarised wave can be represented by
1
  (12)
0
The most general elliptical polarised state and its orthogonal state are given by
 cos(ν) ei 2   − sin(ν) ei 2 
∆ ∆
  and   (13)
 sin(ν) e −i ∆2   cos(ν) e −i ∆2 
   
where ν is the azimuth angle of the major axis of the ellipse and ∆ is the phase difference
between the linear vertical and linear horizontal electric field components, ∆=0 describes
linear polarised states. For examples of common states of polarisation see Table 1.
The Jones Matrix describes the transmission of an optical element and the effect on the Jones
Vector. The optical element is described by a 2 x 2 matrix, M, such that the output
polarisation state EO is related to the input polarisation state EI by
EO = M EI (14)
The matrix, M, can be used to describe polarisers and birefringent elements. The rotation
matrix R(θ) can be used to rotate elements such that an optical element M at an azimuth angle
θ is given by
M(θ) = R (−θ) M R (θ) (15)
where
 cos(θ) sin(θ) 
R (θ) =   (16)
 − sin( θ) cos( θ) 
For example the effect of a quarter waveplate at 45º to a linear polarised wave is given below
 1 −1 
 i4
π
0  2
1 1 
2   1 1
 2  e
1
E O = R (−45º ) M R (45º ) E I = 1 2
1  −i π4  −1
 =
1  
  (17)
 0 2 i
 2 2 0 e  2 2  
ie a right hand circular output polarisation. For a wave with polarisation state EI incident on a
number of optical elements M1 , M2 … MN the output polarisation state EO is given by
E O = M N ...M 2 M1E I (18)
12
For examples of Jones Matrices for common optical elements see Table 2.
2.2 Stokes Vector and Mueller Matrixes

While Jones vectors describe the electric field of a polarised light wave the Stokes vector[3]
describes the intensity and can be directly measured. Stokes vectors are able to describe
unpolarised light but cannot describe coherent combinations of optical waves. The Stokes
vector is a four element vector where the elements describe the intensity of the beam, the
difference in intensity for the beam passed through a horizontal linear polariser and a vertical
linear polariser, the difference in intensity for the beam passed through a linear polariser at
+45º and a linear polariser at –45º and the difference in intensity for the beam passed through
a right hand circular linear polariser and a left hand circular polariser.
 s0   I0 
   
 s1   I H − I V 
S= = (19)
s2 I+ − I− 
   
s   I − I 
 3  R L 
where I denotes the intensity of the wave passed through various analysers. H = linear
horizontal polarisation state, V = linear vertical polarisation state, + = linear polarisation state
at +45º, - = linear polarisation state at –45º, R = right hand circular polarisation state and L =
left hand circular polarisation state. Thus a horizontal linear polarised state of unity intensity
is described by
 s0   1 
   
 s   1
S= 1 =   (20)
s2 0
   
 s   0
 3  
The most general elliptical polarised state and its orthogonal state are given by
 1   1 
   
 cos(2ν)   − cos(2ν) 
 sin(2ν) cos(∆)  and  − sin(2ν) cos(∆)  (21)
   
 sin(2ν) sin(∆)   − sin(2ν) sin(∆) 
   
For examples of common states of polarisation see Table 1.
It is common to use the normalised Stokes vector where the s0 term is dropped, and is taken
to be 1, leaving a three element vector.
The degree of polarisation, DOP, is given by

I Pol s2 + s2 + s2
DOP = = 1 2 3 (22)
I Tot s0
where IPol is the polarised intensity and ITot is the total intensity.
The Mueller Matrix describes the transmission of an optical element and the effect on the
Stokes Vector. The optical element is described by a 4 x 4 matrix, M, such that the output
polarisation state SO is related to the input polarisation state SI by
SO = M S I (23)
13
The matrix, M can be used to describe polarisers and birefringent elements. The rotation
matrix R(θ) can be used to rotate an elements azimuth such that an optical element M at an
azimuth angle θ is given by
M(θ) = R (−θ) M R (θ) (24)
where
1 0 0 0
 
 0 cos(2θ) − sin( 2θ) 0 
R (θ) =  (25)
0 sin(2θ) cos(2θ) 0 
 
0 0 0 1 

For example the effect of a quarter waveplate at 45º on a linear polarised wave is given below
S O = R (−45º ) M R (45º ) S I
 1 0 0 0  1 0 0 0  1 0 0 0  1 
    
 0 0 1 0  0 1 0 0  0 0 − 1 0  1 
= (26)
0 − 1 0 0  0 0 cos( π2 ) − sin( π2 )  0 1 0 0  0 
    
 0 0 0 1  0 0 sin( π ) cos( π )  0 0 0 1  0 
  2 2   
1
 
0
= 
0
 
1
 
ie a right hand circular output polarisation. For a wave with polarisation state SI incident on a
number of optical elements M1 , M2 … MN the output polarisation state SO is given by
S O = M N ...M 2 M1S I (27)
For examples of Mueller Matrices for common optical elements see Table 2.
2.3 Poincaré Sphere

As stated it is common to reduce the Stokes vector to just three elements by normalising to
unit intensity such that s1 = S1/S0, s2 = S2/S0 and s3 = S3/S0. If we consider only fully polarised
light then s12 + s22 + s32 = 1. Thus if we plot the Stokes vector onto a three dimensional graph
we find that all the fully polarised states fall onto the surface of a sphere of radius one. This is
the basis of the Poincaré sphere visualisation of polarised light. The plane containing s1 and s2
cuts the equator of the sphere and all linear states of polarisation fall onto the equator. The
line s1=0, s2=0, s3 pierces the poles of the sphere and circular states of polarisation fall onto
the poles. The rest of the surface of the sphere describes a range of elliptical polarisation
states. Polarisation states which are geometrically opposite on the surface of the Poincaré
sphere are orthogonal states of polarisation.
The effect of a general retardance is to move the state of polarisation in an arc across the
surface of the sphere. The arc will begin at the input state of polarisation, be centred on the
fast axis of the retardance plate (the fast eigenstate of the retardance plate) and will describe
an arc of angular length given by the retardance angle. This allows a quick visualisation of
the effect of birefringent elements.
14
Output State
s3
Retardance
s1
Input State
s2
Waveplate
Axis
Figure 17 Poincaré sphere representation of a horizontal linear input state passing through a
quarter waveplate at azimuth of 45º giving a right hand circular output.
15
2.4 Useful Matrices
State Jones Vector Stokes Vector

H 1 1
Linear Horizontal    
 0 1
 0
 
 0
 
V  0 1
Linear Vertical    
1  − 1
0
 
0
 
+ 1 1 1
Linear at +45º    
2 1  0
1
 
 0
 
- 1 1 1
Linear at -45º    
2  −1 0
 − 1
 
0
 
R 1 1 1
Right Hand    
2 i  0
Circular  0
 
1
 
L 1 1 1
Left Hand    
2 − i 0
Circular 0
 
 −1
 
Linear  cos(θ)   1 
Polarization θ    
 sin(θ)   cos(2θ) 
Azimuth θ  sin(2θ) 
 
 0 
 
General  cos(ν) e i 2 
∆
 1 
Polarisation    
 sin(ν) e −i ∆2 
   cos(2ν) 
 sin(2ν) cos(∆) 
 
 sin(2ν) sin(∆) 
 
Table 1 Some common polarisation states in both Jones vector and Stokes vector format.
16
Element Jones Matrix Mueller Matrix

Ideal Horizontal linear 1 0  1 1 0 0
polariser    
0 0 1  1 1 0 0
2  0 0 0 0
 
 0 0 0 0
 
Ideal Vertical linear 0 0  1 −1 0 0
polariser    
0 1 1 −1 1 0 0
2 0 0 0 0
 
0 0 0 0 

Quarter Waveplate at 45º 1 1 i  1 0 0 0
   
2  i 1 0 0 0 − 1
0 0 1 0
 
0 1 0 0 

Azimuth Rotation, θ  cos(θ) − sin(θ)  1 0 0 0
   
 sin(θ) cos(θ)  0 cos(2θ) − sin(2θ) 0
0 sin(2θ) cos(2θ) 0
 
0 0 0 1 

Linear Retarder, Azimuth  ei 2 1 0 0 0 
0 
δ
0º, retardance δ.   
 0 −i δ2 
e  0 1 0 0 
 0 0 cos(δ) − sin(δ) 
 
0 0 sin(δ) cos(δ) 

Neutral Density Filter K 0 0 0 
Transmission K  
0 K 0 0
0 0 K 0
 
 0 0 0 K
 
Table 2 Some common optical elements in both Jones matrix and Mueller matrix format.
17
3. Polarisation Effects in Optical Fibres

3.1 Birefringence in Optical Fibres
The solution of the electromagnetic wave equation for single-spatial-mode optical fibre
produces two guided polarisation modes. In ideal circular symmetric fibres these two modes
are degenerate. However, any asymmetry in fibre structure or applied external stress leads to
the fibre being weakly birefringent and causes polarisation mode dispersion, PMD. This
PMD leads to pulse broadening and intersymbol interference and ultimately restricts the data
capacity of an optical fibre system.
3.2 Causes of Linear Birefringence

There are a number of intrinsic and external effects that break the symmetry of single-mode
optical fibre and give rise to linear birefringence. For example[4]
• Asymmetrical external stress, e.g. bend, kink or squeezing
• Asymmetrical geometry of the guiding structure, e.g. an elliptical core.
• Asymmetrical internal stress associated with an elliptical core.
• Electric Field
Any external transverse stress across the fibre produces birefringence through the elasto-optic
effect. Figure 18 shows possible arrangements to produce such stress ranging from pressing
the fibre between parallel plates to bending around a mandrel under tension.
f f
2r
r
R
2θ
f f
f
Figure 18 External influences on fibre birefringence.
All such external stresses give rise to a linear birefringence given by the form
ω C
δβl = F(f , r, R , θ) 0 s (28)
c k0
Cs
where F(f,r,R,θ) is a function of the external force, radius etc. k0
is known as the strain-
optical coefficient and is given by:
Cs
= 0.5 n 30 (ρ11 − ρ12 )(1 + ν p ) ~ 0.27 (29)
k0
18
where n0 is the average refractive index of the fibre, ρ11 and ρ12 are the components of the
strain-optic tensor and νp is Poissons ratio.
For a fibre pressed into a V groove as shown in Figure 18 then
f
F(f , r, R , θ) = 2(1 − cos(2θ) sin(θ) ) (30)
πr E
where E is Young’s modulus of the fibre, this takes into account friction between the fibre
and V groove but assumes the forces act as point forces on the fibre. In the case of two flat
plates this simplifies to
f
F(f , r, R , θ) = 4 . (31)
πr E
For a fibre bent around a radius under tension as shown in Figure 18 then
r 2  2 − 3ν P  r  f 
F(f , r, R , θ) = 0.5 2 +    2 . (32)
R  1 − νP  R  π r E 
For a free bend then the second term in Equation 32 is zero. For a standard single-mode fibre
with cladding diameter 125 µm loosely bent around a 5 cm diameter then the birefringence is
approximately 3.4 rad.m-1 such that 3 full loops gives approximately ½π rad birefringence.
This therefore forms is a quarter waveplate.
If a fibre has an elliptical core then there will be differing propagation constants for modes
where the electric field is aligned with the major or minor axis of the ellipse. This leads to a
linear birefringence component with the fast axis along the minor axis of the ellipse. Several
papers give the functional form of this linear birefringence as:
ω a 
δβl = G (V)  − 1∆n 2 . (33)
c b 
where ∆n is the refractive index difference between the core and cladding, a and b are the
major and minor axes of the elliptic core, and G(V) is a function of the normalised frequency
parameter, V. For single mode fibre near cut off where V~2.4 then G(V) ~ 0.2.
Fibres that have an elliptical core will inevitable also have internal stress. The stress results
from the differential expansion and cooling rates between the core and cladding during
manufacture. In circular symmetric fibre this has no net effect but in elliptical core fibre this
will cause a linear birefringence. This stress birefringence is given by:
 U 2  ω CS 1 a−b
δβl = 1 − 2  ∆α ∆T  (34)
 V  c k 0 (1 − ν p ) a +b
where ∆α is the expansion coefficient difference between the core and cladding, ∆T is the
temperature difference between room temperature and the glass softening temperature, a and
b are the major and minor axes of the elliptical core. For a core doped with germanium the
expansion coefficient will be greater then the cladding leading to the fast axis aligned with
( 2
)
the minor axis of the elliptical core. The factor 1 − U 2 is related to the overlap of the optical
V
mode with the stress field and is approximately 0.5 near the cut off wavelength.
A transverse electric field will introduce a linear birefringence through the Kerr effect given
by:
δβl = 2π BK E 2 (35)
where BK is the Kerr Electro-optic constant and E the electric field strength. In silica BK[5][6]
is approximately 2 to 5*10-16 m.V-2 such that a DC electric field of 106 V.m-1 will give a
19
birefringence of just 3 mrad.m-1. For AC electric fields electrostriction effects must be taken
into account particularly when mechanical resonance may enhance any effect.
3.3 Causes of Circular Birefringence

There are a number of external effects that break the symmetry of single-mode optical fibre
and give rise to circular birefringence. For example
• Twist
• Axial Magnetic Field
For a fibre that is twisted the resulting stress gives rise to circular birefringence. That is the
left and right circular modes experience differential phase shift as they propagate through the
twisted fibre. The circular birefringence is given by
δβC = g t (36)
-1
where t is the twist rate (rad.m ) and
g = −0.5 n 02 (ρ11 − ρ12 ) . (37)
g is reported to be ~ 0.14.
An axial magnetic field will introduce a circular birefringence via the Faraday effect given by
δβC = 2 VV H (38)
where VV is the Verdet constant of the fibre material and H is the axial magnetic field. For
silica fibre the effect is small since VV is approximately 4.7*10-6 rad.A-1. Thus for a large
axial magnetic field of say 1 Tesla = 8*105 A.m-1, there is approximately 7.5 rad.m-1 of
circular birefringence. The Faraday effect is very important in the development of non-
reciprocal components such as optical isolators and for this special doped crystals have been
developed with an enhanced VV.
3.4 Local Effects of Fibre Birefringence

As light propagates down a length of fibre the polarisation state will evolve due to the local
birefringence. The local birefringence can be expressed as a vector in the Poincaré sphere
such that the magnitude of the vector gives the total birefringence and the vector intercepts
the Poincaré sphere at the fast polarisation. As with bulk birefringence as the light propagates
through the birefringent media the polarisation state rotates around the birefringence vector,
the rate of rotation being one full revolution for a total birefringence to π radians.
The birefringence vector, W, is given by

 δβ l cos(2θ) 
 
W =  δβ l sin(2θ)  (39)
 δβ C 
 
where δβl is the linear birefringence component at azimuth θ and δβC is the circular
birefringence component. The change of the polarisation state of light propagating through
this birefringence is given by
∂s
= W×s (40)
∂z
20
The solution to this differential equation is a circle on the surface of the Poincaré sphere that
is centred on the point where the vector b pierces the surface of the sphere. The polarisation
state makes one revolution of the circle for a distance, Lb, known as the beat length and given
by
2π
Lb = (41)
δβ
where δβ is the magnitude of the birefringence vector W.
3.5 Global Effects in Fibre, Polarisation Mode Dispersion

If the birefringence of each fibre section is identical then the global birefringence will equal
the local birefringence and the fibre is said to show no-mode coupling. If the local
birefringence varies along the length of the optical fibre, due to internal imperfections and
external perturbations, then these perturbations cause coupling between the two polarisation
modes. This coupling is described by a coupling parameter, h, that is equal to the fractional
power transferred per meter. 1/h is often referred to as a coupling length, LC.
For a medium where there is no polarisation dependent loss the Mueller matrix of a small
section of fibre is a simple rotation matrix, R(W,δz). The rotation matrix is a function of the
birefringence vector and the section length, the birefringence vector depends on its position
within the fibre, z, and the optical frequency, ω. The Rotation matrix for the whole fibre span
is the multiplication of the rotation matrix for each element.
R (ω) = R ( WZ (ω), δz) R ( WZ−1 (ω), δz)...R ( W1 (ω), δz) R ( W0 (ω), δz)
(42)
∏Z R (Wz (ω), δz)
0
=
Such that the output polarisation state, SO, is given by
SO = R (ω) SI . (43)
As there is no polarisation dependent loss and this is a simple rotation of the polarisation state
on the Poincaré sphere any two orthogonal input polarisation states will emerge as two
orthogonal output polarisation states. There exist two orthogonal input states of polarisation
that emerge unchanged. These are the polarisation eigenstates.
The transmission of a polychromatic source through a high polarisation mode coupled fibre
leads to a range of output states and, in general, a depolarising of the light. It has been shown
that over a limited source bandwidth there exist two orthogonal input polarisation states which
lead to two orthogonal output states which are independent of the source frequency (to first
order). These are known as the principle states of polarisation (PSP)[7], they are not the same as
the polarisation eigenstates. To first order there is a phase difference between the states that is
proportional to the source frequency. This is equivalent to a time delay between signals
propagating in the two PSPs. This time delay is the differential group delay (DGD).
For monochromatic light as the optical frequency is changed the output state of polarisation
will describe a circle on the surface of the Poincaré sphere centre on the principle state of
polarisation and making one full revolution for an optical frequency change of 1/(DGD).
21
We define a polarisation dispersion vector, Ω, aligned with the principle state of polarisation,
p, and with magnitude equal to the differential group delay, δτ, such that
 Ω1   p1 
   
Ω =  Ω 2  = δτ p 2  (44)
Ω  p 
 3  3
The change of the output state of polarisation with optical frequency is given by
∂s ∂R (ω)
= Ω×s = R (ω) −1 s (45)
∂ω ∂ω
Similar to the birefringence equation, the solution to this differential equation is a circle on
the surface of the Poincaré sphere that is centred on the point where the vector Ω pierces the
surface of the sphere. The polarisation state makes one revolution of the circle for a change in
optical frequency, δω, given by
2 π 2π
δω = = . (46)
Ω δτ
This can be written in terms of optical wavelength such that the DGD, δτ, is given by
θ λ1λ 2
δτ = c (47)
2π (λ 2 − λ1 )2
Where the wavelength evolves from λ1 to λ2 and the polarisation state rotates through an
angle, θ in radians around the principle state of polarisation on the Poincaré sphere.
sλ1
sλ2
θ
p
Figure 19 Loci of polarisation state on surface of sphere as

wavelength is changed from λ1 to λ2.
To first order where the polarisation vector, Ω, is not a function of the optical frequency the
rotation matrix can be written as[8]
22
1 0 0   p1p1 p 2 p1 p3p1 
R (δω) = cos(δτ δω) 0 1 0 + [1 − cos(δτ δω)] p1p 2
  p2p2 p3p 2 
0 0 1  p1p3 p 2 p3 p3p3 
(48)
 0 − p3 p 2 

+ sin(δτ δω)  p3 0 − p1 
− p 2 p1 0 
where δω, is the optical frequency difference from the central optical frequency to which the
polarisation vector corresponds.
For fibres with no polarisation mode coupling the DGD is a constant with optical frequency
and the principle state of polarisation is a constant. If a pulse is propagated through such a
fibre the polarisation components aligned and orthogonal to the principle axis propagate at
different group velocities leading to two pulses at the exit of the fibre.
For highly mode coupled fibres the DGD and PSP are strong functions of optical frequency.
An optical impulse input into such a fibre will be broadened and in the absence of chromatic
dispersion the output pulse will be a Gaussian pulse. This is the phenomenon known as
polarisation mode dispersion. For an ideal optical impulse (a theoretical pulse with infinite
spectral width) the output pulse will have an RMS width equal to the PMD. The DGD and
PSP are also strong functions of the environmental conditions. However, while the DGD and
PSP will change at a specific optical frequency the PMD is constant. The PMD can be found
from the RMS average of the DGD such that PMD, <∆τ2>½,
N
∑ δτi2
< ∆τ 2 >½ = i =1
(49)
N
where the DGD, δτ, is averaged over optical frequency or environmental conditions.
The DGD is the magnitude of the polarisation dispersion vector, Ω. Perturbation theory has
shown that the elements of the dispersion vector are taken from a normal distribution with
∆τ 2
zero mean and standard deviation α, where α2 = 3 . The statistical distribution of the
magnitude of such a three-element vector is described by the Maxwellian distribution[9],
δτ2
2 δτ 2 − 2α 2
P(δτ) = e . (50)
π α3
This describes the probability of a single measurement or a single sample of DGD having the
value δτ. The long tail indicates that a DGD greater than 3 times the PMD has a probability
of 6*10-5 of occurring (approximately ½ hour per year). It is normal to consider that the DGD
may take on values as high as 5 times the PMD.
23
Figure 20 Distribution of measured DGD values for a highly polarisation mode coupled
fibre.
The alternative definition of PMD based on the linear mean is often used, <∆τ> where
N
∑ δτi
i =1
< ∆τ >= . (51)
N
The difference between these definitions can be calculated from the probability distribution
and is given by
3π
< ∆τ 2 >½ = < ∆τ >= 1.085 < ∆τ > . (52)
8
This factor is only valid in the limit of high polarisation mode coupling.
Figure 21 shows the evolution of pulse arrival times as the amount of polarisation mode
coupling is increased, while maintaining a constant modal dispersion and system length[10]. It
can be seen that the pulse broadening is reduced if there is a large amount of polarisation
mode coupling.
24
1.2
0.8
Relative Probability
0.6
0.4
0.2
0
-1.2 -0.7 -0.2 0.3 0.8
Arrival Time
Figure 21 Effect of polarisation mode coupling on pulse arrival times. Red hz=1, Green
hz=3, Blue hz=10 and Pink hz=100. The curves have been normalised to a
maximum of 1.
The length dependence of the PMD is given by[11]:

δβ'2
[ ]
< ∆τ 2 > = 2 e − 2 h z + 2 h z − 1
2h
(53)
where h is the coupling parameter and δβ' is the modal dispersion given by
∂δβ ∂δβ
δβ' = =c (54)
∂ω ∂k
The length dependence given in equation (53) can be simplified to:
< ∆τ2 >½ = δβ' z = < ∆τ > (55)
in the short length regime where z<<1/h and
z
< ∆τ2 >½ = δβ' = 1.085 < ∆τ > (56)
h
in the long length regime z>>1/h.
25
The description of polarisation mode dispersion based on principle states of polarisation and
differential group delay is suitable to first order only. Higher order polarisation mode
dispersion, in particular second order PMD, is described in terms of the optical frequency
dependence of the polarisation vector, Ω(ω),
 ∂Ω (ω ) ∂ 2Ω1 (ω0 ) 
 Ω1 (ω0 ) + (ω − ω0 ) 1 0 + 12 (ω − ω0 ) 2 + ... 
 ∂ω ∂ω 2

 ∂Ω 2 (ω0 ) 1 2
2 ∂ Ω 2 (ω0 ) 
Ω(ω) =  Ω 2 (ω0 ) + (ω − ω0 ) + 2 (ω − ω0 ) + ... 
 ∂ω ∂ω2  (57)
2
∂Ω ( ω )
 Ω (ω ) + ( ω − ω ) 3 0 + 1 ( ω − ω ) 2 ∂ Ω ( ω )
 3 0 0 0
3 0
+ ... 
 ∂ω 2
∂ω 2

= Ω(ω0 ) + (ω − ω0 ) Ω ω (ω0 ) + ...
where Ωω is known as the second order PMD vector.
This second order PMD vector can be broken into polarisation dependent chromatic
dispersion and a change in the principle state of polarisation. These are the components of the
second order PMD vector that are aligned and orthogonal to the first order PMD vector.
∂Ω ∂∆τ ∂p
Ωω = = p + ∆τ (58)
∂ω ∂ω ∂ω
where p is the principle state. The length dependence of the second order PMD magnitude is
given by[12]
δβ' 4 17 5 1 
< Ω ω 2 >= 4  8
− h z + h 2 z 2 − e − 4 h z − 2e − 2 h z − 2 h z e − 2 h z  . (59)
3h  2 8 
For high polarisation mode coupled fibres the average second order PMD magnitude is given
by[13]
∂∆τ 2 2 ∂p 2
< Ω ω 2 >= 13 < ∆τ 2 > 2 =< p > + < ∆τ 2 >= 1
27
< ∆τ 2 > 2 + 27
8
< ∆τ 2 > 2 , (60)
∂ω ∂ω
from which it can be seen that the first term, the polarisation induced chromatic dispersion, is
considerably smaller then the change in the direction of the PSP with wavelength.
The probability distribution of the second order PMD components, Ωω,i, are taken from a
“Soliton” distribution given by[12]
1
P(Ω ω,i ) = (61)
 π Ω ω,i 
2α 2 cosh 2


 2α 
where, as before, α = ⅓ δβ′2 h z. From this the probability distribution of the second order
PMD magnitude, |Ωω|, is given by[12]
 π Ωω 
π Ω ω sinh  
 2α 2 
P( Ω ω ) =   . (62)
 π Ωω 
2α 4 cosh 2  
 2α 2 
 
Figures 22 and 23 illustrate the probability distributions of the second order PMD
components and the second order PMD magnitude for highly mode coupled fibres.
26
0.6 0.6
0.5 0.5
0.4 0.4
Probability
Probability
0.3 0.3
0.2 0.2
0.1 0.1
0 0
-5 -4 -3 -2 -1 0 1 2 3 4 5 0 1 2 3 4 5 6 7
Second Order PMD Component Second Order PMD Magnitude
Figure 22 Probability distribution of Figure 23 Probability distribution of

second order PMD components for highly second order PMD magnitude for highly
polarisation mode coupled fibre. polarisation mode coupled fibre.
A full analysis of the statistics of second order PMD in the case of high polarisation mode
coupling leads to the following average results shown in table 3[14][12].
Quantity Value
δβ' 2 z
α2
3h
<Ωi> 0
<Ωi2> α2
<Ωi Ωj> i ≠ j 0
<∆τ2> = <Ω2> 3α2
<∆τ>2 = <|Ω|>2 8
πα
2
σ|Ω|2 (3 − 8 π )α 2
σ 2
½α2
Ω2
<Ωωi> 0
<Ωωi2> α4
<Ωωi Ωωj> i ≠ j 0
<Ωω2> 3α4
<(Ω.Ωω/|Ω|)2> 1
3α
4
<(Ω×Ωω/|Ω|)2> 8
3α
4
Table 3 Useful relationships for high polarisation mode coupled systems.
Since the detail of the polarisation mode coupling depends on a large number of small local
mode couplings each of which are determined by internal and external influences on the fibre
the differential group delay changes considerably as the external influences on the fibre
change. Thus if the DGD of a fibre is measured at a given wavelength then, as the external
influences change, the measured DGDs will be samples from the Maxwellian distribution.
However, given sufficient samples the RMS mean will be a constant and equal to the
polarisation mode dispersion.
27
3.6 Polarisation Maintaining Fibre

These fibres are better described by their alternative name of highly birefringent, HiBi, fibre.
The idea is to produce a fibre that will maintain the state of polarisation on transmission and
hold the polarisation under moderate external influence. This is done by giving the fibre a
high intrinsic birefringence such that any external influence does not produce any significant
change. The greater the intrinsic birefringence the greater the fibres polarisation holding
ability under external perturbations. It is typical for such fibres to have beat lengths of order
mm. The polarisation is only “maintained” if the input state of polarisation is aligned with the
birefringence axis. If the input state of polarisation is not aligned with the birefringence axis
then the fibre acts as a long waveplate. The input light is split into the two axes and
propagates at differing velocities, at the output of the fibre the two beams recombine with a
different phase giving the final state of polarisation. Since the fibre is highly birefringent the
polarisation state can undergo a significant change. This change will also depend on external
perturbations on the fibre.
There are several types of HiBi fibre using internal stress and elliptical cores to give the
intrinsic birefringence, see Figure 24.
Oval Inner Clad Panda
Bow Tie Oval Core

Figure 24 Various designs of HiBi fibre.
HiBi fibre is used in relatively short pieces, due to its increased cost, in specialist applications
for example joining a laser source to an electro-optic modulator. In this situation the laser
source is linearly polarised and the electro-optic modulator is polarisation sensitive. If
ordinary fibre were used then as the external perturbations changed the polarisation state of
the light entering the electro-optic modulator would shift leading to a variation in the
modulation depth. The important parameter here is how much light enters the electro-optic
28
modulator in the correct polarisation state. This is described by the polarisation extinction
ratio defined as
P 
ER = 10 log S  (63)
 PF 
where PF is the power propagating in the fast axis of the HiBi fibre and PS that propagating in
the slow axis. The extinction ratio parameter can be applied to a HiBi coupled laser source or
elector-optic modulator or a HiBi patch cord. In systems with connectors then PS and PF
should be replaced with P|| and P⊥ where P|| is the power linear polarised with azimuth aligned
to the connector keyway and P⊥. The extinction ratio is determined by the mechanical
alignment of the sources and components with the axis of the HiBi fibre and also by any
external perturbation (notably during connectorisation).
3.7 Polarisation Dependent Loss

The polarisation effects of the previous sections have been limited to phase and time based
phenomena. In optical components an equally significant phenomena is polarisation
dependent loss. This describes the differences in loss experienced by orthogonal polarisation
states. Polarisation dependent loss (PDL) is defined as the difference between the maximum
and minimum attenuation of the device or component under test. PDL is described by
P 
PDL dB = A max − A min = 10 LOG  max  (64)
 Pmin 
where Amax and Amin are the maximum and minimum attenuation of the device or component
under test and Pmax and Pmin are the maximum and minimum transmitted power as the
polarisation state of the input light is changed over all possible states.
A more detailed description of polarisation loss is in terms of the polarisation loss vector, Γ.
The polarisation loss vector describes both the magnitude and polarisation state for which the
loss is minimised. The linear transmission factor for an input state, sin, is given by
T = Tdepol (1 + sin ⋅ Γ ) (65)
where Tdepol is the transmission for depolarised light and is equal to the average transmission
over all states or the average of the maximum and minimum transmission. From this it can be
seen that
T  1+ Γ 
PDL dB = 10 log max  = 10 log  (66)
 Tmin  1− Γ 
where Tmax and Tmin are the maximum and minimum linear transmission.
In the absence of birefringence an input polarisation state, sin, is transformed to an output

polarisation state, sout, since the components of the input polarisation state aligned to the
polarisation loss vector and those counter aligned will experience different transmission. The
output polarisation state is given by sout, where[15]
s out =
1 − Γ2
sin +
( )
1 + sin ⋅ Γ 1 − 1 − Γ 2 / Γ 2
Γ . (67)
1 + sin ⋅ Γ 1 + sin ⋅ Γ
Thus for a depolarised input polarisation state in the absence of local birefringence the output
polarisation state will be equal to the polarisation loss vector, Γ.
29
3.8 Polarisation Dependent Response

Polarisation dependent response is the equivalent parameter to polarisation dependent loss
applied to optical to electrical converters, for example photodiodes, power meters and optical
spectrum analysers. The polarisation dependent response in (dB) is defined by
R  P 
PDR dB = 10 LOG  max  = 10 LOG  max  (68)
 R min   Pmin 
where Rmax and Rmin are the maximum and minimum responsivity and Pmax and Pmin are the
maximum and minimum displayed power or voltage.
3.9 PMD and PDL Combined

The previous paragraphs have considered PMD in the absence of PDL and PDL in the
absence of PMD. In real systems it is likely that both PMD and PDL will be present. The full
effect of combining PMD with PDL requires a detailed analysis, see for example [16],[17]
and [18]. The physical effect of PDL on PMD is to move the principle states of polarisation
such that the two principle states are no longer orthogonal. We maintain our definitions of
principle state of polarisation and differential group delay as the two output states of
polarisation that are independent of optical frequency to first order and the differential delay
between modes propagating in these two principle states. However, the principle states of
polarisation no longer hold the fastest and slowest mode. There is also the possibility of pulse
broadening in excess of the combined DGD of a fibre system in the presence of PDL. This is
because the non-orthogonal PSPs lead to interference between the polarisation components
propagating in the two PSPs. This can lead to cancellation of the centre of the pulse where the
output from the two PSPs overlap effectively widening the pulse. An alternative view is to
consider that the PDL combined with the PMD leads to a frequency dependent loss that can
suppress the wings of the pulse spectrum. This will effectively broaden the pulse in the time
domain. It is also possible for this filtering process to suppress the centre of the optical
spectrum leading to an effective increase in the spectral width and a pulse compression in the
time domain[19].
30
4. The Effects of Polarisation on Data Transmission

4.1 Impact of first order PMD
The effect of polarisation mode dispersion on a transmission system is to cause eye closure
which reduces the signal to noise ratio and increases the likelihood of transmission errors.
The effect of first order DGD alone is to split the signal into two, delaying one part with
respect to the other. The effect is minimised when the input polarisation state is aligned with
a principle state of polarisation and maximised when the two principle states of polarisation
are equally excited.
A simulation was calculated based on; a pseudo random bit pattern, a first order DGD
transfer function, and ideal intensity detection followed by a low pass 5th order Bessel filter
with a 3 dB frequency set at 0.75 of the bit rate. The eye patterns are shown in Figure 25.
DGD = 0.20 UI DGD = 0.40 UI
DGD = 0.60 UI DGD = 0.80 UI

Figure 25 Eye diagrams for simulated bit pattern and various 1st order DGD. UI = unit
interval = 1 bit period.
The parameter of interest to system designers is the power penalty of any signal degrading
effect. This turns any signal degradation into an effective requirement to increase the signal
to noise. From the eye diagrams the signal quality, Q, can be calculated as
µ − µ0
Q= 1 (69)
σ1 + σ 0
where µ1 is the mean one level, µ0 the mean zero level, σ1 is the standard deviation of the one
level and σ0 the standard deviation of the zero level all taken at the decision point. The Q can
be related to the expected bit error rate based on an assumed Gaussian statistical probability
of the noise such that the BER is approximated by
 Q 
BER ≈ 12 erfc . (70)
 2
For a BER of 10-9 a Q of 6 is required. The question is what is the power penalty due to the
first order DGD degradation. Considering noise free simulations we calculate the basic s/n
that is required to maintain the Q at 6. Assuming the noise adds in quadrature this gives
31
1
Q Re sulting = (71)
1 1
2
+ 2
Q s
n
Thus the signal to noise that can be accepted for a BER of 10-9 is given by
s2 36 Q 2
n
= 2 (72)
Q − 36
and the power penalty is
 Q b−b 2   Q2 
20 log( n b−b ) − 20 log( n ) = 10 log
s s   − 10 log 2  (73)
 Q 2 − 36  Q − 36 
 b −b   
where Qb-b is the back to back Q without the degrading effect under consideration.
Using this method to analyse the simulated data of Figure 25 the power penalties shown in
Figure 26 were obtained. These are worst case penalties since the PSP were equally excited.
7 Simulation
6 F Bruyere
Power Penalty (dB)
L Nelson
5
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
DGD (UI)
Figure 26 Power penalties for first order DGD, measured, simulated and published
F Bruyère[20], L Nelson[21].
While the power penalties depend on the exact system arrangement; modulation format,
bandwidth and filtering there appears to be a significant degradation for first order DGD
greater than 0.5 UI.
4.2 Impact of Second Order PMD

For any optical system containing real fibre there will be significant polarisation mode
coupling that will lead to second-order PMD. The second-order PMD consists of two parts,
the change of DGD with optical frequency and the change in direction of the PSP with optical
frequency as described in equation (58). A full analysis of the effect of second order PMD on
pulse propagation has shown that there are two dominating pulse broadening components[12],
32
an effect equivalent to chromatic dispersion that involves the magnitude of the second order
PMD and an effect which depends on the product of second order PMD and chromatic
dispersion. This second component has a sign that depends on the alignment of the input
polarisation state and the PSPs.
The degrading effects of second order PMD particularly in the presence of chromatic
dispersion are more significant than first order PMD for the highly polarisation mode coupled
regime.
4.3 PMD of Installed Links

The PMD of installed systems can vary considerably depending on the quality of the fibre,
cable design etc. Reported values vary from 0.1 ps.km-½ to 10 ps.km-½. There is also the
question of how fast the PMD vector evolves, as this will determine the speed of any tracking
that may be required for active compensation. A number of studies show differing results
depending on the cable route. For buried cables the evolution is slow with periodic variations
over timescale of 24 hours, while suspended cables show fast variations due to wind.
4.4 Optical PMD Compensation

Unlike chromatic dispersion, PMD cannot be simply compensated with a passive device
since any changes in environmental conditions cause the PMD vector orientation and
magnitude to change. There are a number of optical PMD compensation schemes employing
increasing complexity[22]. Such a compensation system is independent of the optical data
format.
DGD Error Detection
Optical PMD Polarisation DGD Optical PMD

Degraded Input Control Element Compensated Output
Figure 27 Block diagram of an Optical PMD compensation system
There are three basic parts to an optical adaptive PMD compensator, see Figure 27;
1) A PMD detector giving a PMD error signal,
2) A DGD or PMD compensation element,
3) An adaptive polarisation controller to align the compensation element with the
system.
Associated with each scheme is the number of degrees of freedom (DOF) that the scheme can
compensate. The higher the number of degrees of freedom the greater the complexity of the
PMD element and the greater the orders of PMD that are compensated. Three degrees of
freedom will fully compensate first order PMD while six degrees of freedom are required to
compensate first and second order PMD.
33
4.5 Compensation Advantage

Simulations of fibre were performed using the random rotation of the birefringence model[23].
Each element of the fibre is assumed to be a linear birefringent element where the angle, θ, of
the birefringence axis follows a random process
∂θ
= ∂z σθ ν (74)
∂z
where ν is an independent random variable taken from a Gaussian distribution with zero
mean and unit standard deviation. The coupling length, Lc, is determined by, σθ, as
Lc ~ 4/σθ2. The simulation took 200 elements each 50 m long with modal dispersion of 1
fs.m-1, σθ = 0.2 giving a coupling length of 100 m. This gave an uncompensated PMD of
1043 fs with a second order PMD magnitude of 6.301*105 fs2. The ratio of the RMS PSP
rotation to the RMS DGD change was √8.5 compared with the statistical accepted value of
√8.
Using this simulated fibre and providing compensation at the centre optical frequency, the
PMD either side of the compensation point was calculated. The compensation was based on
either a variable DGD element (3DOF) or compensation with 6DOF. Figure 18 and 19 shows
the PMD and 2nd order PMD magnitude as a function of frequency offset. For the PMD it was
found that, for compensation with 3DOF and with 6DOF, the compensated PMD within 386
GHz was lower than the original uncompensated PMD but further away there is degradation
as the PMD compensation acts to increase the total system PMD. The compensated PMD
with 6DOF was lower than that with 3DOF within this bandwidth. A similar simulation for
an uncompensated system PMD of 0.5 ps gave a crossing at 687 GHz suggesting a bandwidth
for improved PMD of approx ~0.35/PMD. Thus the compensated PMD bandwidth appears to
be related to the uncompensated PMD. The 2nd order PMD magnitude shows immediate
degradation for 3 DOF compensation but shows improvement for 6 DOF compensation
within a bandwidth of ~0.15/PMD. Higher order PMD effects may show narrower
bandwidths.
2500
2000
1500
PMD (fs)
1000
Uncompensated
Compensated 3DOF
500 Compensated 6DOF
0
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Optical Frequency Offset (GHz)
Figure 28 Effect of PMD compensation on first order PMD.
34
2.00E+06
1.80E+06
1.60E+06
2nd Order PMD Magnitude (fs^2)
1.40E+06
1.20E+06
1.00E+06
8.00E+05
6.00E+05
Uncompensated
4.00E+05 Compensated 3DOF
Compensated 6DOF
2.00E+05
0.00E+00
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Optical Frequency Offset (GHz)
Figure 29 Effect of PMD compensation on 2nd order PMD magnitude.
Following the work by Frank Bruyère[20] the worst case penalty due to first and second order
PMD is given by
2
Ω Ωω
P(dB) = 10 2
+ 408 C 0 (75)
Tbit Tbit 2
where C0 is the normalised chromatic dispersion given by
 β" 
C0 = 12  2  (76)
 Tbit 
where β″ is the second differential of the propagation constant with optical frequency, ω.
For an arbitrary outage probability of 1 second / day (10-5) we have the maximum DGD equal
to 2.9 times the PMD and the maximum 2nd order PMD magnitude equal to 5.3 time the RMS
2nd order PMD magnitude. Taking the worst case that both these maxima occur together we
can find the PMD that will give a 1 dB penalty for the various compensation schemes, for a
normalised chromatic dispersion of 0.08, see Table 4.
PMD Compensation Maximum Uncompensated PMD (UI)

None 0.09
Fixed DGD, 2DOF 0.10
Variable DGD, 3DOF 0.13
Two fixed DGD, 4 DOF 0.16
Table 4
While the exact PMD values rely on worst case statistics and the work of F Bruyère[20] the
trend suggests that in order for PMD compensation to provide any real improvement high
order compensation will be required and that even then the original system PMD allowable is
only doubled.
35
5. Polarisation Measurement
5.1 Polarisation State
In order to measure polarisation properties it is necessary to know which polarisation state is
incident upon the device under test. It is therefore necessary to either set or measure the
polarisation state of the optical input. An instrument that measures the polarisation state of an
incident optical input is known as a polarisation state analyser or polarimeter.
If we use the Stokes vector method of specifying polarisation state then this can be achieved
by measuring the intensity of the light passed through a number of polarisation analysers. A
minimum of four analysers is required but this can be extended to six analysers. The
analysers must form a Mueller set for example, linear vertical (V), linear horizontal (H),
linear at +45º (+) and right hand circular (R). The intensity of the light transmitted by each
analyser in turn is measured making sure that the polarisation state of the input light remains
constant. Alternatively the input light can be split and the intensity of the light transmitted by
each analyser measured simultaneously in which case the splitting device must not alter the
polarisation state of each part of the light. The insertion loss of each analyser must be equal
or known and a correction applied. Thence the normalised Stokes vector is given by
 IV −Im   IV −IH 
 s1   I m   I V + I H 
 
s =  s 2  =  +I m  =  I + − I − 
I −I I −I
(77)
 m   + −
 s   IR −Im   IR −IL 
 3
 Im   IR −IL 
where Im is the mean Intensity ½(IV+IH), and Ij is the intensity transmitted through analyser j.
For fully polarised inputs it is possible to avoid using the circular polarisation analyser and
2
calculate s3 assuming s = s12 + s22 + s23 = 1 .
An alternative technique is to measure the transmission through a rotating quarter waveplate

and fixed linear polarisation analyser, see Figure 30.
θ
Input Fibre Detector
Measurement
Electronics
and Display
Quarter Linear
Waveplate Analyser
Figure 30 Layout of a rotating quarter waveplate polarimeter.
Taking the Mueller matrixes from section 2.4 and setting the azimuth of the linear
polarisation analyser to zero, the detected intensity is given by
36
 1 1 0 0  1 0 0 0  1 0 0 0  1 0 0 0  s 0  
 1  1     
1 0 0 0 cos( 2θ) − sin( 2θ) 0 0 1 0 0 0 cos( 2θ) sin( 2θ) 0  s1   
I=      
0  s 2 
 2  0 0 0 0  0 sin( 2θ) cos( 2θ) 0  0 0 0 − 1 0 − sin( 2θ) cos( 2θ)

  0 0 0 0  0 0 0 1  0 0 1 0  0 0 0 1  s3  
 0
= s + 1 s + 1 cos( 4θ) s + 1 sin( 4θ) s + 1 sin( 2θ) s
0 4 1 4 1 4 2 2 3
(78)
By rotating the quarter waveplate at a frequency f (revolutions per second) and detecting the
electrical signal at frequency 2f, 4f and DC it is possible to calculate s0, s1, s2 and s3. A small
correction is required if the quarter waveplate is not exactly a quarter wave and this
correction is likely to be wavelength dependent.
It is possible to simplify this if the input polarisation state is known to be fully polarised. The
quarter waveplate is no longer required and the linear polariser is rotated, the detected
intensity is then given by
I = 12 s 0 + 12 cos(2θ) s1 + 12 sin( 2θ) s 2 . (79)
Since it is assumed that the input is fully polarised then s3 can be calculated from
s 0 2 = s12 + s 22 + s 32 , except for the ambiguity of the sign.
In both these polarisation state measurement schemes it is necessary to ensure that there is no
intensity variation due to wobble in the mechanical alignment of the components.
5.2 Birefringence and Beat Length

It is possible to measure the linear retardance of an optical element by rotating that element
and observing the output polarisation state on a polarisation state analyser. For a linear
polarised input state the locus of the output state on the Poincaré sphere will form a figure of
eight with the cross occurring when the linear retardance axis are aligned with the linear
polarised input. The extrema of the loops are related to the retardance. Such a method can
measure a linear retardance up to a half waveplate. Beyond this ambiguities of π in the
retardance occur.
DUT
θ
Polarisation
Polarised Polarisation State
Light Source Controller Analyser
Figure 31 Set up to measure retardance of DUT
Figure 31 shows a system to measure the retardance of a bulk waveplate. Figure 32 shows the
loci of the output state of polarisation for a linear polarised input for three different
retardances. It is possible to calculate the effect for a general input polarisation state but the
technique is most sensitive for near linear polarised inputs.
37
Figure 32 Showing the loci of the output polarisation state for a linear input
polarisation state and three different retardances (1.27 rad, 1.57 rad and 1.87 rad) as
they are rotated. The red vectors illustrate the extremes of the 1.27 rad retardance.
Figure 33 shows the loci of the output state of polarisation for an all fibre quarter waveplate
formed of 4 turns of fibre around a 50mm diameter. The locus is plotted as the azimuth of the
loops is rotated through 180º for three wavelengths of operation. The linear retardance was
found to be near π/4. The loci do not form a closed figure of eight because the connecting
fibre is twisted as the azimuth of the loop is changed and this twist introduces a circular
retardance that effectively reduces the change in azimuth. The pattern is not centred on the
equator since the input state to the fibre loop was not a linear polarised state and the output
state undergoes a random polarisation transformation in the connecting fibre to the
polarisation state analyser.
38
Figure 33 Loci of output polarisation state as loop azimuth is changed from 0º to 180º.
Red Circles = 1540 nm, Green Diamonds = 1500 nm and Blue Squares =1460 nm.
Points are measured, lines are fitted.
The beat length is the period over which the polarisation state of the light propagating in an
optical fibre evolves through one complete cycle. For example with a HiBi fibre if an optical
beam is launched at 45º to the birefringence axis then as the optical beam propagates along
the fibre the polarisation state will evolve through circular to linear at –45º, circular again and
finally back to linear at 45º. Since there is no Rayleigh scatter in the direction of the electric
field if the fibre is viewed perpendicular to the direction of propagation at 45º to the
birefringent axis, the Rayleigh scatter will appear as bands with dark bands where the electric
field of the propagating beam is pointing at the observer[24].
The beat length can also be measured using POTDR, a polarised version of normal optical
time domain reflectometry. The Rayleigh scattered light maintains the polarisation of the
scattering signal such that the backreflected signal evolves at twice the rate of the forward
propagating beam. By passing the backscatter signal through an analysing polariser the
periodic rotation of the polarisation state is converted into an amplitude variation from which
the beat length can be calculated[25]. The OTDR resolution must be greater than the beat
length. In practice for practical HiBi fibre the OTDR technique has insufficient resolution.
39
The optical frequency domain reflectometer technique, with its short distance resolution, has
been shown to give good results[26].
5.3 Differential Group Delay and Polarisation Mode Dispersion

Numerous methods for measuring polarisation mode dispersion have been developed over the
years. These methods fall into two categories, those based on frequency domain
measurements and those based on time domain measurements. Frequency domain methods
include the polarisation state analyser based methods and fixed analyser method. Time
domain methods include; interferometric methods, phase shift methods and pulse time of
flight methods. No one method is ideal for all measurement situations since there is a broad
range of system and components types requiring PMD measurement.
These components can be grouped, for example
• Installed Fibres (large bandwidth, no noise, unstable)

• Laboratory Fibres (large bandwidth, no noise, stable)
• DWDM component (narrow bandwidth, no noise, stable)
• Optical Amplifiers (mid bandwidth, ase noise, stable)
Table 5 below shows these component groups and the PMD measurement methods, marking
with a tick the suitable methods and a cross the least suitable method.
Installed Laboratory DWDM Optical

Fibre Fibre Component Amplifier
Polarimeter Methods × 3 3
Fixed Analyser 3 3 ×
Interferometric 3 3 ×
Phase Shift 3
Table 5 Methods and their suitability.
5.3.1 Polarisation State Analyser based measurement of DGD and PMD
As stated above the polarisation state analyser based methods used to measure DGD and
PMD are frequency domain techniques. The state of polarisation exiting the fibre or
component under test is measured as a function of optical frequency. Measurements are made
for a number of known input states of polarisation as the optical frequency is scanned. The
measurements are analysed to calculate DGD and PMD. A possible experimental
arrangement is shown in Figure 34.
40
Fibre or Component Under Test
Polarisation
State
Analyser
Tunable Polarisation
Laser Rotator
Source
PC
Thermal Enclosure
Figure 34 Possible experimental arrangement for polarisation state analyser based PMD
measurement.
A considerableble number of data analysis methods for polarisation state analyser based
measurements of this type have been developed. These include Jones matrix eigenanalysis,
Poincaré sphere analysis, Poincaré sphere arc method, Poole method and Mueller matrix
method. All of these methods give essentially the same result. However since the calculation
of DGD involves a numerical differentiation with optical frequency some of the methods are
better suited for larger steps of optical frequency. Other methods are able to calculate the full
polarisation dispersion vector, Ω, and not just the DGD magnitude.
For Jones matrix Eigenanalysis[27][28] the input states of polarisation are taken to be the linear
states of polarisation with azimuth 0º, 45º and 90º. These can be easily set using a linear
polariser giving set states that will be independent of optical frequency. The measured output
polarisation states are SH, S+ and SV respectively. The measured output states are converted to
Jones vectors and these are converted to complex ratios using
a
k1 = H x =
1
[SH 2 (1 − SH1 ) + i SH3 (1 − SH1 )]
a H y SH 2 + SH 3 2
2
aVx
k2 = =
1
[SV 2 (1 − SV1 ) + i SV3 (1 − SV1 )]
a V y SV 2 + SV 3 2
2
(80)
a
k3 = + x =
1
[S+ (1 − S+1 ) + i S+3 (1 − S+1 )]
a + y S+ 2 + S + 3 2 2
2
k3 − k 2
k4 =
k1 − k 3
where aHx and aHy are the two complex elements of the Jones matrix for the output
polarisation state corresponding to the linear horizontal input polarisation state. Following the
Jones algorithm the transmission matrix, T, for the system under test is given, to within a
complex constant, by
 k k k2 
T(ω) =  1 4  (81)
 k4 1 
The DGD is then approximated by
41
ρ 
Arg 1 
∆τ =  ρ2  (82)
δω
where ρ1 and ρ2 are the eigenvalues of T(ω+δω)T(ω)-1 and the Arg function is defined by
Arg(α eiθ) = θ.
For Poincaré sphere analysis the output polarisation states SH and S+ are measured for the two
linear input polarisation states at azimuth 0º and 45º. The vector SC is calculated as the cross
product of SH and S+. These three vectors define a rectangular coordinate system that rotates
around the polarisation dispersion vector, Ω, as the optical frequency is changed. By
geometry the magnitude of the polarisation dispersion vector, equal to the DGD is given by
1  ∂SH ∂SC 
2 2 2
∂S+
Ω=  + + . (83)
2  ∂ω ∂ω ∂ω 
For practical measurements only finite differences are known and a more accurate expression
can be obtained given by
Ω=
2
δω
1 1
arcsin [ ] 
∆S2H + ∆S2+ + ∆SC2  . (84)
2 2 
where ∆SH is the magnitude of the change of the vector SH as the optical frequency is
changed from ω to ω+δω.
For Poole[29] analysis the output polarisation states are measured for three input polarisation
states SH, S+ and SL, the two linear input polarisation states at azimuth 0º and 45º and the left
hand circular polarised input. There is no requirement for these exact input states of
polarisation, however the input states must remain fixed as the optical frequency is changed.
The polarisation vector, Ω, is given by
∂Si ∂S
∂ω
× ∂ωj
Ω= ∂Si
. (85)
∂ω
⋅Sj
where Si and Sj are any pair taken from SH, S+ and SL. For output polarisation states that are
close to the principle state of polarisation large uncertainties can arise from noise in the
calculation of the derivatives. Poole suggests the following criteria for selecting the pair of
states used in calculating the polarisation vector
∂Si ∂S j
∂ω
× ∂ω 1
∂Si ∂S j
>
⋅ 2
∂ω ∂ω . (86)
Ω 1
× Si, j >
Ω 2
The DGD can then be calculated from the magnitude of the polarisation vector.
For Mueller matrix analysis[30] the output polarisation states S1 and Sa are measured for two
linear input polarisation states where S1 × Sa ≠ 0. These could be linear polarisation states at
azimuth 0º and 45º. The Mueller rotation matrix is determined from S1, S3 = S1×Sa/|S1×Sa|
and S2 = S3×S1 using
42
 (S1 )1 (S2 )1 (S3 )1 


R =  (S1 )2 (S2 )2 (S3 )2 R 0 . (87)
 (S ) (S2 )3 (S3 )3 
 13
where R0 is an arbitrary rotation connecting the actual input polarisation state to the output
states for output states S1 and Sa. Such that
 (S1 )1 (S2 )1 (S3 )1 
 
S1 =  (S1 )2 (S2 )2 (S3 )2 R 0 S1 _ in
 (S ) (S ) (S ) 
 13 2 3 3 3
(88)
 (S1 )1 (S2 )1 (S3 )1 
 
Sa =  (S1 )2 (S2 )2 (S3 )2 R 0 Sa _ in
 (S ) (S ) (S ) 
 13 2 3 3 3
where S1_in and Sa_in are the input polarisation states which give output polarisation states S1
and Sa. The change in the rotation matrix as the optical frequency is changed from ω to
ω + δω is given by
R δ = R (ω + δω) R (ω) T
1 0 0  p1p1 p 2 p1 p3p1   0 − p3 p2 
. (89)
 
= cos(φ) 0 1 0 + [1 − cos(φ)]p1p 2 p 2 p 2 
p 3p 2  + sin(φ)  p3 0 − p1 
0 0 1  p1p3 p 2 p3 p3p3  − p 2 p1 0 
This can be solved to obtain
cos(φ) = 12 [(R δ )11 + (R δ ) 22 + (R δ )33 − 1] (90)
and
(R δ )32 − (R δ ) 23 
p sin(φ) =  (R δ )13 − (R δ )31 
1
2
(91)
 (R δ ) 21 − (R δ )12 
The polarisation vector Ω = δτ p, where
φ
δτ = (92)
δω
Since all of these measurements involve the change of polarisation state with optical
frequency and measurements will be made at a sequence of optical frequencies the size of the
step will affect the accuracy of the calculated result. The table 6 shows the upper limit of the
product of optical frequency step, δω, and RMS DGD, <δτ2>½ required to give various
uncertainties for each data analysis method. The Jones matrix method, PSA method and
Mueller method show considerably better performance for large optical steps with no-mode
coupled devices as they make assumptions that the differential of the output state of
polarisation will follow an arc on the Poincaré sphere. These methods give good performance
when measuring the PMD, as an average or RMS DGD with an optical frequency step that is
in effect limited by the likely maximum DGD. In the case where the local DGD is required
then all the methods require a much smaller optical frequency step since the higher order
PMD effects mean that the output state no longer follows a simple arc on the Poincaré sphere.
This will be particularly important if the data is being analysed to calculate the PMD vector
with the intention to differentiate this to then calculate the second order PMD.
43
Polarisation Mode Parameter Max value of δω<∆τ2>½ vs Method

Coupling of Jones PSA Mueller Poole
Device Under Test
None Error DGD < 1% 3.14 3.14 3.14 0.5
High Error PMD, <∆τ2>½ < 1% 0.85 0.85 0.85 0.4
High RMS error in DGD < 1% 0.065 0.065 0.065 0.065
High Max error in DGD < 1% 0.007 0.007 0.007 0.007
Figure 6 Optical frequency step requirement.
The method is relatively slow and requires that the fibres polarisation properties do not
change during the measurement process. As such the techniques are best suited to laboratory
conditions where the fibre or component under test can be maintained in stable environmental
conditions. The technique is less suitable for installed cables and unsuitable for aerial cables
where the polarisation properties change in seconds. The technique is able to measure the full
PMD vector, DGD and the principle state of polarisation, PSP, that allows the subsequent
calculation of second and higher order PMD effects.
5.3.2 Fixed Analyser Method to measure PMD
The fixed analyser method used to measure PMD can provide either frequency domain or
time domain measurements depending on the data analysis techniques[31]. A polarised
broadband source is transmitted through the fibre or component under test and the output
signal passed through an analyser to an optical spectrum analyser. The optical intensity as a
function of optical frequency is recorded, this is normalised to the intensity obtained with the
analyser removed or the sum of intensities with the analyser rotated through 90º. A possible
experimental arrangement is shown in Figure 35. As an alternative the source can be tunable
and the optical spectrum analyser replaced by a power meter. The polarisation controller is
only required when the fibre or component under test has low mode coupling and is used to
maximize the intensity variation with optical frequency. However with high polarisation-
mode coupled fibres the measurement uncertainty can be reduced by averaging over the 9
state Mueller set, that is for all combinations of three input states orthogonal on the Poincaré
sphere and three analyser states orthogonal on the Poincaré sphere.
Optical
Broadband L A Spectrum
Source P Analyser
Linear Polarisation Analysing
Polariser Controller Polariser
Figure 35 Possible experimental arrangement for Fixed analyser PMD measurement.
44
Typical measurement scans for a fixed low mode coupled device and a high mode coupled
device are shown in Figures 36 and 37.
0.9
0.8
0.7
Transmission (au)
0.6
0.5
0.4
0.3
0.2
0.1
0
1200 1250 1300 1350 1400 1450 1500 1550 1600 1650
Wavelength (nm)
Figure 36 Fixed analyser spectrum of a low polarisation mode coupled fibre

with 1000 fs of DGD
0.9
0.8
0.7
Transmission (au)
0.6
0.5
0.4
0.3
0.2
0.1
0
1200 1250 1300 1350 1400 1450 1500 1550 1600 1650
Wavelength (nm)
Figure 37 Fixed analyser spectrum of a high polarisation mode coupled fibre

with ~1000 fs of PMD
The data can be analysed through cycle counting or by Fourier transform.
With cycle counting the number of extrema, Ne, (peaks and troughs) or the number of mean
level crossings, Nm, are counted within the measurement window from λ1 to λ2. If extrema or
mean level crossings form the endpoints of the window then the number is reduced by 1.
45
For low polarisation mode coupling the mean DGD is given by[32]
π Ne π Nm N e λ 1λ 2 N m λ 1λ 2
∆τ = = = = , (93)
∆ω ∆ω 2 c (λ 2 − λ 1 ) 2 c (λ 2 − λ1 )
where ∆ω is the width of the optical frequency window. Figure 36 shows Ne = Nm = 130 for a
wavelength window from 1208 nm to 1638 nm giving a DGD = 997 fs.
For high polarisation mode coupling the mean DGD is given by[32]
π Ne π Nm N e λ 1λ 2 N m λ 1λ 2
< ∆τ > = 0.824 = 4 = 0.824 = 4 , (94)
∆ω π
∆ω 2 c (λ 2 − λ1 ) π
2 c (λ 2 − λ 1 )
and thus the RMS DGD is
N e λ 1λ 2 N m λ 1λ 2
< ∆τ 2 > ½ = 0.894 = 1.382 . (95)
(
2 c λ 2 − λ1 ) 2 c (λ 2 − λ 1 )
Thus in the high polarisation mode coupling regime <Ne> = 1.54 <Nm>. Figure 37 shows
Ne = 149, Nm = 94 for a wavelength window from 1208 nm to 1638 nm giving a RMS DGD
of 1021 fs and 996 fs respectively.
An alternative method of analysis for a fixed analyser scan is to use a Fourier transform. For
this the data set must be in equal intervals of optical frequency and not wavelength. Scans
that are measured with equal wavelength steps must first be transformed to equal frequency
steps by interpolation. The DC component of the fixed analyser vs frequency scan is usually
removed by subtracting the mean signal level and the data is often multiplied by a Hamming
window to reduce the noise in the Fourier transform due to the non-repetitive nature of the
scan. The optical power transmission of the fixed analyser scan in the frequency domain,
R(ω) is Fourier transformed and the magnitude of the amplitude in the time domain
calculated as P(δτ). The Fourier transform of the low polarisation mode coupled fibre shown
in Figure 36 is shown in Figure 38.
0.35
0.30
0.25
0.20
P(δτ) (au)
0.15
0.10
0.05
0.00
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Time (fs)
Figure 38 Fourier transform of fixed analyser scan of a low polarisation mode coupled
fibre with 1000 fs DGD.
46
For low polarisation mode coupled fibres the DGD is given by

∞
∫ P(δτ) δτ ∂δτ
∆τ = 0 ∞ , (96)
∫0 P(δτ) ∂δτ
where in practical measurement situations the range over which the integral is taken is
restricted to avoid integrating the noise in P(δτ).
The Fourier transform of the high polarisation mode coupled fibre shown in Figure 37 is
shown in Figure 39. The RMS DGD is given by the second moment of the time domain
distribution given by the Fourier transform. For highly polarisation mode coupled fibres the
distribution is Gaussian and could be fitted to a Gaussian equation.
∞
∫ P(δτ) δτ ∂δτ
2
< ∆τ 2 > ½ = σ R = 0 ∞ . (97)

∫0 P(δτ) ∂δτ
Again in practical measurement situations in order to avoid integrating any residual DC
component and noise the range of integration is restricted. The second moment, σR, is then
calculated by iteration of the following equation,
b  δτ2 
∫a P(δτ) δτ 2 ∂δτ b
−


2
2 1 e  2 σ R  δτ 2
b
= ∫a π σR
∂δτ . (98)
∫a P(δτ) ∂δτ
where a and b are chosen such that a is just greater than the any DC peak and b is just less
than the point where the Fourier transform amplitude is more than 2 times the RMS noise.
0.35
0.30
0.25
0.20
P(δτ) (au)
0.15
0.10
0.05
0.00
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Time (fs)
Figure 39 Fourier transform of fixed analyser scan of a high polarisation mode coupled
fibre with ~1000 fs PMD.
47
5.3.3 Interferometic Method to measure PMD
The interferometric measurement method is a time domain method to measure PMD[33]. A

polarised broadband source is transmitted through the fibre or component under test and the
output signal analysed through a linear polariser and a Michelson interferometer. A basic
configuration is shown in Figure 40. The polarisation controllers allow the input and analysed
states to be adjusted. The input signal is split into the principle states of polarisation and
undergoes PMD. The interferometer provides additional delay and allows interference
between the polarisation modes that fall within the coherence time of the source. The
analyser mixes the output states to generate the interference signal. Interference occurs when
the sum of the delay in the fibre or component under test and the delay in the inteferometer is
equal to zero. For a low mode coupled device this occurs when the delay in the interferometer
is equal to the DGD or when the delay in the intereferometer is zero when the autocorrelation
of the source coherence function is obtained. For low polarisation-mode coupled fibre and
components the polarisation controllers are adjusted to maximize the intensity of the satellite
peaks. With high polarisation-mode coupling the polarisation controllers are used to reduce
the measurement uncertainty by averaging over the 9 state Mueller set, that is for all
combinations of three input states orthogonal on the Poincaré sphere and three analyser states
orthogonal on the Poincaré sphere.
Moving Arm
Fixed Arm
Broadband L
Source P
Linear Polarisation Polarisation
Polariser Controller Controller Polarising
Beam Py
Splitter
Px
Interferometer
Figure 40 Layout of Interferometric measurement of PMD
There are two forms of data analysis; traditional analysis that calculates the standard
deviation of the cross correlation envelope and general analysis that calculates the standard
deviation of the square of the cross correlation envelope less the standard deviation of the
square of the autocorrelation envelope.
In traditional analysis the interferogram Px or Py is analysed, examples of interferograms for

low polarisation-mode coupled fibre and high polarisation-mode coupled fibre are shown in
figures 41 and 42.
48
90 180
80 160
70 140
60 120
Intensity (au)
Intensity (au)
50 100
40 80
30 60
20 40
10 20
0 0
-1500 -1000 -500 0 500 1000 1500 -3000 -2000 -1000 0 1000 2000 3000
Delay (fs) Delay (fs)
Figure 41 Low Polarisation-mode Figure 42 High Polarisation-mode

coupled fibre interferogram coupled fibre interferogram
The envelope of the interferogram is analysed to obtain the PMD, thus the first stage is to
calculate or measure directly the interferogram envelope, as shown in figures 43 and 44.
For low polarisation-mode coupled fibre the polarisation cross correlation peaks are separated
by the DGD from the central autocorrelation peak. The DGD is given by half the delay
between the two satellite peaks.
For high polarisaton-mode coupled fibre the PMD (RMS DGD) is calculated from the second
moment of the polarisation cross correlation envelope, σE, using[34],
3
< ∆τ 2 > ½ = σE . (99)
4
The interferogram is first centred such that the centre of the autocorrelation peak is at zero
delay. The second moment of the cross correlation envelope, σE, is then calculated by
iteration of the following equation,
−a b  δτ2   δτ2 
∫−b PE (δτ) δτ 2 ∂δτ + ∫a PE (δτ) δτ 2 ∂δτ −a
−


2
2 1 e  2 σ E  δτ 2
b
−


2
2 1 e  2 σ E  δτ 2
−a b
= ∫−b π σE
∂δτ + ∫a π σE
∂δτ . (100)
∫−b PE ( δτ) ∂δτ + ∫a PE (δτ) ∂δτ
where PE is the envelope of the inteferogram and a and b are chosen such that a is just greater
than the maximum width of the autocorrelation envelope and b is just less than the point
where the envelope is more than 4 times the RMS noise.
50 90
45 80
40
70
35
60
Intensity Envelope (au)
30
50
25
40
20
30
15
20
10
5 10
0 0
-1500 -1000 -500 0 500 1000 1500 -3000 -2000 -1000 0 1000 2000 3000
Figure 43 Envelope of low polarisation- Figure 44 Envelope of high polarisation-

mode coupled fibre. mode coupled fibre (black) with
calculated Gaussian cross
correlation envelope in red.
49
In general analysis the DC part of each of the x and y interferogram is first removed. Thence
the autocorrelation interferogram, P0, and cross correlation interferograms, P1, are separated
by using
P0 (δτ) = [Px (δτ) − Px (∞)] + [Py (δτ) − Py (∞)] , (101)
and
P1 (δτ) = [Px (δτ) − Px (∞)] − Py (δτ) − Py (∞) . [ ] (102)
The envelope of the autocorrelation and cross correlation interferograms P0E and P1E is
calculated, for the interferograms shown in figures 41 and 42 these are shown in figures 45
and 46. For low polarisation-mode coupled fibre and components the DGD is given by half
the separation of the cross correlation peaks. For high polarisation-mode coupled fibre the
RMS DGD is calculated from the second moments of the square envelopes using
< ∆τ 2 > ½ =
2
(
3 2
σ1 − σ 0 2 ) , (103)
where σ0 and σ1 are given by
∞
∫ δτ P0E (δτ) ∂δτ
2 2
σ0 2
= −∞∞ , (104)
∫−∞ P0E (δτ) ∂δτ
2
and
∞
∫ δτ P1E (δτ) ∂δτ
2 2
σ1 2 = − ∞ ∞ . (105)
∫−∞ P1E (δτ) ∂δτ
2
100 40
90
35
80
30
70
25
60
50 20
40
15
30
10
20
5
10
0 0
-1500 -1000 -500 0 500 1000 1500 -3000 -2000 -1000 0 1000 2000 3000
Figure 45 Cross correlation envelope and Figure 46 Cross correlation envelope and
autocorrelation envelope (blue) gaussian approximation (red) for
for low polarisation-mode high polarisation-mode coupled
coupled fibre. fibre.
This method is quick and is ideal for installed systems where there is no control of the
mechanical stability of the fibres under test.
5.3.4 Phase Shift measurement of DGD and PMD
The phase shift measurement technique provides a time domain measurement of DGD and
PMD. The method measures the phase shift of the modulation on an optical signal as the
optical carrier polarisation state is changed. The phase shift is directly related to the
transmission time so that the DGD can be easily measured. The basic layout is shown in
Figure 47.
50
Tunable Intensity Polarisation Vector

Laser Modulator Controller Detector Voltmeter
φ
RF
Synthesiser
Figure 47 Basic layout of a phase shift system to measure DGD.
The system can be operated in two ways, by scanning the input polarisation state through all
possible states and measuring the maximum and minimum phase and by measuring the
output phase for a number of set input polarisation states. The first technique requires greater
stability of the measurement equipment since the limited number of states in the second
technique can be measured more quickly.
For the random scanning technique the DGD, ∆τ, is given by

φ Max − φ Min
∆τ = , (106)
2πν
where φMax and φMin are the maximum and minimum measured phases in radians and ν is the
modulation frequency. The DGD can be measured as a function of optical frequency by
changing the optical frequency of the tunable laser source; thence the PMD can be calculated
as the RMS average over optical frequency.
For the set state method the output phase is measured for a Mueller set of input polarisation
states. These input polarisation states have three states orthogonal on the Poincaré sphere and
a fourth which is opposite, for example Linear horizontal (H), Linear at +45º (+), Right hand
circular ( R) and linear vertical (V). The respective measured phases are φH, φ+, φR, and φV.
From these the average phase, φ0, for a depolarised signal is calculated as
φ0 = 12 (φH + φV ) . (107)
The DGD is then given by[35]
∆τ =
1
πν
{
arctan tan 2 (φ H − φ 0 ) + tan 2 (φ + − φ 0 ) + tan 2 (φ R − φ 0 ) .} (108)
A slightly better technique is to measure the phase output for six polarisation states that are
orthogonal and opposite on the Poincaré sphere, for example linear horizontal (H), linear
vertical (V), linear at +45º (+), linear at –45º (-), right hand circular (R) and left hand circular
(L). It is preferred to record the phase of orthogonal polarisation states consecutively to
reduce the effect of any system drift. The DGD is then given by
1   φ − φV   φ − φ−   φ − φL  
∆τ = arctan  tan 2  H  + tan 2  +  + tan 2  R  . (109)
πν   2   2   2  
51
The technique is ideally suited to narrow band optical components since the optical
bandwidth required to make a DGD measurement is only that required by the modulation
frequency, ν.
5.3.5 Statistical Uncertainty in measurement of PMD
The uncertainty associated with a measurement of PMD is a combination of the equipment

accuracy and the statistical nature of PMD. For low mode coupled devices this statistical
uncertainty is low but for highly mode coupled fibres and components this statistical
uncertainty will dominate the measurement uncertainty.
The DGD of a highly mode coupled fibre will change with environmental influence such that
any single measurement of DGD will be a sample from the Maxwellian distribution. The
PMD of such a fibre fully describes the distribution and is considered to be independent of
minor changes in environmental conditions. Consider the polarimetric methods to measure
PMD, although the uncertainty applies equally to all measurement methods. The DGD will
be measured as a function of optical frequency and the PMD calculated as the RMS DGD.
Each DGD value is a sample from the Maxwellian distribution and the PMD is the RMS
mean of the distribution. The Maxwellian distribution has an RMS mean, <∆τ2>½, given by,
∆τ 2
∞ ∞ 2 ∆τ 2 − 2α 2 2
∫0 ∫0
2 2
< ∆τ >= P(∆τ) ∆τ ∂ (∆τ) = e ∆τ ∂ ( ∆τ) =3 α 2 . (110)
π α3
and a standard deviation of the square DGD, σ∆τ2, given by
2
(σ ∆τ )2 = ∫0∞ P(∆τ) ∆τ 4 ∂(∆τ) −  ∫0∞ P(∆τ) ∆τ 2 ∂(∆τ) 
2 = 15 α 4 − 9 α 4 = 6 α 4 . (111)
In the measurements there will only be a finite number of DGD measurements and the
uncertainty on the RMS mean, σ<δτ2>½, will be given by
( 2
)
σ< ∆τ 2 >½ =
(
σ< ∆τ 2 > 2
=
)
α 2 < ∆τ 2 >
= , (112)
4 < τ2 > N 2 N 6N
where N is the number of independent measurements of DGD, ∆τ.
N Gisin[36] et al propose the argument that there are two independent measurements points
per cycle in the DGD curve as per the Nyquist sampling theorem, giving the number of
independent measurements as
8 ∆ω 8  2 c (λ 2 − λ1 ) 
N= < ∆τ 2 > ½ = < ∆τ 2 > ½   , (113)
3π π 3π  λ 1λ 2 
where the RMS DGD average <∆τ2>½ is taken over the optical frequency window, ∆ω, or
over the wavelength range from λ1 to λ2. More recent studies of the autocorrelation of the
polarisation dispersion vector[37] suggests that the actual number of independent
measurements, N, is approximated by
∆ω  c (λ 2 − λ1 ) 
N =< ∆τ 2 >½ =< ∆τ 2 >½   , (114)
2.5π  1.25 λ1λ 2 
for 10<N<100 and tends to underestimate N outside this range.
This gives the number of independent measurements within a single measurement scan. In
order to improve the statistical uncertainty it is necessary to make several independent
measurement scans. It is not sufficient to simply repeat a measurement scan as this is unlikely
52
to produce a statistically independent scan. It is necessary to change the mechanical or

environmental conditions of the fibre under test such that the polarisation mode coupling is
altered. It has been suggested that by changing the temperature of the fibre under test by at
least 5 ºC will give an independent measurement however further study is required to
rigorously confirm this. An installed cable measured on separate occasions is likely to give
independent measurement scans. However this should be confirmed by examining the DGD
curves.
5.4 Polarisation Dependent Loss and Polarisation Dependent Response

Polarisation dependent loss and polarisation dependent response require basically identical
measurements, the first being related to the transmission of a passive device and the latter to
the output of an active device. The basic layout is shown in Figure 48. The measurements can
be made in two ways, by random scanning of the input polarisation state or through fixed
known input polarisation states.
Power Meter
/ Active Device
S PC Under Test
Source Polarisation
Controller
DUT
Device Polarisation
Under Test Independent
Reference
Power Meter
Figure 48 Basic technique to measure PDL or PDR.
With the random scanning technique the detected output power through the device under test
or the detected output signal from the device under test is recorded as the input state of
polarisation is scanned through all possible states. The PDL or PDR is then given by
P 
PDL dB = 10 LOG  Max  . (115)
 PMin 
This technique relies on the stability of the input power to the device under test while the
polarisation controller is scanned through all states. Any variation in transmitted power of the
polarisation controller will lead to an error in the measured PDL or PDR. Likewise any PDR
of the polarisation independent power meter will also give an error in the measurement of
PDL. It is possible to assess the combined effect of these uncertainties by connecting the
output of the polarisation controller directly to the polarisation independent power meter. It is
not possible to compensate for this error since depending upon the alignment of the
polarisation axis of the device under test the error may add or subtract from the measured
PDL or PDR. Any instability in the source power and noise in the detected signal will tend to
53
increase the PDL or PDR since these effects will always increase the maximum and reduce
the minimum.
By using a number of known fixed input polarisation states the measurement of PDL and
PDR can be improved by allowing a reference measurement to remove the polarisation
effects of the measurement system. The most common technique is to use a Mueller set of
input polarisation states for example linear horizontal (H), linear at +45º (+), right hand
circular (R) and linear vertical (V)[38]. For PDL the transmitted signal through the device
under test is measured using a polarisation independent detector, for PDR the signal
measured on the active device under test is measured. The reference signal is then measured
for each input polarisation state on the polarisation independent detector. The signal
transmitted or recorded by the device under test is normalised to the reference signal to
remove the variation in transmission of the polarisation controller. The respective normalised
measured signal are PH, P+, PR, and PV, where
Ptrans _ H
PH = . (116)
Pref _ H
From these the normalised measured signals the average transmission, P0, for a depolarised
signal is calculated as
P0 = 12 (PH + PV ) . (117)
The linear PDL magnitude is given by
2 2 2
 P − P0   P+ − P0   PR − P0 
Γ=  H  +  +  . (118)
 P0   P0   P0 
and the PDL or PDR is then given by

1 + Γ 
PDL dB = 10 LOG  . (119)
 1 − Γ 
A slightly better technique is to measure the transmitted signal for six polarisation states that
are orthogonal and opposite on the Poincaré sphere, for example linear horizontal (H), linear
vertical (V), linear at +45º (+), linear at –45º (-), right hand circular (R) and left hand circular
(L). It is preferable to record the signal of orthogonal polarisation states consecutively to
reduce the effect of any system drift. The measured transmitted signal is normalised to the
reference signal as before. The linear PDL magnitude is then given by
2 2 2
 P − PV   P+ − P−   PR − PL 
Γ=  H  +  + P + P  . (120)
 PH + PV   P+ + P−   R L
The PDLdB or PDRdB is then given by equation (119).
This improved technique is better able to withstand instabilities in the source and variations
in the transmission of the polarisation controller. It is possible to improve the measurement
system further by introducing a pick off to monitor the source power and while this pick off
may drift it allows any fast changes in source power to be normalised. The measurement of
orthogonal input polarisation states quickly allows the normalisation of slower drift in the
source power and the use of fixed known input states allows for a reference measurement that
can normalise the changes in transmission of the polarisation controller. The measurement
accuracy relies on the polarisation independent detector having a low or zero PDL since the
device under test is being compared directly with this detector.
54
5.5 Polarisation Extinction Ratio

The polarisation extinction ratio of a source or component coupled with HiBi fibre is a
measure of the quality of the alignment of the birefringence axis of the HiBi fibre and the
strength of any external perturbations on the HiBi fibre. The definition of extinction ratio is in
terms of the ratio of the power in the fast to the power in the slow axis as per equation (63).
In many instances it is not possible to determine the fast and slow axes and so the
measurement is made by looking for the maximum and minimum signal that is transmitted
through a linear polariser.
θ
Input Fibre Detector
Measurement
Electronics
and Display
Linear
Analyser
Figure 49 An arrangement to measure polarisation extinction ratio.
Figure 49 shows an arrangement to measure polarisation extinction ratio. The maximum and
minimum signal is recorded and used to calculate the extinction ratio. This gives the
instantaneous degree of linear polarisation of the beam. The assumption is that a linear
polarised output must have been aligned with the birefringence axis, there is however, the
ambiguity of which, fast or slow, axis. There is also the ambiguity of the length of HiBi fibre
being such that the output phase difference between the light propagating in the fast and slow
axis is zero or 2nπ. If this was the case the output state of polarisation would be linearly
polarised. To overcome this it is necessary to ensure that the phase difference between the
fast and slow axis is changed to explore all possibilities and the measured extinction ratio
taken as the worst case over these phase possibilities. The phase difference between the fast
and slow axis can be change by mechanically stretching the HiBi fibre or by warming or
cooling the HiBi fibre. It may also be possible to explore the phase difference by changing
the operating wavelength of the source or by using a broadband source.
55
6. Calibration of Polarisation Measurement Equipment

6.1 Calibration of Polarisation State Analysers
A polarisation state analyser is essential for the full characterisation of polarisation properties
of optical fibres and devices. In order to make accurate measurements it is necessary to fully
characterise the polarisation state analyser. Polarisation state analysers are relatively complex
measurement instruments with a number of settings and ranges. Thus to fully characterise a
polarisation state analysers it is necessary to check the polarisation states across the whole
Poincaré sphere at wavelengths across the range of operation, for powers across the range of
operation and for input from fibre and or free space beams. In order to reduce this to a
manageable problem the polarisation state is checked across the whole of the Poincaré sphere
at one wavelength, one power and one input condition. The difference between these
reference conditions and other wavelengths, powers and input conditions are then taken into
account.
LP WP
Polarisation α θ Polarisation
Tunable Variable Tunable Controller
State
Laser Source Attenuator Filter Analyser
Mirror #2 Mirror #1
Chopper
Control Chopper
ref MM
Lock in signal
Amplifier Detector
Figure 50 Experimental Layout used to calibrate a Polarisation State Analyser
Figure 50 shows a possible characterisation arrangement for a free space collimated input.
The polarisation state presented to the polarisation state analyser is determined from the
quality of the linear polariser and quarter waveplate and from the azimuth of these two
elements. The extinction ratio of the linear polariser is assessed by measuring the transmitted
power of a similar crossed linear polariser. Referring to the arrangement of figure 50, with
the mirror #1 in position the optical beam is back reflected through the quarter waveplate and
linear polariser before being diverted by mirror #2 to the detector. The azimuth of the quarter
waveplate and the wavelength of the source are adjusted to minimise the back reflected
signal. This occurs when the state impinging on mirror #1 is a circular polarised state. A
small error occurs since the mirror #1 is not at normal incidence to allow the back reflected
beam to be separated at mirror #2. This allows the wavelength of operation to be set such that
the retardance of the quarter waveplate is exactly ½π radians and the relative azimuth
between the linear polariser and the quarter waveplate to be assessed.
The Stokes parameter of the polarisation state generated at the mirror #1 by the linear
polariser at azimuth, α, and the quarter waveplate at azimuth, θ, is given by
56
 s1   cos(2α) cos (2θ) + sin(2α) cos(2θ) sin(2θ) 

2
 
s =  s 2  =  cos(2α) cos(2θ) sin( 2θ) + sin(2α) sin 2 (2θ)  . (121)
 s   sin(2α) cos(2θ) − cos(2α) sin( 2θ)

 3  
The uncertainty on this polarisation state is given by the extinction ratio of the linear
polarisers, E dB, the isolation of the back reflected circular state, I dB, the accuracy of the
angular setting σα & σθ, and the uncertainty due to non-normal reflection from the mirror #1,
σM,
2
σs = σs1 + σs 2 + σs 3 = 13 10 20 + 10 20  + sin 2 (2σα ) + sin 2 (2σθ ) + σ M .
2 2 2 2 −E − I 2
(122)
 
For extinction and isolation ratio > 45 dB, angular uncertainty < 6’ and an angle of incidence
<1º at mirror #1 the uncertainty on the set polarisation state, σs, is less than 0.01 rad.
The azimuth of the linear polariser and quarter waveplate are adjusted to give polarisation
states over the whole sphere that are used to test the polarisation state analyser. The average
deviation of the measured state from the set polarisation state weighted by the area of the
Poincaré sphere marked out by the nearest points is calculated as the uncertainty. At NPL the
azimuth was set from –90º to 90º in 15º steps giving 169 measurements. This gives 62
different states on the Poincaré sphere which are weighted by the area of the sphere marked
out by arcs ±15º from the centre point.
The effect of the input power level on the polarisation state analyser can be easily assessed by
changing the input optical power using the variable attenuator. The actual polarisation state
entering the polarimeter will be fixed provided the azimuth of the linear polariser and quarter
waveplate remain unchanged. It may be necessary to perform this test for a number of input
polarisation states to discover the worst case. For the polarimeter at NPL the worst case input
occurred for an input with s3~½.
In order to assess the polarisation state analyser accuracy with wavelength it is necessary to
present the polarisation state analyser with a known state as the wavelength is changed. This
is relatively easy for a linear polarised state but may not highlight problems with measuring
circular states. By noting the wavelength at which the quarter wavelate had exactly ½π rad of
retardance it is possible within an offset of nπ to calculate the exact thickness based on
published data for the birefringence and the known material of the waveplate. Given this
thickness and the published birefringence data, it is possible to calculate the retardance as a
function of wavelength and thus the polarisation state presented to the polarisation state
analyser.
At NPL a final uncertainty of ±0.03 rad was achieved on the measured polarisation state for a
single mode fibre input in the range –5 to –45 dBm, 1270 to 1640 nm.
6.2 Calibration of PMD Measurement Equipment

For a full and thorough calibration of any PMD measurement equipment it will be necessary
to perform a calibration on each part of the measurement system. Thus for a Poincaré sphere
based method it will be necessary to calibrated the polarisation state analyser, the wavelength
of the source and aspects of the polarisation controller; For a fixed analyser type method it
will be necessary to calibrate the optical spectrum analyser and aspects of the polarisation
controllers; for the interferometric method it will be necessary to calibrate the motion of the
57
moving arm and the linearity of the detector and for the phase shift technique it will be
necessary to calibrate the frequency, phase accuracy and input polarisation states. It is likely
that only the measurement system manufacturer can perform this kind of full calibration,
however the end user may still wish to perform some check to prove the overall measurement
capability. This can be done with standard artefacts.
It is relatively easy to provide standard artefacts that have no polarisation mode coupling and
give a single DGD value, see Figure 51. These could be a single birefringent waveplate or a
short length of HiBi fibre. Such a low mode coupled artefact will give a single well-defined
DGD that is repeatable under moderate environmental changes and between all PMD
measurement methods. A selection of such devices can provide a sequence of calibration
points alone the DGD scale.
Figure 51 A low polarisation mode coupled artefact
The more interesting situation is that of highly mode coupled regime. A fibre based standard
will possess a high degree of polarisation mode coupling but the exact DGD will depend
heavily on environmental conditions. This means that the calibration accuracy will be limited
by the statistical nature of PMD as outlined in section 5.3.5. An alternative highly mode
coupled artefact can be made by creating a stack of birefringent waveplates that have random
azimuth and thickness, see Figure 52. This produces a DGD vs wavelength curve similar to
randomly mode coupled real fibre that is stable under moderate environmental changes, see
Figure 53. This is ideal for the calibration of polarimetric or phase shift based measurement
systems where the DGD vs wavelength curve can be compared to assess both the wavelength
and DGD scales. For interferometric and fixed analyser based methods that measure only an
average PMD the exact average value depends on the wavelength range of the source and the
exact input and analysed states and this leads to a similar statistical uncertainty as for the real
58
fibre. However as the standard is stable under environmental change it is not possible to alter
the statistical distribution to obtain more independent measurements that would allow for a
reduced uncertainty on the average, where as with a real fibre measurements could be
repeated at a number of temperatures to reduce the overall statistical uncertainty.
Figure 52 High Polarisation mode coupled artefact
1000
900
800
700
600
DGD (fs)
500
400
300
200
100
0
1250 1300 1350 1400 1450 1500 1550 1600
Wavelength (nm)
Figure 53 DGD vs wavelength curve for multi-waveplate stack artefact.
6.3 Calibration of PDL Meters

The accuracy of PDL meters relies on the stability of the optical source and the polarisation
independence of the detector. A basic check to assess the uncertainty can be made by
measuring the PDL of a very low PDL artefact for example a good quality patchlead. This
will show any zero offset in the measured PDL due to variation in transmission of the
polarisation controller or PDR of the detector. This zero offset can only be used as an
uncertainty and will limit the minimum measureable PDL. A further check with a PDL
standard, see Figure 54, with PDL in the normal measured range should then be made.
59
Figure 54 A PDL standard artefact.
6.4 Calibration of PER Meters

The calibration of a PER meter requires the presentation of an optical beam with known
polarisation extinction ratio. Many commercial available PER meters are only ably to accept
fibre inputs. At NPL an in house measurement system has been constructed based on a high
quality polariser and polarisation insensitive detectors with the capability of measuring the
PER of light exiting a HiBi fibre. The alignment of the input state of polarisation is adjusted
to give the desired PER which can be accurately measured. This known PER beam is then
connected to the PER meter under test and compared to the PER meter readings.
60
7. Glossary and References

7.1 Glossary of Terms
Autocorrelation The correlation of a function with itself.

Azimuth, θ The angle, measured in the plane perpendicular to the
direction of optical propagation, of the optical axis of
a waveplate or linear polariser.
Beat Length, Lb, m The length over which the birefringent phase shift,
δβ*Lb is equal to 2π rad.
Birefringence, δβ, rad.m-1 The degree to which a material exhibits different
phase velocities for the two orthogonal states of
polarisation, see paragraph 1.3 & 3.2.
Bit Error Rate, BER, bits.s-1 The number of errors per second in a transmitted bit
sequence.
Brewster’s Angle, θP The angle of incidence at which the reflected light
from a dielectric material is fully polarised, see
paragraph 1.2
Circular Birefringence, δβC, rad.m-1 The degree to which a material exhibits different
phase velocities for the two orthogonal circular states
of polarisation. Often referred to as Optical Activity,
see paragraph 1.5 & 3.3
Coupling Length, m The inverse of the coupling parameter, 1/h, see
paragraph 3.5.
Coupling Parameter, h, m-1 The fractional power coupling between polarisation
modes per unit length, see paragraph 3.5.
Cross-Correlation The correlation of one function with another.
Differential Group Delay, DGD, δτ, s The difference in propagation time between the fast
and slow principle states of polarisation, see
paragraph 3.5.
Electro-Optic Effect The physical phenomena where an applied electric
field changes the refractive index of a material, see
paragraph 1.6 & 3.2.
Extinction Ratio, ER The ratio, often in dB, between the optical power in
the slow axis and that in the fast axis of a polarisation
maintaining component or fibre, see paragraph 3.6.
Fourier Transform A mathematical process allowing the transformation
of a signal between the time and frequency domains.
Gaussian 2
Mathematical function of the form f ( x ) = e − x .
61
Half Waveplate A linear birefringent element that provides a phase

difference of π radians, λ/2, between fast and slow
axis at a specified operating wavelength, see
paragraph 1.4.
HiBi Fibre Optical fibre designed to have a large intrinsic linear
birefringence, sometimes referred to as polarisation
maintaining fibre, see paragraph 3.6.
Interferogram The output intensity profile from an interferometer.
Jones Matrix A two by two complex matrix that describes the
transfer of an input state of polarisation to its output
state of polarisation, see paragraph 2.1 & 2.4.
Jones Vector A complex vector that describes the polarisation state
in terms of two orthogonal components of the electric
field, see paragraph 2.1.
Kerr Effect The physical phenomena where an applied electric
field changes the birefringence of a material, see
Linear Birefringence, δβl, rad.m-1 The degree to which a material exhibits different
phase velocities for the two orthogonal linear states of
polarisation, see paragraph 1.4 & 3.2.
Magneto-Optic Effect The physical phenomena where an applied magnetic
field changes the circular birefringence of a material,
see paragraph 1.6 & 3.3.
Maxwellian 2
A mathematical function of the form f ( x ) = x 2e − x .
Mode Coupling The cross coupling of optical power from one
polarisation mode to another, see coupling parameter
and coupling length.
Modal Dispersion, δβ’, s.m-1 The difference in group delay between polarisation
∂δβ
modes per unit length, δβ' = , see paragraph 3.5.
∂ω
Mueller Matrix A transfer matrix describing the transfer of the input
state of polarisation described by a Stokes vector to
the output state of polarisation, see 2.2 & 2.4.
Optical Activity A material that exhibits different phase velocities for
the two orthogonal circular states of polarisation.
Also referred to as Circular Birefringence, see
PM Fibre Optical fibre designed to have a large intrinsic linear
birefringence, sometimes referred to as HiBi fibre,
see paragraph 3.6.
62
PMD Compensation The compensation of PMD by active or other

techniques, see paragraph 4.4.
Poincaré Sphere A visualisation of polarisation states based on
normalised Stokes vectors that fall on the surface of a
sphere, see paragraph 2.3.
Polarisation Analyser A Polariser used to analyse the amount of light within
a given state of polarisation.
Polarisation Dependent Loss, PDL, Γ The variation of loss of an optical fibre or optical
fibre component with the polarisation state of the
transmitted light. The vector Γ is parallel to the least
attenuated polarisation state. See paragraph 3.7.
Polarisation Dependent Response, The variation in the response of a detector system to
PDR the polarisation state of the incident light. See
paragraph 3.8.
Polarisation Dispersion Vector, PDV, The vector in Stokes space that describes the
Ω, s polarisation dispersion. The vector is aligned along
the PSP with magnitude equal to the DGD. See
paragraph 3.5.
Polarisation Mode Dispersion, PMD A general term for all dispersive effects caused by
polarisation properties. Also specifically used to
describe an average DGD, either linear or root mean
square average over optical frequency, time or
sample. See paragraph 3.5.
Polarised Light Light where the electric field vector direction is
defined, see paragraph 1.1.
Polariser An optical component that only transmits light with
one defined state of polarisation. See paragraph 1.2.
Principle State of Polarisation, PSP The output polarisation state of a system that is
stationary with optical frequency to first order, see
paragraph 3.5.
Q Factor The quality factor of a digital optical communication
signal. The signal to noise ratio of the difference
between the one and zero levels at the decision point.
Quarter Waveplate A linear birefringent element that provides a phase
difference of π/2 radians, λ/4, between fast and slow
axis at a specified operating wavelength, see
paragraph 1.4.
Retardance Plate A birefringent element that provides a phase
difference between fast and slow axis, also known as
a Waveplate, see paragraph 1.4
63
Second Moment, σ The root mean square deviation of a signal from the
mean signal.
∞
∫ P(x)(x − x ) dx ,
2
σ2 = 1
P0
−∞
∞ ∞
x = 1
P0 ∫ P(x) x dx , P0 = ∫ P(x)dx .
−∞ −∞
2
Second Order PDV, Ωω, s The vector that describes the change of the
polarisation dispersion vector with change in optical
frequency. It is the differential of the PDV with
optical frequency, see paragraph 3.5.
Stokes Vector A four-element vector (often normalised to three
elements) that describes the polarisation state of an
optical signal in terms of the power transmitted
through polarising analysers. See paragraph 2.2 &
2.4.
Stress-Optic Effect The physical phenomena where applied mechanical
stress changes the refractive index and birefringence
of a material.
Verdet Constant The constant that describes the magnitude of the
magneto-optic effect, see paragraph 1.6.
Waveplate A birefringent element that provides a phase
difference between fast and slow axis, also known as
a Retardance Plate, see paragraph 1.4
64
7.2 NPL Publications

D J IVES, "Calibration of a Polarisation State Analyser for Polarisation Mode Dispersion
Measurements", 4th Optical Fibre Measurement Conference Digest, NPL UK, Poster,
(October 1997).
D J IVES, "Polarisation Mode Dispersion in Chirped Fibre Bragg Gratings", 1998

Symposium on Optical Fibre Measurement Digest, NIST Boulder USA, (September 1998).
D J IVES, “Measurement of the Distribution of Polarisation Mode Dispersion (PMD) along

an Optical Fibre System”, NPL Report COEM55, June 2000.
D J IVES, FOToN UK, “An Intercomparison of Polarisation Mode Dispersion Measurements

and Calibration Artefacts”, 2000 Symposium on Optical Fibre Measurements, NIST Special
Publication 953, (September 2000), pp 91 – 94.
D J IVES, “Measurement of Polarisaiton Dependent Response of Optical Fibre Power Meters

and Optical Spectrum Analysers”, OFMC’01, Girton College, Cambridge, pp 13 – 16.
S HARRIS, D J IVES, “Polarisation Mode Disperision in Restricted Optical Bandwidth: An

Evaluation of Measurement Techniques”, NPL Report CETM40, February 2002.
R F STEVENS, “Polarisation Extinction Ratio Measurement Requirements for Optical

Communications Systems”, NPL Report CETM41, March 2002.
S HARRIS, D IVES, “Polarisation Mode Disperison in Restricted Optical Bandwidth: An

Evaluation of Measurement Techniques”, 2002 SOFM, NIST Special Publication 988,
(September 2002), pp 169 – 172.
R F STEVENS, “Measuring Polarisation Extinction Ratio for Optical Fibres and Components”,
NPL Report CETM43, March 2003.
D J IVES, “Polarisation Mode Dispersion Compensation Techniques and Measurement

Requirements”, NPL Report CETM48, May 2003.
D J IVES, “Measurement Requirements and Methods for Optical Fibre Polarisation

Controllers”, NPL Report DEM-EM001, April 2004.
65
7.3 References
[1] M. BORN, E. WOLF, “Principles of Optics”, Pergamon Press.

[2] E. HECHT, “Optics”, Addison-Wesley.
[3] H. G. JERRARD “Modern description of polarized light: matrix methods”, Optics
and Laser Tech., Dec 1982, pp 309-319.
[4] S. C. RASHLEIGH, “Origins and Control of Polarization Effects in Single-Mode
Fibers”, J of Light Tech., Vol LT-1 No 2, June 1983, pp 312-331.
[5] M. C. FARRIES, A. J. ROGERS, “Temperature dependence of the Kerr effect in
silica optical fibre”, Electron. Lett., Vol 19 No 21, 1983, pp890-891.
[6] A. C. LUI, M. J. F. DIGONNET, G. S. KINO, “Measurement of the dc Kerr and
electrostrictive phase modulation in silica”, J. Opt. Soc. Am. B, Vol 18 No 2, 2001,
pp187-194.
[7] C. D. POOLE , R. E. WAGNER, “Phenomenological approach to polarisation
dispersion in long single-mode fibres”, Elect. Lett., 22, 1986, pp1029-1030.
[8] H. KOGELNIK, L. E. NELSON, J. P. GORDON, “Emulation and Inversion of
Polarization-Mode Dispersion”, J. of Light. Tech., Vol 21 No 2, 2003, pp482-495.
[9] C. D. POOLE, “Statistical treatment of polarization dispersion in single-mode fiber”,
Optics. Lett., Vol 13 No 8, 1988, pp687-689.
[10] N. GISIN, J. P. VON DER WEID, J. P. PELLAUX, “Polarization Mode Dispersion
of Short and Long Single-Mode Fibres”, J. of Light. Tech., Vol 9 No 7, 1991, pp821-
827.
[11] C. D. POOLE , J. H. Winters, J. A. Nagel, “Dynamical equation for polarization
dispersion”, Optics. Lett., Vol 16 No 6, 1991, pp372-374.
[12] P. CIPRUT, B. GISIN, N. GISIN, R. PASSY, J. P. VON DER WEID, F. PRIETO, C.
W. ZIMMER, “Second-Order Polarization Mode Dispersion: Impact on Analog and
Digital Transmissions”, J. of Light. Tech., Vol 16 No 5, May 1998, pp 757 – 771.
[13] G. J. FOSCHINI, R. M. JOPSON, L. E. NELSON, H. KOGELNIK, “The Statistics of
PMD-Induced Chromatic Fiber Dispersion”, J. of Light. Tech., Vol 17 No 9,
September 1999, pp 1560 – 1565.
[14] G. J. FOSCHINI, C. D. POOLE, “Statistical Theory of Polarization Dispersion in
Single Mode Fibres”, J. of Light. Tech., Vol 9 No 11, 1991, pp1439-1456.
[15] N. GISIN, “Statistics of polarization dependent losses”, Optics Comm., 114, 1995,
pp399-405.
[16] N. GISIN, B. HUTTNER, “Combined effects of polarization mode dispersion and
polarisation dependent losses in optical fibers”, Optics Comm., 142, 1997,
pp 119 - 125.
[17] Y. LI, A. YARIV, “Solutions to the dynamical equation of polarization-mode
dispersion and polarization-dependent losses”, J. Opt. Soc. Am. B., Vol 17 No 11,
November 2000, pp 1821 – 1827.
66
[18] R. FECED, S. J. SAVORY, A. HADJIFOTIOU, “Interaction between polarization

mode dispersion and polarization-dependent losses in optical communication links”,
J. Opt. Soc. Am. B., Vol 20 No 3, March 2003, pp 424 – 433.
[19] L. CHEN, X. BAO, “Polarization-Dependent Loss-Induced Pulse Narrowing in
Birefringent Optical Fiber with Finite Differential Group Delay”, J. of Light. Tech.,
Vol 18 No 5, May 2000, pp 665 – 667.
[20] F. BRUYÈRE, “Impact of First- and Second-Order PMD in Optical Digital
Transmission Systems”, Optical Fibre Tech, 2, 1996, pp 269 - 280.
[21] L. NELSON, “Polarisation Mode Dispersion Effects” OFC 2003 Digest, 1, 2003,
TuO1, pp 258.
[22] H. SUNNERUD, C. XIE, M. KARLSSON, R. SAMUELSSON, P ANDREKSON, “A
Comparison Between Different PMD Compensation Techniques”, J. of Light. Tech.,
Vol 20, March 2002, pp 368 – 378.
[23] P. WAI, C. MENYUK, “Polarization Mode Dispersion, Decorrelation, and Diffusion
in Optical Fibers with Randomly Varying Birefringence”, J. of Light. Tech, Vol 14,
February 1996, pp 148 – 157.
[24] A. PAPP, H. HARMS, “Polarization optics of index-gradient optical waveguide
fibers”, Appl. Optics, Vol 14 No 10, October 1975, pp 2406- 2411.
[25] F. CORSI, A. GALTAROSSA, L. PALMIERI, “Beat Length Characterization Based
on Backscatter Analysis in Randomly Perturbed Single-Mode Fibers”, J of Light.
Tech., Vol 17 No 7, July 1999, pp 1172 – 1178.
[26] B. HUTTNER, J. REECHT, N. GISIN, R. PASSY, J. P. VON DER WEID, “Local
Birefringence Measurements in Single-Mode Fibres with Coherent Optical
Frequency-Domain Reflectometry”, IEEE Photonics Tech. Lett., Vol 10 No 10,
October 1998, pp 1458 – 1460.
[27] TIA/EIA-455-122, “Polarisation mode dispersion measurement for single mode
optical fibers by Jones matrix Eigenanalysis”.
[28] B. L. HEFFNER, “Automated Measurement of Polarization Mode Dispersion Using
Jones Matrix Eigenanalysis”, IEEE Photonics Tech. Lett., Vol 4 No 2, September
1992, pp 1066 – 1069.
[29] C. D. POOLE, N. S. BERGANO, R.E. WAGNER, H. J. SCHULTE, “Polarization
Dispersion and Principle States in a 147 km Undersea Lightwave Cable”, J. of Light.
Tech., Vol 6 No 7, 1988, pp 1185 – 1190.
[30] L.E. NELSON, R. M. JOPSON, H. KOGELNIK “Mueller Matrix Method for
Determining Polarization Mode Dispersion Vectors”, ECOC 1999.
optical fibers by the fixed analyser method”.
[32] C. D. POOLE, D. L. FAVIN, “Polarization-Mode Dispersion Measurements Based on
Transmission Spectra Through a Polarizer”, J. of Light. Tech., Vol 12 No 6, June
1994, pp 917 – 929.
optical fibers by interferometry”.
67
[34] B. L. HEFFNER, “Influence of optical source characteristics on the measurement of

polarization-mode dispersion of highly mode-coupled fibers”, Optics Lett., Vol 21
No 2, January 1996, pp 113 – 115.
[35] P. A. WILLIAMS, “Modulation phase-shift measurement of PMD using only four
launched polarisation states: a new algorithm”, Elect. Lett., Vol 35 No 8, September
1999, pp 1578 – 1579.
[36] N. GISIN, B. GISIN, J. P. VON DER WEID, R. PASSY, “How Accurately Can One
Measure a Statistical Quantity Like Polarization-Mode Dispersion?”, IEEE Photonics
Tech. Lett., Vol 8 No 12, December 1996, pp 1671 – 1673.
[37] M. SHTAIF, A. MECOZZI, “Study of the frequency autocorrelation of the
differential group delay in fibers with polarization mode dispersion”, Optics Lett., Vol
25 No 10, May 2000, pp 707 – 709.
[38] R. M. CRAIG, S. L. GILBERT, P. D. HALE, “High-Resolution, Nonmechanical
Approach to Polarization-Dependent Transmission Measurements”, J of Light. Tech.,
Vol 16 No 7, July 1998, pp 1285 – 1294.
68

Polarisation Effects and Measurements in Optical Fibre Systems

Uploaded by

Copyright:

Available Formats

Polarisation Effects and Measurements in Optical Fibre Systems

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Polarisation Effects and Measurements in Optical Fibre Systems

Uploaded by

Copyright:

Available Formats

Polarisation

Effects and Measurements

National Physical Laboratory

Approved on behalf of Managing Director, NPL

1. Polarised Light .................................................................................................................1

2. Mathematical Descriptions of Polarised Light............................................................12

3. Polarisation Effects in Optical Fibres ..........................................................................18

4. The Effects of Polarisation on Data Transmission......................................................31

5. Polarisation Measurement ............................................................................................36

6. Calibration of Polarisation Measurement Equipment ...............................................56

7. Glossary and References ...............................................................................................61

Axis(2) Linear Polariser(1)

Linear Polariser(2) / Analyser

Figure 4 Effect of two linear polarisers crossed at an angle θ.

electromagnetic propagation is supported. At boundaries where the magnetic permeability is

Figure 6 Scattering of polarised light by a small dipole.

Figure 7 Wollaston prism acting as a polarising beam splitter.

1.3 Birefringence in Crystaline Materials

Figure 8 Double images in Calcite

Quarter Wave Rhomb

Half Wave Rhomb

Figure 10 Fresnel Rhombs as achromatic retarders

1.5 Optical Activity and Circular Birefringence

1.6 Electro-optic and Magneto-optic effects

1.7 Effect on Reflection and Transmission

Amplitude Reflection Coefficient

The optical phase change on reflection is given by

2. Mathematical Descriptions of Polarised Light

2.2 Stokes Vector and Mueller Matrixes

The degree of polarisation, DOP, is given by

2.3 Poincaré Sphere

2.4 Useful Matrices

State Jones Vector Stokes Vector

Element Jones Matrix Mueller Matrix

3. Polarisation Effects in Optical Fibres

3.2 Causes of Linear Birefringence

3.3 Causes of Circular Birefringence

3.4 Local Effects of Fibre Birefringence

The birefringence vector, W, is given by

3.5 Global Effects in Fibre, Polarisation Mode Dispersion

Figure 19 Loci of polarisation state on surface of sphere as

The length dependence of the PMD is given by[11]:

Figure 22 Probability distribution of Figure 23 Probability distribution of

Table 3 Useful relationships for high polarisation mode coupled systems.

3.6 Polarisation Maintaining Fibre

Oval Inner Clad Panda

Bow Tie Oval Core

3.7 Polarisation Dependent Loss

In the absence of birefringence an input polarisation state, sin, is transformed to an output

3.8 Polarisation Dependent Response

3.9 PMD and PDL Combined

4. The Effects of Polarisation on Data Transmission

DGD = 0.20 UI DGD = 0.40 UI

DGD = 0.60 UI DGD = 0.80 UI

4.2 Impact of Second Order PMD

4.3 PMD of Installed Links

4.4 Optical PMD Compensation

DGD Error Detection

Optical PMD Polarisation DGD Optical PMD

Figure 27 Block diagram of an Optical PMD compensation system